6th International

Seminar on
Tailings
Management

EDITOR
Hugo Quelopana
Contents

Organizers
Committees
Preface
Acknowledgements
Editor

CHAPTER 1
Geotechnics and Dam Safety
Application of Probabilistic and Deterministic Earthquake
Ground Motions to Tailings Dam Design
Marcelo Martínez and Alan Hull
Breach Modes and Effects of Failure of Barahona 1 Tailings
Dam in 1928
Jorge Troncoso Troncoso and Natalia Pizarro
Released Volume Estimation for Dam Break Analysis
Hugo Quelopana
The Geotechnical Properties of Filtered Tailings and Waste
Rock Blends
Ralph Burden, Ward Wilson, David Williams
and Michael Jacobs
A Different View of Run-Out Distance after Tailings
Impoundment Failure
István Kovács, Claudio Román, Pablo Galdeano
and Marcelo Mussé
Liquefaction Assessment of Deposited Thickened Tailings
using CPTu Soundings Calibrated with Laboratory
Testing: a Critical State Soil Mechanics Application
 Edgar Bard and Guillermo González
Verification of Dynamic Behavior Modeling of Las Tórtolas
Main Dam during 27F Chile Earthquake
Javiera González, Priscilla Torres, José Campaña
and Fabián Leal
Dynamic Characterization of a Tailings Dam Embankment
using a Dense Seismic Array: Preliminary Results
César Pastén, Diana Comte, Gerardo Peña, José Burgos and
Andreas Rietbrock
Dam-Break: Effect of Consolidation Rates in Tailings vs
Run-Out Distance
Diego Inzunza and Jorge León
The benefits of Material Volume Meshes in 3D Slope
Stability Analysis: a Tailings Dam Case Study
Murray Fredlund, Marina Trevizolli and Jon Foster
Geotechnical Characterization of a Limestone Rockfill
used in a Large Rockfill Tailings Dam
José Ccotohuanca, José Ale, Donald East
and Leonardo Dorador
Evaluation of the Stability of the Embankments Dams
of a Tailings Storage Considering Unsaturated
Properties of Soils
Daniel Guzmán and María Teresa Zandarín
Impacts of Topography Quality on Dam
Breach Assessments
Alexandra Halliday and Alfredo Arenas
Comparison of Liquefied Shear Strength of Tailings
Estimated using Standard Empirical Methods and Post-
cyclic Direct Simple Shear Tests
Ganan Nadarajah, David Bleiker and Siva Sivathayalan
Characterization of Old Tailings under
Unsaturated Conditions
 José Ale, César Luna, José Ccotohuanca and Carlos Jiménez
On the Dynamic Properties of Tailings
Rolando Rojas, Miguel Regalado, Renzo Ayala
and Denys Parra
Using Seismic Records to Determine the Predominant
Vibration Frequency of a Tailings Dam Embankment:
First results
Nicolle Correia and César Pasten
Tailings Management at Pueblo Viejo: Accomplishments
of El Llagal TSF
Lenin Montas

CHAPTER 2
Hydraulics, Transportation and Rheology
Dam Runout Sensitivity to the Rheological Formulations
using a Two-Dimensional Flexible-Mesh Model
Reinaldo García, Mario Morales, Pilar Garcia-Navarro,
Javier Murillo and Carmelo Juez
Optimization of Maintenance Plans for Tailings
Transport Systems
Omar Mejías
Tailings Spigots Hydraulic Scale Modelling
Vanessa Veloso, Ray Martinson and Leonardo Olavarría
Characterization of the Effect of Clay Mineralogy
and Content on the Rheological Behavior of Copper
Sulfide Tailings
 Sebastián Contreras, Claudia Castillo, Christian Ihle
and Gabriel Méndez

CHAPTER 3
Seepage and Water Management
Experiences of Geomembranes in Mining Projects
Gabriella Vaschetti, Marco Bacchelli, Alberto Scuero and
Pascual Perazzo
Integrated Methodology and Tools to Estimate and Manage
Infiltration in Tailings Storage Facilities
Agustín Álvarez and Jorge Guarda
Mining and Reusing Tailings: How to Deal with Dewatering
at Low Permeability Media?
Eduardo Ruíz, Jorge Bueno, Lizardo Huamaní, Jairo Paucar,
Carlos Jiménez and Yuri Gallo
Mixed-in-Place Techniques to Improve Dam Safety and
Reduce Seepage
Hans Schwarzweller, Jaime Sobrino and Mario Goulart
Bituminous Geomembrane (BGM) Longevity and Past Use
for Waterproofing Tailings
Emilio Escobar, Bertrand Breul and Bernard Breul
Reducing Leakage Risks on a TSF's by using a Flexible
Ballast Layer as Covering System: Case History
Daniel Fernández and Melchior Briers
Flood Routing over Talabre TSF
 Alfredo Edwards, Francisco Fernández, Tirso López,
Paolo Maldonado and Gonzalo Caro

CHAPTER 4
Innovations in Conventional, Thickened
and Paste Tailings
Resilience in Tailings Management
Sergio Barrera
Slurry Dewatering for the Disposal in Tailing Dams
Rubén Olcay, Christian Hernández and George Valadão
Innovative Process Engineering to Create Better
Tailings Facilities
Christian Kujawa, Jeff Winterton, Rachel Jansen
and Robert Cooke
The Use of Mobile Paste Technology for Crown Pillar
Stabilization in a Flooded Stope
Sue Longo, Thomas Skocir and Jonathon Taylor
High Efficiency Filters for Tailings Applications
Kenneth Rahal, Dave Hanfland and Todd Wisdom
Considerations for Surface Disposal of Filtered Tailings
John Lupo and Marcelo Mussé
Economic Drivers for High-Capacity Tailings
Pressure Filtration
Matthew Pyle, Richard Whittering and Greg Lane
Considerations for Preparing Design Criteria for
Dewatered Tailings
Juan Jose Moreno and Samuel Kendall
An Introduction to Oil Sands Tailings: Innovations
and Challenges
Ward Wilson, Nicholas Beier and Vivian Giang
Optimization of Tailings through Mineral Processing
Päivi Kauppila, Antti Taskinen, Marja Lehtonen, Neea Heino,
Matti Kurhila and Mia Tiljander
 Accelerated Mechanical Consolidation (AMC) to Improve
Tailings Storage Sustainability
David Smirk
Hydraulic Disposal of Sand for Dams
Pablo Vásquez
Tailings Dam Made Redundant using Solid Bowl
Centrifuge
Amol Chinchankar, Sunny Li and Darío Oyanadel

CHAPTER 5
Technologies and Instrumentation for Monitoring
and Surveillance
Operational InSAR Monitoring for Tailings Dams
Ciscu Sánchez, Sara Del Conte and Miguel Ángel Verdugo
Next Step of Remote Sensing Tools for Tailings Dam
Monitoring: the Importance of Predicting and Reporting
Javier Duro, Rubén Iglesias, Fernando Vicente, Dani Monells,
Giuseppe Centolanza and Rodrigo Godoy
Satellite Monitoring of Supernatant Volumes and Beach
Moisture Estimations within Tailings Storage Facilities
Luis Navarro, Gabriel Aravena, Jon Engels and
Fabián Martínez
Using Technology to Identify Seepage Flow Paths Through,
Under and Around Tailings Impoundments
Arturo Fahrenkrog, Ryan Blanchard and Judson Kennedy
Integrated Monitoring of Tailings: EIT STINGS
Karsten Zimmermann, Daniel Schröder, Sebastián Fingerhuth
and Roberto Gesche
Integration of Monitoring Instrumentation for the Laguna
Seca TSF Surveillance: an Innovative Approach
Nicolle Correia, Cristian Roa, Carlo Calderón
and Mauricio Vargas
Improvements on Instrumentation and Monitoring over
Critical Variables in Ovejería Tailings Dam
 Manuel Acuña and Álvaro Gutiérrez
Classification of Tailings by Remote Sensing Methods
by U.A.S
Jimmy Ramos
Integration of Geospatial Data for Monitoring and
Analysis of Tailings Pond “Las Tórtolas”, Anglo American
DLB Company
David Hermosilla
Increasing Tailings Safety and Governance via Real Time
Monitoring Utilising Long Range Wide Area Networks
(LoRaWAN)
Joe Carr
Searching for Precursors of the Córrego de Feijão Dam
Collapse in Brumadinho using Advanced Satellite InSAR
Giulia Tessari, Peter Stegmaier and Paolo Pasquali

CHAPTER 6
Governance and Operational Experiences
Quantification of Risks in Tailings Deposits
Juan Rayo
Construction Quality Assurance Audits: an Effective
Tailings Governance Tool
Jorge Troncoso Boys and Daniel Warner
Approach and Considerations for OMS
Manual Development
Ivette De Jesús and Paul Bedell
Engineer of Record Role and Experience in Chilean
Tailings Storage Facility
Francisco Vera, Marcelo Arenas, Alejandra Neira and
Esteban Barría
Management of Tailings Storage Facility Operational Data:
Lessons Learned
Esteban Barría, Gian Armijo, Marcelo Arriagada
and Alejandra Neira
A Risk-Based Approach to Support the Implementation of
the New Guidelines Related to Tailings Management
Michel Julien, Pascal Lavoie, Edouard Masengo
and Thomas Lépine
Los Diques Tailings Storage Facility: Construction
Quality Assurance
Daniel Warner and Victor Soto
Main Challenges in Tailings Dams Risk Management
Luis Valenzuela
Organizers

Tailings 2019 was organized by the University of Alberta, Canada;

Advanced Mining Technology Center (AMTC), Chile; Universidad Técnica
Federico Santa María, Chile; and Gecamin, Chile.

T h e U n i v e r s i t y o f A l b e r ta ,
G e o t e c h n i c a l C e n t r e , Ca n a d a
The University of Alberta Geotechnical Centre is one of the largest
and strongest group of its kind in Canada and comparable in size to
other major research-intensive groups in North America. Our faculty
are internationally recognized experts in various sub-disciplines
within geotechnical and geoenvironmental engineering. Currently, the
Geotechnical Centre consists of eight faculty members, four professor
emeriti, and over 90 graduate students, postdoctoral fellows, research
associates and technical support staff.
Our commitment to research excellence has led to the establishment
of several specialized facilities, which are well tailored for the innovative
research programs led by the Geotechnical Centre. The Geotechnical
Centre's laboratories are located in the Department of Civil &
Environmental Engineering in the Markin/CNR L Natural Resources
Engineering Facility, which has over 300,000 square feet of floor area,
78 research laboratories, 26 specialized instructional laboratories, and
many high-tech lecture rooms.
The Geotechnical Centre is part of the Faculty of Engineering at the
University of Alberta. The Faculty is committed to ensuring our students
learn in a world-class environment and is home to approximately 200
faculty members who teach, mentor students and conduct leading-edge
research. The Faculty has completed five new buildings, adding 1.5
million square feet of teaching, research, student study and common
areas, and administrative space since 2001 to support expanded
educational and research activities.
Learn more by visiting cort.as/-ImF4

9
 A d va n c e d M i n i n g T e c h n o l o g y C e n t e r , C h i l e
TAILINGs 2019

The Advanced Mining Technolog y Center (A M T C) is the leading

research center in Chile in technology applied to mining and whose
mission is to generate world-class multidisciplinary research, transfer
new technologies and form advanced human capital responding to the
challenges of the mining sector to ensure welfare and development for
Chile and the world.
The AMTC was created in March 2009 at the University of Chile, after
being selected by the Partnership Research Program of CONICYT under
the Baseline Funding Program for Scientific and Technological Centers
of Excellence. The AMTC has managed to consolidate and integrate the
activities of five research teams with extensive experience in the training
of researchers and top professionals as well as in the development of
scientific excellence.
Its 176 scientists are integrated within the five research groups:
Exploration and Deposit Modeling; Mining Design and Planning; Mineral
Processing and Extractive Metallurgy; Mining Automation; Water and
Environmental Sustainability. In addition, A M T C has established
important cooperation, and joint research initiatives with world class
science organizations and technology centers. Among the A M T C
partners are Codelco and BHP Minerals Americas, whose representatives
are also members of the AMTC Board.
Learn more by visiting amtc.cl

10
U n i v e r s i d a d T é c n i c a F e d e r i c o Sa n ta M a r í a , C h i l e
Universidad Técnica Federico Santa Maria was founded in Valparaíso
on 27 April, 1926, when the Federico Santa María Foundation was
established, thanks to his will and testament. The entrepreneur and
businessman from Valparaíso wanted to leave his country with a center of
scientific and technological studies to train professionals at the highest
level. On December 20, 1931, the Crafts School and the Engineering
College, “José Miguel Carrera”, were inaugurated with academic activity
starting in 1932. In 1935, the Foundation was officially named Universidad
Técnica Federico Santa María.
In the 1970's the university started a university technician training
program in the area of science and technology, with the establishment
of the Viña del Mar branch, “José Miguel Carrera”, and the Concepción
branch, “Rey Balduino de Bélgica”.
In 1995, the first Campus in Santiago (Vitacura) opened. In 1996,
the USM Campus Guayaquil opened in Ecuador, making it the first
Chilean University to open a campus abroad. In 2000, USM and LA N
Chile joined to found the Aeronautical Sciences Academy (ACA) at the
Santiago-Vitacura campus. This innovative project links the academic
sector to the aeronautical industry with its mission to prepare top level
professionals for the Latin American market. In 2009 the second campus
in Santiago (San Joaquín) was opened. This new campus will strengthen
the university's presence in the Metropolitan Region.
After 80 years the University is reputed for its prestige and leadership
in engineering, science and technology.
Learn more by visiting usm.cl

11
 GECAMIN, Chile
TAILINGs 2019

Over 20 years promoting the exchange of knowledge and innovation

Gecamin organizes high-quality conferences focused on transferring the
best practices, latest innovations, and technological trends to improve
productivity and excellency in the mining and industrial sectors.
In its 21-year history, Gecamin has organized conferences for
over 27,000 executives and professionals in key areas for the industrial
sector such as water, energy, environment, sustainability, automation,
maintenance, and human capital.
During 2018, Gecamin organized 10 conferences which gathered
2,380 professionals from 50 countries, 800 being delegates from
industrial operations. A book with the technical content was published
for each conference, summarizing over 500 published articles and 600
technical presentations.
Learn more about Gecamin conferences by visiting gecamin.com

12
Committees

ORGANIZING COMMITTEE
Executive Committee

Chair
Carlos Pavissich, Tailings Manager, Copper Projects,
BHP Minerals Americas

Co-organizers
Ward Wilson, Professor, University of Alberta, Canada
César Pastén, Academic, AMTC, Universidad de Chile
Gonzalo Suazo, Geotechnical Professor,
Universidad Técnica Federico Santa María, Chile

Executive Director
Carlos Barahona, General Manager, Gecamin, Chile

Program Chair
Hugo Quelopana, Specialist Engineer, Delfing, Chile

Seminar Coordinator
Oliver Allendes, Gecamin, Chile

Directing Members
Hubert González, Water and Tailings Manager, Minera Sierra Gorda,
Chile · Luis Tejada, Corporate Geotechnical Manager, Minsur, Peru ·
René Orellana, Water and Tailings Corporate Manager, Codelco, Chile ·
John Pottie, Geotechnical Engineering Manager, Teck, Chile

13
 Advisory Committee
TAILINGs 2019

Manuel Acuña, Andina Division, Codelco, Chile · Tatyana Alexieva, BHP,

USA · Marco Bacchelli, Carpi Tech, Switzerland · Raúl Bahamondes,
Minera Centinela, Antofagasta Minerals, Chile · Eldert Besseling,
NETICS, Netherlands · Christian Cáceres, Minera Valle Central, Chile ·
Gonzalo Caro, Project Vice Presidency, Codelco, Chile · Fernando Castillo,
Thrace-Eurobent, Chile · Robert Cooke, Paterson & Cooke, USA · Adel
Cortéz, Fugro, Chile · Michael Davies, Teck Resources, Canada · Mark
Donald, ArcelorMittal Mining, UK · Emilio Escobar, Axter Ibérica, Chile ·
Raúl Espinace, Geotecnia Ambiental, Chile · Julio Faúndes, Hidrica
Consultores, Chile · Daniel Fernández, Huesker, Brazil · Pablo Galdeano,
Ausenco, Chile · Yuri Gallo, Minsur, Peru · Enrique Garcés, Knight
Piesold, Chile · Patricio Gómez, Itasca, Chile · Rodrigo Gutiérrez, Minera
Centinela, Antofagasta Minerals, Chile · Sven Herlitz, AguaEx, Germany,
Chile, Peru · Richard Jewell, Australian Centre for Geomechanics,
Australia · Andrés León, Anddes, Chile · Ray Martinson, Paterson &
Cooke, Chile · José Mello, Wood, Chile · Juan Moraga, M&F Consultores,
Chile · Marcelo Mussé, Ausenco, Netke, Chile · James Obermeyer,
Stantec, USA · Denys Parra, Anddes, Peru · Juan Pastén, Minera Playa
Verde, Chile · Félix Pérez, Hidrica Consultores, Chile · José De Piérola,
Southern Copper Corporation, Peru · Marcos Pinto, Codice Ingeniería,
Chile · Francisco Pizarro, Propex, Chile · Jorge Riquelme, Hatch, Chile ·
Matías Roni, Cía. Minera Doña Inés de Collahuasi, Chile · Arash Roshdieh,
ATC Williams, Australia · Ricardo Ruano, Pascua Lama, Barrick Gold,
Chile · Eduardo Ruiz, Amphos 21, Peru · Leonardo Salas, Anddes,
Peru · Carlos Salinas, Paterson & Cooke, Chile · Antonio Samaniego,
SR K Consulting, Perú · Fred Schoenbrunn, FLSmidth, USA · Fiona
Sofra, Rheological Consulting Services, Australia · Adelson de Souza,
Nexa Resources, Brazil · David Stone, MineFill Services, USA · Ricardo
Valdebenito, RV Ingenieros, Chile · Rubén Vargas, Knight Piesold, Peru ·
Hernán Velásquez, Geo Monitoring, Chile · Carlos Vilches, Codelco, Chile

14
TECHNICAL COMMITTEE
Juan Pablo Aránguiz, Codelco, Chile · Andrés Argomedo, El Soldado,
Anglo American, Chile · Eric Astorga, SNC-Lavalin, Chile · Nicholas Beier,
University of Alberta, Canada · David Benavente, Paterson & Cooke, Chile ·
Cristóbal Binvignat, DLS Ingenieros, Chile · Claudio Blaña, Chuquicamata
Division, Codelco, Chile · Heiner Bueno, Arcadis, Peru · Ralph Burden,
University of Alberta, Canada · José Burgos, El Soldado, Anglo American,
Chile · Aldo Campi, Los Bronces, Anglo American, Chile · Patricio Chacón,
Arcadis, Chile · Diego Cobos, Dynami Geoconsulting, Colombia ·
Diana Comte, Universidad de Chile · Glen Corder, The University of
Queensland, Australia · Nicolle Correia, Minera Escondida, BHP, Chile ·
Jeff Crawford, FLSmidth Krebs, Australia · Alejandro Cruzat, Universidad
de La Serena, Chile · Joaquín Delgado, Universidad Pablo de Olavide, Spain ·
Pablo Durán, Project Vice Presidency, Codelco, Chile · Mansour Edraki,
CMLR, Sustainable Minerals Institute, Australia · Phillip Fawell, CSIRO
Mineral Resources, Australia · Sebastian Fernández, University of Alberta,
Canada · Carolina Flórez, Vale-Verde Mining, Brazil · Silvia Franca, Centre
for Mineral Technology, Brazil · Rodrigo Friz, Geotecnia Ambiental, Chile ·
John Furlan, GIW Industries, USA · Pedro Galleguillos, CICITEM, Chile ·
Alireza Ghirian, Golder Associates, Canada · Salvador Gómez, Fresnillo,
Mexico · Giovanni González, Anglo American, Chile · Noemi González,
FLO-2D Software, USA · Raymond Guang, Golder Associates, Canada ·
César Guzmán, El Teniente Division, Codelco, Chile · Leandro Herrera,
Sernageomin, Chile · Amanda Hevia, Arcadis, Chile · Christian Ihle,
AMTC, Universidad de Chile · Roberto Jamett, Golder Associates, Chile ·
Sadegh Javadi, ATC Williams, Australia · Fernanda Kemeid, Vale, New
Caledonia · Christian Ledezma, Pontificia Universidad Católica de Chile ·
Chris Lee, Golder Associates, Canada · Sven Knutsson, Lulea University
of Technology, Sweden · Marcelo Llano, Klohn Crippen Berger, Australia ·
Sue Longo, Golder Associates, Canada · Andrea López, Antofagasta
Minerals, Chile · Marlene López, Project Vice Presidency, Codelco, Chile ·
Bernd Lottermoser, Aachen University, Germany · John Lupo, Newmont
Mining Corporation, USA · Cristián Marambio, SRK Consulting, Chile ·
Francisco Melo, Mitsui & Co, Chile· Telmo Morales, Mitsui & Co, Chile ·

15
 Juan Moreno, SR K Consulting, Australia · Adam Mouttet-Llanos,
TAILINGs 2019

Conetec, Chile · Juan Ojeda, Minera Las Bambas, MMG, Peru · Claudia
Ortiz, Universidad de Santiago de Chile · Carlos Ovalle, Pontificia
Universidad Católica de Chile · Solange Paihua, Knight Piesold, Peru ·
Miguel Palape, MineBridge Software, Chile · Martín Pérez, Independent
Geologist Consultant, Chile · Behnam Pirouz, ATC Williams, Australia ·
Renmin Pretell, University of California, Davis, USA · Jocelyn Quinteros,
Independent Consultant, Chile · Umme Rima, University of Alberta,
Canada · Edgardo Rivas, Chuquicamata Division, Codelco, Chile · José
Rodríguez, Arcadis, Chile · Rodrigo Ruiz, BHP Minerals Americas, Chile ·
Marco Salazar, Minera Los Pelambres, Antofagasta Minerals, Chile · Haley
Schafer, University of Alberta, Canada · Daniel Servigna, Wood, USA ·
Shahé Shnorhokian, McGill University, Canada · Neeltje Slingerland,
University of Alberta, Canada · João Soares, University of Alberta, Canada ·
Mauricio Tobar, Weir Minerals, Chile · Jorge Troncoso Troncoso, Stantec,
Chile · Nicolás Úbeda, Golder Associates, Chile · Jaime Urquidi, Arcadis,
Chile · Luis Valenzuela, Independent Consultant, Chile · Francisco Vega,
Minera Centinela, Antofagasta Minerals, Chile · Mauricio Villanueva,
Minera Centinela, Chile ·John Wates, John Wates Consulting, South
Africa · Mohan Yellishetty, Monash University, Australia

16
E d itorial C ommittee

Editor
Hugo Quelopana, Delfing, Chile

Copy Editors
Oliver Allendes, International Events Coordinator, Gecamin, Chile
Mimy Mackenzie, Head of Conference and Publications, Gecamin, Chile

Reviewers
Silvia Franca, Centre for Mineral Technology, Brazil · Juan Pablo
Aránguiz, Codelco, Chile · Ghio Arenaldi, Delfing, Chile · Eric Astorga,
SNC-Lavalin, Chile · Cristóbal Binvignat, DL S Ingenieros, Chile ·
José Burgos, Anglo American, Chile · Aldo Campi, Anglo American,
Chile · Patricio Chacón, Arcadis, Chile · Diego Cobos, Dynami
Geoconsulting, Colombia · Diana Comte, Universidad de Chile · Glen
Corder, The University of Queensland, Australia · Nicolle Correia,
Minera Escondida, BHP, Chile · Alejandro Cruzat, Universidad de la
Serena, Chile · Joaquín Delgado, Universidad Pablo de Olavide, Spain ·
Pablo Durán, Projects Vicepresidency, Codelco, Chile · Mansour Edraki,
The University of Queensland, Australia · Raúl Espinace, Geotecnia
Ambiental, Chile · Phillip Fawell, CSIRO Mineral Resources, Australia ·
Carolina Flórez, Minera Vale-Verde, Brazil · John Furlan, GI W
Industries, USA · Pablo Galdeano, Ausenco, Chile · Pedro Galleguillos,
CICI T EM, Chile · Alireza Ghirian, University of Ottawa, Canada ·
Salvador Gómez, Fresnillo, Mexico · Noemi González, FLO-2D Software,
USA · Raymond Guang, Golder Associates, Canada · Amanda Hevia,
Arcadis, Chile · Christian Ihle, AMTC, Universidad de Chile · Roberto
Jamett, Golder Associates, Chile · Christian Ledezma, Pontificia
Universidad Católica de Chile · Chris Lee, Golder Associates, Canada ·
Sue Longo, Golder Associates, Canada · Marlene López, Project Vice
Presidency, Codelco, Chile · Andrea López, Antofagasta Minerals, Chile ·
John Lupo, Newmont Mining Corporation, USA · Cristián Marambio,
SR K Consulting, Chile · Juan Moreno, SR K Consulting, Australia ·
Marcelo Mussé, Ausenco, Netke, Chile · Roberto Olguín, I DI EM,

17
 Universidad de Chile · Claudia Ortiz, Universidad de Santiago de Chile ·
TAILINGs 2019

Carlos Ovalle, Pontificia Universidad Católica de Chile · Solange Paihua,

Knight Piésold, Peru · Martín Pérez, Independent Geologist Consultant,
Chile · Behnam Pirouz, ATC Williams, Australia · Renmin Pretell,
University of California, Davis, USA · Jocelyn Quinteros, Independent
Consultant, Chile · Rodrigo Ruiz, Rodrigo Ruiz, BHP Minerals Americas,
Chile · Daniel Servigna, Wood, USA · Shahé Shnorhokian, McGill
University, Canada · Gonzalo Suazo, Universidad Técnica Federico Santa
María, Chile · Nicolás Úbeda, Golder Associates, Chile · Javier Ubilla,
Universidad Técnica Federico Santa María, Chile · Jaime Urquidi,
Arcadis, Chile · Francisco Vega, Minera Centinela, Chile · John Wates,
John Wates Consulting, South Africa · Claudia Welkner, BHP, Chile ·
Mohan Yellishetty, Monash University, Australia · Esteban Sáez, Pontificia
Universidad Católica de Chile · Juan Ojeda, Minera Las Bambas, MMG,
Peru · Marcos Pinto, Codice Ingeniería, Chile

18
Preface

Sustainability of mining requires not only the economic and technical

viability of its work, but also, to obtain a 'social license' to its development.
One of the aspects of mining that has a greater impact on this social
license is the management of tailings, which will impacts the community
in one way or another. Since the society does not necessarily make a
distinction between different mining companies, it must be borne in
mind that any unfortunate event that happens in a Tailings Storage
Facility will impact –to a greater or lesser extent– not only the affected
mine but the entire industry. This situation can even permeate the
borders between countries and become a global issue. The Brumadinho
dam disaster in Brazil this year is an example of this.
In this context, the 6th International Seminar on Tailings Manage-
ment in Chile held from July 10-12 is an excellent opportunity to discuss
the latest development, innovations and lessons learned related to tail-
ings management. Sharing our experiences helps us to better respond to
our direct clients (authority, shareholders, mining companies, suppliers,
etc.) and to the community, not only from an individual perspective, but
also as an industry.
The more than 60 abstracts included in this book, written by
authors from 15 countries, reflect the great effort and special dedication
from all those involved and show the broad spectrum of subjects that
encompasses tailings management: Geotechnics and Dam Safety
(18); Hydraulics, Transportation and Rheology (4); Seepage and Water
Management (7); Innovations in Conventional, Thickened and Paste
Tailings (13); Technologies and Instrumentation for Monitoring and
Surveillance (11); and Governance and Operational Experiences (9).
Finally, I would like to thank all those whose efforts helped make this
seminar a success: to the authors and their organizations for submitting
papers, to the mining companies for sharing useful experiences, to the
technical committee for sharing their expertise dedicating valuable time
correcting the articles providing insightful comments, to the organizing

19
 and advisory committee for leading and promoting the seminar and
TAILINGs 2019

to the sponsors for their supports. My special thanks to all Gecamin

personnel for their efforts in ensuring all the goals proposed for the
seminar were successfully achieved.

Hugo Quelopana
pro gr a m Chair and editor
Tailings 2019
6th International Seminar on Tailings Management

20
Acknowledgements

The Executive Committee acknowledges with gratitude the efforts of

all the authors for contributing a large variety of high quality, detailed
and innovative papers to the technical program. We also would like
to thank the reviewers and all those involved in the creation of these
proceedings for their assistance. The support of the Organizing, Advisory
and Technical Committees has been greatly appreciated.
The Executive Committee also wishes to thank the following sponsors
(in the order in which they were confirmed) for their generous support:
Platinum: Ausenco
Gold: TailPro, W WL, Geotecnia Ambiental, CGG, PhotoSat, SGS,
Inmarsat and Wood
Silver: Conetec, CarpiTech, Bauer, SMI-ICE-Chile, Orbinox, Walo,
Alfa Laval, Geocom, Diemme Filtration, Coletanche, TRE-ALTAMIR A,
SRK Consulting, Fugro, Bentley and SARMAP
Official Material: Arcadis
Social: Weir Minerals and FLSmidth
Other: Golder
Official Media: Nueva Minería y Energía, Chile
Media Partners: Direcmin, Chile; and Reporte Sostenible, Chile
Institutional Partners: Consejo Minero, Chile; Sociedad Nacional de
Minería (SONAMI), Chile; Comisión Chilena del Cobre (COCHILCO);
Servicio Nacional de Geología y Minería (SERNAGEOMIN), Chile; The
Brazilian Mining Association (IBRAM); Red de Ingenieras de Minas, Chile;
Fundación Minera de Chile; Cámara Chilena Australiana de Comercio
(AUSCH A M); Women in Mining Chile (W IM); Australian Trade and
Investment Commission (AUSTR ADE) and Cámara Chileno Canadiense.

21
 Finally, we would like to thank all the delegates who attended the
TAILINGs 2019

seminar and exchanged their valuable knowledge and expertise, thus

contributing to the great success of this 6th International Seminar on
Tailings Management, Tailings 2019. We are looking forward to seeing
you all again during the next version of Tailings.

Executive Committee
Tailings 2019
6th International Seminar on Tailings Management

22
Editor

hugo quelopana
Mr. Quelopana is a Civil Engineer specializing in tailings management
with over 14 years of experience and has been involved in more than
35 projects related to this mining topic. He has developed studies
of site selection, trade off, schemes and depositions technologies,
water balance, reviews and pilot plants. He has diplomas of project
management, leadership and coaching, and hydrogeology applied to
mining and environment.

23
1
Geotechnics and
Dam Safety
Application of Probabilistic and Deterministic
Earthquake Ground Motions to Tailings Dam Design
Marcelo Martínez1 and Alan Hull2
1. Golder Associates, Chile
2. Golder Associates, USA

ABSTRACT

Current international tailings dam guidelines provide earthquake ground motion selection criteria
for the analysis and design of tailings dams. It is not always clear, however, what ground motion
level is applicable, leading to confusion, inconsistencies, and arbitrary selection of ground motions
for seismic analysis and design. For an operating basis earthquake (OBE) performance level, ground
motions with an annual exceedance probability (AEP) of 1/475 are generally recommended. OBE
ground motions are evaluated using probabilistic seismic hazard analysis (PSHA).
For the higher performance safety evaluation earthquake (SEE), the dam failure consequence needs
to be considered because it influences the final ground motion value. SEE ground motions can be
evaluated using both PSHA and the maximum credible earthquake (MCE) evaluated by deterministic
seismic hazard analysis (DSHA). We suggest that the lower of the MCE 84th-percentile, or the 1/10,000
AEP and 1/5,000 AEP ground motion be used for analysis and design of extreme and high failure
consequence dams, respectively. For a low dam failure consequence, the controlling ground motion
can be the lower of the MCE 50th-percentile (median), or the 1/2,500-AEP.
When DSHA is not possible because of the lack of identifiable fault sources in very low to low seismic-
hazard regions, then engineering judgment should be used to select the most appropriate PSHA-
based ground motion associated with a low AEP.

1
INTRODUCTION

Although international guidelines for tailings storage facility (TSF) seismic analysis and design are
widely cited by designers (e.g., ICOLD 2010, 2016; CDA 2014; NZSOLD 2016; ANCOLD 2017), the
selection of design earthquake ground motions can often appear confusing. Much of the confusion
centers around the appropriate use of earthquake ground motion outputs developed from
probabilistic seismic hazard analysis (PSHA) and deterministic seismic hazard analysis (DSHA)—
two common methods used to develop design-level ground motions for TSFs.
While some designers select ground motions from the seismic provisions developed in local building
codes, these ground motions are not appropriate for design and analysis of TSFs because they are
neither site-specific nor for use beyond life safety design for standard structures. This paper outlines
an approach to the selection of earthquake ground motions for TSF stability analysis and design using
both PSHA and DSHA methods.
Hazard curves presented in this paper are mean horizontal accelerations generalized from the results
of site-specific seismic hazard studies undertaken by Golder Associates in low, moderate and seismic
hazard regions worldwide. Hazard curves presented for very high and high seismic regions are
largely from sites in western Andes of South America, particularly, from Chile and Peru. The ground
motions are representative of what can be expected in western South America, but are not suitable
for any specific site.

EARTHQUAKE HAZARD AND RETURN PERIODS

Current international guidelines for earthquake ground motion selection for tailings dam design are
typically based on discrete annual exceedance probabilities (AEP -or their inverse of return period-),
and for a specified dam failure consequence. Understanding the interrelation between PSHA- and
DSHA-based ground motions and their AEP or return period is critical to the selection of a suitable
ground motion level for the dam design. Furthermore, the use of PSHA and/or DSHA ground
motions for design depends on the regional seismic hazard level where the TSF is located. The
influence of seismic region on the selection of ground motions is descried below.

Low Seismic Hazard Regions

Figure 1 shows a typical probabilistic earthquake hazard curve for mean horizontal peak ground
acceleration (PGA) in a generic low seismic hazard region. A low seismic hazard region is one where
there is a limited number of historical records, the maximum recorded earthquake is less than about
M16.0, and seismogenic faults have not been recognized. In these regions, the identification and

1M is moment magnitude that is a measure of the energy released at the earthquake source based on the area of the
causative fault and the average slip.

2
characterization of seismic sources is typically difficult. When seismogenic fault sources are absent,
DSHA cannot be applied, so PSHA should be used to develop design earthquake ground motions.
In low hazard regions, design-level earthquake ground motions are typically selected for low AEPs
(e.g. AEP of 1/2,500 to 1/10,000) depending on the TSF failure consequence rating. Earthquake ground
motion values are typically low and may not control the TSF dam design, even when the design
earthquake ground motions are for a low or very-low AEP (e.g. 1/10,000).

Figure 1 Typical mean PGA seismic hazard curve for a site in a very-low to low earthquake hazard region
(Disclaimer: ground motion values are referential and for description purposes only)

Moderate-Seismic Hazard Regions

Figure 2 presents the PSHA-based mean horizontal PGA hazard curve typical of a moderate seismic
hazard region.

3
 Figure 2 Typical mean PGA hazard curve for a moderate hazard region
(Disclaimer: ground motion values are referential and for description purposes only)

Such regions usually have an historical record of earthquake occurrence, rare large damaging
earthquakes, and geologic evidence for seismically active faults, although with relatively low average
slip rates (i.e. less than about 1 mm/yr). There is usually sufficient data available to apply the DSHA-
based maximum credible earthquake (MCE 2) to estimate for median, or the 84th-percentile (median
plus one standard deviation) ground motions. In moderate seismic-hazard regions the MCE median
ground motion values are generally closer to and above the 1/475 AEP, and the MCE 84th-percentile
ground motions typically are associated with an AEP of about 1/2,500 to 1/5,000 (Figure 2).

High-Seismic Hazard Regions

Figure 3 shows the probabilistic mean PGA hazard curve for a typical high seismic hazard region.
High seismic hazard regions are located on or close to major plate tectonic boundaries such as the
western regions of South America from Colombia in the north to Southern Chile and Argentina.

2 The conventionally accepted maximum credible earthquake (MCE) definition is used in this paper. The DSHA-based MCE
is understood as the “largest reasonably conceivable earthquake that is considered possible along a recognized fault or within a
geographically defined tectonic province, under the presently known or presumed tectonic framework” (ICOLD 2010 and 2016). By
definition, the MCE does not explicitly consider the earthquake recurrence interval. This MCE definition should not be
confused with the Maximum Considered Earthquake (MCEQ) defined in ASCE 7-10 and the 2015-IBC, probabilistically
evaluated at an AEP of 1/2,475.

4
 Figure 3 Typical mean PGA hazard curve for a very-high to high hazard region
(Disclaimer: ground motion values are referential and for description purposes only)

These regions typically have an historical record of multiple large and damaging earthquakes, and
annual occurrence of many moderate earthquakes. The earthquakes are generated by active tectonic
process at subduction zone and/or multiple, high-slip (>5 mm/yr.) seismogenic faults with evidence
for repeated surface ruptures in the Holocene epoch (i.e. last 11,700 years).
Spectral ground motions (e.g. PGA) are much larger than those in low and moderate seismic hazard
regions. In high seismic hazard regions, the median MCE ground motions are typically closer to or
even below the 1/475 AEP. The 84th-percentile MCE are typically associated with AEPs from about
1/750-year to 1/2,500-year.

PSHA-DSHA Ground Motion Selection Criteria

Figure 4 shows the relationship between mean horizontal PGA hazard curves (Figures 1, 2 and 3)
developed from PSHA and the MCE-84th percentile in the three seismic hazard regions.

5
 Figure 4 Typical mean PGA hazard curves for very low-to-low, moderate and very high-to-high earthquake
hazard regions (Disclaimer: ground motion values are referential, for description purposes only)

As indicated, the MCE scenario represents the largest reasonably conceivable earthquake considered
possible along a recognized fault, or within a tectonic province (ICOLD 2010, 2016). If the earthquake
ground motion value estimated for the MCE is lower than those calculated from the PSHA approach
(e.g. for an AEP of 1/10,000), then the largest reasonably conceivable MCE ground motion will control
design. Otherwise, the PSHA-based ground motions (for the appropriate AEP) will likely govern the
analysis. The earthquake ground motion severity to be adopted will be the MCE ground motion, or
the ground motion for the AEP for the appropriate dam failure consequence, whichever is lower.
Design earthquake ground motion values for very-low and low hazard regions, where the MCE
cannot be deterministically defined (Figures 1 and 4), can be selected for low to very-low AEPs (e.g.
AEP of 1/2,500 to 1/10,000) depending on the dam failure rating.
For moderate seismic hazard regions, the 84th-percentile MCE ground motions are typically at AEPs
of 1/2,500 to 1/5,000, with relatively high spectral values. No further ground motion values linked to
lower AEP (e.g. 1/5,000 to 1/10,000) appear to be physically feasibly for dam design at moderate-
seismic hazard regions as shown in Figures 2 and 4.
For very-high and high seismic hazard regions, large earthquakes are usually (very) frequent and
high ground motion return periods are shorter than low to moderate hazard regions. These strong
earthquakes also trigger very-high ground motions at the site of analysis (Figures 3 and 4).

6
In high seismic regions, the MCE (84th-percentile) ground motions are typically bounded by ground
motions at AEPs of about 1/750 to 1/2,500, and ground motion values for AEPs lower than 1/2,500
appear unnecessary for dam design.
As expected, MCE median (50th-percentile) ground motion values generally have higher AEPs (i.e.
shorter return periods) as shown in Figures 2 and 3.

Period Range of Interest for Dynamic Deformation Analyses

Unlike the pseudo-static PGA-based analyses, earthquake ground motions used in dynamic analyses
are generally developed for a period range of interest that is typically in the range of about 0.2 s to
1.0 s (i.e. near the natural period of the dam structure), for embankment dams used in TSFs.

Figure 5 Typical mean hazard curves and MCE-84th percentile spectral values for discrete spectral periods
(PGA, 0.2 s and 1.0 s) and for high (left), moderate (center) and low (right) earthquake hazard regions.
(Disclaimer: ground motion values are referential, for description purposes only)

Figure 5 illustrates the probabilistic mean hazard curves and the MCE-84th percentile ground motions
for three spectral periods (PGA, 0.2 s and 1.0 s) and the three seismic hazard level conditions. These
relations can be used to evaluate the impact on earthquake return period for spectral periods other
than PGA. For the MCE 84th-percentile ground motions, Figure 5 shows similar values of AEP (i.e.
return periods) for PGA and 0.2 s for both moderate and high seismic-hazard regions with AEPs
ranging from about 1/750 to 1/2,500 for high seismic regions, and 1/2,500 to 1/5,000 for moderate
seismic hazard areas.
For 1.0 s spectral period the MCE 84th-percentile values have typically lower AEPs for higher
frequencies (i.e. ≤0.2 s). AEP values are typically between 1/2,500 and above 1/5,000 for high and
moderate hazard regions, respectively.

7
EARTHQUAKE GROUND MOTIONS FOR TAILINGS DAM DESIGN

Available international guidelines specify several earthquake performance targets. However, the
operating basis earthquake (OBE) and the safety evaluation earthquake (SSE) are the two common
earthquake performance levels used in tailings dam design.

Earthquake Ground Motions for the OBE

The OBE is the ground motion level for which a dam is designed to remain operational, with any
damage being minor and readily repairable following the OBE occurrence. OBE ground motions can
reasonably be expected to occur during the operating life of the dam.
The OBE performance level for tailings dams typically has an AEP of 1/475. An OBE performance
level at a higher or lower AEP can be used based on the needs of the owners, designers and local
regulators. However, any change in the OBE ground motion AEP must still meet the performance
requirement that the dam remain operational, with minimal, easily repairable damage (Wieland 2012,
2016). An OBE performance level may not be necessary in low seismic hazard regions because the
ground motion level is very low and has a minimal impact on dam stability.

Recommended Earthquake Ground Motions for the SEE

Dam Failure Consequence

Definition of the SEE ground motion should be based on the consequence of dam failure. Existing
international guidelines for tailings dam design recommend the failure consequence be based on a
qualitative and quantitative descriptor of the consequence of a hypothetical dam collapse. Guidelines
typically define a minimum of three main failure targets; low, high and extreme, with the ranking
dependent on the human, environmental and economic downstream consequences of dam-collapse
(e.g. number of lives lost, long-term environmental consequences, etc.).

Ground Motion for SEE Performance Level

The SSE is the most severe earthquake ground motion for which a dam must be able to endure
without uncontrolled release of material and for which the dam should be designed. Under the SSE
ground motion, structural damage may be extensive, dam operation may be disrupted, or economic
losses significant (Wieland 2012, 2016). However, the dams’ structural integrity must be maintained,
and uncontrolled release of tailings prevented.
The SEE is usually used for analysis and design for the dam closure and post-closure stages. Common
SEE earthquake ground motion selection criteria are listed in Table 1.

8
Table 1 Ground Motion selection criteria for the SEE Performance Level

Dam Failure
Ground Motion Selection Criteria Ground Motion AEP
Consequence

1/10,000
Extreme The lesser of MCE 84th percentile and AEP
(≈0.5% probability of exceedance in 50 years)

1/5,000
High The lesser of MCE 84th percentile and AEP
(≈1% probability of exceedance in 50 years)

1/2,500
Low The lesser of MCE median and AEP
(≈2% probability of exceedance in 50 years)

CONCLUSION

Current international tailings dam guidelines provide earthquake ground motion selection criteria
for the analysis and design of TSF dams. While these guidelines provide clear definitions of the OBE
and SEE tailings dam performance levels, it is not always clear which specific ground motion level
should be used, particularly for the SSE performance level. The lack of clarity often leads to
confusions, inconsistencies and arbitrary selections of ground motions, because there is not always a
good understanding of the nature and application of the PSHA or DSHA methods.
For the OBE performance level, ground motions with an AEP of 1/475 are generally recommended.
Different PSHA-based AEP values can also be adopted by applying engineering judgment as long as
the dam will remain operational, with any damage being minor and readily repairable.
For the SEE, the dam failure consequence needs to be assessed because it influences the
recommended SSE ground motion value. We suggest that the lower of the MCE 84th-percentile, or
the 1/10,000 AEP and 1/5,000 AEP be used for analysis and design of extreme and high failure
consequence dams, respectively. For a low dam failure consequence, the controlling ground motion
can be the lower of the MCE 50th-percentile (median), or the 1/2,500-AEP.
When MCE ground motions are lower than the PSHA-based probabilistic values for a given AEP, the
controlling earthquake ground motion can be the MCE. Otherwise, the PSHA-based ground motions
should be used for earthquake stability analysis and design of a tailings dam. This approach is
favoured because DSHA-MCE represents the largest conceivable earthquake possible to occur along
a fault or within a seismogenic setting such that larger earthquake ground motions can be expected
to be extremely rare. Consequently, lower AEP values (e.g. 1/10,000) do not apply.
PSHA enables engineers to estimate earthquake ground motions for a specific or range of AEP (e.g.
in some case at AEP of 1/106). However, earthquake ground motions for the analysis and design of a
critical facility like a TSF dam can be limited by the MCE 84th-percentile value that represents the
largest reasonably conceivable earthquake considered possible along a recognized fault or within a
geographically defined tectonic province.

9
When DSHA is not possible because of the lack of identifiable fault sources in very low to low seismic
hazard regions, then engineering judgment should be used to select the most appropriate PSHA-
based ground motion associated with a low AEP.
Although earthquake performance targets for tailings dams are not necessarily the same as used in
seismic design of other mine facilities (e.g. leach pads, waste-rock dumps, and other engineered
structures), the procedure recommended here can be applied based on their failure consequence. The
higher the failure consequence, then the greater the earthquake ground motions that should be used
for engineering analysis and design.

NOMENCLATURE
AEP Annual Exceedance Probability
ANCOLD Australian National Committee on Large Dams
ASCE American Society of Civil Engineering
CDA Canadian Dam Association
DSHA Deterministic Seismic Hazard Analysis
IBC International Building Code
ICOLD International Committee of Large Dams
PSHA Probabilistic Seismic Hazard Analysis
MCE Maximum Credible Earthquake
NZSOLD New Zealand Society of Large Dams
OBE Operating Basis Earthquake
SEE Safety Evaluation Earthquake
TSF Tailings Storage Facility

10
REFERENCES

ANCOLD (2017) Guidelines for Design of Dams and Appurtenant Structures for Earthquake.

ASCE 7-10 (2010). Minimum Design Loads for Buildings and other Structures. American Society of Civil
Engineers.

CDA-Canadian Dam Association (2014) Bulletin: Application of Dam Safety Guidelines to Mining Dams.

IBC 2015 (2015). International Building Code. International Code Council, Inc. USA.

ICOLD (2010) Bulletin 72 Selecting seismic parameters for large dams. Guidelines. Committee on Seismic
Aspects of Dam Design, International Commission on Large Dams.

ICOLD (2016) Bulletin 148 Selecting seismic parameters for large dams, guidelines. Committee on Seismic
Aspects of Dam Design, International Commission on Large Dams.

NZSOLD (2016) New Zealand Society of Large Dams. Dam Safety Guidelines.

Wieland M. (2012) Seismic Design and Performance Criteria for Large Storage Dams. Proceedings of 15
WCEE - Lisbon 2012. Committee on Seismic Aspects of Dam Design, International Commission on
Large Dams (ICOLD).

Wieland M. (2016) Safety Aspects of Sustainable Storage Dams and Earthquake Safety of Existing Dams.
Engineering 2 (2016) 325–331. Elsevier. Committee on Seismic Aspects of Dam Design, International
Commission on Large Dams (ICOLD). http://dx.doi.org/10.1016/J.ENG.2016.03.011.

11
Breach Modes and Effects of Failure of Barahona 1
Tailings Dam in 1928
Jorge Troncoso and Natalia Pizarro
Stantec, Chile

ABSTRACT

The failure of Barahona 1 Tailings Dam occurred as a consequence of the Maule earthquake of
December 1, 1928. The mode the dam breached, and the effects of the following flow failure were
recorded right after the event, mainly by the owner who was in charge at the time of the event, and
the data have been analyzed and complemented by Troncoso et al, in 1991, and in following years.
Several geotechnical investigations have been performed, after repairing the dam and more recently,
to assess and to verify the safety of the actual structure. The purpose of the present paper is to
summarize the findings of the work performed by the authors in relation to: the shear strength
properties of tailings soils that determined the mode of failure; the resulting mechanisms of
progressive ruptures; the extension of the breach within the body of the retaining sand structures
and into the deposit of retained slimes, and resulting lessons derived for dam breach and dam break
analyses. The conclusions include suggestions related to prevention of consequences and mitigation
of risks for similar structures, on the basis of the analyzed results and on the state of knowledge about
surveillance of dams, including technology for monitoring behavior and alerting procedures, for the
protection of Communities.

1
INTRODUCTION ON HISTORY OF THE BARAHONA 1 TAILINGS DAM

The seismic failure of Barahona 1 Tailings Dam (BTD1) occurred after the Maule earthquake of
December 1, 1928, has left valuable information about liquefaction of tailings soils and modes of
rupture of the Tailings impoundment. Troncoso et al (1993) published results of a study of the failure
which included historic details of design and construction of the dam built by the upstream method
by hydraulic fill disposal; the 63 m maximum dam height qualitative seismic response to the Mw =
8.2 event, as described by witnesses; breaching of the dam and consequences of the flow of the tailings
mass. Troncoso et al (1993) also presented preliminary results of a research project conducted in 1991,
on properties of the materials conforming the structure remained in place after the failure of the soils
deposited in the breached sector.
The original deposit was characterized by retaining dykes built with silty sands, partially founded
on the finer soils of the retained slimes, as sketched in Figure 1. As the reservoir filled up, beaches
were left to consolidate, and new raise dykes were added in upstream method. The geometry of the
structure remained after the failures of the breached sector is illustrated in the photograph of Figure
2. One interesting feature of this geometry is the sequence of terraces at different descending levels,
suggesting a progressive type of propagation of the failure modes.

Figure 1 Cross Section of the Dam at Maximum

Figure 2 Detail of Failure Zone. Terraces of Tailings
Height
Soils Left in Place (01.12.1928)

The main objective of this paper is to further analyze the information related to the seismic failure,
especially in relation to the geometry of the breached sector, and to model that with geotechnical
data derived from the 1991 in-situ testing, complemented with some recent tests at the same area,
with the purpose of proposing an explanation for the observed failure modes and to derive criteria
to estimate magnitudes of released volumes of tailings for similar structures exposed to risks of
liquefaction induced by strong earthquakes.

2
MECHANIC AND DYNAMIC PROPERTIES OF FINE TAILINGS

In Situ Measurements

In-situ measurements are the most convenient tool for investigations of actual properties of sands
and silty soils because of the difficulties or virtual impossibility to test representative samples in the
Laboratory. In fact, sands and other non-cohesive soils are vulnerable to densification and other
disturbances of the natural structures when sampled, transported and manipulated before testing.
The resulting uncertainties are big issues for correct interpretations of results and that is the reason
why in-situ testing should generally be prioritized. In complex fabrics as those of BTD1 tailings soils,
where sands and silts prevail and where thin layers or lenses of plastic soils are interbedded with
stiffer sands, this need is obvious.
The mechanic properties of the fine tailings within BTD1 have been investigated in different
campaigns of subsurface exploration mainly with borings and cone penetrations, with in-situ
measurements of SPT and CPTu tests. The residual undrained shear strength, reached after the
seismic loadings, has been calculated from the results of these tests. In particular, the results of the
tests performed in the sector where failure occurred are discussed below. The 1991 penetration tests,
done as part of the research project Fondecyt 833-92, leaded by the senior author, were performed in
the retaining dykes and in the fine soils inside the reservoir, starting from the surface of the upper
layers of the reconstructed and rehabilitated sector. Additional testing has been done by the Owner
in older and more recent campaigns; however, only the results of tests inside or adjacent to the failure
sectors are considered in the present study. The locations of the penetration tests CPTu are shown in
Figure 3 with the identification of each campaign.

a) b)

Failure
Zone

Residual Tailings after

1928 BTD1 Failure

Figure 3 CPTu Tests in Barahona 1 Tailings a) considering the current topography and b) 2018 Google Earth
Capture

3
Shear Strength from CPTu Resistances

CPTu tests generate detailed profiles of qc tip resistance and sleeve resistance, fs. Each parameter is
related to shear strength, however each one is mobilized under different modes of failure and
drainage, therefore leading to different strengths: qc is a general type of mechanism of rupture
developed for bearing capacity under a small cylindrical rod with a conical tip while fs is the lateral
resistance of a thinner surrounding volume of soil strained around the rod. Both parameters are
useful to determine shear strengths of tailings specially in the foundations or close to retaining
structures of the BTD1. The following Figure 4 and Figure 5 summarize the results of the CPTu tests in
the form of cross section of the deposit with profiles of qc and fs versus depth. These profiles have
been constructed with the numerous data points recorded every one centimeter of penetration, at a
rate of 2 cm/s, grouped in segments of average value with lower dispersion.

Figure 4 CPTu Point Resistance (qc) of Fine Tailings

Figure 5 CPTu Sleeve Resistance (fs) of Fine Tailing

4
Both profiles, qc and fs, show clear heterogeneity in strength and interbedding, where qc ranges
between 1.5 and 8.5 MPa while fs ranges between 0.01 and 0.1 MPa. Very soft layers (red) and stiff
layers (yellow to blue) are disposed and form an anisotropic structure. Point resistances qc are lower
in the reservoir than in the dykes, as expected, and the resistance of the slimes also tends to increase
in the proximity of the dykes, also revealing the effect of segregation of coarser soils close to
discharge, which was done until 1928 from the crest of the dam, and predominance of finer silts and
clays further upstream.
The data also show some lower strengths in the tailings deposited after the failure in comparison
with those which remained below the failure surfaces. Although no written information has been
found about the procedures followed to repair the dam after the breach, verbal communications
indicate that fast filling of the depressed areas may have been favored to avoid closer relocation of
ponded waters and risks of new failures. This condition would indeed create weaker structures by
lack of time for more effective consolidation, except at the exposed shallower layers which shall
become consolidated by drying.
The segregation of the tailings soils after discharge is also disclosed when normalized friction ratio,
fs/(qc - σV0), considering σV0 as the total vertical stress, is examined and plotted with normalized qc
in the abacus proposed for classification of soils tested with CPTu (Robertson, 1990), as shown in
following Figure 6. The penetration resistances are summarized for the recorded data of five soil
profiles (91-2, 91-3, 91-4, 91-5 and BH1-5), considering the penetrated tailings strata located between
the elevation 1,565 masl and 1,572 masl, placed below the failure zone according to historic
information. The recorded data of the 91-1 CPTu test was not considered because it is unknown if the
prospected soil correspond to the material used during the breach repair or if it includes tailings
deposited before the dam break.
The coarser soils plot in the upper left region while the finer soils plot in the lower right region. The
soils tested in or closer to the retaining dykes, in profiles 91-5 and 91-4, plot in regions associated with
silty sands and sandy silts, while those in BH1-5, 91-2 and 91-3 plot in regions associated with finer
silty sands, silts and silty clays. The areal extension of the clouds of points, which cover intermediate
regions as they displace from coarser to finer regions does indicate segregation but, most importantly,
the interstratification of the coarser and finer soils in any single profile.

5
 CPTU 91-5 CPTU 91-4
1000 1000
7 8 7 8
9 9

Normalized Tip Resistance

Normalized Tip Resistance

100 100
6 6 5
5

10 4 10 4

1 1 3
3 2 2
1 1
0.1 1 10 0.1 1 10

Normalized Friction Ratio Normalized Friction Ratio

CPTU BH1-5 CPTU 91-2 CPTU 91-3
1000 1000 1000
7 8 7 8 7 8
9 9 9
Normalized Tip Resistance

Normalized Tip Resistance

Normalized Tip Resistance
100 100 100
6 6 6
4
5 5 5
10 10 4 10 4
1 1 1 3
3 2 3 2
2
1 1 1
0.1 1 10 0.1 1 10 0.1 1 10

Normalized Friction Ratio Normalized Friction Ratio Normalized Friction Ratio

Figure 6 Normalized parameters from CPTu Tests performed on 1991 (Classification chart proposed by
Robertson 1990)

MODES OF FAILURE AND MECHANISMS OF PROGRESSIVE RUPTURES

The stratigraphy of the tailings soils disclosed by the records of CPTu measurements allows to
conclude that the modes of failure of Barahona 1 Tailings Dam were determined first by the inherent
weaknesses of the design of the retaining structure, formed by sand dykes built in successive stages
of growth, with axis displaced up and upstream, and partially founded on fine tailings, as shown in
Figure 1. The anisotropic character of the disclosed structures also helps to explain the geometry of
the terraces left in place after the failure, as shown in Figure 1. The weakness of the upstream design
was worsened by the low shear strengths and vulnerability to liquefaction of the retained and
partially supporting slimes.
Based on historic information and on back-analyses of the event, the mode of failure might have been
triggered by increases of pore pressures in the slimes, generated by 2 minutes long duration of strong
accelerations in cyclic loadings, followed by dynamic earth pressures exceeding undrained shear
strength of supporting soils, increasing displacements of upper dyke over weak silts, cracking and
breaching to allow liquefied tailings of upper tailings to flow, overtop and erode underlying dyke.
Loss of confinement allows further liquefaction of second underlain layer of slimes which in turn
flow, overtop and erode third dyke. This progressive rupture would continue until the lower layers
of slimes mobilized shear strengths larger enough to balance the driving stresses corresponding to
the modified geometry. The final geometry of the breach reached a maximum depth of about 40 m.
According with data reported after the failure and quoted by Troncoso et al (1993) and Villavicencio
et al (2014) a total of 2.8 million cubic meter of tailings were discharged, equivalent to 14% of the
impounded volume (20,000,000 m3).

6
The previous analyses indicate that weaknesses of retaining structures built by upstream methods
involve high risks not only for seismic loadings but also for any process which might induce critical
changes of stress condition, which might originate sudden increases of earth pressures, pore water
pressures or shear strains. This is evident after examination of the very low frictional strengths of
layers of soft tailings intruded between stiffer layers and the resulting vulnerability to static
liquefaction.
The results of several limit equilibrium models of the retaining dikes evidence that the most critical
safety factor is related to the sliding wedge failure mechanism. This fact along with the anisotropic
structures of tailings explains the formation of terraces after the dam breach, which advance until
more competent soils and greater confinements are reached.
The stability back-analyses of the upper retaining sand dyke indicates that the minimum required
residual undrained shear strength to avoid the development of the breach was equal to 30 kPa,
considering the sliding wedge mechanism (Equation 1), as shown in Figure 7.

Fs=𝑊 𝐴𝑚 𝑥

T=𝑆𝑢𝑟 𝐿𝐷𝑦𝑘𝑒

Figure 7 Sliding Wedge Failure Mechanism

𝑇 = 𝑆𝑢𝑟 ∙ 𝐿𝐷𝑦𝑘𝑒 ≥ 𝑊 ∙ 𝐴𝑚á𝑥 + 𝐸𝑆𝐻 (1)

Where 𝑆𝑢𝑟 is the residual shear strength of the tailings, 𝐸𝑆𝐻 is the horizontal component of the force
applied by the slimes, 𝐴𝑚á𝑥 is the maximum acceleration estimated for 1928 Earthquake, 𝑊 is the
weight of the dyke and 𝐿𝐷𝑦𝑘𝑒 is the length of the dyke in contact with the tailings.

The flat surfaces of the terraces are consistent with the anisotropic structures of tailings segregated
by sedimentation which follows hydraulic deposition. These anisotropic structures are similar to
stratified varved clays and, therefore, sliding mechanisms of failure occur along the flat contact
surfaces between weaker layers of silts or clays and relatively stronger layers of silty sands. In the
case of existence of a desiccated layer at the surface of the deposit, solid blocks of shrunk tailings may
slide over underlying tailings weakened by seismic movements, overtop through breached crest of
dam and stop close to the foot of the downstream slope, as observed after the failure of Hierro Viejo
tailings, in 1965. The analyses of progressive failure and of the final geometry of the masses of

7
remaining tailings have been also observed after the seismic failures of the tailings dams in 1965,
specially in El Cobre (Dobry and Alvarez, 1967; Troncoso et al, 2017).
The analyzed modes of failure with stable remaining geometry and limited discharge, resulting after
mobilization of available shear strengths, makes it necessary to assess in a more realistic form the
consequences of breaches than those resulting when lower strengths and more pessimistic
mechanisms of failure and stabilization are considered, such as derived from general information of
assorted cases of failures (Rico et al 2008).

EFFECTS OF BTD1 FAILURE

The analysis of historic reports related to the effects of the failure, between the dam toe and the bridge
over the Cachapoal River, as quoted by Troncoso et al (1993), allows to estimate some useful data for
preliminary calibration of dam break models. Relevant data are summarized in following Table 1,
which includes the distances between the dam and three relevant points, as well as the times of arrival
of the flood wave and the height of the wave.

Table 1 Flood Wave Parameters after 1928 Failure.

Place Distance from Arrival Time of Maximum

BTD1 [km] Flood Wave [min] Height of Wave
[m]
Dam Toe 0 3 55
Railroad Station 2.4 6 20
Coya 24 60 6
Bridge of Cachapoal 32 - 1
River
Although data of effects are approximate as based on old reports of witnesses, photographs and the
authors’ reconnaissance of the different sectors of the course, they are nevertheless valuable to
contribute information to the very few case histories of these types of events.
One important fact related to interpretation of effects is to consider the real topography of the courses
where the flow occurred. The courses of creeks and river followed by this flow were mostly narrow
and relatively deep therefore causing the heights of wave to raise and to concentrate the effects in
limited areas. Another feature to consider in relation to the distances to the affected sites is the
availability of high flow rate of natural water courses capable to transport the residues much farther
than in a drier environment.

8
FINAL COMMENTS

The seismic failure of Barahona 1 tailings dam in 1928 has been analyzed as a revealing case history
to understand the mode of failure of an upstream built hydraulic fill structure and the effects of dam
breach and resulting flow of released residues downstream. The shown mechanisms of
displacements, spreading over weakened layers and ending in a set of rather horizontal terraces, at
increasing elevations upstream, indicate the finite extension of the progressive failures and together
with the almost vertical scarps between terraces represent the bounds for limited liquefaction of soil
volumes. The values of shear strength, calculated from CPTu resistances of the structures which
stood in place, result enough to balance the driving stresses and to support the remaining unconfined
structures, stopping the flow after a total of only 14 % of the impounded accumulated volume of
tailings had been discharged. Structural state of the dam, after repairs and refill of the reservoir, plus
construction of operation highway along the crest, 40 year after the failure, has corroborated the
strengths of the consolidated tailings. Part of the satisfactory performance of the present dam has
been linked to the positive effects of evolution of soil properties, mainly due to primary and
secondary consolidation, densification and ageing, as indicated by increases of shear strength versus
depth in the penetration tests performed several years after the failure. Successful rehabilitation with
additional reinforcement berm built in 2009 have made possible to reduce the risks and to reach the
actual safety level of the structure and its capabilities to resist strong dynamic loadings as verified by
the stable performance after the mega-earthquake of Feb27, 2010.
Despite the positive performance of dams like this in recent years there is a need for continuous
attention to evolution of the soil structures, given the inherent vulnerabilities of the original upstream
construction, and to comply with regulatory requests, to be prepared for worse scenarios after
different extreme events, although probability of occurrence may be very low. For this purpose, there
is another lesson learned from the analyses of effects of failures of upstream dams, even under static
loadings, such as occurred recently in Fundao (2015) and Feijao (2019) tailings dams, which relates to
protection of communities living downstream from a tailings dam. For the important purpose of
protection, the review of the records of events that lead to failure, and of the times required for each
event to reach a critical condition, have indicated that it is possible to anticipate occurrence of breach
and to trigger timely responses to mitigate probable consequences. In the case of seismic loadings the
duration of strong motion plus the time required to build up excessive pore pressures, to induce and
propagate deformations along a critical failure surface and finally to breach the dam may be in the
order of few minutes according with compiled history of the 1928 earthquake; additional times are
spent for the wave of liquefied tailings to reach and affect distant locations down the course. For
practical applications such time intervals would allow to alert and to evacuate people from
endangered areas. An adequate seismic instrumentation system should be installed in the structure
to detect earlier arrivals of strong components of ground motion, with accelerometers and dynamic
piezometers to detect corresponding increases of pore pressures, and properly located alarm devices
to trigger emergency protocols.

9
ACKNOWLEDGEMENTS

The authors appreciate the support of El Teniente Division of Codelco for the execution of the in-situ
testing of the research project of 1991 and the permission to add results of more recent geotechnical
investigations in BTD1. Thanks are also extended to our colleagues in Stantec, specially Eng. Gabriel
Troncoso Boys, President of the Chilean chapter of the International Commission on Large Dams
(ICOLD), for their cooperation and review of the manuscript.

REFERENCES

Dobry, R., & Alvarez, L. (1967) ‘Seismic failures of Chilean tailings dams’, Journal of Soil Mechanics
& Foundations Div.

Rico, M., Benito, G., & Diez-Herrero, A. (2008) ‘Floods from tailings dam failures’, Journal of
hazardous materials, 154(1-3), 79-87.

Robertson, P.K. (1990) ‘Soil Classification Using the Cone Penetration Test’, Canadian Geotechnical
Journal, Volume 27.

Troncoso, J. H., Vergara, A., & Avendaño, A. (1993) ‘Seismic Failure of Barahona Tailings Dam’,
International Conference on Case Histories in Geotechnical Engineering. 34.
http://scholarsmine.mst.edu/icchge/3icchge/3icchge-session02/34

Troncoso, J. H., Verdugo, R., Valenzuela, L. (2017) ‘Seismic Performance of Tailings Sand Dams in
Chile’, 16th World Conference on Earthquake Engineering, Santiago Chile, January 9th to 13th 2017, 4230.

Villavicencio, G., Espinace, R., Palma, J., Fourie, A., & Valenzuela, P. (2013). Failures of sand tailings
dams in a highly seismic country. Canadian geotechnical journal, 51(4), 449-464.

10
Released Volume Estimation for Dam Break Analysis
Hugo Quelopana
Delfing, Chile

ABSTRACT

The dam break analysis is a tool in the field of risk studies and emergency plans associated with a
Tailing Storage Facilities (TSF). The models to simulate the development of the breach and its run-
out have been increasingly refined in consideration of the computational advance. However, to date,
the tailings volume released during the failure event – one of the most important input variables in
these dam break models – does not have a standard of how it should be estimated. This situation is
due to the high complexity of the process that is intended to be addressed.
The determination of the released volume can be addressed from two approaches: (i) based on a
definition of the failure surface associated with a certain collapse scenario or (ii) based on historic
tailings dam failures analysis. The first approach requires a thorough analysis of the TSF (including
tailings characterization) and an important discussion of which failure modes are the most
representative. The second approach uses relationships obtained from dam failures databases and
given its simplicity it is a commonly used, or, it is taking into consideration as a reference value for
the first approach.
This paper presents an empirical model to estimate the released volume to be used in dam break
analysis. This model is within the second approach stated above and aims to be an improvement over
the existing relationships. In this context, a key step in the definition and calibration of the model was
the collecting of an updated historical database that included the detailed review of 35 cases of TSF
failures.

1
INTRODUCTION

The fundamental purpose of a Tailings Storage Facility (TSF) is to dispose and contain safely in both
physically and geochemically the tailings material. Despite the advance in the TSF design and
operation, the facts reveal that catastrophic failures continue to occur. In this context, the
improvement of the dam break analysis, which is a highlighted tool in the field of risk studies and
emergency plans associated with a TSF, acquires great importance.
A variable that is key in any simulation of dam break is the tailings volume released during the failure
event. There are two ways to estimate this volume: (i) based on a definition of the failure surface
associated with a certain collapse scenario which requires a thorough analysis of the TSF (including
tailings characterization) and a deep discussion of which failure modes are the most representative
or (ii) based on the use of empirical relationships, such as those developed by Rico, Benito and Díez-
Herrero (2008), Tocher et al. (2014) and Concha & Lall (2018), which are obtained from the analysis
of the historical tailings dam failures.
This paper presents a new empirical model to estimate the released volume and aims to be an
improvement with respect to the empirical relationships previously cited. The development of the
article encompasses three sections: (i) basis of the empirical model that defines the key variables to
be used, (ii) compilation and details of the updated database of failure cases and (iii) presentation of
the empirical model, including its calibration from the database and its application through a
probabilistic approach.

BASIS OF EMPIRICAL MODEL

When analyzing the phenomenon of a TSF failure, a series of variables or aspects that relate to the
volume released during the event can be identified, standing out: failure type (causes), tailings
characterization, dam type, deposit geometric characteristics and situation/size of the pond. These
variables impact the volume released according to their contribution to the capacity of tailings
material to flow. As an example, within the "tailings characterization" aspect, it is expected that a
deposit that has tailings that have been deposited with a low solids content or that have a low level
of consolidation (high rate of rise) will tend to release a volume greater than that corresponding to a
deposit in which the tailings are highly thickened or have experienced a high consolidation.
Unfortunately, there are very few historical cases of failures in which all of the variables already
identified were registered. Then, the empirical model is restricted to the variables that can be obtained
for a minimum number of cases. In consideration of the above, the key variables that have been
considered in this work are the failure type, the dam type and the deposit geometry characteristics.
In relation to the failure type the classification done by ICOLD (ICOLD, 2001) has been used which
focuses on the incident cause: slope instability, seepage, foundation, overtopping, structural,
earthquake, mine subsidence and erosion. This approach is limited because it does not allow to
specify differences between cases of the same group (for example, it does not say anything about the

2
seismic action magnitude in the case of earthquake) but it has as a great advantage that allows, given
its simplicity, to collect an reasonable minimum number of cases for the database.
For the dam type the ICOLD guideline is also adopted, classifying the dam by means of two
components: growth method (upstream, downstream, etc.) and materiality (tailings, earthfill, etc.).
Finally, the deposit geometry characteristics are taken into account by means of the total
impoundment volume “VT” and the dam height "h".

DATABASE

Description

Table 1 shows the database of 35 TSF failures cases that was collected taking into account the capture
of the key parameters set up in previous section and the released volume “VR”. It is noted that VR y
VT are defined for this work as tailings volume not including the volume of water from the deposit
pond. The Table 1 data was obtained from previous similar studies (Rico, Benito & Díez-Herrero,
2008; Concha & Lall, 2018 and ICOLD, 2001) and the sources pointed out in reference section. The
abbreviations used for the failure and dam types can be seen at the bottom of the table.
The data of every failure case was verified revising – in most cases – the initial source of information.
Then the data was contrasted between different sources and a check was made on the reasonableness
of the values (for example, that the total impoundment volume was consistent with the area and
height of the deposit). From this effort the following findings are highlighted:
 Various cases, indicated in previous databases, were not included because the variables VR
or VT are missing (cases of Galena Mine, Itabarito and La Patagua New Dam), the VT does
not match the dimensions of the deposit (cases of Phelps-Dodge and Veta del Agua No1),
or else, the released volume corresponds to, or is like, water (cases of Aitik Mine, Baia Mare,
Bonsal, Buffalo Creek, Churchrock, Omai and Cities Service).
 The values of the variables were updated in some cases the with respect to the previous
databases. As an example, for the Mount Polley case, the values indicated in some
references of ~24 and 74 Mm3 for VR and VT include the 10.6 Mm3 pond that was present
at the time of the failure (WISE, 2019); by subtracting this last volume it obtains what is
indicated in Table 1. The updated cases are indicated in Table 1 with the notation (U) in the
name of the mine/deposit.
 There are 12 new cases (not included in the previous databases) which are indicated in
Table 1 with the notation (N) in the name of the mine/deposit.

3
Table 1 List of TSF failure cases

No Mine/Deposit Country Year Dam Failure VR Mm3 VT Mm3 Hm VR/VT

1 Arcturus (U) Zimbawe 1978 US-T SI 0.07 0.68 25 0.10
2 Bafokeng South Africa 1974 US-T SE 3.00 13.00 20 0.23
3 Balka Chuficheva Russia 1981 US-CS SI 3.50 27.00 25 0.13
4 Barahona (N) Chile 1928 US-CS EQ 2.80 20.00 65 0.14
5 Bellavista (U) Chile 1965 US-T EQ 0.07 0.21 20 0.34
6 Brumadinho (N) Brasil 2019 US-T UN 9.57 11.96 87 0.80
7 Cerro Negro No3 (U) Chile 1965 US-T EQ 0.09 0.26 20 0.33
8 Cerro Negro No4 (U) Chile 1985 US-CS EQ 0.09 1.05 33 0.09
9 Deneen Mica USA 1974 US-CS SI 0.04 0.30 18 0.13
10 El Cobre New Dam (U) Chile 1965 UN-CS EQ 0.50 0.50 19 1.00
11 El Cobre Old Dam Chile 1965 US-T EQ 1.90 4.25 35 0.45
12 Fundao (U) Brasil 2015 US-T SI 32.00 55.00 110 0.58
13 Gypsum Tailings Dam (U) USA 1966 US-T SE 0.14 5.00 11 0.03
14 Hokkaido (U) Japan 1968 US-T EQ 0.09 0.30 9 0.30
15 Huogudu (N) China 1962 US-UN UN 3.30 5.42 19 0.61
16 Kolontar (N) Hungary 2010 DS-F SI 0.65 4.20 22 0.15
17 La Luciana (N) Spain 1960 US-T SI 0.25 1.25 24 0.20
18 Las Palmas (N) Chile 2010 DS-CS EQ 0.17 0.22 15 0.80
19 Los Cedros (N) Mexico 1937 US-T SI 2.50 11.48 35 0.22
20 Los Frailes (U) Spain 1998 DS-E FN 2.00 17.50 27 0.11
21 Los Maquis No3 (U) Chile 1965 US-T EQ 0.02 0.03 15 0.82
22 Merriespruit (U) South Africa 1994 US-T OT 0.55 7.04 25 0.08
23 Mike Horse (N) USA 1975 US-T OT 0.07 0.25 18 0.29
24 Mochikoshi No1 Japan 1978 US-T EQ 0.08 0.48 28 0.17
25 Mount Polley (U) Canada 2014 CL-E FN 13.40 63.40 40 0.21
26 Niujialong (N) China 1985 US-UN OT 0.73 1.10 40 0.66
27 Olinghouse USA 1985 WR-E SE 0.03 0.12 5 0.21
28 Sgurigrad Bulgaria 1966 US-T SI 0.22 1.52 45 0.14
29 Silver King (N) USA 1974 DS-E OT 0.01 0.04 9 0.16
30 Stancil USA 1989 US-E SI 0.04 0.07 9 0.51
31 Stava (U) Italy 1985 US-CS SI 0.20 0.30 28 0.68
32 Unidentified USA 1973 US-E SI 0.17 0.50 43 0.34
33 Veta del Agua No5 (N) Chile 2010 US-CS EQ 0.03 0.08 16 0.34
34 Xiangfen China 2008 US-T SI 0.19 0.29 51 0.66
35 Zlevoto No4 (N) Yugoslavia 1976 US-T SI 0.30 1.00 25 0.30
Average -- -- -- -- 2.25 7.31 30 0.35
(U) Updated data; (N) New case; Dam type: (US) Upstream, (CL) Center line, (DS) Downstream, (WR) Water
retention, (T) Tailings, (CS) Sand Tailings, (E) Earthfill or Rockfill, (F) Fly Ash, (UN) Unknown; Failure type:
(SI) Slope instability, (SE) Seepage, (FN) Foundation, (OT) Overtopping, (EQ) Earthquake, (UN) Unknown

4
Analysis of data

Figure 1, obtained from Table 1, shows the relations between the key parameters and the released
volume values.

100.000 100.000
Released volume (Mm 3 )

Released volume (Mm 3 )

R² = 0.85 R² = 0.48
10.000 10.000

1.000 1.000

0.100 0.100

0.010 0.010

0.001 0.001
0.01 0.10 1.00 10.00 100.00 1.00 10.00 100.00
VT, Total volume (Mm3 ) h, Dam Height (m)

100% 100%
VR/VT

VR/VT

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
EQ SI Others US-T US-CS Others
Failure Type Dam Type

Figure 1 Key parameters against VR or VR/VT a) Total impoundment volume “VT”, b) Dam height “h”, c)
Failure type, d) Dam type

The following clarifications and comments are made with respect to the Figure 1:
 For VT and h, Figures 1a) and 1b), a power trendline is shown and logarithmic axes are
used for better appreciation.
 A strong relationship between VT and VR is observed in Figure 1a).
 As seen in Figure 1b), there is a relationship between h and VR like that between VT and
VR but with a greater scatter.
 The failure type and dam type are contrasted with the VR/VT ratio (fraction of the VT
released) as seen in the Figures 1c) and 1d). In both figures, the 2 main types are indicated
(greater number of cases) and the rest of the types have been grouped as "others", also
indicating for each type the average of VR/VT (thick red lines). The segmented red line
corresponds to the average of VR/VT of the totality of cases (35%, see Table 1).

5
  Figures 1c) and 1d) show, respectively, that the earthquake failure and the “others” dam
type tend to release a greater fraction of tailings than the rest of the types. Care should be
taken in the interpretation of this behavior, firstly because of the limited number of cases
in each category, and secondly because the trend shown is very influenced by the extreme
data. In particular, the case of El Cobre New Dam (VR/VT=100%) significantly increases
the average of the earthquake and “other” dam types; when analyzing this case it can be
seen that the large released volume would be explained in greater proportion by the
tailings characterization – “material less than a year and a half of deposit, unconsolidated
and with high content of moisture” (Dobry, 1965) – than by the failure or dam types.

EMPIRICAL MODEL

Definition of model

Considering the empirical approach applied in this work, the adopted mathematical model is based
on two main assumptions:
 The released volume “VR” is a function of two identified key parameters: the total
impoundment volume “VT” and the dam height "h".
 The influence of each key parameter can be evaluated in a separated way through power
functions.
Then, the proposed function to estimate the VR corresponds to:

𝑉𝑅 = 𝛼 ∙ 𝑉𝑇𝛽 ∙ ℎ𝛿 (1)
Where:
VR = released volume, Mm3.
VT = total impoundment volume, Mm3.
h = dam height, m.
α, β, δ = calibration constants.
The failure and dam types variables have not been included in the Equation (1) because this entails
to develop different relations according the failure and dam type (subdividing the database) and it
has been considered better to perform the calibration of the model with as many cases (together) as
possible. The last comment of the previous section has also been taken into account for this decision.

Model calibration

It was carried out an adjustment by the least squares method, using the database (Table 1) to obtain
the calibration constants of Equation (1). Given that the VR shows a large variation in its magnitude
(0.01 to 32 Mm3), it was decided to perform the calibration on the VR/VT ratio (it targets that the
predicted VR/VT resemble the actual VR/VT). This prevents the calibration from giving importance

6
only to the larger values of VR. Another way to do this, but in a less marked way, is to calibrate the
variable “log VR”, as it is done in Rico, Benito & Díez-Herrero (2008) and Concha & Lall (2018).
The Equation (2) shows the obtained empirical model:

𝑉𝑅 = 0.0612 ∙ 𝑉𝑇 0.809 ∙ ℎ0.544 (2)

The model indicates a proportional relationship between the VR and VT as well as between VR and
h. The values of calibration constants are consistent with the analysis of failure phenomenon which
is also observed in the database: (i) the greater VT, the greater potential volume to be released which
results in a greater volume released in failure case and (ii) the higher h, the greater potential exit area
of tailings when the dam loses its containment capacity.
The application of the model is direct by just entering the values for the key variables in Equation (2)
in the right units. It is noted that h must be measured vertically in the dam crest axis (it is intended
to assess the tailings exit area as indicated in the previous paragraph).

Accuracy of the model

Figure 2 presents the fit of the model to the calibrating data (R2=0,91). The accuracy is similar, though
better, than other previous empirical/statistical approaches. Table 2 shows the performance of 3 other
models in addition to the new one by means of R2 parameter and RMSE (root mean squared error).
When considering “all data” (Table 1), as expected, the empirical model improves the level of
prediction versus previous models. To have a fairer comparison, the comparison considering only
the data that are common (equal) between the models has been added (cases are identified according
to their number in Table 1). With this second comparison, the better prediction of the new model
presented is maintained.

100.00
Predicted VR (Mm 3 )

10.00

1.00

0.10

0.01

0.00
0.00 0.01 0.10 1.00 10.00 100.00
Actual VR (Mm 3 )

Figure 2 Fit of the empirical model

7
Table 2 Predicted performance of empirical model and other approaches

Quelopana 2019 Rico 2008 Tocher 2014 Concha 2018

Data
R 2 RMSE R 2 RMSE R 2 RMSE R2 RMSE
All data 0.91 2.31 0.70 3.21 0.64 5.89 0.70 3.30
2, 11, 24, 27, 28, 30, 32 0.95 0.34 0.91 0.68 - - - -
11, 24, 27, 28, 30, 32 0.86 0.30 - - 0.08 0.66 - -
2, 3, 9, 11, 24, 27, 28, 30, 32, 34 0.88 0.57 - - - - 0.86 1.34

The deviation observed in Figure 2 would be explained for the overlap or mix of following
circumstances:
 The limitations of the key variables considered to explain the VR obtained in the failure
events. In this regard, according to the author, the tailings characterization is a significant
aspect, if not the most important, in explaining the volume of tailings released when a
deposit fails. Unluckily due to the lack of information this aspect could not be included into
the analysis.
 The limitations of the chosen mathematical model due to its structure (power functions).
 Errors in database. The in-depth review of the information made it possible to improve the
database, but it is not possible to ensure that the original sources also have inaccuracies.
Likewise, it was not possible to check that the values of h comply in all cases with the form
of measurement given for this variable in the "model calibration" section.

Probabilistic approach

A probabilistic approach can be carried out to manage the level of deviation from the adjustment
achieved. This approach should be seen as a way to incorporate the uncertainty of the VR estimation
since it was not possible to include in model all the aspects that define its value. As a first step it
defines the “ratio deviation” as show the Equation (3):
𝑉𝑅 𝑉𝑅 𝑉𝑅𝐴 −𝑉𝑅𝑃
𝑅𝑑 = 𝑉𝑇 − 𝑉𝑇 = (3)
𝐴 𝑃 𝑉𝑇
Where:
Rd = ratio deviation.
VR/VTA = actual VR/VT ratio.
VR/VTP = predicted VR/VT ratio.
A mean and standard deviation of 0 and 0.232 respectively are obtained for ratio deviation by
applying the Equation (3) to the database. Then, it is possible to estimate the VR with a certain
probability of occurrence adopting that the Rd has a normal distribution with mean of zero and
standard deviation of 0.232:
𝑉𝑅𝑜(𝑥) = 𝑉𝑅𝑃 + 𝐹(𝑥; 0; 0.232) ∙ 𝑉𝑇 (4)
Where:
VRo(x) = predicted VR with a probability of occurrence of x, Mm 3.

8
VRP = predicted VR according Equation (2), Mm3.
F(x;0;0.232) = inverse of the normal cumulative distribution for the probability of x (mean of 0 and
standard deviation of 0.232).
VT = total impoundment volume, Mm3.
The deviation ratio is used in probabilistic approach, instead of the difference between actual and
predicted VR, because the calibration was performed under the VR/VT ratio and, therefore, the
accuracy of model responds to this last variable. It is also noted that a Gumbel distribution has a
better fit to deviation ratio but numerically the difference with normal distribution is not significant
(both distributions pass the chi-squared test) so it is not justified a more complex application
considering what is attempting to model.
Figure 3 shows the sensitivity analysis carried out to investigate the influence of key variables,
including the probabilistic approach (in figure “Pocc” is probability of occurrence). A base case is
defined as starting point (VT=50 Mm3 and h= 30 m) and for each case the respective key parameter is
varied. It is observed that the probabilistic approach allows have a first knowledge of the variation
expected between forecasted by the model and what could happen in field.

50 50
Predicted VR (Mm 3 )
Predicted VR (Mm3 )

Pocc: 80% Pocc: 80%

40 40
Pocc: 50% Pocc: 50%
30 Pocc: 20% 30 Pocc: 20%

20 20

10 10

0 0
0 10 20 30 40 50 60 70 80 90 100 0 20 40 60 80 100 120
VT, Total volume (Mm3 ) h, Dam Height (m)

Figure 3 Sensitivity analysis regarding the key variables a) VT b) h

CONCLUSIONS

An empirical model has been developed to address the estimation of the volume released during a
TSF failure event, which is one of the most important input variables in dam break analysis.
The presented model allows obtaining quickly a first estimate of released volume. The model is an
improvement with respect to the previous empirical relationships which is explained by the fact that
a second variable is incorporated into the model (the previous relationships only use one variable)
and that the database for calibration was improved and expanded to 35 cases.
The appropriate weighting of particular characteristics of each deposit should be considered to set
the final release volume. Special care should be taken in pondering aspects such as the tailings

9
characterization or when the key variables used are outside of the range of database. In this respect,
the probabilistic approach allows to limit the VR range for decision making.

ACKNOWLEDGEMENTS

Special thanks to Jaime Parra (Delfing) for his support in the deliberations of some topics of this
article.

REFERENCES

Alonso, E., Gens., A. (2006) ‘Aznalcóllar dam failure. Part 1: Field observations and material
properties’, Géotechnique, 56, No. 3, 165-183.

Ayala, F. (2004) ‘La rotura de la balsa de residuos mineros de Aznalcóllar (España) de 1998 y el
desastre ecológico consecuente del río Guadiamar: causas, efectos y lecciones’, Boletín Geológico y
Minero, Instituto Geológico y Minero de España, 115 (4), 711-738.

Bencistà, M, (2011) ‘CFD and GIS based hazard mapping: case study of the red sludge disaster
occurred in Hungary, 2010’, Universitá Degli Studi di Firenze.

Blight, G. (1997) ‘Destructive mudflows as a consequence of tailings dyke failures’, Proceedings of the
Institution of Civil Engineers, Geotechnical Engineering, 125, 9-18.

Brumund, W. (1984) ‘Failure of micaceous waste tailings dam’, International Conference on Case
Histories in Geotechnical Engineering, 16.

Castro, G., Troncoso, J. (1989) ‘Effects of 1985 Chilean earthquake on three tailings dams’, Proceedings
of the Fifth Chilean Conference on Seismology and Earthquake Engineering, Santiago, Chile.

Concha, P., Lall, U. (2018) ‘Tailings dams failures: updated statistical model for discharge volume
and runout’, Environments, 5, 28.

DICTUC (2010) ‘Evaluación preliminar de contingencia en tranque de relaves mina Veta del Agua,
sector Cajón El Sauce, Comuna de Nogales, Región de Valparaíso: Fase I, informe final’.

Dobry, R. (1965) ‘Efectos del sismo de marzo de 1965 en los tranques de relaves de El Cobre’, Revista
IDIEM, 85-107.

Dobry, R., Alvarez, L. (1967) ‘Seismic failures of Chilean Tailings dam’, Journal of the Soil Mechanics
and Foundation Division, ASCE, SM 6, 237-260.

Fernández-Naranjo, F., Rodríguez, V., Rodríguez, R., Alberruche, M., Arranz, J., Vadillo, L. (2015)
‘Liquefaction susceptibility assessment and study of “La Luciana” tailings dam fault (Spain, 1960)
based on historical documents’, DYNA, 82 (189), 189-198.

10
Fourie, A., Blight, G., Papageorgiou, G. (2001) ‘Static liquefaction as a possible explanation for the
Merriespruit tailings dam failure’, Canadian Geotechnical Journal, 38, 707-719.

Grimalt, J., Ferrer, M., Macpherson, E. (1999) ‘The mine tailings accident in Aznalcollar’, Science of the
Total Environment, 242, 3-11.

Hydrometrics, Inc. (2007) ‘Engineering evaluation/cost analysis for the Mike Horse dam and
impounded tailings, Lower Mike Horse Creek, Beartrap Creek and the Upper Blackfoot River,
floodplain removal areas, Upper Blackfoot Mining Complex’.

ICOLD Committee on Tailings Dams and Waste Lagoons (2001) ‘Tailings dams risk of dangerous
occurrences. Lessons learnt from practical experiences’, Bulletin 121. International Commission on
Large Dams.

Ishihara, K. (1984) ‘Post-earthquake failure of a tailings dam due to liquefaction of the pond deposit’,
First International Conference on Case Histories in Geotechnical Engineering, Paper 13.

Macías, J., Corona, P., Sanchéz, J., Martínez, M., Garduño, V., Capra, L., García., F., Cisnero, G. (2015)
‘The 27 May 1937 catastrophic flow failure of gold tailings at Tlalpujahua, Michoacán, Mexico’,
Natural Hazards and Earth System Sciences, 15, 1069-1085.

Penman, A. (2001) ‘Risk analysis of tailings dam constructions’, Conference Proceeding—Safety of

Mining Dams Seminar, Swedish Mining Association, 37-53.

Rico, M., Benito, G., Díez-Herrero, A. (2008) ‘Floods from tailings dam failures’, Journal of Hazardous
Materials, 154, 79-87.

Rourke, H., Luppnow, D. (2015) ‘The risks of excess water on tailings facilities and its application to
dam-break studies’, Tailings and Mine Waste Management for the 21st Century, Sydney, Australia.

Sernageomin (2016) ‘Catastro de depósitos de relaves en Chile (actualización 27-03-2018)’, Ministerio

de Minería, (https://www.sernageomin.cl/datos-publicos-deposito-de-relaves).

Shakesby, R., Whitlow, J. (1991) ‘Destructive mudflows as a consequence of tailings dyke failures’,
Environmental Geology and Water Sciences, Vol. 18 No. 2, 143-153.

Tocher, R, O’Brien, J., Chao, K., Overton, D. (2014) ‘Predicting tailings dam failure and release
volumes’, Proceedings of the 18th International Conference on Tailings and Mine Waste Tailings and Mine
Waste, Colorado.

Troncoso, J., Vergara, A., Avendaño, A. (1993) ‘Seismic failure of Barahona tailings dam’, Third
International Conference on Case Histories in Geotechnical Engineering, Paper 34.

Vale (2016) ‘Barragem de mineraçao’, Diretoria de ferrosos sul – Ger. Geotecnia de Mina de Hidrogeologia
Sul, (https://www.slideshare.net/comcbhvelhas/barragens-de-mineracaovale).

Van Niekerk, H., Viljoen, M. (2005) ‘Causes and consequences of the Merriespruit and other tailings-
dam failures’, Land Degradation & Development, 16, 201-212.

11
Vick, S. (1996) ‘Failure of the Omai tailings dam’, Geotechnical News.

Villavicencio, G., Espinace, R., Palma, J. Fourie, A., Valenzuela, P. (2014) ‘Failures of sand tailings
dams in a highly seismic country’, Canadian Geotechnical Journal, 51(4), 449-464.

Wei, Z., Yin, G., Wang, J., Wan, L., Li, G. (2013) ‘Design, construction and management of tailings
storage facilities for surface disposal un China: case studies of failures’, Waste Management and
Research, 31, 106-112.

WISE (2019) ‘Chronology of mayor tailings dam failures (last updated 13 Mar 2019)’, World
Information Service on Energy – Uranium Project, (http://www.wise-uranium.org/mdaf.html).

Zanbak, C. (2010) ‘Failure mechanism and kinematics of Ajka tailings pond incident’, (http://
www.komplextrade.hu/public_html/failure_mechanism_and_kinematics_of_ajka_tailings_pond_in
cident.pdf).

12
The Geotechnical Properties of Filtered Tailings and
Waste Rock Blends
Ralph Burden1, David Williams2, Ward Wilson1 and Michael Jacobs3
1. University of Alberta, Canada
2. University of Queensland, Australia
3. Goldcorp, Canada

ABSTRACT
Almost all mines produce slurried tailings as a by-product from the extraction process.
Conventionally, these are stored behind dams, which are often constructed from the tailings
themselves. These structures can present a challenge for reclamation and closure, and are a long-term
risk and liability. As large, high-profile failures continue to occur, tailings dams are increasingly
being seen as unacceptable by many stakeholders. To address this problem, an increasing trend in
the industry is deposition of tailings in self-supporting “dry stacks”, typically using pressure
filtration to rapidly de-water the tailings before stacking. Another emerging technology is co-
disposal; blending of waste rock and tailings to create a new, engineered material with favourable
properties. Combining these techniques offers an attractive solution to mine waste management; for
example, the addition of waste rock to a filtered tailings stack has the potential improve stability, and
reduce overall waste volumes. However, very little work has been published on the geotechnical
properties of these blends. This paper presents a rational basis for prediction of the geotechnical
properties of filtered tailings – waste rock blends based upon the blend structure and configuration.
Preliminary results from a series of shear strength and consolidation tests are also presented. It is
shown that the addition of waste rock to filtered tailings stacks significantly increases the shear
strength and reduces the pore pressure response during placement, allowing higher lifts to be stacked
safely.

1
INTRODUCTION

Waste management and its associated problems is arguably the biggest environmental challenge
facing the mining industry. Mining is an inherently waste-producing activity; in open pit mining,
large quantities of overburden material are often stripped before ore can be mined. Furthermore, the
vast majority of the ore also becomes a waste product; even high grade deposits contain only a few
grams per ton of recoverable metal. In typical metal mining operations, ore is recovered in a wet
extraction process. The by-product, “tailings”, is discharged from the mill in the form of a fluid slurry,
typically about 20-40% solids by mass. Conventionally, these are stored behind dams often
constructed from the tailings themselves. These structures can present a challenge for reclamation
and closure, and are a long-term geotechnical risk and liability. As large, high-profile failures
continue to occur, such as the recent Fundão (Morgerstern, Vick et al. 2016) and Córrego do Feijão
disasters, tailings dams are increasingly becoming seen as unacceptable by many stakeholders.
A promising new approach is “dry stacking”; deposition of tailings in a self-supporting pile that has
the potential to eliminate or reduce the need for a dam. This has now been successfully demonstrated
in many commercial-scale metal mining operations (Wickland and Longo 2017). In the vast majority
of cases, creating stackable tailings is achieved by using filtration techniques. Another technology
that is gaining traction in dry-stacking is waste rock and tailings co-disposal. Combining these
techniques offers an attractive solution to mine waste management. Adding co-mingled waste rock
to a filtered tailings stack can improve stability, as well as reduce overall waste volumes.

BACKGROUND
Densified tailings technologies are broadly divided into 3 categories based on mechanical properties
and method of dewatering: Thickened Tailings (TT), typically around 50-65% solids, Paste Tailings
(PT), typically 70-80% solids and Filtered Tailings (FT) (>80% solids). ¡Error! No se encuentra el
origen de la referencia. summarises the properties, and pros and cons, of different densified tailings
technologies.

2
 Figure 1 The tailings continuum (after Davies and and Rice (2001) and Jewell and Fourie (2006))

TT and PT technologies have many potential benefits; namely reduced water consumption, reduced
waste volumes and improved geotechnical stability, whilst still maintaining relatively low operating
costs and good geochemical performance (Jewell and Fourie 2006, Williams, Seddon et al. 2008).
However, some form of containment structure is still required for deposition of TT of PT alone.
Construction of a stable, self-supporting deposit requires filtration or co-disposal technologies.
Filtered tailings “dry stacking” is becoming an increasingly popular alternative to traditional
methods of tailings disposal and is now widespread in practice (Davies and Rice 2001, Jewell and
Fourie 2006, Wickland and Longo 2017). Pressure- or vacuum-filtration is used rapidly dewater the
tailings. In a typical operation, filter cakes are dried so they are in the unsaturated condition,
transported to the impoundment by truck or conveyor, and then spread and compacted using
equipment. This can be a costly and time-consuming process. Perhaps as a result, FT technologies
have only been successfully applied at small- to medium-scale operations. The largest mines
currently using FT technologies have a throughput around 20,000 tonnes per day.

3
CO-DISPOSAL MIX DESIGN AND BLENDING

In general terms, “Co-disposal” refers to the disposal of waste rock and tailings in the same place.
Co-disposal has been applied in many forms with varying degrees of mixing. These include pumped
co-disposal, waste rock inclusions in a tailings impoundment, deposition of tailings into waste rock
cells, and layered co-disposal. A good overview of these techniques is given by Bussiere (2007); a
summary of several co-disposal case histories in given by Habte and Bocking (2017). More recent
research has focussed on producing homogenous blends to create an engineered material which has
favourable properties, known as “paste rock” (Wilson et al. 2008). Blended co-disposal is the focus of
this paper.
¡Error! No se encuentra el origen de la referencia. shows the possible structural configurations of
“paste rock”-style blends. The optimum mix ratio was found to be the “just filled” point, where
tailings slurry fills void spaces between the waste rock particles

Figure 2 Blend configurations (After Wickland et al. 2006)

Wickland et al. (2006) developed a theoretical basis for the design of “paste rock” blends. The
theoretical relationship between mix ratio and density is shown in ¡Error! No se encuentra el origen
de la referencia.. Maximum density occurs at the “just filled” point. The optimum mix ratio that
corresponds to this point is a function of tailings density and the maximum porosity of the waste
rock “skeleton”.

4
 Figure 3 Schematic showing mix ratio – density relationship for fine tailings and waste rock blends (After
Wickland et al. 2006)

The principle behind “paste rock” blends is to create an engineered material which retains the high
shear strength and low compressibility of the waste rock skeleton, combined with the low
permeability and high water retention properties of the tailings. The focus was more on mitigation
of ARD from waste rock dumps, or for use as a cover material, than for dry-stacking. Any non-
segregating tailings, typically paste or thickened, could be used. However, the need for a prescriptive
“optimum” mix ratio, which is often not compatible with the mine plan, and the high costs compared
to traditional methods of tailings disposal, have generally been prohibitive to the commercial
application of this technology.
Filtered tailings and waste rock blends have the potential to be geotechnically stable at a much wider
range of mix ratios, since they do not rely upon a continuous waste rock skeleton. This makes this a
more viable and robust technology than thickened or paste tailings and waste rock blends. In practice,
mines do not usually produce waste materials in the perfect ratio, and the ratio of waste rock and
tailings produced usually varies throughout the life of the mine.
An alternative approach currently under development is co-disposal of filtered tailings and waste
rock. Addition of waste rock has the potential to improve the stability of a filtered tailings deposit by
increasing shear strength, and reducing the build-up of pore pressures during stacking. This may
enable the material to be stacked rapidly in high lifts without prior drying of the filter cakes, which
could make large-scale “dry-stacking” economical at mines with throughputs in excess of 100,000

5
tons per day. Blending could be achieved using the same system of conveyors that is used to transport
the materials.

GEOTECHNICAL PROPERTIES

This section gives a brief overview of some of the results of a programme of laboratory tests on
filtered tailings and waste rock blends. All mix ratios are given as rock : tailings by dry mass. Samples
were blended using a concrete mixer to produce homogenous mixtures. Whilst it is recognized that
this is not always representative of field conditions, “perfect” blending is targeted for laboratory trials
in order to produce good quality, repeatable results. Due to constraints in equipment size, shear
strength tests were scalped to -37 mm and compression tests were scalped to -19 mm. Further details
on blending procedure is given by Burden, Wilson et al. (2018).

Index Properties

Specific gravity of solids and moisture contents for the waste rock and tailings used in this study are
given in Table 1.

Table 1 Index Properties

Material Gs w (%)

Waste Rock 2.76 2.1

Filtered Tailings 2.73 19.3

Particle size distributions for the filtered gold tailings and waste rock used in the study are given in
¡Error! No se encuentra el origen de la referencia.. Waste rock was scalped in the field to -100 mm.

6
 Figure 4 Particle size distributions of waste rock and filtered tailings

Shear Strength

Figure 5 shows the results of large scale direct shear tests on blends at a range of mix ratios, as well
as filtered tailings and waste rock alone. A 300 x 300 mm shear box was used. The blends were tested
at their natural moisture contents. A low strain rate of 0.1 mm/min was used to ensure fully drained
conditions. Tests were carried out at normal stresses of 250, 500 and 1000 kPa.
Addition of waste rock to filtered tailings resulted in a significant increase in shear strength. Some
rock breakage occurred at 1000 kPa. The results suggest that, at higher stresses where rock breakage
may occur, the strength of the blended materials is greater than rock alone. This is probably because
the tailings occupy void space between rock particles, reducing the stress at clast-to-clast contacts
and resulting in less rock breaking. Significance of this effect would be dependent upon the strength
of the waste rock used. It should be noted that for the materials under consideration, the “just filled”
point (shown in Figure 2) occurs at a mix ratio of around 1.8. Thus, it can be seen that addition of a
relatively small amount rock increases the shear strength significantly, even when the blend is still in
the “floating” condition. An increase in strength at the “just filled” point may be expected, however,
in the tests here, no improvement in shear strength is evident for mix ratios above 1:1.

7
 Figure 5 Large Scale direct shear test results for filtered tailings – waste rock blends

Compression Behaviour

The objective of the compression tests was to simulate the stacking of the material at a constant rate
of and measure the settlement and pore pressure response. Controlled Rate of Loading (CRL) tests
were used, using an incrementally increasing load rate to account for the increasing drainage path
length as the stack rises. Preliminary results are presented here, detailed analysis and discussion of
the method is beyond the scope of this paper.
The consolidation cell used in this study was a Slurry Consolidometer, custom built to the
specifications of the University of Queensland. The cell has an internal diameter of 150mm and is
capable of accommodating samples up to 300 mm high. Axial load is applied via a 10 kN high
precision electromechanical load frame. It is equipped with top and base load cells, a base pore
pressure transducer and further pressure transducers located throughout the height of the cell.
Settlement is measured via an LVTD. The device is computer controlled and any combination of load
steps and load rates may be applied. The Slurry Consolidometer is shown in Figure 6. Drainage is
from the top of the sample only.

8
 Figure 6 Slurry Consolidometer

Figure 7 shows the base pore pressure response normalized with respect to vertical stress measured
at the base. The samples were loaded up to a peak stress of 560 kPa, using an initial seating load of
100 kPa, and thereafter incrementally increasing rate over approximately 2 hours. The load was then
maintained and dissipation of excess pore pressure was monitored. Applied stress is plotted in Figure
7 on the secondary axis. Actual measured pore pressure is shown on the second plot.

9
 Figure 7 Pore pressure response for filtered tailings and filtered tailings and waste rock blends in CRL test

The samples are placed loose in the cell at their natural blended moisture content. Initially, the sample
undergoes compaction and removal of large air voids. When a continuous water phase forms excess
pore pressures develop. The results show that increasing rock content reduces the build-up of excess
pore pressures under compressive loading. This suggests that addition of rock to a filtered tailings
stack reduced the build-up of pore pressures at the base of the stack, allowing for faster placement in
higher lifts.

CONCLUSIONS
A dry stack consisting of co-mingled waste rock and filtered tailings has the potential to be more
geotechnically stable than filtered tailings alone. It has been shown that increasing rock content
increases shear strength, and reduces the build-up of pore pressure during stacking. This may allow
dry stacked tailings to be placed rapidly in high lifts, and make large-scale dry stacking economically
viable.

NOMENCLATURE
Gs specific gravity of soil solids
w geotechnical moisture content, defined as mass of water / mass of solids

10
REFERENCES

Burden, R., D. J. Williams, G. W. Wilson and M. Jacobs (2018). The Shear Strength of Filtered Tailings
and Waste Rock Blends. Tailings and Mine Waste 2018. Keystone, CO.

Bussiere, B. (2007). "Colloquium 2004: Hydrogeotechnical properties of hard rock tailings from metal
mines and emerging geoenvironmental disposal approaches." Canadian Geotechnical Journal 44(9):
1019-1052.

Davies, M. P. and S. Rice (2001). "An alternative to conventional tailings management–“dry stack”
filtered tailings." Proceeding of Tailings and Mine Waste’01: 411-420.

Habte, K. and K. Bocking (2017). Co-disposal practice in mind waste management. Tailings and Mine
Waste '17. Banff, Alberta, Canada.

Jewell, R. J. and A. B. Fourie (2006). Paste and thickened tailings: A guide, Australian Centre for
Geomechanics, The University of Western Australia.

Morgerstern, N., S. Vick and B. Watts (2016). Fundão Tailings Dam Review Panel Report on the
Immediate Causes of the Failure of the Fundão Dam.

Wilson, G., B. Wickland and J. Miskolczi (2008). Design and performance of paste rock systems for
improved mine waste management. Proceedings of the 1st International Seminar on Management of
Rock Dumps, Stockpiles and Heap Leach Pads, Perth, Australia.

Wickland, B. E., G. W. Wilson, D. Wijewickreme and B. Klein (2006). "Design and evaluation of
mixtures of mine waste rock and tailings." Canadian Geotechnical Journal 43(9): 928-945.

Wickland, B. and S. Longo (2017). Mine waste case examples of stacked tailings and co-disposal.
Tailings and Mine Waste '17. Banff, Alberta, Canada.

11
A Different View of Run-Out Distance after Tailings
Impoundment Failure
István Kovács1, Claudio Román1, Pablo Galdeano1 and Marcelo Mussé1
Ausenco, Chile

ABSTRACT

Estimates of run-out distance for tailings impoundments are mandatory and are required by
Authorities in many countries over the world. There are widely accepted differences between water
dams and tailings dams with respect to their failure modes and discharge processes. Empirical
methods based on statistical analysis of past events are available to calculate the discharge volume
and the peak discharge and help engineers to predict potential consequences and assess hazard levels
downstream of the dam. On the other hand, simulation tools are available to even include rheological
parameters and generate more sophisticated models if necessary.
Despite this, this kind of analysis still possesses a strong hypothetical approach and has a lot of
uncertainties. Questions arise such as: What are the most common uncertainties of such a simulation?
Is rheological data so important for tailings modelling, especially when tailings slurry and significant
volume of water are considered? What effect can a nearby water body have on the run-out distance?
What about the accuracy of our prediction models? What about the confidence level of alert times for
contingency plan?
The objective of this paper is to present a different view of tailings related flood hazard mapping.
This approach includes a comparison between simple prediction models, real occurrences and
available post-event models. Well documented tailings disasters showing Newtonian behavior are
highlighted and analyzed from the point of view of their use for contingency plans. Particularities of
these events could affect general thinking about run-out distance and may deserve more attention.

1
MANY FACTORS, MANY DOUBTS

There are many factors included in a dam break and flood analysis and this number of variables
mean a high-level of uncertainty which can have significant effects on the results. The main
uncertainty sources for this kind of study include: i) failure mode and breach development; ii) the
discharge volume; iii) hydrologic issues such as the volume of accumulated water on the tailings
surface; iv) the rheological parameters of the tailings at the time of failure; v) and topography.
Though failure modes are in total uncertainty, this issue has effects on flow characteristics, mainly in
the nearest sectors downstream of the dam. The discharge volume could be estimated by the Rico et
al equation. The accumulated process water can be handled only as an estimated value while
rainwater can be calculated in accordance with local hydrologic methods where uncertainties are
built in due to the probabilistic character and empirical coefficients, for instance, rheological
parameters are generally measured in a laboratory and they can differ from the failure moment’s
values where consolidation, age, moisture content and seismic effects should be considered.
Topography is one of the key factors that determine the quality of flood hazard mapping. In general,
high resolution surface data is available only at the project site. Adequate topography information
downstream the project is rarely available (only in form of lower quality public information). Flood
hazard assessment often has the following problems: downstream area until the most vulnerable
sectors can be quite extense. About this potentially affected sector, some information could be gained
using public-domain digital elevation model (DEM) data, typically in a range of resolution with
contours every 30 m. On the other hand, if high-resolution topographic data were available, technical
problems arise about the handling of this significant amount of information particularly when the
objective is to analyze a large area.
Following historical cases are analyzed to compare the accuracy of typical prediction models with
real registered impacts and to discuss some details with respect to contingency plans that may
deserve major attention.

PREDICTION MODELS VERSUS REPORTED CASES

There are at least two recent and well documented tailings impoundment failures where registered
data allow a comparison between prediction models and real occurrences after failures: the Ajka
failure in Hungary, 2010, and the Fundão dam failure in Brazil, 2015. In the case of the Ajka failure
the post-event simulation was carried out by Józsa, Krámer & Baranya, 2011, using a Newtonian 3D
model calibrated in accordance with real flooded area. The comparison for this case was made
considering both post-event numerical results and satellite image. In the case of Fundão dam failure
only the flooded area indicated on satellite images was considered for the comparison. Both
prediction simulations presented in this paper were carried out by a numerical 2D tool.

2
Figure 1 shows the predicted flood extension (yellow area) where the red polygon corresponds to the
real inundation area (defined from the corresponding Google Earth satellite image), while Table 1
includes a numerical comparison of some important parameters.

Torna Stream

Limit of the real

flooded area

Chamber 10

Devecser

Kolontár

Torna Stream

Figure 1 Ajka failure, Hungary: predicted (yellow zone) versus real flooded area (red line area)

According to the compared data, the following observations can be taken from the Ajka failure:
 The prediction model indicates flow arrival time, depths and velocities which show similar
values in comparison with real occurrences.
 Predicted discharge volume is approximately 15% lower than the reported value.
 There are increasing differences in flood path and extension as we move away from the
dam. For example, the prediction model does not indicate impacts on internal zones of the
city of Devecser. This is probably due to that the lower resolution topography does not
include fine details like the Torna streambed that had an important role to conduct flood
wave (towards the Northern part of Devecser).
 Flood hazard level is very high for people living in Kolontár, especially in the first houses
downstream of the dam. In this sector, the predicted 9 minutes of arrival time (Table 1) will
generate difficulties for any contingency plan.

3
The same 2D hydraulic analysis was carried out for the Fundão dam failure, trying to strictly take
into account the data available before the dam failure. Figure 2 and Table 2 show the prediction
results in comparison with post-event registrations. For this second simulation, observations and
conclusions are:
 Predicted discharge volume is 37% lower than reported data.
 Predicted peak discharge is probably lower than the (unknown) real value.
 Despite the lower predicted volume, the model predicts fairly well the real flood path.
 Unacceptable differences are observed mainly between predicted and real arrival times
due to such topographic elements (depressions and other storage facilities) that have
delaying effects on the flow and, therefore, differences between hydraulic parameters are
also generated.

Table 1 Comparison between prediction and verification model data – Ajka failure, Hungary

Parameter Prediction model Reported data Notes

Discharge volume 1.51 ∙ 106 𝑚3 1.75 ∙ 106 𝑚3 Using the Rico et al equation for
the prediction model. Total
discharge volume for the
verification model was computed
by the Ministry of Defence of
Hungary.

Breach 15 min 15-20 min Berm construction material: grey

development time mud. Fast dam break was
considered in the range of
minutes due to construction
material and the reduced
dimensions of the structure.

Peak discharge 3,351 𝑚3 ⁄𝑠 No available information Simplified hydrograph was

adjusted considering the breach
time and dischargeable volume.

Average flow 2-4 𝑚⁄𝑠 2-3 𝑚⁄𝑠 Kolontár is the first populated
velocity in Kolontár area downstream the dam.

Average flow depth 1.0-5.0 m 2.0-2.5 m

in Kolontár

Flow arrival time to 9 min 6-7 min Time to one-foot depth was
Kolontár considered in the prediction
model as arrival time.

4
Despite the underestimated discharge volume and the use of low-resolution topography, the affected
sectors in both prediction models coincide quite well with the real flood area limits with respect to
Kolontár and Bento Rodrigues villages. Figures 3 and 4 demonstrate the affected zones for both
populated areas respectively.
In both cases Newtonian behavior was identified (Józsa, Krámer & Baranya, 2010; and Morgenstern
et al, 2016) and from this point of view the precision of the outputs is not so surprising, due to all the
above presented prediction results were gained from a simple water simulation model. No
rheological data were added to the models. This hypothesis does not reflect general thinking and it
is a very common practice to estimate and use rheological parameters that finally would have
resulted in a much more optimistic or non-conservative scenario.

Fundão Dam

Limit of the
real flood area

Bento Rodrigues

Santarém Dam

Figure 2 Fundão tailings dam disaster, Brazil: predicted (yellow zone) versus real flood area (red line area)

5
Table 2 Comparison between prediction model and reported data – Fundão dam failure, Brazil

Parameter Prediction model Reported data Notes

Discharge volume 20.3 ∙ 106 𝑚3 32 ∙ 106 𝑚3 Using the Rico et al equation for
the prediction model.

Breach 1.8 hrs “in a matter of hours” Slow dam break was considered
development time in the range of a few hours due to
significant dimensions of the
structure.
3⁄
Peak discharge 3,210 𝑚 𝑠 No available information Simplified hydrograph adjusted
according to breach time and
dischargeable volume.

Average flow depth Some non-official sources

1-8 m No available information
in Bento Rodrigues mention even at 10 m depths.

Prediction value is due to the

Flow arrival time to delaying effects of topographic
2.1 hrs 9 min
Bento Rodrigues imperfections and Santarém dam
retaining capacity.

First impact on houses

Figure 3 Kolontár flood prediction model (left side image) versus after-disaster situation (right side image)

6
 Destroyed houses
zone

Figure 4 Bento Rodrigues flood prediction model (left side image) versus after-disaster situation (right side
image)

CONTINGENCY PLANS AND ALERT TIME

Information obtained from flood hazard mapping helps to establish dam contingency plans and to
define alert time and evacuation procedures. The engineer should be aware that the discharge
hydrograph has relevant influence on model outputs used for this issue. This is because discharge
volume and overland roughness govern the flood area extension, while peak discharge value and
breach development time are responsible for velocities and flow arrival time. The discharge
hydrograph is based on three parameters, such as volume, peak discharge and breach development
time, where volume can be estimated with a moderate-good precision, and breach time and peak
discharge with a poor-moderate precision. If volume is conserved for the hydrograph and breach
time or peak discharge is adjusted, flood area extension can be maximized. On the contrary, the flood
wave arrival time will be considered as a more important output. Due to aforementioned uncertainty
factors, flood area should be considered as the main output while flood wave arrival time should be
handled carefully.
When relevant amount of water is accumulated on tailings surface, the initial flood wave will have
the first impact on downstream infrastructure and it will be produced clearly by a Newtonian
behavior fluid. This impact should determine the contingency plan and will make less relevant the
arrival of a second-stage tailings flood wave that could show non-Newtonian behavior.
Overland roughness coefficient has considerable effect on flow velocity and moderate effects on
depths (flood extension area). This issue is relevant in vegetation covered sectors: wheat fields, bush
and tree-covered downstream valleys can distract engineers using higher coefficient values, while no
one can guarantee that cultivated areas will still remain at the moment of dam failure which can occur

7
many years after the first risk analysis. Natural vegetation can be easily lost due to fires or drought,
which are not uncommon events in Chile, for instance. On the other hand, roughness coefficients
vary as the flood occurs (scouring vegetation and changes in bed geometry) and this kind of
modifications cannot be estimated with any level of certainty.

THE ROLE OF WATER

Table 3 shows some selected historical cases where hydrologic conditions could have an important
role in failure modes and the magnitude of damage. Two criteria were defined previously for this
selection: high number of loss of human lives and/or severe environmental damage. Every case has
its own singularities, but it is easy to conclude that higher fatalities are associated with close
proximity to populated areas downstream of the dam. Locations within a distance of some kilometers
downstream from the dam will bear an extremely high risk even with previously established
contingency plans.
The released material may represent a Newtonian behavior due to the presence of water in any form
(i.e., accumulated rainwater, process water, saturated material by rainfall or additional effects by
snow melt or local springs and streams). Without the presence of water, failures will cause less
serious damage. There are several other cases not detailed in Table 3 where causes of the tailings
impoundment failure are clearly related to rainfall and water management problems (i.e., Bafokeng,
1974; Merriespruit, 1994; Arcturus, 1978). Finally, the question of distance is more important for non-
Newtonian behavior fluids while elevations and velocities developed along the flood area is relevant
for Newtonian behavior fluids and the related contingency plans. This latter can be interpreted as a
“too conservative scenario” for the run-out analysis, but the two cases evaluated in this paper do not
underwrite that kind of approach. Nature can often surpass human imagination.

8
Table 3 Some of the worst tailings disasters (authors’ selection criteria)
Year Name/Place Hydrologic- Loss of Environmental damage Impoundment distance (m)
hydrogeologic human of water bodies
From From
condition (1) lives
populated water
area (2) bodies (2)

1966 Aberfan RF, STR 144 (3) - 500 900

(WAL)

1966 El Cobre IW >200 (4) El Cobre and Melón 1,500 2,700

(CHI) Branches, Aconcagua
River

1966 Zgorigrad RF, IW, STR >107 (5) Leva Creek 1,000 ~ 700
(BUL)

1972 Buffalo STR(?), IW 118 (6) Buffalo Creek, 1,000 1,000

Creek (USA) Guyandotte River

1985 Stava (ITA) IW 268 (7) Stava Gully and Avisio ~ 900 ~ 900
River

1995 Omai (GUY) IW - Essequibo River 43,600 < 100

1998 Aznalcóllar IW - Agrio and Guadiamar 2,900 < 200

(ESP) Rivers

2000 Baia Mare RF, SM - Sasar, Lapus Streams, 700 < 100
(RUM) Somos, Tisza, Danube
Rivers, Black sea

2010 Ajka (HUN) RF, IW 11 (8) Torna Creek, Marcal, 900 1,100
Rába and Danube Rivers

2014 Mt Polley IW - Polley Lake, Hazeltine 17,000 600

(CAN) Creek, Quesnel Lake,
Cariboo River

2015 Fundão IW 19 (9) Gualaxo do Norte, 5,800 7,600

Dam (BRA) Carmo, and Doce Rivers,
Atlantic Ocean

2019 Corrego do Still unknown >272 (10) Paraopeba River < 50 6,500
Feijão (BRA)
(1) Caused by: RF – Rainfall, SM – Snow melt, STR – streams and springs, IW – Saturation, seepage and drainage problems.
(2) Downstream distance of the nearest populated area and/or stream, lake (measured on satellite images).
(3) Data extracted from Disaster at Aberfan (1966).
(4) Data extracted from El Cobre desastre (2019);
(5) Data extracted from Sgurigrad Report (1966);
(6) Data extracted from West Virginia’s Buffalo Creek flood (1972);
(7) Data extracted from Stava mudflow (1975);
(8) Data extracted from Bencistà M. (2011).
(9) Data extracted from IUCN (2018)

9
(10) Data extracted from Wise Uranium (2019)

NATURAL WATER BODIES DOWNSTREAM OF THE DAM

Extremely close distances between tailings impoundments and natural water bodies bear a high
environmental damage risk. Once tailings have entered into a natural permanent water body, the
concept of “run-out distance” loses its validity. Mixing with water, the pollution can travel hundreds
of kilometers. The “run-out distance” concept can be used for overland flow and for flow in dry gully
beds. On the other hand, the above presented red mud case simulation shows clearly the importance
of natural streambeds and its role to “conduct” flood waves within a bigger valley. This example
enhances, again, the importance of more accurate topographic details.

CONCLUSIONS
Post-run-out-analysis of tailings dam failures allows a valuable comparison between registered data
and the results of prediction models. It is an excellent opportunity to revise our methodology and the
confidence level of our outputs.
From the selected and analyzed cases, where significant amount of water was stored within the
impoundment area and therefore tailings were not consolidated, it can be concluded the following:
 Water in any form had a very high importance in many historical failures accompanied by
high number of fatalities or extended and severe environmental damage.
 The analyzed Ajka and Fundão failures in this paper show Newtonian behavior due to the
presence of water at the moment of failure. This behavior will result in a more extended
inundation area than a so called sunny-day and rheology-focused run-out simulation.
 For the case of the Ajka and Fundão disasters, using a simple 2D water run-out model with
free and low-resolution topographic data, it was possible to predict the flood wave path
with high similarity to real inundation areas in order to support contingency plans.
 Creating the discharge hydrograph, the volume conservation approach will maximize the
flood area extension. Adjusting the peak discharge and breach development time a flood
arrival time-approach will be given for the simulation.
 Flood area extension and arrival time are the principal factors to develop contingency
plans. However, this latter parameter is very sensitive to topography details, resolution
and hydrograph estimations (volume conservation or arrival time approach), therefore,
special care is required to evaluate it.
 A reasonable distance between dam and any kind of populated sector downstream will
have high importance and it is independent from Newtonian- or non-Newtonian- behavior
based simulations. Populated areas within several kilometers downstream of the dam, hold
an extremely high hazard due to the few minutes that a hypothetical flow would allow for

10
 evacuations and/or any kind of action to avoid or minimize human loss or environmental
damage.

REFERENCES

Bencistà M. (2011) CFD and GIS based hazard mapping: case study of the red sludge disaster
occurred in Hungary, 2010. Tesi di Laurea, Università Degli Studi di Firenze.

Davies, W. E., Bailey, J. F. & Kelly, D. B. (1972) West Virginia’s Buffalo Creek flood: A study of the
hydrology and engineering geology. Geological Survey Circular No. 667, United States Department
of the Interior, Washington.

El gran desastre del tranque y relave El Cobre que dejó cientos de muertos en Chile, Resumen (2019),
visited on 19th of February 2019, https://resumen.cl/articulos/gran-desastre-tranque-relave-el-cobre-
dejo-cientos-muertos-chile

International Task Force for Assessing the Baia Mare Task Force (BMTF), (2000) Report of the Baia
Mare Task Force, visited on 9th of January 2019,

http://viso.jrc.ec.europa.eu/pecomines_ext/docs/bmtf_report.pdf

Józsa, J., Krámer, T. & Baranya, S. (2011) Az ajkai vörösiszap zagytározó gátjának tönkremeneteléből
származó elöntések 2D numerikus modellezése. Szakértői tanulmány. Budapesti Műszaki és
Gazdaságtudományi Egyetem, Vízépítési és Vízgazdálkodási Tanszék.

Sánchez et al (2018) Impacts of the Fundão Dam failure. IUCN (International Union for Conservation
of Nature, Switzerland. Visited on 19 th of April 2019.
https://portals.iucn.org/library/sites/library/files/documents/2018-038-En.pdf

Luino F., & De Graff J.V. (2012) The Stava mudflow of 19 July 1985 (Northern Italy): a disaster that
effective regulation might have prevented. Natural Hazards and Earth System Sciences, p 1029-1044.

Morgenstern, N. R., Vick, S. G., Viotti, C. B. & Watts, B. D. (2016) Fundão tailings dam review panel.
Report on the immediate causes of the failure of Fundão dam. Visited on 9th of January 2019.
http://fundaoinvestigation.com/

Norambuena, R., Sepúlveda, D., Barrueto, C., Santos, E., Godoy, C. & Campaña J. (2018) Distancia
peligrosa, retroanálisis de fallas en depósitos chilenos mediante métodos numéricos. X. Congreso
SOCHIGE, Valparaíso, Chile.

Report of the Tribunal appointed to inquire into the Disaster at Aberfan (1966). No. 553, Evidence.
London, United Kingdom. Visited on 9 th of January 2019.
http://www.mineaccidents.com.au/uploads/aberfan-report-original.pdf

Rico M., Benito G. & Díez-Herrero A. (2007) Floods from tailings dam failures. Journal of Hazard
Management

11
Summary of the official report about the Sgurigrad (Bulgary) failure on 1 st May 1966. Visited on 20th
of January 2019. http://www.cleanriverstrust.co.uk/bulgarian-tailings-dam-failure/

Szépvölgyi J. (2011) A chemical engineer’s view of the red mud disaster. Visited on 9th of January
2019.
https://www.researchgate.net/publication/264476294_A_Chemical_Engineer's_View_of_the_Red_M
ud_Disaster

Van Niekerk, H. J. & Viljoen, M. J. (2005). Causes and consequences of the Merriespruit and other
tailings-dam failures. Land Degradation & Development 16, 201-212, University of South Africa,
Pretoria.

Vick, S. G, (1996) Failure of the Omai Tailings Dam. Geotechnical News. Visited on 9th of January
2019. https://file.ejatlas.org/docs/omai-gold-mine-tailings-dam-
guyana/Failure_of_the_Omai_Tailings_Dam.pdf

Wéber, B., Menedzsment Fórum (2010) Eye witness of the mud disaster: “Vimos como estaba fallando
todo” (available in Hungarian), visited on 19th of December 2018,
https://mfor.hu/cikkek/vallalatok/_Lattuk__ahogy_szakad_be_az_egesz______megszolal_az_iszapo
mles_szemtanuja.html?fbclid=IwAR3KAJCiW6XgzmASVpj_c-
RqqJM6rxfZDz_JRta7iE9cmKv5Ndm72qo7fSI

WISE URANIUM (2019) The Brumadinho Tailings Dam Failure, visited on 19 th of April 2019.
http://www.wise-uranium.org/mdafbr.html

WWF (2008) La catástrofe de Aznalcóllar. X Aniversario: Una lección aprendida? Visited on 9th of
January 2019. http://assets.wwf.es/downloads/_informe_2008.pdf

12
Liquefaction Assessment of Deposited Thickened
Tailings Using CPTu Soundings Calibrated With
Laboratory Testing: a Critical State Soil Mechanics
Application
Edgar Bard and Guillermo González
Golder Associates, Chile

ABSTRACT
Application of thickened tailings technologies sometimes require for the impounded tailings to
perform as a competent engineered fill (e.g. become the foundation of upstream raises, ensure
stability of tailings beach stacked above the dam crest, etc.). In such cases, the geomechanical
behavior of tailings is a key issue, with liquefaction often becoming the most influencing aspect.
When tailings beaches have a high degree of saturation, stability of the impounded tailings must be
demonstrated under post-liquefaction conditions (specially in seismic regions). This requires that
liquefaction susceptibility and undrained residual shear strengths of tailings be well understood.
To evaluate liquefaction susceptibility of tailings deposits, CPTu soundings are currently the most
used tools. Although CPTu can measure the response of tailings associated with their truly
undisturbed in-situ state it does not directly provide geotechnical parameters related to tailings
behavior. In contrast, while laboratory testing does provide those parameters, they often relate to
disturbed conditions. The integration of both CPTu and laboratory data provide a means to
characterize tailings behavior for the undisturbed in-situ condition.
This paper presented a case study where CPTu and laboratory testing data were combined using the
framework of critical state soil mechanics. The differences between completing a liquefaction
assessment purely based on CPTu results (i.e., based on average properties of soils and typical
correlations for CPTu interpretation) and the assessment of the CPTu results combined with
isotropic, oedometer and triaxial compression monotonic paths are presented.

1
INTRODUCTION

Application of thickened tailings technologies sometimes require for the impounded tailings to
perform as a competent engineered fill (e.g. become the foundation of upstream raises, ensure
stability of tailings beach slope projecting above the dam crest, etc.). In such cases, the geo-mechanical
behaviour of tailings becomes a key issue, with liquefaction often becoming the most influencing
aspect. When tailings beaches have a high degree of saturation, stability of the impounded tailings
must be demonstrated under post-liquefaction conditions (specially in seismic regions). This requires
for tailings liquefaction susceptibility and liquefied residual strengths to be well understood.
For existing tailings deposits, CPTu soundings are currently the most used tool for evaluating
liquefaction susceptibility. Although CPTu can measure the response of tailings associated with their
truly undisturbed in-situ state it does not directly provide geotechnical parameters related to tailings
behaviour. On the contrary, while laboratory testing does provide those parameters, they often relate
to disturbed conditions. The integration of both CPTu and laboratory data provide a means to
characterize tailings behaviour for the undisturbed in-situ condition.

METHODOLOGY

Been & Jefferies approach (2016)

The geotechnical approach to understand tailings behavior according to Been & Jefferies (2016) is
based on the principles of soil mechanics critical state, which allows a physical description of the
liquefaction phenomenon.
The main aspect of this approach lies in the state parameter ψ proposed by Been & Jefferies (1985),
which is defined as the difference between the current void ratio of the material 𝑒0 and the void ratio
associated with its critical state (steady state or perfect plasticity) 𝑒𝑐 , for the current confinement or
mean stress level.

𝜓 = 𝑒0 − 𝑒𝑐 (1)

This parameter allows inferring the trend of volumetric changes expected for tailings, where a state
parameter less than -0.05 indicates dilatant behavior, while values greater than -0.05 indicate
contractive behavior.
The work performed consisted on the specific analysis of a series of CPTu soundings to a depth of 25
m carried out at different locations in the impoundment of a tailing dam. This analysis was
complemented with pore pressure dissipation tests considering the relationships presented in
Table 1, where 𝑞𝑡 , 𝑓𝑠 y 𝑢2 are the data recorded during the CPTu tests and correspond to the corrected
penetration resistance, friction resistance and dynamic pore pressure, respectively. On the other
hand, 𝑢0 corresponds to the in-situ pore pressure estimated by pore pressure dissipation tests, 𝑝0 and

2
𝜎𝑣 . are the mean total and vertical stresses, respectively, where the inclusion of an apostrophe
indicates effective stress. 𝑘0 is the lateral earth pressure coefficient.

Table 1 Dimensionless parameter groupings for CPT interpretation (Been & Jefferies, 2016).

𝑄𝑝 = (𝑞𝑡 − 𝑝)/𝑝′ Tip resistance normalized by mean effective stress

𝑄 = (𝑞𝑡 − 𝜎𝑣 )/𝜎𝑣 ′ Tip resistance normalized by vertical effective stress

𝐹𝑟 = 𝑓𝑠 /(𝑞𝑡 − 𝜎𝑣 ) Normalized friction or friction ratio usually expressed as (%)

𝐵𝑞 = (𝑢2 − 𝑢0 )/(𝑞𝑡 − 𝜎𝑣 ) Normalized excess pore pressure

𝑄𝑝 = 3𝑄/(1 + 2𝑘0 ) Relationship between normalized resistances 𝑄𝑝 and 𝑄

The relationship proposed by Been & Jefferies (2016) to infer the state parameter based on
the CPTu results is defined by Eq. (2).

𝑄𝑝 (1 − 𝐵𝑞 ) + 1 = 𝑘̅ 𝑒 −𝑚̅𝜓 (2)

where 𝑘̅ y 𝑚
̅ are defined by Eq. (3) and (4), respectively.

𝑘̅ = 𝑀𝑡𝑐 ⋅ (3 + 0.85/𝜆10 ) (3)

𝑚
̅ = 11.9 − 13.3 ⋅ 𝜆10 (4)

𝑀𝑡𝑐 is the slope of the ultimate state line on the [q – p´] space trajectory obtained from a compression
triaxial test (also called Critical Friction ratio). These parameters were developed and evaluated by
Plewes et al. (1992) and Been & Jefferies (1992).
The most relevant property of the tailings at the time of estimating parameters 𝑘̅ and 𝑚
̅ corresponds
to 𝜆10 , which is the slope of the CSL (Critical State Line) in [e – log (p´)] space, defined on base 10
logarithmic scale. This value can be estimated by Eq. (5), which is the correlation proposed by
(Plewes, 1992), based on the correlation between the normalized friction 𝐹𝑟 obtained from the results
of the CPTu tests (see Fig. 1).

𝐹 [%]
𝜆10 = (5)
10

3
 Figure 1 Relationship between 𝛌𝟏𝟎 and 𝐅 suggested by Plewes et al. (1992) with additional data from Reid
(2012) and Been and Jefferies’s files

Fig. 2 shows the post-liquefaction residual strength ratio normalized by the effective vertical stress
(𝑆𝑟 /𝜎𝑣 ′) with the state parameter for different types of soils and tailings (Been & Jefferies, 2016),
classified according to their compressibility and parameterized by the slope of the critical state line
in the [e - p'] space. The relationships shown in Fig. 2 have the advantage of being determined based
on soil behavior modeling as well as information from historical liquefaction cases, including tailings.
Eq. (6) and (7) correspond to the parameterization of the "Best-practice trend" curve from Fig. 2.

Figure 2 Relationship between initial in-situ state parameter and mobilized steady-state strength from case
history data, for ranges of hardening (strong, intermediate, low) (Been & Jefferies, 2016)

4
 If 𝜓 > 𝜆10 :

𝑆𝑟 1 + 2𝑘0 𝑀𝑡𝑐 − 𝜓/𝜆 (6)

= ⋅ ⋅𝑒 𝑒
𝜎𝑣 ′ 3 2

Else:

𝑆𝑟 1 + 2𝑘0 (7)
= ⋅ 0.0501 ⋅ 𝑀𝑡𝑐 + 0.8 ⋅ (𝜆10 − 𝜓)
𝜎𝑣 ′ 3

Being 𝜆𝑒 the slope of the CSL in the [e – ln (p´)] space

Calibration parameters

The required parameters were calibrated with the results of in-situ vane shear tests performed at
different depths near to three CPTu soundings. Also, isotropic compression, oedometer and triaxial
compression tests were carried out on samples collected from Shelby samplers as well as on
reconstituted samples. The purpose of these tests was to obtain, by means of the laboratory tests, the
intrinsic parameters characterizing the critical state of the tailings (𝑀𝑡𝑐 , 𝜆10 ).

Vane Test
Vane tests were carried out according to ASTM D2573, 2015. The residual undrained shear strength
(𝑆𝑟 )𝑓𝑣 were calculated from the residual torque measurements according to Eq. (8), which considers
the vane geometry.

12𝑇𝑟
(𝑆𝑟 )𝑓𝑣 =
𝐷 𝐷 (8)
𝜋𝐷2 ( + + 6𝐻)
𝑐𝑜𝑠(𝑖 𝑇 ) 𝑐𝑜𝑠(𝑖𝐵 )

Where 𝑇𝑟 is the residual of torque, D the vane diameter, H the height of the rectangular portion of
the vane, 𝑖 𝑇 the angle of taper at vane top with respect to horizontal and 𝑖𝐵 the angle of taper at vane
bottom with respect to horizontal.
The results obtained are presented in Table 2 in which the name of the nearby CPTu is included.

5
Table 2 Vane Test Results.

Vane Name CPTu Name Test Depth [m] 𝝈′𝒗 [kPa] 𝑺𝒓 [kPa] 𝑺𝒓 /𝝈′𝒗 [-]
[m]
5.90 88.06 32.50 0.37
V1 CPTu-1 6.40 104.05 31.70 0.30
6.90 120.05 20.00 0.17
3.00 60.00 13.70 0.23
V2 CPTu-2 6.65 102.69 29.00 0.28

7.49 122.78 26.40 022

4.30 86.00 15.10 0.18
V3 CPTu-3
4.65 92.50 13.10 0.14

𝜎𝑣′ was calculated with a total density (𝛾𝑡 ) of 20 [kN/m3] taking in account the position of the local
water table in each sounding.

Isotropic compression tests

An isotropic compression path was performed on saturated samples obtained from Shelby sampler
up to 1 MPa, applying discrete confining pressure increments. At each increment, volume change
was recorded after stabilization of the volume. The obtained data is plotted in Fig. 3.

Oedometer Test
Oedometer testing were performed using the same initial dry density as the one used to obtain the
isotropic compression path. Both densities where selected to induce, during the loading stage, a
behavior like the normally consolidated behavior of the tailings. The maximum vertical pressure
imposed was lower than 1 MPa to avoid eventual crushing of the tailings particles. The mean
pressure for these results was estimated using a value of 𝑘0 according to the value of 𝑀𝑡𝑐 obtained
from the triaxial tests. The results obtained are plotted in Fig. 3.

Triaxial Compression Tests

Four triaxial compression tests, two CID and two CIU, were performed on tailings samples obtained
from Shelby samplers and remolded. The densities were selected in order to define the position and
the slope of the CSL in [e – log (p´)] space. Samples were prepared using the moist tamping method
and the final void ratio was calculated by squeezing out the interstitial water at the end of each test.
Unfortunately, the looser sample collapsed during the test while the dense sample presented a
kinematic discontinuity that did not allow the development of the dilatancy of the sample. However,
enough data was obtained to determine 𝑀𝑡𝑐 and 𝜆10 and parallel curves (as expected) were obtained
for the isotropic path, the oedometer path and the CSL in the [e – log (p´)] space, as can be seen in

6
Fig. 3. From these three stress paths a value of 𝜆10 = 0.18 and from triaxial tests a value of 𝑀𝑡𝑐 = 1.29
were determined.

Figure 3 Results of isotropic, oedometer and triaxial compression paths on tailings samples

RESULTS AND DISCUSSION

The state parameter ψ and the residual shear strength ratio normalized by the effective vertical stress
(Sr /σv ′) as a function of the depth were estimated based on the results obtained in a series of CPTu
soundings, without applying any correction or calibration. The results are presented in Fig. 4 where
a large dispersion is observed over the entire depth explored.

7
Figure 4 State parameter ψ and residual shear strength ratio normalized by the effective vertical stress (𝐒𝐫 /𝛔𝐯 ′)
as a function of the depth, without calibration

The 𝜆10 values obtained from the isotropic, oedometer and triaxial compression paths and those from
CPTu tests from Eq. (5) are presented in Table 3.

Table 3 Values of 𝝀𝟏𝟎 from isotropic, oedometer and triaxial compression tests and CPTu.

𝝀𝟏𝟎 𝝀𝟏𝟎 𝝀𝟏𝟎 𝝀𝟏𝟎 𝝀𝟏𝟎

Isotropic, Oedometer CPTu-1 CPTu-2 CPTu-3 Mean CPTu
& Triaxial Tests
0.18 0.17 0.20 0.17 0.18

Very good agreement is obtained from the laboratory tests and the CPTu tests for 𝜆10 value using the
proposed relationship in Eq. (5). A value of 𝜆10 = 0.18 was therefore adopted as a calibrated value
for this parameter.
Adopting values of 𝜆10 = 0.18 and Mtc = 1.29, a comparison of the residual shear strength ratio
normalized by the effective vertical stress (Sr /σv ′) of the values of vane tests carried out in the
impoundment and those proposed by Eq. (6) y (7) is reported in Fig. 5. An excellent correlation was
obtained from the in-situ values and those proposed by Been & Jefferies (2016) equations for
𝑆𝑟 /𝜎 ′ 𝑣 ≤ 0.25, while a larger dispersion is obtained for 𝑆𝑟 /𝜎 ′ 𝑣 > 0.3. This is not of major importance
since for greater values of this parameter the potential occurrence of instabilities decreases.
On the other hand, it is necessary to point out that the comparison was made for values of 𝑆𝑟 /𝜎𝑣 ′
obtained for shallow depths, where in general the values obtained by correlations with CPTU tests
are questioned.

8
The state parameter 𝜓 and the residual shear strength ratio, normalized by the effective vertical stress
(𝑆𝑟 /𝜎𝑣 ′) as a function of the depth, were recalculated for the complete set of CPTu soundings
considering the calibration of the parameters 𝜆10 and 𝑀𝑡𝑐 . The results obtained are shown in Fig. 6.
When comparing these curves with those obtained in Fig. 4, it is observed that once the CPTu test is
calibrated the curves present a much smaller dispersion. On the other hand, it is observed that the
ratio 𝑆𝑟 /𝜎′𝑣 reaches a limit value of 0.09 at depths greater than 10 m, which is not clearly observed in
Fig. 4. In addition, tailings present a dilative behavior at shallow depths (ψ>-0.05) and become
contractive at greater depths, which is probably due to the increase of confinement stresses.

Figure 5 Comparison of 𝑺𝒓 /𝝈′𝒗 values obtained with calibrated CPTu tests and Vane tests

9
Figure 6 State parameters 𝝍 and residual shear strength ratio normalized by the effective vertical stress 𝑺𝒓 /𝝈′𝒗
as function of the depth obtained with calibrated parameters 𝝀𝟏𝟎 , 𝑴𝒕𝒄 and 𝑺𝒓 /𝝈′𝒗 .

The tested tailings present a value of the state parameter for the isotropic path 𝜓 = 0.08 and a state
parameter value for the oedometer path 𝜓 = 0.03 (see Fig. 3). While the position of the odometer path
in the [e – log p´] space depends on the approximation used to estimate 𝑘0 , the laboratory values of
the state parameters are representative of the state parameter values obtained from the interpretation
of the CPTu, as shown in Fig. 6. This would imply that tailings deeper than 10 m or 15 m would
present a “normally consolidated” behavior, for these tests.
Therefore, the methodology proposed by Been & Jefferies' (2016), to predict the tailings behavior from
CPTu tests, results representative of tailings behavior and is further justified considering that the 𝜆10
slope obtained from laboratory testing and CPTu tests are also similar.

CONCLUSIONS

Currently, CPTu soundings is likely the most used tool for evaluating liquefaction susceptibility.
Although CPTu can measure the response of tailings associated to their truly undisturbed in-situ
state, it requires calibration with the response and tailings geotechnical behavior obtained through
laboratory testing. In this framework, the methodology proposed by Been & Jefferies (2016) to define
the tailings behavior from CPTu tests is convenient as it relies on calibration procedures based on
intrinsic parameters, that can be determined through conventional laboratory tests.

10
Furthermore, the method by Been & Jefferies (2016) allows estimating the residual shear strength
ratio normalized by the effective vertical stress (Sr /σv ′) through the CPTu sounding, which is an
advantage over other methodologies.

REFERENCES

ASTM D2573 / D2573M-15, 2015, Standard Test Method for Field Vane Shear Test in Saturated Fine-
Grained Soils, ASTM International, West Conshohocken, PA.

Been, K. and Jefferies, M.G. (1985). A state parameter for sands. Geotechnique, 99-112.

Jefferies, M.G. and Been, K. (1992) Undrained Response of Norsand. In Proceedings of the 45th Canadian
Geotechnical Conference, Toronto, October 26–28, 1992.

Been, K. and Jefferies, M. G. (2016). Soil Liquefaction a Critical State Approach (2nd ed.). Taylor and
Francis.

Plewes, H.D, Davies, M.P. and Jefferies, M.G. (1992). CPT based screening procedure for evaluating
liquefaction susceptibility. Proceeding of the 45th Canadian Geotechnical Conference, Toronto, Canada.

Reid, D. (2012) Update on the Plewes method for liquefaction screening. In Proceedings, Tailings and Mine
Waste 2012, Keystone, CO, pp. 337–345.

11
Verification of Dynamic Behavior Modeling of Las
Tórtolas Main Dam during 27F Chile Earthquake
Javiera Gonzalez, Priscilla Torres, José Campaña and Fabián Leal
Arcadis, Chile

ABSTRACT
Currently, the use of numerical dynamic analysis modeling for estimation of deformations in tailings
deposits dams has increased, particularly in Chile in accordance with the requirements of country´s
regulations (D.S. N° 248). Considering the relevance that these analyses have achieved, it is necessary
to validate these models comparing their results with real measurements obtained for large
earthquakes that have occurred in Chile in recent years.
Because of this, the present study performs a three-dimensional dynamic analysis, using FLAC3D™
software, of the main dam of Las Tórtolas tailings deposit, constructed with cycloned sand and
located in the central zone of Chile, 45 km north from Santiago. Analysis was carried out for the
geometry that presented the dam during the earthquake of February 27th, 2010 (27F Chile), magnitude
Mw = 8.8 with its epicenter 360 km from the deposit. In the development of this study, information
obtained from the geotechnical instrumentation that was operative at the time of the earthquake was
used.
Sands that constitute the dam were characterized by an extensive program of laboratory tests, which
included among others: Consolidated Drained triaxial (CID), Consolidated Undrained (CIU and
anisotropic), cyclic, small deformations triaxial and consolidation tests. To represent the behavior of
this material an elasto-plastic constitutive model with a Mohr-Coulomb type failure criteria was used.
At the date of the earthquake, the geotechnical instrumentation of dam included Casagrande
piezometers, fiber optic piezometers and accelerometers, which allowed registering the seismic
acceleration to which the dam was subjected and the variations of water table during the event.
Residual post-earthquake deformations obtained by the model are compared with those induced by
27F Chile earthquake observed in the field.

1
INTRODUCTION

The importance of the behavior of tailings dams against large-scale earthquakes, such as the one
occurred on February 27, 2010 (27F) in Chile, has increased considering the great heights that these
have reach to achieve the tailings storage volumes produced by the increasingly demanding mining
extraction plans. The geotechnical engineering challenge consists of understanding their seismic
performance in order to achieve a correct prediction of its behavior and, in this manner, design
projects that ensure the overall stability of these dams against severe seismic events like the 27F.
It is of great interest to know the real behavior of the tailings dam subjected to this kind of events. In
this context, it is important to compare the instrumentation measures after a strong ground motion
with the deformations obtained from a dynamic numerical analysis.
In this study, the main dam of Las Tórtolas tailings deposit behavior, observed during the 27F
earthquake, is compared with the numerical modeling response, using the information of the
instrumentation installed at the time of the earthquake occurred. To carried out this analysis, the
historical geotechnical characterization of the tailings sand was used, which is supported by an
extensive program of laboratory tests.

LAS TÓRTOLAS MAIN DAM

Location and features

Las Tortolas tailings deposit is located in the
central valley of Chile, 45 km north of Santiago.
This facility started to operate in 1992,
currently consists of three tailings dams (Main
dam, East dam and West dam) and one
earthfill dam (Plant dam). The location of these
dams is presented in the Figure 1.

Main dam geometry

At the date of the 27F earthquake, the main dam Figure 1 General disposal of Las Tórtolas tailings
reached a height of approximately 80 m deposits dams
(elevation 700 a.m.s.l.), with an upstream slope
of 1:2 (V:H) and downstream slope of 1:4
(V: H), as shown in the Figure 2.

Figure 2 Typical section main dam. Elevation 700

a.m.s.l. (2010)

2
Main dam instrumentation

The main dam instrumentation includes accelerometers, Casagrande and optical fiber (static and
dynamic) piezometers. The instrumentation locations are shown in Figure 3, which also shows the
approximate footprint of the dam during 27F earthquake.
The accelerometers that were operating at the time of 27F earthquake correspond to those identified
as TORCA, located downstream of the dam (at the instrumentation cabin), and TORRO located in
the left abutments of the dam, installed in rock. A third accelerometer was installed at the crest of the
dam, however it was not operative at the time of the 27 earthquake. The accelerometers recorded the
ground motion events in three perpendicular components: longitudinal, transversal and vertical.

Figure 3 Location in plant of main dam geotechnical instrumentation and accelerometers installation details

The phreatic level within the dam for the model was assumed according to the measurements
obtained from the piezometers that were operative at the date of occurrence of the 27F earthquake.

GEOTECHNICAL CHARACTERIZATION OF THE FOUNDATION

The main dam is founded on a valley composed by alluvial and fluvial sediments (Unit 1), which are
located mainly in the central area of the valley. These sediments correspond mainly to stratifications
of gravels with thicknesses ranging from 20 meters under the dam up to more than 90 meters to
downstream.
Immediately below this unit is the bedrock (Unit 2) which corresponds to volcanoclastic rocks
(ignimbrites and tuffs), which are supported and alternated with rocks of similar characteristics
(breccias and volcanic agglomerates). The abutments of the dam are founded on this unit.

3
EARTHQUAKE 27F CHILE

General features

The earthquake that occurred on February 27, 2010 in the central

Zone of Chile, was produced by the subduction of the Nazca
plate under the South American plate. The interplate ground
motion, magnitude 8.8 (Mw) produced a zone of rupture of the
order of 450 km length by 150 km wide. The hypocenter of this
event was located approximately 80 km north of the city of
Concepción at a depth of 30 km. The maximum recorded
accelerations (Barrientos, 2010) were 0.65g and 0.58g, horizontal
and vertical, respectively, in the city of Concepción, while in
Santiago city, accelerations of 0.56g (horizontal) and 0.24g
(vertical) were recorded.

Records on the main dam

Table 1 summarizes the main characteristics of the time-history
records measured by the TORCA and TORRO accelerometers Figure 4 Epicenter earthquake 27F
during the 27F earthquake. The Fourier spectrum of the TORCA (red star), aftershocks (blue dots) and
records are presented in Figure 5. location of the Las Tórtolas tailings
deposit (orange star). Modified from
Post 27F earthquake the piezometers installed upstream of the
reference (Barrientos, 2010).
dam recorded an increase in measured values of up to 4.5 m,
while the rest of the instruments showed no significant variations. Figure 6 shows water levels
variations of the sections L1 and L2. It is important, that in most cases, the pore pressures induced by
the earthquake, began to dissipate approximately three months after.

Table 1 Characteristics of measured records in the main dam during 27F earthquake (Campaña,
Bard & Valenzuela, 2016)

TORCA TORRO
Parameter
Long Transv Vertical Long Transv Vertical

PGA (g) 0.170 0.178 0.137 0.241 0.285 0.176

Arias intensity (m/s) 1.576 1.483 0.803 2.868 3.448 1.318

Destructive potential (10-4 g∙s ³) 19.7 17.8 5.7 48.7 60.8 11.4

4
Figure 5 Fourier spectrum of TORCA records measured during 27F earthquake in Las Tórtolas main dam

Section L1

Main Drain axis

Section L2

Figure 6 Pre and post-earthquake water level measured in upstream main dam instrumentation (Campaña,
Bard & Valenzuela, 2016)

NUMERICAL MODELING

Mesh generation
The finite differences grid was created taken
into account the main dam geometry at the
time of the earthquake of the 27F, which
consider a 80 meters high dam with crest
length of 1,100 m. The dimensions of the
mesh (5,200 m x 3,000 m) are considered
appropriate to avoid the influence of the
border on the behavior of the dam, and the
thickness of it (~500 m) is estimated sufficient Figure 7 Finite difference mesh generated for the analysis
to produce a proper transmission of the shear of the main dam
waves from the base of the model. The maximum high of the elements was defined in accordance

5
with literature (Kuhlemeyer & Lysmer, 1973), using the average shear wave speeds of each material
obtained from the degraded shear modulus.

Constitutive model

The constitutive model used for the numerical analysis corresponds to an elasto-plastic model with
Mohr-Coulomb failure criterion. The expressions that relate the elastic-isotropic modulus correspond
to:
𝐸 𝐸
𝐾 = 3(1−2𝜈) and 𝐺 = 2(1+𝜈)
Where:
K and G: correspond to the bulk and shear modulus, respectively.
E and : correspond to the elastic modulus and Poisson ratio, respectively.

In order to incorporate the dependence of the deformation modulus with the tensional state of the
element, the following expression has been incorporated to modelling the foundation, sand and
tailings behavior:
𝐸 = 𝑓(𝜎3 ´)
Where:
𝑓(𝜎3 ´): corresponds to a function of the effective confinement pressure.

It should be noted that in each stage of the dam construction (static analysis) the deformation
modulus is updated according to the new effective tensional state of the element.

Deconvolution analysis
The seismic input motion for the model was obtained using the records of TORCA accelerometer,
that is located downstream of the main dam. This instrument was chosen because its outside of the
potential influence of the dam and no topographic effect was observed that could affect the surface
response.
Tabla 3 summarizes the information used to perform the deconvolution analysis. Because the original
record was measured on the surface and the seismic input motion must be applied at the base of the
model, it was necessary to carry out a deconvolution procedure, which was developed with the Strata
software. Degradation shear modules of the soil foundation was estimated according to Rollins
curves [2], the rock was considered elastic. In general, a Rayleigh damping equal to 0.2% was used.

6
Table 3 Summary of considerations used in the deconvolution procedure of the records measured
during 27F earthquake

Material Thickness (m) Degradation Modulus Damping (%)

Unit 1: Soil foundation 36 Strata: Rollins curves 0.2
FLAC: G/G0 = 0.75

Unit 2: Bedrock 423 G/G0 = 1 0.2

For determination of the vertical acceleration to be applied at the base of the model the Strata software
was used to deconvolve the signal. For this purpose, a linear-elastic method was used considering a
rigidity profile of the materials defined by the compression wave velocities (Vp).
To validate the deconvolution process, an equivalent model (20 m x 20 m column) in FLAC3D was
performed. A Rayleigh damping of 0.2%, around frequency of 0.7 Hz, and a constant value of
degradation modulus ratio (G/G0 = 0.75), obtained for previous analysis in Strata, were used.
Figure 8 shows a comparison between the pseudo-spectra signal recorded and the obtained from the
equivalent model in FLAC3D for each component. It is observed that the adjustment is very good.

Period (s) Period (s) Period (s)

Figure 8 Verification of the deconvolution analysis, using an equivalent model. In blue original records and in
red records obtained from the model

Geotechnical parameters
The geotechnical parameters used to characterize the materials were obtained from field
investigations and several laboratory tests carried out. Table 44.

7
Table 4 Geotechnical parameters for all materials used in the model

t sat
Material ’ (°) c’ (kg/cm²) Su/v’ Es (kg/cm²)
(ton/m³) (ton/m³)
Unit 1: Soil 500∙3’0,50
1.84 - 30 0.25 -
foundation Es min= 450

Unit 2: Bedrock 2.40 - 32 10.00 - 100,000

35, p’ < 28 kg/cm² 0.1, p’ < 28 kg/cm² 420∙3’0.30

Sand (cycloned) 1.70 2.00 0.30
33, p’ > 28 kg/cm² 1.0, p’ > 28 kg/cm² Es min= 420

1 (static) 981∙’3 0.37

Tailings 1.30 1.80 5 -
0 (dynamic) Es min = 1 t/m2

The dynamic parameters considered in the analysis (G/G0, frequencies and Rayleigh damping) are
presented in Table 5. In the case of the soil foundation, the values obtained from the deconvolution
process were used.

Table 5 Dynamic parameters considered in the analysis of the main dam

Material G/G0 f (Hz) Damping (%)

Unit 1: Soil foundation 0.75 0.70 0.2
Unit 2: Bedrock 1.00 0.70 0.2
Sand (cycloned) 0.50 1.00 3.0
Tailings 0.10 1.00 4.0

RESULTS AND DISCUSSION

Input motion verification

At the control point located downstream of the main dam, a horizontal peak acceleration of 0.15g
was obtained, which reasonably fits the measurement obtained in the TORCA equipment during the
27F event (0.18g). Figure 9 shows a comparison between the response spectra obtained from the
numerical model and those calculated from the measured registers, where a filtering of the signal is
observed for periods less than 0.5 s, that is, frequencies over 2 Hz, which is consistent with the chosen
element sizes. In general, for periods greater than this value a good fit is observed.

8
 Period (s) Period (s) Period (s)

Figure 9 Verification of the deconvolution analysis in the final model. In blue original records and in red
records obtained from the model

Accelerations records
Figure 10 shows the maximum horizontal
accelerations obtained from the model for the
control point located at the crest of the dam,
downstream the dam (TORCA) and left abutment
of the dam (TORRO). As expected, acceleration
values increase with the dam height, the
maximum value occurred at the center of the main
TORCA

dam. Acceleration obtained downstream of the

main dam presents a good fit with the field
measurement. In the left abutment, it was not
possible to reproduce the acceleration measured
by the instrument installed in rock (TORRO). This
may be due to the influence of the topographic Figure 10 Maximum accelerations obtained from the
model for the control points defined
effect in the measurement or the fact that this
equipment is not actually installed in rock.

Deformations records

Figure 11 shows the contours of final

displacement magnitude. In the area of the dam
the displacements in general are 4 to 6 cm, with a
concentration in the upper central sector of the
slope with values that reach 10 cm, which are
superficial. These limited displacements obtained
in the model are consistent with the information Figure 11 Contour of final displacement magnitudes
obtained in the main dam due to the 27F earthquake

9
reported for the seismic event, where virtually no deformations were observed in the dam.

Transfer function

Figure 12 shows the transfer function determined between the crest of the dam (top) and three control
points, which are located at the base of the dam, in the cabin and downstream of it, at a point located
in rock. From these results it is obtained that the predominant frequency of the dam is equal to 1.05
Hz, value is within the range of theoretical fundamental periods defined by 4∙H/ Vs, horizontal strata,
and 2.61∙H/Vs, triangular element (dam). These values are calculated using a degraded shear wave
velocity obtained from the model (230 m/s).
(a)

(b) (c)
Top/Cabin Top/Cabin
Top/Base Top/Base
Top/Rock Top/Rock

1,05

2.61𝐻
𝑇= = 0.9𝑠 (1.2𝐻𝑧)
𝑉𝑆
4𝐻
𝑇= = 1.3𝑠 (0.8𝐻𝑧)
𝑉𝑆

Frequency, f (Hz) Frequency, f (Hz)

Figure 12 (a) Location of control points used to calculate transfer functions. (b) Transfer functions. (c)
Smoothed transfer functions (Konno & Omachi [6])

CONCLUSIONS AND COMMENTS

The main conclusions and comments of this article are summarized below.
 The numerical modeling of the behavior of the dam during the 27F earthquake predicts
deformations (4 a 10 cm) that are consistent with those observed in the field after this event
occurred (at the time of the earthquake there was no instrumentation installed to measure
deformations).
 The numerical modeling of the behavior of the dam during the 27F earthquake predicts a
peak of horizontal acceleration (0.15g) consistent with that measured by the accelerometer
installed downstream of the main dam (0.18g). However, the model fails to reproduce the

10
 peak of horizontal acceleration in the sector of the left abutment of the dam, where the
other instrument was installed. This may be due to the influence of the topographic effect
in the measurement or the fact that this equipment is not actually installed in rock.
 The fundamental period of the dam obtained from the model through the transfer function
(crest/base) is consistent with the range of theoretical values. The measurement of
acceleration records in the crest of dams is relevant in order to verify the fundamental
periods assumed in the analyzes.
 A very good fit is observed between the target recorded earthquake and computed
acceleration for the model when performing the deconvolution procedure using an
equivalent linear program (Strata). The deconvolution process for the vertical component
of the earthquake also showed a very good fit, in this case the same procedure was used,
although using a rigidities profile that was defined considering the variation of the
compressional wave velocities in depth.

ACKNOWLEDGEMENTS
The authors wish to acknowledge to Anglo American for the facilities granted for the dissemination
of the information contained in this article.

REFERENCES

Kuhlemeyer, R.L. y Lysmer, J (1973). “Finite element method accuracy for wave propagation
problems. Proc. ASCE. J. Soil Mechs. And Foundations Div. 99 (SM5) 421-427.

Rojas-González, L. F., Ben-Khayal, H. A. y Lewis, K. H. (1985). “Dynamic Properties and Behavior of

Copper Tailings”. XI ICSMFE. Vol. 3, pp. 1289 – 1292.

Rollins, K. M., Evans, M. D., Diehl, N. B. y Daily, W. D. (1998). “Shear Modulus and Damping
Relationships for Gravels”.

Konno, K. and Omachi, T. (1998). Bull. Seism. Soc. Am., 88, 228-241.

Barrientos, S. (2010). “Informe Técnico Actualizado 27 Mayo 2010. Terremoto Cauquenes 27 Febrero
2010”

Valenzuela, L. (2015). “Tailings Dams and Hydraulic Fills”. Casagrande Lecture, XV Pan-American
Conference on Soil Mechanics and Geotechnical Engineering, Buenos Aires, Argentina.

Campaña, J., Bard, E., Cano, C. y Valenzuela, L. (2016). “Registro sísmico en depósito de relaves”. IX
Congreso Chileno de Ingeniería Geotécnica.

11
Dynamic Characterization of a Tailings Dam
Embankment Using a Dense Seismic Array:
Preliminary Results
César Pastén1,3, Diana Comte2,3, Gerardo Peña3, José Burgos, and Andreas Rietbrock4
1. Department of Civil Engineering, Universidad de Chile
2. Department of Geophysics, Universidad de Chile
3. Advanced Mining Technology Center, Universidad de Chile
4. Anglo American, Chile
5. Karlsruhe Institute of Technology, Germany

ABSTRACT
The dynamic characteristics of a tailings dam embankment are relevant for estimating the dynamic
performance of the tailings deposit during large earthquakes. This paper reports the deployment of
a dense seismic array installed over the 80m height cycloned-sand embankment dam of a tailings
deposit located in Central Chile. The network consisted of 28 short-period seismic stations with
triaxial 4.5Hz geophones that continuously recorded ambient seismic noise and earthquakes during
four weeks. The seismic stations were deployed along the embankment crest, the downstream slope,
and the foundation soil. Single-station records were processed using the H/V spectral ratio method
to estimate the embankment dam predominant vibration frequency using both ambient seismic noise
and earthquake records. The results from the H/V spectral ratios from both types of signals are
consistent and show that the foundation soil is stiffer than the embankment dam, that the
predominant vibration frequency of the embankment is about 1 Hz, and there are differences in the
vibration frequency along a cross section of the downstream slope. Finally, the results of the H/V
spectral ratios are compared with standard spectral ratios calculated between the crest and the
foundation of the embankment wall using the available earthquake records. Both methods are
consistent in terms of the predominant vibration frequency and amplification factors.

1
INTRODUCTION

Earthquakes are among the most common triggers of tailings dam failures (WISE Uranium Project,
2019). Villavicencio et al. (2014) reported that 31 out of 38 incidents in Chilean sand tailings dams
were related to earthquake loading. The seismic response of a tailings dam depends upon the
geometry and the stiffness of the materials that compose the embankment wall, as well as the input
ground motion, among other factors. The stiffness of the materials can be estimated through the shear
wave velocity, a parameter that can be measured in-situ with invasive and non-invasive geophysical
techniques. Another dynamic parameter that combines the geometry of the deposit and its stiffness
is the predominant vibration frequency that represents the natural vibration frequency of the earth
structure (Verdugo et al., 2017). If an earthquake induces a ground motion with high energy content
around the predominant vibration frequency, the embankment wall may resonate, increase its
dynamic displacements, and eventually produce damage, which may compromise the physical
stability of the entire tailings deposit.
In earthquake engineering, the most common method to determine the amplification function of a
soil deposit is the standard spectral ratio (SSR) method, computed as the spectral ratio between the
response at the site of interest with respect to the response of a reference point (Steidl et al., 1996).
The reference point can be either the base of the deposit where the input motion is applied or a
reference station that ideally is installed on a site not affected by the local soil conditions, either on
stiff soil or hard rock. The predominant vibration frequency is estimated from the amplification
function as the lower frequency where the amplification function peaks.
Another method to estimate the amplification function is the single station horizontal-to-vertical H/V
spectral ratio (HVSR), also known as the Nakamura’s method (Molnar et al., 2018). This method was
originally developed with continuous records of ambient seismic noise, usually more than 30 mins
long. However, it has been proved in soils deposits that the HVSR curves from ambient seismic noise
and earthquake records yield similar results, hinting that the HVSR is a relatively constant site
characteristic (Fernandez et al., 2019).
Since tailings dams build with cycloned sand in Chile have embankment walls with very gentle
downstream slopes (H:V varying from 3:1 to 4:1), the use of the HVSR method could be a simple
method to estimate its dynamic properties. The use of the method has been tested and validated in
natural sloped terrains (Diaz-Segura, 2016).
In this paper, we show the main features of a seismic array deployed in a tailings deposit in Central
Chile and the preliminary results of the single-station and standard spectral ratios that allow
determining the dynamic properties of the embankment wall.

2
SEISMIC ARRAY

The embankment wall of the studied tailings deposit is constructed with cycloned sand using a center
line method. The current height of the embankment wall is about 80 m with a downstream slope of
H:V = 1:3.5 (Figure 1).
The seismic array consisted of 28 stations, each one equipped with a short period 3-component 4.5
Hz geophone, an Omnirecs DataCube3 Ext datalogger recording at 200 samples per second, a GPS
antenna for time synchronization, and a sealed gel deep cycle battery. The stations recorded
continuously for about one month (from August 29, 2018 to September 26, 2018). Figure 1 shows the
location of the 28 seismic stations deployed in the tailings deposit. Eleven stations were installed in
the crest of the embankment wall, 6 installed in the downstream slope, and 11 installed at the
downstream slope toe. The last group of stations were deployed at an average distance of 5 m apart
from the embankment wall due to restrictions imposed by underground infrastructure. The stations
were buried in a sealed plastic container and the GPS antenna was raised 50 cm to the surface.
Figure 2 shows an example of the daily continuous record at the station T08 in the north-south (NS)
component. The amplitude of the velocity time history (Figure 2a) increases during the operation
time of the day, between 9am and 6pm. Figure 2b shows a normalized spectrogram with the
frequency content of the time history signal. The frequency content remains relatively constant before
the operation work starts at the tailings dam. However, the frequency content changes considerably
between 2 and 15 Hz during the operation period.
During the time that the seismic array was under operation, at least 11 earthquakes of different
magnitude were recorded by the seismic stations. The earthquakes moment magnitudes range from
3.9 to 5.8 and the epicentral distances vary from 10 to 410 km.

(a) (b)

A’ Tailings
impoundment T08
T07 T08 3.5
1 ~80 m
T06
A T06
Crest

500 m A-A’ cross section

Embankment wall
downstream slope

3
Figure 1 Seismic array deployed at the tailings deposit in Central Chile. (a) Schematic plan view of the deposit
and (b) cross section of the embankment wall. Triangles represent the location of the seismic stations

(a) T08 – NS Component

Counts (106)

4
0
-4 Stationary seismic ambient noise

(b)
40 1

30

Normalized Amplitude
Frequency [Hz]

20 0.5

10

0 0
9 pm 2 am 7 am 12 pm 5 pm 9 pm
Time

Figure 2 Example of a continuous daily record at station T08 in the north south component (September 14,
2018). (a) Velocity time history and (b) normalized spectrogram (lighter tones represent larger amplitudes)

DATA PROCESSING AND RESULTS

We processed separately data from ambient seismic noise and earthquake records. First, horizontal-
to-vertical spectral ratios HVSR were calculated using the 12 hrs of stationary ambient seismic noise
recorded from 9pm to 9am, before the operation starts (Figure 2a). We followed the methodology
described in Pastén et al. (2016), using 30 s windows automatically selected from the amplitude
STA/LTA ratio between 0.5 and 2, with a short time period tSTA = 1 s, and a long time period tLTA = 60
s. The horizontal component was calculated as the squared average of north-south and east- west
components. The calculations were performed using the open-source software Geopsy
(www.geopsy.org).
Figure 3 shows the HVSR calculated as the ratio between the combined horizontal component and
the vertical one of 3 stations located at the crest (T08 in Figure 1), at the downstream slope (T07 in

4
Figure 1) and at the embankment toe (T06 in Figure 1) in the cross section A-A’ (Figure 1a). The HVSR
curves indicate a predominant vibration period associated to the peak spectral ratio amplitude. The
station T08 has a predominant frequency of about 1.0 Hz whereas station T06 has it at 1.7 Hz. It seems
that the HVSR at the station T07 is a combination of the curves at the crest and toe of the embankment
wall.

T06 T07 T08

Figure 3 Examples of HVSR calculated from ambient seismic noise at stations T06, T07, and T08 installed on
the cross section A-A’ in Figure 1. The thick black curve is the mean of the curves in color

The eleven earthquakes recorded by the seismic array were used in each station to calculate HVSR.
First, the amplitude Fourier spectra from the velocity records are calculated in each component and
smoothed with a running average window. Then, the north-south and the east-west components are
divided by the vertical one. Figure 4 shows examples of HVSR in stations T08 and T06. The average
curves are similar to the average HVSR computed from seismic ambient noise. Interestingly, the
predominant frequencies using ambient seismic noise and earthquake records agree at 1.0 Hz in
station T08 and 1.7 Hz in station T06.

Figure 4 Examples of HVSR calculated from earthquake records at stations T06 and T08 installed on the cross
section A-A’ in Figure 1

5
Standard spectral ratios SSR were calculated from earthquake records in the station at the crest (T08)
with respect to the station at the downstream toe (T06). Figure 5 shows an example of the SSR of one
of the earthquakes (2018-09-07, Mw 5.8), considering separately the ratio between the horizontal (NS
and EW) and vertical (Z) components.

Figure 5 SSR calculated between stations T08 at the crest of the embankment wall and T06 at the toe

DISCUSSION
The predominant vibration frequency changes with the thickness of the embankment wall where the
stations are located. For instance, at station T08 on the crest of the wall where the underlying cycloned
sand is thicker, near 80m, the predominant vibration frequency is lower than the predominant
frequency at the downstream toe where the station T06 is installed directly on the foundation soil. In
addition, the amplitude of the HVSR increases with the thickness of the underlying cycloned sand
layer. The HVSR of station T08 nearly doubles the amplitude in station T06.
The predominant frequency of the stations installed in the foundation soil may be affected by the
thickness of the foundation sediments in the valley and also by the interaction with the embankment
wall. Decoupling these two effects may require further analyses that will be performed during the
following stages of this project. Ideally, the stations in the foundation soil should be installed at a
distance from the wall that would not influence their response. However, there is not clear rule of
thumb for such distance in this type of earth structures.
The HVSR calculated from earthquake records retains most of the features of the HVSR calculated
from ambient seismic noise. In particular, the predominant vibration frequency and the peak
amplitude are almost identical in both cases. This result suggests that the predominant vibration

6
period can be recovered in tailings dams from earthquake recorded by accelerometers, which is the
type of sensor most commonly used in this type of structures (Campaña et al., 2016).
The SSR in the horizontal components shows that the largest amplification between the two stations
occurs at about 1 Hz and 3 Hz, the fundamental and first harmonic horizontal vibration mode of the
earth structure. The fundamental frequency is similar to the value obtained from HVSR calculated
from both earthquake records and ambient seismic noise. Note that the ratio between these two
vibration frequencies is 1:3, similar to the prediction of the one-dimensional shear wave propagation
theory. The SSR in the vertical component peaks at higher frequencies (2 and 5 Hz) which can be
explained since the vertical component strongly depends on the P-wave velocity of the earth
structure.

CONCLUSIONS AND FINAL COMMENTS

The results from the H/V spectral ratios from ambient seismic noise and earthquake records are
consistent and show that the foundation soil is stiffer than the embankment dam, the predominant
vibration frequency of the embankment is about 1 Hz, and there are differences in the vibration
period along a cross section of the downstream slope.
The results of the H/V spectral ratios were compared with standard spectral ratios calculated between
the crest and the foundation of the embankment wall using the available earthquake records. Both
methods are consistent in terms of the predominant vibration frequency and amplification factors.
The predominant vibration frequency is a parameter that can be used to calibrate numerical models
that are often developed to determine the seismic response of tailings dams when subjected to large
earthquakes.
In the future, the results from this passive experiment can be combined using ambient seismic noise
tomography tools to obtain a 3D shear wave velocity map of the embankment dam, which can be
contrasted and complemented with traditional geotechnical data, such as pore water pressure and
deformation measurements, to improve the understanding of the dynamic performance of tailings
deposits.

ACKNOWLEDGEMENTS

The authors thank Anglo American for the permission granted to deploy the seismic array and to use
the recorded data. We also acknowledge the help provided by F. Hormazábal and I. Garvs with the
data processing.

NOMENCLATURE
HVSR H/V spectral ratio

7
SSR standard spectral ratio

REFERENCES

Campaña, J., Bard, E., Cano, C., and Valenzuela, L. (2016) ‘Registro Sísmico en Depósito de Relaves’,
IX Congreso Chileno de Ingeniería Geotécnica, Valdivia, Chile.

Diaz-Segura, E. G. (2016) ‘Numerical estimation and HVSR measurements of characteristic site

period of sloping terrains’, Géotechnique Letters, 6, 176–181.

Fernández, J., Pastén, C., Ruiz, S., and Leyton, F. (2019) ‘Damage Assessment of the 2015 Mw 8.3
Illapel Earthquake in North-Central Chile’, Natural Hazards, 1-15, (accessed March 10, 2019,
https://doi.org/10.1007/s11069-018-3541-3)

Molnar, S., Cassidy, J.F., Castellaro, S., Cornou, C., Crow, H., Hunter, J.A., Matsushima, S., Sánchez-
Sesma, F.J. and Yong, A. (2018) ‘Application of microtremor horizontal-to-vertical spectral ratio
(MHVSR) analysis for site characterization: State of the art’, Surveys in Geophysics, 39, 4, 613-631.

Pastén, C., Saez, M., Ruiz, S., Leyton, F., Salomón, J., and Poli, P. (2016) ‘Deep characterization of the
Santiago Basin using HVSR and cross-correlation of ambient seismic noise’, Engineering Geology, 201,
57-66.

Steidl, J.H., Tumarkin, A.G. and Archuleta, R.J. (1996) ‘What is a reference site?’, Bulletin of the
Seismological Society of America, 86, 6, 1733-1748.

Verdugo, R., Peters, G., Valenzuela, L., Campaña, J., and Bard, E. (2017) ‘Evaluation of Tailings Dams
Subjected to Large Earthquakes’, Proceedings of the 19th International Conference on Soil Mechanics and
Geotechnical Engineering, Seoul, Korea.

Villavicencio, G., Espinace, R., Palma, J., Fourie, A., and Valenzuela, P. (2013) ‘Failure of sand tailings
dams in a highly seismic country’, Canadian Geotechnical Journal, 51, 449-464.

WISE Uranium Project (2019) Chronology of major tailings dam failures, World Information Service
on Energy Uranium Project, accessed March 10, 2019, http://www.wise-uranium.org/mdaf.html

8
Dam-Break: Effect of Water Content in Tailings vs
Run-Out Distance
Diego Inzunza and Jorge León
Knight Piésold, Chile

ABSTRACT
In current regulatory frameworks, dam breach and tailings run-out modelling of tailing storage
facilities is a requirement and a matter of increasing interest within the regulatory agencies and
practitioners. Increased levels of densification of impounded tailing materials by means of
alternative tailing deposition technologies, by means of installation of wick drains, or by loading
the surface of the impoundment with waste rock, all generate a change in the interstitial pressure of
the tailing column that may lead to densification and an increase of the rheology of the settled
tailings. In the case of run-out analyses, we present a sensitivity analysis of the effect of the degree
of consolidation at the time of a dam breach on the run-out distance potential. The results show a
decrease of up to 16% in the maximum area reached by the outflow due to the increase in the
degree of consolidation. The incorporation of the relative effect of densification methods and the
resulting change in rheology represents a progress in understanding the critical elements that drive
the potential for downstream impact of a dam breach event.

1
INTRODUCTION

Tailings run-out distance refers to the distance that an outflow volume of tailings and water may
travel as a consequence of a breach of the confining dam of the tailings storage facility (TSF). In
Chile, these studies are mandatory for TSFs under Sernageomin regulation DS N°248. To date there
is no specific methodology or official guidelines on the procedure to determine the tailings run-out
distance. Several authors have proposed empirical and geometric methods to evaluate tailings
outflow, such as those proposed by Rico et al. (2008) and Lucia et al. (1981). Martin et al. (2015)
presents a practical guide to assessing dam break outflow modelling by comparing the key
parameters that affect and control the run-out distance. According to the Canadian Dam
Association (2013), three modes of failure are recognized: overtopping; collapse; and contaminated
seepage. This study considers only failure by collapse.
The outflow hydrograph, geometry of the dam breach and rheological properties of the tailings are
among the key variables that control the run out distance. This study evaluates the effects on the
run out distance as a result of improving the tailings rheology by consolidation with wick drains.
The incorporation of wick drains allows the reduction of water content, increasing the solids
content by volume (Sotil et al, 2018). The wick drains are complemented with consolidation load to
reduce the consolidation time. Adams et al (2017) demonstrated the effectiveness of applying a
consolidation load accelerated with wick drains to densify, dewater, and reduce the flowability of
tailings through an instrumented test program installed at a mine site in central British Columbia,
Canada. The objective of this study is to evaluate the potential reduction of the inundation area that
may result from improved tailings consolidation.
The studied TSF is a conventional tailings facility with 3 embankment dams and a total storage
volume of 338.6 Mm3 (Figure 1). The analyses were carried out for the larger dam. The dam is
raised using the downstream raise method, with an upstream slope of 2.5H:1V and a downstream
slope of 2H:1V. The height of the dam is 64 m and the volume of the embankment is 3.8 Mm3.
The results of the inundation area are compared before and after the incorporation of improved
consolidation methods of the tailings by means of wick drains. The methodology includes an
estimation of the outflow volume and hydrograph for both tailings states, with and without wick
drains. Numerical modelling was completed using the Flo-2D software.
A summary of the geometry and main elevations of the dam are those indicated in Table 1. Figures
1 and 2 show the general location of the TSF and a typical embankment section (Dam 2).

2
 Impoundment

Embankment 1

Embankment 2

Embankment 3

Figure 1 General layout

Figure 2 Embankment typical section

THEORY AND METHODOLOGY

The methodology implemented for dam break analyses is currently an ongoing matter of debate
within the engineering community, therefore conservative criteria and parameters have been
utilized. It is noted that the aim of this research is not on the discussion of the methodology for the

3
dam break analysis, but rather quantifying the affected area of the outflow material. The tailings
properties used in the analysis are shown in Table 1.

Table 1 Tailings properties

t t Su
Frictional angle, drained (ϕ) Dry unit weight ( ) Cohesion ( )
m3 m2 σv Liquefaction

26 1.6 0 0.06

The outflow volume was estimated using Equation 1 to obtain the friction angle (ϕ). The outflow
volume therefore is function this friction angle and tailings elevation, the latter is intended to
reduce as it will be explain in the next sections. Has been analysed static liquefaction failure with
the tailings residual strength ratio. Figure 1 illustrates the equivalent friction angle derived from the
residual strength ratio utilized in the model (Blight, 2010).
𝑆𝑢 −1
ϕ = tan (
𝜎𝑣
) (1)

Figure 1 Angle friction on Dam

Piping failure was simulated by means of Bossbreach software, which solves Equation 2 in order to
calculate the outflow as an output hydrograph. The analysis considers the total embankment failure
and the outflow volume passes through the complete width of the embankment.

g b = A[2g(H − Hp )⁄1 + fL⁄D]0.5 (2)

where: g b : flow through the pipe; G= gravity acceleration; A: cross-sectional area of the pipe
channel; H − Hp = hydrostatic head on the pipe; L: length of the pipe channel; D= width of the pipe;
f= Darcy friction factor.
After obtaining the hydrograph, the results were entered as input to the Flo-2D model to evaluate
the affected area from the outflow. This program performs finite difference modeling in a three-
dimensional space and it is possible to determine flow characteristics such as height, velocity and
time of arrival of the flow at different control points downstream of the reservoir

4
Yield stress and dynamic viscosity vs solid concentration are required for the software that
characterise the tailings transportation on the area downstream of the embankment. The Figure 2
(a)-(b) shows the equations for the rheological behaviour. Four samples were tested with a
rotational viscometer and correspond to conventional copper tailings.

Figure 2 Rheology of tailings utilized in the model (a) Yield stress, (b) Dynamic viscosity

The aim of the outflow modelling was to evaluate the effect of accelerated or improved tailings
consolidation by means of wick drain installation within the impounded tailings during the
deposition phase. The purpose of the installation of wick drains is to increase the density of the
tailings materials by accelerating the rate of seepage outflow from saturated tailings. The increased
density is expected to result in an increase of the rheology of the tailing materials.
Wick drain installation within the impounded tailings was considered up to 20 m depth and 1 m
spacing between wick drains. The tailings consolidation was modelled as shown in Figure 3 (a),
corresponding to the ultimate configuration of the TSF with the highest column of tailings. A
loading pressure of 10 kPa was applied over the ultimate elevation of the tailings column, for a
duration of approximately one half year. The wick drains were modelled with a total depth below
the surface of 20 meters, as shown in Figure 3 (b); the final excess pore water pressures after the
installation of the wick drains are also shown in Figure 3 (b). Figure 3 (c) shows the increase of
settlement due to the installation of wick drains, converging to about 2 meters within one half year.

5
 Figure 3 (a) tailings column (b) Excess water pressure (c) Total settlement Tailings

RESULTS AND DISCUSSION

This section presents the model results for the study case. Details for the hydrograph and
parameters of the outflow volume are shown in Table 2. The results show an increase of the total
consolidation as well as a reduction of the water content between the base case (no wick drains) and
the accelerated consolidation case (with wick drains). The accelerated consolidation also increases
the equivalent solids content by weight (Cp) from the reduction of the water content. The behaviour
of the outflow hydrographs, shown in Figure 4, is similar between each case, but there is a
difference in the maximum amplitude between the case base and the accelerated consolidation case.

Table 2 Hydrograph and characteristics parameters

Parameter Base case Consolidated Impoundment

Angle ϕ𝐋𝐢𝐪𝐮𝐞𝐟𝐢𝐞𝐝 (∗) 3.53° 3.53°

Outflow volume (m³)(**) 32,219,857 27,854,071

Cp (%)(***) 60 63

Q peak (m³/s) 8,844 5,625

(*) Value obtained from eq. (1) and resistance S𝑢 ⁄σv Liquefaction = 0.06

(**)The outflow volume was estimated for Base case and Consolidated impoundment using eq. 1 to obtain the friction angle (ϕ), and decrease
of the tailings height (volume reduction) due the densification process for the Consolidated impoundment.

(***) Assumed mean value Tailings Cp.

6
 Figure 4 Outflow hydrographs

The results from the run-out model show that the outflow stream reaches high velocities along the
initial reach of the downstream path, in accordance with the narrow landscape, and a decrease of
the maximum velocity in the flatter landscape. The scenarios evaluated are constrained by the
natural topography and the operating time of the wick drains, in this case one half year.
The deployment of wick drains in the impoundment of the TSF results in a higher degree of
densification and a higher equivalent Cp value and the resulting flooded area decreases. Figure 5
shows the flooded areas from the outflow volume for the study cases and the maximum velocity of
the slurry along the flow path. The flooded area associated to low velocities is reduced in the
improved consolidation case (with wick drains). The results show a decrease of the flooded area by
16% as a result of the higher rheology from improved consolidation by wick drain installation.

7
Table 3 Summary results

Item Flooded Area (m²) Run-out Distance (m)

Base Case 23,855,040 14,097

Consolidated Impoundment 20,113,139 12,941

Reduction 3,741,901 1,156

Reduction (%) 16% 8%

(a) N

m/s
15.0
13.4
11.7
10.0
8.4
6.7
(b) N 5.0
3.4
1.7
<=0

Figure 5 Flooded area and maximum velocity of the slurry along the flow path for liquefied angle. (a) Case
Base; (b) Consolidated Impoundment Case

CONCLUSIONS
This research presents an evaluation of the effects of the installation of wick drains in the
impoundment of a TSF on the increase of the degree of tailings consolidation and the resulting
increase of the rheology and the reduction of the ability to flow downstream of the TSF under a
hypothetical dam breach. The rise of 3% on the Cp parameter implies a densification of the tailings
and a fall of 16% of the flooded area and under the dam-break analyses the run-out distance
decrease 8%. Although the reduction of the run-out distance is not too notorious, the flooded area

8
reduction is significant in the analyses. The wick drains were used in this study because of their
ability to accelerate in a short time the dissipation of excess pore pressure.
The results shown are site-specific and serve as an example, and are therefore limited by the
topography of the study area; consequently, the reduction of the flooded area presented in this
study may vary from site to site. Another important aspect, within the limitations of the results
presented in this study, is the operation time of the wick drains.

REFERENCES

Adams, A., Brouwer, K. & Davidson, S., (2017). Best Available Technologies to Stabilize a Historical
Tailings Impoundment, Proceedings of the Tailings and Mine Waste Conference, Banff, Alberta.

Canadian Dam Association (2007, revised 2013), Dam Safety Guidelines.

Lucia, P.C., Duncan, J.M. & Seed, H.B. (1981) ‘Summary of research on case histories of flow failures
of mine tailings impoundments’, Technology Transfer Workshop on Mine Waste Disposal Techniques,
Information Circular 8857, U.S. Bureau of Mines, Denver, CO.

Martin, V., Fontaine, D. & Cathcart, J. (2015) ‘Practical tools for conducting tailings dam breach
studies’, CDA 2015 Annual Conference.

Rico, M., Benito, G. & Díez-Herrero, A. (2008) ‘Floods from tailings dam failure’, Journal of Hazardous
Materials, Volume 154, Issues 1-3, Pages 79-87.

Sotil, A., Adams, A. & Hall, C. (2018) ‘Tailings Impoundment Closure Enhancement’, Planning for
closure, Santiago, Chile.

Blight, G. (2010) ‘Geotechnical Engineering for Mine Waste Storage Facilities’, CRC Press.

9
The benefits of Material Volume Meshes in 3D Slope
Stability Analysis: a Tailings Dam Case Study
Murray Fredlund1, Marina Trevizolli2 and Jon Foster3
1. SoilVision Systems, Canada
2. SoilVision Systems, Brazil
3. BCG Engineering, Canada

ABSTRACT
One of the greatest challenges in performing 3D geotechnical numerical analysis is in creating a
model of the geometry of a particular site. A well-defined model that accurately represents the site
topography and boundaries of material layers is essential to producing an accurate analysis. For a
site with few material units in which the units are cleanly separated and stacked on top of each other,
the modelling process is relatively simple. However, sites may have numerous features such as
dykes, ore bodies, or other enclosed volumes of material with complex shapes.
The Material Volume Meshes (MVM) is used to handle unusual layered models in situations where
the traditional methods present several limitations, as the most common cases like surfaces folding
over themselves or negative volume layers. The MVM represents one distinct volume of material
composed by a closed mesh shape and any material assigned. Such a material volume could therefore
be subsequently represented by a rock or undrained/drained soil constitutive model. The MVM is
fully independent and over-rides the layer-cake conceptual model and respects the region boundaries
of the model volume during the slope stability analysis.
This paper seeks to demonstrate a case study of a 3D tailings dam numerical model developed using
material volume meshes as a starting point for a slope stability analysis. A conceptual model was
first developed and then imported directly to a 3D limit equilibrium method (LEM) slope stability
software package. The results showed how MVMs can help in obtaining a realistic factor of safety
representative of complex geological subsurfaces. The paper also aims to illustrate the advantages of
a well-defined conceptual model and what type of field information can be utilized to form a material
volume mesh for the purpose of creating a reliable model that adequately represents field conditions.

1
INTRODUCTION

The use of earth dams to create retention areas for the fine mine tailings is commonplace in the mining
industry. The geometry of many tailings dams is complex and typically 3D in nature.
Analysis of slope stability of tailings dams has received increased attention in recent years due to
recent tailings dam failures such as Mt. Polley, Canada and Fundao, Brazil. These failures have
resulted in increased review assessments of the stability of existing tailings dams worldwide by
geotechnical consultants (Fredlund et al., 2017). The most recent failure event of Brumadinho, Brazil
in 2019 has accentuated the need of a technical reliability in conceptual and numerical models, due
to the necessity of understanding potential failure mechanism for future prevision and risk
management.
This paper will examine how 3D Limit equilibrium method (LEM) models are more effective at
modeling complex dam geometry but examining the stability at the right abutment of an earthen
dam, labelled “Dam 1” fin this paper. Specific focus will be given to how material volume meshes
(MVMs) have over come technical limitations to 3D LEM models to accurately replicate geometry
that was previously not possible.

MATERIALS AND METHODOLOGY

Why build a 3D Model

Use of 2D LEM models for assessing the stability of real world slopes is based on two assumptions:
 Plane strain conditions are assumed: all slope strain occurs within the plane of the section.
 The geometry of the section is sufficiently similar either side of the cross section, so that out
of plane geometry has limited effect on the result.
In the case of many tailings dams the above two assumptions are not met. The lithology of the
foundation, geometry of the dam and/or position of the water table can vary across the site resulting
in 2D section geometries that can vary significantly with small adjustments in the section’s position.
These variations can also make it difficult for experienced engineers to anticipate the critical sliding
locations and direction (Fredlund et al., 2017) and thus making it difficult to a line the strains
occurring in the section with the strains of the true failure.
Therefore a 3D model is needed in order to meet the requirements for proper analysis of a tailings
dam at a particular site.

Tailings Dam Model

Like many tailings storage facilities (TSF) the design of “Dam 1” evolved iteratively. After the
completion of the starter dam to a height of 62 m, the required capacity for the TSF was increased
based on further proven ore reserves. The ultimate height of the dam was then changed from 138 m

2
to 145 m. To accommodate this additional 7 m in height, a number of design modifications where
required. Some of the modifications are listed below (the detailed rationale for each modification is
beyond the scope of this paper):
 The neat line inclined core slopes would have to be adjusted to provide sufficient core
thickness above the original crest elevation.
 A bend of the crest alignment at the right abutment (RA) to allow the dam to key into the
natural topography at a higher elevation (Figure 1)
 A transition from an inclined core at the mid-valley to a central core in the abutments (Figure
2)
Dam 1 is a downstream constructed tailings dam. The dam is constructed of a compacted clay core,
supported by engineered rockfill. An internal chimney filter prevents internal erosion of the core into
the coarse rock fill. Internal erosion of the downstream foundation is prevented with a filter blanket
zone. For stability modeling, dam zonation has been simplified based on similarities in material
behavior.

Figure 1. Change in Dam Reference Line and Footprint with Change in Ultimater Dam Height. Original 258
El. Dam in Black, 265 m el. Alignement in Blue

3
 Figure 2. Transition of Core from US inclined Configuration (Top Left) to Central Configuration (Bottom
Right). Note at Station 0+153 (Bottom Left), Downstream Face of Core is Vertical Which is Not Possible to
Model with a 2.5D Surface

Building a 3D Model
To build the Tailings Dam 3D model, information from a number of sources was used to create the
model including: Pre-construction LiDAR, Digitized Foundation approval surfaces, as-build
geometry and 2D and 3D CAD designs.
Geological interpretations and borehole data can be included in the modelling process however, for
this model a simplified geology was assumed as preliminary stability analysis indicated that
upstream (US) stability at the dam bend was controlled by 20 m thick layer of soil immediately under
the dam at the location of the bend.
The 3D model was built entirely within SVDESIGNER, a 3D conceptual modeling software package
developed by SoilVision (SoilVision, 2018). Pre-development LiDAR was used to define the both the
topographic surface and translated 20 m down to bound the bottom of the controlling soil layer in
the simplified geology. A DXF surface of the digitized foundation approvals were imported and
intersected with both the topographic and subsurface surfaces to replicate excavation of the
foundations. The foundation approval surface was also used to create the blanket filter surfaces by
translating the surface up the width of the filter and trimming it to the filter extents. As-built dam
geometries we imported from SurpacTM and AutoCADTM with surface mesh refinement editing
Performed. Future design stages were either extruded from 2D neat line sections or imported directly
from AutoCAD.

4
To simplify their modelling, many of the surfaces described above where modelled using a special
subset of 3D surfaces called 2.5 dimensional (2.5D) surfaces. In 2.5D surfaces are defined in 3D space
with each point having an X, Y and, Z coordinates, but with two limitations: 1) the surfaces must be
continuous between all adjacent points and 2) the surface can only contain one Z coordinate for each
unique X,Y coordinate. This means that unlike a true 3D surface, 2.5D surfaces cannot fold over
themselves or allow true vertical faces.
While many of the dam components can be built up with layers of 2.5D surfaces, there are some dam
components which could not be modeled as 2.5D surfaces. For these surfaces Material Volume
Meshes (MVMs) were needed.

Material Volume Meshes

MVMs are not restricted to 2.5D and exists as enclosed 3D surfaces defining a volume of material.
An MVM overrides the materials defined by the layered surfaces of geotechnical conceptual model.
This feature is useful for modelling geological layers with extremely complex geometry that usually
can’t be represented by 2.5D surfaces in a 3D model (SoilVision Systems, 2018).
As described in Figure 2 the compacted clay core transition from an upstream (US) inclined
configuration in the central of Dam 1 to a central core configuration at the right abutment (RA). This
transition is problematic when using 2.5D surfaces exclusively, for 2 reasons: 1) The downstream face
of the core and filters will invert resulting in a location where the surface will be vertical and 2) The
order in which 2.5D surfaces would be layered from an inclined core (downstream (DS) rockfill first,
then the chimney filters and core followed by US rockfill) is different than the order for a central core
(combined DS and US core and chimney filters first followed by US and DS rockfills).
To solve this problem with the geometry an MVM was constructed the available geometry data. The
top of the filter blanket was used to the define the base of the Core and chimney filter. The previously
designed US and DS surfaces for the core and chimney were imported from AutoCAD and Surpac
and used for the sides of the MVM. The top of each stage was created natively within the software.
Each individual surface was then tied together like assembling the faces that make up a cube to
construct the completed MVM.
Figure 3 illustrates the Tailings dam view with the length of the MVM used.

5
 Figure 3. 3D Model of Dam 1, exterior surface for MVM of LPF and Chimney Filter shown

Slope Stability Analysis

The overall stability analysis of Dam 1 includes a number of 2D cross-sections along the dam profile
that were analysed to asses the stability of the dam under drained, undrained, and seismic loading
cases in both the US and DS direction. The results were compared with established stability criteria.
Where this established criterion was exceeded the design was accepted; where the criteria was not
met, modifications in the design were made until the criteria was met.
In most cases the conservative 2D analysis exceeded the design criteria and were excepted. However,
in the case of US stability at the bend in the RA, 2D sections could not reasonably resolve the pseudo-
static seismic stability and a 3D stability model was required. To analyze the LEM slope stability
analysis in 3D, the software SVSLOPE (SoilVision, 2018) was used which is integrated with the
conceptual modeling software, providing an improved interaction between geometry building and
analysis in the final solution.
The 3D analysis illustrated in this paper aimed to verify the US factor of safety under pseudo-static
seismic conditions. Based on the site-specific seismic assessment available at the time of the analysis,
a seismic yield coefficient (ky) of 0.08 or greater was found to meet/exceed the displacement criteria
regardless of the height or average shear wave velocity of the slip surface when analyzed using Bray
and Travasarou (2017). An orientation analysis was considered to identify the direction of the critical

6
slip surface. Orientations between -40° to 40°from the bisect of the US crest bend angle were
considered.
Pore-water pressures (PWP), were modeled within the software using a spatial function. The values
of PWP were determined based on which portion of the model was being considered and what PWP
primary influence was at that location. e.g. Pond elevation in the US, Filter location in the DS, and
construction and pond loading in the core. Exact details of PWPs are beyond the scope of this paper.
The material strength parameters for the foundations and the fills are based on a large number of in-
situ and laboratory testing completed as part of a number of site investigation and material
characterization programs. Testing includes but is not limited to: Index testing, SPT, direct shear and
triaxial tests. Rockfill properties were estimated using the Average Leps shear normal function (Leps,
1970). Tailings were assumed to liquify for seismic analysis. This meant that it had to be modeled
using the software fluid model type to allow for the correct material density without being
incorporated in the sliding mass.

RESULTS AND DISCUSSION

As shown in Figure 4, the 3D Slope Stability illustrates the slope failure shape for an US analysis
using the Morgenstern-Price calculation method and Figure 5 the 2D Slope Stability under the same
seismic condition of the 3D model. The 2D cross section for analysis was located at the bend inflexion
point.
Results of the 3D stability analysis of the US RA show that the Factor of Safety (FoS) for the pseudo-
static analysis with a horizontal coefficient of 0.08 is greater than 1, indicating that the yield coefficient
is greater that 0.08 and would satisfy the deformation criteria at the location in question.
The MVM technique was compared with 2D LEM Analysis that is limited to the plan strain and
geometric consistency assumptions, mentioned and discussed above. The comparison between the
critical factor of safety of a full 3D model and the 2D section, showed a discrepancy of about 35%,
furthermore the 2D model shows a FoS less than 1 for a seismic coefficient of 0.08, indicating that the
yield co0efficent will be less than 0.08 and the resulting slip surface may not meet the required design
criteria. requiring further design to meet the stability criterial resulting in an overly conservative
design compared to the 3D analysis.

7
 Figure 4. 3D Slope Stability Analysis - US Result Under Pseudo-Static Seismic Condition, Dashed Red line
indicates location section cut for 2D analysis

Figure 5. 2D Slope Stability Analysis - US Result Under Pseudo-Static Seismic Condition, Orange is Tailing,
Light green is Soil, dark grey is bed Rock (impenetrable). Marron is the compacted clay core, Light grey is
chimney and blanket filters. Yellow is rock fill

8
Without the MVM the model geometry would not accurately represent the actual design. which
could limit the extents over which the model could be applied and also limit the complexity obtained
with this 3D model.

CONCLUSION AND RECOMMENDATIONS

Use of 2D stability analysis at the US bend of Dam 1 resulted in an overly conservative stability
model. This conservatism can be acceptable if: the design criteria are being met, the resulting
configuration is reasonable, and/or there are other factors governing the design at the dam bend.
When the stability of the bend its-self is controlling the design, 2D models, which are not able to
properly account for the buttressing affects of concave slopes, will result in overbuild and possible
impractical designs.
It is important to acknowledge how 3D models in themselves are not immune to poor representations
of geometry. Understand the limitations of how the model is constructed is important, just as critical
as choosing a modeling package that provides the tools to accurately represent the geometry being
analyzed.
Use of MVMs is an effective way to overcome the limitations of 2.5D surfaces while still maintaining
the overall simplicity that 2.5D surfaces provide.

ACKNOWLEDGEMENTS
The writers would like to thank BCG Engineering for the case study data provided.

REFERENCES
Bray, J.D. and Travasarou, T. 2007. Simplified Procedure for Estimating Earthquake Induced
Deviatoric Slope Displacements. Journal of Geotechnical and Geoenvironmental Engineering, 133(4):
381-392

Leps, T. M., 1970, Review of Shearing Strength of Rockfill, Journal of Soil Mechanics and Foundation
Division, ASCE 96 (SM4), pp 1159-1170

SoilVision Systems Ltd. (2018) SVOFFICE 5 Help Manual – 4/18/2018, SoilVision Systems Ltd.,
Saskatoon, Saskatchewan, Canada.

Fredlund, M.D.; Lu, H.; Ivkovic, Z. Gitirana, G. (2017) ‘Multi-Plane Slope Stability Analysis of
Tailings Dams’, Tailings and Mine Waste, Banff, Alberta.

9
Geotechnical Characterization of a Limestone
Rockfill used in a Large Rockfill Tailings Dam
José Ccotohuanca1, José Ale1, Donald East1 and Leonardo Dorador2
1. Wood, Peru
2. WoodPLC, Chile

ABSTRACT
The construction of concrete face rockfill dams (CFRDs) has gained popularity in mining projects
worldwide during the last 20 years. Since its current design is still largely empirical and based on
historic performance of other dams worldwide, the study of the physical behaviour of rockfill
materials has a significant importance to improve the understanding of the behaviour of such type
of dams.
Among the available methods of testing to obtain the shear strength using the fine fraction of a coarse
material, rockfill strength is usually characterized by the parallel gradation method (Lowe, 1964).
This methodology is based on reducing the maximum particle size to enable the use of conventional
triaxial shear strength laboratory equipment on samples mainly composed by fluvial granular
materials (Gesche, 2002 & De la Hoz, 2007) and ballasts (Jernigan, 1998 & Cambio and Ge, 2007).
This paper discusses the work undertaken to verify the basis of the parallel gradation method for a
limestone rockfill material in a 150-meter-high rockfill dam by performing large-scale triaxial shear
tests under low and high confining pressures (from 49 kPa to 1960 kPa). The shear strength
parameters obtained by laboratory testing were compared with empirical approaches developed by
Leps (1970) updated by Dorador et al. (2017), Barton & Kjærnsli (1981) and Douglas (2002).

1
INTRODUCTION

The two most relevant factors influencing the design of rockfill structures are the shear strength and
the deformation characteristics. No stability analysis, regardless of how intricate and theoretically
exact it may be, can be useful for design if an incorrect estimation of shearing strength of the
construction materials has been made (Marachi et al., 1969). To achieve that any rockfill property be
meaningful, laboratory tests should simulate the field conditions as closely as possible, however, this
is generally done in the case of earth dams, where the size of particles are adequate to be tested
according to dimensions of laboratory testing devices. In the case of rockfill materials which may
contain particle sizes up to a metre in diameter, it is difficult to simulate field conditions in the
laboratory.
An indirect way to characterize rockfills is to conduct laboratory tests of the “fine” fraction (able to
be tested in a conventional laboratory device) of rockfill materials (called experimental approach).
Some techniques available for shear strength characterization include: the matrix method (Siddiqi,
1984), the scalping method (Al-Hussaini, 1983), the scalping-replacement method (Donaghe and
Torrey, 1979), and the most well-known, parallel gradation method (Lowe, 1964).
This paper presents a review of the limitations and applicability criteria of the parallel gradation
method (Lowe, 1964). This method has been successfully verified for certain types of materials such
as: fluvial granular materials (Gesche, 2002 and De la Hoz, 2007), and ballasts (Jernigan, 1998 and
Cambio and Ge, 2007), despite this, in common practice, the parallel gradation method has been
applied to other type of materials. The paper shows a detailed assessment of the applicability of the
parallel gradation method to a quarried limestone rockfill material, based on a large-scale triaxial test
performed as part of the design of case history of this paper.

CASE HISTORY: ROCKFILL DAM CHARACTERISTICS

The case history is a concrete faced rockfill dam (CFRD), part of a Tailings Storage Facility located in
the central Andes of Peru. The final height of the dam will be 230 m, which will store 570 Mt of
tailings, approximately. Upstream and downstream slopes are 1.4h:1.0v and 1.5h:1.0v, respectively.
The Figure 1 shows the typical cross-section and the materials distribution projected up to final phase.

2
 Downstream Final Dam Crest Dam Crest – Stage 3 Upstream
4730,0 masl 4658,0 masl
Dam Crest – Stage 2
Zones 4637,0 masl
Tailings Level
3 Main Rockfill Dam Crest – Stage 1
Elevation (masl)

4609,0 masl 4648,0 masl

3A Secondary Rockfill
Dam Crest – Starter Dam
2 Transition Layer 4582,0 masl
1 Filter Layer

Limestone Bedrock

Figure 1 Cross section of the rockfill tailings dam

GEOTECHNICAL CHARACTERIZATION
The geotechnical characterization of the case history (see Figure 1) includes 6 geotechnical materials.
For the purposes of this paper, the geotechnical characterization assessment will be focused only in
the rockfill which the dam is made of. The dam is directly founded on competent bedrock.

Parallel Gradation Method

The parallel gradation method consists in scaling the global particle size distribution curve (the model)
to a parallel curve (the prototype), considering the maximum particle size allowed by the testing
device. It was proposed by Lowe (1964), who supposed that the model sample could closely duplicate
the behaviour of the prototype sample, in terms of shear strength, as long as some requirements will
be met: a) the elastic modulus of the model sample must be the same of the prototype sample one (it
implies maintaining the same mineralogy and hardness of particles); and b) the surface roughness,
shape and crushability of particles of prototype sample must be properly represented in the
laboratory-allowed-size specimen. Then, with large-scale laboratory test results reported by De la
Hoz (2007), Gesche (2002) and Dorador (2010), new conditions have been added: The Cu (coefficient
of uniformity) and the Cc (coefficient of curvature). The values of these parameters should be close
between the model and the prototype samples gradation, as well as, the model sample gradation should
not contain more than 10% of fines (passing #200 mesh) due to the importance of fines in the
mechanical response of material.
Despite the requirements mentioned above, in the state-of-the-practice, the parallel method has been
extended to any type of materials with poor verifications, leading to wrong values for shear-
strength/elastic parameters, and as consequence, leading to a wrong geotechnical characterization.
For the case history, the main rockfill is composed of limestone, with maximum particle size of 1.5 m.
Due its large-particle size, the rockfill was reconstituted in the laboratory using the parallel gradation
method (Lowe, 1964). For this paper, 2 parallel curves (for 8” and 2” maximum particle size) were
created, verifying previously the following requirements:

3
  Modelled gradations should not contain more than 10% of fines: The 2 parallel curves of
rockfill (reconstituted in the laboratory) show fines contents lower than 10%. See Figure 2.
 Keep the Cu and Cc values similar: The Cu and Cc values for both the model and the
prototype samples are close (requirement assumed as satisfied). See Table 1.
 Same mineralogy and hardness of particles: The mineralogy and unconfined compressive
strength (UCS) values are the same between the model and the prototype samples.
 Same surface roughness and shape of particles: Both, the model and the prototype samples
are angular and have similar roughness because have same nature. See Figure 3.
Although main rockfill material satisfy the four requirements describes above, a laboratory testing
program was conducted for the case history, in order to verify that shear-strength/elastic parameters
are not affected. This procedure is presented in the following section.

Table 1 Representative diameters (D10, D30, D60) and the Cu and Cc coefficients

dmax (1) Representative diameters (mm) Particle size

Material Curve Cu Cc
(mm) D10 D30 D60 ratio (2)
Prototype 1500 (59”) 4,75 25,4 135,84 28,59 1,00 -
Main
Model 1 (100x180cm) 200 (8”) 0,60 4,06 21,09 35,15 1,30 7,5
rockfill
Model 2 (30x60cm) 50 (2”) 0,097 0,855 4,85 50,00 1,55 30,0
(1) Maximum particle size, (2) Maximum particle size on field (prototype sample) divided with maximum particle size of model.

Gravel Sand
Boulders Clay and Silt
Coarse Fine Medium Fine
1500 300 76.2 25.4 4.76 0.425 0.075
100

Main Rockfill
90
For CID test (100x180cm)
For CID test (30x60cm)
Accumulated percent passing (%)

80

70

60

50 Global particle
size distribution Parallel
40 Protoype Gradations
30 =1,00 Models
=28,59
20 Dmax=1500,0mm
D50=74,0mm
10
fines=2,5%
0
1000 100 10 1 0.1 0.01
Particle size (mm)

Figure 2 Global particle size distribution and parallel gradations of the rockfill materials

4
 (a) (b)

Figure 3 main rockfill particles (a) in field – large size, (b) in laboratory – small size

Large scale triaxial testing program

The rockfill characterization considered an extensive large-scale CD triaxial compression testing
program (see Table 2). The tests were performed with load-controlled method and effective
confinement pressures (49 kPa to 1960 kPa).

Table 2 Large-scale CD triaxial compression testing program

Rockfill Samples Dimensions Tmax Confining pressure (kPa) Density (t/m3)

M-1, M-2, M-3, M-4, 49, 98, 196, 294, 490 (*), 980 (*),
30 cm x 60 cm 2”
M-5, M-6, M-7 1960 (*) 2,25
M-8, M-9, M-10 100 cm x 180 cm 8” 490 (*), 980 (*), 1960 (*)
(*) Samples with different dimensions and same confining pressures to verify the parallel gradation method

Verification of parallel gradation method

On the M-5 to M-10 rockfill samples CD triaxial tests were carried out to verify the parallel gradation
method. The results are shown in Figure 4.
According to the results, the model 1 sample (Tmax 2”) represents successfully the shear strength
behaviour of the sample with coarser particles (model 2, Tmax 8”) as shown in Figure 4 (a, c). However,
at strains greater than 6%, the contractive behaviour is not the same in both samples. This difference
could be due to the occurrence of breakage in the particles of the coarser sample (see Figure 4(b)).
8000 2
Tmáx 2" Tmáx 2" '3 = 1960 kPa
1
7000
Tmáx 8" Tmáx 8" '3 = 980 kPa
0
6000 '3 = 1960 kPa '3 = 490 kPa
-1
Deviator Stress (kPa)

Volumetric Strain (%)

5000
-2

4000 -3

3000 '3 = 980 kPa -4

-5
2000
-6
1000 '3 = 490 kPa -7

0 -8
0.00 5.00 10.00 15.00 20.00 25.00 0.00 3.00 6.00 9.00 12.00 15.00 18.00 21.00 24.00
Axial Strain (%) Axial Strain (%)
(a) (b)

5
 0 1000 2000 3000 4000 5000 6000 7000 8000 9000
6000 48.0
160
'3 = 480 kPa
'3 = 980 kPa

'3 = 1960 kPa

Tmáx 2"
5500 46.0 Tmáx 2"
Tmáx 8"
 140
Tmáx 2"
5000 44.0 Tmáx 8"
Tmáx 8" '3 = 1960 kPa
4500 42.0 120
 = 40,0 E50 Tmáx 2"= 0.567'3 0.707
 = 38,7

Elastic Moduli E50 (MPa)

4000 40.0
 = 39,0
Shear Stress (kPa)

 = 38,5  100
3500 38.0
 = 38,5  = 38,5

 ()
3000 36.0 80
'3 = 980 kPa
2500 34.0
60 E50 Tmáx 8" = 0.359'3 0.751
2000 32.0

1500 30.0 40
1000 28.0 '3 = 490 kPa
20
500 26.0

0 24.0 0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 0 250 500 750 1000 1250 1500 1750 2000 2250 2500
Normal Stress (kPa) Confining Pressure (kPa)

(c) (d)
Figure 4 (a) deviator stresses, (b) volumetric strains versus axial strains, (c) Mohr envelopes, (d) elastic modulus
vs confining pressure

Regarding to Figure 4(d), the model sample suggests greater values of elastic modulus, which would
lead to the estimation of smaller deformations than the actual one, if a numerical deformations
analysis is performed.
It must be recognized that, despite complying with the four requirements of the parallel gradation
methodology, all the mechanical properties are not reproduced effectively.

Rockfill shear strength estimates

Due to the fact that the parallel gradation technique is satisfied for the rockfill of the history case, it
was extended to triaxial CD compression tests in samples with dimensions of 30 cm x 60 cm (M-1 to
M-4). The results in terms of peak internal friction angle for each confining pressure are plotted in
Figure 5(a), where it shows a clearly non-linear behaviour of the rockfill shear strength, represented
3
by the potential equations 5 .𝜎 (kPa) (De Mello, 1977). On the other hand, the elastic
modulus curve in Figure 5(b) suggests two remarkable trends for low and high confining pressures
values.
0.00 1000.00 2000.00 3000.00 4000.00 5000.00 6000.00 7000.00 8000.00 9000.00
6000 52.0 160
 49 kPa = 50,8
5500  98 kPa= 50,7 50.0 E50 = 11.45'30.303
'3 = 1960 kPa

140
'3 = 294

5000 48.0 R² = 0.48

 5  3
kPa

4500 46.0 120

'3 = 196 kPa

4000  196 kPa = 43,7 44.0 E50 = 1.12'30.815

Elastic Moduli E50 (MPa)
Shear Stress (kPa)

100
'3 = 98 kPa

R² = 0.93
3500 42.0
 294 kPa= 40,3
 ()

3000  490 kPa= 40,0 40.0 80

 1960 kPa = 39,0
2500  980 kPa= 38,7 38.0 E50 = 0.567'30.707
'3 = 980 kPa

49 kPa - M-1 60 R² = 0.99

2000 98 kPa - M-2 36.0
'3 = 490 kPa
'3 = 49 kPa

1500 196 kPa - M-3 34.0 40

294 kPa - M-4
1000 490 kPa - M-8 32.0
980 kPa - M-9 20
500 30.0
1960 kPa - M-10
0 28.0 0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
0 250 500 750 1000 1250 1500 1750 2000
Normal Stress (kPa) Confining Presussre (kPa)

(a) (b)

Figure 5 (a) Mohr envelopes and non-linear shear strength envelope, (b) Elastic modulus vs confining pressure

6
Empirical approaches for rockfill shear strength estimate

The shear strength of rockfill materials depends on several parameters such as: the confining
pressure, grain-size distribution, density, fines content, particle strength, roughness, shape, etc. To
take into account these factors, a laboratory testing program can be performed; however, in many
projects, due to the lack of large size devices (for shear strength testing), the rockfill shear strength is
usually estimated using values obtained from existing literature. One of the most commonly cited
literature sources is Leps (1970). Other less adopted approaches are Barton & Kjærnsli (1981) and
Douglas (2002), which use the specific characteristics of rockfill material.
Barton & Kjærnsli (1981) proposed a shear strength prediction equation based on the link between
rock joint and rockfill behaviour, the authors verified their model with experimental data from the
Oroville amphibolite rockfill dam (Marachi et al., 1972) and found a good match, however,
Indraratna et al. (1993) based on the triaxial test data of a greywacke rockfill, concluded that the
Barton & Kjærnsli approach should not be extended to extrapolate the frictional behaviour of rockfill
at high normal pressures. An analysis of the samples tested in both cases, where the first one
(amphibolite) had stronger particles than the second (greywacke), suggests that for the stronger
particles, Barton & Kjærnsli method offers a better correlation. On the other hand, Douglas developed
two equations based on a large database consisting of published and unpublished triaxial testing
results on rockfills strength which were statistically analysed. The equations consider variations in
key factors which could impact on the shear strength of rockfill materials.
A summary of the key factors, shear strength prediction equations and their criteria considered in
each empirical method, as well as, the parameters assumed for case history rockfill, is shown in
Table 3.

Table 3 Summary of empirical approaches to shear strength estimation

Key Factors Prediction Parameters for case history

Method Criteria
G F D U S R L equations rockfill
For low density, poorly UCS testing on samples from quarry
.446′
.
graded and weak particles area produced an average UCS value of
(UCS 3,5 – 17 MPa) 98,8 MPa, indicating that rockfill can be
Leps For average rockfill (UCS 17 classified as average to strong rock
. 3 ′
.
x x x – 70 MPa) particles. It was adopted an average
(1970)
rockfill bound.
For high density, well .
. 3 ′ , ( and ′ en kPa)
. 6 ′
.
graded and strong particles
(UCS 70 – 210 MPa)
For low density, poorly *This updated consider the Leps
′ . 3
graded, weak particles and approach but only with historical data
limestone rock of limestone rockfill.
Average UCS value of 98,8 MPa, well
Leps For average limestone
.45 ′
.
graded ( =1,00 y =28,59) and
x x x x rockfill
updated* limestone rock. It was adopted an
average curve between average to high
For high density, well
limestone rockfill bound.
4′
. 11
. graded, strong particles and . 1
limestone rock .64 ′ , ( and ′ en kPa)

7
 Key Factors Prediction Parameters for case history
Method Criteria
G F D U S R L equations rockfill
Considering the following
𝑅 = 7 (high porosity) to 10 characteristics: 𝑛=19%, quarried origin,
(low porosity) D50=74 mm, UCS=40MPa (minimum)
Barton & 𝑆 = 0,2 (strong particles) to and 98,8 MPa (average), results in:
𝑆
Kjærnsli x x x x x ′ 𝑅 𝑙𝑜𝑔 ( ) + 𝑏 0,3 (weak particles) times For UCS= 40,0 MPa:
′ UCS ′  log( 4⁄′ ) + 3
(1981) 𝑏 =25° (poor rock quality) a For UCS = 98,8 MPa:
35° (good rock quality). ′  log( 5 ⁄′ ) + 3 ,
(′ en MPa)
Considering ANG=2, FINES= 2,5%,
Cc=1,00, UCS=40 MPa (minimum) and
Douglas 𝑎 𝑏: 𝑓(ANG FINES Cc UCS), 98,8 MPa (average), results in:
(2002) c = -0,3974 For UCS= 40,0 MPa:
x x x x ′ 𝑎 + 𝑏′𝑛
1st ANG 0 (rounded particles) ′ 66 + 4 ′ − .3 4
to 3 (crushed angular rock) For UCS = 98,8 MPa:
approach ′ 36 + ′ − .3 4,
(′ en kPa)
Considering n=19%, dmax= 200 mm
(maximum value of Douglas’s data),
Douglas 𝑅𝐹𝐼: 𝑓(𝑒𝑖 ANG dmax FINES ANG=1 (angular), FINES=2,5%,
(2002)  UCS),  = -0,8726 UCS=40 MPa (minimum) and 98,8 MPa
x x x x x 1′ 𝑅𝐹𝐼′3 ANG 1 (angular) and 0 (average), results in:
2nd
(otherwise) For UCS= 40,0 MPa or 98,8 MPa
approach (indistinct):
. 6
1′ 5 5′3 (1′ y ′3 en MPa)
Notes:
G: grain-size distribution, F: fines content, D: density/porosity, U: unconfined compressive strength, S: shape, R: roughness, L: lithology; :
shear strength, ′𝒏 : normal stress, ′𝟏 , ′𝟑 : maximum and minumum principal stress, ′ : peak drained friction angle, 𝑹: equivalent roughness, 𝑺:
equivalent strength, 𝒃 : basic friction angle, ANG: angularity, 𝐅𝐈𝐍𝐄𝐒: percentage of fines content, Cc: coefficient of curvature, UCS: unconfined
compressive strength of the rock substance, 𝐧: porosity, 𝐝𝐦𝐚𝐱: maximum particle size.

A comparison of the rockfill shear strength envelopes based on both empirical and experimental
approaches is shown in Figure 6. For a better representation at small normal stress Figure 6(a) relates
the peak friction angle versus the normal stress; for large normal stress, the Figure 6(b) plots the shear
strength versus the normal stress. Notice that the Leps updated method considers a maximum peak
friction angle of 50 at normal stresses lower than 0,07 MPa (because it corresponds the range without
historical test data).

8
 65 4
Leps (1970) bounds Leps (1970) bounds
Indraratna et al. (1993) bound Indraratna et al. (1993) bound
Leps (1970) Leps (1970)
Leps updated 3.5 Leps updated
60 Barton & Kjærnsli (1981) UCS=40 MPa Barton & Kjærnsli (1981) UCS=40 MPa
Barton & Kjærnsli (1981) UCS=98,8 MPa Barton & Kjærnsli (1981) UCS=98,8 MPa
Douglas (2002) 1st approach UCS=40 MPa Douglas (2002) 1st approach UCS=40 MPa
Douglas (2002) 1st approach UCS=98,8 MPa 3 Douglas (2002) 1st approach UCS=98,8 MPa
55 Douglas (2002) 2nd approach Douglas (2002) 2nd approach
Material 3-test data specimen 100 cm x 180 cm Material 3-test data specimen 30 cm x 60 cm
Material 3-test data specimen 30 cm x 60 cm Material 3-test data specimen 100 cm x 180 cm
Historical tests data on limestone rockfill 2.5 Historical tests data on limestone rockfill
Experimental approach Experimental approach
50
 max (°)

(MPa)
2

45
1.5

40
1

35
0.5

30 0
0.01 0.10 1.00 10.00 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
n (MPa) n (MPa)

(a) (b)

Figure 6 Comparison between empirical and experimental approaches to shear strength of rockfill case history
(a) peak friction angle versus normal stress, (b) shear strength versus normal stress

CONCLUSIONS

A comparative analysis of the triaxial tests results in two modelled samples of limestone rockfill with
different maximum particle size (maximum size of 2" and 8") demonstrated that it is feasible to obtain
the shear strength parameters applying the parallel gradation method.
The deformation parameters are overestimated as the maximum particle size decreases. This could
lead to obtain smaller values of deformation in numerical analyses based on laboratory test results in
small-scale samples.
The empirical methodologies regarding the rockfill shear strength estimation, such as: Leps (1970),
Leps updated, Douglas (2002) and Barton & Kjærnsli (1981), give satisfactory and compatible results
with the large-scale triaxial tests. The relationship of Barton & Kjærnsli (1981) gives the smallest error
range.
Practitioners should consider the additional information in the current literature to reduce
uncertainties in developing shear strength estimates for rockfill materials. Depending on the phase
of the design, a potential approach would be the flow-chart shown in Figure 8.

9
 Figure 8 Potential approach to rockfill shear strength estimates

REFERENCES

Al-Hussaini, Mosaid, 1983 ‘Effect of particle size and strain conditions on the strength of crushed
basalt’, Canadian Geotechnical Journal, 20(4): 706-717.

Barton, N. & Kjaernsli, B. (1981) ‘Shear strength of rockfill’, Journal of Geotechnical Engineering Div.
ASCE, Vol. 107, No. 7, pp. 873 - 891.

Cambio, D. and Ge, L. (2007) ‘Effects of Parallel Gradation on Strength Properties of Ballast
Materials’, Advances in Measurement and Modeling of Soil Behavior.

De La Hoz, K. (2007) Estimación de los parámetros de resistencia al corte en suelos granulares gruesos, Tesis
de Magíster en Ciencias, Universidad de Chile.

De Mello, V. (1977) ‘Reflection on design decisions of practical significance to embankment Dams’,

Géotechnique, 27(3):279-355.

10
Donaghe, R. T. y Torrey, V. H. (1979) ‘Scalping and replacement effects on strength parameters of
earth-rock mixtures’, Proc. Conf. on Design Parameters in Geotechnical Engineering, London, vol.2, pp.29-
34.

Dorador, L. (2010) Análisis experimental de las metodologías de curvas homotéticas y corte en la evaluación
de propiedades geotécnicas de suelos, Tesis de Magíster en Ciencias de la Ingeniería, Universidad de
Chile.

Douglas, K.D, (2002) The shear strength of rock masses. PhD Thesis School of Civil and Environmental
Engineering, The University of New South Wales, Sydney.

Dorador L; Cunning J; Junqueira F; Hawley M (2017). Material characterization in M. Hawley & J.

Cunning (Eds), Guidelines for Mine Waste Dump and Stockpile Design. CRC Press. ISBN
9781138197312. pp 79-98.

Gesche, R., (2002) Metodología de evaluación de parámetros de resistencia al corte de suelos granulares
gruesos. Tesis de Ingeniero Civil, Universidad de Chile, Chile.

Indraratna, B., Wijewardena, L., Balasubramaniam, A. (1993) ‘Large-scale triaxial testing of

greywacke rockfill’, Géotechnique 43(1), pp. 539-543.

Jernigan, R.L., (1998) The Physical Modeling of Soils Containing Oversized Particles. PhD Thesis
Department of Civil, Environmental and Architectural Engineering. University of Colorado at
Boulder.

Leps, T.M. (1970) ‘Review of shearing strength of rockfill’, Journal of the Soil Mechanics and Foundations
Division, 96 (SM4), pp. 1159-1170.

Lowe, J. (1964) ‘Shear strength of coarse embankment dam materials’, Proc. 8th Congress on Large
Dams, Vol. 3, pp 745 - 761.

Marachi, N.D., Chan, C.K., Seed, H.B., Duncan, J.M. (1969) Strength and deformation characteristics
of rockfills materials, Report No. TE-69-5, Department of Civil Engineering, University of California,
Berkeley,

Marachi, D., Chan, C. & Seed, H. (1972) Evaluation of properties of rockfill materials, Journal of the Soil
Mechanics and Foundations Division, Vol. 98, No. 1, pp. 95 - 114.

Siddiqi, F.H. (1984) Strength evaluation of cohesionless soils with oversize particles, PhD thesis University
of California, Davis.

11
Evaluation of the Stability of the Embankment Dams
of a Tailings Storage Considering Unsaturated
Properties of Soils
Daniel Guzmán and María Teresa Zandarín
JRI Ingeniería, Chile

ABSTRACT

Nowadays, the stability analysis performed for embankment dams considers the saturated properties
of the soils, because this properties are measured with no difficult performing consolidated and
saturated triaxial tests. Even though, the soils over the water table remain unsaturated, if climate and
drainage system of the dam allow this. But, it is not a common practice to determine the unsaturated
properties of soils for projects due to the costs and time necessary to perform unsaturated triaxial
tests.
Considering the mentioned above, in this article it is presented a static stability analysis of the
embankment dam modelling the soil using the Extended Mohr Coulomb failure criterion. This
criterion considers the influence of the matric suction on the unsaturated shear strength of soils. To
do this, the unsaturated soils properties were calculated applying the models proposed by (Zapata et
al. 1999), (Fredlund & Xing, 1994), and (Khalili & Khabbaz, 1998). Also, sensitivity analyses were
done, considering different grain size distribution of soils, and considering a rainfall.
Finally, from the model results, a discussion about the advantages of including the unsaturated soils
properties for the safe stability analysis is given.

1
INTRODUCTION

A partially saturated or unsaturated soil is a state of the soil (Gens & Alonso, 1992). The main
difference between this states is the presence of air between their pores, i.e. of negative water
pressures called matric suction. This matric suction caused an increment of the shear strength of the
granular skeleton due to capillary forces.
In an earth compacted dam the pore water pressure are negative with respect to atmospheric pressure
above the water table. Generally, the contributions of these pressures on the shear strength of soils
are ignored to preform stability analysis. It may be reasonable not consider them when most of the
slip surface is below the water table. However, when the water table is relatively deep, or the
environmental conditions are favorable (evaporation is greater than rainfall), these contributions to
the strength should be considered (Fredlund et al. 2012) and (Oh & Lu, 2015).
Then, the stability of the slope and the water flow through it depends on the saturated/unsaturated
condition of the soils. Better results are obtained when the analyses considers the evolution of the
degree of saturation of soils during a rainfall. (Oh & Lu, 2015).
The variation of the degree of saturation of the soil can be represented by the Soil-water characteristic
curve (SWCC). This curve and its value of air entry (AEV) can be estimated from the grain size
distribution of the soil, by means of the correlations of (Zapata et al. 1999). After, the Effective stress
parameter (X) and ∅b are calculated with the model proposed by (Khalili & Khabbaz, 1998), which
considers this parameters dependent on SWCC. Finally, the AEV and ∅b are incorporated into the
stability analysis using a Mohr Extended Coulomb (MEC) criterion.

UNSATURATED SHEAR STRENGTH OF SOILS

To analyze the stability of slopes in unsaturated soils, it is necessary to determine the changes in the
negative pressures of the soil and include them in the calculations of effective stresses. (Bishop, 1959)
introduced the X factor and suggested the following equation for effective stresses.

𝜎′𝑧 = (𝜎𝑧 − 𝑢𝑎 ) + 𝑋(𝑢𝑎 − 𝑢𝑤 ) (1)

Where:
𝜎′𝑧 Normal effective stress (kPa)
𝜎𝑧 Normal total stress (kPa)
𝑢𝑎 Air pressure (kPa)
𝑢𝑤 Water pressure (kPa)
X Effective stress parameter which varies between 0 and 1

2
Extended Mohr Coulomb (EMC) criterion

(Fredlund et al. 1978) proposed an EMC failure criterion for unsaturated soils. They introduced an
additional parameter ∅b that captures the linear increment of the shear strength with the matric
suction (Equation 2). Subsequently, (Escario & Saez, 1986), proposed a non-linear relationship
between ∅b and ∅'as indicated by Equation (3), using the Effective stress parameter.
Then, the EMC criterion is expressed by Equation (4). In the Figure 1 it is show the EMC curve and
the relationship with the matric suction.

𝜏 = 𝑐 ′ + (𝜎𝑛 − 𝑢𝑎 )𝑡𝑎𝑛∅′ + (𝑢𝑎 − 𝑢𝑤 )𝑡𝑎𝑛∅𝑏 (2)

𝑡𝑎𝑛∅𝑏 = 𝑋𝑡𝑎𝑛∅′ (3)
𝜏 = 𝑐 ′ + (𝜎𝑛 − 𝑢𝑎 )𝑡𝑎𝑛∅′ + (𝑢𝑎 − 𝑢𝑤 )𝑋𝑡𝑎𝑛∅′ (4)

Soil-water characteristic curve (SWCC)

The SWCC is defined as the relationship between water content (degree of saturation, gravimetric or
volumetric water content) and the suction of soil (Figure 1). This curve is strongly linked to the grain
size distribution and soil structure, so SWCC varies according to the type of soil.

Figure 1: Relationship between SWCC and shear strength. (Modified after Fredlund, 2000)

3
SWCC by (Fredlund & Xing, 1994)

There are several models that mathematically model the SWCC function, based on data obtained in
laboratory tests, e.g. (Van Genuchten, 1980) and (Fredlund & Xing, 1994). Equation (5) shows the
adjustment proposed by the last mentioned authors.
𝑚
ѱ
1 − ln (1 + ) 1
ѱ𝑟 (5)
𝜃 = 𝜃𝑆 [ ][ 𝑛 ]
1000000
ln⁡(1 + ) ln⁡[𝑒 + (ѱ) ]
ѱ𝑟 𝑎

Where:
𝜃 Volumetric content of water
𝜃𝑆 Saturated Volumetric content of water.
ѱ Soil suction (kPa)
ѱ𝑟 is the residual soil suction (kPa)
𝑎 is the soil parameter that is related to the air entry value (AEV) (kPa)
𝑛 is a model parameter that controls the slope at the inflection point in the SWCC
𝑚 is the parameter that is related to the residual water content

Correlations for estimate the SWCC by (Zapata et al. 1999)

When laboratory data of SWCC are not available, it is possible use a model based on index properties
of a soil: mass-volume and grain size distribution to estimate it, e.g. (Fredlund & Wilson, 1997);
(Aubertin et al. 2003); (Zapata et al. 1999).
In this study was used the model proposed by (Zapata et al. 1999) for cohesionless soils (PI=0) to
calculate the parameters:⁡𝑎,⁡𝑚, 𝑛 and ѱ𝑟 of the model describe above. The equations are:

𝑎 = 0,8627(𝐷60 )−0,751 (6)

𝑚 = 0,1772 𝑙𝑛(𝐷60 ) + 0,7734 (7)
𝑛 = 7,5 (8)
1 (9)
ѱ𝑟 = 𝑎 ( )
𝐷60 + 9,7𝑒 −4
Where:

D60 Diameter in the grain size distribution corresponding to 60% of the cumulative mass

4
Estimating the X Model by (Khalili & Khabbaz, 1998)

The model proposed by (Khalili & Khabbaz, 1998) estimates the Effective stress parameter (X) of the
Equation (4) and the angle φb of the Equation 3, using the parameter "𝑎" of the Fredlund & Xing
model. Following, it is presented the mathematical expression of the model.

𝑢𝑎 − 𝑢𝑤 −0,55
[ ] For 𝑢𝑎 − 𝑢𝑤 > 𝑎
𝑎
X=

1 For 𝑢𝑎 − 𝑢𝑤 > 𝑎 (10)

In this way, Equation (11) shows the EMC criterion for unsaturated soils with the implementation of
Effective stress parameter of (Khalili & Khabbaz, 1998).

𝑢𝑎 − 𝑢𝑤 −0,55 (11)
𝜏 = 𝑐 ′ + (𝜎𝑛 − 𝑢𝑎 )𝑡𝑎𝑛∅′ + 𝑡𝑎𝑛∅′ [ ]
𝑎
The models explained before were used to analyze the stability of an embankment dam of tailings
storage.

EMBANKMENT DAM STABILITY ANALYSIS

Geometry and Geotechnical Properties of Soils

A homogenous 70-meter-high embankment dam has been selected for the analysis. The soil
foundation of the embankment consists on alluvial soil with 30 meters of thickness. And the upstream
slope of the dam is sealed with a geomembrane.

Figure 2: Geometry and Soils of tailings storage

Table 1 shows the mechanical and hydraulic properties of the soils used in the numerical model. Only
the embankment dam was considered partially saturated.

5
Table 1: Geotechnical properties of materials

Material γ (kN/m3) Ф’ (deg) c’ (kPa) Su/s’v KSAT (m/s) θs θr

Foundation 18 36 10 - 1x10-6 - -

Embankment Dam 21 37 0 - 1x10-5 0,30 0,05

Tailings 17 - - 0,10 1x10-7 - -

Drain 18 36 - - 1x10-4 - -

Geomembrane 10 37 0 - 1x10-12 - -

The range of grain size distribution for the embankment dam soil is shown in Figure 3. This range is
a typical for embankment fills. Three soils have been selected (A, B and C), with different D60 (0.6
mm, 5 mm, 10 mm) representing the lower, average and upper limit of the range. This soil is frictional
with a specific gravity of 2.78, void ratio of 0.43; and dry density of 19.5 kN/m3

Figure 3: Grain size distribution for Embankment dam

The Figure 4, shows different SWCC obtained for different D60 using the model of (Zapata et al. 1999).
The Figure 5 shows the relationship between shear strength and matric suction using the model of
(Khalili & Khabbaz, 1998).
The slope of functions (Tanφb) obtained from method of (Khalili & Khabbaz, 1998) is calculated by
averaging the increases greater than AEV up to 650 kPa of suction as the limit of interest, due to the
70 meters of height of the dam. The angle for smaller suctions to AEV is φ'

6
Table 2: Fredlund & Xing parameters

Soil D60 (mm) AEV (kPa) n m ѱ𝑟 (kPa) Фb (deg) Ф’ (deg)

Soil A 0,6 1,3 7,5 0,7 2,1 6,7 37

Soil B 2,5 0,4 7,5 0,9 0,2 5,2 37

Soil C 10 0,2 7,5 1,2 0,02 3,3 37

Figure 4: SWCC for Soil A, B and C according to Zapata et al. (1999)

Figure 5: Shear strength vs Matric suction according to Khalili & Khabbaz. (1998)

7
Analyzed cases

Four stability analysis were performed considering Soil A, B, C in unsaturated conditions and Soil B
in saturated conditions with different flow regimes (with and without rainfall). The analysis were
performed using the software SLIDE v.8 (Rocscience, 2017).
The effect of rainfall infiltration on the stability of the slope using transient regime, was evaluated by
applying a rainfall rate of 135 mm/d, with a return period of 100 years, recorded in the weather station
”Rivadavia” at Coquimbo (Manual de Carreteras, 2018). This rainfall has a duration of 2 consecutive
day.
The safety factor (FoS) is determined by the method of Morgenstern & Price (1965) under the EMC
criterion, on global circular slip surfaces.

Table 1: Cases analyzed

Cases Considerations

Saturated (Conventional analysis) Steady State

Day 1 to 2: Rainfall event (constant 135 mm/d)

Unsaturated Soil A, B and C
Day 3 to 30 : Pore pressure Dissipation

ANALYSIS OF RESULTS
Figure 6 shows the result of the global stability analysis in saturated condition and steady state
regime, without considering the influence of matric suction on the shear strength of the embankment
dam (conventional method). The static safety factor is FoS =1,65 and the hydrostatic negative
pressures do not exceed 650 kPa.

8
 Figure 6: Saturated stability analysis

In the Figure 7 it is shown the unsaturated FoS during day 0, (i.e. before the rain begins). These factors
are always greater than the saturated FoS achieved by conventional analysis. The FoS of the “Soil C”
exceeds by 5% the conventional FoS, however, the “Soil A” reaches a difference of 10%. These results
confirm that the greater matric suction is, greater the shear strength in the type of soil analysed.

Figure 7: Unsaturated FoS vs. Saturated FoS

Figure 8 shows the variation of FoS for the same slip surface that was obtained in the conventional
saturated case. The FoS variation was studied during the 2 consecutive days of rain and its evolution
during 28 after the rain. It can be seen that at the end of the rainfall, all the simulations recorded their
lowest FoS. “Soil C”, reaches lower FoS, however, it is always above saturated FoS (conventional
method). Also, for this rainfall intensity, the fluctuation of the FoS does not exceed ΔFoS=-0,02. Five
days after the end of the rain, the FoS slowly stabilizes close to the initial FoS.

9
 Figure 8: Evolution of FoS during and after a rainfall

Figures 9, 10 and 11 show the variation of negative pore pressures inside the Embankment dams, for
each type of soil. The critical condition occurs during the second day (t = 2 days), here the pore
pressures show the greatest increase in the surface area of the slope downstream. In the lower half of
the dam composed of soil C (Figure 11), a significant loss of suction is recorded, around 300 kPa.
However, this situation does not generate risks in the global stability of the slope.
In addition, after 5 days of analysis (t = 5 days), all the dams begin to stabilize their pore pressures,
mainly due to the high permeability of the soil, causing a gradual increase in the global FoS, as shown
in Figure 8.

Fig. 9. Soil A, D60=0,6mm. Contours of simulated pore-water pressure at different times for the model slope
under projected precipitation

10
Fig. 10. Soil B, D60=2,5mm. Contours of simulated pore-water pressure at different times for the model slope
under projected precipitation

Fig. 11. Soil C, D60=10mm. Contours of simulated pore-water pressure at different times for the model slope
under projected precipitation

Then, considering saturated condition of soils gives conservative results for the stability analysis.
Although this hypothesis is on the side of safety, it is important to consider a model with a correct
calibration of the output flow from drainage system and the height of water table inside the dams,
since a misrepresentation of the water table can yield results even more conservatives.

CONCLUSIONS
A methodology to perform stability analysis considering the unsaturated conditions was developed.
The unsaturated shear strength of soil was calculated using the grain size distribution, (Zapata et al.
1999) and (Khalili & Khabbaz, 1998). This procedure is very useful when there is no laboratory data
on unsaturated soils. For the soil with higher D60, was obtained the lower FoS due to, the residual
moisture is lower. If a rainfall occur the FoS decrease but not reach the saturated FoS. Then, the FoS
obtained for saturated condition is conservative. The approach of the unsaturated stability analysis
is considered more realistic.
The contribution of the negative pressures on the shear strength of the soils should be considered for
develop project of embankment dam of great dimension. This justify the execution of unsaturated
laboratory tests.
It is recommended to continue with this unsaturated approach, mainly analyzing the influence of
different rainfall intensity and permeability of soil on stability analyses. In addition, it is

11
recommended to apply the unsaturated analysis in a fully coupled environment to quantify the
deformations that may occur in the embankments dams under dynamic loads.

REFERENCES

Aubertin, M., Mbonimpaa, M.Bussièreb, B., & Chapuisa, R. (2003). A physically-based model to
predict the water retention curve from basic geotechnical properties. . Canadian Geotechnical Journal
40, 1104–1122. .

Bishop, A. (1959). The principle of effective stress. Teknisk Ukeblad 106, N° 39, 859-863.

Escario, V., & Saez, J. (1986). ‘The shear strength of partly saturated soils. Geotechnique, 36 , 453–456.

Fredlund, & Xing. (1994). Equations for the soil-water characteristic curve. Canadian Geotechnical
Journal 31, 521-532.

Fredlund, D. (2000). The 1999 R.M. Hardy Lecture: The implementation of unsaturated soil mechanics
into geotechnical engineering. Canadian Geotechnical Journal, 37, 963-986.

Fredlund, D. G., Morgenstern, N. R., & Widger, R. A. (1978). The shear strength of unsaturated soils.
Canadian Geotechnical Journal, 15, 313–321.

Fredlund, M. D., Rahardjo, H., & Fredlund, D. (2012). Unsaturated Soil Mechanics in Engineering
Practice. Canada: John Wiley & Sons, Inc.

Fredlund, M., Fredlund, D., & Wilson, G. (1997). Prediction of the soil-water characteristic curve from
grain-size distribution and volume–mass properties. Proceedings of the 3rd Brazilian Symposium on
Unsaturated Soils, Rio de Janeiro, 22–25 april 1997, 13-23.

Gens, A., & Alonso, E. (1992). A framework from de behaviour of unsaturated expansice clays.
Canadian Geotechnical Journal 29, 1013-1032.

Khalili, N., & Khabbaz, M. H. (1998). A unique relationship for X for the determination of the shear
strength of unsaturated soils. Géotechnique 58, N°5, 681-687.

Lu, N., & Likos, W. (2004). Unsaturated soil mechanics. Canada: John Wiley & Sons, Inc.

MOP. (2018). Manual de carreteras. Volumen 3.

Morgenstern, N., & Price, V. (1965). The analysis of the stability of general slip surface. Geotechnique
15 (4), 289–290.

Oh, S., & Lu, N. (2015). Slope stability analysis under unsaturated conditions Case studies of rainfall-
induced failure of cut slopes. Engineering Geology 184, 96-103.

Rocscience. (2017). Slide Version 8.0 - 2D Limit. 2D Limit Equilibrium Slope Stability Analysis.

12
Van Genuchten, M. (1980). A closed-form equation for predicting the hydraulic conductivity of
unsaturated soils. . Soil Sci. Soc. Am. J. 44, 892–898.

Vanapalli, S., Fredlund, D., Pufahl, D., & Clifton, A. (1996). Model for the prediction of shear strength
with respect to soil suction. Canadian Geotechnical Journal 33, 379-392.

Zapata, C. E., Houston, W., Houston, S., & Walsh, K. (1999). Soil-Water Characteristic Curve
Variability. ASCE Geotechnical Special Publication, 84-124.

13
Impacts of Topography Quality on Dam Breach
Assessments
Alexandra Halliday1 and Alfredo Arenas2
1. Golder Associates, Australia
2. Golder Associates, Chile

ABSTRACT
Dam breach assessments (DBA) include estimating the location at which a dam may fail, the volume
and properties of material that may be released and finally predict where the released material may
travel. The purpose of a DBA will vary, however they generally aim to support mine operators in
devising emergency action plans and classifying the consequence of their TSF, assisting in the
implementation of design measures and operational controls on their facilities.
The area impacted during a dam breach will vary depending on the volume released and the flow
properties, but also on the availability and quality of the topographic/bathymetric data downstream
of the TSF. The impacted area can be amplified if tailings find their way into a watercourse but can
be stopped short if a breach occurs on an uphill topography, these changes in topography are critical
to the results of a DBA and need to be captured in the topographic survey for the assessment to be
an accurate representation of the site conditions and potential breach.
Common industry practices adopt the use of numerical modelling software as tool to undertake DBA.
This software allows users to input site specific topography for each DBA. Generally, a survey will
be localized to TSF site boundaries, excluding susceptible receptors and flow channels that a breach
may impact.
Like any modelling software, these packages each have their sensitivities which, for the most part,
are well understood. A sensitivity which has not yet been established is the impact that a low-quality
survey can have on the results of a DBA. This paper presents the impact that a low-quality survey
can have on the results of an analysis by comparing topographies from two site specified aerial
surveys with the corresponding free-sourced surveys.
The results of this study show that the topographic survey quality can have a significant impact on
the dam breach characteristics and inundation area, potentially causing larger runout areas, greater
flow volumes and more rapid flood streams when a free-sourced and low-resolution topography is
utilized as a substitute for a high-resolution site-specific survey.

1
INTRODUCTION

Prior to undertaking a DBA, a common client query is related to the survey they will need to provide.
There are no designated guidelines that specify a level of quality that a topographic survey must be
in order to undertake a DBA, nor are there any guidelines that specify what constitutes a quality
survey. This implies, that theoretically freely sourced survey data could be used for DBA’s, but with
what impact? Limitations exist within software which may also impact the size and quality of the
survey that can be used to carry out a DBA.
This study presents the impact that a low-quality survey can have on the results of an analysis by
comparing the same topography in two different survey qualities; a site-specific aerial survey and a
free sourced survey. To obtain a broader understanding of the impact, two different terrain types
will be compared; a valley terrain with a narrow channel and a valley terrain which widens to a large
and flat area. The detailed topographic surveys have been obtained from site-specific areal
topographies with point densities of 1 m, which for the purpose of this study has been classified as
high-resolution. The free-sourced surveys have been obtained using Geo Meshing – which has the
functionality to generate gridded topographic surveys using Bing Maps. The free-sourced surveys
were obtained from the same locations as the aerial surveys and have a point density of 20 m, which
for the purpose of this study has been classified as low-resolution topography.
A hypothetical dam has been modelled at the top of each of the valleys. The dam dimensions and
material properties have been assumed and have been used to estimate a realistic hypothetical
tailings breach volume for the dam size, the tailings rheology has also been assumed.
FLO-2D, a software used for fluid flow modelling, has been used to assess the downstream
inundation area simulated by the hypothetical breach. Besides the topography, all components input
into the model were consistent for both the low-resolution and high-resolution surveys.
The results of this study aim to provide interested stakeholders with an indication on if there is
importance in investing in a high-resolution site-specific survey when undertaking DBA, or if a free-
sourced survey will provide a similar result.

METHODOLOGY

Often free-sourced surveys have higher point densities in populated areas, with lower density in
more remote locations. Generally, dams are situated outside of largely populated cities and as such
the free-sourced survey is often of low-resolution. For the purposes of this paper, the quality of a
topographic survey has been defined as a function of the point density. A survey with a higher
resolution (greater number of points per unit area) is considered good quality, and a survey with a
lower resolution (smaller number of points per unit area) is considered a lower quality survey.
Once the definition of “high-resolution” and “low-resolution” was established, two locations were
chosen to undertake the assessments. One location is situated in a narrow valley topography
(referred to as Topography A). The second location is also situated in a valley, which widens to a flat

2
wide-open space (referred to as Topography B). Both topographies have a survey resolution of one
point every 1 m.
The geographical location of the two aerial surveys was then used to obtain the free-sourced version
of the topography. The topography was obtained using Geo Meshing and the inbuilt functionality to
obtain surface topographies from Bing Maps.
A dam geometry, and corresponding tailings rheology was then estimated to obtain the hypothetical
breach volume. This volume was estimated using a geometrical approach and the corresponding
breach hydrograph was developed in accordance with this breach volume. The same breach
hydrograph and rheology was adopted for all models.
The breach rheology was estimated to be a ‘homogeneous’ mix of tailings, dam construction material
and water. This hypothetical information, and each of the four topography surveys was then
modelled in FLO-2D to estimate the impact of each survey on the DBA. The key parameters used for
assessing the breach impact were; inundation area, maximum velocity of flow, runout distance,
maximum flow depth, final flow depth and final flow velocity. This information was used to
compare the impact of a high-resolution survey against a low-resolution survey to see if this had a
significant impact on the DBA results.

TOPOGRAPHIC SURVEY

The level of topographic survey required depends on several factors such us; the site terrain, the
phase of the project, the project budget, the project size and potential sensitive receptors. Generally,
all stages will require a topographic survey, the overarching problem is what quality that survey is
required to be. The denser the survey points, the more accurately they will represent the site
conditions and pick up critical changes in the natural terrain. This study has assigned a point density
resolution to each quality of survey (Table 1). The aim of this study is to assess if a low-resolution,
free sourced survey would be acceptable to use for a DBA and if the outcomes of the breach are
significantly altered as a result of the topographic survey quality.

Table 1 Topographic quality (general classification)

Type Resolution (spacing between points) Survey Source Examples

Detailed 1-5m Aerial Survey/LIDAR Survey

Poor > 20 m Free sourced public satellite imagery

3
ASSESSMENT

Topography

The high-resolution topographies (the top row of Figure 1) have a point density of at least one point
every density of less than 5 m and the low-resolution topographies (the bottom row of Figure 1) have
a point density of less than one point every 15 m. The topographies are imported with the equivalent
point density into FLO 2D. Once a topography is imported, a grid is assigned over it, in order to allow
for flow computation, as the software runs by calculating flow balance in individual grid squares. In
this study, the grid element size was defined as 5 m for both the high-resolution topographies and
low-resolution topographies, with the aim of assessing if a low-resolution topography can be
‘improved’ with a denser grid square and detailed elevation interpolation. After the grid size is
assigned, FLO 2D interpolates within each grid element for elevation.

Figure 1 High-resolution surveys (top) and low-resolution surveys (bottom)

Dam Characteristics

The dam dimensions that have been adopted for the breach are presented below in Table 2.

4
Table 2 Dam features

Characteristic Unit Value

Maximum Height m 30

Crest width m 8

Dam Volume m3 234,165

Crest Length m 200

Freeboard m 1

Upstream Slope and Downstream Slopes H:V 2:1

Breach

A geometrical approach has been used to estimate the volume of tailings released. The central
portion of the dam has been selected as the critical section for analysis with a maximum height of
30 m, occurring in the natural low point in the valley.
It is assumed that the breach occurs on a plane defined by a third of the internal angle of friction,
which is projected through the dam wall. An angle of 6° (central and lateral) was used for projecting
the breach through the tailings. A failure length of 80 m, approximately two fifths of the dam width
has been considered. A total outflow volume of ~ 118,000 m3 (approximately half of the total dam
volume) has been estimated. Both the Topography A and Topography B assessments have been
modelled using the same dam dimensions and corresponding breach volume.

Hydrograph

A hydrograph was developed in HEC-RAS using the outflow volume estimated geometrically. This
hydrograph is presented in Figure 2. FLO 2D has functionality to input a mud-sediment hydrograph,
which allows the input of sediment concentration within the hydrograph. The sediment
concentration, by weight, has been estimated to commence at 0% at time zero to allow for the primary
breach phase comprising only of water, followed by an increase in sediment concentration,
representing the gradual increase in solids content during the early stages of the breach up to a
maximum sediment concentration, assumed to be 67% by weight. The sediment concentration
accounts for the volume of tailings and dam construction material combined as one homogenous
outflow volume.

5
 Figure 2 Breach hydrograph and corresponding sediment concentration

FLO-2D uses bulking factors to accommodate sediment concentration in a water hydrograph, the
hydrograph in Figure 2 has been reduced by this bulking factor to account for the sediment
concentration effect.

Rheology and Model Input Parameters

The tailings rheology has been estimated using realistic properties for tailings under the hypothetical
breach scenario. The Manning’s n coefficients for the site terrain have been adopted from estimates,
for Manning’s n for floodplains (Chow, 1958). The parameters adopted for model input are presented
in Table 3 below.

Table 3 Flow Cinematics and Model Input Parameters (dimensionless)

Property Value

Manning’s n coefficient for slopes 0.08

Manning’s n coefficient for flat terrain 0.04

Sediment Specific Gravity 2.72

Laminar Flow Resistance 200

Viscosity vs Sediment concentration: Coefficient, Exponent 0.0001, 18.93

Yield Stress vs Sediment Concentration: Coefficient, Exponent 0.0065, 20

6
RESULTS AND DISCUSSION

Narrow Valley - Topography A

The flow patterns observed are seen in both topographies, for example, the decrease in surface water
depth as the flow moves downstream of the breach is observed in both the low-resolution and high-
resolution topography. Similarly, the high velocity zones for both topographies are also consistent.
There are, however, significant differences in the numerical results of these models, with different
velocities, flow depths, runout distances and inundation areas observed between the two
topographies.
The low-resolution topography resulted in a lower maximum velocity in the first portion of the valley
of ~ 5 m/s, and a higher maximum velocity in the last portion of the valley of ~ 2 m/s. The high-
resolution topography resulted in a higher maximum velocity in the first portion of the valley of ~ 6
m/s, and a lower maximum velocity in the last portion of the valley of ~ 1 m/s.
The low-resolution topography resulted in a smaller overall wetted area with a depth of greater than
0.15 m within the computational domain of ~119,225 m2, and it was observed that the flow mass had
finished flowing, mostly within the domain, at the end of the 10-hour simulation. The high-resolution
topography resulted in a larger wetted area within the computational domain of ~196,900 m2,
however it was observed that there was likely more flow beyond the model domain at the end of the
10-hour simulation (as the flow mass had exited the outflow elements).
The maximum flow depth that occurred during the simulation was similar, but slightly higher for
the low-resolution topography at 11.5 m, and lower for high-resolution topography at 9.5 m.
The maximum depth of flow at the end of 10-hour simulation was ~8.0 m for the low-resolution
topography and ~3.8 m for the high-resolution topography. The low-resolution topography had a
narrow distribution of flow and a depth of less than 0.9 m at the model boundary. The maximum
flow depth was more widely distributed for the high-resolution and at the model boundary had a
depth of less than 0.5 m, (Figure 3 top).
The surface water elevation of the flow was comparable in terms of distance for flow to reduce,
however the width of the impact was significantly reduced for low-resolution topography.
It took ~ 1.1 hours for a small depth of flow to reach model boundary in the low-resolution
topography (likely the initial water breach), the main body then took ~ 2.2-hours to reach the model
boundary, in accordance with the maximum depth of flood occurring at the boundary. The high-
resolution topography showed a similar pattern, the time for this smaller “initial water breach” to
reach the boundary was ~1.1 hours, with the main flood mass following behind as observed with the
high-resolution topography but taking a total time to reach the model boundary of ~10 hours (Figure
3 bottom).

7
 Figure 3 Results – Final flow depth and time for flood to reach 60 cm for both topography resolutions

With all parameters and inputs constant except for the topographic quality, these differences can be
attributed to the interpolated elevation between the two topographies (Figure 4). The low-resolution
topography has a wider valley in the first portion of the model and a much narrower valley in the
end portion of the valley, resulting in higher flowing velocities and larger runout distances. The high-
resolution topography has more points to capture the sharp change in elevation, resulting in a wider
valley, with lower velocities and flow depths.

Figure 4 Variation in valley width at the end portion of the valley caused by interpolation between points on
the low-resolution topography

Wide Valley - Topography B

The low-resolution resulted in a much longer run-out distance within the computational domain,
~2.5 kms downstream of the simulated breach, which likely extended beyond the model boundary
and a total wetted area of 337,200 m2. The high-resolution topography had a slightly wider flood
area and extended only ~ 1 km downstream of the simulated breach and a wetted area of 263,975 m2

8
(Figure 4 top). While the two runout distances are significantly different for these two topographies,
many of the numerical results are similar in magnitude, as followed explained.
The maximum outflow velocities for both topographies were in the order of 5 m/s. After 10-hours of
simulation, the low-resolution topography had almost exited the computational domain, with a
maximum of only 1.5 m/s of flow remaining at the boundary and the high-resolution topography had
stopped flowing.
The maximum flow depth for both topographies was similar at ~ 7.7 m for the low-resolution survey
and at ~ 8.7 m for the high-resolution survey. The final flow depth was also similar for both
topographies at ~3.6 m for the low-resolution survey and ~2.8 meters for the high-resolution survey.
The time to maximum depth for the low-resolution topography is over 10-hours at the model
boundary the time to maximum depth for the high-resolution topography is ~ 4.8 hours (over a
smaller runout length) (Figure 4 bottom).
With all inputs consistent, except for the topography, the variation of results in between the high-
resolution and low-resolution survey is likely a result in the difference in slope. The high-resolution
has an average slope of ~ 8% whereas the low resolution has a significantly steeper average slope of
~ 11%.

Figure 4 Results – ground surface with a wetted area of 0.2 m (top) and time to maximum depth (bottom)

9
CONCLUSION

For a valley profile, like the one used in this study, the low point resolution can cause a narrowing of
the valley resulting in higher velocities, rapid flow times to key infrastructure and populated areas
and longer runout distances. For the wider valley profile, like the one used in this study, although
the numerical results were similar in magnitude, the impact area and runout distance was vastly
different between the two qualities of topographies. The cost of a quality, site specific survey could
far outweigh the cost of implementing rapid evacuation response times or designing for high
consequence breaches. The results of this study suggest that a low-resolution topography is
potentially over-conservative in terms of design and design criteria but that as a result of this over-
conservative estimation of the DBA. Assessments undertaken using low-resolution topographies are
likely satisfactory in terms of dam safety and evacuation considerations. Further studies in different
site terrains should be conducted to assess the impact of low-resolution topographies on different
terrains.

ACKNOWLEDGEMENTS
Nicolas Ubeda, Golder, Chile, for his contribution to obtaining the topographic surveys and
estimating the dam breach volume. Hossein Gildeh Kheirkah Gildeh, Golder, Canada, for his
excellent training in using FLO-2D for DBAs.

REFERENCES
Hungr (1995) ‘A model for the runout analysis of rapid flow slides, debris flows, and avalanches’,
Canadian Geotechnical Journal 32, pages 610 – 623.

Seddon, K.D. (2010). Approaches to estimation of run-out distances for liquefied tailings. Mine Waste
2010, Australian Centre Geomechanics, Perth, Australia.

Olson, S. & Stark. T. (2002). Liquefied strength ratio from liquefaction flow failure case histories.
Canadian Geotechnical Journal 39:629-647 (2002).

Chow, V.T. (1959). Open-channel hydraulics, Edition 2, McGraw-Hill, New York.

FLO-2D (2018) FLO-2D Reference Manual, Build No. 18 2018

10
Comparison of Liquefied Shear Strength of Tailings
Estimated using Standard Empirical Methods and
Post-cyclic Direct Simple Shear Tests
Ganan Nadarajah1, David Bleiker1 and Siva Sivathayalan2
1. Wood, Canada
2. Carleton University, Canada

ABSTRACT
There are a large number of upstream raised tailings dams across Canada and worldwide. While
many of these are historical and closed, there are still several operating tailings dams which are being
raised by upstream construction method mainly due to its lower cost and as continuation of previous
dam raising. Static and seismic liquefaction of foundation tailings is a serious concern in these dams.
The tailings in these facilities are generally deposited by means of spigotting, and thus are generally
in a loose to very loose state, and highly susceptible to liquefaction. The post-liquefaction stability
and remediation design of these dams almost entirely depend on the estimated liquefied shear
strength (Sr) of the tailings.
The Sr of tailings is generally estimated using empirical methods utilizing in-situ penetration test data
such as SPT and CPT data. These correlations have been developed based on back analyses of case
histories involving mostly natural soils. It is routine practice to use these same methods to estimate
Sr of tailings as well. While some practitioners use clean sand equivalent penetration resistance data
with these charts to correct for the fines content, others caution that using fines content correction
could provide non-conservative Sr. This paper presents data and results from two different tailings
dam sites and compares Sr estimated using empirical methods with clean sand equivalent penetration
resistance, (ii) empirical methods without correction for fines content, and (iii) post-cyclic direct
simple shear (DSS) testing of Shelby tube piston samples and reconstituted samples. A relatively new
procedure, involving unload-reload loops, is introduced to study the post-cyclic shear response at
large strain and to assess Sr values in laboratory DSS tests. It has been demonstrated that unlike the
routine post-cyclic DSS test data, which invariably provides much higher Sr values compared to in-
situ test based empirical correlations, the Sr values determined in post-cyclic DSS tests using this
procedure provide Sr comparable with the empirical methods. It is recommended to exercise caution
on relying on Sr values from routine post-cyclic DSS tests results and to explore the post-cyclic
response and Sr using the proposed post-cyclic DSS test procedure. More research and data is needed
before adopting the proposed procedure as a method for design.

1
INTRODUCTION

There are a large number of tailings storage facilities across Canada and the world. These facilities
can generally be grouped into (i) slurry tailings disposal facilities contained by embankment dams
and natural topography, (ii) thickened tailings disposal facilities contained by modest-sized to no
embankment dams, and (iii) dry stack facilities which may not require any embankment containment
structures. The tailings management technology and the type of storage facility are selected by
considering numerous factors, including ore process requirements, water balance, site conditions,
availability of construction materials, capital and operating costs and closure considerations. Due to
relatively lower initial capital cost, many historical and operating mines produce slurry tailings and
store and manage the slurry tailings within embankment dams.
These embankment tailings dams are generally constructed in stages starting with construction of
starter dams. The subsequent dam raises typically follow upstream raise, centerline raise or
downstream raise construction. The upstream dam raise method has been the most commonly used
method due to its relatively lower cost. While many of these upstream tailings dams are historical
and closed, there are still several operating tailings dams mainly due to the lower construction cost
and as a continuation of previous site concept and practices.
The upstream raise method involves construction of staged dam raises over the previously deposited
tailings beach. As the tailings in these facilities are deposited by means of spigotting or end discharge,
the tailings are generally at very loose to loose state and thus highly susceptible to static and seismic
liquefaction. The post-liquefaction stability of these upstream raised dams are almost entirely
dependent on the liquefied shear strength (Sr) of the tailings. It is routine practice to estimate Sr of
tailings using the standard empirical methods developed based on back analyses of case histories
mostly involving natural soils. Some of the widely used empirical methods are Seed (1987), Davis et
al. (1988), Seed and Harder (1990), Olson and Stark (2002) and Idriss & Boulanger (2008) and
Robertson (2010). These methods generally relate the Sr or liquefied shear strength ratio (Sr/vo’) to
CPT and SPT data. It is noted that some of these methods relate the Sr to normalized clean sand
equivalent CPT and SPT data (qc1N-CS and (N1)60-CS) while other methods relate the Sr to normalized
CPT and SPT data (qc1N and (N1)60). The limitations and applicability of these methods for the hard
rock mine tailings are not well understood. Several researchers, including Wijewickreme et al. (2005),
Al Tarhouni (2008) and James et al (2011) have investigated the Sr of tailings mainly through post-
cyclic direct simple shear tests (DSS).
Wijewickreme et al. (2005) conducted a series of cyclic with post-cyclic DSS tests on undisturbed and
reconstituted samples of copper-gold and copper-gold–zinc tailings and measured Sr/vo’ values of
0.42 - 0.64 and 0.13 - 0.53, respectively. These tailings were essentially non-plastic with very high fines
content (> 83%). The measured Sr/vo’ values are generally higher than the range of Sr/vo’ expected
from standard empirical methods.
James et al (2011) conducted series of cyclic DSS tests on reconstituted non-plastic gold tailings and
performed post-cyclic DSS tests on five of those cyclic DSS samples. These tailings sample had fines

2
content of greater than about 70% and were prepared by pluviation of slurry through water. The
measured Sr/vo’ values were between 0.08 and 0.12, which is notably lower than the typical range of
Sr/vo’ reported by Wijewickreme et al. (2005). Al Tarhouni (2008) conducted a series of cyclic and
post-cyclic DSS tests on reconstituted non-plastic silt gold tailings samples. The test samples were
prepared by depositing a thick slurry, and reported Sr/vo’ values ranged between 0.1 and 0.35,
similar to the range of values reported by James et al (2011). The reasons for the relatively high Sr/vo’
values in the laboratory have been attributed to various factors, and there is a risk if the strength
ratios determined from laboratory tailings specimens using the current practice are extended to the
tailings in-situ for design purposes.
This paper presents information and data from two different hard rock mine tailings management
facilities involving spigotted slurry tailings and end discharged thickened tailings. For each tailings
facility, cyclic and post-cyclic DSS tests were performed on undisturbed intact samples and
laboratory reconstituted samples to assess Sr/vo’ values of the tailings. The Sr/vo’ values were also
estimated from CPT and SPT data using the standard empirical methods with and without
considering the correction for fines content. The Sr/vo’ values from post-cyclic DSS tests were
compared against values obtained from the empirical methods and discussed.

CHARATERISTICS OF TAILINGS AND STORAGE FACILITIES

Tailings Characteristics

This study considers information and data from two different hard rock mine tailings facilities
(referred here as Site #1 and #2). The tailings at these two sites are non-plastic to low plastic fine
tailings, and typically contain greater than 80% fines. Table 1 summarizes the characteristics of the
tailings at these two sites. Screening level assessment of liquefaction potential of these tailings
indicates that these tailings are “susceptible” or as requiring advanced laboratory testing to assess
liquefaction potential.

Table 1 Summary of Index Properties of Tailings

Site Mined Metal Fines Content, % Liquid Limit, % Plasticity Index, % Moisture Content, %

Site #1 Gold 95 - 100 (ave.98) - Non-plastic 19 – 44 (ave.32)

Site #2 Copper-zinc 80 - 100 (ave.90) 20 - 31 (ave.27) 0 - 3 (ave.1) 23 - 38 (ave.30)

Characteristics of Storage Facilities

Site #1 is a gold mine and produces conventional slurry tailings which are spigotted within the
Tailings Storage Facility confined by several Perimeter Dams. These Perimeter Dams were
constructed in stages with the Starter Dams founded on native competent foundation. However, the

3
subsequent dam raises involved upstream raise construction method over the beached tailings. As
such, liquefaction and liquefied strength of the tailings is of concern for the dam safety.
Site #2 is a copper-zinc mine and produces thickened tailings which are end discharged within a
Tailings Management Facility confined by a Ring Dam. The Starter Dam was constructed on
previously deposited tailings and the subsequent dam raises were constructed as upstream raise
construction over tailings. Considering that the entire dam is on tailings, liquefaction and liquefied
strength of the tailings is a dam safety concern.

SITE INVESTIGATIONS AND SIMPLE SHEAR TESTING

Site Investigation Programs

Detailed site investigation and laboratory testing programs were conducted for both of these tailings
storage facilities to assess liquefaction potential and liquefied strength of the tailings deposits. The
site investigation programs included in-situ testing such as borehole drilling with Standard
Penetration Testing (SPT), Shelby Tube Piston sampling for laboratory direct simple shear (DSS)
testing, Nilcon Vane testing and Cone Penetration Testing (CPT). The laboratory testing included
index tests and Cyclic and Post-cyclic DSS Testing. The DSS tests were conducted on “undisturbed”
Shelby samples and on slurry reconstituted samples.

Direct Simple Shear Test Equipment

DSS tests on the tailings samples were performed using an NGI type simple shear device (Bjerrum &
Landva, 1966) at the geotechnical research laboratory at Carleton University. This device uses a steel-
wire reinforced rubber membrane to enclose the soil sample and as a result essentially K0 conditions
prevail in the sample during consolidation. Shear loading is applied while maintaining the sample
height constant, and thus constant volume conditions prevail during shear. Vertical, horizontal loads
and displacements are monitored using an automated data acquisition system, and these readings
enable calculation of the resulting stresses and strains. The drop in vertical load during constant
volume loading (Finn 1985) is considered equivalent to the excess pore pressure generated (Dyvik et
al. 1987) in typical undrained tests, and the excess pore pressure data is derived from the changes in
the vertical stress.

Test Procedure and Interpretation

Preparation of Intact samples: Each soil sample was extruded from the Shelby tube using a motorized
hydraulic extruder directly into a lubricated stainless steel specimen ring with a tapered cutting edge.
Sufficient soil was discarded from the edges to minimize the risk of testing disturbed soils. The soil
specimen within the stainless steel ring was trimmed off the top and bottom using a wire-saw and

4
then transferred into a membrane lined cavity, and sealed with O-rings and then transferred to the
CDSS device. A seating load of about 10-15 kPa was applied before taking an initial height reading.
Preparation of Reconstituted samples: Soil extruded from the Shelby tube was allowed to air dry and
an appropriate amount of water was then added to the air-dried soil and thoroughly mixed
producing a uniform tailings slurry sample. The tailings slurry was then placed in the membrane
lined cavity using a spoon and allowed to settle under self-weight. The top loading cap was placed
after about twenty four hours and the sample was sealed using the top O-ring and transferred to the
testing device. Seating load of 10-15 kPa was then incrementally applied keeping track of sample
height.
Consolidation and Cyclic Shearing: Following the initial seating load, consolidation stress is applied
in increments to yield a load increment ratio of about two and allowing limited time for consolidation
between the intermediate load increments. Following the application of final consolidation stress, the
sample is allowed to consolidate overnight. Constant volume conditions are imposed at the end of
consolidation by arresting the vertical displacement. A sinusoidal cyclic shear stress is applied at a
frequency of 0.1 Hz, and care is taken to maintain the cyclic shear stress amplitude at essentially
constant value throughout the loading. The cyclic loading is applied until the sample exceeded 3.75%
single amplitude shear strain.
Post-cyclic Shearing: Strain controlled post-cyclic loading was applied at the end of cyclic loading
without reversal of the last cyclic loading pulse at a loading rate of 60% per hour until the shear
strains reached about 25%. Multiple load-unload-reload cycles were applied in post-cyclic shearing
phase to assess residual shear strength, Sr, of the soil following liquefaction.
Liquefied Shear Strength: The post-cyclic shear strength increases from near zero at zero shear strain
to generally constant shear stress at large strain. This constant peak shear strength has been generally
taken as the liquefied shear strength (Wijewickreme, 2005; Al-Tarhouni et al 2011). However, it
should be recognized that the peak strength obtained in typical post-liquefaction tests may not
represent the residual shear strength. It is believed that multiple load-unload-reload cycles in post-
liquefaction loading might yield a better estimate of the residual shear strength, Sr. The peak “shear
strength” was observed to decrease with multiple load cycles, and was gradually approaching a
residual value. As noted in the discussions below, the lowest value of the constant shear stress
measured during post liquefaction loading cycles is taken as the liquefied shear strength.

Sr/vo’ FROM POST-CYCLIC DSS TEST DATA AND CPT/SPT DATA

SITE #1

A detailed site investigation program involving several boreholes with SPT testing and sampling,
Shelby Tube Piston sampling and CPT test holes was performed. SPT data selected from three
boreholes, CPT data from two test holes and corresponding DSS test results are presented in this

5
paper. The SPT data at these three boreholes generally indicated very loose tailings with normalized
corrected SPT blow count (N1)60 values ranging between 0 and 6. The CPT data also indicated very
loose conditions with normalized cone penetration resistance qc1N ranging between 5 and 40 with
very high excess pore pressures.
The Sr/vo’ values were estimated using the SPT and CPT data. The method proposed by Idriss and
Boulanger (2008) was used to estimate the Sr/vo’ considering correction for fines content. Due to very
high fines content (>95%), the correction for fine content for the SPT data, (N1)60, is 5.5 and for the
CPT data, qc1N, is about 55. This makes the clean sand equivalent SPT (or CPT) resistance notably
higher than the uncorrected penetration resistance. The method proposed by Olson and Stark (2002)
was used to estimate the Sr/vo’ without considering the fines correction. The calculated Sr/vo’ values
from these methods ranges between 0.06 and 0.13, considering the lower bound penetration
resistances.
A total of eight cyclic and post-cyclic DSS tests were conducted on representative intact Shelby tube
samples and reconstituted samples. The consolidated densities of the reconstituted samples were
consistent (14.9 - 15.3 kN/m3) while intact samples’ densities ranged between 12.4 and 15.8 kN/m3.
The intact samples were subjected to only one post-cyclic unload-reload loop while the reconstituted
samples were subjected to four load loops. Typical post-cyclic test results of one of reconstituted
samples is shown on Figure 1. The variation of residual shear strength ratio (Sr/vo’) and excess pore
pressure ratio (U/vo’) with number of post-cyclic load loops are shown on Figure 2. As can be seen
in the plots, Sr/vo’ initially decreases quickly and then stabilizes after about 2 unload-reload loops.
The U/vo’ also initially increases and then stabilizes after about 2 loops which is consistent with the
Sr/vo’ response. The inferred Sr values from the eight tests are shown on Figure 3 together with
residual excess pore pressure ratios at corresponding Sr/vo’ value.
0.4 1.0
Excess Pore Pressure Ratio, U/vo'

0.8
Shear Stress Ratio, /vo'

0.2

0.6
0.0
0.4

1st Cycle 1st Cycle

-0.2
2nd Cycle 0.2 2nd Cycle
3rd Cycle 3rd Cycle
4th Cycle 4th Cycle
-0.4 0.0
-30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30
Shear Strain (%) Shear Strain (%)

Figure 1 Post-Cyclic Shear Response of Site #1 Tailings

6
 1.0 1.0

Excess Pore pressure Ratio, U/vo

Residual Strength Ratio, Sr/vo'

0.9 Intact Sample 0.9

0.8 Slurry Reconstituted Sample 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2 Intact Sample
0.1 0.1 Slurry Reconstituted Sample
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
No. of Unload-Reload Loops No. of Unload-Reload Loops

Figure 2 Variation of Sr/vo’ and U/vo’ with number of Unload-Reload Cycles for Site #1 Tailings

The dependence of post-cyclic response on the density and the residual effective stress following
cyclic loading is widely recognized, but its dependence on the stress/strain history and stress path is
often overlooked (Sivathayalan & Yazdi, 2014). The post-cyclic Sr measured during unload/reload
loops in these tests clearly demonstrate that consideration of the stress/strain history effects is
important. The mobilized Sr during the first load loop was generally high and decreased with each
subsequent loops with increasing excess pore pressure ratio as shown in Figure 2. Four unload/reload
cycles have taken the material through about 400% cumulative 1 shear strain and thus these low Sr
values probably would correspond to the residual shear strength of the material. Even in tests where
the cyclic loading did not cause high excess pore pressures, the subsequent post-cyclic unload/reload
loops triggered high excess pore pressure generation and thus lower Sr. As Vaid & Sivathayalan
(1997) noted, the response following the triggering of liquefaction due to monotonic
unloading/reloading can be used to assess the post-liquefaction behaviour. The post-cyclic tests with
more than one unload/reload loop induced high excess pore pressures and provided lower Sr/vo’
values which are in general agreement with the values from empirical methods.

1 The shear strains during the unload-reload loops are taken as cumulative.

7
 0.6
DSS Test - Intact Sample (1 Load loop)

d= 15.8 kN/m3

Uvo' = 0.00
0.5 DSS Test - Slurry Reconstituted (4 Load loops)
Residual Strength Ratio, Sr  vo

d= 14.7 kN/m3

Uvo' = 0.25

d= 13.1 kN/m3

0.4

Uvo' = 0.75
0.3

d= 12.4 kN/m3

Uvo' = 0.80

d= 15.1 kN/m3

d= 15.3 kN/m3

d= 15.3 kN/m3

d= 14.9 kN/m3

Uvo' = 0.99
Uvo' = 0.99

Uvo' = 0.97
Uvo' = 0.98
0.2

0.1
Range of Sr from Empirical Methods

0.0
0 1 2 3 4 5 6 7 8 9
DSS Test Number

Figure 3 Sr/vo’ from Post-cyclic DSS tests Empirical Methods for Site #1 Tailings

SITE #2
An extensive site investigation program which included CPT test holes, boreholes with SPT testing
and Shelby Tube Piston sampling was completed at this site. However, data and results of only three
CPT test holes and three boreholes with SPT data and Shelby Tube Piston sampling are discussed
here. The SPT data generally indicated very loose tailings with (N 1)60 values ranging between 0 and
4. The CPT data also indicated very loose tailings with qc1N ranging between 5 and over 30 with very
high excess pore pressures.
As discussed for the Site#1, the Sr/vo’ values were estimated using the SPT and CPT data using the
Idriss and Boulanger (2008) and Olson and Stark (2002) methods. Similar to Site#1, due to very high
fines content (> 80%), the correction for fines content for the SPT data, (N1)60, is 5.5 and for the CPT
data, qc1N, is about 55, and once again much higher clean sand equivalent SPT/CPT resistance
compared to the uncorrected penetration resistance. The calculated Sr/vo’ values from these
empirical methods with and without the fines correction ranges between 0.06 and 0.13, considering
the lower bound penetration resistances.
A total of eight cyclic and post-cyclic DSS tests were conducted on representative intact Shelby tube
samples and reconstituted samples. It was noted that the consolidated densities of the reconstituted
samples, while quite consistent (17.5 – 17.7 kN/m3), were higher than density range expected for
loosest deposited specimens (based on the density of intact samples varied from 16.0 to 16.9 kN/m3,
except in the case of one dense specimen that had a density of 19.1 kN/m3). All samples were subjected
to multiple unload-reload loops in post-liquefaction loading. Typical post-cyclic test results of one of
the reconstituted samples is shown on Figure 4. The variation of residual shear strength ratio (Sr/vo’)
and excess pore pressure ratio (U/vo’) with number of post-cyclic load loops are shown on Figure 5.

8
The inferred Sr/vo’ values from the eight tests are shown on Figure 6 together with residual excess
pore pressure ratios at corresponding Sr/vo’ value.
As observed with the Site#1 samples, the mobilized Sr during the first unload/reload loop was
generally high and decreased with each subsequent loops and generally constant after 2 loops. These
post-cyclic tests with more than one unload/reload loop induced high excess pore pressures and
provided Sr/vo’ values which are in general agreement with the values from empirical methods.
0.6 1.0

Excess Pore Pressure Ratio, U/vo'

0.4
0.8
Shear Stres Ratio, /vo'

0.2
0.6
0.0
0.4
-0.2
1st Cycle 1st Cycle
2nd Cycle 0.2 2nd Cycle
-0.4
3rd Cycle 3rd Cycle
4th Cycle 4th Cycle
-0.6 0.0
-30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30
Shear Strain (%) Shear Strain (%)

Figure 4 Post-Cyclic Shear Response of Site #2 Tailings

1.0 1.0
Excess Pore pressure Ratio, U/vo
Residual Strength Ratio, Sr/vo'

0.9 Intact Sample 0.9

0.8 0.8
Slurry Reconstituted Sample
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2 Intact Sample
0.1 0.1 Slurry Reconstituted Sample
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
No. of Unload-Reload Loops No. of Unload-Reload Loops

Figure 5 Variation of Sr/vo’ and U/vo’ with number of Unload-Reload Cycles for Site #2 Tailings

9
 0.6
DSS Test - Intact Sample (3 Load loops)

d= 19.1 kN/m3

Uvo' = 0.25
0.5 DSS Test - Slurry Reconstituted (3 Load loops)

d= 16.0 kN/m3

Uvo' = 0.92

d= 16.3 kN/m3

0.4

Uvo' = 0.82

d= 16.9 kN/m3

Uvo' = 0.72
Survo

d= 17.5 kN/m3

Uvo' = 0.90
0.3

d= 17.7 kN/m3

Uvo' = 0.90

d= 17.7 kN/m3

d= 17.6 kN/m3

Uvo' = 0.94
Uvo' = 0.99
0.2

0.1 Range of Sr from Empirical Methods

0.0
0 1 2 3 4 5 6 7 8 9
DSS Test Number
Figure 6 Comparison of Sr/vo’ from Post-cyclic DSS Tests and Empirical Methods for Site #2 Tailings

RESULTS AND DISCUSSION

There are several versions/revisions of empirical charts/equations to estimate Sr from SPT and CPT
data. All of these empirical correlations were developed based on back analyses of case histories
mostly involving natural soils. Some of the widely used methods are Olson and Stark (2002),
Robertson (2010) and Idriss & Boulanger (2008). These methods relate the Sr or Sr/vo’ to (N1)60 and
qc1N or (N1)60CS and qc1NCS. It is routine practice to use these empirical methods and procedures to
estimate the Sr of hard rock tailings as there are no widely used methods specific to tailings.
Given the significant influence of potential liquefaction of tailings in upstream raised tailings
facilities, more practitioners are turning to cyclic and post-cyclic DSS testing to obtain site specific
data rather than relying on the empirical methods alone. While the DSS tests provide valuable site
specific data, care must be taken to consider possible sample disturbances. Where possible, it would
be good practice to incorporate reconstituted samples into any DSS laboratory test program with
considerations to appropriate sample preparation methods and sample densities.
There is limited literature on the post-cyclic DSS tests data and inferred Sr of tailings. As discussed
earlier, some of these literature data indicate notably higher Sr/vo’ than the expected ranges from the
empirical methods. These higher Sr/vo’ values may be attributed to the routine practice of performing
the post-cyclic DSS tests without reversal of the last cyclic loading pulse. Some practitioners are now
conducting the post-cyclic DSS test phase with multiple load-unload-reload cycles to ensure high
excess pore pressures, large cumulative strains and liquefaction conditions.
The results presented in this paper clearly indicate consideration of the stress/strain history effects,
and deformation over a large range of shear strain is important in assessing the Sr. The mobilized
Sr/vo’ during the first load loop was found to be generally high and decreased with each subsequent

10
loops and high excess pore pressures. The lower Sr/vo’ values determined in DSS after multiple loops
are in general agreement with the values from empirical methods.
Based on the data and discussions presented, it is suggested that attempt to determine liquefied shear
strength (Sr) of tailings from laboratory post-cyclic DSS tests should involve multiple post-
liquefactions load cycles. Three to four unload/reload cycles were found to be sufficient to provide
an essentially constant Sr value for the two tailings tested. Further, it is cautioned that the liquefied
shear strength values obtained without the post-cyclic load reversals may not be reliable and could
be unconservative.

CONCLUDING REMARKS
In-situ and laboratory testing were undertaken to assess the liquefied shear strength of tailings at two
hard rock mine tailings storage facilities. Tailings at Sites #1 was deposited by spigotting of slurry,
and at Site #2 by end discharge of thickened tailings. The liquefied shear strengths of these tailings
are estimated from SPT and CPT data gathered from these two facilities using a widely used standard
empirical methods. The tailings material consisted significant amount of fines (80 – 95%), and thus
whether or not a fines correction was applied to the empirical methods made a significant difference
in the assessment. Since the in-situ test based correlations have generally been derived from back
analyses of natural soils, caution is warranted in their direct applicability to tailings.
A series of cyclic and post-cyclic DSS tests were performed on representative intact and reconstituted
tailings samples from both tailings storage facilities to obtain liquefied shear strength of tailings.
Comparison of liquefied shear strength values estimated from SPT and CPT data using the standard
empirical methods against values inferred from post-cyclic DSS test data indicates that post-cyclic
DSS tests with multiple load loops provides liquefied shear strength values comparable with the
empirical methods. Conversely, use of liquefied shear strength values inferred from DSS tests
without the post-cyclic load reversal, which has been the typical practice in the literature, provides
notably high liquefied shear strength values and could be unconservative. The proposed extension
of routine post-cyclic tests with multiple unload-reload loops could be used as a check. More research
and more data is needed before adopting the proposed post-cyclic DSS test with multiple loading
loops as a method for obtaining liquefied shear strength for design.

REFERENCES

Al-Tarhouni, M. 2008. Liquefaction and Post-liquefaction Behaviour of Gold Mine Tailings under
Simple Shear Loading, MAsc Thesis, Carleton University, Ottawa, ON., 172p.

Bjerrum, L. & Landva, A. 1966. Direct simple shear tests on a Norwegian quick clay. Géotechnique,
16(1):1–20.

11
Davis, A.P., Poulos, S.J., and Castro, G. 1988. Strengths back figured from liquefaction case histories.
Proc., 2nd Int. Conf. on Case Histories in Geotechnical Engineering, University of Missouri-Rolla,
Rolla, Mo., 1693–1701.

Dyvik, R., Berre, T., Lacasse, S. & Raadim, B. 1987. Comparison of truly undrained and constant
volume direct simple shear tests. Géotechnique, 37(1):3–10.

Finn, W.D.L. 1985. Aspects of constant volume cyclic simple shear. In Proceedings of Advances in
the Art of Testing of Soils Under Cyclic Conditions, Detroit, Mich., 24 October 1985. American Society
of Civil Engineers, New York. pp.74–98.

Idriss, I. M. & Boulanger, R. W. 2008. Soil liquefaction during earthquakes. Monograph MNO-12,
Earth-quake Engineering Research Institute, Oakland, CA, 261p.

James, M., Aubertin, M., Wijewickreme, D. & Wilson, W. 2011. A laboratory investigation of the
dynamic properties of tailings. Can. Geotech. J. 48(11), 1587–1600.

Olson, S.M., and Stark, T.D. 2002. Liquefied strength ratio from liquefaction flow failure case
histories. Can. Geotech. J., 39(3), 629–647.

Robertson, P.K. 2010. Evaluation of flow liquefaction and liquefied strength using the cone
penetration test. Journal of Geotechnical and Geoenvironmental Engineering, 136(6):842-853.

Seed, H.B., and Harder, L.F. 1990. SPT-based analysis of cyclic pore pressure generation and
undrained residual strength. Proc., Seed Memorial Symp., BiTech Publishers, Vancouver, BC,
Canada, 351–376.

Sivathayalan, S. and Yazdi, A. 2014. Influence of Strain History on Post Liquefaction Deformation
Characteristics of Sands. Journal of Geotechnical and Geoenvironmental Engineering. 140(3): 1-14.

Vaid, Y. P. and Sivathayalan, S. 1997. Post liquefaction behaviour of saturated sand under simple
shear loading, In Proc. of the 14th International Conference on Soil Mechanics and Foundation
Engineering, pp. 221-224, Hamburg, Germany.

Wijewickreme, D., Sanin, M.V. & Greenaway, G. 2005. Cyclic shear response of fine-grained mine
tailings. Canadian Geotechnical Journal, 42(5): 1408–1421. doi:10.1139/t05-058.

12
Characterization of Old Tailings under Unsaturated
Conditions
José Ale1, Cesar Luna1, José Ccotohuanca1, and Carlos Jiménez2
1. Wood, Peru
2. MINSUR, Peru

ABSTRACT
As a result of new metallurgical processing techniques, an old tailings storage facility located in the
southern Andes of Peru has acquired economic importance. For this reason, it is proposed to
redevelop the mine with a new open pit that will cut through the tailings storage facility. The tailings
will also be reprocessed.
The tailings storage facility is located on the glacial valley, with a tailings thickness of up to 50 m. The
tailings dam is composed of coarse tailings (silty sands), and has been built following the raising
method of upstream embankment. Upstream of the dam, the stored material is mostly composed of
fine tailings (low plasticity silt and clay). The initial groundwater level is at 10 meters depth, which
means that most of tailings are saturated.
The proposed mining method is the excavation by conventional equipment (excavators) and
transport by trucks. Therefore, the tailings material to be excavated must have a water content that
allows the execution of the work. It has proposed a dewatering plan within the tailings storage
facility, to thereby ensure the workability of the tailings material. The tailings dewatering modelling
consider the unsaturated behaviour of the hydraulic properties.
This paper presents the geotechnical characterization for dewatering purposes at different water
content (which ranges from 3% to 25% after dewatering). To perform this characterization, a field
geotechnical investigation and a laboratory program were conducted. The field investigation was
composed of: drillings, SPT, CPT, pumping wells, and Shelby sampling. Laboratory tests program
consisted of: soil-water characteristic curves, grain size by sedimentation, specific gravity, Atterberg
limits, hydraulic conductivity, water content, and triaxial compression tests. The unsaturated
behaviour (soil-water characteristic curve) was obtained from two ways: laboratory tests, and Van
Genuchten equation.

1
INTRODUCTION

Nowadays, the new metallurgical processing technologies have improved the efficiency and the
mineral recovery in the mining activities, making old tailings storage facilities (TSF) acquire economic
importance. The real case used in the development of this paper corresponds to a TSF of tin mine
located in the southern Andes of Peru. The TSF was placed in a glacial valley (as shown in Figure 1)
which is composed mainly of moraine deposits, at an average elevation of 4500 meters above sea
level.

Figure 1 Old Tailings Storage Facility (highlighted) located in a glacial valley. Puno, Peru

Tailings deposition in this old TSF was performed by cyclones. The tailings dam is composed of
coarse tailings (silty sands), and has been constructed following the upstream method. In addition,
upstream of the dam, the stored material is composed mostly of fine tailings (low plasticity silt), as
shown in Figure 2. The TSF thickness is up to 50 meters, and groundwater level is 10 meters deep,
which means that most tailings are saturated.

2
 Fine Tailings
Coarse Tailings

Figure 2 Cross section of the Tailings Storage Facility analysed. Coarse tailings in yellow, and fine tailings in
gray (the location of the section is shown in Figure 3)

Since the old TSF has important tin content, a mining plan was designed to extract and reprocess the
existing tailings. The proposed mining method is to excavate the tailings (by conventional equipment,
excavators) and transport them by trucks. Therefore, the tailings material to be excavated must have
a water content that allows the execution of the work. It has proposed a dewatering plan within the
tailings storage facility, to thereby ensure the workability of the tailings material.
This paper presents the geotechnical characterization of tailings material at different water contents
(ranging from 3% to 25% after dewatering). To carry out this characterization, a field geotechnical
investigation and a laboratory program were conducted.

FIELD GEOTECHNICAL INVESTIGATION AND LABORATORY TESTING

PROGRAM
Log information from more than 100 geotechnical and metallurgical boreholes (within TSF, as shown
in Figure 3) was reviewed. In the geotechnical boreholes, penetration testing (SPT and CPT-u) and
tailings sampling were carried out. In addition, the stratigraphic data from boreholes was used to
know the internal distribution of the tailings types within TSF, and to determine the tailings/natural-
ground interface (see Figure 2).

3
 Figure 2

Figure 3 Geotechnical and metallurgical boreholes in TSF. The line in red, corresponds to the transversal
section of Figure 2

The TSF thickness is up to 50 meters, and groundwater level is 10 meters deep, which means that
most of the tailings are saturated.
In the tailings samples obtained from the boreholes, laboratory testing was performed such as: grain
size distribution (by mesh and sedimentation), limits of consistency, moisture content, specific
gravity, unit weight, triaxial compression strength and soil-water characteristic curve. From the
laboratory results, the geotechnical characterization of the tailings was conducted.

GEOTECHNICAL CHARACTERIZATION OF THE TAILINGS

Physical Characterization

From the grain size distribution test results, it is known that the tailings material within the TSF is
classified as: silty sand (SM), low plasticity silt (ML) and low plasticity silty clay (CL-ML), which
correspond to coarse tailings, intermediate tailings and fine tailings, respectively.
Figure 4 shows the global grain size distribution curves for the tailings material tested, considering
the portion under mesh #200 (silt and clay sizes).

4
 Figure 4 Grain size distribution of tailings material

As Table 1 shows, the average value of specific gravity of the tailings is 2,80 (value results range from
2,66 and 2,89). The value of porosity ranges from 25% and 59%, which corresponds to coarse and fine
tailings, respectively.
According to the Table 1 and Table 2, the average value of the dry unit weight is: 14,21 kN/m3 (from
12,1 to 16,8 kN/m3) for fine tailings; 16,52 kN/m3 (from 13,5 to 18,7 kN/m3) for intermediate tailings;
and 17,43 kN/m3 (from 14,6 to 20,9 kN/m3) for coarse tailings. The minimum value of 12,1 kN/m3
corresponds to a shallow fine tailings zone.
Table 1 lists the samples where in-situ density and porosity information has been available.

Table 1 Physical properties of the tailings

Dry Unit
Boreholes Depth Specific Porosity
Sample USCS Weight
Code (m) Gravity Gs n
(kN/m3)
BHSR-KP10-08 21,75-22,7 ST-05 CL 2,82 14,65 0,47
39,40-39,85 ST-02 CL 2,86 15,09 0,46
BHSR-KP10-09
43,80-44,50 ST-04 CH 2,86 14,52 0,48
4,50-4,95 SPT-03 SP 2,82 16,94 0,39
BHSR-KP11-04 6,00-6,45 SPT-04 SP 2,82 16,94 0,39
12,00-12,45 SPT-08 CL 2,81 12,10 0,56

5
 Dry Unit
Boreholes Depth Specific Porosity
Sample USCS Weight
Code (m) Gravity Gs n
(kN/m3)
13,50-13,95 SPT-09 CL 2,81 12,10 0,56
23,50-23,95 SPT-13 CL 2,76 13,57 0,50
28,00-28,45 SPT-14 CL 2,76 13,57 0,50
16,50-16,95 SPT-10 MH 2,66 18,12 0,59
BHSR-KP11-07
18,00-18,45 SPT-11 MH 2,66 18,12 0,59
BHSR-KP15-05 18,20-18,90 ST-03 CH 2,78 13,58 0,50
13,15-15,40 ST-01 CL-ML 2,89 13,42 0,53
BHSR-KP15-06 13,85-14,65 ST-02 CL-ML 2,83 15,10 0,46
31,95-32,65 ST-07 CL 2,83 16,14 0,42
DH-15-01 15,45-16,15 Shelby-01 SM 2,76 18,57 0,31
DH-15-02 7,5-8,20 Shelby-01 SP-SM 2,79 17,04 0,38
DH-15-04 7,5-8,20 Shelby-01 SM 2,76 16,52 0,39
22,50-23,20 Shelby-02 SM 2,76 17,04 0,37
DH-15-05 36,00-36,70 Shelby-03(1) CL 2,80 16,78 0,39
36,00-36,70 Shelby-03(2) SM 2,79 14,55 0,47
DH-15-06 10,00-10,50 Shelby-01 ML 2,79 15,99 0,42
DH-15-07 10,00-11,30 Shelby-01 SM 2,78 18,44 0,32
6,00-6,70 Shelby-01 ML 2,84 16,75 0,40
DH-15-10
15,00-15,70 Shelby-02 ML 2,84 17,13 0,39
9,00-9,70 Shelby-01 ML 2,81 16,07 0,42
DH-15-11
21,00-21,70 Shelby-02 SM 2,83 20,85 0,25
DH-15-12 6,00-6,70 Shelby-01 ML 2,81 13,52 0,51

For the history case, the tailings were classified as: SM (silty sand) for coarse tailings, ML (low
plasticity silt) for intermediate tailings, and CL (low plasticity clay) for fine tailings.

Shear Strength Characterization

As part of the geotechnical characterization of the tailings materials of the TSF, a program of triaxial
compression strength testing was conducted in saturated conditions. The results are shown in Table
2, in terms of total and effective stresses.

Table 2 Triaxial compression strength results, saturated

Total Effective
Dry Unit
Triaxial Depth Tailings
Borehole USCS Weight c' ф'
Test (m) Material c (kPa) ф (°)
(kN/m3) (kPa) (°)
CU DH-15-07 10.00 SM Coarse 18.44 252.20 34.80 4.40 37.00

CU DH-15-05 22.50 SM Coarse 17.16 63.90 37.60 27.50 36.20

6
 Total Effective
Dry Unit
Triaxial Depth Tailings
Borehole USCS Weight c' ф'
Test (m) Material c (kPa) ф (°)
(kN/m3) (kPa) (°)
CU DH-15-12 6.00 ML Intermediate 13.59 70.00 25.20 47.00 26.60

CU DH-15-10 6.00 ML Intermediate 16.16 157.00 27.40 117.70 28.50

CU DH-15-04 7.50 SM Coarse 16.53 19.90 24.30 10.00 33.00

CU DH-15-02 7.50 SP-SM Coarse 17.05 40.20 32.00 13.60 34.20

CU PR-SR13-02 9.00 ML Intermediate 18.71 183.00 36.00 42.00 39.00

UU DH-15-11 9.00 ML Intermediate 16.07 0.20 35.90 - -

UU DH-15-06 10.00 ML Intermediate 15.99 1.60 35.70 - -

CU DH-15-01 15.45 SM Coarse 18.93 502.60 31.40 186.30 36.50

CU PR-SR13-03 19.00 SM Coarse 16.09 34.00 27.00 2.00 34.00

CU BHSR-KP10-09 21.75 CL Fine 14.65 26.00 16.00 4.00 29.00

CU BHSR-KP10-09 39.40 CL Fine 15.09 77.00 12.00 12.00 27.00

CU BHSR-KP10-09 43.80 CH Fine 14.52 32.00 13.00 23.00 21.00

CU BHSR-KP10-09 43.80 CH Fine 14.52 36.00 14.00 17.00 24.00

CU PR-SR13-04 10.05 SM Coarse 16.99 123.00 28.00 42.00 34.00

CL-
CU BHSR-KP15-06 13.15 Fine 13.42 0.00 15.00 3.00 31.00
ML
CL-
CU BHSR-KP15-06 13.15 Fine 13.42 0.00 23.00 0.00 36.00
ML
CU BHSR-KP11-07 16.50 MH Intermediate 13.50 12.00 12.60 8.00 18.30

CU BHSR-KP15-05 20.10 CH Fine 13.56 23.00 14.00 19.00 22.00

CU BHSR-KP15-05 20.10 CH Fine 13.56 41.00 14.00 21.00 23.00

CU BHSR-KP15-06 20.17 CL Fine 12.88 14.00 14.00 12.00 24.00

CU BHSR-KP15-06 20.17 CL Fine 12.88 13.00 17.00 11.00 26.00

CU BHSR-KP15-06 31.95 CL Fine 16.14 24.00 14.00 14.00 24.00

CU BHSR-KP15-06 31.95 CL Fine 16.14 36.00 16.00 11.00 25.00

CU BHSR-KP11-04 4.50 SP Coarse 16.98 0.00 21.80 0.00 31.00

UU BHSR-KP11-04 4.50 SP Coarse 16.90 0.00 21.00 - -

CU BHSR-KP11-08 11.90 SM Coarse 16.57 6.00 16.40 0.00 29.50

CU BHSR-KP11-04 12.00 CL Fine 12.35 10.00 12.80 8.00 20.10

UU BHSR-KP11-04 12.00 CL Fine 12.10 16.00 5.00 - -

7
 Total Effective
Dry Unit
Triaxial Depth Tailings
Borehole USCS Weight c' ф'
Test (m) Material c (kPa) ф (°)
(kN/m3) (kPa) (°)
CU BHSR-KP11-04 15.00 MH Intermediate 17.57 5.00 9.00 0.00 16.40

UU BHSR-KP11-04 16.50 MH Intermediate 13.60 12.00 2.00 - -

CU BHSR-KP11-04 19.50 CL Fine 13.53 10.00 12.00 12.00 16.60

UU BHSR-KP11-04 19.50 CL Fine 13.57 14.00 2.00 - -

The average values of effective strength parameters are:

 For coarse tailings: cohesion of 31,76 kPa and friction angle of 33,93°.
 For intermediate tailings: cohesion of 42,94 kPa and friction angle of 25,76°.
 For fine tailings: cohesion of 11,93 kPa and friction angle of 24,91°.
And the minimum values of effective strength parameters are:
 For coarse tailings: cohesion null and friction angle of 29,5°.
 For intermediate tailings: cohesion null and friction angle of 16,4°.
 For fine tailings: cohesion null and friction angle of 16,6°.
These tests were conducted in saturated conditions, but helped in the determination of unit weight
and porosity.

Soil-Water Characteristic Curves (SWCC)

For the present study, the soil-water characteristic curves were obtained using the following
methodologies: laboratory tests, analysis of granulometric curves, and review of existing
information.
Figure 5 shows the distribution of the suction of the tailings material according to volumetric
moisture content. This figure has been defined from the results of the laboratory tests.

8
 Figure 5 Suction versus volumetric water content

Now, according to the porosity values showed in the Table 1, the volumetric water content at
saturation must range between 26% and 59%.
Based on the grain size distribution, porosity and laboratory soil-water characteristic curves,
representative soil-water characteristic envelopes have been calculated using the Van Genuchten
(1980) methodology, for coarse, intermediate and fine tailings materials (see Figure 6).
Table 3 shows the parameters used in soil-water characteristic curve models, according to Van
Genuchten's model.

Table 3 Van Genuchten Parameters

θs θr
Model Equation Tailings Material a m n
(%) (%)
𝑚 Fine 48,0 20,0 0,01 0,17 1,21
Van 1
𝜃 = (𝜃𝑠 − 𝜃𝑟 ) ∙ [ 𝑛
]
Genuchten 1 + (𝑎 ∙ 𝜓) Intermediate 35,0 10,0 0,02 0,20 1,30
(1980) + 𝜃𝑟 Coarse 22,0 3,0 0,50 0,44 1,80
Note:
Parameter “a” is expressed in kPa-1.
θs: Saturation volumetric water content.
θr: Residual volumetric water content.

9
θ: Suction ψ (kPa).

Using an average value of specific gravity Gs of 2,80 for the tailings, porosity (n) or volumetric
moisture content (θ) under saturated conditions for each material was calculated. Saturation
volumetric water content values are 43.2%, 31.0% and 26.4%, for fine, intermediate and coarse
tailings, respectively. For the modeling of SWCC envelopes (coarse and fine tailings) the minimum
and maximum porosity values have been used. The average value of porosity was used for the
intermediate tailings.

Figure 6 Van Genuchten envelopes for tailings materials

Normalizing the volumetric water content for each material considering this equation:

𝑚
(𝜃 − 𝜃𝑟 ) 1
=[ ] =𝛳
(𝜃𝑠 − 𝜃𝑟 ) 1 + (𝑎 ∙ 𝜓)𝑛

The volumetric water content response for a suction level can be observed as Figure 7 shows.

10
 Figure 7 Van Genuchten envelopes for normalized volumetric water content

Hydraulic Characterization

As part of the present study, hydraulic conductivity tests (in saturated conditions) were performed
in undisturbed tailings samples. The values of saturated hydraulic conductivity are 1,4E-05 cm/s and
2,0E-04 cm/s, for fine and coarse tailings, respectively.
The Table 4 shows some empirical equations for unsaturated hydraulic conductivity (kunsat), in
function of volumetric water content.

Table 4 Empirical Equations for Unsaturated Hydraulic Conductivity (Fredlund et al., 1994)

Model Equation Comments

𝜭: normalized volumetric water content
Averjanov (1950) 𝑘𝑢𝑛𝑠𝑎𝑡 = 𝛳𝑛
n=3.5
ks: saturated hydraulic conductivity
𝜃 𝑛
Campbell (1973) 𝑘𝑢𝑛𝑠𝑎𝑡 = 𝑘𝑠 ( ) θ: volumetric water content
𝜃𝑠
θs: saturated volumetric water content
Davidson et al. (1969) 𝑘𝑢𝑛𝑠𝑎𝑡 = 𝑘𝑠 ∙ 𝑒𝑥𝑝[𝛼(𝜃 − 𝜃𝑠 )]

For this paper, the unsaturated hydraulic conductivity (k unsat) can be indirectly obtained using
Averjanov (1950) equation, which factorizes the saturated hydraulic conductivity (ks) by the

11
normalized volumetric water content (𝜭) with an exponent of 3.5, for each tailings material. The
Figure 8 shows the unsaturated hydraulic conductivity values for fine tailings and coarse tailings
versus suction.

Figure 8 Normalized soil-water characteristic curve and hydraulic conductivity function for fine tailings and
coarse tailings

As shown in Figure 8 the hydraulic conductivity decreases with suction for both fine and coarse
tailings. The coarse tailings present more sensibility with suction changes.

CONCLUSION
The average value of the dry unit weight is: 14,21 kN/m3 (from 12,1 to 16,8 kN/m3) for fine tailings;
16,52 kN/m3 (from 13,5 to 18,7 kN/m3) for intermediate tailings; and 17,43 kN/m3 (from 14,6 to
20,9 kN/m3) for coarse tailings. The minimum value of 12,1 kN/m3 corresponds to a shallow fine
tailings zone.
The average values of effective strength parameters are: For coarse tailings: cohesion of 31,76 kPa and
friction angle of 33,93°; for intermediate tailings: cohesion of 42,94 kPa and friction angle of 25,76°;

12
and for fine tailings: cohesion of 11,93 kPa and friction angle of 24,91°. These values are in saturated
conditions.
For the unsaturated hydraulic conductivity (as Figure 8 shows), its value could reduce 50% (in respect
to saturated value) at 1 kPa and 200 kPa suction level, for coarse and fine tailings material,
respectively. According to this, the hydrogeological model for a TSF dewatering program must
consider an unsaturated condition for the estimation of dewatering time, pumping flow rates, final
moisture, and so on.
For a good characterization of tailings for dewatering purposes, the obtaining of the porosity must
be done in the best possible way, through the extraction of undisturbed samples. The value of the
porosity influences in the ranges of moisture content, degree of saturation, hydraulic conductivity,
among others.
It is important to mention that, for tailings dewatering projects, the geotechnical investigation
campaign must be programmed in a special way that allows a correct physical characterization of the
materials from the points of view: hydraulic (unsaturated hydraulic conductivity, porosity, specific
gravity, SWCC) and strength (triaxial controlled suction tests).

REFERENCES

Averjanov, S.F. 1950. “About permeability of subsurface soils in case of incomplete saturation”. In
English Collection, Vol. 7. As quoted by P. Ya Palubarinova, 1962. The theory of ground water
movement (English translation by I.M. Roger DeWiest. Princeton University Press, Princeton, NJ),
pp. 19-21.

Campbell, J.D. 1973. “Pore pressures and volume changes in unsaturated soils,” Ph.D. Thesis,
University of Illinois at Urbana-Champaign, Urbana, 1973.

Davidson, J.M., Stone, L.R., Nielsen, D.R., and Larue, M.E. 1969. “Field measurement and use of soil-
water properties”. Water Resources Research, 5: pp. 1312-1321.

Fredlund, D.G., Xing, A., and Huang, S., 1994. “Predicting the permeability function for unsaturated
soils using the soil-water characteristic curve”. Canadian Geotechnical Journal, 1994, 31: pp. 533-546.

Van Genuchten, M.T. 1980. “A Closed-form Equation for Predicting the Hydraulic Conductivity of
Unsaturared Soils”. Soil Science Society of America Journal 1980 Vol.44 No.5, pp. 892-898.

13
On the Dynamic Properties of Tailings
Rolando Rojas1, Miguel Regalado1, Renzo Ayala1 and Denys Parra2
1. Geotechnical Area, Anddes, Peru
2. Anddes and National University of Engineering, Peru

ABSTRACT
Seismic design of tailings storage facilities located at highly active seismic zones such as Peru and
Chile needs to incorporate the proper determination of the dynamic properties of the materials
involves in their seismic stability, such as tailings and foundation materials. There are plenty of
literature dynamic behavior models used to predict the seismic response of tailings, nevertheless, the
majority of these models are theoretical and were proposed for specific soil types, given that a model
developed for a specific soil type could not be applicable for tailings since both of them have different
characteristics starting from its origin and disposal, therefore, applying these literatures models
provides an uncertainty source in the calculation of tailings seismic response. The tailings
experimental data presented is analysed and briefly discussed under different laboratory conditions
such as void ratios, confining stresses, and so on.
This paper presents the results of different resonant column and torsional shear tests performed on
non-plastic coarse-grained and fine-grained tailings executed at Anddes geotechnical laboratory,
including literature review of dynamic laboratory testing on tailings that includes cyclic triaxial tests.
The information has been interpreted focusing on the normalized shear modulus reduction and
damping ratio curves to compare to other existing curves from technical literature proposed by
several authors.
The authors proposed normalized shear modulus reduction and damping ratio curves for tailings
over a range of standard confining stresses. Additional comparisons between the maximum shear
modulus obtained by laboratory testing, geophysical survey and seismic cone penetration tests are
made to provide insights related to representativeness and accuracy of laboratory and field testing.

1
INTRODUCTION

Seismic design of tailings storage facilities (TSF) located at highly active seismic zones need to
incorporate a proper determination of the dynamic properties of the materials involves in their
seismic stability, such as tailings (TL). There are plenty of literature dynamic behavior models
formulated by several authors for different types of soils, such as: Seed and Idriss (1970) for sands;
Seed et al. (1986) and Rollins et al. (1998) for gravels; Sun et al. (1988), Vucetic and Dobry (1991), and
Ishibashi and Zhang (1993) for clays, and current state of the art models, such as Darendeli (2001) for
fine-grained soils and Menq (2003) for sandy and gravelly soils. Nevertheless, the majority of these
models are theoretical and were proposed for specific soil types with different characteristics starting
from its origin and disposal.
Currently, there is limited database available for estimating dynamic properties of TL. Some
important works related to TL present interesting findings; for instance, Savidis (1993) concluded
that uranium TL materials subjected to cyclic triaxial (CTX) tests show a relatively stiffer behavior
than resonant column (RC) tests under cyclic loading. Likewise, Martinez-Rojas et al. (2018) showed
a linear trend between shear modulus and confining stress, regardless the type of TL and shear strain.
According to Winckler et al. (2014) and based on TL dynamic laboratory testing, normalized shear
modulus reduction and damping ratio curves mainly depend on confining stresses. In this study, TL
were evaluated using combined resonant column and torsional shear (RCTS) tests at Anddes
geotechnical laboratory.
TL samples were taken from three different sites, from which five different materials were tested in
laboratory. Of the five samples collected, four came from coarse-grained TL (underflow) and one
from fine-grained TL (overflow). The objective of this paper is to compare different approaches
employed in practice to incorporate the proper determination of the dynamic properties and
proposes normalized shear modulus and damping ratio curves for TL over a range of standard
confining stresses. Those curves may be taken as a preliminary reference to characterize dynamic
properties for seismic analysis of TSF.

METHODOLOGY

Materials and laboratory testing

The TL materials tested in this study were obtained from three different mining projects (labeled
project 1, 2 and 3), and including literature review of dynamic laboratory testing on TL that includes
CTX tests and RCTS testing. The particle size distribution curve (PSDC) of the materials tested
presents similar PSDC shape, unified soil classification system (USCS) and fines content up to a
maximum of almost 50%. The PSDC of all the TL materials used in the laboratory testing are shown
in Figure 1. The TL materials for laboratory testing classify as silty sand (SM); moreover, all these TL
materials are non-plastic. In project 1 and 2, only one sample of overflow and underflow TL (TL-01

2
and TL-02 severally) was tested. In project 3, two samples of underflow TL (TL-03 and TL-04), and
one sample of overflow TL (TL-05) were taken for laboratory testing. The basic index properties of
the specimens tested are shown in Table 1.

Figure 1 Particle size distribution curves of the tailings materials used in the laboratory testing

Table 1 Index properties of TL samples subjected to RCTS tests at Anddes geotechnical laboratory

Sample Material Gs1 D302 D503 D604 Fines WC5 γd6 e7 USCS8
- (mm) (mm) (mm) (%) (%) (kN/m3) - -

Project 1
TL-01 3.49 NF9 0.12 0.15 30.3 5.0 19.13 0.789 SM
overflow
tailings
Project 2
TL-02 3.34 0.08 0.10 0.11 29.8 5.0 17.56 0.866 SM
underflow
tailings
TL-03 Project 3 2.64 NF9 0.08 0.14 48.6 7.5 15.26 0.700 SM
underflow
TL-04 tailings 2.84 0.09 0.18 0.21 26.0 10.0 14.72 0.890 SM

Project 3
TL-05 2.74 NF9 0.09 0.12 43.9 15.0 15.70 0.713 SM
overflow
tailings

Notes: 1) Specific gravity, 2) Average diameter by thirty accumulated percent passing, 3) Average diameter by fifty
accumulated percent passing, 4) Average diameter by sixty accumulated percent passing, 5) Water content, 6) Remolded dry
unit weight, 7) Void ratio, 8) Unified Soil Classification System (USCS) Symbol for Soil Type and 9) Not found.

3
Normalized shear modulus reduction and damping ratio curves

The dynamic behavior at a wide shear strain range, where the specimen exhibits its non-linearity and
strain-dependent behavior, is obtained by joining results from RCTS and CTX tests. This behavior is
performed in terms of normalized shear modulus reduction (G/Gmax) and damping ratio versus the
log of γ (G/Gmax-logγ and D-logγ) plots shown in Figures 2, 3 and 4 for all the materials tested, which
were separated by project, type and deposition method of the TL.
Curves generated from the formulation developed by Darendeli (2001) for fine-grained soils were
used to compare the results of nonlinear dynamic behavior of the TL materials assessed by the RCTS
tests. The equations developed by Darendeli (2001) were chosen due to their ability to predict G/G max-
logγ and D-logγ relationships using different confining stresses. A brief discussion regarding the
G/Gmax-logγ and D-logγ relationships is presented below.
 TL from Project 1 (Figure 2): The normalized shear modulus curves for the range of
confining stresses (100, 200 and 400 kPa) tested fit pretty well with those predicted by
Darendeli (2001). Nevertheless, regarding damping ratio curves, the results obtained in the
RCTS testing shown a slightly stiffer behavior and, therefore, much lower values while
observing the damping increment than the formulated curves by Darendeli (2001).
 TL from Project 2 (Figure 3): The dynamic curves for the range of confining stresses (200,
400 and 600 kPa) tested, are not in good agreement with those predicted by Darendeli
(2001), showing a stiffer behavior for G/Gmax-logγ and D-logγ relationships.
 TL from Project 3 (Figure 4): Both dynamic curves obtained from the testing and predicted
by Darendeli (2001) are in good agreement for the range of confining stresses up to 300 kPa.
However, for the range of confining stresses from 300 kPa up to 1000 kPa, the curves
predicted by Darendeli (2001) do not fit as it did for lower confining stresses. The results
obtained in the RCTS tests shows a much stiffer behavior for both normalized shear
modulus reduction and damping ratio curves than Darendeli (2001) curves. In addition,
the D-logγ relationship subjected to high confining stresses shows lower values for high
shear strains, are in good agreement with Ishibashi and Zhang (1993) curves.

4
Figure 2 Tailings samples tested from Project 1. Normalized shear modulus and damping ratio versus shear
strain and predictions of the dynamic curves from Darendeli (2001)

Figure 3 Tailings samples tested from Project 2. Normalized shear modulus and damping ratio versus shear
strain and predictions of the dynamic curves from Darendeli (2001)

5
 Figure 4 Tailings samples tested from Project 3. Normalized shear modulus and damping ratio versus shear
strain and predictions of the dynamic curves from Darendeli (2001)

Literature review
Currently, there is limited database available for estimating dynamic properties of TL. Winckler et
al. (2014) performed a laboratory testing program for non-plastic sand TL and found that normalized
shear modulus reduction and damping ratio curves mainly depend on confining stresses, as shown
in Figure 5. Furthermore, the study by Borja and Pretell (2012) presented coarse-grained TL results
obtained by means of a CTX device at the University of Utah for confining stresses of 250 and 500
kPa, the normalized shear modulus reduction and damping ratio curves were adjusted to a standard
hyperbolic model using eight points, which these points are shown in Figure 5.
A comparison between these dynamic curves, the results normalized shear modulus obtained in the
RCTS testing fit pretty well with proposed curves by Winckler et al. (2014); however, the results of
CTX tests shown a slightly better fit to 250 kPa than 500 kPa of confining stress. On the other hand,
damping ratio points just for confining stress of 250 kPa are close to the curves formulated by
Winckler et al. (2014); nevertheless, the damping ratio points for a confining stress of 500 kPa obtained
by using the CTX device are not consistent with the curves by Winckler et al. (2014).

6
 Figure 5 Normalized shear modulus and damping ratio versus shear strain from CTX tests performed at the
University of Utah and proposed dynamic curves by Winckler et al. (2014) for sand tailings

Maximum shear modulus

This approach consists on performing an adequate calculation of maximum shear modulus (Gmax)
based on geophysical surveys and seismic cone penetration test (SCPTu) and compared with results
of RCTS testing. A comparison between these approach described of field surveys and evaluated TL
specimens (TL-01, TL-03, TL-04 and TL-05) at laboratory are shown in Figure 6.

Figure 6 Comparison of maximum shear modulus obtained by RCTS, SCPTu and geophysical surveys

As can be seen in Figure 6, the Gmax values obtained by means of the RCTS program testing do not
agree in all the cases. As expected, Gmax increased with increasing confining stresses. In addition to

7
this, a linear trend between small-strain shear modulus and confining stress could be observed, as
suggested by prior research works (e.g. Martínez-Rojas et al., 2018).

RESULTS AND DISCUSSION

Proposed normalized shear modulus reduction and damping ratio

Figure 7 show the summary of all the RCTS testing performed in TL used in this study, adding the
results of the CTX tests (TL-06) conducted at the University of Utah. As shown in Figure 7, samples
TL-03 and TL-06 show a different trend when compared with other samples; however, the rest of
samples shares similar trend behavior for both normalized shear modulus and damping ratio.
According to Table 1, TL samples present void ratio (e) values ranged from 0.70 up to 0.89 and specific
gravity (Gs) ranged from 2.64 up to 3.49. When tailings are disposed in a TSF void ratio and Gs
variability is expected, however its influence on the dynamic properties is not observed in the testing
data shown in Figure 7.

Figure 7 Normalized shear modulus and damping ratio versus shear strain for tailings used in this study

Based on the previous discussion, the authors propose normalized shear modulus reduction and
damping ratio curves for non-plastic TL materials based on the results of the RCTS testing program
and comparison of these to literature review. The proposed curves for normalized shear modulus
reduction and damping ratio for TL materials tested in this study are shown in Figure 8 over a range
of confining stresses from 100 up to 1000 kPa.
The proposed curves for normalized shear modulus reduction are quite similar to those predicted by
Darendeli (2001) up to 0.01% of shear strain for low confining stresses; nonetheless, the formulated
curves by Ishibashi and Zhang (1993) fit very well for higher confining stresses.

8
It is important to mention that the proposed damping ratio curves for small shear strain is based both
on the laboratory testing results, as well as the Darendeli (2001) work. Nevertheless, for higher shear
strain are based on the formulated curves by Ishibashi and Zhang (1993).

Figure 8 Proposed curves for normalized shear modulus and damping ratio for tailings tested in this study

Figure 9 Proposed curves for normalized shear modulus and damping ratio for tailings

The proposed normalized shear modulus reduction and damping ratio curves for TL materials are
presented in Figure 9. These proposed curves were compared to the ones proposed by Winckler et
al. (2014), normalized shear modulus curves of Winckler et al. (2014) plainly showed a much stiffer
behavior for low confining stresses (100, 200 and 400 kPa). However, for higher confining stresses
(600 and 1000 kPa) shown a slightly stiffer behavior for shear strain values ranged from 0.01% up to
0.10%.
As for the case of damping ratio curves, the proposed curves by Winckler et al. (2014) fit pretty well
with predicted in this study, for high confining stresses (600 and 1000 kPa) and shear strain up to

9
0.01%. On the other hand, the curves presented by Winckler et al. (2014) are not in good agreement
with proposed normalized shear modulus reduction and damping ratio curves performed in this
study.
As shown in Figure 9, those curves might be taken as a preliminary and as a reference for obtaining
dynamic curves for seismic analysis of TSF. This should be combined with an adequate calculation
of Gmax based on geophysical surveys, seismic cone penetration testing or accepted empirical
relationships. However, a proper determination of these dynamic properties based on dynamic
laboratory testing always is needed in any case.

CONCLUSIONS
In order to assess the dynamic properties of tailings, samples from five types of these materials were
tested during the design and operation stages of each project and summarized for this study. The
tests performed on non-plastic coarse-grained and fine-grained tailings were evaluated at the Anddes
geotechnical laboratory using a combined resonant column and torsional shear device.
The laboratory testing results of normalized shear modulus versus shear strain fit pretty well with
the curves generated from the formulation developed by Darendeli (2001) for fine-grained soils. A
close fit is observed for low confining stresses up to 0.01% of shear strain, while for higher confining
stresses and shear strains of the materials tested, the results are in good agreement with curves
developed by Ishibashi and Zhang (1993) for sand and clays.
As to the case of the damping ratio curves, the laboratory results show a slightly stiffer behavior for
lower confining stresses than the curves predicted by Darendeli (2001); however, these curves
formulated fit pretty much better than others. For higher shear strains and confining stresses,
Darendeli’s (2001) curves represent more degradation than that of the laboratory results, but these
results rather close to the ones developed by Ishibashi and Zhang (1993).
Normalized shear modulus and damping ratio curves for tailings materials are proposed in this
study over a range of confining stresses from 100 up to 1000 kPa. These proposed curves are intented
to become a useful tool for design of tailings storage facilities in highly active seismic zones, due to it
involves the seismic stability analysis and needs to incorporate a proper determination of the
dynamic properties of the materials involved, such as tailings and foundation materials.
Further dynamic laboratory testing is needed to adjust or confirm the dynamic curves proposed in
this study. That laboratory testing program should include bender element, resonant column
torsional shear, cyclic triaxial, among others sophisticated devices; in order to reach a greater range
of shear strain and its subsequent variety of related behaviors.

10
ACKNOWLEDGEMENTS

The authors of this paper would like to express their gratitude to everyone who has contributed to
this study. Special thanks to the Anddes geotechnical engineers who supported the present work.

REFERENCES

Darendeli, M.B. (2001) Development of a new family of normalized modulus reduction and material damping
curves, Ph.D. Dissertation, University of Texas at Austin, Austin, Texas.

Ishibashi, I. and Zhang, X. (1993) Unified dynamic shear moduli and damping ratio of sand and clay, Soils
and Foundations, 33 (1), 182-191.

Martinez-Rojas, E., Molina-Gómez, F. and Ruge, J. C. (2018) Shear modulus of mining tailings from
resonant column tests, DYNA, 85 (205), pp. 310-318.

Menq, F. H. (2003) Dynamic properties of sandy and gravelly soils, Ph.D. Dissertation, University of Texas
at Austin, Austin, Texas.

Borja, R. and Pretell, R. (2012) Comparison of shear modulus curves with cyclic triaxial results and seismic
response analysis, VIII International Congress of Civil Engineering, Lima, Peru.

Rollins, K., Evans, M., Diehl, N., and Daily, W. (1998) Shear modulus and damping relationship for gravels,
Journal of Geotechnical and Environmental Engineering, 124 (5), 69-81.

Savidis, S. A., Vrettos, C. and Richter, T. (1993) Resonant column and cyclic triaxial testing of tailing dam
material, Soil Dynamics and Earthquake Engineering VI, London, 291-298.

Seed, H. B. and Idriss, I. M. (1970) Soil moduli and damping factors for dynamic response analysis, Report
EERC 70-10, Earthquake Engineering Research Center, University of California, Berkeley.

Seed, H. B., Wong, R. T., Idriss, I. M., and Tokimatsu, K. (1986) Moduli and damping factors for dynamic
analysis of cohesionless soils, Journal of Geotechnical Engineering, ASCE, 112 (GT11): pp. 1016-1032.

Sun, J., Goesorkhi, R., and Seed, H. B. (1988) Dynamic moduli and damping factors for cohesive soils,
Berkeley, CA: Department of Civil and Environmental Engineering, University of California.

Vucetic, M. and Dobry, R. (1991) Effect of soil plasticity on cyclic response, Journal of Geotechnical
Engineering, 117 (1), 89-107.

Winckler, C., Davidson, R., Yenne, L. and Pilz, J. (2014) CPTu-based state characterization of tailings
liquefaction susceptibility, Proceeding of Tailings and Mine Waste 2014, Keystone, Colorado, USA.

11
Using Seismic Records to Determine the
Predominant Vibration Frequency of a Tailings Dam
Embankment: First Results
Nicolle Correia1 and César Pasten2
1. Minera Escondida, BHP, Chile; Universidad de Chile
2. Universidad de Chile

ABSTRACT
A large number of tailings deposits in operation, non-active, and in abandonment are located mainly
in the northern and central Chile where the seismicity threatens the stability of these large and
saturated geostructures. The importance of the physical stability of these deposits is relevant since it
directly influences the life of mine, its closure plan, the impact on the environment, and the
surrounding communities. A common practice is that tailings deposits are instrumented with
different technologies to measure pore water pressure inside the wall, water levels, infiltrations,
deformations, and seismic accelerations. The monitoring and understanding of the information
collected is essential to ensure the stability of the deposit and its operational continuity. This work
seeks to determine the predominant vibration frequency of an instrumented tailings dam
embankment of 45 m height constructed from borrowed material by calculating empirical
amplification transfer functions and by using the horizontal to vertical spectral ratio method (HVSR).
Then, the equivalent linear wave propagation method is used to explain the observed predominant
vibration frequency considering estimations of shear wave velocity profiles of the embankment dam.

1
INTRODUCTION

The most common tailings dam failures in the World are related with slope instability, internal
erosion, and overtopping (ICOLD 2001). In Chile, the most critical threat to a tailings dam is the
occurrence of large earthquakes (Villavicencio et al., 2014). To prevent a failure of the deposit, it is
necessary to develop a deep understanding of the seismic behavior of the embankment and a good
estimation of the relevant geotechnical parameters. The seismic response of a deposit is related with
the geometry and the material properties that compose the wall, such as the stiffness that can be
obtained from the shear wave velocity. Also, the predominant vibration frequency of the structure
allows predicting the behavior of the embankment subjected to seismic loads. The predominant
vibration frequency of the structure can anticipate the occurrence of resonance that could increase
the risk of a catastrophic failure.
To determine the predominant vibration frequency of an earth structure, the horizontal-to-vertical
spectral ratio (HVSR), also known as the Nakamura’s method, can be calculated from continuous
ambient seismic noise records (Diaz-Segura 2016, Molnar et al. 2018). Moreover, the HVSR can also
be estimated by using earthquake records, providing similar results than ambient seismic noise
(Fernández et al. 2019). Using seismic stations installed on a cycloned-sand embankment dam, Pastén
et al. (2019) have demonstrated that the HVSR calculated either with ambient seismic noise or
earthquake records are similar. This result raises the question whether the HVSR method can be used
in embankments constructed with borrowed materials.
In this study, the predominant vibration frequency of a tailings dam embankment is obtained from
earthquake records using the HVSR method and empirical amplification transfer functions. Also, a
1D shear wave velocity profile for the embankment is analyzed with the equivalent linear method to
understand the amplification pattern observed in the earthquake records.

TAILINGS DAM AND SEISMIC STATIONS

The analyzed deposit corresponds to a tailings dam located in the northern Chile, the retained wall
is an earth fill compacted wall that has grown by the downstream construction method, currently has
a maximum height of 47 m, an upstream slope ratio of 1.8: 1 (H: V) and a downstream slope ratio of
2: 1 (H: V).
The tailings dam is monitored with 3 seismic stations, two installed in the crest of the embankment
and one installed in natural ground, on soil foundation, which is considered as the free field station.
In this paper, the data recorded by the basal station and the one located in the section with the highest
elevation of the deposit will be considered. Figure 1 shows a schematic of the tailings deposit and the
approximate location of the stations.
The seismic stations are Obsidian triaxial accelerographs from Kinemetrics with 3 + 1 record channels,
configured with a trigger of 0.01% and sampling frequency of 200 Hz. The stations remain in constant
operation, recording events that have accelerations larger than 0.001g.

2
The records used in this study correspond to low intensity earthquakes that occurred in the vicinity
of the tailings deposit during January 2018. In addition, there is 5.3 magnitude earthquake record
with epicentral distance of 50 km, with the largest acceleration and duration (0.1 g and 200 s). Each
record has three components: one in the vertical direction, the horizontal direction perpendicular to
the wall axis (transverse component) and the horizontal direction aligned with the wall axis
(longitudinal component).

Figure 1 Schematic profile of the deposit and location of the seismic stations

H/V SPECTRAL RATIOS AND EMPIRICAL TRANSFER FUNCTIONS

We calculated the horizontal-to-vertical spectral ratios (HVSR) using the seismic records of the
stations installed on the crest and on the toe of the embankment, using a window of 30 seconds that
contains the largest amplitude of the record, then the Fourier transfer function is calculated for each
component, the curve is smoothed and then divided the horizontal to vertical curve. The geometrical
average is calculating and represented by the black curve. The calculations were performed using the
open-source software Geopsy (www.geopsy.org). The HSVR curves in Figure 2a indicates a lack of a
predominant vibration frequency at the sensor located on the toe of the embankment, which is
characteristic of stiff soil deposits with low impedance contrast with the underlying bedrock. Figure
2b indicates a predominant vibration frequency, according to the peak spectral ratio amplitude, at
approximately 4 Hz for the transverse and longitudinal components. This is the predominant
vibration frequency of the embankment wall.
Figure 3 shows the empirical transfer functions in the transverse and longitudinal components. To
obtain the empirical transfer function, the same analysis is performed considering just the vertical or
the horizontal components. Both average curves in the figure show a peak spectral ratio at about 4
Hz that is consistent with the results from the HVSRs.

3
(a) (b)

Figure 2 HVSR calculated for the toe sensor (a) and for the crest sensor (b). The gray curves are the HVSRs of
every analyzed earthquake record. The black curve is the mean of all the gray curves

(a) 10 (b) 10
9 9
8 8
7 7
H/H longitudinal (-)
H/H transverse (-)

6 6
5 5
4 4
3 3
2 2
1 1
0 0
0 5 10 15 20 0 5 10 15 20
Frequency (Hz) Frequency (Hz)

Figure 3 Empirical transfer functions calculated for the transversal component(a) and for the longitudinal
component (b). The gray curves are the HVSRs of every analyzed earthquake record. The black curve is the
mean of all the gray curves

1D WAVE PROPAGATION ANALYSIS

One-dimensional (1D) equivalent linear analyses were performed, considering two shear wave
velocity profiles that represent the stiffness variation of the embankment wall with depth. The input
ground motion used in the analyses is the earthquake record of the basal sensor with the largest peak

4
acceleration. According with the soil material used in the construction of the wall, the mean curves
for sand proposed by Seed and Idris (1970) were considered in the analysis for the normalized shear
modulus reduction and damping curves. Then, the calculated acceleration time history is compared
with the motion recorded in the crest sensor for the same earthquake. These analyses were performed
using the open-source software DeepSoil (http://deepsoil.cee.illinois.edu).
The 47 m-depth shear wave velocity (Vs) profile assumed for the deposit, according with the lift
stages, is shown in red in Figure 4. This profile is a three-layer model with an increase of the Vs with
depth. The geotechnical properties of the model materials are shown in Table 1. Another multi-layer
Vs profile obtained from MASW and ReMi tests performed in the embankment wall is shown in black
in the Figure 4 for comparison. The results of these tests show an increase of the Vs with depth,
starting with a Vs of about 400 m/s at the crest of the wall and a Vs of 1200 m/s at the base. The high
values for the shear wave velocities and the increase of the Vs with depth agree with the type of
material used in the several lift stages of the wall and the technical specification for soil layer
compaction recommended by de designer.

Table 1 Geotechnical properties of the 3-layers model in Figure 4

Layer Thickness (m) Unit Weight (kN/m3) Shear Wave Velocity (m/s)

1 8 20 600

2 30 20 650

3 10 20 700

Bedrock - 21 1900

5
 2940

2930

Elevation (m)
2920

2910

2900

Multi-layer model
3-layers model
2890
0 500 1000 1500

Vs (m/s)

Figure 4 Shear wave velocity profiles of the embankment wall

The comparison between the actual earthquake record and the acceleration time histories obtained
with DeepSoil for the Vs models presented in Figure 4 are shown in Figure 5. Both Vs models respond
similar in terms of peak ground acceleration (0.09g for the three-layer model and 0.12g for the multi-
layer model); however, the overall amplitudes of the predicted signals are larger than the recorded
one.
0.15

0.1
Acceleration (g)

0.05

-0.05
3-layers model
-0.1
Recorded
-0.15
0 10 20 30 40 50 60 70
Time (s)

(a) (b)

Figure 5 Comparison between equivalent linear method results at the surface of the Vs models in Figure 4 and
recorded motion. (a) 3-layers model and (b) multi-layer model. registered

6
To validate the shear wave velocity models, their theoretical transfer functions were calculated with
the Deepsoil software (Figure 6). The fundamental frequency of the three-layers profile is 4.2 Hz and
the fundamental frequency of the multi-layer profile is 4.2 Hz, values similar than that obtained from
the HVSR method and the empirical transfer functions.

Figure 6 Theoretical transfer function of the three-layer Vs model in Figure 4

CONCLUSIONS

The horizontal to vertical spectral ratio method and the empirical transfer functions calculated from
earthquake records predict a predominant vibration frequency of 4 Hz, which is a value consistent
with the high stiffness of the compacted earth fill dam.
The analysis with the equivalent linear method shows that the peak ground acceleration and the
predominant vibration frequency can be explained with a simple three-layer model. This result
suggests that the seismic response of the embankment wall can be relatively well approximated by a
one-dimensional model when subjected to low magnitude earthquakes. The methodology followed
in this study seems to be valid to estimating the dynamic properties of a tailings dam embankment
and to improve the understanding of its seismic behaviour.

NOMENCLATURE
HVSR H/V spectral ratio
HHSR H/H spectral ratio
T0 predominant vibrating period
Vs shear wave velocity

7
REFERENCES

Fernández, J., Pastén, C., Ruiz, S., and Leyton, F. (2019) ‘Damage Assessment of the 2015 Mw 8.3
Illapel Earthquake in North-Central Chile’, Natural Hazards, 1-15, (accessed March 10, 2019,
https://doi.org/10.1007/s11069-018-3541-3)

International Committee on Large Dams ICOLD (2001) ‘Tailings Dams Risk of Dangerous
Occurrences’

Molnar, S., Cassidy, J.F., Castellaro, S., Cornou, C., Crow, H., Hunter, J.A., Matsushima, S., Sánchez-
Sesma, F.J. and Yong, A. (2018) ‘Application of microtremor horizontal-to-vertical spectral ratio
(MHVSR) analysis for site characterization: State of the art’, Surveys in Geophysics, 39, 4, 613-631.

Pastén, C., Comte, D., Peña, G., Burgos, J., and Rietbrock, A. (2019). Dynamic characterization of a
tailings dam embankment using a dense seismic array – Preliminary results. 6th International Seminar
on Tailings Management, Tailings 2019. Santiago, Chile.

Villavicencio, G., Espinace, R., Palma, J., Fourie, A., and Valenzuela, P. (2013) ‘Failure of sand tailings
dams in a highly seismic country’, Canadian Geotechnical Journal, 51, 449-464.

8
 Tailings Management at Pueblo Viejo:
Accomplishments of El Llagal TSF
Lenin Montas
Pueblo Viejo, Barrick, Dominican Republic

ABSTRACT
The geotechnical design of a mine tailings deposit is made considering the information obtained from
the in situ and laboratory tests, the theoretical analysis and the numerical modeling. The LLD, SD1,
SD2 and SD3 dams of TSF Llagal are heterogeneous earth structures which have among their
functions the storage of waste from the production process of Pueblo Viejo Mine, Dominican
Republic. The maximum level of the Llagal TSF is 265 meters above sea level.
The design of the Llagal TSF covers the behavior of compacted rockfill dams against the action of a
seismic movement, floods, for which purpose it is evaluated. Rigid material is compacted limestone
with few fines to provide robustness, using filter mantle in contact with soil. The low permeability
core is constructed inclined with compacted saprolite of acceptable plasticity, with control of
transverse cracks in breakage of dams by having filters of fluvial origin which are built horizontal to
the core, filters as foundation blankets and rockfill transition zones.
Llagal TSF is designed using de CDA guidelines and it is monitored permanently through a QA / QC
program, ensuring that quality is maintained from its construction, as well as that the guidelines
contemplated in its design and specifications are followed.
Llagal TSF has a world-class OMS guide which highlights established aspects for a sustainable
operation that accounts for constant monitoring of all Llagal tailings facilities: an automated
geotechnical instrumentation system, the measurement of air quality, sediments management,
permanent monitoring of the quality of surface and groundwater, emergency spillways and pumping
system for supernatant water, among others. Site has periodic review of regulators and corporate
senior reviewers and independent board of consultants.
TSF Llagal has been designed taking into account the recommendations of the ICOLD and it is
certified by the ICMI Cyanide Code.

There is no full article associated with this abstract.

1
2
Hydraulics,
Transportation
and Rheology
Dam Runout Sensitivity to the Rheological
Formulations Using a Two-dimensional Flexible-
Mesh Model
Reinaldo García1 Mario Morales2, Pilar García2, Javier Murillo2, Carmelo Juez2
1. Hydronia, USA
2. Universidad Zaragoza, Spain

ABSTRACT

This paper discusses the sensitivity of dam breach tailings runout depending on different
characterisations of the fluids using two alternative rheological formulations. We apply the
RiverFlow2D, a two-dimensional model that incorporates eight non-Newtonian rheological
relationships for hyperconcentrated flow modeling that often occur in tailings dam break simulations
including: Complete and Simplified Bingham models, Turbulent and Coulomb, Turbulent and Yield
Stress, Turbulent, Coulomb and Yield Stress, Quadratic (e.g. Herschley-Buckley), and granular flow.
The model simulates the single-phase flow of hyperconcentrated sediment mixtures in 2D using
flexible meshes formed by triangles that can be adapted to irregular terrain and boundaries, and can
take advantage of Graphic Processing Unit (GPU) hardware to run up to 600 times faster than the
non-parallelized computer code. The RiverFlow2D model was used to perform a dam break tailings
analysis with two rheological formulas, and results show that the runout from each simulation may
change by several kilometres as a function of the parameters and formulas used. The paper remarks
the importance of understanding the differences in the flow behaviour, which it is parametrised in
the rheological models taking into account the physical characteristic of the tailings such as viscosity,
density, yield stress and angle of repose as a critical step in performing dam breach tailings analysis.
Application to a tailings dam break with two different fluid properties and rheological formulation
show that runout distance differ in several kilometres for both simulations.

1
INTRODUCTION
Flooding from tailings dam breaches often involves high speed flows of hyperconcentrated fluids on
complex terrain and urban areas. Computer simulations can provide significant help in flood risk
assessments of these events. In this context, numerical modeling of hyperconcentrated flows entail
significant challenges. The flow resulting from the release of the deposited material exhibit flow
behavior typical of hyperconcentrated or granular fluids. Also, the high flow velocities and the need
to account for significant terrain irregularities downstream the breach, call for models to account non-
Newtonian fluids, and two-dimensions in the horizontal plane capable of adapting to arbitrarily
irregular geometries. Therefore, two-dimensional (2D) models based on flexible meshes can be
important to determine the material extent and depth, since they can capture the topographical
irregularities and also represent realistically the tailings flow dynamics. Due to the similarities with
the tailings dam materials flows, often mudflow models are applied to simulate tailings dams
flooding. However, until recently, the difficulties associated with the numerical implementation of
the non-linear stress terms that represent generalized non-Newtonian fluids in 2D flexible meshes
have made challenging the advance of practical 2D models. There have been some attempts to
develop 2D flow models for tailings dams. USACE (1990) proposed the MF-2D finite element model
for mudflows. This model implemented a Bingham fluid stress term and was based on a movable
mesh, which made the model difficult to apply in real environments with variable terrain, internal
obstacles and irregular boundaries. O’Brien et al. (1993) proposed FLO-2D, a quasi 2D model for
mudflows that uses a quadratic stress term accounting for the material yield stress and viscosity. The
model uses a square-element cells of equal size throughout the grid and one-dimensional (1D)
approximation to compute velocity between elements, which compromises the model ability to
capture realistic velocity fields. Rickenmann et al. (2006) compared three mudflow models with field
events and concluded that the accurate representation of topography is essential to achieve god
replication of observed deposition patterns. Although it has been recognized that some tailings flow
as granular materials (Villavicencio et al. 2011), most 2D models have not incorporated rheological
relationships for granular fluids.
This paper discusses applications of the RiverFlow2D model that is based on the full shallow water
equations for viscous hyperconcentrated non-Newtonian fluids using triangular-cell flexible meshes.
The Methodology section presents the model equations and its eight rheological friction laws. The
following sections describe the RiverFlow2D Graphic User Interface, and the model data
requirements. Finally we present a model validation test, an application to a mud flow event, and a
discussion of a tailings dam flood simulation using two rheological laws.

2
METHODOLOGY

The system of equations used to simulate the flow of hyperconcentrated materials over initially dry
terrain in RiverFlow2D are obtained by coupling the full mass and momentum shallow-water
equations in 2D for viscous fluids with a flexible rheological formulation of the stress term that
considers the fluid viscosity, density, yield stress and friction angle. That system of non-linear partial
differential equations is formulated in a conservative form as follows:
(1) U F(U) G(U)
   S(U, x, y)
t x y
Where U= (h, qx, qy)T is the vector of conserved variables with h representing the material depth,
qx=uh and qy=vh the unit discharges, with (u, v) the depth averaged components of the velocity vector
along the (x, y) coordinates respectively. The flux vectors are given by:

T T
(2)  q y2 1 2 qx q y   qx q y q y2 1 2 
F   qx ,  gh ,  , G   qy , ,  gh 
 h 2 h   h h 2 
  

Where g is the gravitational acceleration. The terms ½ gh2 in the fluxes have been obtained after
assuming a hydrostatic pressure distribution in the vertical, as usually accepted in shallow water
models. The source term vector (S) incorporates the effect of pressure force over the bottom and the
tangential forces generated by fluid viscosity,
T
(3)
 z  z  
S   0,  gh(h b  bx ),  g (h b  by ) 
 x  y  
Where zb is the terrain elevation, ρ is the fluid density and τbx and τby are the bed stresses in x and y
direction respectively that depend on the fluid viscosity, turbulent friction and yield stress.

The formulation allows representing a wide range of viscous fluids including those exhibiting non-
Newtonian behavior that could be characterized with a yield stress, Murillo & García-Navarro (2012).
The RiverFlow2D model performs its calculations on a flexible mesh formed by triangular cells of
different size that can be adapted to virtually any geometry (see Figure 1). Cell data include the
bottom elevation and the surface roughness based on the terrain characteristics. The numerical
engine then determines the velocity vector and depth at each cell throughout the simulation using a
finite-volume scheme, described in detail in (Murillo et al. 2008; Murillo and García-Navarro 2012).
The Gauss theorem is applied to each volume cell allowing the computation of the flux through the
edges of each cell. Each variable is updated using an explicit first order scheme in time. The allowable
time step sizes are assigned dynamically and are automatically controlled by the Courant–Friedrich–
Lewy condition (Murillo & García-Navarro 2008). The model is well balanced and allows a stable
computed solution under all ranges of flow problems ensuring exact mass conservation. The model
has been validated for accuracy using a number of tests (Murillo & García-Navarro 2012). The

3
numerical methods ensure virtually zero error in volume conservation and stability through a
dynamic time-stepping scheme that does not require user-defined adjustments.

Figure 1 Flexible-mesh over an urban area with variable resolution allowing adaptation to irregular
boundaries and complex terrain

Tailings Dams Fluids

Floods generated from failures of tailings dams are unsteady high-speed flow phenomena in which
the properties of the moving fluid are those of a hyperconcentrated viscous mixture water and solid
materials. The flow often exhibits stop and go mechanisms. The global resistance behavior of tailings
dam fluids depends on the relative importance of the shear stresses arising from different sources
that, apart from turbulent shear stress at the river boundary, include viscous stress, yield stress,
dispersive stress and inelastic collisions of solid particles within the fluid mixture (Naef et al. 2006).
Some dry tailings dam materials behave as granular flows (Villavicencio et al. 2011).
The mathematical model adopted here is that of Murillo & Garcia-Navarro (2012), and regards bed
friction modelling in free-surface flows ranging from clear water to hyperconcentrated mixtures of
sediments to granular materials. The model involves the following assumptions:
 The flow is governed by equation (1) with the friction terms evaluated as explained
Rheological Formulation section;
 The mixture of water and solid material is described by using the continuum approach
assuming the same velocity for the liquid and for the solid phase;
 The river bed does not erode;
 The fluid has constant properties throughout the simulation: e.g. density, viscosity, yield
stress, friction angle, etc.

4
Rheological Formulation

The single-phase rheological formulation in RiverFlow2D accounts for different friction terms that
represent a variety of hyperconcentrated non-Newtonian fluids (see Figure 2). Since all terms in
equation (3) are depth-averaged, the equations that describe the tangential forces generated by the
stresses can be lumped into the same mathematical formula despite of having a different nature. In
this way, several shear stresses can be considered to cover a wide range tailings dam break flow

conditions, including turbulent stress  t , dispersive stress  d , Coulomb-type frictional stress  f

, Yield stress y and Viscous stress   .

Figure 2 Rheological diagram of some common non-Newtonian formulations

Table 1 shows the rheological friction laws used in RiverFlow2D model.

Table 1 Rheological flow resistance formulations used in RiverFlow2D model.

Formulation Flow resistance term Input parameters

Turbulent b  t Manning’s n.

Full Bingham  b   0 with f ( , )  0 Bingham viscosity, yield stress, density.

1 0 y

Simplified  b  1.5 y  3  Bingham viscosity, yield stress, density.

Bingham

5
 Turbulent and  b   t  f Manning’s n, friction angle, density.
Coulomb

Turbulent and  b   t  y Manning’s n, yield Stress.

Yield

Turbulent,  b   t  min( y , f ) Manning’s n, yield stress, friction angle,

Coulomb and density.
Yield

 Manning’s n, yield stress, density.

Quadratic b  t  y   
8

Granular b   f Manning’s n = 0, friction angle.

GRAPHICAL USER INTERFACE AND SIMULATION WORKFLOW

To simplify practical model setup and generation of practical mapping, the model is integrated with
the QGIS Geographical Information System (www.qgis.org). This GIS software provides interactive
functions to generate and refine the finite-volume flexible mesh. QGIS uses GIS objects such as points,
arcs, and polygons to construct a high level representation of the model area, facilitating assigning
boundary conditions and roughness values and allowing the user to efficiently manage the modelling
process. The system allows input data in several GIS formats and can generate maps in 3D,
animations and export graphs to Google Earth.

DATA REQUIREMENTS

The data required to run the model includes: Terrain elevations often provided as a set of XYZ ASCII
file, LiDAR, etc., geometry of buildings or other obstacles that may affect flow, surface roughness
(Manning’s n), rheological formulation, material properties corresponding to the selected rheological
formulation and breach location on dam.

MODEL VALIDATION TEST

In this test, model results are compared with experimental dam-break flood measurements from a
laboratory study performed by Recherches Hydrauliques Laboratory Chatelet and University of
Bruxelles (Brufau et al. 2002). The experiment allows to test the model validity to simulate sudden
spills of viscous Newtonian fluids. The test involves the advance of a dam-break wave over a
triangular obstacle located on a horizontal concrete bed channel as indicated in Figure 3. Initial

6
reservoir depth is 0.75 m. The test starts by instantaneously removing the dam-gate. Depth evolution
is measured at gauging points G4, G10, and G11 shown in Figure 3.
The triangular-cell mesh used in the numerical computations has 4770 triangular elements. The
downstream condition is an open outflow boundary. The total simulation time is 40 s. and the time
step is dynamically adjusted by the model to satisfy the model numerical stability condition. The
output data is reported every 1 s.

Figure 3 Experimental model for dam-break flood over a triangular obstacle. G4, G10, and G11 indicate
location of the gauging points

Figure 4 and shows the measured and numerical evolution of depth at 2 gauging stations
respectively. The concordance between the experimental and numerical results is considered
satisfactory. At the points located before the obstacle, the prediction of the arrival time of the wave
as well as the water depth is good. It can be observed that the prediction of the transitions from wet
to dry is correct.

Figure 4 Depth evolution during 40 s at gauging points G4 and G10 in the simulation of a dam-break flood
over a triangular obstacle. Points indicate measured data and solid line model results

7
APPLICATION TO RUDD CREEK MUDFLOW

The model capability to represent flow of high viscosity fluids having a yield stress is compared with
an actual event that occurred in the alluvial fan of Rudd Creek near Farmington, Utah in May 1983
(USACE, 1988). Observations from this event have been used for comparison with other simulation
models (USACE, 1990; O’Brien et al. 1983). The inflow hydrograph was estimated by the USACE and
indicates that the event occurs suddenly reaching a peak discharge in less than 4 minutes and ending
in less than 12 minutes (see Figure 5) and is shown in and the inflow location. The area topography
was digitized from the contour map in USACE (1988). The fluid properties including mud density of
2000 kg/m3, viscosity 3000 Pa.s and yield stress 2390 Pa were obtained after calibrating against
observations. For this illustrative application we assumed a uniform Manning’s n of 0.04 and did not
considered the effect of buildings or streets.

Figure 5 Inflow hydrograph estimated by USACE (1990)

According to observations, the maximum depths occurred near the alluvial fan apex and was close
to 3.7 m. The computed maximum depth ranges from 3.5 to 3.9 m at the fan apex (Figure 9). Mudflow
velocities ranged from 0.3 to 1.2 m/s on the fan (Figure 10) which agrees well with the estimated
velocities of 0.6-1.2 m/s (USACE, 1988). Close to the fan apex the observed velocities were less than 3
m/s while the maximum calculated velocities were around 2.6 m/s.

8
 Figure 6 Rudd Creek Mudflow Simulation Depths after 30 minutes. The yellow polygon indicates the
observed mudflow boundary. Left figure shows material depths while right figure shows velocities

The predicted flood extent is close to the observed affected area which shows that the model is
capable of reasonably reproducing the depositional pattern of hyperconcentrated fluids in complex
terrain.

APPLICATION TO SIMULATE TAILINGS DAM RUNOUT WITH TWO

RHEOLOGICAL FORMULATIONS
The physical properties of mine tailings may vary due to the material water content and other factors.
Therefore the rheological flow characteristic should be accounted for by appropriate rheological
formulation. For instance, dryer tailings tend to behave more like granular materials (Villavicencio
et al. 2011), while tailings with higher water content show more classical non-Newtonian properties
(Nguyen & Boger, 1998). To illustrate the importance of using an appropriate rheology to the material
characteristics, we present an application to simulate a tailings dam runout process using two
rheological formulations. In the example we model the flooding caused by the hypothetical break of
a 90 m tailings dam shown with the initial material deposit shown in Figure 7.

9
 Figure 7 Initial material elevations in the tailings dam application. Colors indicate terrain elevations

A 655,240-cell mesh was generated over the dam deposits and extending 14 km on the downstream
valley. In order to assess the variability in tailings runout distance based on fluid properties, and
rheological formulation we present two cases. The first case involves the use of the Bingham fluid
formulation with the properties listed in Table 2.

Table 2 Material parameters used in Bingham Fluid Scenario.

Parameter Value

Yield Stress 150 Pa

Viscosity 100 Pa s

Density 1845 kg/m3

The second case uses the same mesh as case 1 but uses the granular flow rheological formulation with
the parameters indicated in Table 3.

Table 3 Material parameters used in Granular Flow Scenario.

Parameter Value

Friction angle 38°

Density 1845 kg/m3

Note that both sets of fluid properties presented in Table 2 and in Table 3 data can occur on the dam
tailings depending on the material water content existing at a given time.

10
RESULTS AND DISCUSSION

Figure 8 depicts the simulated flow depth 1 hour after the partial dam-break for case 1 that uses the
Bingham fluid option. Note that the fluid rapidly reaches the lower levels of the river valley that is
about 8 km downstream of the dam.
Figure 9 shows the simulated granular flow depths that reaches only 500 m at one hour after the dam
break and also has reached almost complete stoppage. It is clear that the flooding extent is much
bigger for the Bingham material than for the granular flow.

Yellow polygon indicates flooded

area using granular flow
formulation in Case 2 Figure 9.

Figure 8 Flood extent at 1 hr after the dam breach using Bingham rheological formulation. The total runout
distance is about 14 km from the dam site

11
 Figure 9 Flood extent at 1 hr after the dam breach using granular flow rheological formulation. . The total
runout distance is less than 1km from the dam site

CONCLUSIONS

This paper describes the application of a two-dimensional model (RiverFlow2D) to simulate tailings
dam-breaks with the purpose of assessing the variability in runout distance based on the use of
different rheological formulas and fluid properties. The model uses flexible meshes and includes
eight different non-Newtonian rheological relationships that allow simulation of hyperconcentrated
sediment mixes that involve non-Newtonian and granular fluids. Using two-dimensional triangular-
cell meshes allow modeling hyperconcentrated flows over complex terrain and urban including
buildings and hydraulic structures. Runs with the high-performance GPU-based model option can
be more than 600 times faster than the non-parallelized version. The comparison of model results
with dam-break measurements in a laboratory flume shows that the model can closely replicate the
experimental data. The practical application to the Rudd Creek mudflow event, demonstrates that
the predicted depths, yield-stress fluid stoppage and deposit extent compares well with the
observations in the real event. Finally, a tailings dam-break simulation with two rheological
formulations and two different fluid properties show that runout distance differ in several kilometres
for both simulations. It is important to remark that both sets of fluid properties can occur on the dam
tailings based on the water content at a given time, which stresses the importance of using an
appropriate friction law corresponding to the actual tailing characteristics.

REFERENCES

Brufau P., García-Navarro P. and Vázquez-Cendón M.E. “Zero mass error using unsteady wetting-

12
drying conditions in shallow flows over dry irregular topography”, Int. Journal for Numerical
Methods in Fluids, 45, 1047-1082. 2004.

Lacasta, A. M. Gonzalez-Sanchis, J. Murillo, P. Garcia-Navarro A compromise between

computational load and refinement criteria in two-dimensional hydraulic simulation: a real case,
Hydroinformatics, Hamburg. 2012,

Lacasta A, García-Navarro P, Burguete J, Murillo J. Preprocess static subdomain decomposition in

practical cases of 2D unsteady hydraulic simulation. Computers & Fluids 80: 225-232. 2013.

Lacasta, A., Morales, M., Murillo, J, Garcia-Navarro, P., Garcia, R. High Performance GPU Speed-up
strategies for the Computation of 2D Inundation Models, Hydroinformatics, New York. 2014.

Murillo, J. and P. Garcıa-Navarro, “Wave Riemann Description of Friction Terms in Unsteady

Shallow Flows: Application to Water and Mud/Debris Floods”, Journal of Computational Physics, 231:
1963–2001, 2012.

Nguyen, Q.D. and Boger, D.V. Application of rheology to solving tailings disposal problems. Int. J.
Miner. Process. 54 pp. 217-233. 1998.

O'Brien, J., Julien, P., and Fullerton, W. Two‐Dimensional Water Flood and Mudflow Simulation. J.
Hydraul. Eng. Vol. 119, No. 2. 1993.

Rickenmann, D., D. Laigle, B. W. McArdell and J. Hubl, Comparison of 2D debris-flow simulation

models with field events. Computational Geosciences. 10:241–264. 2006

USACE, Review of the U.S. Army Corps of Engineers Involvement with Alluvial Fan Flooding
Problems. Report TP-124. U.S. Army Corps of Engineers, Hydrologic Engineering Center. 1988.

USACE, MF-2D Two-Dimensional Simulation Model for Mudflows. U.S. Army Corps of Engineers,
Hydrologic Engineering Center. 1990.

Villavicencio, A. G., Pierre Breul, Claude Bacconnet, Daniel Boissier, A. Raúl Espinace. Estimation of
the Variability of Tailings Dams Properties in Order to Perform Probabilistic Assessment. Geotechnical
and Geological Engineering, Volume 29, Issue 6, pp 1073-1084, 2011

13
Optimization of Maintenance Plans for Tailings
Transport Systems
Omar Mejías
Ausenco, Chile

ABSTRACT

The transport of mine tailings through pipelines, whose layout covers great distances, presents a
significant challenge from the operational point of view, since the risks associated with the operation
of these systems have the potential to cause negative impacts on the environment in which they are
inserted as well as on the business in general, and it is in this context that the efforts that can be made
to ensure the availability and overall performance of the system take special relevance. Considering
that environmental regulations are increasingly restrictive and demanding, a failure of these systems
can have very adverse effects, including the revocation of permits to operate.
This work is based on studies carried out to optimize the performance of tailings transport systems,
through the optimization of their maintenance plans. These systems consist of several kilometers of
pipeline with intermediate choke stations. The optimization is developed from the analysis of the
dominant failure modes that can take place during the operation of the system. The aim is to prevent
and eliminate those failure modes through the execution of maintenance tasks, such as inspections,
predictive maintenance and scheduled replacement of components that make up the system.
Supported in fundamentals of reliability centered maintenance (RCM), the appropriate type of task
is defined and the frequency with which it must be executed, considering factors such as the behavior
of failure modes, probability of occurrence, consequences of failure and task feasibility. The main
purpose of the optimization efforts is not to leave dominant failure modes unattended and avoid
using resources in activities that do not add value.
The study presents the impact of optimization on maintenance costs, improvement of the availability
and mitigation of system risks, identifying opportunities for improvement and defining the resources
required to ensure the operational reliability.

1
INTRODUCTION

What is stated in this document is a study carried out to optimize the maintenance plan of a tailings
transport system using a methodology based on the RCM principles. The purpose of the study is to
contribute to the assurance of the operational reliability of the tailings transport systems.
The transport of mining tailings where the route usually covers great distances requires a strategy
and maintenance plans designed in such a way that it allows to manage the risks inherent to its
operation in an effective and efficient way, thus ensuring adequate protection of the environment.
where it operates and business continuity. Considering these factors, it is important to have
maintenance tactics that can take care of the failure modes that can complicate the operation of these
systems.
The study presents the methodology used for the treatment of failure modes and how to build the
maintenance plan focusing on the dominant failure modes. The typical types of tasks that should be
included in the maintenance plan are presented and some results obtained for existing operations
cases are indicated, and finally, some conclusions are exposed in this respect.

METHODOLOGY

The methodology is based on RCM and the optimization is developed from the existing maintenance
plans for tailing transport systems, analyzing the failure modes that are being addressed, the
suitability of the tasks and if there are failure modes to be addressed. RCM is a process used to
determine what should be done to ensure that any physical asset continues to do what its users want
it to do (Moubray, 2004).

Failure Mode Analysis

Understanding the failure mode (FM) is fundamental in any RCM based process. This first and key
step of the optimization process is to understand why a task is performed by defining what needs to
be prevented or detected. The definition of the FM contains the mechanism, which is the condition
we expect to find, combined with its cause. A good understanding of a FM helps identify the
appropriate primary action.
Once defined each FM needs to be tested so only the likely or dominant FMs are considered. This is
done using the following testing process which contains three questions:

2
 Figure 1 Failure Mode testing process

Should the answer to all of these questions be no, then the FM is considered to be unlikely and
therefore the task relating to this FM is classified as redundant and removed from the maintenance
plan.

Maintenance Strategy Decision

Even though a dominant FM has an existing task associated with it, the use of the most appropriate
technique for predicting or preventing this FM needs to be validated. The decision diagram below is
used to determine the appropriate type of task (inspections, fault finding, condition monitoring or
preventive tasks) that can be used for the components of the tailings transport system.

Figure 2 Maintenance strategy decision diagram

3
Where the FM has been determined as dominant and the proper strategy identified, the task
description is verified to ensure there is no ambiguity.
In the case of condition based or fault finding interval inspections for each primary action, a
secondary action must be defined that should be executed if the desired condition of the component
of the system is not met.

Task Frequency

Having determined that a FM is dominant, what type of task is most appropriate and what the most
effective task description is, it is important to determine the appropriate frequency or interval at
which to conduct the primary activity.
For condition-based tasks, the frequency is based on an estimate of the P-F interval. Typically, the
task frequency is set at half the estimated P-F interval duration.

Sources of Information

The existing maintenance plans of the system are used in the first instance and optimised as required.
Where maintenance tasks do not exist for identified failure modes, it is necessary to define new
maintenance tasks. For these purposes, workshops are held with the technicians of the operation of
the tailings transport system.
In addition, sources of information area required, such as:
 Operation and maintenance manuals
 Plans and specifications
 Process diagrams
 Fault history
 Registry of repairs
 List of materials
In the workshops it must be validated the frequency and resources required by tasks.

RESULTS AND DISCUSSION

In order to assess the impact of the optimised maintenance plan, results must be compared with the
existing tactics before the optimisation.

Failure Modes (FMs)

FMs for all the components covered by the study were reviewed to ensure that they apply to the
operating context. Non-dominant FMs were eliminated and only the remaining dominant FMs were
considered.

4
In table 1, there is a list of the most relevant FMs that occur in a tailing transport system. The data
presented in this table have been obtained from real cases studied and represent an average of the
incidence of the different failure modes in the reviewed cases. The percentages represent the portion
of the total failure modes analysed for the different components of the system.

Table 1 Relevance of Failure Modes in Tailings Transport System

Component Failure Mode Percentage

Pipe lining Wear for abrasion 25%

Choke Break for fatigue 15%

Valves Blocked for contamination 10%

Structure Corroded for chemical attack 10%

Pipe Corroded for chemical attack 8%

Bolts Loose for vibration 8%

Seals Degrade for use 7%

Electrical components Short-circuit for contamination 5%

Pipe Break for fatigue 2%

Others Others 10%

From the previous table it is clear that the main failure mode is the wear of the pipes by abrasion, this
confirms that there must be a robust maintenance strategy to have this failure mode under control.
Another of the relevant failure modes is the breakage of the choke discs, and due to its condition of
difficult access to perform inspections, the most feasible is perhaps the replacements at fixed time
with adequate programming.
In the studies carried out, it was identified that of all the failure modes analysed, 8% corresponds to
failure modes that were not being addressed by the maintenance plan before the optimization. In
addition, 5% of the fault modes were redundant.

Maintenance Tasks

For each failure mode, the appropriate primary tasks were defined in order to address the respective
failure mode in the most efficient way. Table 2 shows the distribution of primary tasks by type of
maintenance activity. The data presented in this table have also been obtained from real cases studied

5
and the percentages represent the proportion of the type of task recommended in relation to the total
of tasks associated with the failure modes analysed.

Table 2 Maintenance Task Types for Tailing Transport System

Type of Task Percentage

Condition based 85%

Fixed time 10%

Operate to failure 5%

The tasks that are finally left in the maintenance tactics, in addition to verifying that they address to
a dominant failure mode, should be improved in their description, clearly establish the limits of
acceptance for the case of inspections or measurement of parameters and on the other hand, it should
be validate the amount of resources required for its execution.
After the optimization, there are tasks that were excluded from the maintenance plan, either because
they were repeated, or attended the same failure mode as another task or simply did not add value.
An example of tasks that are excluded from the maintenance plan in real cases, is the inspection of a
valve motor by loose elements, any looseness should be detected by a vibration analysis and it is
considered that this technique is more effective for this case.
It can be seen that most of the tasks that make up maintenance tactics are of the condition-based type.
Among these tasks, the most relevant are those related to the thickness measurement of the tailings
transport pipes. Among the preventive tasks at fixed time, the most relevant correspond to the
changes of choke discs.
It is a proven fact in the industry that the choice of suitable materials and the monitoring of tailings
pipeline conditions are critical aspects to guarantee optimal operation (Minería Chilena, 2012).
Although it depends on maturity degree and efficiency of the existing plans and strategies, the
author's experience allows to estimate that maintenance tasks could be reduced between 10 and 15%
for tailings transport operations after optimization.
Despite addressing a greater number of failure modes after optimization, the number of tasks would
decrease as a result of the elimination of some due to the causes previously exposed and considering
that finally in the maintenance plan will be considered the most effective tasks.
The maintenance plan is finally defined through the grouping and programming of the optimized
tasks, considering the frequency of execution and the resources required. The programming is
fundamental since the laying of the pipes covers long distances so the interventions to the system
must be well taken advantage of.

6
CONCLUSION

The maintenance plan of a tailings transport system is based strongly on predictive or condition-
based inspections. Given its operational regime, the techniques used to prevent the occurrence of
failure modes must be the most effective. Within these activities are the thickness measurement of
the inner lining of the pipes by ultrasound so as to keep the wear under control.
Within the fixed-time preventive tasks of the maintenance plan the most relevant refers to replace of
choke discs, the appropriate definition of the frequency for this type of activity is relevant to avoid
negative consequences.
Other important tasks to perform on these systems is the cleaning of valves, electrical and
instrumentation components. In addition, preventive and predictive inspections to identify corrosion
in structures and systems also becomes important.
In relation to the rupture of pipes, although the probabilities of occurrence are low, the consequences
are catastrophic since the tracings are generally in places of high environmental value. Due to its
normal operation, mainly what is controlled is the internal wear and can set rates to anticipate that
condemnation limits are reached, but if there are manufacturing defects or failures introduced during
assembly is a little more complex to deal with these faults. That is why, the importance that you take
the inspection routes in order to identify as soon as possible some type of event of this type.
The results of the studies carried out are aligned with good practices for the transport of tailings as
the inspection of entire pipeline for wearing, leaks or breaks, perform opportune and scheduled
maintenance and reclaim any disturbed area (Yukon Government Energy, Mines and Resources,
2017).
The number of maintenance tasks is usually reduced after an optimization, and those that remain in
the plan are the most effective. Having effective tasks allow to strengthen the maintenance plan
improving the reliability of the system and better managing the risk that this type of operation entails.
In studies of this type it is usual that despite having fewer tasks, more failure modes are addressed.

NOMENCLATURE
RCM Reliability Centered Maintenance
FM Failure Mode
P-F Potential to functional failure

REFERENCES

Moubray, J. (2004). Reliability Centred Maintenance, 3rd Edn, Aladon LLC, Lillington, North Carolina.

Minería Chilena (2012) Por un sistema eficiente y confiable, viewed 18 December 2018,
<http://www.mch.cl/reportajes/por-un-sistema-eficiente-y-confiable/>

7
Yukon Government Energy, Mines and Resources (2017) Wolverine Mine, Tailings storage facility,
Operation, Maintenance and Surveillance manual, viewed 11 February 2019,
<http://www.emr.gov.yk.ca/mining/pdf/mml-wolverine-tailings-oms-v2017-02.pdf>

8
Tailings Spigots Hydraulic Scale Modelling
Vanessa Veloso, Ray Martinson and Leonardo Olavarria
Paterson & Cooke, Chile

ABSTRACT
The resolution of non-conventional hydraulic problems by means of physical scale modelling is a
common practice in large hydraulic infrastructure designs, related mainly with hydropower
generation. In the design of tailings transportation system, this technique has seen some popularity.
Nowadays the use of computational fluid-dynamic (CFD) calculations is popular in predicting the
behavior of complex systems, but using a physical model is still an excellent tool to confirm the results
and present a realistic perspective of the phenomena.
The behavior of a high rheology slurry and the prediction of how a spigot with different discharge
angles will respond under certain conditions (flow rate, rheology, geometry, velocity) in a
gravitational flow regime is a complex hydraulic challenge. The proper representation and modelling
of the phenomena can improve the functionality of these types of important systems.
This paper presents a case study about a physical hydraulic scale model for a tailings spigot discharge
operating as an off-take of a main pipeline operating in gravity driven slack flow. Different spigot
angle configurations were tested in order to compare the performance of each configuration in
relation to the incoming versus the resulting discharge flow rates. The most interesting aspects,
challenges and results are described. Also a comparison with CFD modelling of the same system is
presented.

1
INTRODUCTION

A Chilean copper mining operation has a tailings gravitational distribution system with four
discharge points operating simultaneously. However, the discharge flows are unequal among them
with the last discharge point having a flow that is much greater than the others. This situation
generates problems with the correct tailings distribution at the dam as well as in the water recovery
system from the tailings pond. Consequently, the operators need to build channels inside the deposit
to correct the tailings distribution.
The tailings distribution system has a nominal diameter of 64 inches (main pipe) with side discharge
points at 90° angles. The side discharge pipes have a nominal diameter of 36 inches (discharge spigot).
Each discharge sector consists of four discharge points. The tailings flowrate of the system in
evaluation is between 3,500 m3/h to 15,000 m3/h.
The physical scale hydraulic model evaluates the effectiveness of the different discharge alternatives
in order to solve the current imbalanced distribution of the existing flows. In this case, the discharge
pipe orientation was modified varied to angles of 30° and 38° relative to the main distribution pipe
axis. In addition, the main distribution pipe was also rotated on its axis so that the discharge pipes
sit at -28°, -35°, and -42° below horizontal. The distribution system currently being used, with the
discharge points set at 90° was also tested with the model and the results were compared with CFD
results and the operational data.

HYDRAULIC SCALE MODEL CONCEPTS

Hydraulic modelling has been a useful tool to have a better understanding of certain physical
phenomena where it is difficult and sometimes impossible to accurately predict the results. A
prototype could be represented adequately by a scale model; therefore, each element of the scaled
model must be similar to the prototype in form, motion aspects and forces involved. It is important
to consider that there are scale effects that can be caused by the forces that are not the predominant
ones, such as fluid surface tension. The scales have been grouped in three categories: geometry, flow
velocity, and viscosity.

Geometric Scale

The geometric scale was selected considering the available space at the laboratory, the
representativeness and the availability of commercial size pipe. The selected nominal scale was
finally 1:20. The model diameter of main pipe is 69.85 mm, and the discharge pipe is 40.48 mm.

Flow and Velocity Scale

In the case of the gravitational flow, Froude number (Fr) is considered to be the more appropriate
(Goosen, P.E.). The similarity between the Froude number of the prototype and the model is the base

2
to ensure a similarity of inertial forces and gravity. The scale result for velocities and flowrate is
detailed below:
1
𝑉𝑚 = 𝜆2 𝑉𝑝 (1)

The ratio of velocities for the nominal scale is:

𝑉𝑚 1 1/2
𝜆𝑉 =
𝑉𝑝
=( )
20
= 0.2236 (2)

Velocity in the model is 22.36% of the actual velocity present in the prototype.
The flowrate relation is as follows:
5
𝑄𝑚 = 𝜆2 𝑄𝑝 (3)

The relation between flow rates for the nominal scale is:

𝑄𝑚 1 5/2
𝜆𝑄 = =( ) = 5.59 ∙ 10−4 (4)
𝑄𝑝 20

Flow rate in the model is 5.59 ten thousandths of the flow in the prototype.

Viscosity Scale

Despite the fact that the real fluid is non-Newtonian with a yield stress of 33 Pa and plastic viscosity
of 0.02 Pa∙s; the selection of flow model utilized the Reynolds number because the test fluid is
Newtonian. The apparent viscosity estimated is 0.99 Pa∙s which was used to emulate the operating
conditions of the system (this approximation was done using the slurry rheogram curve as base). The
relationship between the dynamic viscosities between the model and the prototype is as follows:
𝜌𝑚 3
𝜇𝑚 =
𝜌𝑝
𝜆2 𝜇𝑝 (5)

According to the density values of the fluids, the expression of the relationship between the dynamic
viscosities is obtained as follows:
𝑘𝑔 3
𝜇𝑚 1000 3 1
𝜆𝜇 =
𝜇𝑝
= 𝑚
1460 3
𝑘𝑔
20
2
∙ ( ) = 7.66 ∙ 10−3 (6)
𝑚

Then the dynamic viscosity required for the test fluid is therefore 0.0077 Pa.s.

This approach was confirmed later during experimentation, by comparing the actual condition of
the system with the model results obtained. The results indicated a very good similarity between
both.

SCALE MODEL DESCRIPTION

The hydraulic model required additional facilities to allow the execution of the tests. The liquid to be
tested starts from the pump, goes through the pipe to the inflow measuring box and the test pieces

3
where the flow is split in two flows that go to the outflow and through-flow measuring box. Each of
these elements will be described briefly.

Inflow
measuring box Piece IF: Inflow
tested
Main Pipe

Through flow
measuring box

Spigot

Outflow
measuring box TF: Through-flow
Pump

OF: Outflow

Figure 1 Elements of the hydraulic scaled model and example of piece tested

Pump

A centrifugal pump that was capable of delivering up to 200 l/min of viscous liquid to simulate the
flow of high rheological tailings was used, however, the maximum test flow rate in the pipe is
130 l/min. The pump has a variable frequency drive which helps to cover different flow rate.

Measurement Boxes

The system has three measuring boxes that consist of two or more chambers (Martinson, L.). The first
and second chambers were designed to receive the liquid, dissipate the incoming flow energy and
liberate the air. The last chamber was designed to measure the flow, using a 45° sharp weir which
was specially designed for this application in order to have a controlled flow measurement system.
To establish the curve that relates the fluid height (head) on the weir to the flowrate, volumetric
gauging was carried out (volume and time recording). This allowed the generation of an adjustment
curve (the final curve included the effect of test fluid viscosity).

Pieces Tested

The 3 pieces tested were made according to the specifications of the design drawings (see Figure 2).
The pieces tested are 90°, which is the angle that exists in the current discharge system, and the two

4
others at 30° and at 38°, angles which were selected to agree with a preliminary CFD study. Also,
piece 1 is rotated axially to be oriented with -28°, -35°, and -42° below horizontal.

Piece 1-90° Piece 2-30° Piece 3-38°

Figure 2 Pieces tested to improve the hydraulic discharge (flow left to right)

High viscosity liquid

The last component to describe is the liquid used for the tests. This liquid had to emulate the non-
Newtonian behaviour in the study range and be adequate to represent tailings. The viscosity was
measured during tests using a high precision rheometer and its density was 1 gr/cc. The viscosities
were verified every time the pieces were tested.

It is important to note that the liquid didn’t present any type of hazard to manipulation, as it was a
food-industry additive used to give consistency to certain types of food products.

TEST METHODOLOGY

The challenge of the test execution set in the model require the determination of how much flowrate
goes through a spigot (discharge), depending on a certain input flow.

Definitions

Three flows have been defined and named inflow, outflow, and through-flow.

 Inflow (IF) comes through the main pipe before encountering the discharge.
 Outflow (OF) exits laterally through the spigot.
 Through-flow (TF) continues through the main pipe and does not flow out the spigot. In
other words, it is the IF minus the OF.

The flow rates are measured in the three measuring boxes that have a sharp weir of 45°.

5
Test Matrix

The tests execution required a simple methodology to be followed in order to acquire good quality
data to be analysed and finally conclusive. Table 1 shows all the alternatives analysed and quantity
of flow data measured.

Table 1 Test Matrix

Alternative 1 2 3 4 5 6

Piece Piece 1 Piece 2 Piece 3 Piece 1 Piece 1 Piece 1

Rotation angle 0 0 0 28 35 42

Flowrate six flow rates five flow rates

tested*
between 30 l/min to 130 l/min between 40 l/min to 130 l/min

*5 or 6 different flow rates were tested

Graphic Methodology

Different inflows were tested with the help of the pump's variable frequency drive, covering a flow
rate range to between 30 l/min and 130 l/min in the model (3,500 m3/h to 15,000 m3/h in the prototype).
For each inflow set the weir height was measured in the measuring boxes (inflow, outflow and
through-flow), thus the weir height is related to a specific flowrate. With the recorded data, a graph
for each alternative is obtained.

It should be noted that the model did not include the four discharge spigots that exist in the actual
system because the graphical methodology allows the evaluation of the three discharges in the range

IF1 =85 L/min

Model Flows
130 30
120 OF1 =11 l/min
110 25 Spigot 1
TF1 =74 l/min =IF2
100
90
Tailings Througth-flow (l/min)

20 Main pipe
Tailings Outflow (l/min)

80 TF1 OF2 =10 l/min

TF2 Spigot 2
70
15 TF2 =64 l/min =IF3
60 TF3
OF1
50
OF2
10
40 OF3 OF3 =8 l/min
30 Spigot 3
20 5 TF3 =56 l/min =OF4

10
IF3 IF2 IF1
0 0 OF4 =56 l/min
20 40 60 80 100 120 140 Spigot 4
Tailings Inflow (l/min)

6
of measurements. The through flow of the third spigot must be the outflow of the fourth spigot, as
shown in Figure 3.

Figure 3 Graphical methodology for four spigots set (piece 90°) – model flowrates

RESULTS AND DISCUSSION

Alternatives Evaluated

The flow rate obtained from the scale model testing was proportioned to the prototype scale. The
Figure 4 shows inflow vs outflow relationships for the prototype.

5,000 Alt 1 - Piece 90°

Alt 2 - Piece 30°

Tailings Outflow (m3/h)

4,000

3,000 Alt 3 - Piece 38°

2,000 Alt 4 - Piece 90° Rot 28°

1,000 Alt 5 - Piece 90° Rot 35°

0 Alt 6 - Piece 90° Rot 42°

0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000
Tailings Inflow (m3/h)

Figure 4 Inflow and Outflow by six alternatives tested - prototype flowrates

As can be clearly seen in the graph that as the angle of the piece is decreased, the outflow is increased.
When the rotation of the main pipe is further downward, the outflow is higher. Furthermore,
alternative 1, 2 and 3 show a linear trend. Alternatives 4 and 5 also show a fairly linear relationship,
while some non-linear behaviour is observed at the steepest rotation angle. Further testing is required
to verify the non-linear behaviour at the steeper rotation angles.
It is not only necessary to know how much flow is going to OF but also to find a solution with equal
flow split (25% of the inflow) to each spigot. The Figure 5 shows percentages of outflow in each spigot
considering using an identical spigot arrangement for each of the four spigots. The inflow upstream
of spigot 1 is 10,150 m3/h equivalent to 85 l/min in the model.

7
 70% 65% Alt 1 - Piece 90°
60%
Alt 2 - Piece 30°
Tailings Outflow %

50%
44%
Alt 3 - Piece 38°
40% 25% 37% 35%
31%
31% 19% 21%
27% 26% 28% Alt 4 - Piece 90° Rot 28°
30% 25% 23% 19% 21%
23%
16% 19% Alt 5 - Piece 90° Rot 35°
20% 16%
13% 11% 14%
10%
10% Alt 6 - Piece 90° Rot 42°

0%
25% optimal distribution
Spigot 1 Spigot 2 Spigot 3 Spigot 4

Figure 5 Percentage of outflow at each spigot for 6 alternatives tested (IF1=10,150 m3/h)

Firstly, the 90° tee (piece 1) results are consistent with the real situation where last spigot has the
highest flow. In fact, the graph shows that the outflow in the first three spigots varies between 10%
and 13% and last spigot has flow over 60%. This is very similar to the performance at the plant and
to CFD results of the full scale system.
Alternatives 2 to 6 provide a more equitable distribution of the flows within the four spigots of
discharge. For instance, in alternative 2 and 3 the outflow in the first three spigots varies between
14% and 31%, while in the fourth spigot the outflow corresponds to 37% and 44% of the flow,
respectively. In alternative 4 to 6 the outflow in the first three spigots is between 15% and 31%, while
in the fourth spigot the outflow corresponds to 19% and 46% of the flow.
According to the Figure 5, if the alternative are the same for each spigot, it is impossible to achieve
the equal 25% of discharge from each spigot required. Consequently, options that combined different
alternatives for each spigot are required.

8
Combination Evaluated

The following table shows six options that combined the 6 alternatives analysed.

Table 2 Combination evaluated

Combination 1 2 3 4 5 6

Spigot 1 Alt 3: 38° Alt 3:38° Alt 4: 90°-rot 28° Alt 3: 38° Alt 3: 38° Alt 4: 90°-rot 28°

Spigot 2 Alt 2: 30° Alt 6: 90°-rot 42° Alt 6:90°-rot 42° Alt 5: 90°-rot 35° Alt 4: 90°-rot 28° Alt 4: 90°-rot 28°

Spigot 3 Alt 6: 90°-rot 42° Alt 6: 90°-rot 42° Alt 6: 90°-rot 42° Alt 6: 90°-rot 42° Alt 6: 90°-rot 42° Alt 6: 90°-rot 42°

Spigot 4 Alt 1: 90° Alt 1: 90° Alt 1: 90° Alt 1: 90° Alt 1: 90° Alt 1: 90°

The graph of Figure 6 is constructed from the adjusted curve obtained from Figure 4 which represents
directly the points from the measurements made. For this reason, the estimates are not absolutely
accurate, but they predict the behaviour of the system associated with an alternate configuration and
a specific inflow value. The results obtained are showed in the Figure 6.
The best performance of the combinations evaluated are combination 1, 4, 5, and 6.

Figure 6 Percentage of outflow in each spigot by six combination evaluated (IF1=10,150 m3/h)

9
CONCLUSION

Tailings spigots discharge in gravity regime can be modelled by a hydraulic scale model, which has
demonstrated itself to be a very good predictor of the hydraulic behaviour of a spigot. Also, the model
results are consistent with CFD outcome and the real operation. Particularly, it could become a strong
decision-making tool to improve engineering designs.
For the cases analysed, it was possible to prove that, if the lateral tee angle is sharper the outflow is
increased. Additionally, if the spigot is rotated downward relative to the main pipe axis the out flow
is also is increased. According to the scale modelling result, a combination of different alternatives
for each spigot show the best performance, in order to obtain an even distribution of discharged
tailings.
It is recommended to follow with a trade-off study (TOS) in order to analyse economic and
constructive aspects for the best combinations. It is important to note that the model is a
representation of what should happen under certain conditions and not the representation of exactly
what will necessarily happen. In this sense, implementing one discharge area in the site with the best
alternative from the TOS could be a good tool for optimizing the final configuration and orientation
of the distribution spigots.

ACKNOWLEDGEMENTS

Thanks to all the people involved on this project, the client and especially to the Santiago lab team.

NOMENCLATURE
CFD computational fluid-dynamic
IF Inflow
OF Outflow
TF Through-flow
m3/h cubic meters per hour
l/min litter per minute
Fr Froude number
λ Geometric scale factor
Vm Velocity in the model
Qm Flowrate at the model
Vp Velocity in the prototype
Qp Flowrate at the prototype

10
REFERENCES

Martinson, L., Salinas, C. and Martinson, R. (2018) ‘Hydraulic scale modeling test for a tailings
transfer box‘, 5th International Seminar on Tailing Management, Tailings 2018, July 11 - 13, 2018, Santiago,
Chile.

Goosen, P.E., Cooke, R. (2007) ‘Hydraulic scale model tests for slurry handling equipment’, 17th
International Conference on Slurry Handling and Pipeline Transport, Hydrotransport 17, May 7-11, 2007,
Cape Town, South Africa.

11
Characterization of the Effect of Clay Mineralogy and
Content on the Rheological Behavior of Copper
Sulfide Tailings
Sebastián Contreras1,2, Claudia Castillo3, Christian Ihle1,3 and Gabriel Méndez2
1. Department of Mining Engineering, Universidad de Chile
2. Centre for Biotechnology and Bioengineering, Universidad de Chile
3. Advanced Mining Technology Center, Universidad de Chile

ABSTRACT
In the copper sulfide beneficiation process, recovery can be challenged by the presence of clays in the
feed. Their occurrence has shown to become increasingly important, due to the nature and ageing of
the deposits being currently mined, with subsequent requirement of finer grinding. This has implied
problems in almost every stage of the mineral processing value chain (MPVC). It has been shown
that different kind of clays have different effects on particular stages of the MPVC, suggesting that
special attention must be payed to the mineralogy. As different clay minerals may have differing
properties (e.g. shape and cation exchange capacity and size), their response under similar process
conditions (e.g. solid concentration, milling grade, ion presence and shear rate) will not necessary be
the same, and identifying how they explain the variability on the desired on key output variables of
the operation, would be central for a successful control and operational philosophy. Due to its major
impact on water recovery, in the present work we focus on tailings. Using synthetic copper sulfide
tailings, we explain the role that some clays, exemplified herein with blends of kaolinite and
bentonite, play in the modification of the rheological and physicochemical properties of the tailing’s
suspension. To decouple and characterize the contributions of (1) the particle size and average shape
of the solid fraction, (2) effect of adding the clay species to a pure silica tailing and (3) possible
synergistic interactions between both clays studied, a number of clay content scenarios have been
considered. The rheological behavior and variations of critical parameters such as apparent yield
stress, viscosity and effective packing fraction, were studied in light of existing models. The effect of
kaolinite and bentonite on the suspension rheology was quantitatively different, and it was also non-
additive, suggesting the existence of more complex interactions between the clay species, clay and
silica and/or clay and bulk fluid.

1
INTRODUCTION

As copper-rich deposits have been depleted, it has been necessary to treat lower grade ores and
greater volumes, with the aim of keeping copper production at the present rate. Greater volumes of
mineral imply a need for greater volumes of water, and its limited availability generates problems,
such as an increase of operational costs due to the incorporation of desalinization unit operations
[Cochilco 2018], or undesirable paste-like non-Newtonian behavior of the pulp when reducing the
water supply to the process.
Many operations have also reported the presence of clay within their ROM mineral, thus generating
negative impacts in the different stages of the whole mineral processing route [Connelly 2011]
[Ossandón 2018]. Clays are phyllosilicates with planar geometry and a particle size distribution
ranging from the colloidal size to ~1 µm [Lewis 2000]. They are distinguished among other colloidal
materials by their highly asymmetric and often irregular shape, by their wide distribution in size,
electrical charges that are permanent on the faces and pH-dependent on edges, high cation exchange
capacity and hydration, their ability to disarticulate and the various modes of aggregation and
formation of structures (in coagulation and/or flocculation processes) [Lagaly and Ziesmer 2003].
Different phyllosilicate minerals have different impacts on the mineral processing value chain [Wang
et al., 2015] [Ndlovu et al., 2013]. However, the contribution of each component (mineralogy, size,
concentration, surface chemistry and interaction with the liquid media) is not fully understood in a
purely rheological context. The intent of the present paper is to contribute in that direction. In the
following sections, we present the results of the first stage of a greater study aiming to characterize
the effect of clay mineralogy on the rheological behavior of clay bearing copper sulfide tailings,
exemplified herein with blends of kaolinite, bentonite and silica. A robust sampling and testing
protocol have been developed to evaluate the rheological response of the suspensions under different
conditions. Parameter fitting via solving an inverse problem was achieved using a novel strategy
[Olivera-Nappa et al. 2019], and the obtained parameters, such as yield stress, apparent viscosity,
effective packing fraction and intrinsic viscosity, were analyzed.

METHODOLOGY
The main focus of the present study is on rheology measurement. For the purpose of maximizing the
control of slurry characteristics, synthetic tailings were prepared. Particle sizes were sifted to below
#140 (105 µm)to avoid any bridging interaction within the rheometer. Synthetic tailings have two
main components, namely a coarse fraction, consisting of SiO2 under #140 (80% of the total solid
content), and a finer fraction under #200 (74 µm) accounting for the remaining solids, which can be
either:

 90% pure kaolinite (K),

 bentonite (M), containing 74% sodium montmorillonite, or
 ultra-fine silica (Q), result of a ground and pulverized 97% pure silica sand.

2
All the aforementioned materials were obtained as a dry solid from a local store. A number of
combinations of the fine phases were considered and are shown in table 1.

Table 1 Synthetic tailings composition

Tailings code Fine composition Tailings code Fine composition

KK 80% -#140 Qz, 20% K MQ 80% -#140 Qz, 10% M, 10% Q

KQ 80% -#140 Qz, 10% K, 10% Q QQ 80% -#140 Qz, 20% Q

MK 80% -#140 Qz, 10% K, 10% M SF 100% -#140 Qz

MM 80% -#140 Qz, 20% M SG 100% Q

The suspension’s physicochemical properties were analyzed via their zeta potential, measured using
a Stabino Particle Charge Mapping instrument, following the procedure proposed in [Ndlovu et al.
2014]. Rheology tests were conducted using an Anton Paar RheolabQC concentric cylinders
rheometer with a 1.5 mm gap (inner radius = 19,503 mm, outer radius 21,000 mm). While the cup is
kept fixed, the bob angular velocity has been imposed in order to achieve shear rates ranging between
1 and 400 s-1, following ISO 3219 standard. Suspensions were prepared over a wide range of solid
volume fractions. Measurements were carried out according to a stepwise ramp protocol, and the
values for curve fitting were taken as the steady state of the constant shear curves over a sufficient
long time, thus letting the suspensions approach to a structural equilibrium [Quemada 1998]. The
fitted models were the well-known Bingham, Krieger & Dougherty and Heymann models. Bingham
rheological model was fitted in the [100-400] s-1 range, as suggested in [Reyes et al 2019].

Bingham

(1)
𝜏 = 𝜏0,𝐵 + 𝜇𝐵 𝛾̇

Where 𝜏 and 𝜏0,𝐵 are the shear and yield stresses respectively, 𝜇𝐵 is the Bingham viscosity and 𝛾̇ the
shear rate.

Krieger-Dougherty

𝜙 −⌊𝜂⌋𝜙𝑚
𝜇 = 𝜇∗ (1 − ) (2)
𝜙𝑚

3
Where 𝜇 and 𝜇 ∗ are the suspension and reference viscosities respectively, 𝜙𝑚 is the effective packing
fraction, and ⌊𝜂⌋ is the intrinsic viscosity. In solid suspensions, the intrinsic viscosity is a factor that
accounts for particle shape, in the absence of interactions between particles. As a generalization, it
may be understood, in the context of an equivalent fluid approach, as the extra viscosity that is added
to the suspension because of the presence of a solid phase (and all its colloidal/surface interactions).

Heymann

−2
𝜙 (3)
𝜏0 = 𝜏∗0 ((1 − ) − 1)
𝜙𝑚

Where 𝜏0 is the yield stress, 𝜏 ∗ a fitting parameter with stress units.

RESULTS AND DISCUSSION

Zeta potential measurements are assumed as an indication of the suspension stability. The pH-zeta
potential profiles for the different materials described above were obtained following the procedure
presented in [Ndlovu et al. 2014]. Clear differences on the slope of the zeta-pH profiles are observed
for the different tailings studied (see figure 1), suggesting that even in small proportions with respect
to the total solid content, clays can modify the whole suspension stability. The range of zeta potential
favoring aggregation[Wills and Finch 2016] is represented between the dashed red lines in Figure 1.

Figure 1 Zeta potential of the suspensions at different pH

4
The rheological profiles of the suspensions were obtained at different solid concentrations -(1) in
Figure 2-. Every step of the measuring protocol was analyzed as a single 𝜏 −time profile, and the
reported values (3) were the steady state average of the measure (2), shown as triangles in Figure 2.
The Bingham model was fitted to every curve (4) and their parameters were inferred.

Figure 2 Typical rheological profile of the different samples prepared at Cp 60%

The Bingham parameters were used as input for the fitting of Heymann and Krieger-Dougherty (KD)
models. Using the Heymann model it was possible to infer the maximum packing fraction 𝜙𝑚 , and
the intrinsic viscosity ⌊𝜂⌋ using the KD model. The first parameter (𝜙𝑚 ) represents the physical
threshold where the suspension behaves as a perfect solid, with the yield stress approaching infinity
for effective volume fractions close to 𝜙𝑚 . As expected, the materials usually addressed as being
“problematic clays” exhibit a greater intrinsic viscosity as shown in Figure 3.

5
 Figure 3 Rheological behavior of different synthetic tailings suspensions

As montmorillonite is the main component of bentonite, it is expected that the latter also exhibits
swelling behavior [Norrish 1954]. Such behavior is also observed on the modification of its packing
fraction. The impact of kaolinite is not as important as that of bentonite. Nevertheless, when they are
mixed, the resulting blend modifies the effect of bentonite alone on the suspensions’ properties, as
the MK sample parameters are similar to those of the KK samples (see Figure 3). Concerning
differences between the rheology of same material suspensions, the particle size distribution seems
to be the most important factor, since polydispersity enhances the maximum packing fraction of the
suspension. For instance, our approximately bi-modal QQ tailing shows a packing fraction of 𝜙 ≈
𝑀𝐴𝑋
0.82, consistent with the theoretical maximum of 𝜙𝑚 = 0.8704 for bi-disperse suspensions where
the proportion of large particles is close to 0.75 (our case) [Dörr et al. 2013]. Although there are some
differences between SF and SG samples in terms of their packing fraction, these are better explained
as differences on the particle size distribution profile rather than as differences due to their sizes, as
the maximum packing fraction, in a non-colloidal context, should not have a dependence on size,
unless there exist variations in the mass-length relationship (shape factor) [Farr and Grot 2009].
Another way to analyze the aforementioned effects is to consider the yield stress profile, via
Heymann’s model. Figure 4 shows the fitting of the Bingham-derived yield stress to equation 3 for
the high clay bearing tailings and the silica control. It is possible to distinguish three different zones:
no. 1 corresponds to the packing fraction of the high silica SF tailing, very close to the random close
packing RCP for spheres (𝜙𝑅𝐶𝑃 = 0.637), no. 2 and 3 correspond to the packing fractions of the
kaolinite (KK, KQ & MK) and bentonite bearing tailings (MM & MQ), respectively. It is possible to
see, both in 2 and 3, that when the fine fraction is mixed with quartz the packing fraction increases.
In other words, particles can reach higher concentrations without greatly increasing the yield stress.

6
 Figure 4 Heymann model parameters for the Bingham yield stress, highlighting differences between clay-
doped tailings and the silica control

An economically interesting application of this observation is shown with an example in no. 4: at a

volumetric concentration 𝜙 = 0.36, approximately equivalent to 𝐶𝑝 = 60%, the yield stress of a high
bearing bentonite tailing (MM) could be reduced from 160 to 50 [Pa] if it is blended with quartz (MQ).
This information should be considered in the mine planning, to achieve the desired proportions just
mixing minerals from different zones (if available), hence improving the overall performance of its
processing and its safety margin. It is important to remark that the effect of the different clays on the
resulting yield stress is not additive nor proportional (as shown in Figure 4), and that kaolinite seems
to mask and override the effect of bentonite when blended in the same proportions.

CONCLUSION
It was possible to represent the rheological behavior of the studied tailings suspensions by means of
classical models, and their differences were reflected on differences in their model-derived physical
parameters, decoupling then their individual contribution to the suspension rheology. The effect of
kaolinite and bentonite was different in intensity and it was also non additive, suggesting the
existence of more complex interactions between the clay species, clay and silica and/or clay and the
bulk fluid. Both tested clays lowered the effective packing fraction of the suspensions from the control
𝑆𝐹 𝐾𝐾 𝑀𝑀
integral silica 𝜙𝑚 ≈ 0.63 to 𝜙𝑚 ≈ 0.51 (kaolinite) and 𝜙𝑚 ≈ 0.43 (bentonite). Nevertheless, when
𝐾𝑄 𝑀𝑄
the clay mineral is mixed with quartz that effect is considerably softened (𝜙𝑚 ≈ 0.54 and 𝜙𝑚 ≈ 0.46
respectively), showing that ore blending might appear as a potential solution for protecting the
integrity of the machinery and the transport system by lowering the observed pulp yield stress. The
results for silica bearing synthetic tailings were in good agreement with literature. The data
processing strategy that was used for the parameter recognition allowed an efficient and accurate
handling of uncertainties and will be included in further research. The results of our paper highlight
the need of an information bridge between the mine and the mineral processing plant, which could
enhance the overall plant performance with an adaptative operational plan.

ACKNOWLEDGEMENTS

The authors would like to thank the Laboratory for Rheology and Fluid Dynamics of the Mining
Engineering Department of Universidad de Chile (LRF), for their support in providing the equipment
and protocols required for the experiments carried out in the present research. The authors also
acknowledge support from Fondecyt Project 1160971 and INNOVA CORFO Project CSIRO Chile
10CEII-9007.

7
NOMENCLATURE

𝛾̇ Shear rate [1/s]

𝜏 Shear stress [Pa]
𝜇 Viscosity [Pa⋅s]
𝜙 Volume fraction [-]
𝐶𝑝 Solid concentration [-]
𝜏0,𝐵 Bingham’s yield stress [Pa]
𝜇𝐵 Bingham’s viscosity [Pa⋅s]
𝜇∗ Reference viscosity (Krieger-Dougherty Model) [Pa⋅s]
𝜙𝑚 Maximum effective packing fraction [-]

REFERENCES

Cochilco (2018). Consumo de agua en la minería del cobre al 2017. Technical report.

Connelly, D. (2011). High clay ores–a mineral processing nightmare. Australian Journal of Mining, 24,
28-29.

Dörr, A., Sadiki, A., & Mehdizadeh, A. (2013). A discrete model for the apparent viscosity of
polydisperse suspensions including maximum packing fraction. Journal of Rheology, 57(3), 743-765.

Elshkaki, A., Graedel, T. E., Ciacci, L., & Reck, B. K. (2016). Copper demand, supply, and associated
energy use to 2050. Global Environmental Change, 39, 305-315.

Farr, R. S., & Groot, R. D. (2009). Close packing density of polydisperse hard spheres. The Journal of
chemical physics, 131(24), 244104.

Ndlovu, B., Forbes, E., Farrokhpay, S., Becker, M., Bradshaw, D., & Deglon, D. (2014). A preliminary
rheological classification of phyllosilicate group minerals. Minerals Engineering, 55, 190-200.

Ndlovu, B., Farrokhpay, S., & Bradshaw, D. (2013). The effect of phyllosilicate minerals on mineral
processing industry. International Journal of Mineral Processing, 125, 149-156.

Norrish, K. (1954). The swelling of montmorillonite. Discussions of the Faraday society, 18, 120-134.

Olivera-Nappa., Contreras, S., Medina, D., & Conca, C. (2019). Mathematical modelling of the
glucose-insulin system for appropriate inference of physiological parameters derived from OGTT.

Ossandón, J. M. (2018). Efecto de minerales arcillosos contenidos en concentrados de cobre sobre el

comportamiento de escorias y el desgaste de refractarios durante la etapa de fusión. Master's thesis,
Universidad de Chile

Quemada, D. (1998). Rheological modelling of complex fluids. I. The concept of effective volume
fraction revisited. The European Physical Journal-Applied Physics, 1(1), 119-127.

8
Reyes, C., Álvarez, M., Ihle, C. F., Contreras, M., & Kracht, W. (2019). The influence of seawater on
magnetite tailing rheology. Minerals Engineering, 131, 363-369.

Wang, Y., Peng, Y., Nicholson, T., & Lauten, R. A. (2015). The different effects of bentonite and kaolin
on copper flotation. Applied Clay Science, 114, 48-52.

Wills, B.A., & Finch, J. A. (2016). “Dewatering,” Wills’ Miner. Process. Technol., pp. 417–438, 2016.

9
3
Seepage and Water
Management
Experiences of Geomembranes in Mining Projects
Gabriella Vaschetti1, Marco Bacchelli1, Alberto Scuero1, and Pascual Perazzo2
1. Carpi Tech, Switzerland
2. Carpi Tech, Latin America

ABSTRACT

Watertight geomembranes, and in particular thermoplastic geomembranes of special formulation,

have been adopted as water barriers in all types of water retaining dams since 1959, in new
construction and in rehabilitation. Especially in new construction, such geomembrane systems made
feasible or improved projects that would not otherwise have been feasible at acceptable costs.
Since a decade, the use of such systems has been extended to tailings dams. A paper presented at
Tailings 2018 described the concepts and a state-of-the-art system with extruded curbs and
polyvinylchloride (PVC) anchor “wings”, adopted among others at the first two stages of Las Bambas
tailings dam in Peru. This paper, through recent case histories in mining projects, amplifies the
subject by discussing the reasons for selecting a geocomposite largely used in water dams since 1976,
and by presenting other available anchorage solutions. Such solutions relate to embankment dams
raised without the extruded curbs, and to hardfill dams.
In embankment dams, the selection of the water barrier is based on assessment of materials
requirements and sources, on required quantities, on budget rates for civil works, on comparative
costings, on efficiency and serviceability. Especially in regions with high seismicity, a geomembrane
upstream water barrier can provide higher performance and safety, at competitive costs. An
anchorage system consisting of geomembrane bands embedded in trenches, to which the
geomembrane liner is heat-seamed, is in itself flexible and contributes to the overall capability of the
system to accommodate settlements.
Hardfill dams, becoming popular in recent times, also require an upstream water barrier.
Geomembrane systems are substituting concrete facings that are costlier, longer to construct, and can
be subject to cracking. An anchorage system using deep grouted anchors, which can also be retro-
fitted in case not included in the project since the design, has been used in such dams.
The paper presents the two technologies and gives an update of the Las Bambas project.

1
INTRODUCTION

The use of thermoplastic composite geomembranes of special formulation, after sixty years of
applications, is an established technology in waterproofing of new and existing water dams. In
mining projects, they can be used to impound tailings, or for water-retaining dams for mining use.
Objective of this paper is to describe the characteristics and experience of such composite
geomembranes and of recent anchorage techniques used in mining projects.
When adequately designed, geomembrane water barriers increase safety because they avoid in the
long term the uncontrolled presence of water in the dam and accommodate settlements/differential
movements that would exceed the resistance of other types of water barriers. In mining projects, the
geomembrane is exposed to the tailings or to the water, therefore an appropriate geomembrane, a
sound anchorage system, and adequate quality control, are essential for good performance of such
water barriers. The case histories of the paper intend to present the main issues related to
geomembrane selection and the face anchorage systems that are available in addition to the one
already discussed at Tailings 2018 (Bacchelli et al., 2018), which was related to embankments formed
with the Itá method (upstream extruded porous concrete curbs). The face anchorage systems
presented here depend on the type of dam and method of construction: in embankment dams
constructed without curbs, face anchorage can be provided by embedding PVC anchor bands in
trenches formed at the upstream face and by heat-seaming the waterproofing liner to the anchor
bands, while in hardfill dams face anchorage can be provided by deep anchors driven into the hardfill
and grouted. An update of the present status of the 148 meters high Las Bambas tailings dam is also
presented.

EMBANKMENT DAM AND ANCHORAGE BY TRENCHES

The first example is related to a water dam that is part of the expansion and optimization project of
an open pit coal mine. The designer of the project identified a potential surplus of mine water and
the resultant need for a 2 million cubic meters water dam within a broad, natural valley to the north
of the existing mine. The purpose of the dam is to provide water storage for dewatering of open cut
coal mining pits and replacement of other water storages on the mine site that will in the future be
removed due to pit expansion. The dam is equipped with a controlled discharge system. ‘Clean’ and
‘dirty’ water will be separated by diverting run-off water from the catchment around the dam, thus
reducing the required capacity of the dam and the risk to the downstream environment.
Two options were considered for the design, i.e. a zoned earthfill dam comprising clay and lime-
stabilized clay, and a zoned earthfill embankment fitted with an upstream waterproofing
geomembrane sealing system. Preliminary stability analyses were performed, based on the criteria
of static stability of the embankment upstream face under “rapid drawdown”, seismic stability under
earthquake loading, and foundation seepage. Comparative costings were developed based on

2
assessment of materials requirements and sources, quantities, budget rates for civil works, efficiency
and serviceability. The option with an upstream geomembrane sealing system was selected.

Waterproofing liner selection

Especially in regions with high seismicity, the selection of the correct type of geomembrane is crucial
for performance. The requirements derive from the loads that the geomembrane will have to sustain:
 Loads applied by the subgrade: puncture and burst over irregular subgrade, subsidence,
displacements between deformable embankments and concrete structures
 Loads applied by construction activities, generally in the category of puncture and burst
 Loads applied by service conditions: environmental aggression, impact by floating debris,
ice, boats etc., action of waves and wind, backpressure due to water table.

The characteristics required for a geomembrane in any type of dam are low hydraulic conductivity
(i.e. watertightness), good mechanical properties (tensile behavior in particular), good endurance
properties (UV resistance, oxidation resistance, thermal behavior), workability, easy and reliable
welding, minimization of folds.
As far as watertightness is concerned, all modern synthetic geomembranes are essentially watertight.
The permeability of the geomembranes presented by the paper is as low as 6.25·10 -14 cm/s.
Appropriate tensile behavior plays a major role in future performance. The flexibility and elongation
capability of a polymeric geomembrane are illustrated by its tension-elongation curve. In dams and
reservoirs application, only the monotonically increasing part of the curve should be considered, and
not the tension-elongation curve until failure, which does not represent the real behavior of the
material in the field, as will be discussed below.
The two geomembranes most widely used respectively in tailings dams and in water dams are
compared in Figure 1: a 2.5 mm thick high density polyethylene (HDPE) geomembrane (in black)
and a composite geomembrane (geocomposite) formed by a 2.5 mm special compound of
polyvinylchloride (PVC) plasticized with high molecular weight branched plasticizers (in red). For
HDPE, the curve is limited to the range of admissible strains in the field, i.e. an allowable elongation
up to the yield point, which is at about 12% elongation; beyond this point the geomembrane thins
down locally and elongates like gum, presenting a plastic elongation under essentially constant
tension up to the elongation at break (700%), and ceases to function from a mechanical standpoint.
According to ICOLD, the International Commission on Large Dams (ICOLD, 2010), “Mainly for
HDPE, for a stress higher than the yield point, significant partially irreversible deformations (creep) occur after
the stress has ceased.” Actually, since all geomembranes are scratched in the field, and scratching
causes a reduction of thickness, the admissible elongation should be lower (Giroud, 1984).
Therefore, HDPE geomembranes should be used only well below the yield elongation, and with a
substantial factor of safety. International literature (Seeger & Muller, 1996, Peggs et al., 2005, and an
article to be published by J. P. Giroud), indicates that to be on the safe side the allowable elongation
of HDPE geomembranes should not exceed 3 to 5%, even lower for textured geomembranes. For

3
elongations greater than 3% the “creep” phenomenon is important and cannot be neglected. Low
linear density polyethylene (LLDPE) geomembranes have a tension-elongation curve quite like that
of HDPE geomembranes. The co-energy concept (Giroud, 2005), a powerful tool to rank
geomembranes according to their ability to withstand differential settlements, clearly shows the
lower factor of safety that is attained when using HDPE and LLDPE geomembranes.

Figure 1 Tension-elongation curve of an HDPE geomembrane (black) in the admissible range (up to yield) and
of a SIBELON® geocomposite (red) before breaking of the backing geotextile (after which the geomembrane
will still be functional)

The diagram of the geocomposite in Figure 1 presents the initial part up to breaking of the backing
geotextile, after which the geomembrane will present the characteristic behavior until failure. The
geocomposite shows a monotonically increasing tension-elongation diagram, and no yield.
Allowable elongation is > 70%, which means higher safety in respect to settlements and to differential
movements between the embankment and concrete structures intercepting it.
HDPE geomembranes can be subject to stress cracking because of their weight and of sliding on the
slopes. Stress cracking, responsible for numerous failures of installations with HDPE geomembranes,
is related to the dimensional stability, creep and crystallinity characteristic of HDPE. Workability and
seamability are also crucial for behavior. An HDPE geomembrane, with higher coefficient of thermal
expansion and stiffness, will be more difficult to place flat, especially under high temperature
excursions. Higher folds will form: the height of the folds in an HDPE geomembrane is at least 4
times larger than in a SIBELON® geocomposite. High folds make proper seaming more difficult and
time consuming and amplify the stresses due to the dynamic action of water and wind. Failure of

4
HDPE geomembranes frequently occurs near the seams, due to weakening caused by excessive
temperature or pressure during seaming, or in case of elongation caused by wind suction or thermal
expansion generating excess of tension and reduction of thickness.
HDPE geomembranes are generally placed on a separate anti-puncture geotextile. In geocomposites
the geotextile that is part of the liner increases the friction angle of the liner, thus allowing placement
on steeper slopes with consequent reduction of fill material, and reducing installation times (only one
layer to be placed).
Concerning durability, such geocomposites had been tested by independent laboratories and proven
by successful field installations in challenging conditions. Tests on specimens extracted from
geomembranes installed > 30 years ago (oldest installation in 1976) in the Alps at altitudes above 2000
meters showed excellent residual mechanical properties and little degradation (Cazzuffi, 2013).
Summarizing, unless chemical resistance against particularly aggressive components should require
using a PE geomembrane, such a geocomposite is more technically performing and provides a higher
factor of safety. The slightly higher cost of the material is largely compensated by the technical and
by the construction advantages (anti-puncture geotextile bonded to the geomembrane, quicker
placement, quicker and easier welding). The material had successful references in mining projects:
Sar Cheshmeh in Iran (Scuero and Vaschetti, 2009), and Las Bambas in Peru.

Face anchorage

In embankment dams the technologies for anchoring the waterproofing liner depend on the
procedure of construction. If the embankment is raised placing the fill against an upstream layer of
extruded lean concrete curbs, the face anchorage system is the one already mentioned (Bacchelli et
al., 2018). When the upstream face of the embankment has a granular finishing layer like in this case,
anchorage can be provided by heat-seaming the waterproofing geocomposite on an anchor line made
by the same material. The anchor line is constructed by excavating anchor trenches in the finishing
layer and embedding inside them a geocomposite band having enough width to allow a flap exiting
from the trench, which is then backfilled with porous ballast material (Figure 2 at left). One
geocomposite sheet is then placed and heat-seamed at its selvedge to the flap, followed by placement
of the adjacent geocomposite sheet, which is in turn heat-seamed to the first installed sheet (Figure 2
at right). The strength and continuity of the seams, checked by adequate quality control procedures,
provides sound anchorage to the anchor bands, and watertightness of the seams adjoining adjacent
geocomposite sheets.
The technology had a pioneer application at Kouhrang head pond in Iran (Jafarzadeh et al., 2005),
where it was selected among three variants (zoned embankment + clay core, homogeneous
embankment + selected shell material, and homogeneous embankment + upstream geomembrane
system), because deemed “... the best solution for project target… it can save up to 50 % in costs of pond
construction”. The system was adopted in the lower part of Runcu rockfill dam in Romania (2015),
and at some slopes and invert of the 18 Water Saving Basins of the Panama Canal Expansion (2015).

5
 Figure 2 Face anchorage system by SIBELON® CNT geocomposite bands embedded in trenches

Project details
The dam is 460 m long and 17 m high, with a 10 m wide crest. The upstream slope has a 2H:1V
inclination. The bulk of the embankment consists of conditioned and compacted Zone 3A and 3B
sandstone earthfill. The base layer for the geomembrane is a Zone 1A lime-stabilized clay layer. In
such layer were excavated vertical anchor trenches at 6 meters spacing. SIBELON® CNT 4400
geocomposite bands (the same material used for the waterproofing liner described below) were
embedded in the trenches leaving a protruding flap (Figure 3 at left) for face anchorage of the
waterproofing liner. The trenches were then backfilled with porous concrete.
A face drainage system is provided between the waterproofing liner and the clay surface. The system
consists of a drainage geocomposite (a triaxial drainage net heat-bonded at fabrication to a filter
geotextile) placed at the upstream face (Figure 3 at right) and connected to a slotted collector pipe at
the upstream toe, in turn connected to an outlet pipe discharging into a sump downstream of the
embankment. The drainage system will ensure that, in the unlikely event of a concentrated leak,
seepage water will be safely drained away to avoid erosion of the clay base layer.

Figure 3 At left SIBELON® CNT geocomposite bands embedded in trenches to act as face anchorage system
for the waterproofing geocomposite, at right the drainage geocomposite, placed between the bands, with the
filtering geotextile facing the clay surface to avoid clogging of the drainage geonet

The waterproofing liner is the geocomposite SIBELON® CNT 4400, a 3 mm thick geomembrane

6
formed by a special compound of polyvinylchloride plasticized with high molecular weight
branched plasticizers, and heat-bonded during fabrication to a 500 g/m2 nonwoven, needle-punched,
continuous filament polypropylene geotextile. Figure 4 at left shows the geocomposite sheets being
seamed to the anchorage line. At the upstream toe the geocomposite is extended into a nominal 4 m
deep trench (Figure 4 at right), which is backfilled with clay to provide the dual function of perimeter
anchorage and foundation seepage minimization. At crest, the geocomposite is placed in a trench
and anchored by a mass concrete anchor beam. At abutments, a shotcrete cover sprayed on a mesh
protects the geocomposite against fire. Waterproofing works started in May 2016 and were
completed in July 2016, in total 12,149 m2 of geocomposite installed in 6 weeks. Due to a severe
drought ever since the dam was built the reservoir has not yet been filled.

Figure 4 At left SIBELON® CNT geocomposite sheets seaming, at right the geocomposite placed in the bottom
peripheral trench before backfilling

HARDFILL DAMS AND DEEP GROUTED ANCHORS: AMBARAU, DR OF CONGO

2017

Face Symmetrical Hardfill Dams (FSHD) belong to the family of Cemented Material Dams (CMD),
present similar technical and cost advantages, and in recent times are gaining increasing
appreciation. They use local aggregates with low cement content (typically 50-80 kg/m3) and no joints.
The low-strength material is sufficient for stability requirements but, being semi-pervious, requires
an upstream facing. Until 2014, the upstream facings of FSHDs were typically made with
conventional concrete (slabs or grout enriched hardfill) and provided with waterstops. Concrete
facings require heavy and costly equipment, long construction time, are expensive, and frequently
require post construction maintenance. Geomembrane facings are a viable alternative.

Face anchorage and project details

The available technologies are the system with geocomposite wings on curbs or, if construction of
the dam has already started without curbs, deep grouted anchors, first adopted for rehabilitation of
concrete facings that had deteriorated to such a point that they could no more provide enough pull-
out resistance for more “classical” anchor systems. Examples are canals in Germany and Austria, and

7
more recently Filiatrinos hardfill dam in Greece. With a slightly different configuration, deep anchors
gave excellent performance at Vaité earthfill dam in French Polynesia (2011), where they were
designed to resist wind hurricane conditions of 204 km/h.
The system foresees driving into the completed hardfill (or embankment) deep anchors bars at site-
specific patterns, grouted with special fittings to provide appropriate pull-out strength. The
SIBELON® geocomposite is then deployed and punched over the anchors. A stainless-steel washer
distributing the stresses on the geocomposite, an anti-puncture geotextile, and a waterproofing
geomembrane washer, complete the anchor (Figure 5 at left). This patented system was installed in
the first stage of Ambarau hardfill dam that will produce power for the Kibali gold mine, owned by
Randgold Resources, AngloGold Ashanti and Sokimo, and designed by ARQ.
Ambarau is 21 m high and 255 m long, has 1V: 0.75H symmetrical slopes, and was originally designed
with a water barrier consisting of concrete slabs with waterstops, placed over precast elements of
grout enriched RCC forming a stepped facing and then filled with porous concrete to provide an
inclined slope (Figure 5 at right). When construction of the dam had already started, the designers
decided to assess a new design for the upstream water barrier, which could be more reliable, quicker
to construct allowing earlier impoundment, more “forgiving” in respect to defects in construction,
and less expensive. The selected waterproofing system was the system with geocomposite anchor
wings. However, when the design was modified to a geomembrane facing instead of a concrete
facing, the right abutment of the dam had already been completed with the finishing layer shown in
Figure 5 at right. Consequently, another face anchorage system that could be installed on the already
completed part at the right abutment was needed.

Figure 5 Ambarau dam: At left the patented deep anchor system, at right the stepped facing of the hardfill
filled with porous concrete during the first phase of construction

8
Geomembrane systems are very flexible also in the sense that they can be customized to be retro-
fitted when construction of the dam has already started with a different system. It was thus possible
to adopt for the completed right abutment a deep grouted anchors system, and for the central part
and left abutment still to be constructed the anchor wings + curbs system. Changing the concrete
facing to a synthetic flexible geomembrane facing, anchored with two different systems, did not
require any alteration of the original design of the dam. At the right abutment, the SIBELON® CNT
4400 geocomposite liner, the same geocomposite used in the afore mentioned dam, was anchored
with the deep grouted anchors system described. The porous concrete that fills the stepped facing
forms a continuous drainage layer under the geocomposite, to avoid uplift build-up behind the
geocomposite by collecting and conveying to discharge any infiltration water.

Figure 6 Ambarau dam: At left and middle the patented deep anchor system, at right view of the dam

The waterproofing system is extended to cover the plinth and fastened with a mechanical seal
watertight against water in pressure and tested to withstand water heads up to 250 m. In the top
vertical wall that was under construction, and at the central part of the dam and at the left abutment
that were still to be constructed, the system with anchor wings was adopted. In total, 2,770 m2
installed.

UPDATING ON LAS BAMBAS TAILINGS DAM, PERU

As discussed in a previous conference (Bacchelli et al., 2018) adopting a PVC geocomposite instead
of an HDPE geomembrane allowed modifying the upstream slope from 1V:2H to 1V:1.7H, with large
reduction of materials and of construction times and costs. The same system adopted in the first two
stages of the project, i.e. exposed geocomposite anchored on PVC anchor wings embedded in
extruded porous concrete curbs, is being adopted for the stages that have followed. Las Bambas dam
is now in Phase 3. At present the embankment is 148 m high, reaching 4080.2 m elevation. The
waterproofing system of Phase 3 accounts for a total of 214,173 m2 of geocomposite, which added to
the square meters installed in the previous phases gives a total of 564,349 m2. Figure 7 shows
installation of Phase 3 geocomposite on the anchor wings on May 31, 2018. Stage 4 is expected to

9
reach elevation 4110 m, using the same system. Up to now the dam is behaving as planned, and more
stages should come in the future.

Figure 7 Las Bambas dam on May 31, 2018

RESULTS AND DISCUSSION

The geocomposite presented has characteristics that make it suitable and advantageous in tailings
dams. The tensile properties highlighted in the paper demonstrate its technical superiority in respect
to long-term resistance to loads and to the benefits provided at construction. References of more than
a decade in mining projects, and successful > 40 years’ experience in water dams, can be a basis for
increasing confidence of the designers in these materials, especially in times when high performance
can be of essence to improve safety of tailings dams. Additionally, only this liner can provide such a
wide range of face anchorage systems. Anchorage by trenches requires a very flexible material, not
prone to folds, and that can be seamed to the flaps with a reliable method – i.e. heat-seaming and not
extrusion-seaming. The watertightness and long-term behavior of anchorage by deep anchors relies
on resistance to stresses concentrated at the stainless-steel washers, which will be higher for a
material with higher admissible elongation, and on the watertightness at the geomembrane washer,
which again asks for heat-seaming and not extrusion-seaming.

CONCLUSIONS

The discussed materials and anchorage techniques contribute to the overall capability of exposed
geomembrane systems to accommodate settlements and differential movements that may occur at
the dams. They allow completing in shorter times and at lower costs the water barrier that is essential
for embankment dams and hardfill dams. New projects incorporating this technology are under
design.

10
REFERENCES

Bacchelli, M., Scuero, A., Vaschetti, G., and Scarella, M. (2018) ‘Geomembranes as impervious element
to prevent accidents in tailings storage’, Tailings 2018, 5th International Seminar on Tailings management,
Santiago July 11-13, 2018.

Cazzuffi, D. (2013) ‘Long-time behaviour of exposed geomembranes used for the upstream face
rehabilitation of concrete and masonry dams’, 9th ICOLD European Club Symposium 2013, Venice April
10-12, 2013.Giroud, J. P. (1984) ‘Analysis of stresses and elongations in geomembranes’, International
Conference on Geomembranes, Vol. 2, Denver, June 1984, pp. 481-486.

Giroud, J. P (2005) ‘Quantification of geosynthetic behavior’, Geosynthetics International, Special Issue

on Giroud Lectures, Vol. 12, No. 1, pp. 2-27.

ICOLD, the International Commission on Large Dams. (2010) Bulletin 135: Geomembrane Sealing
Systems for Dams – Design principles and review of experience, ICOLD, Paris.

Jafarzadeh, F., Javaheri, H., Heidari, T. (2005) ‘Kouhrang HPP. head pond, first hydraulic
geomembrane project in Iran: comparison with other variants’, ICOLD 73rd Annual Meeting, Tehran
May 1-6, 2005.

Peggs, I.D. (2005) ‘Maximum Allowable Strains in HDPE, LLDPE, and PP Geomembranes’,
GeoFrontiers 2005, IFAI Publisher, Roseville, USA.

Scuero, A. and Vaschetti, G. (2009) ‘Unconventional Design in Dam Raising: Sar Cheshmeh Tailings
Dam’, ICOLD 23rd Congress, Brasilia May 25-29, 2009.

Seeger, S, and Müller, W. (1996) ‘Requirement and Testing of Protective Layer Systems for
Geomembranes’, Geotextiles and Geomembranes, Volume 14, 365-376.

11
Integrated Methodology and Tools to Estimate and
Manage Infiltration in Tailings Storage Facilities
Agustín Álvarez1 and Jorge Guarda1
Ausenco, Chile

ABSTRACT

In recent years, changes to the legal framework surrounding mining facilities and their contact waters
have placed new emphasis in adequately assessing the efficacy and design of seepage barriers and
contact water management works to minimize and manage the potential environmental impacts.
These generally stem from stability issues from rising pore water pressures and seepage into the
underlying or downstream hydrogeological systems. A detailed and integrated analysis is rarely
carried out by companies due to the technical complexities of the problem, and when done, it is
generally isolated from interactions with other aspects of the operation.
An interdisciplinary top-down approach would allow for the gradual reduction of uncertainty
derived from limitations of the conceptual models and numerical imprecision in incremental
modeling iterative steps. As more data and results from initial more straightforward modeling efforts
become available during the stages in the life cycle of the facilities, complexity is added to the
conceptual and numerical models used in the reiterated impact assessments. The optimization is
aimed at variables such as pore pressure distributions within the tailings and dam structures,
seepage, characterization of groundwater flows, interaction with surface waters, and the site-wide
water balance. All these are fundamental in the efficient design of drainage structures, hydraulic
barriers, and the physical and chemical stability risk analyses of the facilities. As a methodological
framework, the proposed methodology integrates the analysis of the bounding hydrological system
and the climatic and meteorological drivers, with the operational constraints and site-wide water
balance analyses. This is achieved with an encompassing watershed model to estimate net yields into
the facilities, detailed modeling of surface layers using 1-D and 2-D profiles including the interactions
with the unsaturated zone, and the subsequent modeling of groundwater flows using more complex
2-D or 3-D modeling. The purpose of this paper is to show the advantages that this approach brings
to better ascertain environmental and operational risks related to tailings storage facilities.

1
INTRODUCTION

In recent years, changes to the legal framework surrounding the Environmental Impact Assessment
System (SEIA) in Chile concerning mining facilities and their contact waters, have established the
compulsory use of hydraulic seepage barriers and other contact water works to prevent the
occurrence of any significant environmental impacts downstream of these installations, placing an
added emphasis on their adequate design and assessment of their efficacy. Although this involves
defining the geological and hydrogeological system with a thorough site investigation and
understanding of the constitutive behavior of the materials and soils through field testing and
laboratory analysis, the most prominent feature in this conceptualization and integral part of the
engineering process is the numerical mathematical modeling. Just like Burland in 1996, we base this
definition on the idea that modeling is a process and not a prediction or forecast per se. The purpose
of this process is not only to ascertain the effects of the installations and the hydrologic conditions,
but also to be used as an interpretive tool to better understand the interactions or impacts of the
Mining Facilities on the surrounding environment.
Providing these answers requires much specific hydrogeologic information and analyses that go
beyond the capabilities of initial conceptual and analytical mathematical models, although these are
still an essential part of the Conceptual Modeling Process. The complexities of the problem lead to
modeling efforts focused on the use of deterministic, process-based numerical models. As computer
power has become cheaper and readily available, there are a plethora of computer codes available
today, many of which have become industry standards, depending on the specifics of the problem.
These process-based numerical models consist of governing equations that describe the physical
processes within the problem domain; boundary conditions that specify heads or flows along the
boundaries of the problem, and for time-dependent problems, initial conditions that specify flows or
heads within the problem domain at the beginning of the simulation.
The process of numerical modeling enhances engineering judgment and provides a basis for
understanding complicated physical processes from stability issues from rising pore water pressures
to the seepage into the underlying or downstream hydrogeological systems. Companies rarely carry
out detailed and integrated analyses due to the technical complexities of the problem, and when they
do, it is generally in isolation from interactions with other aspects of the operation.
The objective of this document is accordingly to provide a workflow for the modeling efforts using
an interdisciplinary top-down approach. These are undertaken to identify and assess the impacts of
mining project’s Tailings Storage Facilities (TSF) in Chile, with specific emphasis on integrating the
analysis of the bounding hydrological system and the climatic and meteorological drivers, with the
operational constraints and site-wide water balance, seepage analyses and groundwater flows. We
also provide some key definitions and concepts to guide the selection of the appropriate numerical
model characteristics and code from the pool of commercially available software.

2
METHODOLOGY

The typical sequence involved in a modeling workflow has been covered extensively in the literature
regarding development of numerical models for all aspects of mine operation and facilities. These
promote an approach to the model development that is underpinned by a progression through a
series of interdependent stages with frequent feedback loops to earlier stages that account for
complex independent and interrelated processes. These are shown in Figure 1.

Planning

Conceptualization

Data collection/analysis

Model construction

Model testing/calibration

Model Predictive Simulations

Model Uncertainty Analysis

Figure 1 Typical Model development workflow

Details for most of these steps are well known and don’t need much explanation besides the subtleties
derived from the specific application to Tailings Facilities. The steps in a typical workflow for a
modeling effort described in more detail in the following paragraphs.

Planning

In the Planning stage, the modelers and key stakeholders must define the Purpose of the model,
aimed at answering a specific set of questions. Although constrained by the initial quality and
availability of data, the purpose or intended use of the model is the primary factor when deciding
the level of simplifications and assumptions that define the Conceptual Model, and subsequently
determine the characteristics of the numerical model, code selection and model design.
In the case of TSF applications, typical model purposes include:

3
  Evaluation of seepage rates from TSF into the soil;
 Improving hydrogeological understanding (synthesis of data);
 Designing practical solutions to meet specified goals (engineering design);
 Optimizing designs for economic efficiency and account for environmental effects
(optimization of the hydraulic barrier);
 Evaluating impacts of changes in rheology, mining plan, TSF recharge, discharge and
aquifer storage processes (water resources assessment);
 Predicting impacts of alternative hydrological or development scenarios (to assist
decision-making);
 Quantifying the sustainable pumping and injection rates of the hydraulic barrier;
 Sensitivity and uncertainty analysis (to guide data collection efforts, refinements or
simplifications of the numerical model, revisiting the conceptualization, and to guide
risk-based decision-making);

Conceptual Model

Development of a valid conceptual model is the most crucial step in a computer modeling study
(Middlemis, 2000). However, the conceptual model should be parsimonious, such that the model is
as simple as possible while retaining enough complexity to represent the physical elements of the
system adequately and to reproduce system behavior. The level of detail within the conceptual model
should be chosen, based on the modeling objectives, the availability of quality data, knowledge of
the hydrological and hydrogeological systems and their complexity. Alternative conceptual models
should be considered to explore the significance of the uncertainty associated with different views of
how the system operates. The conceptual model should be developed based on observation,
measurement, and interpretation wherever possible. Quality-assured data should be used to improve
confidence in the conceptual model. The hydrogeological domain should be conceptualized to be
large enough to cover the location of the critical stresses on the groundwater system (both the current
locations and those in the foreseeable future) and the area influenced or impacted by those stresses.
It should also be large enough to adequately capture the processes controlling groundwater behavior
in the study area. There should be an ongoing process of refinement and feedback between
conceptualization, model design, and model calibration to allow revisions and refinements to the
conceptual model over time. Some of the main elements that are addressed in the Conceptual
Modeling Stage are the following:
 Boundaries/Hydrogeological domain, flow direction, and sources and sinks
 Hydrological characterization and domain
 Hydrostratigraphy and hydrogeological properties
 Rheological characterization of the TSF material and its geomechanical and
hydrogeological properties
 Surface water (SW) - groundwater (GW) interaction
 Stresses like evapotranspiration and recharge (imposed by climate), or anthropogenic in
nature like drains, reservoirs, or groundwater extractions via pumping

4
  Groundwater budget components and preliminary water balance
 Ancillary information

Construction of the Numerical Model – The Top-Down Approach

A top-down approach (also known as stepwise design and in some cases used as a synonym of
decomposition) is essentially the breaking down of a system to gain insight into its compositional
sub-systems in a reverse engineering fashion. In a top-down approach an overview of the system is
formulated, specifying, but not detailing, any first-level subsystems. Each subsystem is then refined
in yet greater detail, sometimes in many additional subsystem levels, until the entire specification is
reduced to base elements. A top-down model is often specified with the assistance of "black boxes",
which makes it easier to manipulate. However, black boxes may fail to clarify elementary
mechanisms or be detailed enough to realistically validate the model. Top-down approach starts with
the big picture and breaks it down from there into smaller segments.
The size, discretization, and the dimensionality of the model domain should be chosen to reflect the
modeling objectives, conceptual model and target confidence level classification. Spatial
discretization of the model domain should start relatively coarse and be reduced iteratively so that
model run times can be kept reasonably short but also provide enough refinement to be able to
adequately represent the problem geometry, including the layout of proposed developments and the
processes of importance. The model is initially evaluated in steady-state, and if the temporal variation
is essential in either the groundwater stresses to be modeled or the model results being sought,
transient simulations are required. Initial conditions in a transient simulation are obtained, wherever
possible, from the previous model runs (e.g., a steady state solution).
One of the critical challenges in modeling GW-SW interactions is the significant time-scale differences
between surface water and groundwater processes. The groundwater processes underlying and
within the TSF can be orders of magnitude slower than the surface water movements, especially if
the connecting processes within the TSF involve variably saturated phenomena. The responses of the
groundwater systems to hydrological and management drivers such as climate variability (spatially
and temporally), land use change, and groundwater extractions can be very damped and lagged.
Hence, an essential requirement in modeling GW-SW interactions in systems involving TSFs is to
account for these time lags in the recharge and infiltration processes.
Numerical methods most commonly used in groundwater modeling today are the finite-difference
(FD), finite-element (FE) and the control volume finite-difference (CVFD) methods. In the FD method,
nodes are located in 3D space using indices (i,j,k) to assign relative locations within a rectangular
grid. In the FE method, the locations of nodes are designated using spatial coordinates (x,y,z) in a
finite-element mesh. In a CVFD model, spatial discretization includes a combination of cells of any
shape, and the spatial discretization can be different even in each vertical layer.
Traditionally, modeling efforts involving TSFs opted for the FE method (e.g., FEFLOW) given its
advantages over the more rigid FD grids (e.g., MODFLOW) as they can more efficiently detail the

5
complex geometries of the TSF elements while being able to maintain coarser meshes for the
surrounding and underlying groundwater system. However, the recently developed codes like
MODFLOW-USG (Panday et al., 2013) and MODFLOW 6 (Langevin et al., 2017) use CVFD to make
a version of the FD code MODFLOW that supports unstructured grids (USG) allowing much of the
same flexibility in spatial discretization as the FE method while maintaining a more straightforward
numerical solution scheme (i.e., much shorter run times).
Still, the most widely used code for solving groundwater flow problems currently is the FD code
MODFLOW by the U.S. Geological Survey (USGS). Given the use of MODFLOW, complexities
involving the TSF interactions have received a variety of treatments, ranging from including them as
simple head and/or flow boundary conditions to modeling them explicitly with either saturated or
variably saturated flow approaches (e.g., using MODFLOW complementary packages like
SURFACT). This is in part due to MODFLOW being freely available with detailed documentation,
although the most popular Graphical User Interfaces (GUI) like Groundwater Vistas and Visual
MODFLOW are proprietary. Although no single FE code has attained the popularity of MODFLOW,
the proprietary code FEFLOW is also widely used given its efficient meshing. But, as mentioned
before, the CVFD code MODFLOW-USG (Panday et al., 2013) has been gradually gaining ground
(specially in Chile) and is now widely supported by all GUIs. The latest addition from the USGS,
MODFLOW 6, contains most of the functionality of all previous versions of MODFLOW but it is still
very new to the modeling community, and current experience is limited.
Table 1 shows, as an example, some of the commonly used GUIs and groundwater modeling codes
for the modeling of TSF.

6
Table 1 Some commonly used groundwater modeling software in the context TSF evaluations

Spatial Dimension Type Type of flow Code

1D, 2D, 3D FE Saturated, Unsaturated, Transport, Hydrus

Heat, 1D/2D/3D

3D FD Saturated, Transport, Heat, Groundwater Vistas

CVFD Variable Density (SEWAT), & Visual MODFLOW-FLEX
Unsaturated (SURFACT) using MODFLOW,
MODFLOW-USG,
MODFLOW-SURFACT,
MT3DMS, SEWAT, etc.

2D FE FE Saturated, Unsaturated, Transport, SEEP/W, VADOSE/W

Heat, Variable Density

3D FE Saturated, Unsaturated, Transport FEFLOW

Heat, Variable Density and Fractured
Media

Integration with Hydrological Model

The domains of surface hydrological and hydrogeological systems function together as a coupled
system as losses to infiltration from precipitation and surface flows enter the hydrogeological domain
and act as a source of recharge, and evapotranspiration extracts water from the soil. Surface
hydrological modeling must be done considering also the operational domain of the mining
installations, which may go beyond the boundary of the hydrogeological domain. These interactions
can sometimes be adequately represented using boundary conditions in a groundwater-flow model
with built for purpose packages (e.g., the River, Stream, Lake, packages in MODFLOW-USG), while
in other cases it is necessary to link or couple surface hydrological models with groundwater models
dynamically.
In this paper, we propose the use of GoldSim for the integrated mine water management planning
and hydrological modeling. GoldSim is a powerful and flexible Windows-based computer program
ideal for carrying out probabilistic simulations of complex dynamic systems. The program is highly
graphical, extensible, carries out stochastic simulations (using Monte Carlo simulation) allowing the
user to directly represent the uncertainty (parameter and hydrological) of the hydrological system
and associated site-wide water balances. It can be programmed to read outputs from and write inputs
to the groundwater model.
The site-wide water balance model is constructed then to represent the existing surface water
movement, including the recharge to the groundwater component of the model. It considers the

7
rainfall, evaporation, evapotranspiration, infiltration, storage and the movement of surface water
around the site at any timescale that may be suitable to the coupling of both systems (e.g., daily,
monthly). By combining surface water model components with results from the groundwater model,
GoldSim can be extended to include the results of groundwater modeling, and therefore provide a
useful predictive tool by which to compare options for surface and groundwater management. If run
times are short, the groundwater model can be run stochastically. If run times make the use of Monte
Carlo type simulation impractical, results from the groundwater model can be used to establish
functional relationships to simulate subsystems empirically within GoldSim.
Finally, GoldSim can simulate the surface impoundments used for long-term storage of mine tailings,
including changes in the location of slurry deposition is over time to control the slope and shape of
the tailings beach as it forms, resulting in a continually changing geometry of the tailings beach and
the associated water reclaim pond. GoldSim can be used to simulate the dynamic geometry (stage-
storage-area) of a typical TSF, incorporating initial geometric configuration based on available survey
information. GoldSim can even be used to dynamically simulate the TSF as a mixed-media storage
model accounting for changing quantities of solids and water in the facility over time, simulating, for
example, the reduction in water capacity as the solids accumulate and account for entrained water in
the solids so that it adequately accounts for the pooled water.

8
Calibration and Validation of the Numerical Model

Calibration of the models is generally done through a history matching process. This refers to
matching field measurements (including at least heads and fluxes) in both steady-state and transient
simulations to the respective simulated values in previously defined targets by adjusting the model’s
parameters within reasonable conceptual ranges in sequential forward runs of the model until the
model produces an acceptable match. In its most general form history matching includes these steps:
 Select calibration targets from the set of field observations;
 Run the model using best estimates of input parameters (material property parameters and
hydrologic parameters);
 Compare simulated outputs to the targets;
 Adjust values of input parameters to obtain better fits of simulated values to targets;
 Select the model with the best fit possible given limitations on time and resources.

This process of history matching can be done manually by trial-and-error or by using software, where
an objective function, based on target and residual values, is optimized. Ample literature exists
regarding manual and automated calibration (i.e. optimization) of groundwater models where the
inverse problem is solved automating the manual trial-and-error process where properties are
estimated iteratively using statistical regression and computer algorithms. Many of these advanced
methods are available to groundwater modelers through the PEST software suite. The problem is
called “inverse” because what is known (heads and/or flows) must be inverted to find what is not
known (e.g., the parameter set).

Uncertainty Analysis

Ideally, the uncertainty analysis stage is aimed at delivering a range of model prediction scenarios
with associated likelihoods, each plausible in that it is consistent with the available information and
data (Middlemis, 2018). The uncertainty analysis will also indicate the main sources of uncertainty
and by how much the uncertainty in outcomes could be reduced by incorporating further data into
the model, guiding, in principle, future data acquisition efforts.
Uncertainty analyses range from the relatively simple scenario modeling to more complex and
advanced methods like Linear Uncertainty Analysis, nonlinear methods, Monte Carlo methods and
even multiple conceptualization methods, but these advanced uncertainty methods tend to be
computationally demanding and require a higher level of user interaction with specialized software
like PEST (Doherty, 2015).

9
SAMPLE PRACTICAL APPLICATION

A real case is analyzed in the following section, where we derive certain conclusions and lessons
learned from the implementation or lack thereof of the recommendations included in the presented
workflow. This case corresponds to a downstream sand dam (DS) type of TSF (see Figure 2) that is
emplaced in Chile’s central region. There are important environmental commitments that must be
met in relation to minimizing the impact of the TSF on the catchment’s yield downstream which has
focused the different modeling efforts over time on the operational aspects and water management
of the TSF.
Upstream
runoff Tail Reservoir

Tailings Pond
Sand Dam
Tailings

Seepage
Collection

Figure 2 Downstream sand dam (DS) type of TSF

The TSF is planned to operate for a total of 40 years, and it is projected to house in the order of 1,800
Mm³, with a sand dam height of about 260 m and a total footprint of about 19 km². The discharged
tailings have a Cp in the order of 57%.
Application of the proposed workflow was partially carried out, and the described experience does
not include analysis of the system from its inception and initial modeling efforts. Since it is rarely the
case that only one consultant to be involved in the whole lifetime of a project, we consider this to be
an especially applicable example. Previous modeling efforts included hydrological and
hydrogeological characterizations, a water balance model done in Excel for the TSF design and
subsequent MODFLOW modeling of the TSF to estimate infiltrations and the design of the hydraulic
barrier, to prevent downstream discharge of the infiltration from the tailings, contour channels to
prevent excess runoff from entering the TSF and the design of the tailings deposition scheme and
design volume of the tailings pond. The system is designed to be “isolated” from the rest of the
catchment and to recirculate water to the processing plant. The contour channels system includes a
tail reservoir redirect runoff and bypass the TSF and help mitigate its impact on the catchment’s yield.

10
Hydrogeologic Modeling

The purpose of the modeling effort was to estimate the lifetime of pumping wells located
downstream the contour channels that would pump water that was coming from subsurface flow
coming An initial hydrological characterization and catchment water balance was carried out to
estimate the recharge (as a percentage of precipitation) based on a precipitation-runoff model built
with 20 years’ worth of data from a nearby weather station and a flow monitoring station at the only
exit point of the catchment.
A groundwater flow and transport model was built using the MODFLOW-NWT and MT3DMS codes
(on the Visual Modflow platform). The grid of the model was defined by square cells of uniform size
of 50 x 50 m, in order to include the drainage network of streams and the TSF’s sand dam drains but
still maintain reasonable execution times for the model.
The resulting grid contained 32,553 active cells, distributed over three parallel layers, which were
defined from the topographic mesh and the average thickness of the hydrogeological units present
in the study. The maximum thickness of the model reached 1,810 m, considering a flat bottom at
550 masl and that it includes the basal rock as a permeable hydrogeological unit.
The surface discharge zones were represented in the numerical model by a drain-type edge condition.
The cells with a drain have been distributed in the bottom of the valleys and along the main
tributaries in order to reproduce the point base flow that occurs in the streams. The tailings deposit
was represented indirectly as a general head boundary condition (GHB) that characterizes the total
head generated on the hydrogeological system by the tailings deposit, generating the resulting
infiltration from the tailings and the tailings pond into the underlying terrain.
As it is generally the case, a stationary calibration was first done in order to represent the undisturbed
state of the groundwater system simulating the conditions of pre-TSF, and subsequently, a transient
calibration was carried out, covering a period of 12 years simulating the historical behavior of the
groundwater system with the incorporation and growth of the TSF. Calibration included the
adjustment of the average yearly recharge and hydraulic properties of the hydrogeological units to
match simulated groundwater levels to data from 22 observation wells and estimated surface base
flows. The model achieved a satisfactory RMSN of 2.4%.
An uncertainty analysis was not done, however, and although a sensitivity analysis was carried out,
it mainly focused on the transport model parameters as there was insufficient data to calibrate the
transport model and it was important to understand the uncertainty in the range of time before
contacted water would start to be pumped by the wells.
Although it was not the main objective of the model, water budget analysis of the resulting
groundwater flows shed light on the temporal evolution of several important variables: infiltration
from the tailings, changes in the water storage in the ground under the tailings, and flows into the
drainage system elements as a function of time and the total head inside the tailings, all of which
would be used in the subsequent modeling efforts to better understand the system.

11
Hydrologic modeling and water balance

Hydrologic modeling of the catchment upstream of the tailings dam was first carried out in order to
refine the operation of the tail pond that discharges fresh water from runoff downstream of the
tailings dam. This operational model was built in GoldSim as one of the main purposes was to
evaluate risks using a Monte Carlo simulation approach. This hydrologic precipitation-runoff model
served as the basis of an integrated water balance model of the TSF, where the main purpose was to
evaluate the availability of tailings pond water for recirculation back to the process plant.
The balance model shed light on the temporal variability of many of the main hydrologic flows of the
system, including principally ingress from direct precipitation and runoff and losses from
evaporation from the tailings and tailings pond. Other variables that were modeled directly where
the water losses from water retained in the tailings and evaporative losses from the construction of
the sand dam. Calibration of this model was done by the history-matching process using measured
runoff flows and the volume of water stored both in the tailings pond and the tail reservoir. Losses
from infiltration from the tailings into the underlying groundwater system were first estimated as a
function of the groundwater modeling results, but then had to be modified or corrected in order to
accommodate the measured flows in the drainage system, recirculation to the process plant and the
calibration of the pond volume.
The need to modify the head-infiltration relationships derived from the hydrologic modeling, shed
light on certain shortcomings of the hydrogeological modeling that derived from assuming a constant
recharge rate as a direct percentage of precipitation and the simplification of modeling the tailings as
a GHB ignoring the spatial variability of the recharge.
A more exhaustive sensitivity analysis was carried out using GoldSim’s built-in tools for these
purposes. Results showed that the water balance model was very sensitive to the Cp parameter of
the tailings. The water balance calibration also resulted in nonuniqueness of the parameter set, driven
by the uncertainty in the evaporative losses from the tailings and the precipitation-runoff model.
Parameter nonuniqueness can lead to predictive nonuniqueness (Doherty, 2003). This problem has
been addressed extensively in the literature in relation to groundwater modeling applications and
the use of automated parameter optimization software like PEST, but experience in conjunction with
the hydrologic simulation like the one done in GoldSim is scarce at best. The extent of model
predictive uncertainty should be investigated if management decisions are to be based on model
projections.
Another way in which model predictive uncertainty can be addressed is by the iterative use of model
results over time as the methodology proposes in this paper. The update of groundwater models is
common practice and updates should be used as an opportunity to refine our understanding of the
previous modeling effort’s results in the context of using the qualitative and quantitative uncertainty
analysis of the previous predictions to drive future modeling efforts.
Just as an example, an update of the groundwater model of the previously described TSF was done,
and this time, it used the GoldSim water balance model to generate recharge series with a monthly

12
temporal variability and for different hydrologic conditions to use in the calibrations and predictive
simulations of the updated model. The model was updated to MODFLOW-USG with a Quadtree
refinement of the grid, which allowed to keep a coarser mesh (100x100 m) in less relevant parts of the
model and refine down to 25 x 25m in the more sensitive areas like the bottom of the streams,
pumping wells and drains. Calibration was extended with years of new data and additional
monitoring wells and the model’s four layers better represented the updated hydrogeological
conceptual model. Although in this model the tailings were again not explicitly simulated (they were
incorporated again as a boundary condition), the use of the recharge series derived from the water
balance model and the posterior comparison of results from the drainage system for both models
allowed to further constrain certain parameter values in the hydrological model, thus diminishing
parameter uncertainty, and in consequence, the predictive uncertainty of both models.

FINAL COMMENTS
Although the preceding chapters have described a typical TSF layout and a modeling approach
limited to the use of a 3D FD numerical model like MODFLOW (NWT and USG) coupled to the
hydrological and water balance modeling in GoldSim, it is our belief that this general experience can
be extrapolated to any type of modeling efforts with varying degrees of detail.
The different modeling efforts over time related to a TSF must be encompassed by a coupled
hydrological model that both feeds and receives feedback from the other models. This approach
allows for the gradual reduction in the predictive uncertainty of all models by being able to constrain
the uncertainty in the parameter sets of each independent model in their respective calibration
processes. This is particularly useful when, for example, the groundwater models are complex, and
their run times make the use of more advanced automated uncertainty analysis impracticable.
Eventually, this results in the ability to better ascertain the environmental and operational risks
related to the TSF.

13
REFERENCES

Anderson et al., (2015) Applied Groundwater Modeling Simulation of Flow and Advective Transport,
Second Edition. Academic Press Elsevier, San Diego.

Barnett et al., (2012). Australian groundwater modelling guidelines. Waterlines report, National Water
Commission, Canberra.

Burland, J.B. (1996), Closing Session Discussions. In: Proceedings of the First International Conference on
Unsaturated Soils, Paris, 6-8 September 1995. Edited by E.E. Alonso and P. Delage. A.A. Balkema. Vol.
3, pp. 1562-1564.

Doherty, J. and Johnston J.M., (2003), Methodologies for Calibration and Predictive Analysis of a Watershed
Model.
J. of the American Water Resources Association (JAWRA) 39(2):251-265.

Doherty, J. (2015), Calibration and Uncertainty Analysis for Complex Environmental Models. Watermark
Numerical Computing, Brisbane, Australia.

GEOSLOPE International Ltd. (2017). Heat and mass transfer modeling with GeoStudio 2018 (Second
Edition). Calgary, Alberta, Canada.

Langevin, C.D. et al., (2017), Documentation for the MODFLOW 6 Groundwater Flow Model. U.S.
Geological Survey Techniques and Methods, Book 6, Chap. A55.

Middlemis et al., (2000) Groundwater Flow Modelling Guideline. Murray-Darling Basin Commission Nº
125, Western Australia.

Middlemis H and Peeters LJM (2018), Uncertainty analysis—Guidance for groundwater modelling within
a risk management framework. A report prepared for the Independent Expert Scientific Committee on
Coal Seam Gas and Large Coal Mining Development through the Department of the Environment
and Energy, Commonwealth of Australia.

Servicio de Evaluación Ambiental, SEA (2012), Guía para el Uso de Modelos de Aguas Subterráneas en el
SEIA.

Valenzuela, L. (2016), Design, construction, operation and the effect of fines content and permeability on the
seismic performance of tailings sand dams in Chile. Obras y Proyectos 19, 6-22

Wels, C., Mackie, D., Scibek, J., Robertson GeoConsultants Inc. & SRK Consulting (Canada) Inc. (2012)
Guidelines for Groundwater Modelling to Assess Impacts of Proposed Natural Resource Development
Activities, British Columbia Ministry of Environment, Water Protection & Sustainability Branch.

14
Mining and Reusing Tailings: How to Deal with
Dewatering at Low Permeability Media?
Eduardo Ruiz1, Jorge Bueno1, Lizardo Huamani1, Jairo Paucar1, Carlos Jimenez2 and Yuri
Gallo2
1. Amphos 21 Consulting, Peru
2. Minsur, Peru

ABSTRACT

San Rafael mine is one of the largest underground mines and rich grade tin mine located in the
southern part of Peru, in the department of Puno. It crops out at elevations between 4,500 and 5,100
m, on the flanks of the glacier-capped Quenamari Mountain. The mine is owned and operated by
Minsur S.A. and has been worked for more than 40 years. Currently, this mine produces over 10 %
of the world tin, being considered as the leading Tin producing mine in South America and the fourth
largest in the world.
As part of the future San Rafael’s expansion projects there were an evaluation of alternatives, being
selected the plan to reuse the existing tailings from the tailing storage facility (B2) troughs a process
of dewatering. This selection was considered as the most suitable and cost-effective plan because this
plan allows the physical stability of the unit and the economic viability of the tin recovery in tailings,
because of the important volume and high tin concentration.
The present paper describes the evaluations carried out in the pre-feasibility and feasibility stages,
including the hydrogeological characterization of the dam tailings, simulation of scenarios, hydraulic
tests, detailed engineering design of the dewatering plan and the construction strategy. Additionally,
the paper will describe the real-time follow-up implementation program that will allow knowing the
current state of advance of the deposit desaturation, before the execution of early works.

Finally, as a preliminary result the simulation of different scenarios supported the plan to drain the
aquifer and facilitate the mining with trucks at the lowest possible tailing’s saturation pathways.
Additionally, the simulation helped to determine the time that will take to drain the entire
dewatering of the area depending on the permeability of the zone.

1
INTRODUCTION

Modern society could not function without the products of the extractive mining industries, which
are vital components in, for example, aeroplanes, ceramics, computers, construction materials, metals
and paint. On a global scale, the mining industry also provides direct employment to over 40 million
people, and indirect support for c. 200–250 million people (Kossoff et al, 2014; Azapagic, 2004).
Nevertheless, the mining industry produces enormous volumes of waste. The amount of mine waste
produced is of the same order of magnitude as that of fundamental Earth-shaping geological
processes, some several thousands million tonnes per year (Fyfe, 1981; Förstner, 1999). This volume
is frequently deposited in tailing storage facilities called dams, which can fail if there is not an
adequate design to ensure a physical stability. The causes of failure are diverse depending if the dam
is active or inactive (Rico et al, 2008), and the phase of management (construction, operation or
closure). Between the factor there are: intrinsic as tailings granulometry (particle size distribution),
hydraulic conductivity, consolidation or infiltration; and other extrinsic, such as geology of the
bedrock, precipitation, disposal and operation of new tailings, water drainage, etc.
In this sense, the mining industry continually lives with this concern, and for decades, we have
witnessed numerous economic and environmental disasters. Examples include the collapse of the
tailing dam in Aznalcóllar mine (Spain) in April 1998 (Eriksson et al, 2000) and more recent disasters
with loss of human lives associated with stability failures, such as the recent rupture of the
Brumadinho tailings dam (Brazil) in January 2019.
One of the priorities of mining organization must be to proactively isolate the tailings from
groundwater, rivers and lakes (Kossoff et al, 2014), considering the water as a key factor in the
physical stability of the dam. Take for example, the groundwater and capillary water stored in the
unsaturated zone of tailings ponds or dams, they are so variables that directly influence in the safety
and physical stability of these mining facilities and, eventually, contribute in favour of fault
mechanism. Therefore, monitoring water levels and moisture evolution within the tailings along the
time provide fundamental parameters for both design and safety control. Water inside dikes and
slopes of tailings significantly influences the stability of these structures, reducing the safety factor,
especially in seismic risk areas where processes such as liquefaction plays a relevant role. Among
groundwater control methods, dewatering has been widely used in large underground engineering
with an important efficiency (Huang et al, 2017), not mention the economic value that the re-mining
of old tailings material can have for the important ore grades contents.
This article will present a practical case of dewatering in a tailing dam that belongs to San Rafael
mine in Peru. This dewatering process has the objective to ensure the physical stability of the dam
and to reprocess the existing tailings with economic ore grade value. No one is missing the point that
dealing with tailings dewatering can be a titanic task, due to the low permeability associated with
tailings materials, making this case as an engineering challenge.

2
METHODOLOGY

Description of site of study

The San Rafael mine exploits an unusually high grade, lode-type Sn-Cu deposit in the Eastern
Cordillera of the Peruvian Central Andes at an altitude between 4500 and 5000 meters above level of
sea. This mine belongs to Minsur mining company. Currently, this mine produces over 10 % of the
world tin, being considered as the leading Tin producing mine in South America and the fourth
largest in the world (Minsur S.A., unpubl. data, 2000, Carlin 2005).
The San Rafael mine lies within the Cordillera Carabaya which consists mainly of a thick sequence
of marine sedimentary rocks. Throughout the Phanerozoic, this region underwent convergent plate
interaction, orogeny and intermittent, mixed, mantle- and crust-derived bimodal magmatism (Clark,
1990).
As part of the productive strategy, San Rafael mine planned a project to reprocess the existing tailing
located in a tailing storage facility called B2 (Figure 1). The engineering was planned and developed
in 2013, in order to begin the operation in 2019 and reprocess the existing tailings in a new treatment
plant.

Figure 1 General view of B2 of San Rafael mine (Peru)

Tailings’ hydrogeological characterization and conceptualization of the dewatering plan

Tailings storage facilities usually present a great spatial variability of hydraulic conductivity and the
physical properties of the tailings materials that they contain. Often, a lagoon is formed away from

3
the dam wall, according to the tailing’s disposal plan. In these cases, it is very common that those
tailings with finer granulometries are settled in furthest sectors.
In order to characterize tailings to plan its draining, a wide network of piezometers were installed
and several tests were conducted, both in the laboratory and through these points. The geometric
model was based on various boreholes. That served to establish and to size the pond with regards to
resource and ore grades.
To stablish the physical properties of B2, several laboratory tests were carried out from soil samples
obtained through boreholes. These were mainly granulometry curves and tailings water
characteristics tests (relationship between soil water content and suction; see Figure 2).

Figure 2 Suction in the samples in relation to their volumetric moisture content

To determine the hydraulic properties (hydraulic conductivity and storage coefficient) of the tailings
deposited in B2, hydraulic tests were carried out during the different phases of well drilling (Lugeon
and Lefranc tests), and later, pumping tests were carried out using the wells and piezometers
installed in the area. The results of these tests were fundamental data to understand the
hydrodynamic behavior of the site and to define the most suitable dewatering system.
From the results of the different tests carried out and considering the information collected and
analyzed (geology, evolution of groundwater levels, recharge and discharge of the system, surface
runoff, hydrochemical and geochemical characteristics, etc.), the hydrodynamic performance of the
tailings was defined.
During storage facility studies, numerical modeling of the site allowed to simulate and analyze
different scenarios of drainage, according to number of wells, times and rates of pumping, sectors

4
and levels with different physical and hydraulic properties. These results were considered to design
the network of wells to drain B2 (See figure 3). The updated modeling established the need to
implement 23 pumping wells (of different diameters and depths) distributed in a convenient way
around the tailings pond. These wells are equipped with submersible pumps. According to the
strategy of dewatering management, this pumping net discharged water by means of PVC pipes and
provide the dewatering requirements.

Figure 3 Dewatering System – B2 (plant view)

In addition to the dewatering system, in the Northern sector of B2 (sector with low permeability
materials) a WellPoint drainage system was installed. In B2, this pumping system is composed of
lances or filter tips separated 2 meters. These lances have of 5 - 6 m length standing in the ground
around the area to be drained and are joined by means of corrugated coupling hoses Ø 1 ½" to a
perimeter collector that pumps water with vacuum methodology. This system is removable and can
be reinstalled in different areas as the mining operation progresses (see Figure 4).

5
 Figure 4 A pumping well (left) and a wellpoint line (right) located in B2

Construction, implementation and monitoring

During the first months of 2018 the construction of wells and implementation of equipment was
carried according to the defined dewatering system. It was planned that the wells operate in a
controlled manner according to the needs during the 9 years of material extraction. The main aim of
the system is to assure that the water level will descend and be maintained at an objective level at
every moment, such that no problems arise during the mining operation. Wellpoint system serves as
temporary support in the less permeable sectors.
In order to assure the system operation and verify that dewatering objectives are achieved in a timely
manner, a full hydrogeological monitoring of the operation is carried out, as well as daily preventive
maintenance of all equipment involved. The data collected daily allows analysing the evolution of
water levels and drawdown by sector (see Figure 5). This information allows to adjust the most
efficient pumping flow in each well and is incorporated periodically to the existing numerical model
for its updating, which allows to carry out a new estimation and forecast of drawdowns by sectors.
This methodology will anticipate any potential deviation from the original plan.

6
 Figure 5 Water level evolution (average of existing piezometers for southern and central sector) and
forecasting

RESULTS AND DISCUSSION

According to the hydrogeological model, the time required to the entire dewatering of the area
depends on the permeability of the tailings. Around 8 months will be needed in the central sector
(the most permeable formed mainly with sand and silt sizes) and around 18 months in the sectors
with fine materials of low permeability (northernmost sector and SE sector, silt and clay sizes) to be
completed dewatered.

CONCLUSION

Tailings dewatering activities are highly challenging and should be properly planned and integrated
into the design of projects that evaluate the potential for reprocessing and recovery of tailings.
Further, the dewatering of San Rafael mine’s B2 is in the initial stage, some key aspects and difficulties
were already identified to achieve an efficient drainage system. The following considerations, among
others, should be considered for similar projects:

7
Tailings characterization

It is important to carry out an adequate tailings characterization, in order to identify sectors and levels
with different physical properties and hydraulic conductivity, in order to plan correctly the
dewatering strategy. Understanding of non-saturated parameters, porosity and granulometry are
essential. Performing previous long-term pumping tests will ensure feasibility.

Adequate dewatering strategy

Numerical modelling simulations with reliable information is an essential aspect in order to define
the better dewatering strategy (n° of wells, times and rates of pumping, etc), as well as to anticipate
problems and difficulties during the operation.

Water levels monitoring

A continued monitoring of wells and piezometers allows analysing the evolution of the water levels
and the drawdown by sectors (see Figure 6), and additionally allows to adjust the pumping flow in
each well to make it as efficient as possible and prevent equipment failures.

Figure 6 Main hydrogeological section of B2 (North-South)

8
Use of additional support systems

In areas with fine particle and relatively shallow water levels; wellpoint drainage systems can help
to lower water levels and reduce lateral water entries (recharges). Of course, well point need to be
properly defined in order to work with low permeability materials.

Reduction of water incomes

In areas with fine particle (high content of silts and clays) and relatively shallow water levels;
wellpoint drainage systems can help to lower water levels and reduce lateral water entries
(recharges). Of course, wellpoint distances, need to be properly defined in order to work with low
permeability materials.

Hydrochemical implications on equipment maintenance

As is often expected in a TSF, waters in this case study have shown an acidic behavior (~pH 4) and
high concentrations of sulfate, arsenic, manganese and iron. This quality of waters causes important
oxidation processes in the pumping equipment and accessory materials, as well as clogging processes
in pumps and pipes (see Figure 7).
Due to the aggressive conditions of the water, the continued pumping through wells produces a
partial blockage of the pipes, as well as in the pumping equipment itself, which leads to a loss of
efficiency of the drainage system. In this sense, it is important to carry out a preventive maintenance
of both the water drainage system and the welllpoint system, particularly those related to pumps,
valves, pipes and instrumentation (measurement sensors). Effluent need to be properly managed to
ensure any environmental impact.

Figure 7 Accumulation of oxides in pumping equipment

9
Tailings disposal to ensure best dry conditions

It is recommended to dry the extracted material (water content according to material porosity) to
reduce the residual moisture content prior to processing. Need of some space, as a stock for this, it
must be considered.
Finally, the future forecasts of the desaturation of the deposit using the automated monitoring of the
control network and the updating of numerical models, are an essential part of the strategy to
anticipate the problems and difficulties to develop the project successfully.

ACKNOWLEDGEMENTS

Special thanks to Minsur´s team for their support and authorization to prepare this publication, both
for the corporate personnel involved in the project and for the outstanding personnel of San Rafael
mine. Since good time, both teams are participating with great effort in the project and making this
challenge a reality in the B2 dewatering.

REFERENCES
Azapagic, A., 2004. Developing a framework for sustainable development indicators for the mining
and minerals industry. J. Cleaner Prod. 12, 639–662.

Carlin, J (2005): Tin. In Mineral Commodity Summaries. U.S. Geological Survey. Periodic Pulication,
174-175).

Clark AH, Farrar E, Kontak DJ, Langridge RJ, Arenas Figueroa MJ, France LJ, McBride SL, Woodman
PL, Wasteneys HA, Sandeman HA, Archibald DA (1990) Geologic and geochronologic constraints on
the metallogenic evolution of the Andes of southeastern Peru. Econ Geol 85:1520–1583

Eriksson, N., and P. Adamek. (2000) "The tailings pond failure at the Aznalcóllar mine,
Spain". Environmental Issues and Management of Waste in Energy and Mineral Production, Balkema,
Rotterdam: 109-116.

Förstner, U., 1999. Introduction. In: Azcue, J.M. (Ed.), Environmental Impacts of Mining Activities:
Emphasis on Mitigation and Remedial Measures. Springer, Heidelberg, pp. 1–3.

Fyfe, W., 1981. The environmental crisis: quantifying geosphere interactions. Science 213, 105.

Kossoff, D., Dubbin, W. E., Alfredsson, M., Edwards, S. J., Macklin, M. G., & Hudson-Edwards, K. A.
(2014). Mine tailings dams: characteristics, failure, environmental impacts, and remediation. Applied
Geochemistry, 51, 229-245.

Rico, M., Benito, G., Salgueiro, A.R., Díez -Herrero, A., Pereira, H.G., 2008b. Reported tailings dam
failures: a review of the European incidents in the worldwide context. J. Hazard. Mater. 152, 846–852.

10
Zhai, Q., & Rahardjo, H. (2012). Determination of soil–water characteristic curve variables. Computers
and Geotechnics, 42, 37-43.

11
Mixed-in-Place Techniques to Improve Dam Safety
and Reduce Seepage
Hans Schwarzweller, Jaime Sobrino and Mario Goulart
BAUER Fundaciones, Panama

ABSTRACT
Due to an increasing awareness, regarding dam safety and environmental concerns, related to
seepage of contaminated fluids of tailings dams, tailor-made and cost-effective solutions to upgrade
existing dams are becoming ever more important.
By mixing the existing soil strata of dikes or dams in-situ with binding agents – usually cement-
bentonite slurry - using different techniques, improvements can be achieved in a fast and economic
way with very little or no spoils.
The article will describe two different technical solutions – Mixed-in-Place (MIP) by augers and
Cutter-Soil-Mixing (CSM) and gives examples of executed projects.
The Mixed-in-Place (MIP) method is ideal in non-cohesive soils and to depths up to 24 m
approximately. In the MIP method, the surrounding soil is broken up with a single or triple auger, it
is then mixed, and the pores filled in with the binder suspension. The individually treated areas are
combined to form walls by the step back method. In this, it is the size of the auger(s), which ultimately
dictates the dimensions of the newly formed cut-off-wall.
The Cutter-Soil-Mixing (CSM) method combines features of the diaphragm wall trench-cutter
technique and the MIP method, as well as offering some additional benefits: the use of modified,
high-performance cutters means soil mixing is possible even at greater depths and in densely packed
soils. The method also enables even greater wall thicknesses and higher levels of reinforcement to be
attained if required. Accordingly, in addition to the applications described for the MIP method, the
CSM method is suitable in particular for reinforcing high dams and dikes. The main advantages and
limitations of both methods are analyzed and quality control issues addressed.
In summary, soil-mixing solutions are highly attractive in economic and ecological terms and can be
adapted to a great number of different site situations.

1
INTRODUCTION

Seepage control technologies such as cut-off walls are widely used in civil engineering, in dam and
dike construction, for temporary and permanent groundwater barriers. Innovative construction
methods and cut-off wall equipment have been developed over the last 40 years increasing
application of soil mixing techniques for cut-off walls to a large variety of construction sites with
vastly different constraints and subsurface conditions, for both civil and in mining projects. Cut-off
walls are used as well to enclose contaminated areas or to direct contaminated ground water to a
central water treatment facility.

METHODOLOGY
To minimize or even prevent disposal of excavated material during cut-off wall construction and to
decrease the amount of slurry being used, active soil mixing systems are nowadays a state of the art
technology. Best results in quality and performance can be achieved in uniform soil formations. The
application is limited in heterogeneous soils with boulders, very compacted soils and rock
formations. The Cutter Soil Mixing (CSM) system is technologically the most advanced system, using
a modified trench cutter technique for constructing a cut-off wall. Deep Auger Mixing Systems are
also frequently used but face various limitations compared to CSM. For each project, specific binder
needs to be developed considering the in situ soil conditions and the required characteristics of the
cut-off wall.

Mixed-in-place (MIP)
The term 'Mixed-In-Place' describes the process of mixing soil in-situ with binder. During this
process, the pores within the soil structure are filled with the binder slurry. The result of this process
is a strengthened body of soil in the shape of which is defined by the geometry of the augers (Figure
1):

Figure 1 Mixed-In-Place (MIP) construction sequence

2
Construction of a MIP trench is carried out by drilling a triple continuous flight auger to the required
depth (Figure 2). During drilling and withdrawal of the augers, the soil structure is broken-up,
thoroughly mixed and binder slurry is introduced through the hollow stems of the augers. In order
to achieve complete homogenization of the soil during mixing, the rig operator is able to vary the
direction of rotation for each individual auger and at the same time raise or lower the sledge mounted
augers (Figure 3). Construction of the MIP panels is carried out in bites (Figure 4) by adopting the so-
called 'back-step' construction sequence. Thus ensuring a continuous wall without joints. If required,
project specific reinforcement can be installed by lowering it into the freshly mixed wall.

Figure 3 Movement options during mixing

Figure 2 Mixed-In-Place equipment

Figure 4 ‘Back-step' construction sequence

Cutter Soil Mixing (CSM)

The two counter rotating CSM cutter wheels penetrate into the ground at a continuous rate. The
cutting wheels break up the soil matrix and at the same time, a fluid is pumped to the nozzles set
between the cutting wheels, where it is mixed thoroughly with the loosened soil. Adding compressed
air stream can improve the breaking and mixing process during the downward movement of the
cutter. The rotating cutter wheels push the soil particles through vertically mounted shear plates thus
completely mixing the materials. The penetration speed of the cutter and the volume of fluid pumped

3
in are adjusted by the operator to create a homogeneous, plastic soil mass, which permits easy
penetration and extraction of the machine (Figure 5):

Figure 5 Cutter Soil Mixing (CSM) construction sequence

Having reached the design depth, the soil-mixing cutter (Figure 6) is slowly extracted while cement
slurry or cement-bentonite continuously added. The mixing and withdrawal speed in combination
with the required binder-slurry volume is carefully planned and controlled. Homogenization of the
liquefied soil mixture is ensured by the rotation of the cutter wheels. A continuous wall is formed in
a series of overlapping primary and secondary panels. Overcutting into fresh adjacent panels is called
«fresh-in fresh method», but also the «hard-in-hard method» is possible, whereby the panel is cut
into already hardened primary panels. Installation of reinforcement into fresh panels is possible as
well.

4
 Figure 6 Kelly-guided CSM and rope suspended Tandem Side Cutter on MT 75

QUALITY CONTROL

An extensive testing program according to a Quality Management Plan shall be undertaken to

monitor and control the complete production process and to check the final product. This includes
all steps of the entire process like preliminary preparations, site set-up, materials delivery, slurry
production, storage and transport, and in-situ mixing thus ensuring the quality of every single panel
and the envisaged wall-continuity. In general, samples are taking of every delivery of cement and
bentonite and tests for the slurry density, Marsh time and temperature are done in the usual way
directly after mixing. In addition, the slurry volume is measured with a flow meter and recorded
with electronic devices (B-Tronic). Slurry samples are taking during each shift to control the curing
process of the slurry. The homogenization of the in-situ mixed materials has to be verified by
sampling directly from the wall immediately after the mixing process has been completed. The final
product can be checked by conventional core taking.

RESULTS AND DISCUSSION

MIP or CSM cut-off walls have limited structural function. To improve the stability of the dams or
dikes, individual soil-mixing-elements can be reinforced by steel-elements. In addition further soil-
mixing-elements orthogonal to the barrier wall can be installed. Number and lengths of such
stabilizing elements depend on the project’s requirements and the prevailing soil. With the
techniques described soil permeability less than 1 x 10 -8 m/s can be achieved minimizing seepage,
improving the overall long-term dam stability. Compared to other cut-off wall solutions soil-mixing
techniques are both economical and fast solutions with only limited volume of material transported
to and from site. No concrete is placed as the in-situ soil and binder-slurry is used as a construction

5
material and limited drill spoil is generated. Other existing barrier systems, for example those
constructed with conventional grouting methods, often fail to meet the degree of efficiency and
durability and other performance specifications over the life of the structure as required by the
design. Both soil-mixing techniques use relatively light equipment when compared to conventional
dig and replace techniques like grabs and cutters.

Figure 7a Existing tailings dam with seepage problems

Figure 7b Rehabilitation of existing dam by installation of cut-off wall

Depending on the depth of the installed cut-off wall, the level of the downstream water table can be
influenced. In addition, the reduced seepage of eventually contaminated tailings liquids will improve
the ground water quality in the areas downward the dam.

Case studies:

1. Cutter-Soil-Mixing Wall at Fort McMurray, Alberta, Canada:

The project was governed by two main factors: schedule and temperature. The specialist foundation-
engineering contractor had been approached to execute a cut-off wall that was approximately 900 m
long and up to 13 m deep at the South West Sand Storage (SWSS) at the Syncrude site. The main
criteria were the execution during severe winter conditions, as the Fort McMurray region is known

6
for its cold conditions from November to March. The construction of the wall in these winter
conditions created many unique problems with handling equipment and materials that specialist
foundation-engineering contractor had to solve. A common soil-bentonite wall was not possible due
to the freezing of the slurry in the open trench. Frost would also prevent the mixing of the excavated
material with the slurry and bentonite powder to create the homogenous soil bentonite mixture as
usually used. The solution for this project was a Cutter-Soil-Mixing (CSM) wall. The Bauer CSM-
method allows the bentonite-cement-binder-slurry to be injected directly into the soil at the location
of the treated wall. Only the first 0.5 m to 1 m of frozen soil had to be removed to create a small trench
to contain any overflowing suspension. The specialist foundation-engineering contractor’s solution
for handling the slurry preparation equipment was to set up the mix plant inside a large heated tent.
To guarantee that the slurry would not freeze while being pumped from the tent through a heated
booster station on to the machine up to 500 m away, the slurry line had heated connections and
insulation wrapping the line. The soil mixing was performed by a BAUER RTG 19 with a BCM 10
cutter attachment. To speed up the production process of the cutter pre-drilling was performed by a
BG 18 drill rig and a continuous auger. This enabled the contractor to treat the sandy soils and to
finish the project on time allowing the client to continue operations without any impact on the
production.

Figure 8 Cutter Soil Mixing equipment

2. Mixed-in-Place (MIP) Wall at Sonthofen, Germany:

After floods in 1999 caused extensive damage along the River Iller in southern Germany (Figure 9),
the embankments were post sealed by vertical cut-off walls constructed by the triple-auger Mixed-
in-Place method with steel elements installed as reinforcement. During the floods of August 2005, the
embankments were subjected to a severe test earlier than anticipated. The result: the MIP-stabilized
and reinforced embankments withstood the lateral pressures caused by the extremely high water
table of the river as designed even when being overtopped.

7
 Figure 9 Mixed-in-place triple-auger technique. The MIP – wall, while being overtopped prevents levee to
collapse during flood event

CONCLUSION
The article demonstrates that Mixed-In-Place (MIP) and Cutter Soil Mixing (CSM) are suitable
methods to install cut-off walls within existing dams and dikes in a fast and economical way. The
design can be adapted to a great number of different site conditions according to the requirements of
the project. Whilst cut-off walls executed in conventional ways similar to diaphragm walls have been
used for the rehabilitation of tailings dams the authors are not aware of the execution of mixed-in-
place cut-off walls in this field. In consideration of the benefits mentioned in this article the techniques
mentioned here could become of interest for a number of tailings rehabilitation projects in future.

NOMENCLATURE
CSM Cutter soil mixing
MIP Mixed in place

REFERENCES

Arnold, M., Beckhaus, K., Wiedenmann, U. (2011), ‘Cut-off wall construction using Cutter Soil
Mixing: a case study’, Geotechnik 34 (2011), Heft 1.

Banzhaf, P., Colmorgen, B. (2011) ´Reliable seepage control by plastic concrete cut-off walls´, USSD
21st Century Dam Design – Advances and adaptations, 31st Annual USSD conference, San Diego, CA, USA.

Bauer Maschinen GmbH (2016) ‘CSM Cutter Soil Mixing, Process and Equipment’, Schrobenhausen,
Germany.

Bauer Maschinen GmbH (2016) ‘Mixed-In-Place’, Schrobenhausen, Germany.

8
Banzhaf, P. (2016) ´Dam Safety for new and aging dams – seepage control using advanced
construction techniques – tailor-made to meet the individual project and owner’s needs’,
Schrobenhausen, Germany.

Stefan, S. (2016) ‘Cut-off Wall Technologies in Mining’, Proceedings IMWA 2016, Freiberg, Germany.

9
Bituminous Geomembranes (BGM) Longevity and
Past Use for Waterproofing Tailings
Emilio Escobar1, Bertrand Breul2 and Bernard Breul3
1. Axter Iberica, Chile
2. Axter, France
3. Bituminous Geomembrane Expert, France

ABSTRACT
The paper starts with a description of the structure of a bituminous geomembrane (BGM) which is
manufactured by impregnating a polyester geotextile with a compound based on elastomeric
bitumen. The geotextile provides high mechanical resistance and high puncture resistance. Using a
BGM manufactured in a plant eliminates the issues associated to waterproofing using natural
materials like clay, including but not limited to finding enough quantities of materials with
appropriate water tightness characteristics, , having the suitable weather conditions to place them
and the difficulties to follow a stringent QC/QA.
The paper presents the information regarding the long durability of BGM and it addresses the specific
BGM characteristics that help explain this longevity. We start with results of field-testing naturally
aged BGM under covered conditions after 15 and 35 years of service at an ICOLD dam in Corsica
(France). Then we explain the life expectancy studies carried out by the French company Andra for
the use of BGM for the containment of low and medium intensity nuclear wastes. These studies
concluded, after seven years of testing, that biodegradation of BGM in confined conditions is longer
than 300 years. And finally, we develop the results of BGM testing in North America done by the
Department of Nuclear Energy in the Brookhaven National Laboratory (Upton, NY) and the Batelle
Pacific Northwest Laboratory (Richland, WA) that validate a life expectancy of BGM of 300 and even
1,000 years in confined conditions.
At the end, we present some case studies of BGM in tailings dams and reservoirs where BGM has
been installed in remote and very harsh conditions: the Rio Tinto Diavik Diamond Mine in Northwest
Territories of Canada; the Agnico Eagle Kiitilä Gold Mine in northern Finland within the Arctic
Circle; the Kinross Kupol Gold Mine in Siberia.

1
1 INTRODUCTION

Bitumen has been used as a construction and a watertight product for centuries. From the fourth
millennium BC, the Sumerians, the Babylonians and the Assyrians exploited bitumen as hydraulic
mortar for the construction of terraces, canals, dams, etc., including the famed hanging gardens of
Babylon. Many ancient achievements, including the Tigre dikes at Assur (Mesopotamia), still bear
witness to the exceptional longevity of bitumen. Bitumen emerged again as a standard 20th century
waterproofing material in civil and water engineering in the form of bituminous concrete, asphalt
and bituminous geomembrane (BGM).

2 GENERAL PRESENTATION OF BGM

The first applications of a BGM prefabricated in a plant, took place in Avoriaz (France) in 1975 for
the waterproofing of potable water storage reservoirs at a 2000 m altitude in the Alps, under the
supervision of Doctor JP Giroud.
A BGM is a geomembrane manufactured by impregnating a non-woven, needle punched, long fiber
polyester geotextile (up to 400 g/m2) and a glass fleece with a bitumen compound at a temperature of
approximately 180°C. The top of the structure is sanded to enhance UV resistance and provide grip
for installation workers. On the bottom, there is a polyethylene terephthalate (PET) film to limit
penetration of any vegetation roots through the membrane. Please refer to Figure 1.

Figure 1 Coletanche Structure

The geotextile provides the mechanical resistance and the high puncture resistance. The elastomeric
bitumen provides the waterproofing properties of the geomembrane. The bitumen has a large
resistance to UV radiation, a notable cold bending capability and it ensures the longevity of the
composite by totally impregnating and coating the fibers of the geotextile.

2
BGM thicknesses typically range from 2.2 to 5.6 mm according to the ASTM D5199 standard, with a
mass per unit area ranging from 2.5 to 6.4 kg/m2 as per ASTM D3776.

3 DURABILITY OF BGM

The durability of a geomembrane is measured in terms of how it keeps its physical, mechanical and
hydraulic properties over time. The paper will review the durability of a BGM based on actual testing
of samples from job sites after 5 to 40 years of service life as well as long-term durability projections
based on studies done by scientific authorities to use a BGM to waterproof nuclear waste storage sites
where a durability of 300 years was required.

3.1 Laboratory testing

3.1.1 Testing from actual job sites after up to 20 years in service

La Technologie Routière, a French laboratory carried out testing on BGM sample material exhumed
from two test sites in France, as follows:

− A potable water reservoir in Guazza, Corsica (France). Testing was performed after 10 years
of service,

− A motorway ditch along highway A81 in Le Mans (France), where the geomembrane was
installed on a sand layer of 10 cm and covered with 25 cm of topsoil in 1979. Testing was
performed after 7, 14 and 18 years, of service.

Figure 2 Guazza Reservoir and Highway A81

Test results from both cases show that the coefficient of permeability according to Darcy’s law
decreased by only one power of 10 after 20 years to 4.0 x 10-13 m/s for a BGM exposed. In addition,
noteworthy is that there was no significant reduction in tensile strength for any of the samples as the
tensile properties of BGM are derived primarily from the polyester non‐woven geotextile that is well
protected from ageing and degradation processes by being totally impregnated with bitumen.

3
CEMAGREF, Laboratory of the French Ministry of Agriculture, followed up on the conditions of
some dams waterproofed with exposed BGM for more than 20 years. The results published by Prof.
Girard (IRSTEA Bordeaux, Rencontres Géosynthétiques 1993, France) conclude that the life cycle of
BGM is greater than 20 years without any maintenance.
3.1.2 Testing from Ospédale Dam in Corsica (France) after 40 years in service
The Ospédale dam was the first large rockfill dam worldwide (more than 20 m high) to use a
geomembrane sealing system (GSS) using a BGM. A testing program undertaken by CEMAGREF in
2007 concluded that the reduction of the tensile strength after 40 years in service is 17,5%, well within
the 25% allowed by European standards and that the BGM has maintained its waterproofing
capability (flow occurs along the periphery of the dam and not through it). Prof. Gourc at the 2017
Gecamin Geosynthecs Conference in Santiago presented detailed results.

Figure 3 Overview of the Ospédale dam in Sept. 2018 and under construction

3.2 Durability projections

3.2.1 Required CE Testing
The CE marking requires the evaluation of the durability of geomembranes by measuring the
mechanical characteristics of geomembranes after artificial ageing in UV and after oxidation
according to EN 12224: the geomembrane is exposed to an UV brilliance during 3000 hours with the
cycle phase wet/dry following:

− 5 hours of exposure in heat dries, with a temperature of 50°C,

− 1 hour of watering in the water, with a temperature of the black sample of 25°C.
The energy exposure is 50 MJ/m2. After oxidation, the values of tensile strength and elongation
measured on the BGM are equal on average to 90 % of the measured value of a new
geomembrane. The results of mechanical tests after ageing by the effect of UV and oxidation allow
to estimate the life expectancy being superior to 60 years.

4
3.2.2 ANDRA Studies
The French National Radioactive Waste Management Agency (ANDRA) is a public body in charge
of the long‐term management of all radioactive waste in France. It initiated a program to seal the
radioactive waste produced everywhere in France (laboratories, hospitals, etc.) for at least 300 years,
which represents the life of this low and medium radioactive material. The key criteria to select the
capping liner for this project were:

− Imperviousness: the amount of rainwater that may penetrate the cap and thus meeting the
waste must be kept to a strict minimum to avoid leaching and leakage of radionuclides,

− Longevity: the cap must be flexible in the time in order to move with the substrate but must
also remain,

− Watertight throughout the surveillance phase, i.e. 300 years,

− Protection: in addition to water infiltration, the cap must be able to resist to different stresses
such as erosion, temperature fluctuations, chemical attacks, living organisms, etc.
A BGM was finally selected for the capping of the “La Manche” facility, which took place between
1992 and 1994. Please refer to Figure 4.

Figure 4 “La Manche” Radioactive Waste Storage Facility (ANDRA, 2009)

In order to determine the long‐term ageing and effectiveness of existing potential covering
membranes, ANDRA developed a mathematical model and calibrated it with test results on samples
taken from existing structures sealed by BGM since 5, 10, 15 and 20 years. All liner samples were
collected by Coyne et Belier, a well‐known international consulting firm in dams and water
management, from different existing sites. To ensure BGM met the criteria, ANDRA asked for
numerous tests in various laboratories. After seven years of testing (conducted by ANDRA and the
French Nuclear Safety Commission), the conclusion was that under the worst-case scenario, the
maximum biodegraded thickness of the BGM is 1,5 mm on either side (i.e., a total of 3,0 mm out of
5,6 mm) after 300 years of service. The remaining thickness of BGM would still serve as an effective
waterproofing material.

5
3.2.3 American studies
In this paragraph, we will mention only results of studies done and published in USA
3.2.3.1 Biodegradation Testing of Bitumen studies done by Department of Nuclear Energy in Upton
An estimate of the rate of biodegradation of bituminous material is necessary to predict the long-
term stability of low and intermediate level radioactive waste. Data from a series of experiments have
been analysed to determine the rate of degradation of BGM samples under a variety of conditions.
Among the variables investigated were the effect of soil type (Barnwell, Richland soil, Upton) soil
moisture (usually 5.6%), sample surface area and microbial strain. The tests were done in chambers
with BGM samples and in measuring the metabolic CO2 release from these chambers under the
action of very active bacteria. The duration of these tests is between 60 and 197 days. Using these
data, it was found that, for degradation in soil, a mean rate of 5.5 x 10-4 cm/yr represented all data
within a factor of about two. One other conclusion was that the environmental factors studied have
a small and surprising effect on the rate of biodegradation.
3.2.3.2 Studies done by Batelle Pacific Northwest Laboratory (PNL)
A study on BGM, one of the components in a design for a cap to isolate wastes for 1000 years by Bob
Romine, Research Scientist at Battelle Pacific Northwest Laboratory, Richland, Wash, H. Freeman,
Research Scientist at Battelle Pacific Northwest Laboratory, Richland, Wash. These researchers
studied the asphalt’s long-term performance to learn how it can be used in protective barrier systems
designed to permanently isolate radioactive, hazardous and mixed wastes (Freeman et Al. 1994).
Because of the 1000-year design life of these facilities, conventional 30-year Resource Conservation
and Recovery Act (RCRA) closure covers are considered inadequate. A team of engineers and
scientists on the Hanford Site Permanent Isolation Surface Barrier Development Team has worked to
create a design for a cap that could isolate wastes at this site for 1,000 years. The team includes
engineers from Westinghouse Hanford Co., Kaiser Engineers Hanford Co. and scientists from PNL.
Their conclusion is polymer-modified asphalts offer the distinct advantages of elongation resiliency
and a service life believed to surpass other geomembrane materials with a thickness of 4 mm and
fulfill the life expectancy of 1,000 years and furthermore since the membrane is self-bonding repair.

4 CASE HISTORIES

4.1 Diavik Diamond Mine

The Diavik Diamond mine in the Northwest Territories of Canada is located on the East Island of Lac
de Gras which is approximately 300 km northeast of Yellowknife (64°31’ North, 110°20’ West). Rio
Tinto is the operator of the project. The Diavik Mine site lies within the Arctic Climatic Region where
daylight reaches a minimum of 4 hours per day in winter and a maximum of 20 hours in
summer. The climate is extreme, with long, cold winters and very short, cool
summers. Temperatures are cool, with an average mean monthly temperature in July of 10°C and in

6
January of -50°C. Winds are moderate to strong. Snow falls in every month. On average, there are
260 days with temperatures below zero per year.
BGM was primarily installed in three water management ponds in 2005 and following years until
now due to the weather conditions under which the BGM can be installed, particularly under
extremely cold conditions, areas of cost and schedule savings that the change allowed are identified.
The period of construction of Pond 14 would be constructed during freezing conditions. In addition,
Esker Sand would not be available during this time, so that the liner bedding and cover materials
would require crushing and screening to produce the required volumes. Alternative dam sections
were evaluated to allow efficient cold weather construction and to minimize the amount of crushing
and screening required. A section incorporating BGM liner in the upstream section was selected. The
decision to use a BGM liner was based on the elastomeric bitumen in the liner remaining flexible
below –20°C and having a low coefficient of thermal expansion (1 x 10 -6 cm/cm/°C). This meant that
the material would be relatively insensitive to temperature changes during installation and
service. As well, the BGM is a heavy material, and not prone to being lifted by light winds during
installation. This type of liner had been installed at other sites at temperatures down to about –20°C
and productivity during installation at low temperatures was reported to be similar to productivity
at above zero temperatures.
A 50 mm minus crushed gravel was selected as the liner bidding and cover material. Tests during
construction showed that a crushed 150 mm minus cobble, gravel and sand material could be used
without damage to the liner, and this material was used as the bedding and cover. Pond 14 was
constructed during April and May 2005, with 12,700 m² of liner installed during the period from May
7 to May 13, 2005. During this period the temperature varied from –18°C to –4°C, winds were
typically in the range of 10 km/h to 20 km/h and 1 to 2 cm of snow fell on a number of days. The rate
of liner installed varied from 1,300 m²/day to 2,400 m²/day. Quality control testing and quality
assurance testing involved vacuum box and ultrasound testing of seams to confirm the quality of the
welds and removal of samples for destructive testing. Vacuum box testing proved to be more
effective than ultrasound testing in the cold conditions during installation. Liquids with anti-freeze
product are required to prevent the vacuum box from freezing to the liner.

Figure 5 Tailings Reservoir, Diavik mine

7
4.2 Kittilä Gold Mine

This mine is owned and operated by Agnico-Eagle Mines Ltd. based in Toronto (Canada). This mine
is the second mine located north of Finland. Environmental permission and the use of BGM for the
mine operations was granted by Northern Finland Environmental Permit Authority in 2002
Key factors of the mine: mine area 860 ha, annual ore mining 1 000 000 tones, annual gold production
5 tones. The Kittilä Mine site lies within the Arctic Climatic Region (150 km north). Temperatures are
cool with an average mean monthly temperature in July of 15°C and in January of -15°C. Two ponds
were designed initially with a total area to cover of 540 000 m² and until this time every extension or
new tailings ponds are designed with BGM. Now, over a million square meters of BGM are installed.
The purpose of these ponds of are to store the water that comes from the process. Water from the
ponds is circulating back to the process plant. Therefore, the water circulation is practically a closed
system.

Figure 6 Aerial view Kittila Gold Mine

Dam walls are made from blasted rock with compound structure. This means that the main dam wall
is made from maximum 600 mm diameter grain size blasted rock. The “wet” sides of the dam walls
are made also from blasted rock (dam max. 300 mm grain size) and gravel (grain size 55 mm). On the
horizontal part of the ponds. BGM is laid directly on a mineral sealing layer made by using local
excavated moraine, which is compacted. Thickness of the mineral sealing layer is 1000 mm and water
permeability less than 5x10-8 m/s. Mineral sealing layer and BGM act as a composite liner as with
BGM there is no wrinkles. It means that BGM and support stay in intimate contact creating a double
liner system. For the bottom, a 3.5 mm bituminous liner was used and for the slopes, it was a 4.0 mm
thick BGM.
BGM was chosen because:

− It can be installed directly over the prepared mineral sealing bottom layer instead of a layer
of sand and geotextile for HDPE. This provided a large cost savings for transporting and
screening these extra layers and a large economy in CO2, very good for limiting impact on
the environment,

8
 − Installation is much less dependent on weather conditions. Due to the extended construction
season of BGM. The reduced construction period means that the equipment used for the
ponds can be used earlier in mine construction. For investors, this could have the mine
operation up and running much sooner, and the financial impact could be huge.

4.3 Kupol Mine

The Kupol deposit is in the Chukotka Autonomous Okrug (Russia). The total distance between the
Kupol and Bilibino the nearest major city, is approximately 200 kilometers.
The climate of the region around the Kupol site is extremely severe, consisting of long and cold
winters (8-8.5 months). The average annual temperature is -13°C, ranging from - 58°C to 13°C.
BGM in this project has promoted the ease to be installed and welded in extreme weather conditions
(rainfall, wind, cold temperatures down to -40°C) permitting to earn more than 1 year in the delay
means 1 year of production of gold and the use of wider range of rough materials for the liner bedding
and cover zones.

Figure 7 Kupol dam (Siberia, Russia)

5 TAILINGS CAPPING

The Tundra mine is a large mine located 240 km north of Yellowknife Noth West Territories (Canada).
From 1962 to 1968, the site was exploited as an underground gold mine. Restoration activities at
Tundra focused on the protection of health, public safety and the environment.

9
 Figure 8 Tundra mine, Northwest Territories, Canada

The challenges that remain on the site include the treatment of water in the tailings containment
areas, the treatment of hydrocarbon-contaminated soils, covering tailings and waste rock and
decommissioning of dams.
The Client was Aboriginal Affairs of Canada, the location is in North West Territories, Canada. The
Product used was a BGM of 4 mm thick with inside a geotextile of 250 gr/m². The surface are 335 088
m². The Installation was between 2010 to 2012. BGM was installed to control leachate and protect
groundwater. It is used as cover to prevent rain water infiltration into the waste. In time, BGM will
allow the rehabilitation of the site with revegetation.

6 CONCLUSION
As we have presented in this work, the laboratory testing on brand-new samples according to
European standardization for anticipating life duration lead to anticipate a minimum duration of 60
years in covered conditions. Testing of exhumed samples from actual job sites showed limited loss
of their properties (mechanical and permeability) after 40 years under covered conditions. In
addition, testing done to assess the long-term durability of the BGM have allowed the experts to
project that BGM biodegradation under covered conditions is a very slow process and thus water
tightness can be maintained for more than 300 years and up to 1,000 years.
The tailings reservoirs where BGM has been used under extremely harsh weather conditions have
been operating flawlessly for more than 15 years.
All of the above, coupled with the fact that a BGM can have a reliable, well-documented QA/QC
process from its manufacturing through its deployment and installation, lead us to conclude that a
BGM can be safely used to maintain the water tightness of tailings reservoirs for periods well beyond
the useful life of the mines.

10
REFERENCES

Eloy-Giorni, C., Pelte, T., Pierson, P., and Margarita, R., 1996, « Water Diffusion through
Geomembranes Under Hydraulic Pressure », Geosynthetics International, Vol. 3, No. 6, pp. 741-769
(Conference Paper)

Giroud, J.P., and Bonaparte, R., 1989a, « Leakage through Liners Constructed with Geomembranes,
Part I: Geomembrane Liners », Geotextiles and Geomembranes, Vol. 8, No. 1, pp. 27-67. (Book)

Giroud, J.P., and Bonaparte R., 1989b, « Leakage through Liners Constructed with Geomembranes,
Part II: Composite Liners », Geotextiles and Geomembranes, Vol. 8, No. 2, pp. 71-111. (Book)

Giroud, J.P., Khatami, A., and Badu-Tweneboah, K., 1989, « Evaluation of the Rate of Leakage
through Composite Liners », Geotextiles and Geomembranes, Vol. 8, No. 4, pp. 337-340. (Book)

Giroud, J.P., Badu-Tweneboah, K., and Bonaparte, R., 1992, « Rate of Leakage through a Composite
Liner due to Geomembrane Defects », Geotextiles and Geomembranes, Vol. 11, No. 1, pp. 1-28. (Book)

Giroud, J.P., and Morel, N., 1992, « Analysis of Geomembrane Wrinkles », Geotextiles and
Geomembranes, Vol. 11, No. 3, pp. 255-276. (Erratum : 1993, Vol. 12, No. 4, p. 378.) (Book)

Giroud, J.P., and Soderman, K.L., 1995, « Design of Structures Connected to Geomembranes »,
Geosynthetics International, Vol. 2, No. 2., pp. 379-428. (Book)

Gourc, Emer.Prof.,IGE-LTHE / University Grenoble-Alpes / France Back to the first large rockfill dam
with an upstream Geomembrane Sealing System (Ospedale,1979) (Conference Paper)

Alonso, E., Degoutte, G., Girard, H. G. (1990) ‘Results of seventeen years of using geomembranes in
dams and basins’ , Proceeding of the 4th International Conference on Geotextiles, Geomembranes
and Related products, The Hague, NDL , pp437-442(Conference Paper)

Cazzuffi, D., Giroud, J.P., Scuero A., Vaschetti, G. (2010) ‘Geosynthetic barriers systems for dams,’
Keynote Lecture, Proceeding of the 9th International Conference on Geosynthetics, Guaruja, Brasil,
Vol.1, pp115-163 (Conference Paper)

Giroud, J.P., (2009) ‘An explicit equation for strain in geomembrane uplifted by wind’, Geosynthetics
International, Vol.16, N°6, pp 500-502 (Report)

Gourc, J.P., Delmas, Ph., (2016) ‘The behavior of “alive” earthworks with geosynthetics after several
decades’ Keynote Lecture, Proceeding of the 6th European Conference on Geosynthetics, EuroGeo 6,
Ljubljana, Slovenia, 43p(Conference Paper)

Gautier, J.L., Lino, M., Carlier, D., (2002) ’ A record height in dam waterproofing with bituminous
geomembrane: La Galaube dam on Alzeau river’, Proceeding of the 7th International Conference on
Geosynthetics, Nice, France(Conference Paper)

11
ICOLD-CIGB (2010) ‘Geomembrane Sealing System for dams-Design principles and return of
experience’, Bulletin 135, International Commission of Large Dams-Comittee International des
grands Barrages, Paris, France, 464p(Conference Paper)

Lambert, TOUZE-Foltz, N.,(2000) ’ A test for measuring permeability of geomembranes’, Proceeding

of the 2nd European Conference on Geosynthetics, Bologna, Italy 10p(Conference Paper)

Identification of The True Elastic Modulus Of High Density Polyethylene From Tensile Tests Using
An Appropriate Reduced Model Of The Elastoviscoplastic Behavior A. Blaise, Stéphane André,
Patrick Delobelle, Yves Meshaka, C. Cunat

Mafra, J.M.Q., Mello, J., Eldridge T., Breul, B., (2008) ‘Two case histories of dams waterproofing with
bituminous geomembrane’ , Proceeding of the 1st Pan American Geosynthetics Conference, Cancun,
Mexico ,10 p. (Conference Paper)

Peggs, I.D. (2015) ‘prefabricated bituminous geomembrane: a candidate for exposed geomembrane
caps for landfill closures.’ ,Florida, USA(Conference Paper)

Royet, P., Girard, H., Poulain D., Tisserand, C., (2002) ‘Comportement des géomembranes dans les
barrages français en remblai’, Ingéniéries-E.A.T.pp3-15 , <hal-00465446->(Conference Paper)

R.E. Barletta, B.B.Bowerman, R.E.Davis, C.E.Shea; Department of Nuclear Energy, Brookhaven

National Laboratory, Upton, NY 11973.“ Biodegradation testing of bitumen”, Published in Scientific
Basis for Nuclear Waste Management X, Symposium held December 1-4,1986, Boston, Massachusetts,
U.S.A. (Conference Paper)

Paul Morgan, Alan Mulder Engineers at Thornton Shell Research Center Shell Bitumen, “The Shell
Bitumen Industrial Handbook,” 1995. (Book)

Sageos Laboratory Test Report # S 549-005 (Report).

Den Haag, Geomembrane in dams and basins. Geotextiles, Geomembrane and Related Products,”

Alonso, E., Degoutte, G., and Girard, H. (1993), “Réalisation et comportement d’ouvrages
hydrauliques munis d’une géomembrane,” Comptes rendus, Rencontres 93, pp. 285-294. (Conference
Paper)

W. Drent and H. Eschrich, Bituminization of Radioactive Wastes - Bibliography, ETR-274, June, 1972.
(Book)

R. Roffey and K. Hjalmarsson, Microbial Processes in the Final Repository, the Silo Part. Theoretical
Approach and Preliminary Experiments on the Biodegradation of Bitumen. Part 1 . , F0A-C-40172-
B4, 3 May, 1983. (Book)

L. P. Buckley, B. C. Clegg and W. K. Oldham, Radioactive Waste Management and the Nuclear Fuel
Cycle, 6(1), 19 (1985) (Book)

12
R. W. Traxler, in Bituminous Materials: Asphalts, Tars, and Pitches, Volume I : General Aspects,
edited by A.J. Hoiberg (Huntington, New York, 1979), pp 323 - 346. (Book)

C. E. ZoBell and M. A. Molecke, DOE Report No., SAND78-1371, 1978. (Report)

T. B. O'Neill, "The Biodegradation of Oil in Seawater for Naval Pollution Control", NCEL N-1195,
January 1972. (Book)

N. A. Abdellah and B. M. Pederson, in Seminar on Testing, Evaluation and Shallow Land Burial of
Low and Medium Radioactive Waste Forms, Ceel Belgium, 1983, R. Krisher and R. Simon, Eds.,
Harwood Academic Publishers, 1984. (Book)

P. L. Piciulo, C. E. Shea and R. E. Barletta, U.S. Nuclear Regulatory : Commission Report No.
NUREG/CR-4200, 1985 (Report)

B. S. Bowerman, K. J. Swyler, D. R. Dougherty, R. E. Davis, B. Siskind and R. E.-Barletta, U.S. Nuclear

Regulatory Commission Report No. NUREG/CR-3829, 1985. (Report)

U. S. Nuclear Regulatory Commission, Technical Position Paper, "Technical Position on Waste Form,"
Rev. 0, May 1983.(Report)

R. Bartha and D. Pramer, Soil Science, 100(1), 68-70 (1965). (Book)

P. L. Piciulo, C. E. Shea, and R. E. Barletta, U.S. Nuclear Regulatory-Commission Report No.

NUREG/CR-4083, 1985 .(Report)

Giroud, J.P., 1997, « Equations for Calculating the Rate of Liquid Migration Through Composite
Liners Due to Geomembrane Defects », Geosynthetics International, Special Issue on Liquid
Migration Control Using Geosynthetics Liner Systems, Vol. 4, Nos. 3-4., pp. 335-348. (Book)

Giroud, J.P., Soderman, K.L., Khire, M.V., and Badu-Tweneboah, K., 1998, « New Developments in
Landfill Liner Leakage Evaluation », Proceedings of the Sixth International Conference on
Geosynthetics, Vol. 1, Atlanta, Georgia, USA, March 1998, pp. 261-268. (Conference Paper)

Giroud, J.P., 2000, « Lessons Learned from Failures and Successes Associated with Geosynthetics »,
Keynote Lecture, Proceedings of Eurogeo 2, the Second European Conference on Geosynthetics,
Bologna, Italy, October 2000, Vol. 1, pp. 77-118. (Conference Paper)

Rowe, R.K., 1998, « Geosynthetics and the Minimization of Contaminant Migration through Barrier
Systems Beneath Solid Waste », Keynote Lecture, Proceedings of the Sixth International Conference
on Geosynthetics, Vol. 1, Atlanta, Georgia, USA, March 1998, pp. 27-102. (Conference Paper)

ICOLD / CIGB. 1991. Watertight geomembranes for dams -State of the art / Etanchéité des barrages
par géomembranes - Technique actuelle. Bulletin technique CIGB, n° 78: 140p, Paris (France) (Book)

Lambert, S. & Pierson, P. & Rollin, A. 2002. Géomembranes : Guide de choix, Presses Internationales
Polytechnique. Montréal, Commercialisation pour la France. Tec & Doc Lavoisier: 312 p (Book)

13
Davis, C.A. & Lew, M. & Ponnaboyina, H. & Perez, A. 2013. Interface Friction Testing Between Soil
and a Bituminous Geomembrane. Proceedings of Geosynthetics April 1-4. Long Beach, CA,
(Conference Paper)

14
Reducing Leakage Risks on a TSF´s by Using a
Flexible Ballast Layer as Covering System: Case
History
Daniel Fernandez1 and Melchior Briers2
1. HUESKER, Brazil
2. HUESKER, USA

ABSTRACT
In most of the mining regulations around the world, a relative thin ballast layer (“cover soil”) is
required in order to ensure both UV and mechanical protection to the primary barrier system
(geomembrane). Several case studies conducted over the years, provided evidence that leakage of
hydraulic barriers is caused during the construction of the conventional ballast layer. This involves
the use of granular soils that are placed and compacted on top of the geomembrane surface, piercing
and stressing the geomembrane while heavy equipment operate on top of it.
Due to the above mentioned challenges on site, a flexible ballast layer has been developed, resulting
in a mattress containing tailings as filling material, confined by two or more geotextiles. The system
is designed so that the ballast layer can be installed with no heavy equipment operating on the
geomembrane, therefore reducing the risk of possibly damaging the primary barrier and leakage. A
second positive effect is the use of the mine waste as a fill/construction material, as this reduces the
CAPEX of the pond whereas increasing its capacity in approx. 3 to 4%. Since no special granular
material needs to be imported or produced/crushed on-site, the carbon foot print reduction also plays
an important role in the analysis.
Several on-site testing were performed to prove the systems feasibility on a TSF in South Africa. This
case history is shown as an example of what innovation can achieve, when the performance is a
priority in accordance to local regulations.

1
INTRODUCTION

Since the promulgation of the National Norms and Standards for disposal of waste to landfill in
August 2013, only waste facilities storing inert waste (Type 4) have been exempt from including a
composite liner system which is formed by the combination of an HDPE geomembrane and
underlying clay layers.
The thickness of the HDPE geomembrane ranges from 1.5 to 2.0mm thick, depending on the
classification of the waste. The thin layer is prone to mechanical damage which leads to leakage
through the liner. In order to restrain the leakage locally around the damaged geomembrane,
intimate contact between the underlying clay layers and geomembrane is required. Intimate contact
is achieved by applying a confining stress to the composite lining system.
Also, awareness of UV degradation of exposed HDPE has recently been investigated in South African
conditions. Msiza and Shamrock (2014) considered four case studies of facilities where geomembrane
that had been exposed for several years could be compared to unexposed geomembrane, say from
an anchor trench. They found that in each case degradation occurred. In one of the case studies, the
degradation was used as motivation to include a ballast layer above the primary geomembrane in
the next phase of the facility.
Thermal expansion of HDPE leads to the formation of waves or wrinkles which prevent intimate
contact and leads to high stresses in the geomembrane. The ballast layer both insulates the
geomembrane preventing these folds from forming and restrains the geomembrane from expanding.
However, folds that are entombed in the ballast layer remain entombed as found by Soong and
Koerner (1998) and do not flatten out as may be expected. Rather the crest of the fold sharpens causing
increased tensile stress in the crest. Typically a ballast layer is included to prevent both mechanical
damage and UV degradation while applying a confining stress to ensure intimate contact. The ballast
layer may form a protective layer but it may also serve as a drainage layer in facilities that generate
leachate.

TYPES OF BALLAST LAYER

The following types of ballast layer have been used in waste facilities:
 Rock: typically used in landfill facilities in combination with a sand protection layer or
protection geotextile. The rock acts as a drainage layer, allowing leachate to gravity drain to
the low points of the basin, ensuring that hydrostatic pressure does not build up within the
ballast layer which would be a driving force for leakage through any damaged areas.
 Sand: where circumstances allow it, sand may be used instead of rock. It does not require a
protection geotextile if the particles are below a specific size (say 3.0mm). Sand may also be
available on-site leading to cost-savings for the project.

2
  Cement stabilised sand: In facilities that will be storing liquid, the drainage function of the
ballast layer is not required. Stabilising the sand that forms the ballast layer with cement
provides a strong surface that prevents damage during periods where the facility is dry or
being maintained.

 Precast concrete blocks: Precast blocks, in combination with a protection geotextile, provide
a solid protection layer for desilting of dams. The blocks can also be placed manually,
reducing traffic driving on the ballast layer during installation. However, this is a costly
solution and the size of the facility dictates its feasibility.
The thickness of the ballast layer is typically 300mm as stated in Designing with Geosynthetics
[Koerner, 6th Edition]. Specifically when using compaction equipment on a cement stabilised ballast
layer, a thickness of 300mm allows larger machinery to be used in the installation.
The examples above include the use of aggregate (sand, stone, cement, concrete) in order to create
the physical boundary between the geomembrane and the elements that may cause damage.
However, the most risk to the geomembrane is during the ballast installation - this negative aspect of
the ballast layer will be discussed in a further section.
The use of a Geosynthetic Ballast Layer (GBL), although still requiring filling, provides less risk
during installation due to no machinery being used on top of the ballast layer.
The following GBL alternatives are available:
 Protection geotextile: A non-woven geotextile with a mass of 1000g/m2 would be able to
provide protection to the geomembrane. However, if used in isolation, the geotextile would
be significantly degraded by UV in a short period of time and therefore is not a permanent
solution. Its confining stress would also be insufficient to insure intimate contact.
 Concrete filled geocells: Concrete placed in geocells provides high compressive strength
ensuring protection of the geomembrane. However this is also costly and requires
specialized works to construct.

 Fibre reinforced concrete: This product is formed by a thin layer of concrete, approximately
5 - 13mm thick, reinforced by fibres. It is supplied in a roll, 1.0m wide, and can be installed
without the excessive use of machinery. It would serve as a good ballast layer offering good
compressive strength. It’s mass, at 7.5kg per m2, would be sufficient to ensure intimate
contact. However, the convenience comes at a cost and the size of the application is directly
linked to its feasibility.
 Double woven tubes: formed by two woven polypropylene geotextiles seamed together by
weaving the woven threads together. In order to combat the threat of UV degradation, a third
non-woven geotextile layer is sewn on the double woven. Being internationally
manufactured, the cost of supply and transport is high and dependent on the exchange rate.

3
 Figure 1 Schematic cross section of a GBL as double woven tubes

A case study of using the international double woven tubes filled with tailings is presented in the
following section.

METHODOLOGY: THE SITE TRIAL

A hazardous waste facility is currently being constructed in KwaZulu-Natal. The Client, who has
constructed waste facilities before, requested that the ballast layer should not be constructed using
large plant driving on the ballast layer if an alternative was available. This lead to the research and
development of a site specific GBL that will be filled with the same tailings that the facility is being
constructed to store.
As the design of the facility was completed and approved by the authorities without the inclusion of
the GBL, it was decided that a full scale site trial should be conducted to ensure that the GBL would
be compatible with the tailings while serving all the functions of a ballast layer.
Four 15m long by 4.8m wide samples were delivered from the manufacturer in Germany to carry out
the trial. Each sample was formed of five 93cm tubes with 8cm weaves between. It was decided to do
two trials on the side slopes and two on a horizontal surface to imitate the ballast layer on the basin.
The first trial was carried out on the side slope of an existing facility. The gradient of the slope was
1:2 and the supporting layer was a smooth HDPE geomembrane. The GBL was anchored at the crest
using duck-billed anchors as there was insufficient space to excavate an additional anchor trench.

4
 Figure 2 GBL filling trial preparation on site slope

After the GBL was anchored at the crest, it was rolled down the side slope and pumping points were
inserted at the crest. The solids stored within the facility were at a sufficient height that the GBL was
horizontally supported for approximately 2.5m at the toe. The GBL was then filled with a slurry
coming from the plant that contained approximately 15% solids by volume. The bags immediately
filled with liquid. The solids, however, slowly started to build up in the GBL at the toe of the slope.
After various stages of filling, the bags were effectively filled with solid tailings. The consolidation of
the solids within the bag after 24 hours, allowed pedestrian traffic on the bag, indicating a bearing
pressure of at least 15 KPa without noticeable deformation. Therefore the GBL was considered
technically viable on the side slope.
The width of the bags experienced shrinkage during the filling process. As the bags filled and
increased in height, the sides of the bags were forced inwards. The bags filled to a height of
approximately 450mm. Over the 4.8m width, a reduction of 26% was experienced. This information
is required in the estimation of quantities for the full facility.
The next trial took place on the horizontal level of an earlier waste facility. The bag was rolled out
and pumping points were installed in one end. The GBL was then filled with slurry and again the
bags immediately filled with liquid.
However it was comparatively difficult to fill the horizontal bags with solids than the bags that were
placed on the side slope. It was found that after a certain duration, the solids that were being pumped
into the bag were equal to the solids being forced out of the side seams due to the pressure at which
the bags were being filled.

5
 (a) (b)

Figure 3 Trials on site. (a) side slope, (b) horizontal surface

During the second horizontal trial, different pressures were used as well as different locations of the
pumping points (moved from centre to side of bag). The GBL was effectively filled with solid tailings
indicating technical feasibility for the basin. It was decided that a pressure regulator would be
required to control the pressure during the installation.
In terms of shrinkage, both horizontal bags experienced a 5% reduction in width. The bags did not
fill to the same height as the bags on the side slope but filled to a height of approximately 200mm.

Installation of the ballast layer

Typical sand ballast layer installation process

The liner is rolled out on the prepared clay layers during the cool hours of the day.
While the geomembrane is still flat, the ballast sand is carefully placed on the geomembrane, on the
same day using a small wheel mounted loader.
The sand or cement stabilised mixture is spread above the geomembrane using a small track mounted
dozer – no turning of the dozer is permitted.
Compaction equipment is used if the sand is cement stabilised to ensure a sufficient strength is
obtained.
Folds may be generated due to the “pushing” nature of the spreading process. As the sheet is not yet
welded, the aim is to push the fold outwards towards the sides of the sheets being installed.
No folds should be entombed during the cover the process.
Once a sufficient area is covered, the ballast for the following geomembrane panel is heaped on
constructed ballast in preparation for moving it onto the next sheet after it has been welded to the

6
previous sheet. This implies that the trucks carrying in the ballast material will be driving on ballast
on installed geomembrane.
Sheets are only welded together when they have reached the same temperature.

GBL (double woven tube alternative) installation process

The liner is rolled out on the prepared clay layers during the cool hours of the day;
The GBL is laid out above the geomembrane and is secured with UV stabilised sand bags and rope;
The GBL will prevent UV degradation of the geomembrane as well as mitigate its thermal expansion
causing fewer and smaller waves to occur.
If the GBL is going to be filled with tailings, the filling process cannot commence until the entire
geomembrane installation is completed. If it is going to be installed with sand slurry then filling could
proceed immediately. Either process will require the ability to pump the filling into the bag.
If the GBL is going to be installed with slurry, there will be liquid generated at the low points of the
facility which will need to be continuously pumped away. If the slurry is tailings, the liquid will be
contaminated and will need to be pumped to another facility. Either way, operations of filling the
bag will require constant management of liquid.

Figure 4 The geosynthetic ballast layer installed on the side slope and suitably ballasted with UV resistant
sand bags and rope. The GBL is white due to the exposed non-woven geotextile being stitched to the
underlying black woven geotextiles.

7
ANALYSIS AND DISCUSSION: THE POSITIVE AND NEGATIVE ASPECTS OF
USING A GEOSYNTHETIC ALTERNATIVE

 The main positive aspect of using a GBL alternative is keeping machinery off of the liner
system. Nosko and Touze-Foltz (2000) found that when considering the results of electrical
leak detection surveys carried out over 300 sites covering a lined area of 3 250 000m2, that
77.8 % of the damage occurred on the basin and that 15.59% of the damage was due to heavy
equipment driving on the cover layer above the geomembrane. Removing this risk in the first
place should lead to less damage occurring during installation.
 The GBL can be filled with tailings which means that it is not taking up valuable airspace in
the facility. This can be viewed as waste re-use and is aligned with Goal 1 of the National
Waste Management Strategy of 2012.

Figure 5 TSF covered with confined and dewatered tailings on a GBL

 A calculation made by Cilliers and Briers (2018) showed a comparison on Embodied Carbon
emissions among two ballast solutions: cement stabilized sand layer (CBL) and a flexible
geosynthetic mattress ballast (GBL) determining that the second one produces only 4% of the
carbon emissions than a traditional CBL solution (Table 1)

Table 1 Total calculated Embodied Carbon emissions (Cilliers and Briers 2018)

Embodied Carbon emissions (tCO2e)

Solution Embodied Transport Construction Total
CBL 504 0.468 1.66 506.3
GBL 19.5 1.162 0 20.7

8
The negative aspects of using a GBL are the following:
 UV degradation: This is more of a concern on the side slopes than on the basin which is
usually covered shortly after commissioning. The UV exposure consumes the antioxidants
within the geosynthetic which leads to the engineering properties of the GBL breaking down.
 Veneer stability: Textured geomembranes have increasingly assisted in ensuring that liner
systems are stable on the side slope. However, introducing a geosynthetic interface where
there used to be a sand interface leads to less friction being transferred specifically if the GBL
is going to be filled with slurry. Koerner and Narejo (2005) report that the typical interface
residual friction angle between a textured geomembrane and a woven geotextile is 18° which
is slightly lower than the angle of a 1:3 slope (18.4°) typically used in waste facilities,
suggesting instability. The typical interface residual friction angle of a textured
geomembrane and a gravel material, reported in the same document, is 31°, which would
have a high factor of safety against instability. Ideally site specific shear interface testing or
full scale tests should be carried out.
 Anchorage: The reduced veneer stability is less of a concern if the GBL is suitably anchored
at the crest. As it will be installed at the same time as the primary geomembrane, it may share
the same anchor trench.

Figure 6 Anchorage detail of GBL at the crest

 Time to fill bags: Filling the bags may be a longer process than constructing a conventional
ballast layer. Instead of placing and spreading the ballast layer with large machines, the GBL
is manually unrolled and then each bag is individually filled. This will require on-going and
dedicated operations.
 Cost: The cost of the GBL can vary significantly depending on where it is manufactured and
what filling material is used. The thickness of the ballast layer also has an effect on cost as it
is consuming valuable airspace. A cost comparison of options from the case study is
presented in the following section.

9
The costs comparison

Table 2 shows the construction cost comparison of four different ballast layer options that was carried
out for the case study facility:

Table 2 Comparative costs of different ballast layers

No. Alternative description: Cost per m3 of % more Cost per m2 % more

waste stored * than of liner than
lowest footprint * lowest
1 300mm cement stabilised sand ballast layer R 129 8.4% R 885.11 4.4%
(site sand used), protection geotextile of
1000g/m2 included
2 Fibre Reinforced Concrete Ballast Layer R 189 59% R 1347.99 59%
(thickness negligible)
3 Double woven tubes filled with tailings R 119 - R 847.44 -
(thickness negligible)
* R = south african Rands (exchange considered €1 = R 14.35)

Long Term Performance

Samples test were carried out on geosynthetic tubes used to contain sediments at the Husum
Harbour project in Germany from September 2013 until April 2014 (Wilke 2017). The tests
undertaken provided the following data for analysis; soil density; water content; ignition loss;
particle size distribution; density of solid particles; lime content and field vane sheer tests. The
overall dry solid (DS) content observed was 51.61%. It was witnessed that over time the rainfall does
not adversely affect the DS content of the encapsulated material.

CONCLUSION

The advantages of a GBL solution mentioned above may not necessarily justify the adoption of this
technique for every project. A proper feasibility analysis must be done in order to consider costs,
carbon emissions, environmental licenses, local conditions, etc. In those countries where
environmental regulations demand a risk analysis on different failure modes, the leakage through
the barrier system may become a priority. In this context, every step of the barrier system installation
will be taken into account, including the high damage risk on the primary barrier, which may be
produced by equipment traffic during a cover layer placement and compaction. Therefore, a leakage
risk reduction through the adoption of a GBL solution increases its feasibility on a certain project.
When considering strain control on barrier systems, every geomembrane may suffer stress cracking
if no proper conditions were designed, considerably reducing the project life. Most of the barrier
systems are not designed to control long strains when the peak shear strength value on an interface
decreases to its residual shear strength value. Therefore, having a covering layer which is anchored
at the crest of a slope eliminates the risk of stress cracking on the primary barrier.

10
The state-of-the-art trend is to have a better control on uncertainties for every possible risk (leakage).
The use of GBL´s is a very innovative engineering solution that allows for the safe containment of
pollutants resulting in environmentally friendly mining practices.

REFERENCES

Brundtland, G. (1987). Our common future: The World Commission on Environment and
Development. Oxford University Press, Oxford, UK.

Cilliers C. (2015). The use of geosynthetics in the installation of ballast layers. Proceedings of the 2015
Landfill Conference, September 2015, Tulbagh, Western Cape, South Africa.

Cilliers C., Briers, M. (2018). A new safe and sustainable approach for constructing ballast layers for
waste containment facility liner systems. Proceedings of the 2018 Tailings & Mine Waste Conference,
September 2018, Keystone Resort, Colorado, USA. Environmental Product Declaration Number EPD-
HUE-20160014-IAC1-DE, Huesker Synthetic GmbH, 2016. Institute Construction and Environment
e.V. (IBU), Germany.

Giroud, J.P., Pelte, T. and Bathurst, R.J., 1995, Uplift of Geomembranes by Wind, Geosynthetics
International, 2, (6), pp. 897-952.

Hammond, G.P. & Jones, C.I. (2011). Inventory of (Embodied) Carbon & Energy (ICE) V2.0.
Department of Mechanical Engineering, University of Bath, UK

Huesker. (2016). Geosynthetic Ingenuity in Mining. Gescher, Germany, Huesker Synthetic GmbH.

Huesker. (2017). Incomat® Geosynthetic Concrete Mattress System for Hydraulic Engineering.
Gescher, Germany, Huesker Synthetic GmbH.

Koerner, R. M., (2016). Geotextiles: From design to applications, 1st ed., Woodhead Publishing,
Sawston, Cambridge, U.K.

Koerner, GR and Narejo, D. Direct shear database of geosynthetic-to-geosynthetic and geosynthetic-

to-soil interfaces. GRI Report 30. Geosynthetic Research Institute, Pennsylvania, USA.

Nosko V. and Touze-Foltz N., (2000). Geomembrane liner failure: modelling of its influence on
contaminant transfer. Proceedings of the Second European Geosynthetics Conference, Pàtron
Editore, Bologna, Italy, pp 557-560.

Raja, J., Dixon, N., Fowmes, G., Frost, M., Assinder, P. (2015). Sustainable Construction Solutions
Using Geosynthetics: Obtaining Reliable Embodied Carbon Values, Geosynthetics International 22
(5) pp 393-401.

RSA (Republic of South Africa) (2013c) National Environmental Management: Waste Act (59/2008):
National norms and standards for disposal of waste to landfill. Government Gazette 36784 No.R. 636
of 23 August 2013.

11
U.K. Department for Transport. (2012) Fuel consumption by HGV vehicle type in Great Britain, 1993
to 2010. Table RFS0141. Domestic road freight activity (RFS01). Available from:
https://www.gov.uk/government/statistical-data-sets/rfs01-goodslifted-and-distance-hauled.

Wilke, M., Cantre, S. (2017). Port maintenance operations by geosynthetic dewatering tubes -
Working principles and practical experience. Proceedings of the Coasts & Ports Conference, June
2017, Cairns, Australia.

12
Flood Routing over Talabre TSF
Alfredo Edwards1, Francisco Fernández1, Tirso López2, Paolo Maldonado3 and Gonzalo
Caro3
1. EIC Ingenieros Consultores, Chile
2. Knight Piésold, Chile
3. Codelco, Chile

ABSTRACT
Tailing Storage Facilities (TSF) are usually located at a zone where it is exposed to natural
watercourses, hence they must have the capacity to regulate floods of high return periods. This
includes the probable maximum flood (PMF).
Generally, the reservoir flood routing uses storage-discharge methods, however, the unsteady flow
over deposited tailings and the sediment transport have not yet been examined quantitatively.
This paper addresses PMF at Talabre TSF (Atacama desert, Northern Chile), with a 65 km2 interiorly
sectorized area. A two-dimensional unsteady flow calculation was applied considering sediment
transport and mobile bed.
The results indicate an attenuation of the inflow hydrograph as it flowed over the reservoir. An
increase of water level was detected when the flow moves from one sector to another. The latter, can
be explained by the dimension of the watercourses formed by the tailings and the uses of sectorized
distribution inside the TSF. This produces a temporal decrease of the freeboard and it brings the
water closer to the perimeter embankment.
Prediction indicates the breach of the interior dykes (used to sectorized the TSF), which triggers a
sudden evacuation of the accumulated water to the adjacent sector located at a lower elevation. This
result implies a decrease of the water level, which increases the freeboard and the distance of the
flood to the perimeter embankment. At the same time, this implies a local increase of erosion.
Results suggest that in order to improve the flow from one sector to another, the filling plan and
internal dykes design must consider these effects in order to distance the flood from the perimeter
embankments and to ensure the structural safety against eventually extreme weather conditions in
Atacama Desert.

There is no full article associated with this abstract.

1
4
Innovations in
Conventional,
Thickened and
Paste Tailings
Resilience in Tailings Management
Sergio Barrera
Delfing, Chile

ABSTRACT

The concept of resilience is very old and for a long time was applied mostly to living systems, which
have the potential for self-regeneration. Recently this concept has been extended / applied to other
types of non-self-regenerative systems - including tailings deposits - with the aim of evaluating the
system's ability to cope with large demands or requirements without affecting the final objective of
the system. Under this approach, resilience ensures that in light of any event or request the system
will mostly retain its main properties. In other words, once the event is over, the work or system will
continue to fulfil the purpose for which it was built and functioning as if nothing significant or greater
had occurred. In addition, resilience is a relevant capacity / property of the system when evaluating
the safety of a work and, also, to compare alternatives in the option selection phase. Bearing in mind
the growing importance of tailings management works due to their implications for the environment
and the mining business, this paper examines this concept and proposes a methodology for
evaluating the resilience of these works, describing the different relevant factors and a way to apply
it in the particular case of tailings deposits.

1
INTRODUCTION

Failure in tailings impoundments can have major effects not only for the operations of the
impoundment itself (by not having a structure for tailings storage) but also in the community and
the environment which may lead to the loss of the social license to operate. The concern for this
potential damage has increased in recent times as a consequence of the recent large failures (Mount
Polley, Fundão), the rapid and ample spread of the consequences, and the increasing social pressure
associated with the event. This has led to an advance in safety standards (design, construction, and
operations) and governance in mining companies.
Tailings impoundments are in general large civil works that store low strength / low stability
materials and that, therefore, require structural containment elements which normally are dams. In
addition, these facilities sometime intercept water bodies and must, therefore, manage large volumes
of water which mix with or accumulate on the surface of the stored tailings. The mix of low strength
materials (tailings) and the accumulation of water present a challenge for the safety of the structure
and, in particular, to the embankment. The concept of resilience is born from trying to assure the
stability of these structures. The concept of resilience is very old and for a long time was applied
mostly to living systems, which have the potential for self-regeneration. However, this concept has
been recently extended / applied to other types of non-self-regenerative systems as is the case of
tailings deposits.
The purpose of this paper is to describe the concept of resilience, specifically applied to tailings
management, including the tailings impoundment and its vicinity.

WHAT DO WE UNDERSTAND ABOUT RESILIENCE IN TAILINGS?

Resilience is appreciated because it represents the “capacity of a material, mechanism, or system to
recover its initial state when the perturbations to which it was under cease to exist” (one of the
definitions of the Real Academia de la Lengua {Royal Spanish Academy}). The above implies a
capacity for adaptation of living organisms (which is not the case of a tailings impoundment).
Another vision of the concept of resilience has been posed by McLellan et al, 2012, which is applied
to structures related to energy. In this case, the resilient system must comply with the following
conditions:
1. “The amount of change the system can undergo and still retain the same controls on function and
structure. 

2. The degree to which the system is capable of self-reorganization to accommodate external changes. 

3. The ability to build and increase the capacity for learning and adaptation”.
McLellan et al, 2012 also point out that the concept of resilience is an approximate antonym of
vulnerability, implying that a structure is vulnerable when it presents the following components: (i)
“sensible to perturbations or external stresses; (ii) lack of capacity to cope or adapt; and (iii) expose
to perturbations”. Finally, McLellan et al, 2102 define resilience as “the ability of an energy system, or

2
component thereof, to withstand damage due to a natural disaster, to have a benign or negligible impact on
sustainability in case of damage or loss, and to contribute in a positive way to societal recovery post-disaster.”
A similar definition of resilience is posed by Robertson & Caldwell (2017) as “resilience is the ability of
a system to manage change without failure, or recover with low consequences”. In all these definitions,
resilience of a tailings impoundment is associated with minimizing the failure or consequences and
there obviously is no reference to self-regeneration (since these aren’t living organisms according to
a classical definition).
Resilience aims to identify the strengths or weaknesses of a system with the purpose of evaluating
the stability (safety) of it. The risk assessment (or cost estimate by probability of occurrence of a
failure) differs from the resilience because its objective is to evaluate the potential economic damage
of the failure of a structure / system. This implies that the risk assessment evaluates the resilience of
the structure or system as part of the process to estimate the probability of failure.

WHY IS RESILIENCE IMPORTANT?

As indicated above, resilience is key to avoiding or minimizing the occurrence of large failures in
complex structures / systems such as tailings impoundments where there is a complex configuration
of different elements that form part of the system or different parts of the structure. Therefore,
resilience must focus in identifying the weak links / parts of a system that work together or
identifying the existence of replacement capacities.
In other words, if the capacity of a part or element is exceeded, other elements will provide the
resistance or capacity to maintain operations or the resistance of the system or the integrity of the
system.
The above implies that systems or structures where the failure can only occur due to the concurrence
of several simultaneous weaknesses are significantly more resilient than those systems whose
resilience is supported by the resistance of only one element or factor. A classic example of the
difference in resilience is found in the upstream and downstream construction methods of
embankments. Robertson & Caldwell (2017) present this case in Figure 1. The downstream method
has several elements to control the water: core, filter, and drain. In the case of tailings sand dams
there are also several elements: the sand that acts as the filter of the deposited tailings and also drains
because it is permeable (maximum fines content of 20% for copper tailings) and an extensive base
drainage system to eliminate the water of the sand. In this manner, the body of the embankment is
not saturated and in addition sand has high density (compacted in layers). On the other hand, the
resistance of the upstream construction method is totally based on the placement of a reduced
horizontal thickness layer of compacted, permeable sand which is placed on very low density,
saturated tailings. Any deficiency in the quality of this sand makes the stability of this structure come
into question.

3
 Figure 1 Resilience of different construction methods

FAILURE MODES

Resilience is associated with a potential for failure with failure defined as is “a major uncontrolled and
unintended release of retained water / tailings, or an event whereby a dam is rendered unfit to safely retain
water or tailings because of a total loss of structural integrity” (Almog et Al, 2011). The above implies the
occurrence of an undesired event that significantly affects operations and the environment. Figure 2
presents the iconic failures that have occurred worldwide since the beginning of the 20th century.

Figure 2 Iconic failures in the last century

4
The causes of these failures have been varied and studies (ICOLD, 2001; Concha, 2018;
http://www.wise-uranium.org/mdaf.html) have identified the following main causes:
1. Water overtopping over the dam (large floods or spillway with low capacity)
2. Large earthquake
3. Weak foundation strata
4. Poor quality sand
5. High phreatic level within the sand dam (low permeability sand or significant decrease in
the capacity of the drainage system)
As may be seen, the geotechnical factors seem to be the most significant, but they are not the only
ones. It is also worth noting that the construction method used was the upstream method in the
majority of the iconic failures shown in Figure 2.

RESILIENCE AREAS

Resilience encompasses all the facilities or areas in the impoundment, including those located
downstream that may be impacted by tailings release or seepage. Keeping this in mind, four
resilience areas are identified that are associated with physical areas and with the management of
design / construction / operations / closure:
• R1 – Resilience of the embankment
• R2 – Resilience of the impoundment
• R3 – Resilience of the downstream environment
• R4 – Resilience of the management
Figure 3 presents a schematic view of these resilience areas.

Figure 3 Areas of resilience

5
Resilience R1 – of the embankment

It encompasses all aspects that have to do with chemical and physical stability. From a physical point
of view, the following aspects are noted:
• Construction materials. In particular, it is relevant to resilience depends on the type, the
geotechnical characteristics of the initial material, the placement, the volume or quantity, the
variability, and the stability (durability) with time.
• Foundation soil. The foundation soils are key to the stability and must be characterized by a
proper geological investigation based on field tests and laboratory tests. This investigation
must provide ample coverage of the entire footprint of the foundation of the embankment
(including potential expansions).
• Growth / construction method. The selection of the method to raise the embankment must
take into account all the critical aspects for stability: seismicity, rainfall, accumulation of the
water in the impoundment, geotechnical characteristics of the material of the embankment,
quality control during construction, and operations. The upstream method of construction
is not permitted in the majority of high seismicity areas (Chile, Peru and Japan) and has
recently been banned in Brazil. The upstream method is also not recommended in
impoundments where there is the potential for the tailings pond to be located close to the
embankment.
• Drainage system. The presence of a phreatic surface (pore pressure) inside a dam is a
destabilizing factor. For this reason, it is necessary that the embankment must have a robust
drainage system that assures the rapid collection and elimination of water that may seep into
the part of the embankment that provides for structural stability. This is most critical in
embankments that are constructed with fine granular material (e.g., tailings sand) that can
weaken on shearing.
• Geotechnical instrumentation. The geotechnical instrumentation form part of the devices
that are used to monitor the behavior of the embankment. Maintenance, monitoring, and
analysis of the data derived from these instruments is a key function that provides early
warning of deviance from design performance.
Resilience also encompasses chemical stability. The generation of acid does not generate physical
instability, but it could become a major problem for the downstream environment in the long term.
If the material used to construct the embankment could be a source of acid drainage, its mineralogy,
permeability and drainage characteristics must be defined as well as the local climatic conditions.

Resilience R2 – of the impoundment

The management of the tailings water pond is one of the most critical aspects given its impact on the
available freeboard of the embankment and on seepage. The following specific aspects have been
identified as being important factors impacting on the capacity to manage the tailings water pond:
• Tailings disposal scheme. The manner in which the tailings are deposited defines the type
of embankment and its growth strategy. In addition, it defines the location of the tailings
pond and, therefore, the water management aspects that are discussed in the following item.

6
 • Water management and, in particular, the volume of water retained during normal
operations and during floods must be understood. The characterization of the spillway (in
cases where a spillway is necessary) with regards to its location and capacity is included in
this topic.
• Type of tailings. The geotechnical behavior of the deposited tailings depends on the parent
ore body and the process used to recover the commodity. In the majority of the cases, the
tailings are of medium to low permeability, but there are exceptions. The same applies to the
consolidation rate, which may be significantly affected by the drainage capacity of the base
of the impoundment.
All of the above-mentioned factors have an enormous impact on the tailings condition during the
course of operations and at the end of operations and on the time for stabilization after
decommissioning. Moreover, these factors also influence the properties of the tailings for studies of
dam failure.

Resilience R3 – of the downstream environment

If we accept that “resilience is the ability of a system to recover with low consequences”, the
downstream environment plays an important role. For an impoundment that is already in operation
the conditions of the downstream environment are already fixed. This does not however preclude it
from being evaluated as part of the current conditions. The downstream environment should
however play an important role during site selection where a correct evaluation of its vulnerability
is key to making the correct choice of site. In general, the following aspects must be considered:
• Topography
• Flora and fauna (presence of protected or endangered species)
• Cultural patrimony (archeological remains, ancient paths, artisan centers or of ancestral
cultures, etc.)
• Medium or large sized population centers
• Infrastructure (large public works)
• Industries

Resilience R4 – of the management

Every manmade system or operation requires an organization that manages it. For effective and
efficient work, organizations have an internal hierarchal structure, definition of roles, KPIs, and
procedures, among others. Tailings impoundments are operations that have the following
peculiarities that require special care by organizations:
 A long operational life that may exceed 50 years
 During the long life of the impoundment the downstream conditions may change
significantly
 The ongoing construction (raising) of embankments (and other internal facilities) forms part
of the operations
 The impoundment can be expanded, significantly increasing its capacity and its footprint

7
  Regulations and best practice change over time
The management of a tailings facility must consider all these aspects in addition to those normally
applicable to civil work of this magnitude. In summary, the main aspects of management are listed
below and must include:
• Conservation, adequate and systematic ordering of background information for the project
and operation (Memory)
• Retention of an Engineer of Record who is responsible for the safety of the work with regards
to its design and operation
• Reviews / internal audits / external independent audits. The annual Dam Safety Inspection
and the Dam Safety Review, which are periodically done every three to five years
• Internal organizational structure including functional definition of roles / responsibilities,
KPIs, procedures, etc.
• Operations Manual with a section on monitoring of key variables for the stability / safety of
the impoundment. It is important that this manual includes or provides guidelines for the
analysis of the information recorded and routine conformance assessment by the Engineer
of Record
• Quality assurance. Every construction phase must explicitly include quality assurance and
quality control performed by a company that is independent of the contractor. The Engineer
of Record must verify conformance of construction to the objects of the design with reference
to the QA/QC information and as built drawings and must visit the site from time to time.

ESTIMATE OF RESILIENCE
The assessment of relative resilience of a tailings deposit is a two-staged process. The mechanics of
the process is similar to the one applied in the Multiple Accounts Analysis (Robertson & Shaw, 1999).
The first stage consists in the identification of the evaluation items in each one of the four areas of
resilience with their respective indicator. The second stage corresponds to the process of assigning
values to each indicator and the weighting between indicators and between areas. This entire process
includes the participation of a group / team comprising the professional staff responsible for the
management of the tailings and for the design. This process has to be facilitated by a multifaceted
analysis or risk analysis expert. In this manner, a sensibility analysis may be also performed by
modifying the values of the indicators and their weight for each area and between areas.

FINAL COMMENTS
The tailings deposit resilience evaluation methodology described in this paper has appeared fairly
recently and – in the author’s experience – has not been widely used. There is also no technical
literature on the subject that is available for reference from public sources. The author only knows
of one application in Chile, and this was for a tailings deposit with a sand embankment, in which the
author participated. In the aforementioned case, the physical stability resilience assessment was done
in conjunction with the people responsible for the operation of the deposit. The result of the

8
assessment had the double effect of quantitatively estimating the resilience and of identifying the
most critical areas / activities for assuring the safety of the deposit. Additionally, the combined
analysis provided an invaluable opportunity for those in charge of the operation to identify /
understand the key factors influencing the safety of facilities and their relative magnitude and
importance.

REFERENCES

Almog, E., Kelham, P. and King, R. 2011. “Modes of Dam Failure and Monitoring and Measuring
Techniques the Environment Agency”. Project: SC080048/R1

Concha, P & Lall, U. 2018. Tailings Dams Failures: Updated Statistical Model for Discharge Volume and
Runout. Environments 5 (2), 28

ICOLD. 2001. Bulletin 121, Tailing Dams – “Risk of Dangerous Occurrences, Lessons Learnt from Practical
Experiences."

McLellan, B. *, Zhang, Q., Farzaneh, H.; Utama, A. and Ishihara, K. 2012. “Resilience, Sustainability
and Risk Management: A Focus on Energy”. Challenges 2012, 3, 153-182; ISSN 2078-1547;
www.mdpi.com/journal/challenges

Real Academia Española. http://dle.rae.es

Robertson, A. & Caldwell, J. 2017. 3RT – Approach to Assessment and Management. MIG – Inaugural
Conference, London, UK. November 21, 2017

Robertson A. MacG. and Shaw S.C., 1999, “A Multiple Accounts Analysis for tailings site selection”. In:
Sudbury '99 Conference Proceedings, Mining and the Environment II, vol. 3, pp. 883-891.

9
Slurry Dewatering for the Disposal in Tailing Dams
Rubén Olcay1, Christian Hernández1 and George Sales2
1. Universidad Arturo Prat, Chile
2. Universidade Federal de Minas Gerais, Brazil

ABSTRACT

The final disposal in tailing dams using more concentrated slurries presents several advantages when
compared with the conventional disposal due to some aspects: greater operational safety of the dams
located in seismic regions, greater water recovery from the mineral process, reduction of the water
evaporation in arid regions, greater ease of recovery of the land disposal and less environmental
impact. The objective of the present work is to compare three different dewatering techniques applied
to tailings obtained from the industrial process of phosphatic ores. The techniques evaluated were:
laboratory paste thickener, deep cone thickener and hyperbaric filtration (pilot scale). The main
results obtained showed: a) specific gravity of the sample is 3.06 g/cm3; b) the size distribution
indicated 100% passing in 30 microns; c) the chemical analysis showed that magnesium, calcium and
phosphorus are the more important elements; d) mineralogical analysis indicated predominance of
calcite and dolomite; e) the sample presented a high value of specific surface area reaching the value
of 0.976 m2/g; f) considering the techniques used, were obtained the following underflow
concentrations in mass (w/w): 70% for paste thickener, 63% for deep cone thickener and 91% for the
hyperbaric filter; g) the overflow turbidity levels obtained from paste thickener and deep cone
thickener did not exceed the value of 100 NTU, while for the filtrate of the hyperbaric filter the value
was 800 NTU; h) the yield stress values obtained were: 200 Pa for paste thickener and 50 Pa for deep
cone thickener. The technique of dewatering more efficient for final disposal in tailings dams was the
hyperbaric filter followed by the thickeners of paste and deep cone respectively.

1
INTRODUCTION

The disposal of mineral tailings in the form of mineral paste is an alternative technology that may
allow better utilization of the disposal areas, provision of greater stability of disposal, greater reuse
of process water, operational optimization and other factors (Olcay, Hernández & Valadão, 2014).
The use of thickeners to produce of high density slurry or mineral paste is already practiced in several
countries such as Australia for "red mud" disposal, Canada and the United States for disposal of
phosphatic clays (Tão, Parekh & Honaker, 2008; Oxenford & Lord, 2006). Higher density underflow
solids reduce water losses or water usage in many instances. This must be tempered with
consideration to the underflow rheology (Summerhaysa & Gaspar, 2016). In many industries there is
also the filtration with a technique alternative for solid-liquid separation. The filters have a variety of
geometries and mechanisms (Stickland et al., 2018). In the northern region of Chile is an important
issue mainly due to lower availability of water.
Yield stress, which is defined as the minimum stress that must be applied to induce flow, is a key
design parameter in the industrial application of surface disposal of thickened tailings (sometimes
termed paste) (Gao & Fourie, 2015). Several studies relate to the yield stress with a colloidal stability
and surface characteristics of the particles (Brader, 2010; Boger, 2013; Coussot, 2005). The yield stress
of tailings increases exponentially with increasing solids concentration so a small change in
concentration can result in a large change in the yield stress, consequently, control of material
consistency in waste disposal operations is critical (Gao & Fourie, 2015).
The beneficiation of phosphate ores generates large quantities of tailing and by-products. For
example, in Florida (USA) the phosphate mining and beneficiation produces about 100.000 tpd of
waste clays. Waste clay disposal represents one of the most challenging problems for the phosphate
industry. Due to their colloidal nature and ultrafine size (91.5% by weight -325 mesh or 44 µm), the
phosphatic clays are usually very stable in water suspensions and settle extremely slowly. Large areas
for disposal in conventional dam system are required (Tão, Parekh & Honaker, 2008).
The consolidation of phosphatic clays to a state suitable for reclamation and reuse of the land may
take a decade or more, and suitability for support of constructed facilities may not occur at all. Efforts
to solve the problems associated with phosphatic clay disposal and reclamation have included testing
of mechanical, electrical, biological, and chemical processes. These efforts have included the addition
of coagulants and flocculants, mechanical dewatering, the use of enzymes and bacteria and various
other methods of thickening and stabilization (Les, Leon & Jhon, 2011).
That sludge disposal technology is facing challenges and opportunities simultaneously and can still
be improved. Sludge dewatering is an essential process in sludge disposal and is very important for
effective reduction of final processing costs (Wei et al., 2018).
The objective of the present work is to compare three different dewatering techniques: laboratory
paste thickener, deep cone thickener and hyperbaric filter (pilot scale) applied to tailings obtained

2
from the industrial process of phosphatic ores, considering the possible final disposal using more
concentrated slurries.

METHODOLOGY

Mineral Sample

The mineral sample used are tailings obtained from the industrial process of phosphatic ores of
Brazil. The Table 1 presents the different techniques and equipments used to determine the physical
and chemical characterization.

Table 1 Techniques and equipments used to characterize the mineral sample

Property Technique Used Required Equipment

Specific gravity Gas pycnometry Pycnometer and gas
Size distribution Laser diffraction Cilas
Specific superficial área Air permeability Blaine apparatus
Chemical composition X–ray fluorescence Dispersive X-ray spectrometer
Mineral composition X-ray diffraction X-ray spectrometer

Settling Tests
The settling tests were carried out using graduated cylinders (0.5 L) at 10% solids in mass with dosage
of 50 g/t for selecting the best performing flocculant. The types of flocculants used are: cationic (CF),
anionic (AF) and non-ionic (NIF) prepared at concentrations of 0.1% W/V. The flocculant which
presented the best performance was selected for thickening tests.

Laboratory Paste Thickener Tests

This equipment in scale of laboratory is made of acrylic with three cylindrical modules with
maximum slurry volume of 10 L capacity for continuous feeding, use of peristaltics pumps for
feeding and recirculation of slurry, discharge cone angle 45 degrees and others. The rake system,
with low speed rotation, can be used. The underflow discharge of the mineral paste is carried out on
the bottom of the equipment and the overflow exits through the top. All the tests in the paste
thickener aimed to obtain the maximum possible percentage of solids in the underflow. Figure 1
shows the general view of the assembly of the equipment.

3
 Figure 1 General view of operational assembly of laboratory paste thickener

Pilot Scale Deep Cone Thickener Tests

In this case, a pilot scale equipment was utilized. The operational assembly consists in system of
preparation flocculant, feeding pumps and static mixer. The main characteristics of the deep cone
thickener are: diameter 2.0 m, total height 3.7 m and discharge angle 70 degrees. The static mixer has
a diameter of 0.152 m and total length of 4.0 m that allows the addition of flocculant in differents
points. Figure 2 shows the general view of the assembly of the equipment.

Figure 2 General view of operational assembly of the pilot scale deep cone thickener

Pilot Scale Hyperbaric Filtration Tests

The pilot scale tests in the hyperbaric filter were carried out to determine the dry cake loading for
different filter cake thicknesses, blowdown times and filtration rate of the slurry. Compressed air is
continuously applied following cake formation to dry the filter cake. The system mainly consists of:
pulp stirring, feed pump, air compressor and membrane wash water. The objective was to determine
the shortest total cycle time that would produce the largest amount of filter cake at the target moisture
content (20%). Figure 3 shows the general view of the assembly of the equipment.

4
 Figure 3 General view of operational assembly of the pilot scale hyperbaric filter

Yield Stress Measurements

Measurements of yield stress were performed on slurries and pastes using a digital rheometer with
interface for constant sending data to the computer. The methods Bingham and Break-through were
used to determine the yield stress (Olcay, 2012).

Turbidity Measurements

Measurements of turbidity were performed with the overflows obtained of the thickeners and the
filtrate of filter by using portable turbidimeter.

RESULTS AND DISCUSSION

Mineral Sample Characteristics

 Specific gravity: The value obtained by gas pycnometry (GP) of the sample was 3.06 g/cm3.
 Particle size distribution: The sample is very fine with 100% passing in 30 µm. The value of
the median size (d50) is about 5 µm.
 Chemical composition: The results indicating the predominance of CaO (48.22%), MgO
(4.60%), SiO2 (3.03%), P2O5 (3.34%) and Fe2O3 (1.54%) with the presence of small quantities of
Al2O3, K2O, SO3 and Ti2O3. The loss on ignition (LOI) is 37.5% principally due to the presence
of carbonates.
 Mineral composition: The analysis X-ray diffraction was carried out using the method of
Rietveld that indicated as the main mineralogical phases: calcite (CaCO3) and dolomite
(CaMg(CO3)2). It was also detected the presence of phlogopite (K2O∙6MgO∙Al2O3∙6SiO2∙2H2O)
and magnetite (Fe3O4).
 Specific superficial area: Determined by Blaine techniques was of 0.976 m2/g confirming the
fine size.

5
Settling Tests

Figure 4 shows the settling rate obtained for slurry with cationic, anionic and non-ionic flocculants.

Figure 4 Settling rate tests for cationic (CF), anionic (AF) and non-ionic flocculants (NIF)

According in Figure 4 the flocculant selected was the CF3 (cationic) for subsequent tests in the
thickeners. The settling rate reached was 0.480 cm/s at the dosage of 50 g/t. It represents a significant
increase (~40 times) of the settling rate when compared with the test without addition of flocculant.

Laboratory Paste Thickener Tests

The best operational conditions found for slurry feed at 10% solids in mass were the followings: total
time in continuous operation of 3 hours, dosage of the selected flocculant 80 g/t, total volume of
operation 10 L, use of mixer slurry/flocculant, recirculation of the slurry in the compaction zone and
rake rotation in 5 rpm. Figure 5 shows the concentration underflows to different times of operations.

Figure 5 % Solids in mass of the underflow as function of operation time in the laboratory paste thickener

According Figure 5 in the best operation condition were values achieved around 70% solids in mass
after 2 hours of continuous operation. The overflows turbidity levels obtained did not exceed the 100
NTU. Values that are considered suitable for reuse of water in the process.

6
Pilot Scale Deep Cone Tests

The best operational conditions found for slurry feed at 10% solids in mass were the followings: total
time in continuous operation of 3 hours, dosage of the selected flocculant 750 g/t to decrease turbidity
of the overflows, total volume of operation 5 m3, use of static mixer (slurry/flocculant), completely
open underflow valve. Figure 6 shows the concentration underflows obtained to different times of
disposal.

Figure 6 % Solids in mass of the underflow as function of disposal time in the pilot scale deep cone

According Figure 6 was values achieved of 63% solids in mass for 3 hours of continuous operation
and 70% solids in mass after of 4 days of disposal (natural dewatering). The overflows turbidity levels
in the tests did not exceed the 100 NTU.

Pilot Scale Hyperbaric Filter Tests

The slurry at 10% solids in mass was pumped into the chamber and the compressed air was
introduced into the sealed chamber of 5 to 7 bar. The filtration tests were performed at a slurry
temperature of 25-30 °C. Figure 7 shows the dry cake weight as function of cake thickness and cake
moisture versus production rate.

Figure 7 (a) Dry Cake Weight versus Cake Thickness; (b) Cake Moisture versus Production Rate

7
According Figure 7 were obtained productions rates around 170 Kg/hr.m2 for moisture cake of 10%.
The filtrate turbidity levels obtained were of 800 NTU. Values that are considered not suitable for
reuse of water in the industrial process of phosphatic ores, being necessary the use of coagulants with
dosage of 50 g/t for the adequacy of turbidity.

Yield Stress Measurements

Figure 8 shows the yield stress of the slurry for different concentration of solids in mass obtained
from tha tests realizaed in the laboratory paste thickener and pilot scale deep cone.

Figure 8 Yield stress versus concentrated solids in mass

According Figure 8 were obtained high density slurries and pastes with values of yield stress between
50 and 200 Pa.

Economics

Are shown in Table 2 a estimative summary of the capital costs, operating costs and cost per m3 of
each dewatering technique utilized considering as base case 200 m3/h of slurry. The capital costs
considered the acquisition and installation of the equipments. In the operating costs were considered
consumption of reagents and electric energy without considering staff costs, maintenance of
equipments and transport of the tailings. For the determination of dewatering efficiency (Equation 1)
was considering as target 90% solids in mass as value adequate for final disposal in the dams.

Dewatering Real
Efficiency (%) = ( ) ∗ 100 (1)
Dewatering Target

8
Table 2 Summary of capital cost, operating cost and dewatering efficiency

Key Performance Indicator Paste Thickener Deep Cone Hyperbaric Filter

(KPI) (Cost in US$) (Cost in US$) (Cost in US$)
A. Capital cost of plant 3.5 million 2.3 million 6.0 million
B. Operating cost for year 0.07 million 0.65 million 0.20 million
C. Average cost per m3 processed 0.04 per m3 0.41 per m3 0.12 per m3
D. Dewatering efficiency 78% 70% 100%

According Table 2, the hyperbaric filter requiere a higher investment capital but has an excellent
dewatering efficiency followed for the thickeners. Possiblely, the operational costs of deep cone
thickener can be reduced evaluating other types of flocculants with modifications of the system.

CONCLUSION
The hyperbaric filter showed an excellent dewatering efficiency for the disposal in tailings dams but
the quality of the filtrate requires treatment to reach the turbidity values obtained by the other
techniques. Among the tests of thickening in stationary state the best technique was the paste
thickener due mainly to the dewatering efficiency and flocculant consumption, the water quality in
terms of turbidity was similar. The total costs of implementation and operating of each dewatering
equipment will depend on the reality of each project. The global trend in the area mining is the final
disposal with tailing dewatering based on aspects such as operational safety, reuse of process water,
less environmental impact among others.

REFERENCES
Boger, D. V. (2013) ´Rheology of Slurries and Environmental Impacts in the Mining Industry´, Annual
Review of Chemical and Biomolecular Engineering, Vol. 4, pp. 239-257.

Brader, J. M. (2010) ´Nonlinear Rheology of Colloidal Dispersions´, Journal of Physics, Condensed

Matter 22, pp. 9-15.

Coussot, P. (2005) Rheometry of Pastes, Suspensions and Granular Materials: Applications in Industry and
Environment, General and Introductory Chemical Engineering, John Wiley & Sons.

Gao, J. & Fourie, A. (2015) ´Using the Flume Test for Yield Stress Measurement of Thickened Tailings´,
Minerals Engineering, Elsevier, Vol. 81, pp. 116 - 127.

Les, G.B., Leon, C.S. & John, E. (2011) Development and Pilot-Scale Demonstration of Deep Cone Paste
Thickening Process for Phosphatic Clay Disposal, Final Report-Phase II, Institute of Phosphate Research,
University of Kentucky, Florida, USA.

9
Oxenford, J. & Lord, E.R. (2006) ´Canadian experience in the application of paste and thickened
tailings for surface disposal´, Proceedings Ninth International Seminar on Paste and Thickened Tailings,
Australian Centre of Geomechanics, Limerick, Ireland, pp. 93-105.

Olcay, R.H., Valadão, G.E.S. & Hernández, C.A. (2014) ´Caracterisitcas Principales de Pastas
Minerales de Relaves de Manganeso Preparadas en Espesador en Escala de Laboratorio´, Revista del
Instituto de Investigaciones de la Facultad de Geologia, Minas Metalurgia y Ciencias Geográficas, UNMSM-
Perú, Vol. 17, n° 34, pp. 25-34.

Olcay, R.H. (2012) Desenvolvimento de Equipamento Produtor de Pasta Mineral Para Aproveitamento dos
Rejeitos das Lamas Calcárias e Diminuição do Impacto Ambiental. Tese de Doutorado, Universidade
Federal de Minas Gerais, Belo Horizonte, Brasil.

Summerhaysa, R. & Gaspar, A. (2016), ´SYMPHOS 2015: 3rd International Symposium on Innovation
and Technology in the Phosphate Industry. Thickening, Filtration and Clarification in the Phosphoric
Acid Industry´, Procedia Engineering, Elsevier ScienceDirect, Vol. 138, pp. 164 – 173.

Stickland, A.D., Skinner, S.J., Cavalida, R.G. & Scales, P.J. (2018) ´Optimisation of Filter Design and
Operation for Wastewater Treatment Sludge´, Separation and Purification Technology, Elsevier, Vol. 198,
pp. 31-37.

Tão, D., Parekh, B.K. & Honaker, R. (2008) Development and Pilot-Scale Demonstration of Deep Cone Paste
Thickening Process for Phosphatic Clay Disposal, Final Report-Phase I, Institute of Phosphate Research,
University of Kentucky, Florida, USA.

Wei, H., Gao, B., Ren, J., Li, A. & Yang, H. (2018) ´Coagulation/Flocculation in Dewatering of Sludge:
A Review*´, Water Research, Elsevier, Vol. 143, pp. 608-631.

10
Innovative Process Engineering to Create Better
Tailings Facilities
Christian Kujawa, Jeff Winterton, Rachel Jansen and Robert Cooke
Paterson & Cooke, USA

ABSTRACT

This paper explores technologies and approaches for creating better tailings facilities through
innovative process engineering. Historically the optimum tailings solution for a mining operation
has largely been based on economics. With the recent tailings facility tragedies, there is increasing
acceptance that storage facility stability and safety must take priority over cost. This change in
emphasis will allow for the development and implementation of technologies previously considered
cost prohibitive.
Apart from a drive to eliminate tailings altogether, as for example with in-situ mining, there are a
range of technologies which have already been implemented to reduce the volume of tailings stored
at the mine site; these include selective mining and pre-concentration. Increasingly in the future, the
volume of tailings may be reduced when tailings are viewed as a resource to generate commercially
valuable by-products.
There are several approaches to enhance the tailings properties in the deposit. These include coarse
grind processing, amendments to improve tailings dewatering and to sequester carbon.
The concept of Engineered Tailings has been developed to design flowsheets which optimize tailings
dewatering, transport, and storage characteristics. The concept has wide application for improving
tailings management. Classifying tailings and de-sulfurizing tailings to improve their properties are
discussed.

1
INTRODUCTION

The mineral resources industry faces challenges as ore bodies become lower grade, more complex
and finely disseminated. The trend is for larger scale operations producing higher tonnages of tailings
with finer particle size distributions than past operations. This trend is combined with pressure to
improve the stability and safety of tailings facilities, reduce environmental impacts and waste
generation, ensure safe and secure mine closures and operate with a social license while reducing
mine energy and water consumption.
This paper explores how better tailings facilities can be created through innovative process
engineering applied to both processes typically considered as part of the tailings system (i.e. tailings
dewatering and the downstream processes) and to upstream processes which have generally not
been contemplated previously when developing tailings management strategies (e.g. selective
mining and optimizing the grind).
In the past, the optimum tailings solution for a mining operation was largely based on economics.
With the recent tailings facility tragedies, there is increasing acceptance that storage facility stability
and safety must take priority over cost. This change in emphasis will allow for the development and
implementation of technologies previously considered cost prohibitive.
This paper focus on approaches which are novel, or which have not yet had wide implementation.
Accordingly, the scope excludes:
 Conventional tailings approaches. In many cases conventional methods, such as a cyclone
sand centerline or downstream construction, are well proven and safe and will continue to
be used for tailings facilities in the future.
 Paste tailings (defined as tailings with minimal bleed water following deposition). While
there have been several successful relatively small-scale implementations of surface paste
technology, it has not scaled well to large tonnage operations. Due to the inherent
technology limitations, we do not expect widespread implementation of paste tailings for
large tonnage operations.
 The standard implementation of filtered tailings where the whole tailings stream is filtered.
The approach is well proven for relatively small tonnage mining operations but has not
been implemented yet for large scale operations. While we believe the challenges of
equipment and operation scale-up will be overcome in the future, in this paper an
alternative Engineered Tailings approach is offered which will allow for more effective
implementation using currently available filtration technology.
The paper scope also excludes methods such as underground mine backfill and in-pit disposal to
reduce the amount of tailings to be stored in the surface tailings facility. The Mount Polley Expert
Panel (Morgenstern et al, 2015) consider these methods as “Best Available Technology” and they
should always be considered as part of the tailings management plan where feasible.
As a final introductory comment, the industry widely refers to the tailings continuum ranging from
un-thickened through to filtered tailings with decreasing water content in the tailings. We note there
is a discontinuity in this “continuum” where the water content results in tailings which are too

2
concentrated for cost-effective pump and pipeline transport yet too wet to be transported
mechanically using a belt conveyor and stacker system.

PROCESS OPPORTUNITIES

Selective Mining

Historically selective mining has been described as high-grading or ‘robbing’ the mine to process the
highest grade possible. Increasingly, the term is being used to describe mining based on grade as well
as other ore characteristics that affect recovery and processing performance. The resultant tailings
properties are not always factored into the mine sequencing or economic models because of the
perceived lower value compared to other key indicators such as mineral grade and deportment.
However, with the increasing emphasis on tailings facility stability, these techniques can be extended
to allow mining operations to control tailings property variability for simplifying tailings dewatering,
transport and deposition processes. Selectively excluding or scalping materials known to create
challenges with the tailings system (e.g. high clay mineral content) from the mill feed could result in
significant simplification of the tailings system. The viability of this approach needs to be assessed
by weighing the benefits against the possible loss of recoverable minerals, but this practice will likely
increase in importance as mine operations face increasing pressure to shift to lower water content
tailings. The value and utility will depend on the scale of mineralogical variability in the ore body,
mining scale and rate, as well as the alignment of long-and short-term operational goals.

Pre-Concentration
Generally new ore deposits have lower grades requiring mining, moving and processing larger
volumes of material per ton of mineral recovered. Rejection of liberated barren gangue from the feed
by pre-concentration reduces the mass that needs to be processed (i.e. contacted with water) with the
consequent reduction in the mass of tailings to be processed and stored.
The aim of pre-concentration is to remove barren material as early in the process as possible. The
valuable minerals do not need to be fully-liberated and often it is only necessary for barren gangue
to be available for rejection (Duffy et al., 2015). There are several technologies appropriate for pre-
concentration including gravimetric, magnetic, optical, and radiographic techniques. The suitability
in each case depends on the ore properties. Pre-concentration has the added advantage that it can
complement selective mining at the bench or stope by determining grade or other mineral properties.
More valuable metal may be extracted from the resource while the processing plant treats less
tonnage at higher feed grade and generates less tailings for disposal.

3
Coarse Grind Processing

Traditionally, the optimum grind has been established by considering grinding costs (capital and
energy) versus recovery (mining operational income). As comminution technology has advanced
becoming more cost effective and metal prices have increased with demand, this trade-off has
generally led to finer grinding to increase recovery. This trend is further exacerbated by the fact that
remaining ore reserves are of lower head grade and often more finely disseminated. Finer ground
ores require more water for processing, the tailings are more difficult to dewater, and are more likely
to remain in a “fluid” state within the tailings storage facility.
There is now recognition that in establishing the optimum grind consideration needs to be given to
the implications for tailings dewatering, transport and storage. When these factors are included, it is
probable that coarser grinds will result in an economic optimum for the mine life (and a safer tailings
storage facility).
The difficulty of dewatering tailings is related to particle surface area, which increases exponentially
with decreasing particle size:
 Water absorption and adsorption increases with finer particle size. Clay particles, which
tend to accumulate in the ultra-fine -2 µm fraction with comminution, absorb and adsorb
even more water because of their high surface activity.
 The interstitial flow paths between particles required for dewatering become narrower and
more torturous as particles become finer.
 With finer grinds, it is more likely that minerals which have a negative impact on
dewatering are liberated (e.g. clay minerals).
 The concentration of very fine particles has a direct impact on the tailings rheology with
the concomitant impact on associated processes (e.g. effective flocculation).
Rather than simply selecting a coarser grind for the total ore stream, there are other options which
may more effectively limit the fines generated by the extraction process while improving pay-mineral
recovery:
 Designing flowsheets to facilitate early recovery of liberated mineral particles from the
comminution circuit to minimize their exposure to grinding (for example flash flotation in
the milling circuit). This prevents over grinding which may result in poor recoveries.
 Recovering middling (mixed mineral) particles early on for later selective grinding for
liberation and recovery. This results in only a fraction of the feed being ground finer
reducing the quantity of fines being generated. With this already common approach, the
focus is generally flotation of middling particles that can be captured with traditional
flotation technologies. However, the concept can be extended to even coarser middling
particles using flotation technologies specifically suited to such coarser particles.
 Alternatively, utilization of technology requiring only partial mineral surface access to
achieve full recovery (e.g. leaching), could be used.

4
Methods to Engineer Tailings

Paterson & Cooke have developed the Engineered Tailings concept in which the tailings flowsheet is
designed to optimize tailings characteristics for dewatering, transport, and storage. The concept has
wide application to improving tailings management:
 For systems where delivery of consistent solids concentration (or rheology) tailings slurry
is important, this can be achieved by blending thickener underflow with either thickener
feed or filter cake.
 EcoTailsTM, developed by Goldcorp and FLSmidth, combines waste rock and filtered
tailings to create a stable stack which is resistant to acid generation.
 For applications where it is important to chemically stabilize the tailings, improve
dewatering or develop early strength, cementitious binder, chemicals or other waste
streams can be mixed with the tailings slurry prior to deposition.
The following sections discuss two strategies for achieving an Engineered Tailings solution.

Classified Tailings
Tailings can be classified to separate the fines fraction from the bulk tailings, which can then be
treated separately to lower capital and operating costs. Desktop studies completed by Paterson &
Cooke have shown potential savings up to 50% for capital and operation costs when dewatering
classified tailings instead of whole tailings.
This is achieved by adding hydro-cyclones, hydro-separators or dewatering screens ahead of
dewatering. Many tailings operators are already familiar with classification technologies as they
operate cyclones to create construction sand from tailings for dam building. Paste backfill operations
also often uses classification to de-slime tailings before using in backfill recipes to optimize binder
consumption.
By splitting the tailings stream into two fractions, options such as filtered tailings become more
attractive. The de-slimed coarse fraction has a smaller tonnage, coarser particle size distribution and
minimal clays. This change in characteristics can significantly increase filtration rates and decrease
the relatively high capital costs associated with filtration plants. In some instances, the classification
step produces de-slimed tailings with a high enough solids concentration to feed directly to filtration,
eliminating the need for thickening. Removal of fines and clays reduces filter cake compressibility
which increases the filtration rate allowing simpler equipment to be used (such as recessed chamber
filters). Stackable filtered tailings are produced with less process variability, more consistent cake
moisture content and at significantly lower cost than if the whole tailings stream was filtered.
Classifying tailings does not necessarily result in an increase in water recovery. The slimes fraction
still needs to be handled, and this stream will have slower dewatering rates than the whole tailings.
However, if this fraction is relatively small, technologies that were previously considered impractical
due to low capacities become more feasible, such as screw press filters, centrifuges and geotubes.

5
The tailings storage strategy may be a combination of filtered tailings stacking and slurry tailings
stored in separate facilities. Alternatively, filtered coarse tailings could be used to build “paddocks”
and the thickened fines deposited within the interior.
The authors expect that classified tailings dewatering circuits will become a viable option for large
tonnage, low grade operations, especially in conjunction with the development of filters with larger
capacity to reduce the number of units required and thereby the costs.

Tailings Desulfurization
Acid generating (AG) or potentially acid generating (PAG) tailings limit the options for storing
tailings. Preventing oxidation and future acid mine drainage issues can significantly increase tailings
storage costs. Substantial savings can be realized if tailings classified as non-acid generating (non-
PAG) are separated from the main tailings stream.
Typically, the acid generating potential is due to the presence of pyrite and other sulfur bearing
minerals. Generally, they are in either the coarser or the finer particle size fractions following
flotation. In the coarser size fractions, only middling (mixed mineral) particles with some exposed
sulfur bearing surfaces are of concern, as fully encapsulated sulfur bearing minerals are not exposed
to oxygen and leaching. Sulfur bearing minerals in the finer size fractions are likely fully liberated.
Desulphurization is best achieved through flotation using tailored flotation equipment and
conditions. Prior to flotation, the tailings are split into coarse and fine fractions using hydro-cyclones
or hydro-sizers. Depending on the distribution of sulfur bearing minerals they are concentrated in
either the fine or coarse size fractions. Size specific flotation is used to recover the sulfur bearing
minerals:
 For coarse middling particles, typically hydro-cyclone underflow product, flotation
technologies which can operate at high solids concentration and are capable of “lifting”
coarse particles into the froth phase are employed.
 For the fine particle fraction, typically at lower solids concentration, flotation is done with
flotation machines capable of high shear and therefore capable of capturing the fine
particles.
With careful hydro-cyclone design, often both sulfur bearing particle categories can be concentrated
in the cyclone underflow, simplifying the de-sulfurization.
Often additional valuable pay-minerals are recovered during desulphurization; the economic
realizable benefit depends much on the proximity to the main plant, concentrate grade, other
minerals recovered and how much additional processing is required.

Distributed Tailings Deposition Systems

While thickened and paste tailings systems are most often implemented to achieve water savings as
the prime objective, there are numerous applications where a key objective is the development of
steeper tailings beaches to create a “stacked” deposit to maximize the tailings stored on the facility.

6
Early implementations of stacked hydraulically placed tailings were based on the Robinsky (1978)
concept where the deposit is developed by depositing the tailings through one or more centrally
located discharge points. These systems rely on the production of high solids concentration (i.e. high
rheology) tailings to achieve the stacked deposit. The concept has achieved mixed success for smaller
tonnage implementations (less than 30,000 tpd) but has not been successfully implemented for large
tonnage systems.
McPhail (2018) developed a stream power methodology for predicting tailings beach slopes and
recognized steep beach slopes could be achieved with low to moderate yield stress tailings if a
distributed tailings distribution system is employed. Stowe et al (2018) describes the process and
hydraulic requirements for a distributed tailings deposition system comprising multiple
simultaneously operating near equal flow rate discharges which are spaced sufficiently far apart to
ensure the individual discharges do not combine on the tailings beach. The system must
accommodate the inherent variability in process conditions and tailings properties (Martinson et al,
2015). Stowe et al (2018) present two large tonnage (≈ 100,000 tpd) case studies showing how the
implementation of distributed tailings systems increased beach slopes from 0.5% to above 2% for
tailings with low (11 to 17 Pa) and moderate (35 to 55 Pa) yield stresses.
The development of the distributed tailings distribution system is a significant advancement as it
allows high-rate or high-density thickeners to be used instead of paste thickeners which have not
operated reliably for high-tonnage applications.

In-line Flocculation

The high shear in centrifugal pumping of thickener underflow to the tailings storage facility generally
destroys the flocculant chains so that little benefit remains of the initial flocculation to promote
consolidation after deposition into the tailings storage facility. Yet flocculation after deposition will
promote water disengagement and consolidation.
To promote dewatering and consolidation of tailings after deposition, specifically designed polymer
is injected into the tailings pipeline before the discharge point. The turbulence in the pipeline is used
to achieve contact and flocculation. The polymer chemistry creates large macro structures of polymer
chains combined with tailings particles in a loosely held matrix which stacks well. The solids
concentration of the released water, generally well below half a percent of solids, is a function of the
initial solids concentration and the polymer chemistry selected for the application. Because of the
macro structure and the stacking, subsequent dewatering through self-weight consolidation is faster
than for the same material not flocculated.
While flocculation is more effective at lower solids concentrations, in-line flocculation technology can
be used for slurries with relatively high solids concentrations with the caveat generally, the higher
the solids concentration, the higher the polymer dosage. In-line flocculation dosages tend to be higher
than required for dewatering the same tailings in a thickener (with a dilute feed stream).

7
The macro structure created allows the tailings to be stacked which facilitates dewatering and
consolidation of the tailings pile. This stacking characteristic of inline flocculated tailings may make
it unsuitable where a long shallow tailings beach is required. The degree of stacking can be controlled
by tailoring the polymer dosage.
The macro structure improves the dewatering rate allowing the tailings to become trafficable sooner
than for conventional tailings. For small operations, the dewatering of an in-line flocculation facility
and subsequent land reclamation can be within the span of a few months.
While the capital costs are low, the operating costs of an inline flocculation operation are typically
higher than for tailings thickened to the same solids concentration due to the higher reagent costs
and labor requirements for the operations. In-line flocculation is an ideal tool for use during times of
emergency when the main dewatering process is unavailable. Further work is required to better
understand and quantify the long-term impact of the polymer on tailings consolidation.

In-line Tailings Amendment

At the Horizon oil sand operation, Canadian Natural has developed a method for injecting CO₂ into
their tailings pipelines to enhance the tailings settling rate. The addition of CO₂ reduces the tailings
slurry pH, thereby reducing the tailings rheology and so increasing the settling and consolidation
rate. This allows for quicker recycle of the process water reducing the volume of water stored on the
tailings facility; CNRL claim this process reduces the pond to half the size it would have been without
CO₂ injection (Canadian Natural, 2019). Canadian Natural has installed a CO₂ recovery plant at
Horizon’s hydrogen plant with the capacity to capture 438,000 tonnes of CO₂ annually. Research is
continuing to confirm that the CO₂ remains sequestrated within the tailings.

Carbon Sequestration

Power et al (2014) note that while the mining industry is a major greenhouse gas (GHG) emitter, due
to their mineralogical composition and reactive surface area the tailings from some mines have
potential to mineralize carbon and so offset GHG emissions. Research has focussed on tailings from
ultramafic and mafic ores which are amendable to carbon mineralization due to their high content of
magnesium-silicate and hydroxide minerals. Felsic ores may also be suitable for carbonation
reactions but due to their low reactivity little attention has been placed to date on these ores.
An investigation at the Mount Keith Nickel Mine in Western Australia has demonstrated that about
11% of the GHG emissions from the mine are currently being passively sequestrated by the tailings
facility (Wilson et al 2014). The tailings have the potential to sequestrate about ten times the mine’s
GHG emissions. The authors propose various strategies for accelerating carbon mineralization which
include changing the current tailings management practice to extend tailings exposure to the
atmosphere. This will require a larger surface area for the tailings facility and likely increase water
losses. Despite these challenges it is evident there is significant potential to utilize tailings to offset
not only the mine’s GHG emissions but also potentially the emissions of other industries.

8
Emerging Technologies

There are several technologies being developed to enhance tailings dewatering, these include
microwave, microbial and electro-kinetic dewatering. While these technologies have not yet been
implemented on a large scale, they will increasingly be considered for tailings projects where
reducing water content is a key-driver. Their application appears promising for the dewatering of
the fine tailings with Engineered Tailings.
There is opportunity to reduce the volume and/or improve the quality of tailings stored through
additional processing to create commercially valuable by-products. This is done successfully with
platinum and gold tailings (chromite and uranium) and possibly in the future with oil sand and
phosphate tailings (heavy minerals and uranium).

PROCESS TECHNOLOGY SELECTION

Clearly there is no single process solution for creating better tailings facilities and increasingly we
have more options in our “tool box” from which to develop appropriate tailings systems. It is
important that process, geotechnical, geochemistry and environmental disciplines work together in
an integrated team to ensure the best overall solution is identified for each unique operation. In
assessing the available options, the following factors should be considered from a process
perspective:
 Grind and mineralogy. These factors define not only which process options are suitable for
dewatering and transport the tailings, but also the geotechnical and geochemistry
performance of the tailings in the deposit.
 Tonnage. Some technologies are too expensive or complex for implementation for large
tonnage operations. Other technologies, such as paste thickening, have not scaled reliably
to large tonnage applications.
 Water and Energy. The quality, scarcity and cost of water is a key factor in assessing the
degree of dewatering that is economically feasible. However, energy requirements increase
with the degree of dewatering and so water usage and energy consumption need to be
appropriately balanced.
 Technology. There needs to be a recognition that implementation risk for new technologies
can be high and that this risk generally increases with tonnage. When possible, the simplest
solution is generally the best solution.

CONCLUSIONS
This paper has reviewed a range of technologies and approaches to create better tailings facilities in
terms of stability, environmental impact and societal acceptance. Specifically, the Engineered Tailings
concept is seen as the way forward to optimize the tailings dewatering, transport and storage

9
characteristics. There is no doubt that these methods will continue to improve because the future of
mining is dependent on how well we manage our tailings facilities.

REFERENCES

Canadian Natural (2019) Managing Tailings, viewed 31 Jan. 2019, https://www.cnrl.com/corporate-

responsibility/advancements-in-technology/managing-tailings.

Carneiro, A., A.B. Fourie (2018) ‘A conceptual cost comparison of alternative tailings disposal
strategies in Western Australia’, Proc. Paste 2018 – RJ Jewell and AB Fourie (Eds).

Duffy, K., W. Valery, A. Jankovic and P. Holtham (2015) ‘Integrating bulk ore sorting into a mining
operation to maximise profitability’, Proc. Metallurgical Plant Design and Operating Strategies, Perth,
Australia, 7-8 September 2015.

Martinson, R., J. Engels, and G.I. McPhail (2015) ‘High density thickened tailings - New design
challenges in transportation and deposition’, Proc. 3rd International Seminar on Tailings Management,
Santiago, 19-21 August 2015.

McPhail, G.I. (2018) ‘Beach prediction experience to date: further development and review of the
stream power-entropy approach’ Proc. 21st International Seminar on Paste and Thickened Tailings, Perth,
April 11-13, April 2018.

Morgenstern, N.R., S.G. Vick, D. Van Zyl (2015) “Independent Expert Engineering Investigation and
Review Panel Report on Mount Polley Tailings Storage Facility Breach”, Province of British Columbia.

Power, I.M., J. McCutcheon, A. L. Harrison, S. A. Wilson, G. M. Dipple, S. Kelly, C. Southam and G.

Southam (2014) ‘Strategizing carbon-neutral mines: A case for pilot projects’, Minerals 2014, 4,
pp. 399-436.

Robinsky, E.I. (1978) ‘Tailings disposal by the thickened discharge method for improved economy
and environmental control’, Proc. 2nd International Tailing Symposium, Miller Freeman Publications
Inc., San Francisco, pp. 75–91.

Stowe, J.Q, R. Cooke, I. Farrell and R. Martinson (2018) ‘Design considerations for distributed tailings
deposition systems’, Proc. Tailings and Mine Waste 2018, Keystone, Colorado, USA, September 30 – October
2, 2018, pp. 35-42.

Wilson, S.A., A. L. Harrison, G. M. Dipple, I. M. Power, S. L.L. Barker, K. U. Mayer, S. J. Fallon, M.

Raudsepp and G. Southam (2014) ‘Offsetting of CO 2 emissions by air capture in mine tailings at the
Mount Keith Nickel Mine, Western Australia: Rates, controls and prospects for carbon neutral
mining’, Int. J. of Greenhouse Gas Control 25, pp. 121–140.

10
The Use of Mobile Paste Technology for Crown Pillar
Stabilization in a Flooded Stope
Sue Longo, Thomas Skocir and Jonathon Taylor
Golder Associates, Canada

ABSTRACT

The use of paste for tailings disposal and stabilization of underground voids in an active mine is a
well-established technology practiced the world over. Paste can also be used for mine site closure
and remediation. In 2018, a mobile paste plant was operated at a closed mine site in Ontario, Canada.
Decades of mining activity at the site had left some of the near surface stope crown pillars with long
term stability concerns. A backfill program was designed to include the use of historic tailings that
were excavated from the tailings impoundment area. The tailings were used to produce paste for
backfilling the stopes from the surface via a series of targeted boreholes. One stope posed unique
challenges in that it was flooded with mine water. While mobile technology has been employed at
other closed sites, backfilling flooded voids is rare. Evaluation of the underground voids on site was
done with a Cavity Monitoring Survey (CMS) using a laser scanner. This technology, however, is not
suitable for flooded applications. The flooded void had to be evaluated using sonar technology.
Typical paste recipes had to be modified for subaqueous placement by using admixtures. A strategy
for capturing and dealing with water displaced out of the underground void by paste was required
to avoid contamination of the local watershed. These modified methods can be applied not only to
other closed sites, but also to active sites trying to deal with flooded stopes.

1
INTRODUCTION

At a closed mine located in the Province of Ontario, Canada, ongoing site remediation activities have
been performed over the last 10 years. The mine saw two phases of activity over a 30-year period and
was permanently closed in the 1980s. As part of the closure program, all surface infrastructure was
removed leaving only the underground workings – adits shafts and raises to surface – and the tailings
impoundment areas. All the surface openings were capped to the standard of the day during the site
closure in the 1980s. Upgrading of the closure measures on site began in the 2000s.
As part of the site remediation program, an investigation was completed of the near surface stope
crown pillars using a series of NQ (3-inch) diamond drill holes from which rock core was captured,
and a series of 4-inch percussive holes. CMSs were performed from each hole that intersected the
stope void in order to access the crown thickness at various points. The data from the investigation
was used to complete a stability assessment, from which seven stopes were identified as having
potential long-term stability issues.
It was decided that the best approach to stabilizing the crown pillars was to backfill the stope void
space with paste from the surface. Historic tailings excavated from the tailings impoundment area
were the primary source of backfill material. A series of 6-inch boreholes were drilled to intersect
each stope at key areas which would be used to deliver backfill.
Six of the seven voids requiring backfilling were situated on a ridge above the adit level of the mine.
The last stope, situated below the adit level and below the water table, was completely flooded. While
backfilling voids from the surface at closed mine sites using mobile paste technology is not
commonplace, it has been employed successfully at a number of different mine sites with established
technology and techniques. Backfilling a flooded void posed several unique challenges that required
modifications to mapping, recipes, equipment, and the backfill strategy.

METHODOLOGY

Flooded Stope Mapping

During the geotechnical investigation, existing mine drawings and maps were used to locate voids
on the surface that were to be investigated. However, these historic mine drawings provided only
one section view of the flooded stope.
For the dry stopes, boreholes were placed at various locations that would allow multiple overlapping
CMSs to be performed. The CMS consists of a laser scanner mounted on a dual articulating head to
allow for a full horizontal and vertical range of scanning. The resulting output is a point cloud that
can provide accurate details of the stope features, such as locations of ore passes, raises, and
interconnecting features with other stopes. These can be used to create a wireframe representing the
void space, from which the stope volume can be estimated. This device, however, does not work in
water. The flooded stope had to be scanned with a sonar unit developed for surveying this type of

2
flooded void space. One of the shortcomings of the sonar unit employed is that, unlike the CMS, it
does not have a dual articulating head, thus limiting the scans to horizontal only. To achieve a
3-dimensional point cloud, the unit is lowered an appropriate distance between horizontal scans. In
this stope, the sonar surveying was particularly difficult. Even with overlapping scans from multiple
boreholes, only the lower portion of the stope could be surveyed. Figure 1 shows sonar scan data
overlaid on top of an original section drawing of the stope. The sonar point cloud shows the locations
of six boxholes or draw points, not fully detailed on the original drawings. Missing from the sonar
scan is the top section of the void near the water level at approximately the adit level elevation, which
is a result of the limitation of the sonar technology. The adit level is presented by the grey dotted line
in Figure 1, and the water level in the mine is approximately 5 m below this elevation. The visible void
space from the sonar survey was estimated to be 3,900 m3.

Figure 1 Section view of the flooded stope with sonar survey data overlaid on top of historic mine drawings

Modified Paste Recipes

Two recipes were used on this project for backfilling voids. The first was a standard paste recipe
consisting of tailings, water, and binder which would be the primary recipe for bulk filling most of
the void volume. The other was a barricade recipe that would be placed into draw points or other
“leaky” areas. The barricade recipe is a low slump blend of tailings, water, binder, and 15% – 25%
aggregate. The aggregate in the recipe helps generate low mobility and stackability, along with some
early strength gains allowing for lifts to be placed in quick succession. The low slump combined with

3
the aggregate creates material that is less likely to flow out of the leak points and can choke off access
points relatively quickly.
These recipes had to be modified for subaqueous deposition in the flooded stope. After passing
through the borehole breakthrough, the paste would have to travel about 20 m through water to
reach the bottom of the stope. Previous testing had shown that a well-blended, homogenous paste at
a medium to high slump mix will tend to disperse when travelling through water. The binder,
tailings, and aggregate (aggregate for barricade recipes) will tend to separate and stratify on the
bottom. A binder such as Normal Portland Cement (NPC) is quite fine and will washout of the matrix
and stay in suspension the longest. The resulting lift will have layers of inconsistent strength.
In order to produce a paste mixture suitable for subaqueous placement, admixtures were added to
the recipe. Two admixtures were used:
 MasterRoc MF 510
 Master Matrix UW 450
MasterRoc MF 510 is a water reducing agent specifically designed for backfill applications. For a
given slump, it reduces the water content of the mix resulting in higher strength. For a given water
content, it also improves the workability or rheological properties of the mix aiding in the pumping
and placement of the material.
Master Matrix UW 450 is an anti-washout agent that prevents cement fines from being washed out
of the paste matrix. Originally designed for placing concrete under water, this admixture produces
similar results with paste. This admixture must be used in conjunction with a water reducing
admixture. The two admixtures combined produce a more cohesive paste with a lower water content
for a given slump, resisting washout of fines during placement in water.

Modifications to Mobile Paste Plant Equipment

Paste was produced on site using a continuous mixer truck. These trucks are self-contained plants
that typically produce concrete on site, and are equipped with an aggregate bin, binder bin, and water
tank. Materials are metered from the belt and hoses into a bowl and fed into a mixing auger. A picture
of a mixer truck can be seen in ¡Error! No se encuentra el origen de la referencia.. The truck was
used as a mobile plant with all materials pre-loaded on board to pour small barricade lifts at different
locations, but with limited capacity before it ran out of material.

4
Figure 2 A continuous mixer truck used to produce paste backfill on site. The mixing auger is positioned over
a borehole and is gravity pouring into the void below

Modifications to the trucks included upgraded hydraulics that allowed higher throughputs, and the
addition of a control valves on the cement auger for more precise control over binder addition. The
mixer truck has built-in admixture tanks, but for this application the viscosity of the admixtures made
it difficult to dose. Totes of admixtures were tied into the truck tanks to provide a steady feed. The
water injection point was relocated in the bowl to ensure mixing happened at the bottom of the bowl
and not on the edge of the belt.
For increased efficiency of large pours, the truck was positioned at an optimal location and all raw
materials were delivered to it.
From the paste delivery perspective, the mixer truck discharged paste from a chute at the end of the
mixing auger. In order to reach multiple boreholes without relocating, a boom pump was used to
pump paste from the discharge chute of the mixer truck to the boreholes. Boom pumps are used in
the construction industry to pump concrete where delivery by truck is not possible.
In this setup, the mixer truck can run continuously, if material can be supplied uninterrupted. This
equipment arrangement required significant space near the void with good access roads for material
delivery. This setup was used to fill the flooded stope and can be seen in ¡Error! No se encuentra el
origen de la referencia..

5
 Figure 3 Flooded stope paste production site

Flooded Stope Backfilling Strategy

The requirements for stabilizing the crown pillar/stope walls of the flooded stope were to tight fill
the void with paste with a compressive strength of 600 kPa. For tight filling to be successful, the paste
backfill must be in direct contact with the base of the crown pillar (i.e., no void pockets).
The first objective of the backfilling plan was to seal off the bottom of the stope from the rest of the
flooded mine. To accomplish this, four 6-inch paste delivery boreholes were drilled over several
boxholes to direct material to those specific locations. Other boreholes were centrally positioned and
were to be used for bulk filling the stope. The boreholes ranged in length from 20 to 30 meters.
The plan was to let paste mound underneath the delivery borehole in a boxhole. When this boxhole
was sealed, it would continue to mound up and spill over into the next one, and so forth. The boxholes
would be targeted with multiple small lifts of barricade material which utilized aggregate in its blend.
Each lift would be allowed to set up for a minimum of 3 - 4 hours before the next one would be placed
on top. This strategy would limit leakage of paste out of the boxholes into the drifts connecting the
rest of the mine.
Once this was accomplished, and the stope was sealed off from the mine, the stope could be pumped
out and a CMS performed on the void. The CMS would provide more accurate details of the void,
including the upper section which was not able to be scanned by the sonar unit. The CMS would also
identify any other openings in the stope that could be potential leak points of paste into other areas
of the mine.
After confirmation of the upper stope geometry, bulk filling of the remaining void could begin. If the
stope could be pumped dry, the admixtures used for making the paste suitable for subaqueous
deposition would no longer be required. The admixtures significantly increased the complexity and
cost of the backfill material.

6
RESULTS AND DISCUSSION

Flooded Stope Backfilling Campaign

Backfilling began as per the plan with barricade blend being poured into the flooded stope with the
objective of sealing the bottom openings, starting with Boxholes 1 through 3. The progress of the
backfilling activities was monitored by two methods: sonar scans of the void and depth sounding
using weighted tape measures. Depth sounding was performed in all boreholes to determine the
current location of the paste floor in the stope. This method proved to be quicker and less
cumbersome to perform than sonar scans.
Dewatering of the stope was a critical part of the backfilling plan. Soon after backfilling activities had
begun, a small submersible turbine pump was lowered down an inclined borehole. This inclined
borehole would locate the pump in an upper section of the void, well away from the paste backfill
activities. Pumping water out of the stope was started even before the bottom openings were sealed
off. This was done to ensure that the paste would not displace water in the stope through a borehole
to the surface, when the bottom openings were finally sealed off. Monitoring of water levels in the
stope during dewatering would give an indication of when the stope was sealed off, as the refresh
rate would drop below the pump out rate.
Sampling and testing of the water from the flooded stope showed that it was not appropriate to
discharge on surface and therefore a system was required to prevent any release. Preventing any
release from getting into the local watershed was a critical mandate of the closure program. The mine
water from the stope was pumped into the production site water storage tank. This water was used
for paste production to re-hydrate the dry tailings to the required slump.
After a few weeks, progress monitoring showed that accumulated paste volume was much lower
than the placed volume, beyond the contingency for estimated lost material. It was suspected that
material was being lost through one of the boxholes. It was decided that additional boreholes would
be required to provide more direct coverage over boxholes 4 through 6.
Later analysis showed that the settling/stacking of the paste did not follow a typical sub-aerial
regimen and, in general, was quite unpredictable in the flooded environment. Additionally,
inconsistency of weighted tape measurements due to the hard versus soft paste surfaces made for
difficulty in determining the paste floor and likely skewed the placed volume number.
After the new boreholes were drilled, backfilling resumed. Shortly thereafter, the water level in the
stope began to drop, signifying that the stope was now sealed off from the rest of the mine. After the
stope was dewatered to below the level of the highest borehole, a CMS of the upper void space was
performed. The CMS showed a much larger upper extent of the void that resembled the mine
drawings. Additional boreholes were added based on the CMS to allow for a better view of the
remaining stope that had not been visible from the first deployment. The complete view can be seen
in Figure 4.

7
 Figure 4 CMS (upper left) of the flooded stope imposed on top of the sonar survey (bottom right)

The new survey showed a relatively complete upper portion of the stope, at a different location than
what was illustrated on the historic mine drawings. Based on the new CMS data, two additional
boreholes were placed to intersect the highest point in the stope. One borehole would be used for
paste delivery to ensure that the backfilling operations would not create a pocket. The other borehole
would be used for dewatering and venting, allowing the remaining water to be pumped out of the
stope so that tight filling could be achieved.

Quality Control Monitoring

As part of the closure program, it was necessary to provide evidence to the regulating authority of
backfilling activity and progress. Of critical importance was testing of the paste backfill to show that
the required strength of 600 kPa at 28 days was being achieved. Since in-situ testing was not possible,
daily samples of paste were taken and cast in to 3-inch diameter, by 6-inch high cylinders.
In order to replicate the conditions underground which were approximately 10˚ C as measured
throughout the summer, the cylinders were stored submersed under water at 10˚ C. The cylinders
were cured for different durations (variations of 1, 3, 7, 28, 56 days), and then placed in a hydraulic
frame and compressed to failure. This gave the Unconfined Compressive Strength (UCS) of the cast
paste cylinders.

8
 Figure 5 UCS vs Curing Days for samples with and without anti-washout

Figure 5 shows the data from three different sets of cylinders. The two solid blue lines show paste of
the same recipe and slump, from different days, both with plasticiser and anti-washout admixtures.
The corresponding strengths are very similar. The dashed red-line represent material of the same
recipe and slump, but only using plasticiser. There is a marked overall decrease in UCS of this sample.
This illustrates the effectiveness of anti-washout admixture in subaqueous applications.
Over 350 cylinders were cast and broken for the flooded stope. Initial UCS data showed lower than
expected strengths. Binder content was increased from 7.5 wt% to 8.5 wt% to compensate. The curing
under water and at cold temperatures resulted in a delayed onset of strength in the paste. The 600 kPa
target strength was not being achieved after 28 days curing, but rather after 56 days. Except for a few
early lifts at 7.5 wt% binder, most of the cylinders broken exceeded their target strength after 56 days
curing.
Additional monitoring of the backfilling effectiveness was performed using post-production
confirmatory boreholes drilled into the ceiling of the lower segment of the void. Camera inspection
of these boreholes confirmed that the paste was tight to the back of the void and that no air pockets
existed.

CONCLUSION

Backfilling of the flooded stope took 10 weeks to complete. Initial early work required small lifts of
barricade recipe each day, followed by setup time which resulted in slow progress early on.
Preliminary evaluations estimated the void volume at 3,900 m3. After the backfilling was completed,
the actual volume of backfill placed in the flooded stope was 7,940 m3. The discrepancy between these

9
values can be attributed to larger than anticipated void volume due to the inaccuracy of the initial
sonar scans, and some leakage of paste through open boxholes into other areas of mine.
Paste backfill can be effectively utilised for backfilling flooded stopes but requires a different
approach than one applied to dry stopes. In this project, the water in the stope greatly impeded the
collection of data and understanding of the void characteristics.
If barricades or bulkheads separating the stope from the rest of the mine do not exist, construction of
one will be required. There are several ways to do this and, in our example, it was accomplished with
a careful application of a low slump aggregate paste. To achieve the required strength of the backfill
in a flooded stope, the use of plasticiser and anti-washout admixtures are necessary recipe
supplements to achieve consistent strength results.
Contingency and flexibility must be applied to the backfilling strategy for flooded voids to account
for the lack of and difficult to comprehend data inherent to this type of situation.

REFERENCES
Longo, S., Crossley, C. (2018) ‘Coal and Gold - How to Rehab Historic Mines for Urban Development’,
MineClosure 2018, September, Leipzig, Germany

Longo, S., Quintero, A., Kennard, D. (2015) ‘Mobile Paste Backfill Systems - A Decade of Work’,
Paste 2015, Cairns, Australia

Longo, S., Kennard, D., Taylor, J., Cull, J. (2017) ‘Abandoned Mine Rehabilitations - Case Studies’,
Tailings and Mine Waste 2017, November, Banff, Alberta

Ahmed, I., Francoeur, R., Anderson J., Brown, R. (2010) ‘Paste Benefits and Application from Mining
Industry to Land Reclamation’, 2010 International Solidification/Stabilization Technology Forum, June,
Sydney, Canada, 153 – 159

Crooks, J., Raine, M., Brown, R. (2004) ‘Paste Technology Overcomes Coal Mining Legacy’, Mining
Environmental Management, July 14 - 15

10
High Efficiency Filters for Tailings Applications
Kenneth Rahal, Dave Hanfland and Todd Wisdom
FLSmidth, USA

ABSTRACT

As more miners investigate filtration technologies for the dewatering of their process tailings, it is
important to understand the impact of filter efficiency on plant production and costs. Filter Presses
have seen increased interest due to their large capacities, which can have a significant CAPEX and
OPEX for any tailings facility. For filter presses, efficiency is defined as the ratio of productive time
vs unproductive time. Productive time is time the filter is dewatering slurry and unproductive time
is when the filter is preparing for the next cycle, typically called mechanical time, plus the time the
filter is in maintenance. Different filter press technologies include different operating philosophies,
different levels of automation as well as different mechanical designs which will impact efficiency.
Different filters also have different maintenance philosophies that effect availability which is when
the filter is not able to dewater slurry due to maintenance of the equipment. These combined
differences lead to different levels of CAPEX and OPEX, which need to be matched with the process
requirements to optimize return on investment. This paper will review the different filter press
designs, technology, operating and maintenance philosophies and use installation examples to
highlight the optimization of filter efficiency and costs.

1
INTRODUCTION

The mining industry has shown increased interest in filtered tailings in recent years due to notable
tailings dam failures and increased water scarcity. Filtered tailings has been seen in the past as a cost-
prohibitive process for larger throughput, lower grade, mines. However, as larger pressure filters are
developed, the economic case for filtered tailings has gained traction. The reduced risk and
consequence of failure of a filtered tailings impoundment versus a traditional wet impoundment also
aids in acquiring a “social license” to operate, which can reduce time for permitting. In addition, the
reduction in land usage and closure costs along with the possibility of progressive closure further the
economic case for filtered tails.
As companies investigate filtered tailings the focus has largely been placed on the correct design of
the tailings storage facility and required properties of the filtered material (Davies 2011, Ulrich and
Coffin 2015). There are many filtration options for dewatering tailings prior to disposal. The four
major categories of filtration equipment employed are centrifuges, belt presses, vacuum filters, and
filter presses. The maximum production for each technology is given in Table 1.

Table 1 Filtration Technologies.

Dewatering Technology Centrifuge Belt Press Vacuum Filter Filter Press

Highest Production per unit (MTPD) 1,500 1,500 7,200 20,000

The advantage of a filter press is that it can achieve much higher driving forces compared to a
centrifuge, belt press or vacuum filter. Centrifuges and Belt Presses also only produce a saturated
cake whereas with vacuum and pressure filters you have the option of de-saturating the filter cake.
Most pressure filters today will reach a driving force of 15 bar or greater. This allows it to achieve
lower moisture contents as well as higher throughputs. Lower moisture content is often needed for
stacking of tailings.
The disadvantage of a filter press is the batch type process which it uses whereas the other equipment
options are all continuous in nature. This batch discharge of filter cake makes it essential to
understand the efficiency of filters and to match them to process requirements for each individual
application. Efficiency can be broken into two components, process and maintenance (availability).
Process efficiency can be defined as productive time divided by the total time required for the
filtration cycle. The filtration cycle is made up of the process and mechanical components. Process
Components include filling the press with slurry, consolidating the slurry into a cake, washing the
cake (if needed), and air blow of the cake to reduce moisture content. Mechanical Components
include inflating/deflating membranes (if needed), opening the drip tray doors, opening the filter
press to discharge the cake, shaking the filter cloth to remove any hung-up cake, closing the drip tray
doors, washing the filter cloth and sealing surfaces, and closing the filter press.

2
The mechanical components of the filter cycle time do not contribute to the dewatering process.
There, the process efficiency can be defined as:

𝐹𝑖𝑙𝑡𝑒𝑟 𝐶𝑦𝑐𝑙𝑒 𝑇𝑖𝑚𝑒 − 𝑀𝑒𝑐ℎ𝑎𝑛𝑖𝑐𝑎𝑙 𝐶𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠 𝑇𝑖𝑚𝑒

𝑃𝑟𝑜𝑐𝑒𝑠𝑠 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 =
𝐹𝑖𝑙𝑡𝑒𝑟 𝐶𝑦𝑐𝑙𝑒 𝑇𝑖𝑚𝑒

The different design aspects of filters can also affect the maintenance strategy and procedures, which
in turn affect the efficiency of the filter. This paper will focus on cloth change and plate maintenance.
These aspects affect the efficiency due to their impact on the down time. Increased downtime in turn
increases the number of filters required to process a given throughput.

DISCUSSION
When considering the type of pressure filter to select, several factors must be considered related to
the slurry, production, product moisture as well as equipment specifics. The overall filter
characteristics typically drive the decision toward the type of pressure filter, however, mechanical
time and equipment efficiency should also be considered. The three classifications of filter presses
available on the market are a single plate shifting type, a multi-plate or pack shifting type and a full
stack opening type.
Tailings processing or dewatering of the tails requires the site to specify typical feed and output
conditions. Based on the specified conditions, the process for selecting the equipment can begin.
Below are criteria for consideration that impact the rate at which the slurry will dewater and its
impact on the cycle time or filtration area required to meet the specified conditions.
 Settling Rate – Filtration rates trend with settling rates. When filtration rates are higher,
the filter selection will be geared toward equipment with fast cycle times and low
mechanical times.
 Feed Solids – As the feed solids increases, the amount of liquid to be removed decreases.
Filtration cycle times will be reduced as feeds solids is increased.
 PSD – The particle size distribution or PSD will provide insight into how the slurry will
react to pressure filtration. A narrow band can impact cake formation by packing tightly,
limiting flow. Having a disproportionate number of fine particles smaller than 20 microns
can cause poor cake formation, with the fines limiting the flow through the cake.
 Particle Shape – When the aspect ratio of the face to edge increases, the particles are more
platelet in nature (Clays, Mica, etc.). This type of particle creates bridging and blinding,
drastically reducing filtration rates and limiting the effectiveness of the air blow step.
 Target Moisture – As the target moisture, liquid remaining in the cake, decreases it
typically requires a greater amount of energy and time. Low target moisture rates often
result in the need for more filtration surface area.

3
Once the basics of the slurry are understood and tested, the selection of the proper equipment
becomes an easier exercise. The filtration rate at the target moisture and tonnage determines the
direction of the filter selection. The trade-off is the relationship between the filtration cycle time and
the mechanical time for each type of filter equipment. The lower the overall cycle time, the more
important the filter efficiency becomes to the filtration process. There are three types to consider –
single plate shifting, multi-plate pack shifting or full plate stack opening.

Single Plate Shifting Filters

Single plate shifting filters have high mechanical time, as each plate is shifted individually for cake
discharge as well as for cloth washing. The type of cloth washing can vary; however, the typical
washer is designed with high pressure and is utilized to clean and rejuvenate the filter media on each
plate. Cloth washing is tailored to the application, but due to the time required to complete the wash
it may only perform one wash per day. Of the options to consider, the single plate shifting type filters
typically require the least amount of CAPEX, but when considering process efficiency, may require
more installed filters.
After completing the process portion of the sequence, the filter will open only one plate at a time to
discharge cake. Due to the small opening required a single plate shifting filter has the smaller
equipment length, 14.4 meters for 100 chambers. A small shifter will grab each plate and index it
toward the tail end of the filter. As it moves, the cake will release and the shifter moves on to the
next plate. In some cases, this type of filter could contain more than 150 plates, requiring over 30
minutes to index all 150 plates. Typically, a single plate shifter type of filter will not be selected unless
the total cycle times are greater than 30 minutes. The overall process efficiency for the single plate
shifting type filters is the lowest of the dewatering options.
The two primary design options available for single plate shifting filters are to use either recessed
chambers or membranes. Membranes are typically used for applications with compressible cakes
(pulp/paper, kaolin, etc.) or to consolidate the filter cakes if an uneven filter cake is formed inside the
filter chambers. Typically, feeding a filter from the upper portion of a filter plate will result in uneven
cake formation while feeding a filter from the lower portion of filter plate will not. Membrane plates
work by squeezing the filter cake, resulting in cakes with lower volume than the filter chamber.
Therefore, it is more efficient to produce consolidated cakes using feed pressure, which can be
accomplished with the use of higher feed pressures that can be accomplished with plates that are
designed with lower feed eye positions. If membrane plates are selected, FLS recommends that a non-
compressible fluid such as water is used to inflate the membranes to reduce safety risks. The use of
membranes increases the mechanical time of the filter, as shown in Table 2 below. This example is
for a single plate shifting filter with 100 chambers with a 90 minute process time, which would not
be uncommon for a single plate shifting filter. The filter media is washed every 10 cycles and this
time is divided across each cycle to obtain the average mechanical time. The efficiency is calculated
based on time only and does not include the loss in efficiency from reduced cake volume in
membrane filters.

4
Table 2 Single Plate Shifting Efficiency Comparison.

Design Process Time (min) Mechanical Time (min) Efficiency (%)

Recessed 90 18 83.3

Membrane 90 22 80.4

As can be seen in Table 2, the cycle time is 4 minutes longer for a filter with membrane plates,
resulting in a 3.4% loss in efficiency. If a process has longer process times, this loss in efficiency would
decrease even further, becoming less significant.

Multi-Plate Pack Shifting Filters

Plate pack shifting filters have similarities with the single plate shifting design. They tend to have a
high number of plates and utilize minimal space for opening and shifting the plate stack. This type
of filter typically utilizes a pack of approximately six filter plates. This type of filter could contain
more than 150 plates, requiring over 5 minutes to index all 150 plates, which is less than that required
by the single plate shifting filter. With the reduction in mechanical time, the typical process cycle time
is between 15 to 45 minutes. In this type of filter, shaking of the plates and low pressure washing to
flush the seal surface is more common. The plate pack shifting filter also utilizes a high-pressure
wash to restore the media, but will clean the pack of six plates at a time with one pass of the washer.
More frequent washing will lower the process efficiency requiring more filtration area and
potentially more filters. The overall CAPEX would fall between the single plate shifter and the full
stack opening filter.
Once the filtration sequence is complete, the filter will open and start to shift the plate packs. A shifter
will be indexed to each pack of plates. Each plate pack is shifted to discharge the cake and can
incorporate shaking to aid with stick cake removal. The shifter will index through the entire plate
stack, signalling the end of the cycle. A multi plate shifting filter requires slightly more equipment
length for 100 chambers, 18.2 meters. If a wash sequence is required, the plate stack is closed and then
opened again one plate pack at a time. A high-pressure washer engaged with the shifter will wash
one plate pack at a time. Overall, the operation efficiency is improved by shifting and washing packs
of plate.
Plate pack shifting filters are also available with either recessed and membrane chambers. A
comparison of the efficiencies for a plate pack shifting filter with 100 chambers, with a typical process
time of 45 minutes, is shown in
Table 3. The efficiency is calculated based on time only and does not include the loss in efficiency
from reduced cake volume in membrane filters.

5
Table 3 Plate Pack Shifting Efficiency Comparison.

Design Process Time (min) Average Mechanical Efficiency (%)

Time (min)

Recessed 45 10 81.8

Membrane 45 14 76.3

As can be seen in

Table 3, the cycle time is 4 minutes longer for a filter with membrane plates, resulting in a 5.5% loss
in efficiency. This drop in efficiency is more significant than that of the single plate shifting style due
to lower process time.
One of the differences with plate pack shifting that can improve efficiency and availability is to
consider a shifter that incorporates a shaker into its design. As the plate pack is shifted to fully open
each plate, the filter cloth is shaken. The design of the shaker can change the amplitude and
frequency to maximize cake discharge. Sticky cakes may require operator interaction to ensure full
cake release.

Full Plate Stack Opening Filters

Full plate stack opening where all the plates are connected to each other via links is the final type of
filter to consider. With this design the plates open all together, typically less than one minute. This
requires the filter to be much larger. A full plate stack opening filter requires significantly more
equipment length, 28.3 meters for 100 chambers. This type of filter normally includes shaking after
the plate stack is open to assist with cake discharge as well as a full plate wash after every cycle. The
opening speed, shaking and washing the complete filter in one step, all reduce the total mechanical
time, greatly improving the operating efficiency. It is common for excess capacity to be planned into
the filter sizing, due to the high availability, and this type filter may not require a spare filter.
The full plate stack opening filters operate essentially the same as the others. Once the cycle is
complete, the entire plate stack opens. After opening, the plates are shaken to free any remaining
pieces of filter cake. The drip trays close and the entire plate stack is washed. This can be done on
the plates or from the structure around the filter. The key is to sweep away solids from the seal
surfaces and to clean the filtration area. After a wash of 15-20 seconds, the press is closed and the
next sequence is ready to begin. Total mechanical time for a full plate stack filter is typically less than
four minutes. The highest operational efficiency and availability are achieved with this type of filter.

6
Full plate stack opening filters are also available with either recessed and membrane chambers. A
comparison of the efficiencies for a full plate stack opening filter with 100 chambers is shown in Table
4. The efficiency is calculated for time only and does not include the loss in efficiency from reduced
cake volume in membrane filters.

Table 4: Full Plate Stack Opening Efficiency Comparison.

Design Process Time (min) Mechanical Time (min) Efficiency (%)

Recessed 3 3.5 46.2

Membrane 3 6.5 31.6

As can be seen in Table 4, the cycle time is 3 minutes longer for a filter with membrane plates,
resulting in a 14.6% loss in efficiency. This drop in efficiency is very significant and can result in the
need for additional filters. Due to the low process times, it is important to minimize all non-
productive times such as opening and closing the filter, valve opening and closing, etc.

Selection

When it comes to selecting the appropriate filter for a given application, it generally comes down to
a trade-off of efficiency needs against CAPEX. Table 5 shows the different efficiencies for the 3
pressure filter technologies for different process times. Mechanical times are based on recessed plate
filter plates.

Table 5 Filter Type Efficiency Comparison.

Process Time (min) Mechanical Time (min) Efficiency (%)

Full Pack Single Full Pack Single

3 3.5 10 18 46.2 23.1 14.3

45 3.5 10 18 92.8 81.8 71.4

90 3.5 10 18 96.3 90.0 83.3

The full plate stack opening filter always has a greater efficiency than any partial plate pack shifting
filter. If the drop in efficiency is significant it can result in a greater number of filters required to
process a given throughput. This increase in the number of filters would result in increased operation
and maintenance costs. As can be seen in the table, the biggest changes in efficiency occur for short

7
process times where mechanical times make up a large portion of the total cycle time. The CAPEX
for the full plate stack opening filter is generally in the range of 1.25 to 1.75 times more expensive
than the plate pack shifting filter, which is in turn typically 1.25 to 1.5 times more expensive than the
single plant shifting filter. Therefore, it must be determined if the use of less efficient filters resulting
in an increase in the number of filters results in a higher CAPEX and OPEX than the use of fewer,
more expensive and efficient filters. An example of this trade-off is given for a process that has two
tailings streams as described in Table 6.

Table 6 Stream Properties.

Stream P80 TPH

Fines 10 200

Coarse 200 2000

Lab filter testing revealed two different process times which are shown in Table 7 along with the
corresponding cycle times for each filter technology.

Table 7 Pressure Filter Process and Cycle Times.

Fines Coarse

Filter Mechanical (min) Process (min) Cycle (min) Process (min) Cycle (min)

Single 18 18.5 36.5 4.5 22.5

Pack 10 18.5 28.5 4.5 14.5

Full 3.5 18.5 22 4.5 8

The resulting quantities of filters needed to process each stream is shown in Table 8.

Table 8 Quantity of Filters.

Filter Fines Coarse

Single 3 20

Pack 3 13

8
 Full 2 7

The process of determining which technology is most economic must include the length of mine life.
Therefore, the trade-off analysis was done for 5, 10, 20 and 30 years of mine life, no discounting for
NPV is included. The cost is normalized against the single plate shifting filter and shown in Figure 1
and Figure 2 for the fine and coarse streams, respectively.

2,00

1,80

1,60
Normalized Cost

1,40

1,20

1,00

0,80

0,60
0 5 10 15 20 25 30 35
Years

Single Pack Full

Figure 1 Cost Trade-off for Fine Tailings

1,10
1,00
0,90
Normalized Cost

0,80
0,70
0,60
0,50
0,40
0,30
0 5 10 15 20 25 30 35
Years

Single Pack Full

Figure 2 Cost Trade-off for Coarse Tailings

9
Figure 1 for the lower tonnage fines tailings reveals that the single plate shifting filter is more
economic for mines with less than 10 years of operation. The full plate pack opening filter is more
economic for mines with more than 10 years of operation. This switch is due to the higher labour
costs of the single plate shifting filter. As such, the point of change in which filter to select can change
between high and low labour cost regions.
Figure 2 for the higher tonnage coarse tailings reveals that the full plate pack opening filter is more
economic regardless of mine life. This is due to the significant difference in quantities of filters needed
to dewater this tonnage.

Maintenance

Another aspect to consider is the loss of availability due to maintenance. Most of the maintenance on
a pressure filter, approximately 80%, is associated with the filter cloth and filter plates. Along with
the cloth on filter plates the wash bars, plate rollers, membranes (if required for operation), and the
plates themselves require maintenance. Due to the high availability needs for tailings filtration, filter
design and maintenance practices should focus on reduction in filter downtime.
Cloth life should be selected to minimize the number of unplanned cloth failures. The goal should be
such that 99% of cloth changes are planned, and 1% is unplanned. Unplanned cloth failures have
significant flow-on effects to filter availability as failure of a cloth leads to solids behind the cloth,
these solids then pass through the filtrate ports in the plate (path of least resistance) and then build
up behind the adjacent cloth on the other side of the plate, the build-up of solids also causes the plates
to become deformed when the solids build up on the sealing surfaces and the plate pack is closed,
this then leads to premature slurry leakage at the sealing surface, failure the adjacent plate and failure
of the adjacent cloth, and this process continues until potentially all the cloths and plates are damaged
in the filter.
A key element to consider for full and multi plate stack opening filters is the ability to remove plates
for cloth change and/or plate repair. The filters with the highest availability utilize a design with
sidebars that allow the plates to be removed from the top of the filter. This is typically done with
quantities of 10, or even more, plates at a time. Removing the filter plates from the filter allows easy
access to the cloth on filter plates, the wash bars, plate rollers, membranes, and the plates themselves
which all require inspection and maintenance. As soon as the dirty filter plates are put into the
maintenance rack, new clean plates are lifted into the filter allowing the filter to operate while the
dirty plates are inspected and maintained. As the cloth and plate maintenance is performed outside
the filter, while the filter is operating, high availability of the filter is achieved. The total downtime
to remove a set of plates and reinstall a set of plates has been measured in an operating filter to be as
little as 10 minutes. The time to inspect a dirty plate and change the filter cloth on each side of a plate
has been measured to be 10 minutes per filter plate.
Overhead beam filters require plates to be removed from the side of the filter or cloth change inside
the filter. Other restrictions to plate removal include fixed flood wash assemblies and shifting

10
mechanisms incorporated into the structure. Managing filter cloth changes outside of the filter is
safer and improves the mechanical availability of the filter.

CONCLUSION

As more miners investigate filtration technologies for the dewatering of their process tailings, it is
important to understand the impact of filter efficiency on plant production and costs. Filter Presses
have seen increased interest due to their large capacities, which can have a significant CAPEX and
OPEX for any tailings facility. The process variables that affect the selection of filter presses are
Settling Rate, Feed Solids, PSD, Particle Shape, and Target Moisture. There are three types of pressure
filters to consider that have different levels of efficiency; Single plate shifting which are the least
efficient, Multi-plate pack shifting, and Full plate stack opening which are the most efficient.
Within these filter types, the plate design of recessed versus membrane impact the efficiency of the
filters. This impact on efficiency is the most significant for shorter process time applications. Due to
the differences in efficiencies and subsequent numbers required to process a given tonnage, the costs,
CAPEX and OPEX, must be evaluated to determine the appropriate filter type for each application.
Most of the maintenance on a pressure filter, approximately 80%, is associated with the filter cloth
and filter plates. Due to the high availability needs for tailings filtration, filter design and
maintenance practices should focus on reduction in downtime. The filters with the highest
availability utilize a design with sidebars that allow the plates to be removed from the top of the
filter.

REFERENCES
Davies, M. (2011) “Filtered Dry Stack Tailings – The Fundamentals” Proceedings Tailings and Mine
Waste 2011, Vancouver, BC Canada.

Ulrich, B. and Coffin, J. (2015) “Combined Tailings and Mine Waste” Proceedings Tailings and Mine
Waste 2015, Vancouver, BC Canada.

11
Considerations for Surface Disposal of Filtered
Tailings
John Lupo1 and Marcelo Mussé2
1. Newmon, USA
2. Ausenco, Chile

ABSTRACT
Interest in filtered tailings has grown significantly over the last several years, driven by social,
environmental, and financial concerns around tailings storage facility (TSF) stability, water
availability and water quality. Filtered tailings can provide benefits, such as 1) reduce water
consumption; 2) potential to stack on the surface, reducing the facility footprint; 3) no reclaim water
pool to manage, reducing the potential for accidental release and 4) potential to stack on top of
existing TSFs or waste dumps, thereby reducing the “real-estate” issue by requiring construction of
a new TSF.
However, the surface disposal of filtered tailings also come with their own risks which must be
considered in all phases of design, construction, and operation. While filtered tailings are often
perceived as being fully unsaturated, the tailings produced from filter presses and vacuum filters are
actually close to saturation (70 to 90 percent saturation) and undrained conditions can develop within
a filtered tailings stack if they are loaded too quickly (e.g. rate of rise), which could compromise its
stability. Filtered tailings placement must be scheduled to avoid development of undrained
conditions and ensure stability.
In addition, interruptions in filter plant availability require temporary storage of unfiltered or “out-
of-spec” tailings that must be handled separately and seamlessly to avoid impacting the mine
operation. Temporary storage of “out-of-spec” tailings may require a large, engineered basin, similar
to that for traditional slurry tailings. If the filter plant is out of operation for several weeks (or longer),
the mine operation is faced with managing two different tailings disposal facilities, which is an added
operational complication. Design of a filtered tailings stack must also consider the geometry to
manage surface water from precipitation, erosion, and reclamation.
This paper discusses considerations for the design, construction, and operation of facilities for surface
storage of filtered tailings, which can have significant influence on social, environmental, and
financial aspects for a mining operation.

1
INTRODUCTION

Interest in filtered tailings has grown significantly over the last several years. Much of this interest
has been driven by recent catastrophic tailings storage facility (TSF) failures (e.g. Mt. Polley in 2014,
Fundão in 2015, and Córrego do Feijão in 2019) raising social, environmental, and financial concerns
around surface storage of saturated tailings. In addition, global concerns around water availability
and water quality have become strong drivers toward maximizing water recovery in ore processing
and dewatering tailings.
While a number of technologies have been studied for dewatering tailings (MEND, 2017), filtration
of tailings using either mechanical press or vacuum belt, are the most common technologies being
employed today. Davies (2011) provides an excellent overview on the basics of tailings filtration.
Filtered tailings can provide a number of benefits, such as 1) reduce water consumption; 2) potential
to stack on the surface, reducing the facility footprint; 3) no reclaim water pool to manage, reducing
the potential for accidental release and 4) potential to stack on top of existing TSFs or waste dumps,
thereby reducing the “real-estate” issue by requiring construction of a new TSF. However, the surface
disposal of filtered tailings also come with their own risks which must be considered in all phases of
design, construction, and operation. These risks are mostly related to certain degree of saturation and
drainage of the filtered tailings mass and the potential to generate undrained conditions, leading to
instability.
In addition, to saturation and drainage risks, there are other external factors that need to be
considered in the design for surface disposal of filtered tailings. These external factors include:
 Interruptions in filter plant availability due to mechanical or other issues;
 Changes in ore mineralogy (e.g. increasing fines content) resulting in poor filtration;
 Changes in mill reagents (to increase recovery) resulting in poor filtration; and
 Temporary increase in mill throughput (i.e. financial pressures) resulting in decreased
residence time in the filter.
All of these external factors are likely to result in delivery of unfiltered or “out-of-spec” tailings that
must be handled separately and seamlessly to avoid impacting the mine operation. Typically, this
will require construction of a dedicated basin to store out-of-spec materials. The handling system for
these materials must also be addressed in the design and operation of the facility.
This paper discusses considerations for the design, construction, and operation of facilities for surface
storage of filtered tailings, which can have significant influence on social, environmental, and
financial aspects for a mining operation.

2
FILTERED TAILINGS BENEFITS

Reduce water consumption

The process of filtration converts slurry tailings (typically with 25 to 60 percent solids) to a “cake”
with approximately 80 percent solids, thereby recovering water that would have been sent to the TSF.
However, filtration systems (vacuum or mechanical) cannot recover all of the water in the tailings
stream, and generally produce materials with water contents of 15 to 30 percent [by weight] (Krizek,
2004, Kaswalder et al 2018, MEND 2017), equating to saturation levels of 70 to 90 percent. Since the
field capacity, the water content at which free-drainage ceases, of tailings may range from 8 to 25
percent (MacRobert, 2012, Swanson et al, 1999), seepage from some filtered tailings masses will likely
occur, requiring some level of collection system. Amoah (2018) noted that magnetite tailings from the
Karara Mine in Australia, drained to field capacity very quickly, developing very low air-entry values
to prevent further infiltration of precipitation. The drainage characteristics of filtered tailings is highly
variable depending on the ore mineralogy, particle size, and chemistry, and should be quantified as
part of the design process.
Davies (2011) noted that the non-recoverable amount of water in filtered tailings is approximately 25
to 50 percent less than with a conventional slurry tailings impoundment. In jurisdictions where water
resources are scarce and/or expensive to develop, filtration can provide a significant advantage over
traditional slurry disposal.

Stacking

Dewatering of tailings not only provides beneficial recovery of water, but it also results in an increase
in the shear strength of the tailings mass, thereby allowing the material to be stacked. As an example,
Lupo and Hall (2010) present strength data from several filtered tailings showing that, when
compacted, the effective friction angle may approach 35 degrees and the material behaviour becomes
dilatant. Additionally, the characteristics of filtered tailings are often such that it can be used as a
construction material (e.g. placed in compacted lifts at a specified density and moisture content),
thereby allowing stable landforms to be developed.
The ability to stack filtered tailings provides the following benefits:
 Significant decrease in the required footprint for storage of tailings;
 Concurrent reclamation of lower lifts as the tailings are stacked; and
 Development of a stable landform with “engineered” fill (e.g. compacted tailings) that
sheds rather than collect surface water from precipitation.
The ability to stack filtered tailings has also provided opportunities to re-use or retro-fit existing TSF’s
for additional storage. Case histories of converting existing TSFs to filtered tailings storage are
presented in Veillette et al (2009), Lupo and Musse (2014), and Meyer and Williams (2014).
3
While existing TSFs may be converted to filtered tailings storage, the behaviour of the tailings beneath
the filter stack must be thoroughly evaluated with respect to bearing capacity and liquefaction (static
and dynamic). In addition, water management within and around the tailings stack must be
addressed as the accumulation of excess water may lead to stability concerns.

Water management

A significant advantage of filtered tailings is the recovery of the majority of excess water during the
filtration process, resulting in the elimination of the reclaim pool or storage of process water on the
tailings surface. Elimination of the reclaim pool reduces the potential for accidental release by
removing one of the mechanisms for transporting tailings. Additionally, the need for return water
pumps, pipes, and other infrastructure is eliminated, thereby simplifying the overall tailings storage
approach.
While tailings filtration can eliminate the need for a reclaim water pool, this does not mean that water
cannot accumulate on the tailings surface. Precipitation will accumulate on the surface of the filtered
tailings stack. If left unchecked, these conditions could lead to erosion and/or physical instability. A
successful filtered tailings stack design will minimize accumulation of precipitation by grading the
tailings surface to drain in an engineered channel/spillway configuration.
From the Author’s experience, there is a perception within the industry that adopting filtered tailings
eliminates the need to conduct a water balance for the tailings circuit. In our opinion, a water balance
is required, regardless of what type of tailings disposal methodology is selected. While tailings
filtration recovers much of the free excess water at or near the mill, some water is still contained
within the filtered tailings mass that will likely report as seepage. In addition, some amount of
precipitation is likely to infiltrate into the filtered tailings mass which must be accounted in a water
balance so that overall water demand for ore processing can be estimated.

FILTERED TAILINGS RISKS

Saturation and undrained conditions

Filtered tailings are often perceived as being fully unsaturated, however, as noted earlier, the tailings
produced from filtration systems typically range from 70 to 90 percent saturation, with moisture
contents of 15 to 30 percent (by weight). As filtered tailings are stacked, the lower part of the stack
can become saturate, as noted in the case history presented by Ulrich and Coffin (2017). Filter tailings
stacks may become saturated from the following:
 During the stacking process, the tailings mass compresses from self-weight, resulting in
a decrease in porosity. The reduction in porosity may result in the development of

4
 saturated conditions at the bottom of the stack, where the greatest compression would
occur;
 As noted by MacRobert (2012) and Swanson et al (1999), the field capacity of tailings
typically ranges from 8 to 25 percent (though it can vary considerably depending on the
ore mineralogy). If filtered tailings are delivered with a moisture content greater than the
field capacity, excess moisture (moisture content minus field capacity) will slowly move
to the bottom of the stack, driven by gravity drainage conditions. This effect may take
several months or longer, as the unsaturated hydraulic conductivity is quite low and
decreases with decreasing moisture content; and
 Precipitation landing on the surface of filtered tailings will infiltrate into the mass. The
amount of infiltration will depend on the saturated and unsaturated hydraulic properties
of the tailings. Filtered tailings that are close to saturation (say 90 percent) can quickly
become fully saturated from infiltration of precipitation.
From a design and stability stand-point, it is critical to evaluate the potential for saturation to develop
within the stacked tailings from compression, drainage, and infiltration. The presence of saturated
conditions within the filter stack suggests that, depending on the loading (e.g. stacking) rate,
undrained or partially-drained conditions may develop near the base of the stack. Therefore, tailings
placement must be scheduled to avoid development of these conditions and ensure stability.

Out-of-spec tailings

Out-of-spec tailings are defined as filtered tailings that have a moisture content greater than the
design specification. Out-of-spec tailings are often generated due to one or more of the following:
 Loss of filtration efficiency due to partially clogged or damaged filters;
 Filtration plant is inoperable due to system failure;
 Change in ore mineralogy (e.g. increased fines, increased plasticity, etc.); or
 Financial pressures to temporary increase in throughput, reducing residence time in the
filter.
Any of these conditions will require temporary handling and storage of unfiltered or out-of-spec
tailings in order to avoid negative impacts to the mine operation. Temporary storage of out-of-spec
tailings typically require a large, engineered basin to be constructed, similar to that for traditional
slurry tailings. If the filter plant is out of operation for several weeks (or longer), the mine operation
is faced with managing two different tailings disposal facilities, which is an added operational
complication.

Surface water management

As noted earlier in this paper, stacking of filtered tailings must consider the management of surface
water from precipitation to avoid:

5
  Ponding of excess water on the tailings stack, which could lead to overtopping/flow
failures;
 Erosion of the tailings stack side-slope. Filtered tailings are generally readily erosive and
the ability to stack the material, resulting in steep slopes (some greater than 18 degrees),
give rise to conditions where rapid and deep erosion of side-slopes can develop.
 Damage to concurrent reclamation. Uncontrolled discharge of surface water can damage
existing reclamation works, thereby increasing cost and delaying concurrent reclamation
(one of the benefits of filtered tailings).

CONSIDERATIONS FOR DESIGN, CONSTRUCTION AND OPERATION

The following considerations are presented for the design, construction and operation for the surface
storage of filtered tailings:
From a design standpoint, it is recommended the following engineering work be completed:
 Pilot-scale filtration program be conducted on the full range of anticipated ore
mineralogies and thickener discharge (percent solids). The information from these pilot
tests will inform the design and potential variability in filtered products for storage. The
filtered products should be tested for dry density, moisture content, particle size
distribution, and Atterberg limits;
 One-dimensional, drained, compression tests should be conducted on filtered tailings
samples. These should be conducted in a large-scale (100 mm diameter or more) rigid-
mold (not small consolidometers) frame. The compression displacement versus load
should be measured and converted to dry density, void ratio, and saturation versus
stress. These results will provide information on settlement over time (for storage
capacity estimation) and the development of saturation within the stack (from self-
weight);
 Completion of moisture-density tests to develop dry density – moisture curves. These
curves can be used for construction of the facility;
 Unsaturated hydraulic property tests, including field capacity, conducted at the
anticipated dry density and moisture content in the stack;
 Saturated hydraulic property tests under load, conducted at the anticipated dry density
and moisture content in the stack;
 Triaxial shear tests, consolidated-undrained (CU with pore pressure) and
unconsolidated-undrained (UU). CU tests should be conducted at the dry density and
moisture content of the compacted tailings and at the anticipated dry density and
moisture content in the stack (e.g. uncompacted tailings). UU tests should be conducted
at the anticipated dry density and moisture content in the stack (e.g. uncompacted
tailings);

6
  Direct shear tests conducted at the dry density and moisture content of the compacted
tailings and at the anticipated dry density and moisture content in the stack (e.g.
uncompacted tailings);
 As with any tailings storage facility, a clear design criteria and design basis must be
established, including acceptance criteria for performance (e.g. piezometric heads,
settlement or deformation, etc). In addition, the operational criteria (e.g. stacking rate,
stacking plans, slopes, compaction, etc.) must be clearly defined prior to construction of
the facility. Finally, the design must include detailed construction specifications and
construction quality assurance/quality control program so that the design intent is
properly translated into the constructed facility;
 If an existing TSF is to be converted for filtered tailings storage, characterization of the
existing tailings via cone penetration testing (CPT), piezometer installation, tube
sampling and laboratory testing must be conducted. Characterization work should focus
on overall stability, liquefaction potential, and local bearing capacity.
From a construction stand-point:
 Implementation of the construction quality control (QC) and quality assurance (QA)
program;
 Periodic reviews and inspection report by the Engineer of Record (EoR);
 Internal (owner) periodic compliance reviews on the QA/QC program; and
 Tracking of design changes or modifications during construction. Any change or
modification to design must be approved by the EoR as well as the owner.
From an operation standpoint
 Implementation of the operational criteria;
 Periodic reviews and inspection report by the Engineer of Record (EoR);
 Internal (owner) development of critical controls for the facility. The critical controls are
controls that significantly influence the likelihood and/or consequence of an event. With
filtered tailings, critical controls may include: rate of rise, piezometric levels within the
stack, moisture content or saturation;
 Internal (owner) routine review on performance and reporting on the critical controls.
Finally, at any stage of the project, best available practices and techniques (BAT and BAT) must be
considered (see Garbarino et al 2018).

CONCLUSIONS

Filtered tailings can provide a number of benefits over traditional slurry, thickened, and paste tailings
such as greater reduction in water consumption, reduced footprint by stacking and/or conversion of
an existing TSF for stacking. However, it is equally important to recognize potential risks with surface
disposal of filtered tailings. Filtered tailings are often perceived as being fully unsaturated, the
7
tailings produced from filter presses and vacuum filters are actually close to saturation (70 to 90
percent) and undrained conditions can develop within a filtered tailings stack if they are loaded too
quickly. Additionally, temporary storage of unfiltered or “out-of-spec” tailings that must be
addressed in the design and operation. Finally, the design of a filtered tailings stack must also
consider the geometry to manage surface water from precipitation, erosion, and reclamation.

REFERENCES

Amoah, N., Dressel, W., and Fourie, A.B. (2018) Characterisation of unsaturated geotechnical
properties of filtered magnetite tailings in a dry stack facility, Paste 2018, Perth, Australia.

Davies, M. (2011) Filtered dry stacked tailings – the fundamentals, Tailings and Mine Waste 2011,
Vancouver.

Hore, C & Luppnow, D (2014), Karara iron ore TSF – Design considerations for a unique large scale
dry stack facility’, Proceedings of Tailings and Mine Waste 14, Colorado State University.

Garbarino, E., Orveillon, G., G. M. Saveyn, H., Barthe, P. and Eder, P. (2018). Best Available
Techniques (BAT) Reference Document for the Management of Waste from Extractive Industries,
European Union 2006/ 21/ EC.

Kaswalder, F., Cavalli, D., Hawkey, A., and Paglianti, A. (2018) Tailings dewatering by pressure
filtration: process optimisation and design criteria, Paste 2018, Perth Australia.

Krizek, R.J. (2004) Slurries in Geotechnical Engineering, 12 th Spencer J. Buchanan Lecture, Texas A&M
University, College Station, Texas.

Lupo, J.F. and Musse, M. (2014), Converting slurry tailings facilities to filtered dry stacks - a case
history, Paste 2014, Vancouver, Canada.

Lupo, J.F. and Hall, J. (2010), Dry stack tailings – design considerations, Tailings and Mine Waste,
2010.

MacRobert, C.J., (2012), A field study of beaching behaviour and the in–situ moisture regime of
tailings during active deposition, University of the Witwatersrand, in fulfilment of the requirements
for the degree of Master of Science in Civil Engineering.

Meyer, I. and Williams, C., (2014), Stacking dewatered tailings on a conventional slurry
impoundment, Paste 2014, Vancouver, Canada.

Swanson, D.A., Savci, G., Danziger, G., Mohr, R.N., Weiskopf, T., (1999), Predicting the soil-water
characteristics of mine soils, Tailings and Mine Waste 1999.

8
Ulrich and Coffin, (2017) Characterization of unsaturated tailings & its effects on liquefaction,
Tailings and Mine Waste 2017, Banff, Alberta.

Veillette, M.F., Martin, T.E., and Larrete, S.A. (2009), Stabilized upstream tailings dam and converted
into a filtered tailings facility, Tailings and Mine Waste, 2008.

MEND, 2017. Study of Tailings Management Technologies, MEND Report 2.50.1.

9
Economic Drivers for High-Capacity Tailings
Pressure Filtration
Matthew Pyle1, Richard Whittering2 and Greg Lane1
1. Ausenco, Australia
2. Ausenco, United Kingdom

ABSTRACT

Storage of process plant tailings in conventional wet tailings storage facilities is one of the most
significant environmental liabilities for mine permitting, operation and closure. Several recent
tailings dam failures with poor environmental and political legacies have raised the profile and public
awareness of tailings management for mines.
Filtration and dry stacking of process plant tailings can improve the social and environmental
outcomes of projects by reducing the risk of loss of containment, minimising site water consumption
and enabling progressive rehabilitation. Other benefits include smaller site catchment areas, reduced
runoff, reduced leachate production, less contamination of groundwaters and less onerous ongoing
monitoring.
In addition to the environmental and social benefits, high capacity tailings filtration can also provide
for improved business case outcomes, but recognising value requires a broad project view that
considers incremental savings across many project areas and cost centres. In this sense a ‘traditional’
project framework, where geology, mining, processing and tailings handling scopes are treated
independently (often by different sub-consultants at different times) can fail to identify and optimise
the key economic drivers for tailings filtration and draw an unfavourable conclusion.
This paper provides insight into the key economic drivers for tailings filtration projects, outlines
which projects are most suited to tailings filtration and discusses the project framework required to
assess and maximise the business case versus conventional tailings storage methods. Design
decisions around the filtration plant and dry stack are also outlined.

1
INTRODUCTION

Two key recurring issues in the public domain for existing and new mining projects are; tailings
stewardship and water management. These issues affect the profiles of the mine, company, industry,
community and country. The introduction of tailings filtration and dry stacking for some projects has
demonstrated the concept of tailings filtration, including the benefits of improved environmental
performance and reduced water consumption. Tailings filtration and dry stacking is now considered
a key benchmark for environmental best practice, with many new projects now requiring dry tailings
management to receive permits and their ‘social license to operate’.
For many projects that have already implemented tailings filtration, the economic trade-offs are
simple, the benefits are clear, and the business cases have been demonstrated by one or two
predominant factors, such as high costs of conventional tailings storage facilities (TSF) or water
limitations in arid environments. Other projects where tailings filtration is more easily justified tend
to treat high-value ores, with a high economic return for every tonne of fine tailings produced.
Examples include high grade, lower throughput operations (< 2Mt/y) often from underground mines.
For projects which are larger scale and treat lower-value ores, tailings filtration and dry stacking may
not be justified by one or two factors alone. For these sites, filtered tailings may be concluded to be
less cost-effective than other tailings disposal options, but this “has sometimes resulted from
evaluations that do not incorporate all the costs that may potentially affect the finances of a project”
(Carneiro & Fourie, 2018). To maximise the value of tailings filtration and dry stacking, a more
detailed consideration of the range of benefits, project drivers and synergies across multiple project
areas is required. Understanding the trade-offs can become more complex and require an approach
that integrates a wide range of skill sets. Such expertise is not always available within a single
company. Furthermore, the economic drivers that maximise the value for a project with tailings
filtration may influence the project to select different mining strategies, mine plans, processing
strategies and circuit designs than a project without tailings filtration, complicating the evaluation
and decision-making process. The outcome can be incorrect decisions with the consequence of poor
economic, social and environmental legacies.
This paper distils the key economic drivers and discusses the required project framework to assess
tailings projects with high capacity tailings pressure filtration.

ECONOMIC DRIVERS FOR TAILINGS FILTRATION PROJECTS

Projects most suited to high-capacity tailings filtration have the following characteristics;
 Tailings which are amenable to filtration
 Limited availability or high cost of water
 High costs of conventional tailings storage, including embankment raises
 Synergies between tailings filtration and other processes, such as metal or reagent
recovery, or downstream heap leaching

2
  Commonality with other infrastructure (i.e. waste rock storage, in the event of co-
disposed tailings).
 Low operating cost drivers (power, labour)
 Low capital cost drivers (efficient engineering and execution)
 High economic margins per tonne of fine tailings, based on the orebody and project.

The following sections outline key economic drivers that affect the assessment of tailings filtration
and dry stacking compared to conventional tailings storage methods. These drivers are related to:
 Financial assessments
 Water management costs
 Dry stack design decisions
 Filter plant design decisions
 Commodity and grade
 Particle size and clay content
 Reclamation and closure
 Risk-based costs.

Financial Assessments
Tailings filtration and dry stacking trade-offs compete against more conventional methods of tailings
placement, including slurry and thickened tailings deposition within tailings dams. The cashflow
profile for tailings filtration and dry stacking requires significant upfront capital expenditure, but can
provide operating cost savings throughout the project life and closure costs that are a fraction of those
for conventional tailings storage facilities. Conventional dam storage cashflows have low up-front
capital costs, and operating costs that progressively increase throughout the project life, followed by
large closure costs and ongoing monitoring and maintenance for many years.
Even if the life of mine costs in real terms for conventional tailings deposition and tailings filtration
and dry stacking were identical for a project, methods for financial assessments (including NPV,
NPV/CAPEX, IRR, or payback period) overwhelmingly favour the cashflow profile of conventional
tailings storage due to the discount rates applied to deferred expenditure. The source of project
financing and selecting a discount rate (typically between 5% and 18%) is consequently a key
economic driver, with tailings filtration and dry stacking heavily penalised by expensive capital and
high discount rates.

Water Management Costs

Where the supply of water limits the throughput of the operation, tailings filtration is a viable method
to directly increase revenue by reducing water consumption and this yields a highly positive business
case. If water can be sourced but the costs of supply are high, tailings filtration provides an
incremental cost saving.

3
In arid regions, concentrator water consumption for conventional tailings can average over 0.7 m 3/t
in a given year. Tailings filtration presents an opportunity to reduce water consumptions by
approximately 0.5 m3/t to less than 0.2 m3/t. For large South American copper concentrators where
water is desalinated and pumped up the Andes, water supply costs can exceed US$3/m3. This
effectively offsets the costs to filter tailings by US$1.50/t and reduces the capital costs associated with
water supply infrastructure.
For operations in cold climates, winter months can cause tailings dams and water storage facilities to
freeze which limits the availability of water. For these projects tailings filtration simplifies water
management and also reduces water consumption.

Dry Stack Design Decisions

There are many contributing and influencing factors that affect the cost of the filtration plant and dry
stack facility. Some early questions that need to be resolved include:
 What is the stack design and stacking plan?
 What material will be stacked within the dry stack?
 How will material be placed, and how much desiccation will occur?
 What filtered tailings moisture is required to satisfy geotechnical requirements?
Two common methods for tailings storage are filtered tailings only contained either by compacted
tailings or waste rock buttresses, or co-disposal of filtered tailings and coarse waste rock. In mine
waste management, “co-disposal” refers to the disposal of tailings and mine rock streams in one
integrated disposal facility. When filtered tailings are stored without waste rock, tailings are typically
dewatered to within 3% (dry basis) of the optimum proctor moisture (or lower if required to satisfy
geotechnical and saturation constraints). When filtered tailings are co-disposed with waste rock,
filtered moistures can be as high as practicable. This is typically around saturation and is constrained
by cake release from the filter and materials handling.
Co-deposition of tailings may provide for some capital cost savings in the filtration plant through the
relaxed target moistures. However, additional infrastructure for waste rock handling may be
required, particularly if the waste rock and tailings production profiles aren’t well matched.
The dry stack design must suit the optimum placement method(s) (trucked or machine stacked),
which is a function of transportation distances, topography and cost trade-offs for the life of mine.

4
Filter Plant Design Decisions

Once a target moisture has been selected, consideration of the following questions can help determine
the optimum filtration strategy:
 What is the optimum thickener technology and filter feed density?
 What is the optimum filter feed pressure?
 What is the optimum chamber thickness?
 Is membrane pressing required, and if so to what pressure and for how long?
 Is a cake air blow required, and if so for how long and what is the air consumption?
 Are wash cycles required (to recover soluble metals or reagents)?
 What is the best media that optimises filtration performance, filter availability and
minimises cost?
These questions are inter-related and rely on test work and engineering trade-offs to optimise the
answer in conjunction with the dry stack design for a given project and filter design.
Two common filtration strategies offered by different vendors are:
 Single stage pumping (nominally 8 bar), with membrane squeeze (nominally 10 bar) and
concurrent air blow as required
 Two stage pumping (nominally 15 bar), without membrane squeeze, air blow as required.

Key influencing factors driving the filtration strategy are material filtration characteristics, feed
density, required moisture, climate, filter design, filter cake properties, operating flexibility, cloth
type and plate design, availability, reliability, maintenance, altitude, power cost, and capital and
operating costs.
Maximising the filter feed density by selecting high compression thickeners in lieu of high rate
thickeners can be a significant value driver for minimising the total thickening and filtration facility
capital and operating costs, particularly for slower filtering ores. Membrane pressing or high-
pressure pumping can also reduce the air consumption required for blowing of the cake, which is
important for projects which have high power costs or are at altitude. For other projects, the cost of
labour is a key value driver and filters that have filter cloths which are free-hanging and removed
from above the filter rather than cloths which are affixed to the plates or removed from the side can
reduce the site labour costs.
Filter cloths are typically fabricated from felt or woven synthetics, however new media types are
under development by vendors and third-party suppliers. Media cost can vary significantly and life
can range from 800 to 5,500 cycles, with most vendors suggesting 3,000 cycles for base metal tailings
applications. Some applications benefit from intermittent high-pressure cloth washing to rejuvenate
the cloths and extend cloth life.

5
Commodity and Grade

Projects where each tonne of tailings enables high revenue are more likely to afford tailings filtration
and dry stacking projects compared to those where the margin per tonne of tailings is lower. For
instance, for thermal coal projects, the operating margin for each tonne of filtered tailings (excluding
coarse reject) is relatively high at US$300 per tonne. Therefore, the additional incremental cost of
tailings filtration (at say $1.50/t) is relatively insignificant, representing only 0.5% of the operating
margin. High grade copper, gold and zinc projects and to some extent alumina refineries can also
operate at sufficiently high margins such that tailings filtration circuits do not impose a large impact
on project margins even with no other contributing factors. These projects can justify more
conservative filter selections and filter plant designs.
However, iron ore, nickel and lower grade copper and gold deposits generally rely on higher
throughputs and economies of scale to maintain operating margins. For these projects, thickening,
filtering and stacking tailings at an incremental operating cost of US$1.5/t or greater has a significant
effect on project economics. Projects at lower grades are, therefore, unlikely to support tailings
filtration without offsetting savings in other areas such as water, power or alternative tailings
handling costs. Low grade projects, where economies of scale and low operating costs are required
to make the project viable must work much harder to identify business case benefits, optimise the
filter design and selection and minimise operating costs as much as possible.

Particle Size and Clay Content

There can be significant variability in the amenability of different ores, or even different lithologies
within a single orebody for filtration. A good proxy for filtration performance is observing the
amount of material passing 10 µm in the filter plant feed. Tailings with <15% passing 10 µm generally
filter at high rates and can achieve low filter cake moistures. Tailings with higher fines contents, say
>30% passing 10 µm can show poor filtration rates and high moistures with substantially longer
filtration times required.

6
 Figure 1 Impact of particle size on tailings filter throughputs

Classical filtration theory, including the Darcy and Kozeny-Carman equations relate filtration
amenability to the effective capillary size within the filter cake, which is related to particle size
distributions. For increasing fines contents, the average capillary size reduces, as does the flow of
filtrate through the filter cake for a given filtration pressure.
Two key influencing factors for the proportion of fines in plant tailings are ore mineralogy and
processing. Aluminosilicates (clays) are fine grained and quickly comminute into micron-sized
particles, and therefore often directly related to the percentage of fines in tailings. Fines may also be
generated through grinding and regrinding the ores, with the type of comminution device and circuit
design also contributing to fines generation.
The type of clays is worth considerable attention. Clays are colloidal particles and form network
structures within slurries, increasingly so at higher densities. Clay types and concentrations typically
vary based on geological events, oxidation and depth from the surface or structural controls.
Different clay types can form very different types of network structures, which can serve to physically
and chemically lock water within the tailings filter cakes and introduce materials handling problems.
The behaviour of clays can also change significantly for different water chemistry; pH and ionic
strength are two key drivers.
Clay contents can be indicated through a range of measurement techniques, including XRD,
QEMSCAN and CEC, although not all measurement methods are accurate. Even without measured
clays, graphs of thickener underflow density versus yield stress often give an excellent indication of
the propensity of the clays to form network structures and the likely amenability to filtration. It is not
unusual to observe strong correlation between the % passing 10 micron, clay contents, yield stress
curves and filter throughputs.

7
Other ore and processing factors that can negatively affect filtration amenability of ores include
particle shape (particularly if fine, needle-like particles are present) or precipitates that can blind filter
cloths, reducing filtration throughputs and increase operating costs.

Reclamation and Closure

When compared to conventional tailings storage facilities, compacted filtered tailings can achieve
higher in-situ densities, enabling storage within difficult topographies and reducing the footprint of
conventional tailings storage facilities. For the Greens Creek Mine, the required footprint was
reduced by 78% when compared with a slurry-type tailings impoundment (Condon & Lear, 2006).
Closure costs for conventional TSFs can be four times that for an equivalent dry stack (Carneiro &
Fourie, 2018) or higher if extensive reshaping of the TSF structure is required for long term landform
stability.

Risk-based Costs

As evidenced by several recent and high-profile tailings dam failures, the cost of large tailings dam
failures includes government fines, compensation to families and community, environmental clean-
up costs, image and reputation, company market value, loss of production, as well as wider reaching
costs associated with regulation changes driving increased compliance costs for other dams. The total
cost of a large dam failure can easily run into tens of billions of dollars (USD).
The risk of tailings dam failure is typically considered to be so low that the cost of tailings dam failure
is not considered when assessing alternate tailings storage methods. However, studies have shown
failure rates of tailings dams over the last 100 years to be 1.2% (Azam, 2010). Dams with upstream
construction, unusual rain conditions and poor management were identified as key factors.
Although not all failures are high-profile and cause significant loss of containment, the risks and
consequences are high enough that they should be included in any assessment of tailings storage
methods. For example, a recent high-profile upstream iron ore tailings dam failure could be assessed
to have caused approximately USD 50 Bn worth of lost value. If we consider an annual thoughput of
40 Mt/y and an average tailings dam failure rate of 0.012% per year, the risk cost of an average
conventional tailings dam failure works out to be US$0.15/t. This is a significant contributor to the
overall economics and indicates that probabilities and consequences of tailings storage failure should
be included in trade-offs between filtration and dry stacking and conventional options based on key
factors.

PROJECT FRAMEWORKS TO EVALUATE TAILINGS FILTRATION

As tailings filtration projects transition from low throughputs and high grades to lower grades and
higher throughputs, the mining industry needs to continually challenge and optimise the business
case for tailings filtration projects to maximise margins. Conservative design criteria, engineering and

8
assumptions, as well as failing to include synergies with other areas of the project will erode the
business case value and lead to suboptimal management decisions, with potentially tragic legacies.
A ‘traditional’ project development pathway starts with exploration drilling, geology, basic
metallurgical test work (often on composite samples) and mine planning for the resource. Pre-
feasibility studies are then completed using basic engineering test work on composite samples
coupled with in-house benchmarking data from engineering companies to determine the financial
viability of the deposit and bring resources into reserves. As more detailed feasibility studies are
completed, comminution variability tests are completed, geometallurgical comminution proxy
models are developed, and the mine design is optimised in parallel with the plant design to maximise
value.
Projects under development may not have considered tailings filtration and dry stacking, but are
now assessing the technology given potential economic benefits, permitting constraints and social
licensing considerations. Specific focus on clay types and concentrations by geologists and mining
engineers is typically not considered a key project value driver and clays are generally not included
in the block model and mine plan. Ore typing based on comminution parameters, metal associations
or geological features may also mask filtration responses, and grind sizes and flotation schemes are
selected solely on metallurgical performance without considering tailings filtration and dry stacking
impacts.
Composite samples used for basic engineering test work in early study phases by different
independent labs or vendors with different methods, at different feed pressures and feed densities
are difficult to compare, may not represent the optimum filtration strategy and may not be useful to
support proxy model development. Tailings samples used for filtration test work should be
representative and tested using flotation tailings slurry, however often tailings samples from float
circuits don’t include reground cleaner-scavenger tailings, or are filtered for shipping and
reconstituted with lab water for filtration testing, losing fines and changing the slurry behaviour.
Many vendors offer to conduct filtration test work free of charge but then provide test work results
that focus on their equipment sizing rather than the true test work results. Vendors conduct
equipment sizing using different assumptions and apply different levels of conservatism. Key
information required to fundamentally understand filtration characteristics is sometimes withheld,
such that test work results cannot be easily interpreted and used by engineering companies or other
vendors.
High capacity tailings filtration can be the highest capital cost area for a project. The mine
geometallurgy, mine planning, ore scheduling, blending and tailings logistics should be redefined
and re-optimised to maximise value considering tailings filtration as a revenue constraint. This
requires simultaneous and high-level expert input on all project aspects, to trade-off and balance
mine design, mine plan, tailings stack design, stacking plan, geotechnical constraints, process design,
operations, logistics and financial modelling to achieve the best project outcome.

9
The consequences of a ‘traditional’ project approach can limit the efficacy of project development for
tailings filtration due to a lack of focus on key information, limited useful datasets, lack of specific
expertise, and staggered (rather than simultaneous) involvement of experts in a siloed fashion with
limited ability to communicate and conduct cost trade-offs across project areas. Mining companies,
project managers and engineers should therefore;
 Ensure geologists and mining engineers are aware of, and focus on the likely proxies for
tailings filtration
 Conduct well-managed filtration test work on a range of variability samples
 Seek to develop proxy models that predict tailings filtration throughputs and feed those
back into the mine block models
 Engage early with filtration experts, filtration vendors, stacking system vendors, tailings
and geotechnical engineers
 Conduct trade-offs and optimisations that include the mine plan, stack design, process
design, filtration strategy and filter design
 Evaluate and include all potential benefits and synergies that tailings filtration and dry
stacking affords.

CONCLUSION

High capacity tailings filtration and dry stacking has the potential to achieve improved economic,
social and environmental project outcomes, but evaluating and successfully implementing the
technology raises a number of emerging challenges for projects looking to filter and dry stack lower
grade ores at higher throughputs. Knowledge of the key economic drivers and an assessment
framework that considers incremental savings across many project areas and integrates experts
across multiple disciplines simultaneously to optimise the business case is fundamental to justifying
tailings filtration. Total lifetime costs for construction, operation and closure of filtered tailings dry
stacks can be competitive and provide a solution that reduces TSF footprint and enables progressive
reclamation, simplifying permitting and EIA approvals.
Based on the key value drivers, tailings filtration projects should optimise towards higher mine strip
ratios to maximise concentrator head grade, pre-concentration and waste rejection at coarse particle
sizes where possible. Selecting comminution technologies that liberate at coarser sizes, and
separation technologies that preconcentrate values at lower levels of liberations, minimises the
amount of fines in the tailings stream and can add significant value to a project
For new mines, social and political pressures and regulations are likely to increase the costs of
conventional tailings storage, whilst driving projects to minimising water consumption, minimising
tailings dam risks and increasing focus on progressive rehabilitation. In parallel, filter technology
development is reducing the capital and operating costs of tailings filtration. These factors are likely
to continue and drive the economic evaluation of tailings storage methods away from conventional
tailings storage and toward high capacity tailings filtration and dry stacking.

10
REFERENCES

Azam, S., Li, Q. (2010) Tailings Dam Failures: A Review of the Last One Hundred Years. Waste
GeoTechnics. Geotechnical news.

Carneiro, A & Fourie, AB. (2018), 'A conceptual cost comparison of alternative tailings disposal
strategies in Western Australia', Proceedings of the 21st International Seminar on Paste and Thickened
Tailings, Australian Centre for Geomechanics, Perth, pp. 439-454.

Condon, P & Lear, K. (2006), ‘Geochemical and Geotechnical Characteristics of Filter-Pressed Tailings
at The Greens Creek Mine Admiralty Island, Alaska’, Proceedings 7th International Conference on Acid
Rock Drainage, ICARD, St. Louis, pp. 350-363.

11
Considerations for Preparing Design Criteria for
Dewatered Tailings Facilities
Juan José Moreno and Samuel Kendall
SRK Consulting, Australia

ABSTRACT
Dewatered tailings facilities (thickened, paste, filtered) are generally considered safer alternatives to
conventional tailings disposal, primarily due to the lower amounts of water stored as a pond or
interstitially within the tailings mass. While it is acknowledged that a risk-based approach to TSF
design should always be followed, the pressure to deliver projects quickly and often within limited
budgets creates a tendency to rely on standards-based design criteria which may not always be
directly applicable to dewatered tailings facilities.
Key design criteria commonly adopted from conventional TSF design are driven by the Population
at Risk (PAR) and Dam Failure Consequence, slope stability Factors of Safety (FOS), Stormwater
storage and Earthquake Loading. For dewatered tailings facilities it is important for stakeholders to
recognise that prescriptive design elements, typically adopted from conventional TSFs, may not
represent the most critical sources of risk, and therefore selection of design criteria should be case-
specific.
There is arguably a void in international guidelines with respect to incorporating a design
flowchart to streamline the design of dewatered tailings facilities. This paper discusses current
practice for evaluating key TSF design criteria, taking as a base leading guidelines such as MAC
(2019), ANCOLD (2012) and CDA (2007, 2014), and relates this to the context of dewatered tailings
facilities. It identifies important elements to consider, issues likely to be encountered and areas of
improvement, and provides a basis for continuing research and discussion.

1
INTRODUCTION

The devastating impact of catastrophic tailings dam failures has brought increased focus to risk-
based design processes, especially for high-risk facilities. The implementation of risk-based criteria
at the design stage is considered industry best practice. The aim of a risk based approach is to align
the TSF design with the level of risk acceptable to the designer, the owner and the community
impacted by the construction of the facility.
In the search for alternative tailings storage solutions that carry lower risk and less severe failure
consequences, dewatered tailings management systems such as thickened/ paste and filtered
tailings present as an attractive option because the reduced moisture content and smaller volumes
of free water promote increased water efficiency and result in more stable landforms.
Despite being internationally recommended, risk-based design is not always practical to achieve,
with some dam owners and designers choosing to implement prescriptive criteria defined by
international standards and guidelines as a quick and convenient design methodology that is
unlikely to be challenged by peers or by government agencies.
This paper discusses the prescriptive design criteria for TSFs, considering current tailings
management guidelines from relevant mining jurisdictions, and relates this to the context of
dewatered tailings facilities. It identifies important elements to consider, potential issues to be
addressed and possible areas of improvement.

METHODOLOGY
This paper is based on a limited review of current leading design tailings guidelines and design
codes from around the world, including the best practice guidelines from agencies across a range of
jurisdictions where mining is a relevant industry listed below:
 Australian National Committee on Large Dams (ANCOLD)
 Canadian Dam Association (CDA)
 The Mining Association of Canada (MAC)
 Supreme Decree (DS) No. 248/2007 (Chile) – abbreviated DS248.
The following section provides a summary of the standards-based design criteria from each of the
international agencies relating to:
 Population at Risk (PAR) and Dam Failure Consequence
 Factor of Safety (FOS)
 Stormwater Management
 Earthquake Loading.

2
POPULATION AT RISK AND DAM FAILURE CONSEQUENCE

Where a standards-based design approach is adopted, prescriptive design loads and contingencies
are based on PAR and impacts/ losses resulting from a potential dam break. Of the guidelines
reviewed, only ANCOLD and CDA present a framework to define PAR and dam failure
consequences.

ANCOLD

Definition of the PAR involves considering a dam break situation under “sunny day” (no flooding)
and extreme flood events (ANCOLD, 2012a). The dam break PAR is defined as the total PAR minus
the PAR affected by a natural flood event immediately prior to the dam break (flood dam break
scenario only). Specific criteria outline scenarios in which people can be discounted from the PAR,
this includes scenarios where it can be proven that people within the breach flow zone can be
evacuated by means of an effective and reliable Emergency Action Plan.
ANCOLD (2012a) suggests that for simple cases the dam break could follow empirical/ qualitative
methods or make a general assumption that the tailings are replaced with water. Where a more
complex dam break assessment is warranted, a 2D mudflow can be completed. However, the
guidelines note that considerable judgement is needed to determine a realistic mudflow scenario.
In some scenarios, an assessment using an incremental Potential Loss of Life (PLL), defined as the
PLL after an event that causes dam failure minus the PLL for the same event in the absence of a
dam failure, is recommended. The dam break PAR/ PLL is assigned an order of magnitude
grouping in a range from <1 (0) to >1000.
The ‘Severity of Damage and Loss’ is evaluated across a range of impact areas (business
importance, public health, social dislocation and environmental consequences) and condensed to an
overall severity level based on the impact area with the highest severity rating (minor, medium,
major or catastrophic). The facility’s dam break PAR/ PLL grouping and overall severity level are
then used to define the consequence category.

3
 Severity of Damage and Loss
Population at Risk
Minor Medium Major Catastrophic

<1 Very Low Low Significant High C

>1 to 10 Significant Significant High C High B

>10 to 100 High C High C High B High A

>100 to 1,000 High B High A Extreme

>1,000 Extreme Extreme

Replicated from ANCOLD 2012a (Table 2: Recommended consequence category); supplementary notes are not shown.

CDA

During the dam classification process, each of four hazard rating components – PAR, loss of life,
environmental and cultural values, and infrastructure and economics – is considered
independently. The overall dam hazard rating is defined by the component which suffers the
highest incremental loss. Similar to ANCOLD (2012a, b), CDA (2007) defines the dam break PAR
(based on the total flood PAR) as None, Temporary or Permanent. In dam break scenarios with a
‘Permanent’ PAR, the facility’s classification is based on the PLL. CDA (2007) stipulates that “the
potential for flow of the impoundment contents beyond the perimeter containment is a determining
factor on whether the containment needs to be considered as a dam”. Such terminology may
provide potential for re-classification of dry stack facilities if it can be proven that flow failure is not
a foreseeable consequence of an eventual slope failure.

4
 Incremental losses
Population
Dam Class Environmental and Cultural
at Risk Loss of Life Infrastructure and economics
values

Low economic losses; area

Minimal short-term loss
Low None 0 contains limited
No long-term loss
infrastructure or services

No significant loss or deterioration

of fish or wildlife habitat Losses to recreational facilities,
Temporary
Significant Unspecified Loss of marginal habitat only seasonal workplaces, and used
only
Restoration or compensation in transportation routes
kind highly possible

Significant loss or deterioration of High economic losses

important fish or wildlife habitat. affecting infrastructure,
High Permanent 10 or fewer
Restoration or compensation in public transportation, and
kind highly possible commercial facilities

Significant loss or deterioration of Very high economic losses affecting

critical fish or wildlife habitat important infrastructure or services
Very high Permanent 100 or fewer
Restoration or in kind possible but (e.g) facility, storage facilities for
impractical dangerous substances)

Major loss of critical fish or wildlife Extreme losses affecting critical

habitat infrastructure or Services (e.g.
More than
Extreme Permanent hospital major industrial complex,
100 Restoration or compensation in major storage facilities for dangerous
kind impossible substances)
Replicated from CDA 2007 (Table 2-1: Dam Classification)

FACTOR OF SAFETY

The slope stability FOS is defined as the ratio of available shear resistance along a potential failure
plane to the activating shear forces along the same plane. The FOS is commonly used to account for
uncertainties in design and construction elements such as material properties, foundation
variability and construction quality.

ANCOLD

ANCOLD notes that there are no rules for acceptable FOS as it is incumbent upon the design
engineer to make considerations and judgements on the selection of the minimum FOS. While the
suggested minimum FOS under static loading is 1.5, a lower value of 1.3 may be adopted for short-
term undrained conditions (i.e. construction) where it is deemed the critical failure surface does not
have potential to cause loss of containment, and the available strength will increase with time.
ANCOLD (2012) no longer recommends the use of pseudo-static analysis as a screening tool for
slope stability under seismic conditions, but instead recommends an assessment of liquefaction
potential followed by deformation estimates using numerical models (Finite Element/ Finite
Difference) or simplified methods. The designer should exercise discretion when defining the

5
minimum FOS for post-seismic stability based on the reliability of residual/ liquefied shear strength
estimates.
Recommended Minimum for Shear strength to be used
Loading Condition
Tailings Dams for evaluation

Long-term drained 1.5 Effective strength

Short-term undrained 1.5 Consolidate undrained

(potential loss of containment) strength

Short-term undrained (no 1.3 Consolidate undrained

potential loss of containment) strength

Post Seismic 1.0 – 1.2 Post seismic shear strength

Replicated from ANCOLD 2012a (Table 8: Recommended factors of safety); supplementary notes are not shown.

CDA
Prescriptive TSF slope stability criteria summarised in CDA (2014) have been derived from CDA
(2007) criteria for water dams. Similar to ANCOLD, an FOS of 1.3 is proposed as a lower limit for
safe slopes during construction/ operation where the failure consequence is minor and mitigation
measures can be readily implemented. A minimum FOS of 1.5 is stipulated for downstream slopes
where long-term, steady-state seepage conditions are expected to prevail. A pseudo-static FOS of >1
is used to indicate that the performance of the dam under earthquake conditions is acceptable; CDA
(2007) notes that the prescriptive FOS criteria are applicable to both limit equilibrium and
numerical finite element shear strength reduction factors (SSRs). Full or partial drawdown
scenarios may not be directly applicable to all TSFs.
Recommended Minimum for Shear strength to be used
Loading Condition
Tailings Dams for evaluation

> 1.3 depending on risk Effective strength

During or at end of
assessment during
construction
construction 1.5

Long term (steady state 1.5 Consolidate undrained

seepage, normal reservoir strength
level)

Full or partial rapid 1.2 to 1.3 Consolidate undrained

drawdown strength
Replicated from CDA 2014 (Table 3-4: Target Factors of Safety for Slope Stability in Construction, Operation, and Transition
Phases - Static Assessment)

Loading Condition Minimum Factor of Safety

Pseudo-static 1.0

Post-earthquake 1.2
Replicated from CDA 2014 (Table 3-5: Target Factors of Safety for Slope Stability in Construction, Operation, and Transition
Phases - Seismic Assessment)

6
MAC

The intent of the MAC guidelines is to address the specific needs of facility owners and tailings
facilities by using critical risk control. The MAC guidelines therefore avoid defining generic FOS
criteria, but make reference to the ‘Detailed technical guidance’ provided in CDA (2007, 2013).
ANCOLD and ICOLD are also referenced as recommended resources.

Chilean Tailings Practice and Regulations

The DS248 guidelines outlines a phased stability assessment, with each phase introducing increased
complexity in the stability analyses, dependent on the TSF’s business importance, risk profile and
the level of the study. No guidance is provided with respect to assignment of categories to each of
these aspects.
Static and pseudo-static methods are recommended for Phase I and Phase II analyses, with Phase I
including pseudo-static analysis and assuming all tailings liquefy following a seismic event. Phase
II incorporates simulation of pore water pressures. For both phases, the recommended minimum
FOS is 1.2; compliance with Phase I and II is sufficient for low embankments (less than 15 m high);
further analyses are required for the remaining phases. Phase III applies to embankments higher
than 15 m and should incorporate stress-strain dynamic analyses. In Phase IV, closure criteria are
applied to the analyses, where loads correspond to maximum credible events. There is no minimum
recommended FOS for static analyses, the reason for this may be related to Chile falling in a zone of
high seismicity, which requires the static FOS to be high enough to ensure that the FOS for pseudo-
static analyses exceeds 1.2. The FOS values for Phases III and IV are not stipulated, and there is no a
definition of allowable deformations.

STORMWATER MANAGEMENT

Water management is a fundamental contributor to the overall success of tailings management and
therefore receives good coverage in design and operations guidelines.

ANCOLD

As outlined in ANCOLD (2012), the definition of stormwater storage criteria is based on the
facility’s spill consequence category, which considers the consequence of the release of stored water
from the dam, as opposed to the consequence category which considers failure of the embankment.
Fall-back methods for ‘non-spill’ facilities where water quality is unsuitable for release define
allowances for extreme storm storage in terms of Annual Exceedance Probability (AEP),
contingency freeboards and in some cases, wet season water storage. Even where a facility is
classified as a ‘non-spill’ facility, implementation of an emergency spillways is recommended. The
magnitude of the design flood adopted for spillway design of ‘spill’ facilities is based on the dam
failure consequence category. Additional spillway freeboard allowances are required to account for

7
wave action and wind set-up conditions. The estimation of closure spillways and freeboard
allowances should always be based on the Probable Maximum Flood (PMF).
Dam Spill Minimum Wet Season Extreme Storm Storage Recommended Contingency
Consequence Water Storage Allowance Allowance Freeboard
Category

1:5 notional АЕР wet season Determine by risk nil nil

Low
runoff assessment

1:10 notional АЕР wet 1:100 AEP, 72 hr flood 1:10 AEP wind 0.3
Significant
season runoff

High C 1:100 notiona1 АЕР wet 1:100 AEP, 72 hr flood 1:10 AEP wind 0.5
season runoff
High B 1:1000 AEP, 72 hr flood 1:50 AEP wind 0.5

1:1000 notional АЕР wet 1:10000 AEP, 72 hr flood 1:50 AEP wind 0.5
High A / Extreme
season runoff
Replicated from ANCOLD 2012a (Table 4: Minimum Extreme Storm Storage – Fall-back method and Table 5: Recommended
Contingency Freeboards)

Dam Failure Design Flood AEP Wave Freeboard Allowance

Consequence Category

Low 1:100 Wave run-up for 1:10 АЕР wind

Significant 1:1000 Wave run-up for 1:10 АЕР wind

High 1;100,000 Wave run-up for 1:10 АЕР wind

or PMF None

Extreme PMF То be determined by risk assessment

Replicated from ANCOLD 2012a (Table 6: Recommended minimum design floods for spillway design and wave-freeboard
allowance during operation phase); supplementary notes are not shown.

CDA
CDA defines three functions of the TSF in relation to water management:

 Temporary storage of water during operations

 Temporary storage of the Environmental Design Flood (EDF)
 Storage and/ or passage of the Inflow Design Flood (IDF) runoff to ensure the integrity of
the facility.
Management of the TSF pond under normal conditions requires definition of a safe range within
which the pond should be maintained. This range is defined by the Low Operating Water Level
(LOWL) and the Normal Operating Water Level (NOWL), which are based on site-specific
constraints, such as beach development, required decant depth and critical phreatic conditions, and
does not include a storm allowance.
The EDF is a water storage allowance above NOWL when a ‘no spilling’ condition is prescribed.
While the CDA recommends estimating the EDF based on a storm event with a return period
between 1:50 years and 1:200 years, it is also suggested that site conditions and the quality of the

8
stored water be considered in consultation with the facility owner and regulators, which may
require a more severe storm to be adopted.
The IDF is the most severe flood condition that a TSF design must be able to accommodate. The
designer should indicate in its water management strategy whether the IDF volumes are to be
safely contained above the EDF volumes, or will be passed through a spillway. It is recommended
by CDA that even when IDF volumes are to be stored within the TSF, the design should always
include an emergency spillway. The CDA recommends IDF target levels be increased in line with
the Dam Classification.
Annual Exceedance Probability – Floods Annual Exceedance Probability –
Dam Classification
Wind setup - Freeboard

Low 1/100 1/100

Significant Between 1/100 and 1/1,000 1/10

High 1/3 between 1/1,000 and PMF 1/2

Very High 2/3 between 1/1,000 and PMF 1/2

Extreme PMF 1/2

Replicated from Technical Bulletin Application of Safety Guidelines to Mining Dams, CDA (Table 3-2: Target Levels for
Flood Hazards, Standards-Based Assessments, for Construction, Operations, and Transition Phases)

MAC

The MAC guidelines consider water management as an inherent component of tailings disposal. It
is implied that a High consequence category is typically assigned to facilities that hold a significant
amount of water, and the category is lowered as the amount of water decreases. A site-specific
water management plan should include common elements such as design floods (assigned
considering requirements of Best Available Technology), dynamic water balance models, life of
mine surface water management plans, and effluent criteria. No specific criteria or exceptions are
indicated for water management in the case of dewatered tailings facilities.

Chilean Tailings Practice and Regulations

The DS248 mainly defines basic documents and design elements that should be presented for
regulatory approval purposes. While the estimated should align with local hydrological conditions,
a minimum freeboard allowance of 1 m is stipulated.
DS50 typically regulates the preparation of water management plans, but explicitly excludes
dewatered tailings facilities. DS50 indicates that in cases where dewatered facilities are placed on
existing basin drainage, diversions should be constructed using a return period of 50 years and the
freeboard must be able to accommodate a 100-year return period.

9
EARTHQUAKE LOADING

Design guidelines and codes provide recommendations for estimating applicable seismic loads;
some are more focused on technical aspects, while others discuss failure mechanisms and
allowances that should be included in the analyses. These are discussed as follows.

ANCOLD

ANCOLD recommends implementing a flowsheet for conducting seismic stability analyses for
tailings dams. The earthquake severity is determined from the dam failure consequence category
that applies. The liquefaction potential is evaluated using semi-empirical methods based on tailings
properties, and depending on the outcome, appropriate post-seismic strengths should be used to
run static limit equilibrium stability analyses. Pseudo-static stability methods to estimate the FOS
due to dynamic loads are no longer recommended. The tailings liquefaction potential is a critical
element when tailings is part of the embankment or where portions of the tailings beach raise above
the embankment’s maximum level.

Dam Failure Consequence Operations phase Post Closure

Category OBE MDE

Low 1:50 1:100 MCE

Significant 1:100 1:1000 MCE

High / Extreme 1:1000 1:10000 MCE

Replicated from ANCOLD 2012a (Table 7: Recommended Design Earthquake Loadings (AEP))

CDA

Definition of seismic criteria should consider the Dam Classification. While CDA (2007) provides
recommendations with respect to target earthquake levels, the facility owner may adopt more
stringent criteria commensurate with its risk tolerance.
The CDA indicates that criteria for post-seismic crest deformation should be established in
accordance with the type of facility. For tailings stacks without a pond (dry stack scenario or facility
with external ponds), CDA (2014) stipulates that the seismic analysis should include a review of the
liquefaction potential of the materials as part of the design of the facility, and that this should be
confirmed through Dam Safety Reviews undertaken during operations.

10
Dam Classification Annual Exceedance Probability – Earthquakes

Low 1/100

Significant Between 1/100 and 1/1,000

High 1/2,475

Very High ½ between 1/2,475 and 1/10,000 or MCE

Extreme 1/10,000 or MCE

Replicated from Technical Bulletin Application of Safety Guidelines to Mining Dams, CDA (Table 3-3: Target Levels for
Earthquake Hazards, Standards-Based Assessments, for Construction, Operations, and Transition Phases)

MAC

There is no specific guide for defining seismic conditions; where applicable, MAC makes reference
to ANCOLD, ICOLD or CDA guidelines.

Chilean Tailings Practice and Regulations

DS248 indicates the use of pseudo-static stability analysis as part of the engineering analyses
required for project approval. The design earthquake corresponds to the maximum credible
earthquake and must be estimated using a region-specific database. The minimum allowable FOS is
1.2. Although there is no specific TSF failure consequence categorisation, it is indicated that
depending on the severity or consequence of TSF failure, additional analyses are required –
including dynamic analyses based on stress-strain techniques using dynamic properties of the
tailings in which case potential deformations should be estimated.

RESULTS

PAR and Dam Failure Consequence

When considering dewatered tailings facilities, modelling of dam breaks using water flows may
produce unrealistically large breaches which do provide an adequate evaluation of actual risk and
potential impacts in the event of a dam break. The use of strength-based modelling approaches,
such as those proposed by Seddon (2010) and Martin and Fontaine (2015), is considered a more
appropriate dam break modelling method for thickened tailings facilities where appropriate
rheological and/ or strength parameters are adopted.
Where there is no potential for liquefaction and no pond, tailings are theoretically non-flowable and
the dam break runout can therefore be modelled based on a slope failure (Small et al., 2017 as
referenced by Orman, M, 2017). This scenario is considered most relevant to filtered tailings dry
stacks where unsaturated, dilative material conditions are often achieved in peripheral structural
zones through mechanical compaction effort. In circumstances where forecast of the failure mode of
a TSF resembles a ‘non-dam’ facility, such as in the case of slumping on an unsaturated facility with
no pond, there may be scope to consider the facility a ‘non-dam’ landform. Potential for dam break

11
outflow increases where filter under-performance causes an unexpected build-up of pore pressures
in contractive, saturated areas at the base of the landform. For this reason, it is imperative that the
variability of the tailings mineralogy and gradation is considered at the design stage so that
appropriate engineering and operational contingencies can be implemented and monitoring of the
facility focuses on meeting key performance indicators related to those parameters.
The tailings flow behaviour resulting from a basal liquefaction failure is more difficult to predict
due to the complex relationship between liquified strength and kinetics; at the onset of liquefaction,
the driving shear stress exceeds the tailings peak shear strength and is considerably larger than the
liquefied shear strength. Acceleration of the tailings mass starts mobilising the tailings down the
slope and as the driving shear stress reduces, reaching a point lower than the liquefied tailings
shear strength, the flow velocity would approach zero (Olson and Stark, 2002).
Regardless of the dam break impact, given the importance of tailings facilities in mining operations,
dewatered tailings dams are still likely to be assigned a High Consequence Category, driven by the
suspension of operations and financial/ reputational impact on the owner’s business.

Factor of Safety

Unlike earthquake loading and hydrological criteria, fall-back FOS criteria are based on a set of
universal loading conditions, independent of the dam failure consequence and the reliability of
stability parameters. By comparison, stability criteria adopted for unsaturated waste dumps and
stockpiles (CSIRO, 2017) are based on the failure consequence as well as level of confidence in
stability parameters. As outlined by Herza et al. (2017), the use of prescriptive FOS criteria does not
necessarily ensure a lower probability of failure, as the stability analysis depends on the reliability
of the input parameters.
ANCOLD (2012) states that increasing the shear strength of unsaturated materials to account for
suction effects should generally be avoided. In semi-arid climates where dewatered tailings
technologies are implemented for water conservation, the influence of suction can be considerable
and can result in significant increases in shear strength (Herramen, 2016). The practical benefits of
steepening the landform slopes by accounting for unsaturated strengths need to be considered
against increased regulator scrutiny and perception as well as the facility’s closure requirements.
Crystal et al. (2018) have shown that where compacted tailings are placed in a peripheral ‘structural
zone’ of the dry stack, it is possible for the tailings to behave as a saturated material, especially in
areas of high rainfall. It is therefore imperative that the design includes appropriate contingencies
to mitigate the development of a phreatic surface.

Stormwater Management

References to specific water management requirements for dewatered tailings are generally limited
in the guidelines reviewed for this paper. Given that conventional tailings facilities carry high risks
due to maintaining a pond on the facility, the stringency applicable to the design criteria for water
management is related to the dam failure consequence category. When designing dewatered

12
tailings facilities, there are essentially two ways to manage runoff – store runoff in the facility or
divert it to an external pond.
Dry stacks are typically built up in levels, where the lowest point would vary according to the
development of the stack. For facilities developed in arid regions, runoff may be accumulated in
low spots, preferably away from the TSF perimeter. In this case, design criteria would be based on
the dam failure consequence category of the TSF similar to a conventional facility; the storm storage
capacity and the freeboard allowances would be used to define the gradient towards the centre of
the facility that should be considered in the stacking operation. A water balance would be required
to estimate the capacity of the stack to absorb and evaporate the transient pond formed after
seasonal and extreme storm events.
For some dewatered facilities, it may be impractical to maintain a reclaim pond on the TSF and
runoff would often be diverted to an external pond. In these cases, criteria for the design of the
external pond may be dictated by the dam failure consequence category of the stormwater pond;
the effect of this is that design criteria for water management would be less onerous.
While the authors recognise that water management is a site-specific design element, given the
reduced amounts of water available in dewatered facilities, water management strategies to achieve
a cost-effective solution should be explored rather than using prescribed solutions that
predominantly apply to conventional tailings facilities. MAC (2019) provides a comprehensive
framework to assist in defining strategies using a risk-based approach, which has the additional
benefits of minimising costs while maintaining acceptable risk levels.

Earthquake Loading

Typically for the design of conventional tailings facilities, it is necessary to demonstrate the stability
of the embankment and peripheral materials only. In scenarios where tailings are stacked without a
retention structure (dry stack) or if only a small starter dam is placed at the toe of a slope as in a
thickened central discharge facility, the stability of the tailings mass itself requires to be evaluated.
In particular, this assessment should carefully consider liquefaction, be it static or dynamic.
ANCOLD and CDA guidelines provide a robust framework to evaluate seismic impacts on a TSF
that is mainly focused on the stability of the retaining structure. ANCOLD recommends the
assessment of the tailings mass when it forms part of the embankment (upstream raises). The
recommended approaches are the comparison of cyclic resistance ratio (CRR) to the seismically
induced cyclic stress ratio (CSR), and the use of critical state-based liquefaction assessment (Jeffries
and Been, 2006).
While these approaches may be applied with some adaptation to estimate the stability of dry stacks,
they are reliant on a significant assumption regarding the forecast statistics of the in situ material.
In the case of the critical state approach, given that dewatered tailings are delivered with reduced
amounts of water, it can be assumed that deposited tailings are at or slightly above the critical line
(Been and Li, 2009). However, the stack does not always evolve through the life cycle estimated in
the design, because overloading of the tailings without reaching appropriate desiccation or self-

13
weight consolidation may drive the designed state path below or at the critical state line. It is
therefore important to define operating conditions of the stack and introduce critical controls to be
monitored throughout the life cycle of the facility.

CONCLUSION

The pressure to deliver projects quickly and often within limited budgets creates a tendency to rely
on standards-based design criteria which may not always be directly applicable to tailings dams.
The focus of this review was on the applicability of the standards-based design approach and how
this relates to the design and operation of dewatered tailings facilities.
The extent of physical impact associated with a potential dam break must be fully understood at the
outset of the facility design, and the consequence classification of the facility must follow on from
this understanding. For this reason, it is essential that plausible failure modes and runout scenarios
are considered when conducting dewatered tailings dam break assessments – doing so will avoid
misalignments in perceptions of acceptable risks.
There are two ways of managing stormwater on stacked tailings; if the facility is designed to store
internal runoff onto the facility, flood criteria are based on the consequence category of the tailings
facility. Alternatively, runoff can be managed by designing a water-shedding landform which
would require construction of an external pond; flood criteria in this case would be based on the
consequence category of the stormwater pond.
Although pseudo-static analyses are still accepted as a valid assessment of the stability for seismic
conditions under CDA and DS248 guidelines, it can be concluded that the stability of a beach slope
or a stacked slope may not be optimally assessed using this approach; a focus on liquefaction
potential and assessment of the likely loss of shear strength should be given preference.
While a number of gaps and opportunities to streamline and guide a more efficient design of
dewatered facilities have been identified, the practical benefits of seeking approval of lowered
criteria would need to be balanced against increased regulatory scrutiny and design effort to
supersede fall-back design criteria. It is considered that the fall-back methods may lead to increased
dam construction and operating costs, while the design of mining dams should be based on a risk-
based approach.
The MAC, ANCOLD and CDA all provide excellent good practice guidance (ICMM, 2019) and are
in general agreement that the design of a tailings facility should follow a risk-based approach. The
Chilean case is different, as there is no specific framework to assess risk to define design criteria in
agreement with the risk tolerance of the facility owners and other stakeholders.

14
REFERENCES

ANCOLD (2003) Guidelines on Risk Assessment, Australian National Committee on Large Dams
Incorporated, Australia.

ANCOLD (2012a) Guidelines on Tailings Dams: Planning, Design, Construction, Operation and Closure,
Australian National Committee on Large Dams Incorporated, Australia.

ANCOLD (2012b) Guidelines on the Consequence Categories for Dams, Australian National Committee
on Large Dams Incorporated, Australia.

Been, K, Li, A. (2009) ‘Liquefaction and Paste Tailings’, in R Jewell, Ab Fourie, S Barrera and J
Wiertz (eds), Proceedings of the 12th International Seminar on Paste and Thickened Tailings, Australian
Centre for Geomechanics, Perth, pp. 281-289.

CDA (2007, Revised 2013) Dam Safety Guidelines, Canadian Dam Association, Canada.

CDA (2014) Technical Bulletin: Application of Dam Safety Guidelines to Mining Dams, Canadian Dam
Association, Canada.

Crystal, C, Hore, C, Ezama, I (2018) ‘Filter-Pressed Dry Stacking: Design Considerations Based on
Practical Experience, in Proceedings Tailings and Mine Waste 2018 Conference, pp. 209-2019.

Fundación Chile (2018) ‘Avances y retos para la gestión de los depósitos de relaves en Chile’,
Programa Tranque, Santiago de Chile.

Gobierno de Chile (2007) Reglamento para la aprobación de proyectos de diseño, construcción, operación y
cierre de los depósitos de relaves, DS248, Ministerio de Minería, Santiago de Chile.

Golder Associates (2016) Review of Tailings Management Guidelines and Recommendations for
Improvement, report prepared for International Council on Mining and Metals (ICMM) (12/03/2019,
https://www.icmm.com/website/publications/pdfs/tailings/161205_review-of-tailings-management-
guidelines.pdf).

Herramen, R (2016) ‘Getting soil strength parameters for stiff unsaturated clay’, paper presented at
AGS-SA 2016 Seminar – Design considerations for cuttings, revetments and retaining structures for
unsaturated clay, Adelaide, 19 September.

Herza, J, Ashley, M and Thorp, J (2017) ‘Factor of Safety?-Do we use it correctly?’, in Proceedings
ANCOLD 2017, viewed 10/03/2019,
https://www.ghd.com/en/services/resources/PDF/ANCOLD_2017_Ashley_Factor-of-Safety-Do-we-
use-it-correctly.pdf.

Herza, J, Phillips, J (2015) ‘Anomalies in the design of mining dams’, in Proceedings ANCOLD 2015,
viewed 10/03/2019, http://www.ghalivigunn.com/en/services/resources/PDF/ANCOLD_2015-
Anomalies_in_design_for_mining_dams-HERZA_PHILLIPS.pdf.

15
MAC (2017) A guide to the Management of Tailings Facilities – Third Edition, Mining Association of
Canada, Canada.

MAC (2019) A guide to the management of tailings facilities, version 2.1, Mining Association of Canada,
Canada.

Martin, V, Fontaine, DD and Cathcart, JG (2015) ‘State of Practice for Conducting Tailings Dam
Breach Studies’, in Proceedings of the Tailings and Mine Waste 2015 Conference, Vancouver, BC,
Canada.

Olsen, SM, Stark, TD (2002) ‘Liquefied strength ratio from liquefaction flow failure case histories’,
Canadian Geotechnical Journal. Vol. 39: 629–247.

Orman, M, Small, C, Al-Mamun, M (2017) Tailings Dam Classification and Breach Analyses, Perspective
from the Canadian Dam Association, presentation outlining the current state of practice for dam
consequence classification and dam-break assessment outlined by the Canadian Dam Association,
Amec Foster Wheeler, viewed 13/03/2019,
https://www.mtech.edu/mwtp/2017_presentations/wednesday/marc-orman.pdf

16
An Introduction to Oil Sands Tailings: Innovations
and Challenges
Ward Wilson, Nicholas Beier and Vivian Giang
University of Alberta, Alberta, Canada

ABSTRACT
The oil sands industry is one of Canada’s key industries, providing a stable energy supply across
North America. However, the environmental footprint of Alberta’s oil sands operations continues to
be a challenge to the industry and the energy regulator. The University of Alberta, Canadian Oil
Sands Innovation Alliance Tailings Environmental Priority Area (COSIA Tailings EPA) and Alberta
Innovates – Clean Energy have joined forces to establish the NSERC/COSIA Industrial Research
Chair in Oil Sands Tailings Geotechnique to address the challenges of managing oil sands tailings
and remediating tailings pond sites.
The topics presented will describe the key challenges currently faced by the oil sands industry.
Several new processes and techniques that offer promise for application in the oil sands industry, as
well as in other mine waste operations, will be discussed. Future directions and alignment of the
research programs to meet the needs of the Tailings Management Framework for the Mineable
Athabasca Oil Sands and the Alberta Energy Regulator’s Directive 085 (AER, 2017a) for Fluid Tailings
Management for Oil Sands Mining Projects are also examined.
Through this research partnership, the collaborators aim to produce greater resource efficiency,
productivity and sustainability in the oil sands industry and provide fundamental contributions to
the stabilizing of tailings in a timely manner for reclamation. The ultimate goal of the research is to
provide new technologies that can be readily implemented with the optimum cost/benefit success in
terms of sustainable, restored landscapes in Canada.

1
INTRODUCTION

The development of the Canadian oil sands in Alberta may be considered the largest energy project
worldwide that is currently in progress, helping Canada’s economy stay strong while providing a
stable energy supply across the nation and the world. However, the environmental footprint of
Alberta’s oil sands operations continues to be a challenge to the industry and its regulator. With an
estimated 1.2 billion cubic meters (m³) of oil sands mature fine tailings within a greater footprint of
220 km2 of land in northern Alberta (Kent, 2017), the oil sands industry poses $21 billion in
reclamation liabilities (Alberta Auditor General, 2015). Enhancing tailings management and the
ultimate reclamation of these deposits is more crucial than ever.
In 2017, the Canadian oil sands industry produced approximately 2.7 million barrels per day out of
the total 4.2 million barrels of oil per day that Alberta produces. The Alberta Energy Regulator (AER,
formerly the Energy Resources Conservation Board, ERCB) predicts that oil production will increase
to 3.7 million barrels per day within the next decade, with gross revenues expected to grow to $100
billion per year. Half of this output is generated by surface mining operations (in situ methods
produce the remainder). The oil sands are viewed as a vital component to the Canadian economy. At
the same time, the environmental impacts of oil sands operations are also growing.
Water-based processes are used to extract bitumen from mined oil sands ore. Tailings impoundments
are used almost exclusively for waste management. The combined footprint for these impoundments
is currently 220 km2 (Kent, 2017). Syncrude’s Mildred Lake Settling Pond has an ~18 km perimeter
tailings dam and may be considered one of the largest engineered fill earth structures in the world
(Morgenstern, 2010). Processed oil sands ore yields a tailings slurry composed of water, sand, silt and
clay (or “fines”), and a small amount of residual bitumen. The tailings are discharged into a settling
basin, allowing the sand fraction to settle quickly on the beach with the balance of fine tailings settling
in the pond.

2
The oil sands industry has always operated under conditions of zero discharge for process affected
water, and the clarified water produced near the surface of the pond is recycled for the extraction
process. Fluid fine tailings (FFT) with a low solids content varying between 15 and 30% forms below
the clarified surface water. The clay fraction within these tailings is dispersed due to a high sodium
content that resists settling and densification. Prolonged settling eventually produces tailings with
increased solids content (greater than 30%) that are frequently termed Mature Fine Tailings (MFT).
MFT is essentially in the liquid state with practically zero shear strength, thus rendering it non-
trafficable. Experience has shown that further consolidation to achieve soil-like consistency in the
plastic state will take more than a century.

Figure 1 Expected fluid fine tailings (FFT)

inventory as per regulatory approvals (AER,
2017b, c, 2018a-e)

The current combined inventory of MFT for the oil sands industry is estimated to be more than 1
billion m³ and could grow to approximately 1.7 billion m³ within the next decade (Figure 1). Tailings
operations and tailings management for the oil sands industry have attracted global attention. In
February 2009, the ERCB enacted its first Directive, Directive 074, that stipulated tailings performance
criteria and requirements for oil sands mining schemes and identified several long-term objectives
for oil sands tailings management. Three of these key objectives are: 1) to minimize and eventually
eliminate long-term storage of fluid tailings in the reclamation landscape; 2) to create a trafficable
land. At the earliest opportunity to facilitate progressive reclamation; and 3) to eliminate or reduce
containment of fluid tailings and external tailings disposal area during operations. In 2016, Directive
074 was replaced by Directive 085 (AER, 2017a), which will be further discussed in the next section.
These objectives motivated the oil sands industry to form the Oil Sands Tailings Consortium (OSTC,
now officially known as the Tailings Environmental Priority Area of the Canadian Oil Sands
Innovation Alliance, COSIA Tailings EPA) with the specific goal to significantly advance the rate at
which tailings ponds can be reclaimed. COSIA Tailings EPA is comprised of member companies from
five of Canada’s largest oil sands operators (including CNRL, Imperial Oil, Suncor, Syncrude and
Teck).

3
THE NSERC/COSIA INDUSTRIAL RESEARCH CHAIR

Energy is a key strategic research area at the University of Alberta: over 200 faculty members
specialize in energy areas across 23 departments and 10 faculties. Through the award of a $75 million
Canada First Research Excellence Fund to establish the Future Energy Systems research initiative, the
University of Alberta has been recognized as a leading energy research institution in Canada.
Moreover, the University of Alberta is an international leader in natural resource engineering
research with an established track record of engaging in collaborative research initiatives with the oil
sands industry. For example, in 2003, stakeholders in the oil sands industry recognized the need to
mitigate the vast amount of disturbed land and quantity of oil sands tailings being stored and
supported the establishment of the $2.2 million Oil Sands Tailing Research Facility (OSTRF) with the
University of Alberta.
In 2013, COSIA identified an opportunity to combine all of its industrial research efforts to reduce
the environmental footprint of the oil sands tailings and to collaborate with academic researchers at
the University of Alberta to find a solution for this problem. COSIA Tailings EPA and University of
Alberta joined forces to establish the NSERC/COSIA Industrial Research Chair in Oil Sands Tailings
Geotechnique to address the challenges of managing oil sands tailings and remediating tailings pond
sites. The establishment of the Chair strengthened the industrial–academic collaborative research
efforts between the University of Alberta and oil sands industry that began as early as 1980 and
intensified the core research in oil sands tailings remediation conducted at the University of Alberta.
Upon consultation with industry, four major areas of innovative research were determined for
Phase I of the IRC program (2013-2019) as follows: 1) investigating the unsaturated soil mechanical
properties of oil sands tailings; 2) investigating consolidation processes for various forms of fluid fine
tailings, MFT and amended MFT; 3) assessing and improving tailings deposition; and 4)
understanding the long-term geotechnical behaviour of fine tailings through laboratory testing and
simulations. Throughout the course of the current IRC program, 18 highly qualified personnel have
been trained to conduct research under these research areas, with objectives primarily aimed at
assisting the industry achieve the mandate of Directive 074. The collaborative research resulted in the
development of new field investigation techniques, tailings simulation models and unsaturated soil
mechanical models, which have been transferred to Canada’s oil sands tailings industry.
While the industry made progress under Directive 074, experience demonstrated that further
advances were needed. A new Directive issued by the AER in 2016, Directive 085, specifies that new
FFT deposits must be ready to reclaim ten years after the end of mine life, and that all legacy tailings
must be ready to reclaim by the end of mine life (AER, 2017a). Traditionally, tailing ponds are used
to settle FFT, and recover water for recycling in oil sands operations (Chen et al., 2013). However, it
has been estimated that FFT will take 125-150 years to completely consolidate though gravity settling
(Eckert et al., 1996), rendering dryland reclamation extremely difficult under Directive 085.
Alternatively, tailings with aqueous covers, also known as end pit lakes, offer long-term FFT
containment and closure with low initial costs (Charette et al., 2010). These mine closure landscapes

4
contain FFT deposits under a water cover within decommissioned mine pits. The water cap provides
the habitat for an aquatic ecosystem while the FFT is allowed to naturally dewater over time (Charette
et al., 2010). The expectation is that tailings with aqueous covers will develop into self-sustaining
aquatic ecosystems through fresh water inputs, in situ biogeochemical processes and engineering
interventions. Currently, over 1 billion m3 of FFT are expected to be reclaimed in this manner (AER,
2017b,c, 2018a-d).
The strong research relationship between the University of Alberta and COSIA during Phase I of the
IRC program resulted in a commitment to renewing the program for another five years (Phase II,
2019-2024) to address these challenges. The research to be undertaken during Phase II takes a holistic
view of the oil sands tailings as part of Canada’s energy system. A major aspect of the research
includes the design of future tailings stacks that are well suited to progressive and rapid require
reclamation, while providing the critical research needed to reduce the environmental impact and
quicken the remediation of current tailings deposits.
The next phase of the research proposes the development and application of new and innovative
geotechniques to design dry stacked deposits for the oil sands industry while simultaneously
investigating geotechnical and bio-geochemical aspects of current oil sands tailings management and
reclamation strategies. The proposed research program is focused on: 1) engineered mixtures of
tailings and overburden for dry stacked deposits; 2) the influence of bio-geo-chemical interactions on
the geotechnical behaviour of soft tailings; and 3) capping soft tailings deposits. These
complementary themes aim to conduct research that looks to the future of tailings deposits, while
providing research for immediate/interim challenges. All themes include some research focused on
the impact of clays in terms of geotechnique.
Phase II of the IRC program aims to understand whether tailings with aqueous covers an
environmentally responsible reclamation approach are by investigating the geotechnical and
biogeochemical evolutions within tailings with aqueous covers. The IRC program will also
investigate new methods for surface disposal of FFT, including dry stack deposits and capping of
soft tailings. The development of novel technologies for the capping of soft tailings is a critical
component of the research, as all legacy tailings must be ready to reclaim by the end of mine life
(AER, 2017a). The development of methods for the deposition of dry stack tailings is growing rapidly
in the global mining industry (Wickland and Longo, 2017), and the opportunity to develop and apply
this technology to the oil sands industry will be capitalized on within the IRC program. The
anticipated outcomes of the research will contribute to the more effective utilization of our natural
resources and for provision of research leadership within the engineering, natural resources and
environmental areas. This innovative partnership between academia and industry maximizes
technology transfer, bolsters novel discoveries in natural resource engineering and management, and
trains top quality students, researchers and collaborators.

5
CANADA’S OIL SANDS TAILINGS

The oil sands found in the shallow McMurray formation beneath northern Alberta are at low
temperature (i.e. in the range of 5°C) and thus, unlike conventional oil reservoirs that are at much
higher temperatures, the high viscosity of the cold bitumen prevents extraction with conventional
pumping technologies. However, oil sands operators have turned to steam assisted gravity drainage
(SAGD) technologies to extract the bitumen found at greater depths. This in situ production method
is increasing and now accounts for almost 50% of output. Nevertheless, conventional mine methods
have traditionally been and continue to be widely used to mine the oil sands, followed by water-
based processes with heat to extract the bitumen from the ore. Above grade tailings impoundments
provide storage for all of the process water and tailings.
Tailings are typically discharged to a settling basin as shown in Figure 2 below. The sand fraction
settles rapidly on the beach with the remaining fine tailings settling in the pond. Low solids content
clay tailings called thin fine tailings (TFT) form below the surface. Since the clay fraction in the tailings
is dispersed by the high sodium content in the process water, settling, thickening and consolidation
is exceptionally slow.
Oil Sands from Mine Bitumen to Upgrading
Extraction
Water + Process aids

Extraction Tailings

Reclaim Water
Sand Sand
Dykes Dykes
Mature Fine Tailings
Sand Beach Sand Beach

Figure 2 Typical oil sands tailings settling basin (Beier and Sego, 2007)

In 2012, the COSIA Tailings EPA published a Technical Guide for Fluid Fine Tailings Management.
The Guide was prepared by a working group of COSIA member companies along with two expert
panels comprised of some of the most experienced and distinguished geotechnical engineers,
academics, scientists and consultants in the field of oil sands tailings. The key issues targeted for
managing FFT during operation of oil sands mines pivot on the creation of sustainable terrains for
mine reclamation. The Technical Guide for Fluid Fine Tailings Management specifically states, “the
methods for transformation of fluid find tailings into stable, sustainable elements of a closure
landscape are all in various states of development, from preliminary research to commercial practice.
None can be considered as mature, i.e. proven practice, with performance fully demonstrated for
operation and closure”.
The COSIA Tailings EPA (2012) Technical Guide defined four deposit types that will be created using
the FFT management practices currently under active development for commercial use. These are: 1)

6
thin layered - fines dominated deposits; 2) deep - fines dominated deposits; 3) fines enriched sand
deposits; and 4) water capped fines deposits.
The major performance criteria for these deposit types can be categorized in terms of cost, area
footprint, containment requirement, energy and time to capping and reclamation. Novel and
innovative techniques to implement, monitor, evaluate and advance the functional performance of
these disposal schemes are needed so that oil sands tailings repositories can be constructed to form
geotechnically secure landforms that progress toward mine closure. A general description for these
deposit types is provided in the following section.

DEPOSIT DESCRIPTIONS
Thin layered - fines dominated deposits. Thin layered - fines dominated deposition was trialled
and/or implemented by CNRL, Suncor and Syncrude, and has attracted considerable attention. This
deposit type is constructed by discharging fine tailings sub-aerially in lifts that vary between 100 and
500 mm in thickness and can be carried out with MFT or thickened tailings (TT). The method involves
the injection of flocculant into the pipeline upstream from the discharge point, which allows the
required contact time, mixing and the formation of clay flocs to take place within the pipe.
Dewatering of the thin layered MFT deposits occurs through initial decant and runoff, drainage to
the foundation layer and by atmospheric evaporation. Freeze thaw cycles can further dramatically
accelerate dewatering. Wilson et al. (2011) provides a detailed description of a field trial completed
by Syncrude at its Mildred Lake Settling Basin. Once sufficient dewatering has occurred, the tailings
lift can be transported to an overburden disposal location. Alternatively, each dewatered lift can be
left in place followed by further sequential deposition of multiple lifts.
Suncor is currently operating a commercial scale in-line flocculated MFT thin lift disposal operation
known as TRO (tailings reduction operation). Shell has also established a large-scale prototype at its
Muskeg River mine. In addition, Syncrude is currently testing a large-scale prototype for MFT
centrifugation (known as MFT-C or Centr-Cake). The primary performance indicator for thin layered
- fines dominated deposits is the development of acceptable undrained shear strength. In general,
this method of deposition is considered to be viable at commercial scales provided sufficiently large
land areas are available that allow adequate surface drying and the development of target shear
strength values.
Deep - fines dominated deposits. Deep - fines dominated deposits are created using an FFT discharge,
similar to those for thin layered deposits, which is placed into a deep disposal site. The thickness of
the deposit is allowed to accumulate to a significant depth over time. Initial water decant occurs due
to a polymer flocculant process followed by continuous water release through self-weight
consolidation and environmental dewatering associated with evaporation and freeze thaw. In
general, three processes can be used to create deep fines dominated deposits. These include: 1) the
use of in-line flocculated MFT followed by deposition in a deep containment cell (accelerated

7
dewatering), 2) the deposition of TT from sand depleted tailings streams that have been processed
with a thickener, and finally 3) MFT-C, which is a process to dewater FFT that uses solid-bowl scroll
centrifuges following the addition of flocculants.
The primary performance measure for deep - fines dominated deposits is the increase in solids
content over time. Once sufficient volume reduction and surface crust development has occurred, the
deposit can be capped with sand to provide additional loading and further consolidation, followed
by the placement of topsoil. Deposits are best suited where in pit area and storage volume is available.
Fines enriched sand deposits. Fines enriched sand deposits are formed using composite or
consolidated tailings (CT) or non-segregating tailings (NST). CT is produced by blending MFT with
high solids content sand slurry and coagulants. Alternatively, TT can be used as the source of FFT to
produce NST. Fines enriched sand deposits typically have sands to fines ratios (SFR) between 3 and
5 with solids content well above 70% after deposition. CT is the process currently in use for this type
of deposit and has been implemented at commercial scale by both Syncrude and Suncor. NST is
currently implemented at commercial scale at CNRL. The primary performance measures for the
successful implementation of fines enriched sand deposits is a satisfactory SFR distribution upon
deposits that provides an increasing solids content with time under increasing surcharge.
Water capped fines deposits. Water capped deposits are envisioned for large volumes of fluid fine
tailings. The method involves placing naturally densified MFT or “pre-treated” MFT (i.e. using in-
line flocculation) with a solids content greater than 30% into an engineered mine pit. Densification
and full consolidation of the MFT is currently expected to take centuries. The amount and rate of
consolidation is a key issue of concern. The primary performance factor for this approach is to obtain
acceptable water quality for the inflows and outflows from the surrounding terrain coupled with the
capacity to emulate natural lakes. While subaqueous disposal with a water cover is considered a
leading practice for acid generating tailings both for Canadian and international jurisdictions, the use
of a water caps to establish pit lakes has not yet been accepted for closure of oil sands fluid find
tailings. However, extensive biogeochemical characterization has been performed in Syncrude’s Base
Mine Lake, the first field-scale test of water capped deposits in the Athabasca oil sands region and
will continue to be investigated through the IRC program.
Comparison of the deposit types. Table 1 provides a summary of the characteristics and key attributes
for the four methods described above. Clearly, individual operators will need to consider a wide
range of factors when selecting methods including land availability, site geology, geotechnical
conditions and mine development. The Technical Guide for Fluid Fine Tailings Management
recommends an adaptive management approach to implement and balance the use of the methods
outlined here. Considerable technology development, research, monitoring, observation, analysis
and experience will be required for stakeholders and decision-makers to implement these new
technologies for tailings management, reclamation and closure.

8
Table 1 Summary of deposits (after COSIA Tailings EPA, 2012)

Method Cost Area Containment Energy Time to

($) Footprint Requirement Intensity/GHG Capping/Reclamation

Thin-Lift Drying High Very Compacted Medium Extended placement

Large shell time; rapid when
complete

Thin-Lift Drying Very Large None when High Rapid

with Rehandling High placed

Centr. High Small Low High Slow

Deep
AD Medium Moderate Low Low Slow
Cohesive
ATT Moderate Low Very Low Slow Moderate
(Energy
recovery)

Fines-Enriched Medium Small High Medium Fast

Sand Deposits

Water-Capped Low Moderate Yes Low Water-capping time

In-pit Deposits Densified – depends on source rate
Low of water fill. Time for
MFT – water quality attainment
Medium likely 10 years.

CONCLUSIONS
Management of fluid fine tailings has become a paramount issue for the oil sands industry. The oil
sands industry has an environmental footprint of 220 km 2 in northern Alberta (Kent, 2017),
representing $21 billion in reclamation liabilities (Alberta Auditor General, 2015). These imminent
and important industry issues guide the research collaboration and focus of the proposed IRC
program. The research themes outlined in this paper describe novel processes and techniques that
may hold promise for immediate application in the oil sands industry and, more importantly, address
long-term reclamation issues for the industry. These research challenges also illustrate the need to
develop multidisciplinary research teams to address the significant issues facing the oil sands
industry. The teams must collaborate with knowledgeable industrial representatives to ensure the
research programs remain focused on the actual industrial issue, while the important fundamental
understanding is compiled or developed to properly solve the issue. The research undertaken via the
IRC program will produce greater resource efficiency, productivity and sustainability in the oil sands
industry and provide fundamental contributions to the stabilizing of tailings in a timely manner for
reclamation.

9
The IRC program enables the oil sands industry to combine its efforts with those of leading
researchers at the University of Alberta to develop novel technologies and methods to manage oil
sands tailings in northern Alberta. The IRC program will advance the scientific community’s
fundamental understanding of the behaviour of fluid fine tailings and create new designs for dry
stackable tailings to increase the sustainability of the oil sands operations. These advances are aligned
with the industry’s regulatory requirements and commitment to improve tailings management by
reducing land disturbance through tailings management facilities and earlier restoration of tailings
deposits. The IRC program will provide industry with the ability to responsibly develop the oil sands
and reduce the impact of its tailings. Once the fundamentals of each issue outlined in the proposal
are well understood, demonstration field tests and verification will be used to guide industry for
future development of these processes.
The development of the Canadian oil sands has attracted intensive global attention. Canadians and
the global community are concerned with the scale and impact of developing this vital resource. They
demand that proven sustainable technologies be developed and implemented to reclaim and
replenish the land that has been disturbed by mining and the associated waste repositories. Both the
environmental and financial liability of the Canadian oil sands industry are enormous. In addition to
optimizing the operation of the oil sands tailings industry, the IRC program will enable the industry
to responsibly manage their post-production tailings and create to more environmentally-conscious
solutions to reduce the industry’s footprint. The ultimate goal of the NSERC/COSIA IRC in Oil Sands
Tailings Geotechnique is to provide new technologies that can be readily implemented with the
optimum cost/benefit success in terms of sustainable oil sands operations and restored landscapes in
Canada. Further, the collaboration between industry and academia will enable fundamental
discoveries for the behaviour and improved management of fluid fine tailings, benefitting the
engineering discipline, and bolster the University of Alberta and Canada’s position as an
international leader in mining reclamation and oil sands tailings geotechnique.

ACKNOWLEDGEMENTS

The authors would like to thank Alan Fair, whose expertise and experience in the oil sands industry
greatly contributed to this paper. The authors also thank the Natural Sciences and Engineering
Research Council of Canada (NSERC), Canada’s Oil Sands Innovation Alliance (COSIA) and Alberta
Innovates – Clean Energy for their financial support.

REFERENCES

Alberta Auditor General. (2015, July) Systems to Ensure Sufficient Financial Security for Land
Disturbances from Mining. Available at https://www.oag.ab.ca/reports/environment-and-parks-and-
alberta-energy-regulatior-systems-ensure-sufficient-financial-security-land-disturbances-mining-
july-2015.

10
Alberta Energy Regulator. (2017a) Directive 085: Fluid Tailings Management for Oil Sands Mining
Projects.

Alberta Energy Regulator. (2017b) Decision 20171025A: Suncor Energy Inc; Application for
Operational Amendment and Base Plant Tailings Management Plan.

Alberta Energy Regulator. (2017c) Decision 20171218A: Canadian Natural Resources Limited;
Application for Operational Amendment and Base Plant Tailings Management Plan.

Alberta Energy Regulator. (2018a) Mineable Oil Sands Fluid Tailings Status Report 2016.

Alberta Energy Regulator. (2018b) Decision 20180523A: Canadian Natural Upgrading Limited;
Application for Muskeg River Mine Tailings Management Plan.

Alberta Energy Regulator. (2018c) Decision 20180523B: Canadian Natural Upgrading Limited;
Application for Jackpine Mine Tailings Management Plan.

Alberta Energy Regulator. (2018d) Decision 20180613A: Syncrude Canada Ltd; Application for
Aurora North Tailings Management Plan.

Alberta Energy Regulator. (2018e) Decision 20180716A: Imperial Oil Resources Limited; Application
for Kearl Management Plan.

Beier, N. and Sego, D. (2007) The Oil Sands Tailings Research Facility. 60th Canadian Geotechnical
Conference and 8th joint IAH-CNC Groundwater Specialty Conference, October 22-24. Ottawa, ON.

COSIA Tailings EPA (2012). Technical Guide for Fluid Fine Tailings Management. Calgary, AB: Oil
Sands Tailings Consortium and Canadian Oil Sands Innovation Alliance.

Charette, T., Castendyk, D., Hrynshyn, J., Kupper, A., Mckenna, G. and Mooder, B. (2010) End Pit
Lakes Guidance Document 2012. Cumulative Environmental Management Association. Available at
http://www.cemaonline.ca

Chen, C.S., Lau, Y.Y., Mercer, S.M., Robert, T., Horton, J.H. and Jessop, P.G. (2013) The effect of
switchable water additives on clay settling. Chemsuschem 6: 132-140.

Eckert, W.F., Masliyah, J.H., Gray, M.R., and Fedorak, P.M. (1996) Prediction of sedimentation and
consolidation of fine tails. American Institute of Chemical Engineers, 42(4): 960-972.

Kent, G. (2017, September 28) Tailings ponds a critical part of Alberta’s oilsands legacy. Calgary
Herald. Available at https://calgaryherald.com/business/energy/tailings-ponds-a-critical-part-of-
albertas-oilsands-legacy

Morgenstern, N. R. (2010) Keynote Address: Improving the safety of mine waste impoundments.
Tailings and Mine Waste ’10 (pp. 3-10). London, ON: CRC Press Taylor & Francis Group.

Natural Resources Canada. (2018) Crude oil facts. Available at

https://www.nrcan.gc.ca/energy/facts/crude-oil/20064

11
Wickland, B.E. and Longo, S. (2017) Mine waste case examples of stacked tailings and co-disposal.
Proceedings of the 21st International Conference on Tailings and Mine Waste, November 5-8. Banff,
AB.

Wilson, G. W., Kabwe, L. K., Donahue, R., & Lahaie, R. (2011). Field performance of in-line flocculated
fluid fine tailings using thin lift deposition. In A. Fourie, M. Tibbett & A. Beersing (Eds.), Mine
Closure 2011 (pp. 473-481). Perth, WA: Australian Centre for Geomechanics.

12
Optimization of Tailings through Mineral Processing
Päivi Kauppila1, Antti Taskinen2, Marja Lehtonen2, Neea Heino1, Matti Kurhila2 and Mia
Tiljander2
1. Industrial Environments and Recycling, Geological Survey of Finland, Finland
2. Mineral Processing and Materials Research, Geological Survey of Finland, Finland

ABSTRACT
Hazardous elements and minerals in tailings limit their usability and pose challenges for their
disposal. Removal of these hazardous substances prior to the tailings disposal would increase the
usage potential of waste materials and the raw material value of the ore. In addition, it would
decrease the costs of waste management, as the amounts of the most hazardous waste requiring
costly facilities would decrease.
One option to increase the raw material value of the ore is to optimize the tailings properties during
the mineral processing simultaneously with the recovery of valuables from the ore. This would tackle
the challenge of waste management already before the waste becomes waste, and would also
facilitate production of new mineral products from ores.
In this study, a cyclical model for the optimization of tailings was first developed to improve the
environmental performance of tailings and to decrease the amounts of hazardous waste. The model
was then tested with samples from the Kopsa Au-Cu ore deposit located in Western Finland. In the
tests, the main focus was to reduce As and sulfide minerals from the tailings into a smaller waste
fraction. The testing included detailed mineralogical investigations, mineral processing tests and
comprehensive environmental characterization of the waste materials.
The results showed that optimization of tailings requires multiple data on tailings properties to
evaluate their raw material potential as well as environmental and management performance.
However, with this type of predictive approach it is possible to separate the tailings into several
fractions taking into account their properties instead of just disposing all the material as one. The case
study of the Kopsa ore further showed that modification of tailings properties can be made with
mineral processing without jeopardizing the recovery of the valuables.

1
INTRODUCTION

Marked amounts of tailings are generated annually in mining operations throughout the world. Their
management is challenging and costly, since they often contain hazardous elements and minerals,
which also limit their usage. This applies particularly wastes, which contain sulphide minerals that
are prone to cause acid mine drainage leading to deterioration of water quality of recipient waters as
the sulphides oxidize. Removal of hazardous compounds would ease the management of tailings,
decrease the environmental impacts and also increase the usability of wastes, but also increase their
raw material value, since often the hazardous compounds may also be the residual valuable metals
in the wastes. As such, it would also improve the profitability of mining operations.
One option to decrease hazardous compounds from tailings and to increase its raw material value is
to use mineral processing to modify tailings properties. For example, Lv et al. (2014) have used froth
flotation and Liu et al. (2018) bioleaching to recover metals from tailings. In addition, flotation has
also been applied to remove As from tailings (e.g. Choi et al., 2013), and high-gradient magnetic
separation (HGMS) has been tested to recover valuables from mine dumps and tailings (Watson &
Beharrel, 2006).
In this study, the objective was to use mineral processing to optimize environmental properties of
tailings simultaneously with the recovery of valuables from the ore. At first, a cyclic model was
developed for the optimization, and then the model was tested with the Kopsa Au-Cu ore, located in
Western Finland (Kauppila et al., 2018, Taskinen et al., 2018). In the Kopsa case, the aim was to
decrease As and sulphide mineral content from the tailings by modifying the mineral processing flow
sheet. Tailings were characterized before and after the modifications made in the processing for their
environmental properties such as mineralogy, chemical constituents, potential to produce acid mine
drainage and leachability of elements. The focus of this paper is to describe the changes achieved in
the environmental performance of the tailings.

METHODOLOGY
To improve the properties of tailings, a cyclic model for their optimization was first developed. This
included review of the requirements for the environmental characterisation in Finland and in Europe,
and evaluation of the methodology needed in the optimization.
The model was then tested with the Kopsa Au-Cu ore, located in Western Finland. The Kopsa ore
deposit is a porhyric Au-Cu mineralisation, which is hosted by tonalite and mica schist. The major
ore minerals include arsenopyrite, chalcopyrite and pyrrhotite with accessory e.g. löllingite,
marcasite, pyrite, sphalerite, and gold (Gaál & Isohanni, 1979). The estimated ore reserves in the
deposit are 13.6 Mt with 0.81 g/t Au, 0.15% Cu and 2.15 g/t Ag (SRK, 2013). The deposit contains
notable arsenic (0.1-1.2 %, Gaál & Isohanni, 1979, Nurmi et al. 1991). Belvedere Mining Oy was
planning to start to exploit the deposit, but went bankrupt in 2015.

2
As a starting point for the testing, detailed mineralogy and chemical content of the ground ore
samples were measured to identify main contaminants and hazardous minerals and to plan mineral
processing. Mineralogical investigations were carried out using a scanning electron microscope
(SEM) with two energy dispersive detectors (EDAX Genesis) equipped with a mineral liberation
analyser (MLA). Electron probe microanalyser (EPMA) was applied for more detailed identification
of some specific mineral phases and their quantitative composition. Chemical analyses of the ore
included total element concentrations measured with XRF, and Au and Ag determinations with fire
assay (FA).
Original mineral processing was carried out in accordance with the plan made by Belvedere Mining
Oy, and it included crushing and grinding of the ore followed by a two stage froth flotation to
separate Cu concentrate and sulphide concentrate including Au and Ag. The main variables and
methods tested to modify tailings were grind size, grinding media, flotation time, dosage and type
of flotation chemicals, pulp pH, flotation gas, Knelson gravity concentration and magnetic
separation. During mineral processing, the success of the modifications was monitored by measuring
tailings’ total concentrations of elements with XRF, Au and Ag grades by FA completed with flame
atomic absorption spectroscopy (FAAS), and sulphur assays using an Eltra carbon/sulphur analyser.
Detailed characterization of tailings was carried out from the original process and from the finalized
modified process. Characterization included mineralogy (MLA and EPMA methods described
above), chemical composition (XRF, aqua regia (AR) leach; pyrolytical measurement of total-S, total-
C, sulphide-S, CO3-C), potential to produce acid mine drainage (ABA test in accordance with Finnish
Standards Association, 2012), and leaching of harmful elements. AR digestion was used to measure
the concentrations that are most likely of concern in the drainage from waste (e.g. Fosso Kankeu et
al., 2015) and to evaluate whether mining waste is inert (Government Decree, 2013a). Leaching of
harmful elements from the tailings was studied using standardized leaching tests (Finnish Standards
Association, 2002, 2017) as required by the Finnish mining waste Decree (Government Decree, 2013a).

RESULTS AND DISCUSSION

Model for the optimization

Based on the review of the legislative requirements for the characterization of tailings and the
objective to increase the recovery of raw materials from the ore, a cyclic model for tailings
optimization during mineral processing was drafted (Fig. 1). In the model, it was seen important to
consider the characteristics of tailings already during bench-scale mineral processing tests, because
in bench-scale it is possible to study in detail the adjustments to the process needed to have positive
influence on the tailings. In addition, the characterization of process water was added in the model
to evaluate influence of the modifications on water quality to identify water treatment requirements,
and also to evaluate possible loss of valuables in the waters.

3
 Figure 1 Model for the optimization of tailings in mineral processing

Certain waste characteristics need to be determined due to legislative requirements for defining their
waste classification and official requirements for long-term disposal. In Europe, the environmental
classification and characterisation of mining wastes is regulated by the European Union directives
(EU, 2006, 2009a, 2009b), and national legislation (Government Decree, 2013a, 2013b). The required
characterisation includes geochemical, mineralogical, geotechnical, and radiological properties of
mine wastes together with the determination of process chemical residues (cf. also European
Committee for Standardization, 2012). All these were included in the model (Fig. 1). Based on these
properties, the class of the waste is defined in accordance with the waste decree (Government Decree,
2012, cf. also European Council, 2003) and the management actions are defined (cf. EU, 2018).
The cyclic approach requires suitable data from the tailings and process water characterization to
evaluate the influence of the modifications, and whether additional actions are still needed. In
particular, mineralogical investigations are crucial to identify hosts for both the contaminants and
valuables in the waste, and also their mode of occurrence (e.g. associations, degree of liberation).
After finalizing the process flow sheet in batch tests, a mini-pilot or pilot scale testing is needed to
provide optimum conditions for removing the target minerals in a continuous, commercial plant.
Optimally, new raw materials can be produced or waste fractions with variable environmental
properties generated with a limited amount of the most hazardous waste (Fig. 1). As a result, eco-
efficiency of raw materials can be increased, management of mining wastes improved, and costs of
both mining and waste management decreased.

4
Testing of the model to improve environmental performance of the Kopsa tailings

The proposed model was tested to remove As and sulphides from the Kopsa tailings. Based on the
chemical and mineralogical analysis, the Kopsa Au-Cu ore contained As (0.74%), S (0.59%), and Cu
(0.10%), with arsenopyrite (1.7%), pyrrhotite (0.48%), chalcopyrite (0.32%), pyrite (0.20%), goethite
(0.15%), and löllingite (0.01%) as the main hosts. In addition, various arsenate minerals such as
pharmacosiderite, arseniosiderite, and scorodite (totally 0.04%) hosted As.
The original process flow sheet generated a non-inert, non-acid producing tailings with elevated S
(0.09%), As (0.03%) and Cu (0.015%). The primary sources for As in the tailings included arsenopyrite,
löllingite and arsenates (e.g. scorodite/parascorodite, and pharmacosiderite or arseniosiderite type
mineral). Also goethite contained minor As (As2O5 0.04–5.9%). The main hosts for S, aside from the
As-sulphides, were pyrrhotite, chalcopyrite, pyrite, sphalerite, covellite and goethite. Cu originated
mainly from chalcopyrite and covellite, and small amounts of Cu occurred also in goethite.
Based on the results, the mineral processing flow sheet was modified to increase removal of arsenic
minerals and sulphides. After a series of testing, a combination of adjusted froth flotation with high-
gradient magnetic separation (HGMS) appeared the most promising. It decreased the total As and S
content of tailings by ca. 67% (from 0.03% to 0.01%) and 89% (from 0.09% to <0.01%), respectively, as
a result of successful removal of arsenopyrite, arsenates and Fe sulphides (Fig. 2a, 2b). The decrease
in the sulphide content additionally reduced the acid production potential (AP) of the tailings (from
2.7 to <0.3 kg CaCO3/t). In addition, the amount of the final tailings decreased from 90% to 45% of the
total processed mass.

Figure 2 a) Total concentrations of S and As, b) amounts of As minerals and sulphides, and c) leaching of SO 4
and As from the original and HGMS (modified) tailings

Modifications had an even higher impact on the leaching of As and SO4 from the tailings than their
overall contents. The results from the leaching tests showed that leaching of As decreased by ca. 85%
(from 26 to 3.8 mg/kg) and that of sulphate by ca. 94% (from 162 to 9.9 mg/kg) from the original
tailings to the HGMS tailings (Fig. 2c). In addition to their lowered content, this was due to changes

5
in the degree of liberation of As minerals and sulphides. In the HGMS tailings, these mineral phases
occurred in somewhat higher portion associated with silicates than in the original tailings.
Despite the notable improvements, the tailings were still classified as non-inert, since the As content
still exceeded the threshold value (5 mg/kg) for inert waste, even though S content was below the
limit (0.1%) (Government decree, 2013a). However, the leaching of As no longer exceeded the limit
for hazardous waste to be deposited at a hazardous waste landfill (25 mg/kg; Government decree,
2013b).
With the improved properties, the modified tailings could most likely be used for example as a cover
material in waste facilities. This however requires further research on their long-term behaviour.
Overall, even though the tailings properties were improved, the recovery of gold was not jeopardized
in the process. Preliminary results from additional tests, where classification was added after HGMS,
indicated that an even lower arsenic content (0.001%) could be achieved with further modifications
to the process.

CONCLUSION
The developed model for the optimization of tailings enables improving the environmental
performance of tailings by tailoring mineral processing methods in order to increase utilization of
tailings and to minimise the amount of hazardous waste. Testing of the model with the Kopsa case
study showed that it is possible to have a positive impact on the tailings’ environmental performance
through modifying the originally planned process flow sheet. The content of As in the Kopsa tailings
decreased by more than 60% and S almost 90% by adapting froth flotation and adding high-gradient
magnetic separation in the process. In addition, leaching of As and SO4 was decreased by ca 85% and
94%, respectively.
The Kopsa case study underlined the importance of mineralogical investigations for monitoring and
designing the beneficiation processes as well as for characterising the environmental properties of
tailings. To fully benefit from the operational model and to increase the raw material value of
excavated ores, an overall change in the approach of the mineral sector is needed to consider the raw
material as a whole already at the beginning of the mine planning.

ACKNOWLEDGEMENTS

The authors thank Belvedere Mining Oy for providing the ore samples for the study and ERDF
funding programme, Boliden Kevitsa Mining Oy, FQM Kevitsa Mining Oy, Kemira Oyj and
Endomines Oy are acknowledged for co-funding the research.

6
REFERENCES

European Committee for Standardization (2012) Characterization of waste. Overall guidance document
for characterization of waste from the extractive industries, CEN/TR 16376, European Committee for
Standardization.

Choi, J., Park, K., Hong, J., Park, J. & Kim, H. (2013) ‘Arsenic removal from mine tailings for recycling
via flotation’ Materials Transactions, vol. 54, pp. 2291–2296.

EU (2006) Directive 2006/21/EC of the European Parliament and of the Council of 15 March 2006 on the
management of waste from extractive industries and amending Directive 2004/35/EC, Official Journal of the
European Union L 102/15, viewed 23 May 2018, https://eur-lex.europa.eu/oj/direct-access.html.

EU (2009a) 2009/359/EY, Commission Decision of 30 April 2009 completing the definition of inert waste in
implementation of Article 22(1)(f) of Directive 2006/21/EC of the European Parliament and the Council
concerning the management of waste from extractive industries (notified under document number C(2009)
3012), Official Journal of the European Union L 110/46, viewed 25 May 2018, https://eur-
lex.europa.eu/oj/direct-access.html.

EU (2009b) 2009/360/EC. Commission decision of 30 April 2009 completing the technical requirements for
waste characterisation laid down by Directive 2006/21/EC of the European Parliament and of the Council on
the management of waste from extractive industries (notified under document number C(2009) 3013), Official
Journal of the European Union L 110/48, viewed 15 May 2018, https://eur-lex.europa.eu/oj/direct-
access.html.

EU (2018), Best Available Techniques (BAT) Reference Document for the Management of Waste from
Extractive Industries, JRC Science for policy report, viewed 5 January 2019,
https://ec.europa.eu/jrc/en/publication/eur-scientific-and-technical-research-reports/best-available-
techniques-bat-reference-document-management-waste-extractive-industries.

European Council (2003) Council decision on 19 December 2002 establishing criteria and procedures for the
acceptance of waste at landfills pursuant to Article 16 of Annex II to Directive 1999/31/EC (2003/33/EC),
Official Journal of the European Communities L11/27. viewed 10 May 2018, https://eur-
lex.europa.eu/oj/direct-access.html.

Finnish Standards Association (2002), Characterisation of waste, Leaching, Compliance test for leaching of
granular waste materials and sludges. Two stage batch test at a liquid to solid ratio of 2 l/kg and 8 l/kg for
materials with a high solid content and with a particle size below 4 mm (without or with size reduction), SFS-
EN 12457-3, Finnish Standards Association.

Finnish Standards Association (2012), Characterization of waste. Static test for determination of acid
potential and neutralization potential of sulfidic waste, SFS-EN 15875:en, Finnish Standards Association.

Finnish Standards Association (2017), Characterization of waste. Leaching behaviour test, Up-flow
percolation test (under specified conditions), SFS-EN 14405:2917:en, Finnish Standards Association.

7
Fosso-Kankeu, E., Waanders, F., Mulaba-Bafubiandi, A. & Sidu, S. (2015) ‘Leachability of Suspended
Particles in Mine Water and Risk of Water Contamination’ In: Brown, A., Bucknam, C., Burgess, J.,
Carballo, M., Castendyk, D., Figueroa, L., Kirk, L., McLemore, V., McPhee, J., O’Kane, M., Seal, R.,
Wiertz, J., Williams, D., Wilson, W. & Wolkersdorfer, C. (eds) 10 th ICARD & IMWA 2015 Conference
– Agreeing on solutions for more sustainable mine water management”, Santiago, Chile 2015, viewed
30 January 2018, https://www.imwa.info/docs/imwa_2015/IMWA2015_Fosso-Kankeu_286.pdf

Gaál, G. & Isohanni, M. (1979) ‘Characteristics of Igneous Intrusions and Various Wall Rocks in Some
Precambrian Porphyry Copper-Molybdenum Deposits in Pohjanmaa, Finland’ Economic Geology, vol.
74, pp. 1198–1210.

Government Decree (2012) Government Decree on waste 179/2012, viewed 20 June 2019,
https://www.finlex.fi/fi/laki/kaannokset/2012/en20120179.pdf.

Government Decree (2013a) Government Decree on Extractive Waste 190/2013, viewed 15 March 2018,
https://www.finlex.fi/fi/laki/kaannokset/2013/en20130190_20150102.pdf.

Government Decree (2013b) Government Decree on Landfills 331/2013, viewed 18 March 2018,
https://www.finlex.fi/fi/laki/kaannokset/2013/en20130331_20160960.pdf.

Kauppila, P.M., Lehtonen, M. & Heino, N. (2018) ‘Optimisation of raw material potential and
environmental properties of tailings: an operational model’ Geological Survey of Finland, Bulletin, vol.
408, pp. 97–108, http://tupa.gtk.fi/julkaisu/bulletin/bt_408.pdf.

Liu, Y.-G., Zhou, M., Zeng G.-M., Wang, X., Li, X., Fan, T. & Xu, W.-H. (2008) ‘Bioleaching of heavy
metals from mine tailings by indigenous sulfur-oxidizing bacteria: Effects of substrate concentration’
Bioresource Technology, vol. 99, pp. 4124–4129.

Lv, C. C., Ding, J., Qian, P., Li, Q. C., Ye, S. F. & Chen, Y. F. (2015) ‘Comprehensive recovery of metals
from cyanidation tailing’ Minerals Engineering, vol. 70, pp. 141–147.

Nurmi, P. (1991) ‘Geological setting, history of discovery and exploration economics of Precambrian
gold occurrences in Finland’ Journal of Geochemical Exploration, vol. 39, pp. 273–287.

Räisänen, M.L., Kauppila, P.M. & Myöhänen, T. (2010) ‘Suitability of static test for acid rock drainage
assessment of mine waste rock’ Bulleting of the Geological Society of Finland, Volume 82, pp. 101−111.

SRK Consulting (2013) Preliminary economic assessment for the Kopsa copper-gold deposit, Finland, Report
by SRK Consulting (Sweden) AB, October 2013, 168 p.

Taskinen, A., Kauppila, P.M., Tornivaara, A., Heino, N., Kurhila, M., Tiljander, M. & Korhonen, T.
(2018) ‘Improving the environmental properties of arsenic and sulphide rich Kopsa Au–Cu ore
tailings through optimised mineral processing’, Geological Survey of Finland, Bulletin vol. 408, pp. 59–
82, http://tupa.gtk.fi/julkaisu/bulletin/bt_408.pdf.

Watson, J. H. P. & Beharrell, P. A. (2006) ‘Extracting values from mine dumps and tailings’ Minerals
Engineering, vol. 19, pp. 1580–1587.

8
Accelerated Mechanical Consolidation (AMC) to
Improve Tailings Storage Sustainability
David Smirk
Phibion, Australia

ABSTRACT
The overall purpose of a tailings storage facility is to provide a safe, stable landform with enduring
containment of tailings and with minimal impact on the local environment and community. However,
as sustainable operations only exist if financial benefit is realised, then it follows that processes
delivering these benefits must also be both technically feasible and cost-effective.
Accelerated Mechanical Consolidation (AMC) is an integrated process where planned layer deposition
of tailings is combined with mechanical working of these over-saturated materials to expedite the
transition from slurry to solid. The process is primarily delivered through enhancing the particle
packing density through repeated passes utilising specialised equipment to improve drainage and
consolidation. This process is proven and highly predictable and delivers dewatering rates orders of
magnitude faster than self-weight processes.
AMC can also be delivered as an operating function without modification to existing tailings storage
facilities or require supporting capital expenditure. Water loss is primarily through drainage and
consolidation processes with only limited dependence on evaporative loss. Performance approaches
levels only achieved by filtration and centrifugation processes but with a potential benefit of being
operating expenditure only.
The AMC process needs to be designed based tailings properties, but reductions in tailings volume
exceeding 30% with increases in tailings density in excess of 10 - 20% are common. Additionally,
increases in density of this magnitude lead to increased water recovery and significant tailings strength
often in excess of 30 kPa. These improvements will lead to a reduction in the required operational
tailings management footprint with its associated environmental and financial benefits.
This paper will discuss some of the underlying aspects of the AMC process and show several examples
of its successful application.

1
INTRODUCTION

The concept of sustainable tailings management is not new. It has been explored in detail by the
Ministerial Council on Mineral and Petroleum Resources (MCMPR) & Minerals Council of Australia
(MCA) (2003) and Department of Foreign Affairs and Trade (DFAT) (2016). However, these concepts
tend to focus on the tailings delivery system as the decision point on technology rather than what is
achieved in the Tailings Storage Facility (TSF) and how that matches with the desired operational and
closure goals. In doing so, this forces the operator to consider potentially high capital cost solutions for
tailings conditioning prior to deposition as the only option to develop a sustainable tailings
management system
Accelerated Mechanical Consolidation (AMC) is a low-cost in-TSF management approach that can
achieve operational and closure goals without the high capital expenditure required of other advanced
tailings technologies. By removing the upfront CAPEX burden, AMC reduces overall tailings
management costs and hence contributes to enhancing the profitability of an operation, maximising the
value that can be extracted from an orebody along with sustaining employment and community values.

METHODOLOGY
In response to recent tailings dam failures, increased environmental awareness, land access restrictions
and community expectations, there has been a progressive trend in industry to:
 Improve confidence in tailings dam management;
 Reduce operational footprints and land acquisition costs;
 Reduce potential environmental impacts;
 Improve water recovery to reduce the demand on virgin water sources;
 Reduce business risk of failure; and
 Leave a completed structure that can be safely rehabilitated to support community supported
post-closure land use.
This has given rise to a range of tailings technologies with a primary focus on delivering higher tailings
deposition densities. Putting the technology to one-side, the financial end points have transitioned from
low capital and operating cost to high capital and operating cost. DFAT (2016), adapted from Davies &
Rice (2004) provides an overview of the tailings continuum in Figure 1.

2
 Figure 1 Tailings Continuum – DFAT (2016), adapted and from Davies and Rice (2004)

Like any aspect of a mining/processing operation, a tailings technology must satisfy a technical
requirement and a financial requirement. A cost prohibitive tailings management technology,
regardless of technical excellence, does not ensure sustainability. If the economics do not support a
tailings technology, then it is likely that the project will not proceed or it will force early closure of an
established project. This also has significant social and community impacts.
The goal of a sustainable tailings management solution is to adopt a tailings technology that:
 responsibly manages the tailings stream to support the operation, ensure operational safety,
delivers enduring containment and a stable deposit, protection of the environment, maximum
recovery of water resources and progressive rehabilitation;
 achieves performance levels that will be accepted by all stakeholders (community);
 is scalable to an operator as the business ebbs and flows; and
 targets a minimum capital and operating cost, during operations, closure and post-closure, to
maximise the economic value of the development.
Below we describe a tailings technology and philosophy that embraces the lower capital and operating
costs offered by slurry and paste systems, while delivering performance that permits efficient water
recovery, minimal water content, negligible seepage losses, progressive rehabilitation and a stable
tailings mass. This approach is counter-intuitive and breaks the link between conventional
understanding that high tailings discharge density is required to deliver a high final tailings density.

ACCELERATED MECHANICAL CONSOLIDATION (AMC)

Accelerated Mechanical Consolidation (AMC) or “Mud Farming” is the mechanical working of tailings
to separate the solids tailings from the interstitial water. It involves repeated ploughing of a tailings

3
layer from the point of deposition to the point of water recovery to encourage drainage, consolidation
and evaporation. The AMC process is described in detail in Munro & Smirk (2012), Munro & Smirk
(2013) and Munro et al. (2014).
AMC is a proven, established technology and is used in a wide range of commodities. AMC is listed as
a best practice technology by the International Aluminium Institute (IAI 2014) and is provided as an
example of Best Available Technology (BAT) by Brett et.al (2015) as part of a review of the Mt Polley
Tailings Dam Failure.
Whilst seemingly a simple method to improve tailings density, the AMC process is a complex
combination of science/engineering disciplines including geotechnical principles, rheology,
meteorology, information technology, mechanical and process engineering.
AMC is most efficient and effective when partnered with an Archimedes Screw Tractor (MudMaster ®).
The Archimedes Screw Tractor is not a new invention. Pioneering work in the USA, Holland and Russia
in the mid 20th century highlighted the capability and robustness of this technology. Newer versions
have improved the reliability of this design. However, the primary purpose of these designs is to access
soft areas that cannot be reached by any other vehicle. They were never designed as a zero-shear vehicle
hence are not directly suitable for efficient tailings dewatering.
However, using the Archimedes Screw Tractor as a starting point, the MudMaster® is very different in
that it is specifically customised to the required tailings properties to ensure safe, stable operation,
maximum productivity and minimal shear. It obtains its mobility through the action of the counter-
rotating scrolls allowing the machine to safely traverse soft terrain, without shear. This unique, shear-
less operation (shear strength < 2 kPa), when carefully matched to both the properties of the tailings,
the depth of deposition and the design of the machine, allows highly productive dewatering of tailings
at rates many times faster than natural consolidation processes.

Figure 2 MudMaster®

4
AMC is a mineral processing operation not a civil solution. For best results it needs to be managed like
a processing operation. This means a defined operations schedule, continuous monitoring of
dewatering progress and clearly identified key performance criteria for best results.
AMC operations require the ploughing of the tailings along the angle of repose from the point of
tailings deposition to the designated point of water collection. The machine scrolls engage on the
underlying dewatering layer to achieve maximum traction. The action of the MudMaster® forces
entrained tailings water to the surface where it collects in the shallow surface drains created by the
passage of the machine. The accumulating water then drains along the scroll lines to the designated
decant collection point. This water recovery continues with diminishing recovered water volumes with
each pass and a gradual increase in tailings density. Once this water is removed, the MudMaster®
reaches a critical buoyancy point where the tailings begin to resist the presence of the machine and the
machine begins to ride up on top of the tailings thus providing compaction, further enhancing density
improvement. Generally, 70% of overall dewatering effectiveness from AMC is attributable to drainage
and consolidation.
While tailings will always dry to some degree under evaporation in non-AMC managed operations,
the effectiveness of this process is usually limited to the surface layers, requiring very thin layer
deposition to be placed over very large areas to be an effective management option. Under AMC,
thicker layers can be deposited (in layers up to 1,200 mm) thereby reducing the area requirements. After
several passes of the machine, nearly all interstitial water has been removed by drainage and the
tailings begin to exhibit some strength allowing the tailings to open up (and remain open) along the
scroll lines. This behaviour significantly expands the area of the evaporating surface leading to
enhanced dewatering by evaporation at this point. Overall AMC moisture losses associated with drying
rarely exceed 30%. However, this highlights the importance of effective drainage. In the absence of
well-maintained and adequate drainage systems performance can quickly deteriorate and the time
required for dewatering increases often by a factor of 3x.
Importantly, the AMC dewatering performance is largely dictated by the tailings properties with the
process controls aligned to harness these properties. Only with the application of the correct controls
will an AMC process deliver optimum results. Ongoing monitoring of multiple data streams collected
during the process via telemetry (such as productivity, machine attitude, traction and torque demand)
enable real time mapping and optimisation of the AMC process.
Once the tailings have met the required water recovery, target density and/or strength, another layer
of tailings can be deposited to the designated depth and the process repeated. The time required for
full dewatering is a function of both the particle size distribution, tailings mineralogy, hydraulic
conductivity, flocculation strategy (if applied) and settled and target final density. The number of
machine passes required is not prescriptive nor comparable with more commonly available civil
earthmoving equipment.

AMC OUTCOMES

Rapidly removing the tailings water increases water recovery, tailings density and generates tailings
strength. Using this process:

5
  AMC process can be designed and applied as an operations-only model fully supported by an
integrated operational plan. This means that in most cases these improvements can be realised
without any capital expenditure.
 AMC can be applied to existing TSF’s without modification to the slurry distribution or water
recovery network.
 AMC is a predictable, robust process that is applied with high discipline. Commissioning
requirements are minimal and tailings volume reduction benefits accrue immediately.
Additionally, operational costs are a minimal and predictable.
 Progressive rehabilitation can commence the moment AMC operations cease. There is no
further maturing or drying required. In most cases, operations with low ground pressure
equipment can safely commence as ground conditions are already above 30 kPa.

CASE STUDIES

AMC was originally developed in the alumina industry but has since been expanded into copper, coal,
mineral sands, rare earths, oil sands, nickel and phosphate tailings. Some examples are as follows:
Alumina
A Jamaican based alumina refinery was facing a key tailings storage decision regarding sustainable
tailings technology to support the operation for the foreseeable future. Critical issues included severe
restrictions on available capital, difficulties associated with TSF footprint expansion and the time
required to commission a suitable alternate solution.
Additionally, Jamaican bauxite residue tailings are recognised as extremely difficult to manage due to
ultra-fine particle size distribution. Extensive flocculation research identified a suitable flocculant that
addressed these issues and resulted in a tailings that was not dominated by the presence of these ultra-
fine materials.
Tailings are deposited at 4,600,000 m3/year at a density of 1.29 t/m3 (P50 <1μm). If left to natural
consolidation and drying processes, this tailings volume reduces to 2,260,000 m3/year for a final density
of 1.5 t/m3 (bulk). However, at this density there is minimal tailings strength limiting access and
rehabilitation timing.
Using AMC, final tailings dry densities have been increased to 1.8 t/m3 (+20% improvement) by
reducing the tailings volume 30% to 1,570,000 m3/year within 42 days of deposition. At this density, the
undrained shear strength of the tailings exceeded 30 kPa allowing access on the tailings surface for
sampling, construction and rehabilitation. The relationship between tailings density and undrained
vane shear strength is shown in in Figure 3.

6
 Figure 3 Undrained Shear Strength vs Tailings Density

With the higher density, operational footprint has reduced to 80 ha (from 250 ha) further reducing
capital expenditure pressures and allowing improved water balance control. This transition is shown
in Figure 4.

Figure 4 AMC Timelapse – Alumina (42 days from time of deposition to 30 kPa)

AMC was selected as the preferred technology because:

 Zero changes to tailings properties or operations were required;
 Zero capital expenditure was required;
 Operations could commence immediately; and
 No engineering design, construction implementation or commissioning period was required;

7
Rare Earths

A rare-earth operation was seeking a sustainable solution to tailings management. Mined ore required
exceptional grinding resulting in an ultra-fine, hard-rock tailing that resists densification efforts. This
meant that required TSF structures were large and cost prohibitive.
Tailings are deposited at 850,000 m3/year at a density of 1.34 t/m3 (bulk) (P80 <4μm). If left to natural
consolidation and drying processes, the tailings volume reduces to 570,000 m3/year for a final density
of 1.5 t/m3 (bulk). However, at this density there is minimal tailings strength, limiting access and
rehabilitation timing.
Using AMC, final tailings densities have been increased to over 1.7 t/m3 (bulk) (+13% improvement) by
reducing the tailings volume 30% to 400,000 m3/year within 49 days of deposition. At this density, the
undrained shear strength of the tailings exceeded 30 kPa, allowing access on the tailings surface for
sampling, construction and rehabilitation. This has reduced the size of the TSF structures significantly
reducing capital cost as well as slowing the rate of capital expenditure. Before and after photographs
of this transition are shown in Figure 5.

Pre – Operations Pass #15

Day 0 Day 49

Figure 5 – AMC Timelapse – Rare Earths (49 days from time of deposition to 30 kPa)

AMC was selected as the preferred technology because:

 Zero changes to tailings properties or operations were required;
 Zero capital expenditure was required;
 Operations could be applied flexibly with no financial penalty; and

 Operational costs associated with AMC were less than the cost of construction allowing
operating expenditure to recover capacity within the existing TSF for less than the cost of
constructing new facilities.

CONCLUSIONS
The application of AMC to fine tailings provides a potential alternative to high CAPEX dewatering
technologies. AMC embraces the performance levels demanded by sustainable tailings management

8
while providing a more flexible cost structure. Additionally, the technology can be designed and scaled
to any sized operation.
While AMC can be applied to new operations, its greatest advantage is that it can be applied to
established operations that have already invested in slurry-based tailings management systems. No
additional capital expenditure is required, and in most cases, it can be applied to an existing system
with minimal operational changes required to ensure success.
Sustainable tailings management demands continual improvements in tailings management to ensure
best efforts are applied to improve the safety of both operating and closed TSF’s, support progressive
rehabilitation of completed facilities and ensures containment and protection of our precious
environment. A key factor in this approach is ensuring that the cost of the tailings solution does not
become a barrier to its implementation. AMC is an innovative solution that challenges the current
methodology and allows a compromise to incorporate low-cost, low-risk, robust technologies while
delivering best-practice performance.

ACKNOWLEDGEMENTS
The author is grateful for the extensive assistance provided by Ecolab/Nalco both for Jamaican
laboratory trials and in-field support.

REFERENCES
Brett, D., Longey, R, & Herza, J. (2015). ‘Abracadabra – the Disappearing Tailings Dam’. ANCOLD
Proceedings 2015: Contemporary Challenges for Dams

Davies, MP, Rice, S (2004). ‘An alternative to conventional tailings management: ‘dry stack’ filtered
tailings’, Proceedings of Eighth International Conference on Tailings and Mine Waste (pp. 411–422), Vail,
USA, 10–13 October.

DFAT (2016). Tailings Management – Leading Practice Sustainable Development Program for the Mining
Industry. Canberra Australia: Department of Foreign Affairs & Trade.

IAI (2014) Bauxite Residue Management Best Practice Guide, International Alumina Institute, World
Aluminium and European Aluminium Association.

MCMPR & MCA (2003) Strategic Framework for Tailings Management. Canberra: The Ministerial

Council on Mineral and Petroleum Resources, and Minerals Council of Australia.

Munro, L.D. and Smirk, D. (2012) Optimising bauxite residue deliquoring and consolidation, in
Proceedings 9th International Alumina Quality Workshop, March 2012, Perth, Australia.

Munro, L.D. and Smirk, D. (2013) ‘Mud farming of fine tailings – application and benefits of
MudMaster® technology’, in Proceedings 17th International Conference on Tailings and Mine Waste, G.W.
Wilson, D.C. Sego and N.A. Beier (eds), 3–6 November 2013, Banff, Canada, pp. 15–23.

Munro, L. D, White, R. W. & Smirk, D. D., 2014, ‘Post-deposition management - the next step in paste
and thickened tailings management.’ R. J. Jewell, A. B. Fourie, P. S. Wells and D. van Zyl (eds), Paste
2014 – Proceedings of the 17th International Seminar on Paste and Thickened Tailings June 8 – 12, 2014,
Vancouver, Canada.
9
Hydraulic Disposal of Sand for Dams
Pablo Vásquez
SHIMIN Ingeniería, Chile

ABSTRACT
Tailings disposal in a "conventional" deposit endowed with a resistant dam, constructed with sands
coming from the classification of the complete thickened tailings, is the most used and proven
technology with good results in Chile.
The sand dams built up to the present, in their period of beginning of operations, have presented
deviations whose direct cause is the "youth" of the dam itself and the drainage system whose
"priming" takes a certain time to complete. This is the case of the start-up of large-scale mining
projects, with the initial dam rising at a very high speed, (> 3 m / month). The author has observed
the application of certain criteria that have mitigated the occurrence of such deviations in the start-
up, but with a great impact on already normalized operations.
This article contains a well-founded list of criteria to be considered during the construction of sand
dams (in start-up and normal operation), with the hydraulic method. These criteria are applied both
to the design, for example, the sizing and deployment of discharge pipes, initial parapets), and in the
procedures and competences to be acquired by the operations staff.

1
INTRODUCTION

Tailings management resulting from the minerals benefit process, is a matter that without having a
direct economic value for the company, constitutes a peripherical process that provides continuity,
safety and soundness to the company’s business, for these reasons, safe tailings disposal is vital for
the plant operation.
Currently, the company, apart from managing its tailings according to the highest safety standards
in its operations, is required to comply with authorizations, commitments and the community, as
well as the strict risk controls, which are increased in larger sized deposits, according to the volume
of the mining waste.
Tailings disposal technology, in medium size and large mining operations, mainly operates with
thickened slurry, at weight concentrations from 52-55%, in most concentrate plants, up to weight
concentrations of 65% and over, among them, paste of 65%-70% density and very high rheologic
parameters can be distinguished. The latter, even though they have shown important advantages in
water recovery and a lower direct impact on the area required for the disposal, at least for medium
size and large mining operations, have shown difficulties, whose solutions are being developed and
consolidated, having an impact on high investment and operations costs.
Tailings disposal technology in deposits provided with resistant dams built out of the thick fraction
(sands) resulting from the tailings classification, constitutes the most used and proven technology
with good results in Chile.
Considering that the dam construction process (production, placement and certification) is a
continuous process that must meet a specification and a schedule, it must systematically ensure the
quality defined to comply with the dam stability and safe behavior, in front of hydrological and
geotechnical contingencies that may occur around the location.
This article contains a list of operational and design criteria, based on the author’s operational
experience (Quillayes 1999-2008 and El Mauro 2008-2009 for MLP Chile, and Ovejería 1998-2003 for
CODELCO, Cerro Verde 2005 for MCV Peru), which the author proposes to consider in sand dam
construction, start up and normal operations.
Among the methodologies applied in the different deposits currently in operation, the hydraulic
method has shown to be the simplest and involving lower operational costs. These criteria shall be
considered in the design and operation procedures.

METHODS OF SAND DISPOSAL ON DAMS

Method of Sand Placement in Dams

The three following methods of sand disposal in dams are the most commonly used in the mining
industry.

2
1 Mechanized Method: Consists of a sand stockpile outside or inside the dam, transportation by
truck to the destination point (Tranque Talabre, Chuquicamata, Chile), and the massive use of
machinery (Quebrada Honda, Cerro Verde 1, both in Peru), (bulldozers) and trucks for sand
movement.
2 Paddocks Method: Placement of dams by means of pipes in confined areas of 50m x 50m
(terraces), which are being reproduced according to the growth. It considers the intense use of
machinery (bulldozers) and trucks for sand movement.
3 Hydraulic Method: Placement and transportation of sand by means of a pipeline located at the
existing dam slope/incline (all Chilean major dams). The use of machinery is reduced only to
compacters and the paddocks formation at the beginning of the dam construction.

Figure 1 Methods of sand disposal on dams

In the next section, a description of the hydraulic method is presented, including its advantages and
disadvantages, and the relevance it has had in the development of safe operations and mining,
particularly in Chilean/Peruvian contexts.

HYDRAULIC METHOD OF SAND DISPOSAL IN DAMS

Context and method description

The occurrence of serious and high impact accidents (Canada, Brazil), larger volumes to be stored in
dams, and dam/impoundment relations that demand an accelerated dam growth, impose higher
demands of safety in operations. In the Chilean context, law and authority controls have become
more demanding (DS248).
Besides, mineral exploitation of lower purity increases the processing throughput, the tailings to be
managed and the volume of sand to be disposed in the dam.

3
Under these conditions, the hydraulic method is the most efficient and less costly. It uses the
centralized classification outside the dam and the discharge by means of pipes and spigot from the
consolidated capping, becoming the most adequate methodology for the new requirements.
The hydraulic method considers the use of sands (thick tailings fraction) as a confining and resisting
structure in most deposits. Sand production is carried out by separating the thick fraction (+ #200
Tyler) of fresh tailings by means of cyclones.
The sand dam construction itself can be done through mechanized or hydraulic manner, being the
latter the method that limits the use of machinery and minimizes costs (CAPEXs and OPEXs).
They cycloned sand hydraulic disposal method has become the most preferred technology in Chile
and Peru, where its relevance has grown, and it has consolidated as a safe and inexpensive
technology.

Advantages and limitations

Among the advantages of the hydraulic method, the following stand out:
 Maximizes the rate of sand placement.
 Minimizes operations idle time.
 Increases the deposit efficient area.
 Eliminates risk of down water contamination.
 Minimizes costs (<US$0.4/m3, profiling and compaction)
The main limitations of the hydraulic method are shown below:
 Insufficiency of deposit area (fields) in the initial dam at start up.
 High slope at the ravine or start up dam (-50to-35%).
 Insufficient or incomplete “priming” drainage.
 Tendency to the mechanization of the method (cut – loading – transportation – compacting).
 Lack of experience, especially at the initial period.

Tailings classification for dams

Continuous operation, high flow of tailings to be classified and frequent changes of sand placement
areas, make it essential a centralized classification and transport of the sands to the wall by
pressurized pipes. Conventional tailings cyclones are used with sand/slime proportion, at a range of
35/65 to 50/50 and solids at a concentration range of 30-36%. In case of excess of fine sands to be
deposited, a two-stage cyclone process or Recyclon® is considered. In order to minimize the use of
water, the Ciclowash® type is considered. In this case, the tailings can be fed in concentrations of a
45–50% range.
The maximum content of fine sands depends on the specific characteristics of the sand available.
Parameters such as permeability, cut resistance and cohesion must be considered.

4
Table 1 Fine particle content at sands in different dams

solids % < 74 mm
Dam
(maximum)
Mauro MLP Chile 19%
Ove je ría CODELCO Chile 12%
Las Tórtolas Anglo Ame rican Chile 14%
Linga SMCV Pe rú 14%

Figure 2 Sands deposited hydraulically in the dam

Sand transportation/pumping

The sand is usually transported up to the deposit area on dam by means of pressurized pipelines. A
head loss (j%) above 4% was observed for 9”, 10” and 14” pipes, for solid concentration above 70%
and flow velocity over 2.1 m/s.
Depending on the available charge (slope < 4%) pumping may be required for transportation:
 For shear stress below 120 Pa, horizontal centrifugal pumps are normally considered, in
series if required (Ovejería, Quillayes, Linga).
 For higher TDH and for sand slurry shear stress over 120 Pa, BDPs are used in parallel for
higher flows, (Ovejería, El Torito).
If slurry density is lower than 2 t/m3, for operating pressure below 70 fluid m, HDPE pipes are used
(Ovejería, Quillayes); for higher pressure, steel lines are used (Quillayes, Mauro, Linga).
If energy dissipation in the pipelines is necessary, reducer rings are used (Quillayes, Mauro).
Next, the rheology associated to Tranque Ovejería – CODELCO Andina Chile sand slurry is shown:
 Slurry concentration to dam 72 % a 74%
 Solids specific gravity 2.77 t/m3
 D50 200 mm

5
Head loss at pressurize pipes

Next, the head loss estimate for sand slurry in Chilean dams is shown (Ovejería Codelco Chile and
Quillayes /MLP)
For head loss estimate, pumping tests are performed (BDP strokes-volume flow, manometers and
hydraulic profile, friction loss). The selection of the calculation model for frictional loss will depend
on the rheologic characteristics of the slurry: particle size, viscosity, shear stress, solids concentration,
specific gravity.

Table 2 Rheological characteristics of Tranque Ovejería – CODELCO Andina Chile sand slurry.

Solid weight % pH Viscocity µ [cP] Shear Stress [Pa]

58,9 8,6 13,4 0,8

64,6 8,6 46,2 0,6
68,6 9,1 221,0 1,5
69,2 8,7 199,0 0,6
74,3 8,8 931,0 20,1

Figure 3 shows the adjustment for four models of frictional loss estimates (calculation of friction
coefficient), two of laminar regime and two of turbulent regime: Laminar Newtonian (Laminar),
Buckingham (Laminar), Colebrook (Turbulent) and Slatter (Turbulent)

Figure 3 Adjustment for model’s loss estimates Tranque Ovejería

Models were calibrated by adjusting dynamic viscosity. Laminar Newtonian model with the lowest
τf measured is better adjusted. Mean square error (MSE) was evaluated. Table 3 and Figure 4 show
these results.
A lower MSE was observed for the calibration of laminar regime models.

6
Main finding:
 Hydraulic transportation of sands at high concentration (over 70% of solids) fits better with
the laminar regime flow model.
 Tailings sands behave like Newtonian fluids, except for very concentrate special cases (over
74 %solid weight).

Table 3 Mean Square Error (MSE) for different models of frictional loss Tranque Ovejería.

Frictional Loss Model MSE

Laminar Newtonian 0,3

Buckingham 0,3
Slatter 0,5
Colebrook 0,5

Considerations-limitations:
 Analysis performed for a selected range of concentrations (72%-74%)
 Findings do not address the segregation of solids inside the pipeline.
 There is conviction that the formation of a bottom mobile bed and possibly temporary, may
produce the reduction of the flow area. A supplementary modeling would be convenient for
better accuracy.
These results are relevant when defining an adequate design, as they provide a safe operation and
reduction of works and additional costs.
6.0

5.5

5.0
J modelado (%)

4.5

4.0

3.5

3.0
3.0 3.5 4.0 4.5 5.0 5.5 6.0

J medido (%)
J Laminar Newtoniano J Buckingham
J Slatter J Colebrook

Figure 4 Calibration of frictional charge loss (J) for turbulent and frictional models. Tranque Ovejería

7
Sand disposal: practices in Chile
All findings and criteria proposed below are based on and are the exclusive reference of records and
processing of empirical data obtained from the operation in the indicated dams.
 Sand is placed in the deposit by means of pipes provided with multiple discharges (spigots)
separated between 2.5 m and 3.5 m which operate with pressure.
 Essential for dams of great velocity of growth (> 2 m per month).
 Each discharge (spigot) can be provided with a wear piece made of rubber or PU with steel
core.
 Spigots can be made of steel lined in rubber or HDPE.
The initial period of sand disposal is characterized by:
 Considering shapes of sand confinement at the ravine over the drainage system.
 Arriving with direct sand discharge lines to the confinement area.
 It is essential that initial sands meet with high and strict selection standards.
Hydraulics of deposition of sands during startup of dam construction is shown at Figure 5.
Considering charge loss of 5% in 1000 m, a range of 50 m can be lost due to friction, then:
 Boxes for gravitational operation can be considered (in pressurize flow) up to discharge point
at 60-70 m below, up to 1000 m of distance (experience in Quillayes) and also a centrifugal
pump that contributes with those 60-70 m of charge.
 For large charges available (>150 to 200 m) and distances < 1000 m, pressurized lines should
be considered, if necessary, provide them with dissipation rings or transfer boxes,
pressurized for distribution.
Mud area
(cracks)

GROWTH HYDRAULICS

1
Natural slope

Stable Growth

Start Dam
2
0

Unstable Growth

Figure 5 Sands deposit diagram. Dam evolution is observed

 Capacity of sands placement (Quillayes MLP 2001-2008, Ovejería 1998-2003 Chile)

o Throughput: 25 to 52 ktpd
o Operating fields: 2 to 3
o Sand per field: 12 to 26 ktpd
o Compacted sand volume: 18000 – 50000 m3/day
o Field width: 200 m
o Discharge holes: 30 a 40 per each 100 m

8
 o Holes diameter: 0.75” to 1”
o Downstream displacement: 300-400 m
o Area required per field (cancha): 40000 m2.
o Initial drainage time: 24 hours
o Drainage, time month 4: 12 hours
o Drainage, time month 6: 4 to 5 hours
o Drainage, time 2th year: 0 to 5 hours

o Thickness of layers before compacting: 0.40 to ,0.60 m

o Compactor’s passes for 10 tons: 2 to 3
o Minimum density compacted: 95% standard Proctor
 In relation to the charge available for transportation and discharge of sands:
o Normally, sand transport systems to the dam are gravitational (at least most part
of the first years).
o Charges available for dams in mountains: > 170 m.
o Initial dam’s length < 600 m.
o Distance cyclones - dam < 500 m.

CONCLUSIONS
The use of cyclone tailings sands in the construction of conventional tailings dams, coming from
copper sulfide ore concentrating plants is a proven and successful experience for medium and large
mining.
The pressurized flow in pipes is normally the adequate and most economical method for transporting
the sand pulp from cyclones to the discharge area in the dam.
The hydraulic method of sand discharge in the dam considers the slope of natural runoff (25-35%) of
these, using the continuous cyclic sequence: discharge, drainage, compaction, certification, has been
suitable for high growth speed (5m / month, 1000000 m3/month).
The hydraulic method of sand disposal in dam construction is operationally simpler, faster and easier
to control compared to other methods (loading and transport, paddocks).

REFERENCES

Observatorio Latinoamericano de Conflictos Ambientales, febrero 2018. Tranques de relaves se

duplicarían en 20 años y 47% de los depósitos existentes enfrentan conflictos.
http://olca.cl/articulo/nota.php?id=107145.

Uranium Project, Tailing dam Safety. May 2019. (http://www.wise-uranium.org/mdaf.html).

Chronology of major tailings dam failures. Records of 1965 to May 2018 failures occurred in Chile.

9
Decreto Supremo 248, Reglamento para aprobación de proyectos de diseño, construcción operación y cierre
de los depósitos de relaves, Ministerio de Minería, Gobierno de Chile, diciembre 2007.

Vásquez, P. 2010, Muros de Arena en Tranques de Relaves/Disposición Hidráulica, VII Congreso de

concentraductos, mineroductos y relaveductos. Viña del Mar Chile 2010. hydraulic analysis and
estimation by Valencia, P. SHIMIN Ingeniería.

10
Tailings Dam Made Redundant using Solid Bowl
Centrifuge

Amol Chinchankar1, Sunny Li2 and Dario Oyanadel3

1. Alfa Laval, Sweden
2. Alfa Laval, Australia
3. Alfa Laval, Chile

ABSTRACT
A member of an Australian owned and operated group of private companies with interests in
mining, the Coal Handling and Preparation Plant (referred to as CHPP henceforth) is an open cut
mining operation that produces both thermal coal and high quality semi-soft coking coal for overseas
customers. Located in New South Wales, it is based in the heart of the Hunter Valley, one of
Australia’s largest coal mining regions.
The CHPP had a vision to address their limited tailings dam capacity and had recognised for some
time that a suitable dry tailings disposal method had to be found. The existing tailings dam used on
site was nearing capacity and there were no suitable alternative sites for a replacement dam.
Within this operational context, they set out to find the best solution for dewatering thickener tailings.
The CHPP questioned if there was a dewatering technology that could handle tailings containing
high levels of clays.
Four different technologies were trialled at pilot plant scale on-site for dewatering their thickener
tailings from their Coal Washing Plant. Eventually the Alfa Laval Solid Bowl Centrifuge was selected
based on the outcomes of the trials.
Three Alfa Laval 720mm Solid Bowl Centrifuges were installed at the CHPP to handle part of the
tailings flow. In 2018, two more new larger Solid Bowl Centrifuges were installed to handle the total
volume of thickener underflow. This new installation will remove the need for a new tailings dam
when the existing dam capacity runs out and can be combined with the existing 720 mm units in case
additional dewatering capacity is required.
This paper will outline the process results and decision making that drove the CHPP over a few years
to install Solid Bowl Centrifuges. These centrifuges are also known as Decanter Centrifuges.

1
 INTRODUCTION

A member of an Australian owned and operated group of private companies with interests in
mining, the Coal Handling and Preparation Plant (referred to as CHPP henceforth) is an open cut
mining operation that produces both thermal coal and high quality semi-soft coking coal for overseas
customers. Located in New South Wales, it is based in the heart of the Hunter Valley, one of
Australia’s largest coal mining regions.
The CHPP produces an estimated 2.58 M tonnes of raw coal and 1.49 M tonnes of saleable thermal
and semi-soft coking coal each year for use in the steel and power industries.
With the existing tailings dam nearing capacity, the CHPP decided to pursue a dry tailings
management alternative.

TAILINGS BACKGROUND

The CHPP have 29 different seams that contain hard-to-dewater fines, such as bentonite clay and
volcanic sediments.
For the life of the mine to date, the tailings from the CHPP have been deposited in several tailings
emplacement facilities or tailings dams. This involves pumping the material to the dam, allowing
natural settling of the solids to occur and recycling water back to the CHPP.
By using Solid Bowl Centrifuges to dewater tailings at the CHPP, 62% to 70% of the water from the
thickener underflow would be recoverable without it leaving the plant. It would contain the water
within CHPP processing circuit, reducing the cost of pumping tailings to the tailings dam and return
water back to the CHPP.
The CHPP aimed to co-dispose the dewatered solids produced by the SBC’s by placing it back in the
mine with the overburden as part of the rehabilitation of the site. Sometimes this is also mixed with
coarse reject material.
Typical samples of the Thickener Underflow material from the CHPP were taken to determine the
Particle Size Distribution of the tailings (refer Table 1 below).

Table 1 Typical PSD of thickener underflow at the CHPP (feed to SBC)

Size (microns) Mass (%ad)

less than 25 59.23

25 – 38 3.53

38 – 63 7.60

2
 63 – 125 11.23

125 – 250 11.53

250 – 500 5.73

500 – 710 0.50

710- 1000 0.27

Above 1000 0.37

LOOKING FOR A SOLUTION

The CHPP identified the following key evaluation criteria for dry disposal of coal tailings:
• Spadeable cake with approximately 60% dry solids to be co-disposed with coarse
reject material into the pit to assist with mine rehabilitation
• Clean centrate for re-use in the coal washing process
• Sound operating economy
Based on the above criteria, they set out to find the best solution for dewatering thickener tailings. It
was decided to evaluate various technologies by conducting a research program including:
• Laboratory experiments
• Pilot plant testing
• Preparing a comparative report
Along with the technical feasibility assessment, an internal commercial assessment of the various
options was conducted.

Four different technologies were considered or trialled on-site for dewatering the thickener tailings.
A summary of their evaluation is given below.
Geo-bag technology: The CHPP did not find this technology too appealing as they wanted to produce
a product that could be trucked away for depositing it in the mine pit.
Belt Press Filter: This device was being used by some of the coal washing plants but the CHPP
wanted to explore if they could achieve better results. They realized that it was difficult to dewater
coal tailings that had high content of fine clay and also that the device consumed high polymer
dosage.
Recessed Plate-and-Frame Filter: The CHPP did a lot of work with this technology and were almost
convinced that they were going to use it in their plant. However, the test work made them realize
that it was difficult to achieve consistent results with this technology and moreover it had a high
CAPEX.

3
 Solid Bowl Centrifuges: Initially the CHPP thought of using the Solid Bowl Centrifuge for pre-
treatment of the thickener underflow before feeding it to a Plate-and-Frame filter to try and dewater
it further. They found that the SBC could make a spadeable solid product along with a clean centrate
that could be recycled.

After careful evaluation the CHPP decided to go ahead with Solid Bowl Centrifuge technology owing
to following reasons:
• Reduction in tailings dam capacity requirement (if there is an existing tailings dam that
the mine wishes to continue using)
• Avoiding the necessity of having a tailings dam
• Reduced capital investment compared to above-referred mechanical dewatering
technologies
• Dry solids in excess of 60% (by mass) contribute to open cut rehabilitation
• Water savings and water re-use in the plant
• Continuous, enclosed and clean operation
• Smaller footprint compared to other technologies
• Reduced operator intervention compared to other mechanical dewatering technologies
• Life-cycle cost for the dewatering system using SBC expected to be 12-18% lower

SOLID BOWL CENTRIFUGES

Solid bowl centrifuges have for many years been known for their separation efficiency based on the
enhancement of sedimentation in a high centrifugal field of 1000-3000 times the earth’s gravity.
Sedimentation velocities increase with a similar factor and the footprint or area needed for a given
sedimentation duty is greatly reduced. Dewatering of tailings will in most mining applications
involve both sedimentation and dewatering. In recent years, we have seen an increased focus on
water recovery and a push for dry stacking instead of tailings ponds. This has increased the demand
for good dewatering, while maintaining high capacity and clean liquid. The original solid bowl
centrifuges used in the mining industry during the 1970s and 1980s relied on drainage of liquid from
the solid particles to generate a high concentration of the discharged solids. Screen bowl centrifuges,
where the solids are conveyed across a section with a perforated outer rotor wall were also
introduced to improve the drainage capacity to reach a high solids concentration. Drainage is
relatively slow process when the solid fraction contains significant amounts of fine particles. In many
of these applications other technologies including plate and frame filter presses have been the
preferred solution for a number of years presumably due to the low dewatering capacity of the
drainage process in the solid bowl centrifuges. In recent years, a new design of the solid bowl
centrifuge coupled with a change of operating mode, where solids are compacted below the liquid
surface in the decanter, have shown an impressive increase in solids handling capacity to the extent
that solid bowl centrifuges are once again becoming an attractive alternative to the other dewatering
solutions. (1)

4
 Figure 1 Schematic view of a Solid Bowl Centrifuge

The horizontal centrifuge rotor (bowl) with red cross-section is driven from the main motor on the
right side to provide fast rotation where the feed suspension is accelerated to form a liquid ring inside
the rotor after being pumped into the centrifuge through the stationary feed tube. The particles in the
suspension will settle to the inside of the rotor while the liquid flows through the centrifuge and
leaves by overflowing weir plates at end wall to the left. The decanter centrifuge has an internal
conveyor (inner rotor with a helical flight) which is connected to the outer rotor through a high torque
gear box (shown in orange). The central shaft of the gearbox is connected to the motor to the left,
whereby it is possible to establish and control a differential speed between the conveyor and the
rotor. The differential speed of the conveyor will transport the separated solids towards the conical
end, where it leaves the centrifuge at a radial position unreachable for the liquid. (1)
For a deeper understanding of historic development of the Solid Bowl Centrifuge technology, please
see (2).

SOLIDS CONSOLIDATION/DEWATERING

In the original configuration of decanter centrifuges the solids (cake) were dewatered when the
conveyor transported the solids out of the liquid pond on it’s way along the conical section towards
the solids discharge openings and liquid could drain off and return to the pond. This situation is
shown in the upper half of figure 2 wherein the pond depth was very small (shallow pond) allowing
for a long drain zone. For large particles with high permeability the drainage is efficient, but capacity
will be reduced if we have fine particles in the cake.

5
 Figure 2 The upper half of the picture shows a shallow pond design used for mining applications. The lower
half shows a deep pond configuration where the solids are compacted below the liquid surface by the load
from the overburden of solids

For suspensions with a content of fine particles it is better to operate with a deeper pond as shown in
the lower half of figure 2. With a deeper pond the load from the overburden of solids can be used to
compact or consolidate the solids before it is conveyed into the conical section and we no longer need
the drainage to achieve the high solids concentration.
A further development of the deep pond operation is shown in figure 3, where a baffle disc has been
added to be able to operate with a liquid surface level radially inside of the solids discharge radius.
In this way, we can stack the solids in an even thicker layer for more radial compaction and in
addition we get a hydraulic pressure supporting the mechanical conveying of the solids towards the
solids discharge. This solution is particularly useful for cakes with a low permeability and this is also
the reason the liquid cannot pass through the cake and escape through the solids discharge openings.
(1).

Figure 3 A circular disc, known as baffle disc (shown in red), allows operation with very deep pond

ROBUST DESIGN

Alfa Laval has designed and developed the Solid Bowl Centrifuge for tough duties such as in Mining.
It has the most robust construction that is suitable for handling erosive solids. Highly wear resistant
material such as Tungsten Carbide are used as the material of construction in the feed zone, the solids

6
 discharge and the conveyor (shown below as a, b and c, respectively). In addition to that, the bowl is
protected from the inside by using Stellite strips.

(a) (b) (c)

FIRST SBC INSTALLATION AT THE CHPP

Three Alfa Laval 720mm Solid Bowl Centrifuges were installed at the CHPP to handle part of the
tailings flow.

7
 Table 2 Typical Mass Balance for the 720mm Solid Bowl Decanter

8
 Figures 4 & 5 Solids cake discharged on the conveyor from Solid Bowl Centrifuges at the CHPP

Performance of the three SBCs has been closely monitored over a number of years by the CHPP.
Table 3 below shows typical data. It will vary depending on which coal seam is being washed in the
CHPP.

Table 3 Typical results obtained at the CHPP

SG of the Thickener Underflow 1.1 to 1.32 (Average 1.22)

Particle size D50 less than 25 microns (see typical PSD under
‘Tailings Background’ section of this paper)

Mass flow of solids per 720 mm SBC 20 – 25 dry tons/hr

Solids content of the Thickener Underflow 25 – 40% w/w

Product (cake) Moisture 30 – 38% w/w

9
 Figure 6 Dewatering station for coal tailings at the CHPP including 3 off 720mm Solid Bowl Centrifuges
supplied by Alfa Laval

NEW INSTALLATION OF LARGE SOLID BOWL CENTRIFUGE

In 2018, the CHPP decided to dewater 100% of their coal tailings rather than sending a portion of the
wet tailings to the tailings dam. In order to achieve this, they decided to add two more new larger
sized Solid Bowl Centrifuges (SBC), each with a bowl diameter of 1000mm, to the existing SBC
installation. The plan was to get the new SBC’s into operation in Q4 2018 with the intention to remove
the need for a new Tailings Dam as the existing dam capacity runs out.

10
 Table 4 Typical Mass Balance for the 1000mm Solid Bowl Decanter

Figure 7 1000 mm Solid Bowl Centrifuge supplied by Alfa Laval

11
 Typical performance of the two new 1000 mm SBC’s is outlined in the table below. It should be noted
here that the performance will depend on the composition of the specific seam being processed
through the CHPP.

Table 5 Expected performance from the 1000 mm Solid Bowl Centrifuges at the CHPP

Performance parameter Range

Product (cake) moisture 32% to 38% (w/w)

Solids capture Above 95% (w/w)

Solids capacity (dry solids) 55 to 70 dry tonne/h

Feed Solids 25 to 40% (w/w)

Volumetric feed capacity 110 to 190 m3/hr

The two new 1000 mm Solid Bowl Centrifuges (model P3-10070) were commissioned in Q4 2018. The
results from the initial commissioning period are tabulated below under Table 6. It may be noted that
both the SBC’s have been handed over for taking into regular operation at the CHPP.

Table 6 Initial commissioning data with new 1000 mm SBC at the CHPP

24-10- 24-10- 25-10- 26-10- 26-10-

Date (dd/mm/yy) 18 18 18 18 18
Time (hh:mm) 10:13 13:27 10:59 10:30 13:36
MACHINE CONDITIONS
N (Bowl speed) rpm 1100 1100 1100 1100 1100
n (Conv. diff.) rpm 7,82 7,77 8,48 11,66 12,07
T (Conv. torq.) kNm 39,70 42,37 35,31 25,40 24,99
Control Mode T T T T T
FEED CONDITIONS
Feed Flow Rate m3/h 85,24 101,76 115,16 130,00 140,00
Solids Conc (oven drying) %w/w 32,10 26,76 25,86 27,58 31,37
Dry Solid Flow kg/h 33929 33222 35736 43017 54019
Density from Density Meter kg/m3 1240 1220 1200 1200 1230
POLYMER CONDITIONS
Floc Solution Flow m3/h 7,80 9,50 10,20 7,20 7,20
Floc Conc'n (w/o dilution) %w/w 0,25 0,25 0,25 0,25 0,25
CAKE CONDITIONS
Recovery % 98,10 97,57 98,64 98,58 98,66
Solid Concentration (oven
drying) %w/w 66,3 62,4 62,3 63,8 62,4

12
 Dry Solid Flow kg/h 33302 32438 35275 42425 53314
Cake Density kg/m3 1284 1263 1262 1270 1263
Cake Volume m3/h 39,13 41,17 44,85 52,36 67,63

CENTRATE CONDITIONS
Solid Concentration %w/w 1,021 0,989 0,531 0,631 0,769
Solid Concentration ppm 10210 9890 5310 6310 7690
Solid Flow kg/h 645,96 807,65 487,35 610,19 722,88
Centrate Flow ton/h 63,27 81,66 91,78 96,70 94,00
Centrate Density kg/m3 1003 1003 1002 1002 1003
Centrate Flow m3/h 63,05 81,39 91,62 96,50 93,76

CONCLUSION
By utilizing the deep pond design coupled with robust materials of construction, solid bowl
centrifuges can be utilized to deliver high cake concentration, even at high flow rates. A better
understanding of scaling to high capacities will allow large diameter SBC’s to be used for high
capacity duties in mine tailings applications. While applying this technology it could be helpful to
carry out field tests using pilot-scale SBC’s as the tailings characteristics including mineralogy,
particle size distribution, etc, might affect the dewatering performance from one site to another.
It is expected that Solid Bowl Centrifuges with their low hold-up volume, small foot-print, automated
operation, reduced operator intervention as well as lower CAPEX and OPEX as compared to some
other mechanical dewatering technologies will become an attractive choice for dewatering duties in
mining applications in the future.

ACKNOWLEDGEMENTS
The authors of this paper acknowledge the assistance and co-operation of the CHPP and it’s
management in preparing this paper.

REFERENCES

1. Madsen B. (2017): ‘Re-introducing solid bowl decanter centrifuges for dewatering in the
mining industry’. (Conference paper presented at SYMPHOS 2017, 4th International
Symposium on Innovation and Technology in the Phosphate Industry)

2. Records A., Sutherland K. (2001): Decanter Centrifuge Handbook. 1st edition, Elsevier, NY

13
5
Technologies and
Instrumentation
for Monitoring and
Surveillance
Operational InSAR Monitoring for Tailings Dams
Ciscu Sánchez1, Sara Del Conte2 and Miguel Ángel Verdugo1
1. TRE-ALTAMIRA, Chile
2. TRE-ALTAMIRA, Canada

ABSTRACT

Recent catastrophic failures have highlighted that continuous monitoring of surface deformation is
essential to ensure the safety of tailings dams and storage mine waste. As part of a comprehensive
risk assessment strategy, information should be collected not only from the dam itself, but from the
surrounding areas as well, in order to ensure any potential hazards to dam integrity are identified
before they can impact worker/environmental safety and mine operations.
Satellite SAR interferometry (InSAR) is a valuable operational tool for monitoring the integrity and
surface deformation of tailings storage facilities at mine sites worldwide. InSAR is a remote sensing
technique which uses radar satellite imagery to measure millimetric level movement occurring over
time. Radar systems are active; they collect information from natural reflections of the signal bounces
on the ground without the need to install any equipment. Furthermore, single satellite images extend
thousands of square kilometres in size, meaning they can provide information not only on the
stability of tailings infrastructure, but also the entire mine site and possibly entire regions. The launch
of high resolution and short revisit time satellites has vastly improved data collection capacities.
Furthermore, the development of processing techniques and tools which support near real time
delivery have made it possible to use InSAR as an operational tool.
The benefits of InSAR monitoring for mine sites and tailings are discussed in this paper, with focus
on recent improvements of satellites and processing techniques geared towards operational needs.

1
INTRODUCTION

Despite the many advances made in the mining sector and increased geotechnical engineering
knowledge, tailings dam failures still occur (Roche et al, 2017). The overall number of annual tailings
dam failures has been in decline in the last years, but the number of serious failures has increased
(Bowker & Chambers 2015), causing extensive environment damages and loss of life.
Tailing facilities or tailing dams are designed and built slowly in stages over many years where
conditions may also change with time, causing unexpected failures with high social, environmental
and economic loss (ICOLD 2001, Chambers & Bowker 2017, Roche et al 2017).
In many cases, dam failures are preceded by a series of surface deformation anomalies, such as
tension cracks and accelerating trends (Vanden Berghe et al., 2011). If detected in their initial phase
and well interpreted, these anomalies could support warning and alarm systems in the activation of
risk mitigation procedures. Monitoring surface displacement is fundamental in the prevention phase
as it highlights areas that need high priority investigations.
The objective of monitoring the displacements is to follow the dam reaction to the continuous loading
(Vanden Berghe et al., 2011) and to monitor the deformation in closed mine environments to preserve
the integrity of the facility and avoid future failures. Jointly with other geotechnical monitoring
instrumentation; such piezometers, inclinometers, levelling, GPS and other types sensors, InSAR
measures surface displacement. InSAR is one of the key players in detecting surface deformation and
acceleration, which may indicate an imminent risk of failure.
Over the last decade InSAR data has been successfully applied over number of mine sites (Carnec &
Delacourt 2000, Colesanti et al. 2005, Jung et al. 2007, Herrera et al. 2007, Herrera et al. 2010, Iannacone
et al. 2014, Colombo & MacDonald 2015, Sanchez et al 2016, Falorni et. al, 2015, Athayde, et. al, 2017,
Velasco et. al, 2018), and has become one of the most suitable technologies to measure surface
deformation. Beneficial features of InSAR surface monitoring for mine sites include:
 The provision of information without the need to install ground instrumentation (no need to
access remote or hazardous sites);
 The wide coverage and high density of information, not achievable with in-situ
instrumentation;
 The capability to detect both slow and fast movement (from millimetres to metres) by
integrating different InSAR techniques and satellites.
Recent advances in the satellite space segment and processing algorithms have significantly reduced
computational time. Whereas the advent of newer satellites, with increased spatial resolution and
acquisition frequency have increased information density. Near-real time InSAR monitoring is now
widely applied in different mining applications to highlight possible incipient movements in areas
not visible to in-situ instrumentation.

2
INSAR OVERVIEW

Synthetic Aperture Radar (SAR) satellites acquire images of the Earth’s surface by emitting
electromagnetic waves and analysing the reflected signal. Interferometric Synthetic Aperture Radar
(InSAR), also referred to as SAR Interferometry, consists of the phase comparison of SAR images,
acquired at different times with similar looking angles from space or airborne platforms (Gabriel et
al., 1989; Massonnet and Feigl,1998; Rosen et al., 2000; Bamler and Hartl, 1998).
The phase difference between two SAR images is proportional to the target motion occurring along
the sensor-target line-of-sight (LOS) direction (Figure 1) during that time interval. As SAR satellites
are continuously circumnavigating the globe, a number of SAR images can be collected for the same
area over time and information about the deformation of the earth’s surface can be extracted.

Figure 1: An illustration showing the relationship between ground displacement and signal phase shift. This is
the basic principle of InSAR for measuring ground movement

Advanced DInSAR (A-DInSAR) techniques are based on the data processing from multiple image
acquisitions to form time series of deformation. Permanent Scatterer Interferometry, the first A-
DInSAR technique, identifies and monitors point-wise permanent scatterers (PS), pixels that display
stable amplitude and coherent phase throughout every image of the dataset (Ferretti et al. 2000, 2001).
Permanent scatterers are related to natural radar targets such as rocky outcrops, un-vegetated Earth
surfaces, boulders, manmade structures (buildings, street lights, transmission towers, etc.), and any
linear structure that can reflect a signal back to the satellite. As these types of targets are common in
non-urban areas, a new technique known as SqueeSAR® was presented by Ferretti et al. in 2011 that
extracts information from distributed scatterers (DS). This extends measurement point coverage to
areas with limited infrastructure and that are lightly vegetated.
The existence of low-resolution SAR data archives going back to the 1990s initially led to the extensive
use of InSAR data to perform historical ground deformation analyses. More recently, the use of high-
resolution sensors has considerably increased the density of detectable information, up to thousands

3
of measurement points per square kilometre (Figure 2). In conjunction with this, the launch of
satellites with a high frequency of acquisitions (up to a few days), combined with the development
of sophisticated automatic processing algorithms made it possible to continuously provide reliable
surface deformation measurements with each new satellite image.

Figure 2 Example of SqueeSAR results obtained from Sentinel (low resolution) and TerraSAR-X (high
resolution) radar imagery over the same site, by processing around 50 images acquired within 2 years

INSAR IN MINING APPLICATIONS

Active mines have large range of ground movement rates affecting different assets. Monitoring plans
have to be defined in order to maximise the information extracted from the satellite imagery, to be
flexible, and to be rapid in case of emergencies. The availability of SAR data with different resolution
and acquisition frequencies, as well as, the combination of near real-time and long-term stability
monitoring solutions, makes it possible to monitor all ranges of movement, while continuously
providing reliable surface deformation measurements over the entire mine site.
Two main InSAR monitoring tools are prevalent within the mining industry, they can be integrated
in a monitoring plan in combination with other monitoring techniques, they are:
 InSAR Bulletins for a rapid delivery of surface information at every new satellite acquisition.

4
  Periodic and/or continuous Advanced InSAR analyses to provide deformation time series for
a detailed understanding of mm-scale up to metric-scale surface deformation.

InSAR bulletins (Figure 3) are generated within hours of every new satellite acquisition. Using the
two most up to date images, bulletins provide operational information about movement occurred in
the last few days. Bulletins provide an overview of the mine site to highlight early detection of
changes in ground movement, and fast slope movements.

Figure 3 Sample bulletin over tailings facilities

Advanced InSAR analysis can be performed on a continuous monitoring basis or with periodic
updates (annual, semi-annual, quarterly, monthly). By analysing a collection of images, these
techniques can identify natural targets on the ground and can detect millimetric deformation over
wide areas by provide accurate time series of deformation. An example over a tailings dam is shown
in ¡Error! No se encuentra el origen de la referencia., along with a deformation time series of an
individual point and a surface profile section generated along the crest of the dam.

5
 Figure 4 Example of SqueeSAR results over a tailings dam with a deformation time series of an individual
point (top graph) and a surface deformation profile across the A-A’ cross-section (bottom graph)

In case of rapid movement, advanced InSAR analyses are integrated with speckle tracking
techniques, capable of tracking movement to meter per day, by exploiting the amplitude component
of the SAR image. When applied to a set of images speckle tracking can also provide deformation
time series with centimetric accuracy. An example of SqueeSAR® results and speckle tracking results
combined coverage is shown in Figure 5.
Operational monitoring analysis in active tailings dams requires frequent updates of surface
deformation data. The assessment of the embankments stability, as well as, the timely identification
of changes and accelerations in the deformation rates during operational and loading activities is
essential to ensure the safety and continuity of mining operations. Advanced InSAR analysis are
particularly suited to detect precursor slow accelerations over tailings facilities. In particular,
deformation time series produced from long-term monitoring can be investigated with inverse of
velocity methods (Fukuzono in 1985) to predict the timing of the failure weeks/months in advance
before a failure (Figure 6).

6
 Figure 5 Overview of SqueeSAR results and advanced speckle tracking results combined coverage over a
mine site and sample of deformation time series

Figure 6 Time series acceleration detection and inverse velocity collapse forecasting, using the displacement information of
one measurement point

7
CONCLUSION

There are many factors that could compromise tailing embankments stability, and most of them are
detectable as clear signs of surface displacements. InSAR is as a widely proven technique used to
detect surface displacements in near real time. The recent launch of high-resolution and high
frequency satellites has promoted the improvement of processing algorithm and the development of
new InSAR products capable of supporting operational schedules.
The wide availability of SAR data with different resolution and acquisition frequency, and the
combination of near real-time and long-term stability monitoring solutions makes it possible to
monitor all ranges of movement affecting the mine site. With it, continuous reliable surface
deformation measurements over the tailings facilities can be provided. Given the high accuracy of
the measurements provided, and the timely identification of changes in trends, InSAR can be used
as a powerful forecaster of imminent dam failures.

REFERENCES
Adam, N, Parizzi, A & Crosetto, M 2009, ‘Practical Persistent Scatterer Processing Validation in the
Course of the Terrafirma Project’, Journal of Applied Geophysics, vol. 69, pp. 59–65.

Athayde, C. Sánchez, C. Carneiro, S. De Matos, B. (2017) ‘O uso da tecnología InSAR (Interferometría

de radar de abertura sintética) para o monitoramento da estabilidade de barragens.’ XXXI – Seminário
nacional de grandes barragens – SNGB del Comité Brasileiro de Barragens. Belo Horizonte – MG, 15 a 18
de maio de 2017.

Bowker, L.N. and Chambers, D.M. (2017). In the Dark Shadow of the Supercycle Tailings Failure Risk
& Public Liability Reach All Time Highs. Environments 4:75. doi 10.3390/environments4040075.

Carnec, C., Delacourt, C. 2000. Three years of mining subsidence monitored by SAR interferometry,
near Gardanne, France. Journal of Applied Geophysics, 43, 43-54.

Chambers, D.M., Bowker, L.N. (2017) Tailings Dam Failures 1915 – 2017. Available online:
http://www.csp2.org/tsf-failures-1915-2017 (Accessed on 16 August 2017).

Colesanti, C., Mouelic, S. L., Bennani, M., Raucoules, D., Carnec, C., Ferretti, A. 2005. Detection of
mining related ground instabilities using the Permanent Scatterers technique: a case study in the east
of France. International Journal of Remote Sensing, 26 (1): 201-207.

Colombo D & MacDonald B. 2015. Using advanced InSAR techniques as a remote tool for mine site
monitoring. Proceedings, Slope Stability Congress, Cape Town, South Africa.

Costantini, M., Falco, S., Malvarosa, F., Minati, F., 2008. A new method for identification and analysis
of persistent scatterers in series of SAR images. In: Proceedings of IGARSS 2008, Boston.

8
Ferretti, A., Prati, C., and Rocca, F. 2000, Nonlinear subsidence rate estimations using permanent
scatteres in differential sar interferometry. IEEE Transactions on Geoscience and Remote Sensing,
38(5):2202–2212

Ferretti, A., Prati, C., and Rocca, F. (2001). Permanent scatterers in SAR interferometry. IEEE
Transactions on Geoscience and Remote Sensing, vol. 39, no. 1. pp. 8 -20.

Fukuzono T. (1985). A new method for predicting the failure time of a slope. In: proceedings of the
fourth international conference and field workshop on landslides. Tokyo: Japan Landslide Society.
(p. 145-50).

Gabriel, A.K, Goldstein, R.M. and Zebker, H.A. (1989) Mapping small elevation changes over large
areas: differential radar interferometry. Journal of Geophysical Research, 94, 9183–9191

Herrera, G., Tomás, R., Lopez-Sanchez, J. M., Delgado, J., Mallorqui, J. J., Duque, S., & Mulas, J. 2007.
Advanced DInSAR analysis on mining areas: La Union case study (Murcia, SE Spain). Engineering
Geology, 90, 148-159.

Herrera, G., Tomas, R., Vicente, F., Lopez-Sanches, J.M., Mallorquí, J.J., Mulas, J. 2010. Mapping
ground movements in open pit mining areas using differential SAR interferometry. International
Journal of Rock Mechanics & Mining Sciences, 47 (2010): 1114-1125

Iannacone, J.P., Corsini, A., Berti, M., Morgan, J., & Falorni, G. 2015. Characterization of Longwall
Mining Induced Subsidence by Means of Automated Analysis of InSAR Time-Series. In Engineering
Geology for Society and Territory - Volume 5, International Publishing

ICOLD (2001). Tailings Dams Risk of Dangerous Occurrences: Lessons Learnt from Practical
Experiences. Bulletin 121 (pp. 145). Paris, France: International Commission on Large Dams (ICOLD).

Jung, H.C., Kim, S. W., Jung, H. S., Min, K.D., Won, J.S. 2007. Satellite observation of coal mining
subsidence by persistent scatterer analysis. Engineering Geology, 92 (2007): 1-13.

Macdonald, B. C., Sharon, R., Muir, S., Iannacone, J. P., Falorni, G., & Michaud, J. S. (2016, September).
InSAR derived synoptic ground surface deformation map of the Porgera Mine, Papua New Guinea.
In Proceedings of the first Asia Pacific Slope Stability in Mining Conference APSSIM (pp. 6-8).

Massonnet, D., & Feigl, K. L. (1998). Radar interferometry and its application to changes in the Earth's
surface. Reviews of geophysics, 36(4), 441-500.

Roche, C., Thygesen, K., Baker, E. (Eds.) 2017. Mine Tailings Storage: Safety Is No Accident. A UNEP
Rapid Response Assessment. United Nations Environment Programme and GRID-Arendal, Nairobi
and Arendal, www.grida.no ISBN: 978-82-7701-170-7

Sánchez F., Conde A. and Salvá B., Colombo D. (2016) Use of SAR radar satellite data to measure
ground deformation in underground and open pit mine sites, El Teniente case study, Chile. The First
Asia Pacific Slope Stability in Mining (APSSIM) Conference 2016, At Brisbane, Australia 6-8
September 2016.

9
Vanden Berghe, J-F., Ballard, J-C., Wintgens, J-F. and List, B. (2011). Geotechnical risks related to
tailings dam operations. In Proceedings Tailings and Mine Waste Conference, 2011. Vancouver, BC,
University of Britsh Columbia Library.

Velasco, V. Sanchez, C. Couso, C. and García, M. 2018. InSAR technology for salt mines monitoring.
Proceedings of the SaltMech IX conference, Hannover, Germany 2018.

Werner, C., Wegmuller, U., Strozzi, T., & Wiesmann, A. (2003, July). Interferometric point target
analysis for deformation mapping. In Geoscience and Remote Sensing Symposium, 2003. IGARSS'03.
Proceedings. 2003 IEEE International (Vol. 7, pp. 4362-4364). IEEE.

10
Next Step of Remote Sensing Tools for Tailings Dam
Monitoring: the Importance of Predicting and
Reporting
Javier Duro, Rubén Iglesias, Fernando Vicente, Dani Monells, Giuseppe Centolanza and
Rodrigo Godoy
Dares Technology, Spain

ABSTRACT
Tailings dams’ failures constitute a risk for safety and operations losses in mining areas. The inability
of detecting precursors on time may lead to operational shut-downs with economic losses, accidents
with human fatalities, and or environmental damages of different magnitude.
Nowadays, mines are often facing an incomplete surveying, relying on expensive in-situ sensors. In
this framework, radar sensors, mounted on space-borne platforms, represent a complementary
alternative. They can measure a wide range of deformations over time with millimetric precision and
independently of day-night cycles and weather conditions.
DARES TECHNOLOGY exploits radar interferometry techniques combined with optical data and in-
situ sensors to provide useful information about slope instabilities over pit areas, settlement of waste
dumps, deformations in tailings dams and subsidence in infrastructures areas.
The recent development of change trend indicators of InSAR deformation time-series, such as the
inverse velocity and/or the transects analysis, which contain spatio-temporal information about the
deformation process, are improving the usability of InSAR-derived products. Furthermore, the
combination of these indicators with in-situ information allows the rapid identification of
discontinuities over the retaining walls before a crack appears to avoid collapses or leakages.
Finally, the provisioning of interactive web visualization tools is becoming a must to manage and
specially understand this huge amount of information. This service highly improves conventional
reporting, allowing end-users to easily zoom in and out over detailed optical images as background
to precisely identify where deformation is happening and make further analysis comparing this
information with in-situ information, computing prediction indicators and/or generating transects
over affected areas.

1
INTRODUCTION

Data acquired from Synthetic Aperture Radar (SAR) sensors allow obtaining all-day, all-weather
high-resolution reflectivity images of large-scale areas. Using SAR images taken at different times,
SAR Interferometry (InSAR) techniques exploits the phase differences to retrieve the displacement
information of affected areas with millimetric precision (Massonet et al. 1998).
In this framework, InSAR analysis allow the measurement of surface deformation for a wide range
of ground displacement types (slopes, tailings, dewatering, injection heave, etc.) and at different
scales (faults, slope areas, entire mine, etc.). The processing of SAR data can provide measurements
with high degree of spatial detail than can be integrated with other in-situ information sources. In
this way it is possible to get a unified global view of ground deformation effects and better
understand their causes. This is crucial to control aspects of safety in operations in different sectors,
such as mining, infrastructures or oil&gas. During the last decade open-pit and underground mining
are increasingly using data collected by SAR sensors. Geotechnical managers are taking advantage
of this data to map precursors over entire mining areas to prevent accidents and to take decisions
about excavations and/or tailings dam’s constructions.
Traditionally, the main detractors of InSAR products, complain about the difficult to deal with large
amounts of data and, therefore, to detect areas prone to have geotechnical problems, about the lack
of a frequent update of information and, finally, about the difficulties to understand SAR geometry.
In this paper, advanced InSAR tools are presented in order to demonstrate its utility in the monitoring
of tailings dams compared with traditional results. Finally, the integration of these tools with in situ-
information will be put forward.

FROM STANDARD INSAR RESULTS TO THE FORECASTING OF AREAS PRONE

TO SUFFER STRUCTURAL PROBLEMS

Conventional Mid-Term InSAR Results

As presented in the introduction, differential phase between different satellite acquisitions is

sensitive to ground deformation. Unfortunately, not all the points within the interferograms have the
enough phase quality to provide reliable information. Furthermore, interferometric phase of
differential interferograms is compromised by inaccuracies in the Digital Elevation Model (DEM)
employed during the compensation of the topographic component and by the impact of atmospheric
phase screen (APS). In order to compensate the topographic height errors and separate atmosphere
from displacement, multi-temporal stacks of SAR data are exploited using Persistent Scatterers
Interferometry (PSI) techniques (Ferretti et al. 2001, Mora et al. 2002, Bernardino et al. 2002).
DARES TECHNOLOGY employs the Coherent Pixels Technique (CPT), developed since 2002 by the
Universitat Politècnica de Catalunya (Mora et al. 2002). This PSI technique is based in three main
blocks. The first block corresponds to the selection of reliable measurement points to be included in

2
the latter PSI processing. For this purpose, the quality estimator, called coherence, is employed.
Quality values vary in the range [0, 1], with the two extremes accounting for the two limit situations
of data totally uncorrelated and fully correlated, respectively. A coherence threshold is fixed in order
to identify the reliable points for the following steps. Once a set of reliable points are identified, a
triangulation process is carried out in order to minimize APS and remove the offset between
interferograms. In order to estimate the residual topographic component, a linear model is defined
for each link of the triangulation. This model is then adjusted to the data through the minimization
of a cost function and, finally, the absolute values of topographic height error for each pixel are
obtained through an integration process using one reference point as seed. Once the topographic
error is removed from the interferograms, APS is compensated by exploiting the statistical differences
between this component and the displacement one. APS behaves smoothly in the spatial domain and
completely random along the temporal axis, while displacement is assumed to change smoothly
along time. Using a spatial low-pass filtering, jointly with a temporal filtering over the phase
unwrapped interferograms, both components can be adequately separated (Mora et al. 2002). At this
stage, displacement time-series of each measurement point are obtained.
Figure 1 shows PSI results over a mine in the Atacama Desert (Chile) using data collected by the
European Space Agency Sentinel-1 (S1) satellite in the descending acquisition orbit mode. In the
Figure, red points refer to displacements moving away from the radar sensor while blue ones indicate
a movement on the surface approaching it. On the right side, two interesting points showing a change
in its trend in the last period of observation, are highlighted.

Figure 1 Displacement rate results at a mine in the Atacama Desert (Chile) using data collected by the
European Space Agency (ESA) SENTINEL-1 (S1) satellite. Points A and B are located in the open pit and in a
waste dump showing a clear change in its trend during the last period of observation, concurrent with a
period of heavy rains in Chile.

3
Thanks to the global coverage of SAR satellites, whole mines can be monitored with millimetric
precision: slope stability over the pit areas, settlement of waste dumps, deformations in tailings dams
and subsidence in infrastructures areas.
Despite this, PSI results can be highly improved. On the one hand, mid-term reports are not enough
to prevent failure events such as the one that occurred in Cadia gold mine on 9TH March 2018 or
Brumadinho on 25TH January 2019. On the other hand, PSI standard products can be difficult to
manage and to understand, especially over areas with topographic changes, such as mining areas.

Advanced InSAR Products

SAR sensors are only sensitive to displacements produced along the line-of-sight (LOS) direction and,
therefore, the measured displacements are a projection of them. For the current polar-orbiting SAR
sensors, the look direction is either East or West, for the ascending or descending pass, respectively.
For this reason, SAR sensors are only sensitive to vertical displacements and/or horizontal ones along
the East or West direction. If the displacements obtained by both acquisition modes are combined,
the Up-Down (UD) and East-West (EW) displacement components can be obtained from the
incidence angle and the track heading of both geometries solving a system of linear equations
(Hanssen 2001). Once the UD and EW components of displacement are obtained, the magnitude of
the displacement is closer to the real one and more advanced products can be generated over the
areas of interest:
 Rapid Reports. The generation of rapid reports through the computation of classical InSAR
is firstly proposed. Instead of using PSI derived displacement time-series, a single differential
interferogram is updated each satellite pass in order to identify the increase in magnitude
and/or in extension of deformation areas. Indeed, the accuracy of this product is lower than
the one provided by mid-term PSI products, i.e, around one twentieth of the wavelength
(which corresponds 0.5 cm for sensors at C-band such as S1), but is extremely useful to
identify new areas of deformation.

 Acceleration indicators and cross-sections. Once new hot spots of deformation are identified
with rapid reports, PSI time-series can be analyzed. The problem at this point is that checking
the large number of points provided by InSAR products could be a difficult task, sometimes
impossible to fulfill. In first instance, the computation of the acceleration over the last
samples of the time-series is proposed in this paper in order to highlight the most vulnerable
areas. Moreover, further analysis can be done in order to predict a potential failure date. The
so-called inverse-velocity approach [Carlà et al. 2019] allows to forecast the date of failure,
under a purely cinematic point of view (without accounting for the geology or the
geotechnical features of the site). First the velocity should be computed as a derivative of the
displacement 𝑣 = 𝑑𝜌/𝑑𝑡, and then, a linear regression over the inverse velocity 1/𝑣 is
computed in order to find the intersection point with zero. The potential failure date can be
hence obtained through the coefficients of the linear regression (Carlà et al. 2019).

4
 Furthermore, the computation of cross-sections allow for spatio-temporal analysis of ground
deformation along specific spatial profiles. An increase in the separation between
consecutive graphs indicates an acceleration in the deformation process.

 Visualization and Integration Services. Finally, it is proposed the reporting of InSAR-derived

products with in-situ information. In this framework, webGis services represent an excellent
alternative to allow end-users easily zoom over detailed optical image, as background, and
precisely identify ground deformation areas and compare it with in-situ information. Finally,
the computation of prediction indicators or the generation of transects can be easily
incorporated on top of these services.

RESULTS AND DISCUSSION

In this Section some results are presented in order to show the improvements proposed in the paper.
In Figure 2 a rapid report corresponding to a tailings dam located in a Peru mining area is depicted.
The representation corresponds to the accumulated deformation between January 9TH and 21TH 2019
using ascending and descending orbit geometries. Two different deformation focuses, revealing a
deformation increase in the last 12 days of around 2 centimetres, are detected. The analysed period
coincides with heavy rains in the area. The different deformation detected in ascending and
descending modes indicates that a significant horizontal displacement towards East direction, which
can be appreciated in the EW component of the decomposition, is affecting the tailings dam. A strong
vertical component can be also appreciated. In the case of dam’s safety taking into account, not only
the magnitude of displacement, but also the direction is extremely important.

Figure 2 Rapid report corresponding to a tailings dam located in a Peru mining area. The represented isolines
corresponds to the accumulated deformation between January 9 TH and 21TH 2019 using ascending and
descending orbit geometries. Two different deformation focus revealing an increase in deformation of 2
centimetres in the last 12 days is detected. The UD and EW decomposition revels a strong horizontal
component towards the East of the tailings dam.

5
As it can be seen in the right image on Figure 2, InSAR is a unique and very powerful tool to map
vertical deformation or compaction on tailings dam. Indeed, this vertical deformation is very difficult
to measure by ground-based radars as they usually look from the front side being mainly sensitive
to horizontal displacements of the retaining walls. On the other hand, PRISM data have also a
reduced sensitivity on the up-down axis due to the lack of vertical baseline on the observations.
Therefore, InSAR is a useful tool to monitor the whole tailings dam area, with a complete coverage,
continuous sampling and high sensitivity to the ground settlement.
Indeed, the comparison of rapid reports along time allows to detect precursors in tailings dams for
guaranteeing safety and production efficiency on time, nonetheless, further analysis can be retrieved
from data to ease the understanding of the deformation dynamics. As described in the previous
Section, in order to include the temporal axis in the analysis of the areas of interest, the generation of
cross-sections are proposed.
In this context, a cross-section over the tailings dam located in the Peru mining area presented
previously, is depicted in Figure 3. The deformation dynamics during the last months (from
25/07/2018 to 21/01/2019) shows a clear acceleration in the last 12 days. Furthermore, the identification
of peaks in the cross-section can be associated to specific locations in the tailings dam areas prone to
suffer structural problems.

Figure 3 Cross-section over a tailings dam located in a Peru mining area corresponding to the period from
25/07/2018 to 21/01/2019. Two peaks with a clear acceleration in the last 12 days are clearly identified

At this stage, the areas presenting a high increment of deformation between the last acquisitions, and
its spatio-temporal behaviour have been deeply analysed. The next step is looking for time-series
presenting an acceleration in the last samples of the time-series and forecast potential failures. In
order to illustrate this, the back-analysis of the tailings dam failure of Cadia gold mine, Australia, on
9th March 2018, is presented.
Figure 4 shows how time-series revealed a precursor displacement of the tailings dam centred over
the region prior to the failure. In the Figure the accumulated displacement field measured before the
event (from January to March 2018) is superimposed over an optical image obtained with Sentinel-2.

6
Notice that the period of acceleration started about one-and-half months before the collapse. As
explained in the previous Section, an inverse velocity analysis can be carried out in order to forecast
the date of failure (see Figure 4). In that case the error was of about 4 days. This analysis can be
performed over all InSAR measurements showing accelerations in order to provide an extremely
useful early-warning system to help geotechnical responsibles in the decision-making process to
ensure the operational continuity.
At this stage, the combination of all InSAR-derived tools showed in this Section with other remote
sensing sources and in-situ information, clearly help in the proactive identification of discontinuities
over the retaining walls to avoid collapses or leakages. In that way, in-situ instrumentation provides
value information that can be exploited in an assimilation and integration scenario as well with
multiple benefits towards a transversal approach. For instance, piezometric measurements have
demonstrated to be an important asset in the previous scenario where the dynamics of water or fluids
induce surface motion in the operations area. The simultaneous and near-real time access to surface
dynamics and piezometric level variations allows to detect correlation between both deformation
and the cover dynamics with the water level variations. Derived from this correlation susceptibility
maps or heatmaps can be derived to emphasize the relevant areas that can be affected for future
events. Other sources of information, such as ground-based radar, PRISM or extensometers data can
be integrated with InSAR products.

Figure 4 Accumulated deformation and inverse velocity back-analysis over Cadia mine tailings dam failure

Finally, the reporting of InSAR services with in-situ information can be favoured from webGis
services. In this framework, DARES TECHNOLOGY has developed DARES MAPPER. This way of
presenting results highly improves conventional reporting, allowing end-users easily zoom in and
out over a detailed optical image as background in order to precisely identify where ground
deformation is happening and make further analysis, changing the temporal span for the
deformation represented, comparing this information with in-situ information, computing prediction
indicators or generating transects over affected areas. Having access simultaneously to all sources of
information unlocks a hidden understanding without one or the other source.

7
Figure 5 DARES MAPPER configuration showing the correlation between InSAR time-series and piezometers

Figure 5 shows a configuration of DARES MAPPER providing the accumulated vertical deformation
from 19/08/2018 to 15/03/2019 for an area corresponding to a tailings dam, and the time-series
evolution of the point selected in the visualizer (highlighted with white circle). As illustrated, the
accumulated deformation is represented jointly with the piezometric information for the same
temporal span. The simultaneous analysis of the temporal evolution of the piezometric level and the
deformation reveals a clear correlation pattern. Despite the lack of continuous piezometric
measurements (light blue line), it is clearly visible that both deformation trends (orange and dark
blue lines) behave in a similar way with a detectable delay.

CONCLUSIONS
The increasing focus on tailings dam safety brings with it an increasing awareness of the importance
of a reliable monitoring program to ensure the operational continuity of the tailings dam.
The advanced InSAR-derived products presented in the paper offers alert systems with
weekly/monthly frequency, providing timely information about potential dam failures, and enabling
the control and mitigation of dam events that can cause interruptions in mining operations. The
papers demonstrates how InSAR can provide the vertical and horizontal deformation components
over tailings dams, giving a clear figure of the geotechnical conditions every week. The use of cross-
sections allows for spatio-temporal analysis of ground deformation along specific spatial profiles that
ease the understanding of displacement dynamics. Additionally, time-series acceleration indicators,
such as inverse velocity, allow the prediction of failures or discontinuities over the retaining walls
before a crack appears. All this information can be visualized using webGis services, such as DARES
MAPPER, improving user experience and allowing the comparison with other remote sensing tools
and in-situ information.

8
REFERENCES

Berardino, P.; Fornaro, G.; Lanari, R.; Sansosti, E. A new algorithm for surface deformation
monitoring based on small baseline differential SAR interferograms. IEEE Trans. Geosci. Remote
Sens. 2002, 40, 2375–2383.

Carlà, T.; Farina, P.; Intrieri, E. ; Ketizmen, H. ; Casagli, N.. Integration of ground based radar and
satellite InSAR data for the analysis of an unexpected slope failure in an open-pit mine. Engineering
Geology. 235, 39-52, 2018.

Ferretti, A.; Prati, C.; Rocca, F. Permanent Scatterers in SAR interferometry. IEEE Trans. Geosci.
Remote Sens. 2001, 39, 8–20.

Hanssen, R.F. Radar Interferometry: Data Interpretation and Error Analysis; Kluwer Academic
Publishers: Dordrecht, the Netherlands, 2001.

Massonnet, D.; Feigl, K.L. Radar interferometry and its application to changes in the Earth’s surface.
Rev. Geophys. 1998, 36, 441–500.

Mora, O.; Mallorqui, J.J.; Duro, J. Generation of deformation maps at low resolution using
differential interferometric SAR data. In Proceedings of 2002 IEEE International Geoscience and
Remote Sensing Symposium, IGARSS ’02, Toronto, ON, Canada, 24–28 June 2002.

9
Satellite Monitoring of Supernatant Volumes and
Beach Moisture Estimations within Tailings Storage
Facilities
Luis Navarro1, Gabriel Aravena1, Jon Engels1 and Fabián Martínez4
1. Tailpro Consulting, Chile
2. Anglo American, Chile

ABSTRACT

The past few decades have seen a shift towards high density tailings with advances in thickening and
filtering technologies thus improving water recovery and reducing the volumes of water deposited
within a tailings storage facility. However, the vast majority of tailings facilities worldwide still
generate a supernatant pond as the tailings naturally consolidate and bleed, which need to be
controlled in their size and volume, particularly on the larger throughput operations. In some
countries it is a requirement to routinely report to the authorities this volume.
This article presents a study to monitor the development of supernatant water ponds, and to monitor
beach moisture in tailing deposits using remote sensing techniques. Principally this relates to a
multispectral analysis of satellite bands, which by means of various corrections and applying
combinations of these bands through mathematical algorithms, the depths of relatively shallow
supernatant ponds can be determined reasonably accurately. The importance of the study is to
provide an alternative to current in-situ monitoring techniques, such as sonar instrumentation, which
is limited to a minimum depth. The added advantage is that satellite monitoring delivers information
quickly and cost effectively and more importantly avoids the requirement for human presence thus
reducing operational risks.
The study demonstrates that it is possible to monitor remotely water bodies, to an acceptable degree
of accuracy, at a revisit frequency that would be very demanding if carried out using conventional
onsite techniques. To improve the accuracy of the satellite surveying techniques discussed, it is
necessary to calibrate the measurements and compare with data obtained in the field, both for the
bathymetry and beach moisture content generating a soil moisture index. This represents the next
stage of development of this research.

1
INTRODUCTION

In Chile the use of water in the mining industry represents only 3% of the total supply of the country.
However, many mining operations are located in the Atacama Desert, the driest desert in the world,
areas where water scarcity is a limiting factor for development of alternative industries in the region
(COCHILCO, 2018). The quality and quantity of water resources represents a very sensitive issue in
any mining project, since water is involved in almost every area of the operation, which can often
lead to a conflict with other users of the resource. The demand for water in mining is increasing due
to the global demand for metals and the increased number of projects processing lower grade ore at
higher throughputs. The focus of modern-day mining is looking to reduce consumption of water and
the impacts on the surrounding environment, as for example the development of high density
thickened or filtered tailings disposal.

APPLICATION OF SATELLITE REMOTE DETECTION FOR MONITORING WATER

BODIES AND SOIL MOISTURE
The application of satellite bathymetry is used to detect the depth of water, and subsurface profile of
water bodies, such as rivers, lakes, oceans and even small bodies of water such as supernatant ponds
on tailings facilities. On the other hand, through the calculation of the NDWI (Normalized Water
Differential Index), NDVI (Normalized Vegetation Differential Index) and other algorithms, such as
the BI2 index (second Brightness Index), we can identify water masses and zones of high humidity
saturation. In this way we can use the index as a unit of measurement to determine water stress, soil
moisture saturation or direct demarcation of water bodies such as lakes and reservoirs, and in our
case tailings deposits.
The precision and detail of these measurements are controlled by the detectors of the satellite and
their operating resolution whereby commercial satellites have much higher resolutions compared to
the free satellite data available. For this paper the Sentinel 2 satellite has been used, which is defined
as a terrestrial observation mission developed by the European Space Agency (ESA) within the
Copernicus Program. It is composed of two earth orbital satellites named Sentinel 2A and Sentinel
2B offering a wide range of detection wavelengths and the data is freely available, normally within
24 hours of an acquisition.

SATELLITE BATHYMETRY METHODOLOGY AND CASE STUDY

As part of this paper, a case study for a tailings facility was developed and compared with measured
bathymetry data collected from a conventional sonar-based system. Multispectral images of the
Sentinel 2 satellite were obtained, using the Copernicus platform of ESA and processed using the
Sentinel Application Platform (SNAP) software. The architecture of this software is ideal for the
processing and analysis of terrestrial satellite images due to its versatility and large number of specific
tools for Sentinel and other satellites.

2
Post-processing algorithm

The physical principle of the energy received by the satellite can be divided into four basic
components. Lb Lower radiance, Lv Subsurface volumetric radiance, Ls Specular radiance and La
Trajectory atmospheric radiance. Equation (1) identify these components in further detail.

𝐿 𝑇𝑂𝐴 = 𝐿𝑏 + 𝐿𝑣 + 𝐿𝑠 + 𝐿𝑎
(1)
where:

LTOA = Radiance registered by the detector of each band.

To obtain the depth of the body of water observed by the satellite, it is necessary to extract the lower
radiance (Lb) and the volumetric luminosity from the total radiance.
There are two fundamental models to obtain the satellite bathymetry eliminating Lb and any specular
effects. This is achieved by assuming that the bottom of the water is equal to zero and considering
the optically deep reflectance (L∞ o R∞) represents the combined effect of volumetric glare, specular
glare and trajectory of atmospheric radiance.

Ratio method

The first model was proposed by Stumpf & Holderied. (2003) and is based on the following
logarithmic equation:

ln(𝑛𝑅𝑤 (𝜆𝑖 ))
𝑍 = 𝑚1 ∗ − 𝑚0
ln (𝑛𝑅𝑤 (𝜆𝑗 ))
(2)
Where:

Z = depth

m0, m1 and n = constant coefficients for the model

Rw(λi) and Rw(λj) = observed radiances for the optical bands λ i and λj (after the atmospheric
and solar corrections).

For this model, the background depth is estimated based on the attenuation of light and short-wave
radiation that varies spectrally.

3
Analytical method

Assuming that the lower reflectance ratio between two spectral bands is constant for all background
types and the variability of light attenuation caused by atmospheric effects is negligible for a given
area, the estimated depth with the use of the model can be expressed as (Lyzenga 1978, 1985):

𝑍𝑒𝑠𝑡 = 𝛼0 + ∑ 𝛼1 𝑙𝑛[𝐿(𝜆𝑖 ) − 𝐿∞ (𝜆𝑖 )]

𝑖=1
(3)
Where:

 N = Number of spectral bands

 αi (i=1, 2, ….N) = Constant coefficients obtained during the calibration of the
model
 L(λi) = Radiation obtained after the atmospheric and solar corrections. The
logarithm makes the expression linear.

From these methods and carrying out tests with known bathymetries as part of this paper, the
following new algorithm is shown in Equation 5.

𝑍 = 𝛼 ∗ ∑[𝑙𝑛(𝜆𝑖 ) − 𝐿∞ (𝜆𝑖 )] /(𝜈 + 𝜌)

𝑖=1
(4)
Where:

 z = Estimated depth for a selected Region of Interest (ROI)

 α = Coefficient of penetration of luminosity according to studies by Secchi for
attenuation in waters with turbidity (Secchi, 2018)
 ln(λi) = Radiation obtained after the atmospheric and solar corrections
 ln∞(λi) = Original band without corrections
 (ν) y (ρ) = The spectral bands corrected to attenuate surface undulations

Summary of work flow

The work flow shown in Figure 5 for satellite bathymetric estimation, and used in the case study
presented in this paper, is as follows:
 Identification of the study site (date and geographic location)
 Optimal image selection nearest the proposed date (e.g. neglecting clouds and haze)
 Transformation of TOA-BOA atmospheric correction

4
  ROI work area generation (only visible bodies of water)
 Subset of ROI
 Application of mathematical algorithms in spectral bands
 Band generation resulting from post-processing
 Generation of deliverables (contour lines, depth plane, sections)
 Generation of final report

Case Study

For this paper, freely available spectral bands sets of the Sentinel 2A and 2B satellites were used over
the Las Tórtolas tailings facility belonging to Anglo American’s Los Bronces mine in Chile. As a check
to the methodology of the work presented, bathymetric surveys using sonar equipment were used to
compare with the data obtained from the satellite imagery processing (Figure 1 and Figure 2). As part
of the analysis, all satellite bands were used since each of them presented information of interest,
although the first bands blue, green and red presented more detailed information related to the depth
of water.

Figure 1 Satellite image of the original band 2 of the Figure 2 Satellite image of the processed band 2 of
L1C Sentinel 2 TOA (Top Of Atmosphere) the L2A Sentinel 2 with atmospheric correction and
radiance applied with Sen2cor BOA (Bottom Of
Atmosphere).

Results

Four Sentinel 2 acquisitions during July 2018 were analysed and processed. The dates selected were
those that were the closest to the sonar bathymetric surveys for improved comparison purposes. The
bands were analysed with SNAP software and Global Mapper (Blue Marble Geographics, 2002-2017)
and the results presented for each dataset in Figure 3.

5
 PROFILE 1

PROFILE 2

PROFILE 3

Figure 3 Depth map and bathymetry profiles 08-07-2018

Figure 4 presents a comparison between the supernatant pond volumes measured using the satellite
bathymetry method and the sonar data provided by Anglo American. An analysis of these results is
presented in the following section of the paper.
*Sonar data provided by Anglo American

Figure 4 Comparative graph of sonar vs satellite bathymetry measurements

6
Analysis

As the results of the analyses show, the calculated bathymetry volumes are greater for each particular
date compared to the sonar collected data. Although these values are within the expected errors to
be obtained, the following are important to consider:
 The pixel resolution of the Sentinel 2 satellites is 10m, which reduces the accuracy of the
defining the edge of the water body to be analyzed compared to higher resolution satellite
data.
 For the sonar measurements, the boat can only navigate in depths greater than 0.75m. The
non-navigable area was estimated in their analyses.
 The scale factor used for the attenuation of luminosity by penetration in the body of water
(Sechhi disc) can vary between a range of 1.7 to 2.5 or more, for these cases it was estimated
to use a factor of 2.5 that is an average for waters with suspended sediments or high turbidity.
 Climate conditions on the day of capture by the satellites can affect the spectral bands and
the process of atmospheric correction.
 There is a minor difference in the dates between both compared survey methodologies.

SOIL MOISTURE MONITORING METHODOLOGY AND CASE STUDY

The mapping of soil moisture to a high spatial resolution is very important for several applications
in hydrology, agriculture, risk assessment in projects, and in mining. With the arrival of free Sentinel
data at high spatial and temporal resolutions, it is now possible to develop studies related to soil
moisture, which can better meet the needs of users. In this context, the main objective of the study is
to develop an operational approach for the mapping of soil moisture in tailings deposits, generating
a characterization of the beach to identify and quantify the surface of bodies of water, active beach
and inactive, beach with high humidity saturation and dry beach.

Post-processing algorithm

The second Brightness Index (BI2) algorithm represents the average of the brightness of a satellite
image. The result looks like a panchromatic image with the same resolution as the original image.
This index is therefore sensitive to the brightness of soils which is highly correlated with the humidity
and the presence of salts on the surface (Escadafal, 1989).

(𝐵4 2 )+(𝐵32 )+(𝐵82 )

√ (5)
3
where:

 B3, B4 and B8 = Sentinel 2 multispectral bands.

7
Summary of work flow

To carry out the analysis, an area of interest (AOI) was defined inside the Las Tórtolas tailings deposit
(upstream of the walls) of an area of 9,863,278 square meters (9.86 km2). This area must be maintained
constant over time, in all measurements.
From the analysis, 5 ranges of beach moisture have been defined, according to the following criteria:
 Water
 Saturated beach
 High humidity (active beach)
 Low humidity (beach not active)
 Dry beach
The final objective of this analysis is to quantify the variation of the area over time, for the different
degrees of humidity classified according to the previous criterion.
The study considers the analysis of 18 captures of Sentinel 2 images, obtained between November 5,
2018 and February 18, 2019. The work flow shown in Figure 14 for soil moisture characterization, and
used in the case study presented in this paper, is described as follows:
 Identification of the study site (date and geographic location)
 Optimal image selection nearest the proposed date (e.g. neglecting clouds and haze)
 Image resample and Subset of ROI
 Mathematical algorithm application (BI2/NDWI/NDVI)
 Import boundary vector polygon and mask out the TSF
 Pixel classification on 5 defined ranges and pixel count
 Generation of final report (statistics, graphs and figures)

Case Study

As a result of the work previously carried out in the bathymetric monitoring of the TSF, and given
the current condition of the supernatant pond during the driest months of the year (October to
March), Anglo American Chile, Las Tortolas, had requested a study to characterize the beach
moisture in the tailings facility using remote sensing techniques, in order to quantify the variation of
the area over time.
As in the previous case study relating to the satellite bathymetry, multispectral images of the Sentinel
2 satellite were obtained, using the Copernicus platform of ESA and processed using the Sentinel
Application Platform (SNAP) software.

8
The use of 18 Sentinel 2 image captures, obtained between November 5, 2018 and February 18, 2019
were considered. As an example, figures 5 to 7 show the analysis for the capture of the 24th of January
2019. For comparison purposes, the analysis for the other dates are shown in the graphs of Figure 8.

Figure 5 24th Jan 2019, RGB Satellite image L1C Sentinel 2 TOA (Top Of Atmosphere)

Figure 6 24th Jan 2019, processed image on BI2 Figure 7 24th Jan 2019, pixel classification on
Index algorithm processed image

9
Table 1 24th Jan 2019, soil moisture characterization from pixel classification on BI2 algorithm

Total
# Region Area (m2)
pixels
Water 4.783 478.300

Saturated beach 22.446 2.244.600

Active Beach (high humidity) 43.426 4.342.600

No active beach (low humidity) 15.172 1.517.200

Dry beach 12.814 1.281.400

Results
Using the methodology presented above, 18 sets of images were analysed, captured between
November 5, 2018 and February 18, 2019. The following graphs show the characterization of the
tailings pond and supernatant water pond, and its variation in the defined time interval.

Figure 8 General characterization area variation between 05/11/18 and 18/02/19

Analysis
Based on the analysis presented in this paper, and the results obtained, the following comments are
given:
 During the last week of January 2019 and the beginning of February 2019, the highest
temperatures of the year were registered in the central zone of Chile, which could be related

10
 to the decrease in the area of the saturated tailings areas and the area of the pond, with high
losses due to evaporation and temperature.
 The pixel resolution of the Sentinel 2 satellites is 10m, which reduces the accuracy of defining
the edge of the water body to be analysed compared to higher resolution satellite data.
However, the error in the measurement of the area is negligible for large-scale studies. The
area of the study site measured through a polygon in Autocad was of 9,863,278 m2, while the
sum of the area of the pixels of the analysis was of 9,864,100 m2. The difference between these
two areas represents an error less than 1% (0.83%).
 The different humidity ranges were defined and visually selected based on the experience
acquired in tailings management at similar sites. These can be adjusted to coincide with real
beach moisture data if this exists.

CONCLUSION

Satellite bathymetry
This paper has presented bathymetry measurements for supernatant pond volumetric analysis using
freely available satellite data. A case study that compares two independent methods for calculating
the volume of a supernatant pond of an actual tailings storage facility has been presented with
reasonable accuracy being achieved via satellite bathymetry when compared to a conventional sonar
measurement using a boat.
The methodology presented can provide a guide to assessing the volume of a supernatant pond,
without the need to intervene or access the supernatant pond using manual techniques. There are
important limitations to the use of satellite bathymetry, such as the turbidity of the supernatant water
in a tailings facility which can impact the accuracy and interpretation of the true pond depth and is
a factor to assess for a particular facility prior to evaluating the use of satellite bathymetry compared
to conventional methods. Based on the case study presented, having reasonably clear supernatant
water, measurements of more than 12 m could not be detected. Another important limitation is the
local climate for a particular tailings facility location, where frequent cloud cover can prevent
acquisition of images. In this case commercial satellite services would be required to repeat shots
daily until a clear acquisition is made (there are no costs associated for failed acquisition attempts).
Based on the case study presented, an average error of less than 10% was achieved between the
satellite and sonar-based surveys. According to the studies reviewed as part of this paper, satellite
monitoring is the next step to conventional bathymetry processing and accuracy is likely to improve
as the resolution and technology develops further.
As part of the continued development of this paper, the following future work is proposed:

 Analysis using commercial satellites having a better resolution compared to Sentinel 2 and
other free data sources. Funding is currently being acquired to use the Worldview 2 Satellite

11
 or higher, having 8 high resolution satellite bands. Worldview 2 offers a panchromatic band
of 0.46m and multispectral resolutions of 1.8m (DigitalGlobe, 2018).
 According to studies relating to light attenuation in turbid waters (Secchi, 2018), the
constant used predominantly appears in a range between 1.7 and 2.5 and is commonly
derived for open water bodies not related to mining. For supernatant ponds of tailings
facilities, this light attenuation factor requires a more customised approach and is specific
to a particular tailings facility (depending on the turbidity and colour of the supernatant
water). The use of Secchi discs in the field can assist in the determination of this attenuation
factor.

Soil moisture characterization

This paper has also presented a methodology for tailings beach moisture characterization analysis
using freely available satellite data. The methodology presented is an interesting alternative in the
control and monitoring of tailings deposits, which can easily complement day to day operational
activities and review water balance models for a project.
The margin of error obtained from the measurement of area through the quantification of pixels is
considered negligible for large TSFs such as the one presented in this paper.
The second Brightness Index (BI2) was selected as part of this study other algorithms, since this
allows to clearly differentiate the interface of the water body, and the interface of the tailings beach.
As part of the continued development of this paper, the following future work is proposed:
 Analysis using commercial satellites, similar to the work considered as part of the
bathymetry analysis.
 A more in-depth study of the software and programming codes are required in order to
improve and automate post-processing algorithms.
 Complement the study with tailings samples taken in situ, to analyse the correlation
between the graphic characterization and the true moisture content of the tailings on the
beach.

ACKNOWLEDGEMENTS
The authors wish to thank Anglo American for providing bathymetry reports used as part of the
research presented in the paper.

REFERENCES
Comisión Chilena del Cobre (COCHILCO) 2018, Consumo de agua en la minería del cobre al 2017,
Santiago de Chile.

12
Digitalglobe, 2018, https://www.digitalglobe.com/es

Escadafal R. 1989, Remote sensing of arid soil surface color with Landsat Thematic Mapper,
Advances in Space Research Volume 9, Issue 1, Pp 159-163.

European Space Agency 2018a, ESA eduspace, viewed 5 Oct 2018,

http://www.esa.int/SPECIALS/Eduspace_ES/SEMZ3YD3GXF_0.html

European Space Agency 2018b, Sen2Cor configuration and user manual, viewed 5 Oct 2018
http://step.esa.int/thirdparties/sen2cor/2.5.5/docs/S2-PDGS-MPC-L2A-SUM-V2.5.5_V2.pdf

European Space Agency 2018c, Sentinel Online, viewed 5 Oct 2018,

https://earth.esa.int/web/sentinel/technical-guides/sentinel-2 msi/msi-instrument

Global Mapper software 2017-2018, http://www.bluemarblegeo.com/index.php.

Lyzenga, D 1978, Passive Remote-Sensing Techniques for Mapping Water Depth and Bottom
Features.

Lyzenga, D 1985, Shallow-water bathymetry using combined lidar and passive multispectral scanner
data.

Secchi, A 2018, Method for measurement of water transparency or turbidity in bodies of water
Source: https://en.wikipedia.org/wiki/Secchi_disk

Stump, R & Holderied, K 2003, Determination of water depth with high-resolution satellite imagery
over variable bottom types.

13
Using Technology to Identify Seepage Flow Paths
Through, Under and Around Tailings Impoundments
Arturo Fahrenkrog1, Ryan Blanchard2 and Judson Kennedy2
1. HydraMetrix, Chile
2. Willowstick Technologies, USA

ABSTRACT

Tailings impoundments continue to undergo failures at an unacceptably high rate, including

failures at operations owned by high profile mining companies. These failures are often the result
of a combination of design, construction and operation actions that are controlled by humans and
must be better coordinated and managed in the future. The consequence of failure can be the
widespread flow of tailings, environmental damage, social license to operate, company value and
mining industry sustainability. Any additional technology and information that enables an owner
of a tailings impoundment to be more certain of its condition, and thereby reduce the risk of failure
is of tremendous value to tailings and mine water management. One method that can provide
valuable information involves the injection of a low frequency electrical current in the ground
between two electrodes generates a magnetic field that can be measured at the ground surface with
sensitive magnetic sensors. The map of the magnetic field, measured at the frequency of the injected
current, can be used in turn to determine the position of the current paths in the ground. When the
current is channeled in conductive paths, the Magnetometric Resistivity (MMR) method can be
used to detect these paths. Conductive current paths are then interpreted as preferential seepage
flow paths. A case study, from a tailings impoundment at a Zinc Mine in Tennessee, USA, is also
shown. Several preferential paths of electric current were identified and interpreted as preferential
seepage flow paths out of the impoundment. One in particular, identified the source and path of a
spring downgradient of the impoundment (referred to as the “Big Seep”) that appeared decades
ago without any explanation. This method has been shown to be an effective tool to model
preferential seepage flow paths and assist in targeted remedial works.

1
INTRODUCTION

Tailings impoundments are unique and are different to water supply dams for many reasons: they
store tailings (waste liability) versus water (resource asset); their construction and design evolve as
the facility develops and tailings volume increases while codes, regulations and technologies
change; they embrace tailings properties, disposal methods and water management in
impoundment design; and they store tailings in perpetuity.
Properly planned, designed, constructed, operated and maintained tailings impoundments provide
safe and effective structures. Tailings impoundments use a host of tailings and water management
methods to optimize storage space, reduce operation costs, protect the environment and maintain
indefinite impoundment stability and safety.
Despite the experience with tailings impoundments in the mining industry, they continue to fail at
an unacceptably high rate. Tailings dam failures due to seepage and poor management accounted
for 45% of all failures from 2000 to 2009 (Azam & Li 2010). Poor management included
inappropriate dam construction procedures, improper maintenance of drainage structures and
inadequate long-term monitoring programs (Rico et al. 2007). Seepage technologies should be used
to understand and remediate preferential flow paths at depth through, under and around tailings
impoundments to help in their stabilizations, expansions and closures.

METHODOLOGY

This paper describes the Magnetometric Resistivity (MMR) method which uses a low-voltage, low-
amperage, alternating current to energize subsurface water by strategically placed electrodes
(Jessop et al. 2018). At each study area one or more independent surveys are required. Each survey
uses electrodes, connected by an insulated copper wire, and a power supply to energize the circuit
(Figure 1).

Figure 1 Survey setup

2
One electrode is placed in the upstream tailings and the other is placed downstream in contact with
the water below the tailings impoundment (in a well, pond or stream).
The power supply uses about 270 V and 1.5 amps. The exact voltage and amperage varies
depending on the conductivity of the subsurface materials.
As with all alternating electrical currents, this circuit generates a magnetic field (Biot-Savart Law)
that is measured and mapped from the surface using a Willowstick instrument. The instrument
uses three magnetic sensors oriented in orthogonal directions (x-, y-, and z-axes); an integrated
digital signal processor is used to collect, filter and process the sensor data; a high resolution GPS is
used to spatially define the measurement locations; and, a handheld computer with mapping
software is used to couple the GPS data with the magnetic field data and store it for subsequent
reduction and interpretation. All of this equipment is attached to a surveyor’s pole and hand
carried to each field measurement station, and each measurement is collected in eight seconds. This
collected data is used to create two- and three-dimensional (2D and 3D) maps and Electric Current
Distribution (ECD) models of seepage flow paths (Figure 2). The technology maps and models
preferential groundwater flow paths like an angiogram that allows doctors “to see” blood vessels in
a human body.

Figure 2 Steps to create data results and 2D maps

The application of the technology to tailings impoundments is based on the principle that water in-
creases the conductivity of earth materials through which it flows. As the signature electric current
travels between electrodes strategically placed upstream and downstream of the tailings
impoundment, it concentrates in the more conductive zones, where tailings water preferentially
flows out of the tailings impoundment as seepage through, under and around the impoundment.
An electric circuit is established in the water of interest. Measuring the resultant magnetic field at
the surface reveals the electric current flow and distribution. Data is processed and compared to a

3
predicted magnetic field from a theoretical homogenous earth model to highlight deviations from
the “uniform” model. 2D maps and 3D models are generated and combined with known sub-
surface data to enhance preferential seepage path definitions. The purple shading in the ratio
response map (Figure 2) shows actual flow that is less than flow predicted by the “uniform” model.
Green shading shows actual flow that is more than flow predicted by the “uniform” model, so it
likely represents a seepage path. Once seepage paths have been identified by the technology, sheet
pile walls or grout curtains (Kofoed et al. 2011) can be used to remediate them.
This technology has been applied to tailings impoundments all over the world, and it has some
limitations. With each problem there have been ways to work around them to figure out where
water preferentially seeps out of the impoundment. One common problem is conductive culture
(man-made metal objects, metal fences, buried power lines) which can influence the magnetic field
readings. Fortunately, the data can be filtered and the effects of the culture can be modeled out to
provide the needed results. Sometimes, due to property boundaries the circuit wire cannot be
placed far enough away from the study area. And finally, at some sites the groundwater isn’t
conductive enough to establish a good current.
This paper uses a tailings impoundment at a Zinc Mine in Tennessee, USA, for a case study. Several
preferential paths of electric current were identified and interpreted as preferential seepage flow
paths out of the impoundment. This example is used to illustrate the procedures, findings and
benefits of this technology.

RESULTS AND DISCUSSION

Two sinkholes appeared in the interior of a Zinc mine’s tailings impoundment, causing the mining
company to receive a citation. Additionally, water down-gradient of the tailings impoundment
drains to a centralized area called the “Big Seep” as shown in Figure 3. A dye test confirmed that
water from a nearby settling pond was connected to the Big Seep, but the client did not know how
it was connected. The client was interested in knowing if the sinkholes posed a risk to the integrity
of the impoundment. The client also wanted to know the source of water to the Big Seep.

4
 Figure 3 Tailings Impoundment at a Zinc Mine in Tennessee with sinkhole problems

Using the MMR method Willowstick completed an investigation on the tailings impoundment.
There were four individual surveys completed each having a unique electrode configuration. Each
survey had measurement stations on a 33m by 33m grid which provided the resolution needed for
this project. The surveys required a total of 5km of wire and 1543 magnetic field measurement
stations. The field work took a total of two weeks to complete. The MMR investigation identified
five major preferential seepage flow paths in and around the impoundment. In Figure 4, the green
shading shows areas of greater magnetic field intensity and the yellow lines indicate the
preferential seepage flow paths Willowstick identified. The sinkholes were observed above the
preferential seepage flow paths. The Willowstick technology also identified several flow paths that
were converging at the Big Seep.

5
 Figure 4 ECD model slices with preferential seepage flow paths

Due to the successful application of the MMR technology, the client understood what caused the
sinkholes and the source of the Big Seep and this allowed them to begin their remedial efforts
accordingly. Thus saving time and money. The mine was still in the process of making repairs at
the tailings impoundment when this paper was written.

CONCLUSION
The results from the Zinc mine tailings impoundment show how the MMR method can be used as a
non-intrusive approach to supplement known geological, geotechnical, hydrological and
groundwater information to enhance the knowledge of the tailings impoundment’s seepage,
stability and integrity conditions.
This information can then be used to cost-effectively support the design of tailings impoundment
stabilizations and future tailings impoundment raising, ultimate closure and post-closure and to
provide long-term safety and stability.

ACKNOWLEDGEMENTS
The authors gratefully acknowledge the support of HydraMetrix and Willowstick Technologies in
the preparation of this paper.

6
REFERENCES

Azam, S. & Li, Q (2010) ‘Tailings dam failures: A review of the last one hundred years’, Geotechnical
News, December 2010.

Jessop, M. Jardani, A. Revil, A. & Kofoed, V., (2018) ‘Magnetometric resistivity: a new approach and
its application to the detection of preferential flow paths in mine waste rock dumps’, Geophysical
Journal International, Volume 215, Issue 1, Pages 222-239.

Kofoed, V.O. Jessop, M.L. Wallace, M.J. & Quan, W. (2011) ‘Unique applications of MMR to track
preferential groundwater flow paths in dams, mines, environmental sites, and leach fields’, The
Leading Edge, February 2011, Vol. 30, No. 2.

Rico, M. Benito, G. Salgueiro, A.R. Diezherrero, A., & Pereira H.G., (2007) ‘Reported tailings dam
failures: a review of the European incidents in the worldwide context’, Elsevier, Netherlands.

7
Integrated Monitoring of Tailings: EIT STINGS
Karsten Zimmermann1, Daniel Schröder1, Sebastián Fingerhuth2 and
Roberto Gesche3
1. Engineering Surveying and Geomonitoring, Germany
2. Pontificia Universidad Católica de Valparaíso, Chile
3. Universidad de Chile

ABSTRACT
The following article details the work and results of STINGS - an innovation project funded by EIT
Raw Materials to establish a ground- and space borne remote sensing and analysis system to
effectively and cost-efficiently monitor critical ground infrastructure stability and content, primarily
focusing on mining tailing dams. It is dedicated to increase the safety standards related to tailing
operations and to deliver the related mining sector, government, citizens and all stakeholders
affected by previous and current activities with an extended monitoring and early warning system
for identification of operational impact and environmental risk.
The technical feasibility of STINGS developments strongly depends on the underlying technical
components. In this project, in particular differential and persistent scatterer interferometry (dInSAR
and PSI) is of interest in order to detect and monitor surface deformation occurring at tailings.
Furthermore, radar absorption methods are investigated to identify soil moisture changes by multi-
temporal analysis of SAR-Data from Sentinel 1 applied for the detection of ground moisture changes
due to leakages of liquids from the tailings pond.
In-situ sensor development and sensor application to detect movements and instabilities at tailing
sites is another major task in STINGS. While various sensors (GNSS) and data logging technology are
available on the market - within the project the focus is on the development of low cost equipment
with automated communication technology (ICT) for automated sensor data transfer.
STINGS will provide an important part corresponding to the integration of the various systems for
the proposed interdisciplinary monitoring of tailings. The data management and early warning
system developed during the project will accumulate all information and create safety information
about the supervised tailings. The system is based on DMT’s monitoring system SAFEGUARD that
provides a technical foundation and proofs the technical feasibility.
The main regions for system development within the project are South America with a tailing site in
Chile and Eastern Europe with a tailing site in Romania.

1
INTRODUCTION

STINGS is an innovation project funded by EIT Raw Materials to establish a ground- and space borne
remote sensing and analysis system to effectively and cost-efficiently monitor critical ground
infrastructure stability and content, primarily focusing on mining tailing dams. It is dedicated to
increase the safety standards related to tailing operations and to deliver the related mining sector,
government, citizens and all stakeholders affected by previous and current activities with an
extended monitoring and early warning system for identification of operational impact and
environmental risk.
As example for the significance of tailings in the mining industry Figure 1 show the enormous
number of tailing facilities in Chile divided into active, inactive and abandoned operations [1].
Tailings are designed in different types such as material embankments or piled up material both with
large ground footprint.

Figure 1: Number of mining related tailing operations in Chile and example pictures of tailing operation types

Global data on failures of both operated and closed tailing dams are extremely scarce and incomplete,
much of this due to differences in legislative reporting requirements and to some degree as a
reflection of willingness of failure reporting. Nevertheless, available data shows that there has been
a significant amount of tailing dam failures during the last decades (Fig. 2). Data shows that the
number of serious (> 100 Km³ mass release) and very serious (> 1 Mm³ mass release) failures are

2
increasing (both including loss of life). This is mainly driven by larger tailings becoming necessary to
allow economic extraction of lower grades of ore. This tendency is predicted to continue at a rate of
at least 1 - 2 major failures every year from 2017 - 2020. Increasing global mining operations need the
availability of new technologies to monitor tailings safety effectively and cost-efficient. Here, a key
value STINGS will provide is the intersection between the increasing risks of tailing management
and the improved technologies of monitoring.

Figure 2: Reported global tailing dam failures and severity over time [1], [2]

INTEGRATED MONITORING APPROACH

By integrating and analysing a broad range of complementary ground and space borne sensor data
STINGS offers a unique information service to identified risks, prevent disasters and to capture
previously undetected secondary values. The monitoring system will focus on different information
types, different sensors and combines them in order to generate the most reliable information related
to tailings stability, chemical and mineralogical content (Fig. 3). It will implement satellite radar
information as well as ground based sensor data detection of mechanical movements. An additional
integration of optical remote sensing information enables to detect observable exits of substances
from tailings. By using these different kinds of sensor information, data analysis and modelling the
system will create safety information relevant for early warning. STINGS will offer cost efficient
performance improving automation for monitoring of both active and closed tailings sites while
providing a holistic risk management tool helping monitoring operations to ensure both measuring
as well as analyses and reporting.

3
 Figure 3: Integrated monitoring approach of STINGS

The technical feasibility of STINGS developments strongly depends on the underlying technical
components. Satellite radar interferometry is a cornerstone of remote sensing. The European Union
(ERS 1, ERS 2, ENVISAT, and now Sentinel 1A and B) and Germany (SRTM, TerraSAR-X, TanDEM-
X mission) are leading players in this field of technology. In this project, in particular differential
interferometry (dInSAR) is of interest in order to detect and monitor surface deformation occurring
at tailings. For this application, the capabilities of the current operational European and German
radar satellites are complementing quite well. Furthermore, radar absorption methods are
investigated by STINGS to identify soil moisture changes by multi-temporal analysis of SAR-Data
from Sentinel 1 as an indicator for the potential dry out of tailings. The results can be used as an early
warning signal for particle emissions by wind from the tailing pond. In addition, the same
methodology is foreseen to be applied for the detection of ground moisture changes due to leakages
of liquids from the tailings pond. Optical Remote Sensing with data namely from the Copernicus
Sentinel 2 mission (Fig. 4) and hyperspectral missions (e.g. Hyperion, EnMap) represents a keystone
of the planned system.

4
 Figure 4: Sentinel-2 image of Mina La Africana, Cía Minera Pudahuel (ESA)

Optical remote sensing is already applied for exploration and monitoring tasks in the mining sector.
Recent results from R&D-projects clearly demonstrated the potential of new optical space-borne data
from missions like Hyperion and Sentinel 2 e.g. for change detection of water bodies and soil
moisture due to mining-related ground movement.
In-Situ sensor development and sensor application to detect movements and instabilities at tailing
sites is another major task in STINGS. While various sensors (GNSS) and data logging technology are
available on the market, within the project we focus on the development of low cost equipment with
automated communication technology (ICT) for automated sensor data transfer.
System and data integration is a main challenge of the digital edge and specifically for what is called
“Industry 4.0“ - a collective term embracing a number of contemporary automation, automatic and
intelligent data exchange and fully digitized manufacturing technologies. STINGS will provide an
important part corresponding to the integration of the various systems for the proposed
interdisciplinary monitoring of tailings. The data management and early warning system developed
during the project will accumulate all information and create safety information about the supervised
tailings (Fig. 5). The system is based on DMT’s monitoring system SAFEGUARD that provides a
technical foundation and proofs the technical feasibility.

5
 Figure 5: STINGS service illustration

PROJECT PARTNER NETWORK

The partner composition of STINGS combines user partners to determine market requirements,
industrial partners for commercialization and the thematic expertise of scientific institutions [3]. The
partner consortium includes DMT (Germany), EFTAS (Germany), University of Stuttgart (Germany),
Technical University of Freiberg (Germany), Ovidius University of Constanta (Romania), LTU
Business (Business Consultation, Sweden), Geological Survey of Slovenia (Slovenia) and the
University of Valparaiso (Chile).

6
The main regions for system development within the project are South America (focusing on Chile)
and Eastern Europe (focusing on Romania). In Chile there are a substantial amount of mines and
associated tailing operation as well as interesting mining associations [e.g. ENAMI], where both
medium sized and larger mining companies participate, which offers an starting point for the
STINGS service. In Romania there are several potential test sites that are currently closed down, but
could be re-opened, which is the case for many currently closed down mining operations in Romania.
In this case it would be possible for STINGS to support in consulting and setting up monitoring
equipment and practices in both Romania and the region in general.
To improve tailing monitoring specifically in Latin America one of the core project partners is PUCV
(Pontificia Universidad Católica de Valparaíso) from Chile. PUCV has already conducted intense
research focusing on improving operational controls and generating technologies for recovering
areas occupied by tailings deposits. Within STINGS, the university will give valuable technological
input, enables to select a suitable site for the system developments at large local tailing operations
and will evaluate the results of the system for use in Chile.
In Europe STINGS partner Ovidius (University of Constanta) will specifically contribute to the
technology development for observation of chemical content of tailings, emission of particles by air
(dust), moisture of the tailings inside the tailing basin, chemical content of liquids inside the tailing
and detection of potential leakages. The activity will be carried out mainly on the Romanian Moldova
Noua site in selected tailings and other relevant geological spots.

ACKNOWLEDGMENT
STINGS project activities have received funding from EIT Raw Materials (https://eitrawmaterials.eu),
the European Institute of Innovation and Technology (EIT), a body of the European Union, under the
Horizon 2020, the EU Framework Programme for Research and Innovation.

REFERENCES

[1] Tailings dam failures 1915-2017” Xls sheet. Compiled by David Chambers of CSP2, and Lindsay
Newland Bowker of Bowker Associates. http://www.csp2.org/tsf-failures-1915-2017 [2017-06-10]

[2] Bowker Newland Lindsay, Chambers Davis. In the Dark Shadow of the Supercycle Tailings
Failure Risk & Public Liability Reach All Time Highs, Environments 2017

[3] Project website, www.tailing-monitoring.com

7
Integration of Monitoring Instrumentation for the
Laguna Seca TSF Surveillance: an Innovative
Approach
Nicolle Correia1, Cristian Roa1, Carlo Calderón2 and Mauricio Vargas2
1. Minera Escondida, BHP, Chile
2. Timining, Chile

ABSTRACT
During 2018, the Chilean government anounced that it will work on a new National Tailings Policy,
focused on safety and the enviromental impacts associated with the more than 710 tailing storage
facilities (TSF) identified throughout the country. In the specific case of active TSFs, companies are
expected to report the status of their facilities online and in real time.
At Minera Escondida Ltda. we are initiating a process to integrate the instrumentation used to
monitor dam stability in order to understand the state of the infrastructure by implementing sensor
telemetry, data centralization and visualization, data analysis within its operational context, all this
for the purpose of providing timely decisions and preparing a platform for appropriate
communication with the company and authorities.
Currently, the Laguna Seca TSF’s instrumentation system is based on vibrating-wire, fiber-optic,
and Casagrande piezometers, accelerometers, clino-extensometers, satellite interferometry (InSar)
and survey monuments. The challenge is to automatically integrate the data coming from the
different sensors, together with the topographic and field inspeccion components.
Over the second half of 2018, we have started a project that seeks to provide an innovative response
to the challenge posed by the corporation and government authorities, using the TIMINING-Aris
platform to monitor dam stability. In this paper, we present the purpose of the project, the main
challenges faced during the implemention process and the first results of the integrated
survelliance.

There is no full article associated with this abstract.

1
Improvements on Instrumentation and Monitoring
over Critical Variables in Ovejería Tailings Dam
Manuel Acuña and Alvaro Gutiérrez
Andina Division, Codelco, Chile

ABSTRACT
The geotechnical instrumentation is part of the safety and operational program of Ovejería Tailing
Dam – CODELCO Andina. Then, its maintenance and monitoring have been established as an
essential issue in the dam operation.
In the past two years the Superintendencia de Depositos de Relaves (Tailings Dam Superintendence)
have developed new displays to monitor the vibrating wire piezometers (VW) installed in 13
transversal sections of Ovejería Tailing Dam. The 42 piezometers have been configured to record
water levels in steps of 1 hour and then the data is transmitted via wireless to a centralized server
located 3 km away.
The advantage of these new display, available in the software PI Processbook for all the users of
Andina division, is that it includes the VW and Casagrande piezometers in the same display. The
foregoing permits to verify the appropriated data measurement, or even better, to see flaws,
improving the decision-making in cases of relevant water level variations.
During the last year have been installed new devices with the ability to monitor pH, conductivity
and turbidity, as a trial, in the Ovejeria Tailing Dam drainage system. These devices have the
capability to record, in steps of 15 seconds, the variables mentioned above, allowing to identify an
increasing amount of solid contents of the drainage water, preventing eventual piping inside the
dam.
The benefits observed so far, are the accuracy and quantitative data recorded from such devices,
which are stored in the server for future investigation and queries (long term data). These kind of
devices are recording data in natural conditions, capturing the data as needed for the loading
conditions (sand deposition). Also, these devices can be used to generate alarms to notify the workers
as needed.

There is no full article associated with this abstract.

1
Classification of Tailings by Remote Sensing
Methods by U.A.S
Jimmy Ramos
Minera Los Pelambres, Chile

ABSTRACT
The present study originates from the search for continuous improvement in order to increase the
recirculation of water in the mining processes in order to minimize the consumption of fresh water
in the long term, that is to say, reduce the strategic risk that turns out to be this limiting factor in the
development of the industry.
The aim of the project carried out, is the use of remote sensing techniques for the classification and
the monitoring and control of the tailings. Using the methodology to determine the coverages of
interest and quantify the surfaces present in the basins of tailings through the use of Unmanned
Aerial Systems. It processes the information generated by the flights, using methods
Photogrammetric ,obtaining Orthophoto, DSM, DTM, then the process of Teledección is performed
with the software ILWIS to get the classes of the tailings, using the classification algorithms
(supervised and non-supervised). As a by-product was obtained from the map of classification of
the tailings, the areas of classification and temporal analysis of images with the comparison of the
months monitored. With the use of the System unmanned Aerial eliminates the risk of human in
the dams of tailings and facilitates efficient results in the classification, the methodology allows to
classify the tailings and water the tailings, with the following classes: tailings high moisture,
tailings wet, tailings dry with moisture, tailings dry without moisture; in the case of water, we get
the following classes: deep, high deep, medium-depth , shallow , giving the state and the
availability of water reuse in the metallurgical process, and by monitoring bathymetric robotic body
of water to complement the results obtained.

There is no full article associated with this abstract.

1
Integration of Geospatial Data for Monitoring and
Analysis of Tailings Pond “Las Tórtolas”, Anglo
American DLB Company
David Hermosilla
Linkapsis, Chile

ABSTRACT
The tailings pond Las Tórtolas of Los Bronces’s mining has been in operation since year 1992 and has
a total surface area of 10,000,000 m2. Currently, this reservoir has a humid surface of approximately
320,000 m2 and is projected to give operational continuity to the mine until year 2060.
The company Linkapsis executes the contract of “Tailings Pond Topography Services and STP” since
year 2016 for the Tailings Pond Operation Management of Anglo American Mining Company, Los
Bronces Division. The main objective of the contract is to obtain the geometry of the deposit and its
surroundings on a daily basis. For this purpose, innovative mapping and geospatial surveying
technologies (LiDAR, UAV, USV, Satellite Imagery) are integrated to get volume calculation of wet
areas with availability of water resources, for re-entry into mining processes, as well as specific
surfaces for analysis.
This presentation shows preliminary results of a methodology that involves the processing,
classification, geostatistical modeling and spectral analysis of high spatial resolution satellite
imagery, treated with atmospheric corrections of surface reflectance, in order to establish an estimate
of the depth of the water resource in a reservoir. Mathematical modeling of spectral variables of the
image is applied, calculating the degree of spatial correlation between this estimated value and the
real data measured by means of unmanned bathymetric surveys - USV (Unmanned Surface Vehicles),
which nowadays replace or complement the use of manned platforms.
With this experience, it is expected to have a future integration of predictive artificial intelligence
models, which can translate, using geospatial data from satellite platforms, and multiple variables
that need to be managed and analyzed within a reservoir. In this way, current lengthy methodologies
can be optimized, and also a positive impact on safety and productivity in mining surveying works
can be obtained.

There is no full article associated with this abstract.

1
Increasing Tailings Safety and Governance via Real
Time Monitoring Utilising Long Range Wide Area
Networks (LoRaWAN)
Joe Carr
Inmarsat, United Kingdom

ABSTRACT
Instrumentation data from Tailings Storage Facilities is often collected on-site and mining companies
are largely reliant on human activity for the collection, storage and analysis of data. This process has
many potential risks, including poor data collection practices, inconsistent storage and interpretation
of data and crucially alarms or risks being missed or not understood which could lead to a breach or
failure of a Tailings Storage Facility.
Inmarsat has developed over the last 18 months a system which combines existing on site
instruments, LoRaWAN technology, edge computing and standalone satellite for real-time data
collection and analysis which is in agreement with the game-changing improvement in how
companies manage their current tailings storage facilities.
The solution collects data from a range of industry standard sensors via edge connectivity before it
is transferred across the network to a cloud dashboard. This enables mining companies and national
regulators to gain a comprehensive view of the status of their dams, with metrics such as pond
elevation, piezometric pressures, inclinometer readings and weather conditions displayed in one
place, no matter where the mine is located and critically all in real time. The system has a range of
features including digitally signed, searchable audit logs, customisable alarms which also include
escalation hierarchies; ensuring an alarm is never missed, searchable and exportable alarm logs. The
monitoring system also operates with standalone power and communications ensuring the same
system can be used on any site, anywhere on the planet. The aim of the system is to remove the
human element from the monitoring equation ensuring consistent, compliant, reliable monitoring
This presentation outlines the challenges faced by global mine tailings management, before
introducing a new remote tailings dam monitoring solution which enables smarter and faster
decision-making, improved safety standards and helps to support regulatory compliance.

There is no full article associated with this abstract.

1
Searching for Precursors of the Córrego de Feijão Dam
Collapse in Brumadinho using Advanced Satellite
InSAR
Giulia Tessari1, Peter Stegmaier2 and Paolo Pasquali1
1. Sarmap, Switzerland
2. CreaLinQ, Switzerland

ABSTRACT
Brumadinho mine complex, located South-West to Belo Horizonte, Brasil, has become sadly popular
for the dramatic disaster of a tailing dam collapse, Córrego do Feijão, occurred on 25 January 2019.
The dam collapse caused a catastrophic mud flood, involving 12 million cubic meters of released mud
and causing more than 200 confirmed deaths.
This work aims to shed new lights on possible precursory deformations affecting the tailing dam.
Space-borne Synthetic Aperture Radar (SAR) Sentinel-1 and COSMO-SkyMed acquisitions have been
analyzed through multi-temporal differential interferometric techniques, the Small-Baseline Subset
(SBAS). The great potential of these remote sensing techniques is due to availability of large data
archives and the capability to retrieve displacement with millimeter accuracy. Analyses of results
could demonstrate the importance of these techniques on mining monitoring practices. Moreover,
this case can help on illustrating the potential and limitations of this technique, highlighting the
prerogative of several SAR sensors working with different wavelengths and resolutions and
discussing the applicability on tailing dam and mining context.

1
6
Governance and
Operational
Experiences
Quantification of Risks in Tailings Deposits
Juan Rayo
JRI Ingeniería, Chile

ABSTRACT
Tailings management unit operations include dewatering (thickening, filtration), transportation
(slurry, conveyors, trucks) and disposal (saturated, dry). All these operations have inherent risks that
could reduce mining production, negatively impact the environment, harm people, among others,
and all of them mean financial losses for the mining industry. The magnitude of the risks associated
to tailings management is wide, from minor losses in production or remediable impact in the
environment, to catastrophic situations such as collapse of walls causing a massive flood of slurries,
with fatalities and subsequent reputation damage, losses in production and financial crisis. In the
light of recent failure of dams in Brazil, Canada and China, in addition to the disheartening climate
change predictions, a strong methodology to evaluate risks in tailings disposal is required by the
mining industry.
For more than 20 years, JRI has developed a deep and accurate methodology to evaluate risks in the
Chilean mining industry. In this paper, JRI’s risks calculation methodology is shown, but applied
only to tailings disposal operations.
To clarify the methodology, two real cases (not identified) are explained.

1
INTRODUCTION

Human life is a permanent challenge, with several dangers every single day (car crashes, crime,
terrorism, wars, sickness, earthquakes, storms, etc.), but life expectancy has grown from 50 years in
the 19th century, to 70 years in the 20th century, and currently is over 100 years for babies born after
year 2000. That means humans could survive most of the dangers if we could evaluate and control
them, (Penman A, 2001).
In the mining industry, risks are present in all of the operations, from the mine to the smelter, and of
course in tailings disposal. Major tailing dams review is shown in table 1 (Danihelka P, Červeňanová
E, 2007).

Table 1 Examples of historical accidents

Mine name/location Incident date Impact

Buffalo Creek, USA 26.02.1972 125 deaths, 500 homes destroyed

Mufilira, Zambia 25.09.1970 89 deaths, 68.000 m3 into mine workings

Omai, Guyana 19.08.1995 4.2 million m3 cyanide slurry released

Placer, Philippines 02.09.1995 12 deaths, 50.000 m3 released

Los Frailes, Spain 24.04.1998 released 4-5 million m3 of toxic tailings

slurries

Stava, Italy 19.07.1985 269 deaths, tailings flowed up to 8 km

Aitik mine, Sweden 09.08.2000 1.8 million m3 water released

Baia Mare, Romania 30.01.2000 100.000 m3 cyanide contaminated water with

some tailings released

Baia Borsa, Romania 10.03.2000 22.000 t of tailings contaminated by heavy

metals released

Merriespruit, South Africa 22.02.1994 17 deaths, 500.000 m3 slurry flowed 2 km

Chile has more than 300 tailings deposits, 1/3 in operation, 2/3 closed, built over a period of more
than 120 years, with a total amount of dry solids storage of more than 16.000 MTon deposited in land.
The catastrophic failures in Chilean tailing dam, reported by Sernageomin and other sources, that
includes collapses, with fatalities and/or considerable environmental damage, are around 12 cases. A
rougher statistical analysis indicates that, in Chile, one tailing dam collapses every 10 years and the
average probability that a single tailings disposal facility fails during its 30 years of life is close to 1%
per year (value not negligible). Current design is stricter, with regulations related to foreseen unstable
walls (upstream method), better geotechnical simulations, higher quality controls, etc. But tailings
operations are growing, walls are higher (over two hundred meters), climate is more adverse, and
eventual lack of controls by operators in order to minimize costs could reduce the advantage of better

2
knowledge. The recent failures of 3 big tailing dams in educated countries demonstrate that the level
of risk in tailings management is practically maintained during the last decades.
Failure in a tailing deposit is not a single or cataclysmic event that occurs overnight. Instead, failure
stems from the addition of errors in judgement repeated during long terms (few years). Proactive
companies better evaluate the risks and live longer, face less crisis, do better in financial terms and
have a better reputation. Unfortunately, less than 15% of compared companies could be considered
proactive (Fortune). Most of the Chilean mining companies have made quality evaluations of risks
in tailing dams, with standard tables, but final results normally only show results as follows:
- Weak quality of dams : Probability medium / consequence high
- Water infiltration issue : Probability medium / consequence medium
- Erosion by wind & dust : Probability medium / consequence medium
With that information, the executives are normally reluctant to provide financial resources to do
mitigation work and only define focus in more controls by the operators. Additionally, most of the
Chilean mining executives have only few years in top positions and they are focused in short time
profits. But, if the executives are really informed that all of the 3 risks shown above could mean
hundreds of millions of dollars in exposure during their period of management, they would probably
take actions to mitigate the level of risks.

METHODOLOGY FOR RISK QUANTIFICATION

The amount of money involved in risks is the main factor that allows defining actions and work to
prevent and control risks.

𝑅𝐼𝑆𝐾 = 𝑃𝑅𝑂𝐵𝐴𝐵𝐼𝐿𝐼𝑇𝑌 𝑂𝐹 𝐻𝐴𝑍𝐴𝑅𝐷 ∙ 𝐶𝑂𝑁𝑆𝐸𝑄𝑈𝐸𝑁𝐶𝐸 (1)

A unified and transparent approach to evaluate risks and consequences is possible. If a tailing dam
has important values of expected risks, the costs to mitigate the risks are profitable. As an example,
during the 90’s, a large tailing dam in Spain, named Los Frailes, collapsed and immediately the
shareholders of the mining company lost 95% of their share value, that was partially recovered after
8-10 years. Recently, the failure of Samarco dam will mean at least hundreds of millions of dollars of
losses for BHP-Billiton.

RESULTS AND DISCUSSION

Risk qualifications

There are two kinds of risks depending on the primary event that could produce the failure:
• Risks from nature (earthquakes, floods, global warming, severe winds, others).
• Operational risks (geotechnical, inadequate design, bad practices, low control, others).

3
Most of the risks have statistical values that could be considered. Some useful data is shown in table
2.

Table 2 Statistical values of risk

Risk Statistical values

Earthquake with high destructions potential 1/50 to 1/200 per year (by region)
(Zonal or cortical)

Floods with high destruction potential 1/100 to 1/1000 per year (by region)
(Affected by global climate change)

Strong winds with erosion potential 1/5 to 1/100 per year

(Depending on the area)

Geotechnical failures (Lack of site data, weak 1/50 to 1/2000 per year (depending start date)
designs, lower construction standards)

Failure by bad operational practices (Low or no 1/10 to 1/200 per year

controls, major design change, works
postponed)

The ten major events that produce risks in Chilean tailing dams are:
• Strong earthquake in the tailing area
• Big flood discharging over the tailing disposal
• Snow or mud avalanches over the tailing deposit
• Strong winds that erode dams
• Geotechnical failures of site
• Lack of storage capacity by operational weaknesses
• Water infiltration from the tailings trough the soil
• Lower dam stability
• Lower quality of drainage systems
• Collapse of water diversion systems

4
Hazard probability

All the calculations for probabilities of a big failure in a tailings deposit during its operational life are
based in available statistics in addition to specific evaluation of the project definitions and operation
performance, which is summarized in equation 2.
𝑃=𝐹∙𝑌 (2)
where:
P: probability
F: yearly probability of failure
Y: expected life
If P is 1.0 or more, it is expected that the tailing dam will suffer the failure during its life. If P is lower
than 10-6, the hazard probability is negligible and no action should be considered.

Probability calculation
It is possible to calculate the probability for each event that could take place on the tailing dam as
follows:
𝑃 = 𝐹 ∙ 𝐹𝑠 ∙ 𝐹𝑑 ∙ 𝐹𝑐 ∙ 𝐹ℎ ∙ 𝑌 (3)
F: statistical data base (events per year)
Fs: site factor (0.5 to 5.0)
Fd: design factor (0.5 to 5.0)
Fc: construction factor (0.5 to 5.0)
Fh: human factor (0.5 to 5.0)
Statistical database is a gross evaluation of published data regarding similar events. In the case of a
combination of events, the independent probability should be multiplied. Site factor (Fs) depends on
the characteristics of the site where the tailings deposit is installed, as shown in table 3.

Table 3 Values of site factor (Fs)

Value Definition

0.5 Very dry zone, without creeks and high resistance of soil

1.0 Reasonable area and low hydrological risks

2.0 – 5.0 Narrow valleys, large hydrogeological area, low quality of soil (salts/clays)

5
Design factor (Fd) depends on the quality of the tailing dam project, as indicated in table 4.

Table 4 Values of design factor (Fd)

Value Definition

0.5 Well defined geotechnical site data, high safety factor for drainage and dams design,
water drainage systems carefully designed

1.0 Normal design, according to current standard and best practices

2.0 – 5.0 Design with weaknesses in different levels and areas

Construction factor (Fc) depends of the competence of constructor and quality controls from owner
and it is valued as indicated in table 5.

Table 5 Values of construction factor (Fc)

Value Definition

0.5 Very well constructed dams and first-rate civil works, high level of compaction, low
permeability, good materials, records, etc.

1.0 Constructed following all design criteria, with reasonable controls

2.0 – 5.0 Bad construction practices, low level of QA/QC, lack of records, delays, etc.

Human factor (Fh) is the behaviour of the mining company regarding the tailing disposal operation
and it is valued as indicated in table 6.

Table 6 Values of human factor (Fh)

Value Definition

0.5 Very good operators, supported by experts, with updated operational standards and
emergency procedures.

1.0 Normal operator capacity, adequate governance, with procedures and controls.

2.0 – 5.0 Bad operational practices, untrained operators, no procedures, lack of interest from
executives.

It should be noted that negative situations have higher values than positive situations. To illustrate,
a very well designed tailing dam, yet poorly operated, leads to higher risk.
In the case of a tailing dam with 20 years of operation, in a flat deserted zone, with a reasonable
design, maintained with restrictions and with low interests from executives, its risk of failure is:
𝑃 = 0.01 ∙ 0.5 ∙ 0.1 ∙ 2.0 ∙ 3.0 ∙ 20 = 0.60 (4)

6
This deposit has a 60% chance of having a collapse during its useful life. The estimated impact is 100
MUS $.

Consequence calculations

The economic impact of a dam tailing deposit failure could be calculated as follows:
𝐶 = 𝑉 + 𝑅𝑝 + 𝑅𝑚 + 𝑇 + 𝑀 (5)

The operational losses are obtained from the equation 6.

𝑉 = 𝐿 𝑥 𝑈 + 𝐹𝑐 (6)
Where:
C: Total economic loss (MUS $)
V: Operational losses (MUS$)
L: Lack of production (TF Cu)
U: Non-recoverable profit (US$/TF Cu)
Fc: Fixed costs to be maintained (MUS$)
Rp: Repair costs (MUS$)
Rm: Remediation of environmental affected areas (MUS$)
T: Compensation to third parts (MUS$)
M: Penalties from Authority (MUS$)
As an example, an important failure in a tailing transport system for a big mining company, with
thousands of tons of tailings discharged to agricultural lands, would probably mean 6 days of lack of
production, a long time remedial work plus heavy penalties from authorities.

𝐶 = 6 ∙ 1.5 + 3 + 10 + 5 + 5 = 32 𝑀𝑈𝑆$ (7)

The cost of this failure is 32 MUS$. This type of event normally has a probability of 3% per year.

Accuracy in calculations
According to JRI, using FEL1 level of engineering for the analysis, the method has an accuracy of -
50% / +80% for probability and -35% / +50% for consequences.

Tolerance limits
Each dangerous event could be defined as a combination of ranges of probability and impact. The
mining companies should define where the borders between acceptable and non-acceptable risks are.
Acceptable means to take only control related measures and non-acceptable means to define actions,
works and expenses.

7
JRI uses the following criteria to define the tolerance limits, for all of the expected life of the tailing
deposit:
 A high probability event (10% per year) with low economic impact (could be 2% of total
yearly cost of the mining company) is deemed acceptable.
 A very low probability event (0.01%) with high economic impact (20% of company VAN) is
deemed acceptable.

With both criteria, for a typical copper mine of 100 KTPD, 0.5% Cut, 20 years of life, copper price 3.0
US$/lb, 88% global recovery, opex (C1) 2.0 US$/lb, the tolerance parameters would be:
 Probability: 10% Impact: 15 MUS$
 Probability: 0.01% Impact: 350 MUS$

Graphical analysis
The calculation of probability and consequences, including errors, plus the tolerance curves, must be
shown in a log-log graphic. Previous examples are shown in Figure 1.

Figure 1 Logarithmic relationship between consequence and probability

8
Analysis of real cases
Case N° 1 (unidentified company). A long time ago, a mining company had two big tailings in a
narrow valley and a permanent river was diverted through a poorly designed, constructed and
maintained tunnel.
Probability calculations showed high values:
𝑃 = 0.001 ∙ 4 ∙ 2 ∙ 2 ∙ 2 ∙ 14 = 0.45 (8)
This is equivalent to P= 45%. Unfortunately, the tunnel collapsed, with a partial plugged, after 8 years
of operation and 2 years after the risk evaluation took place. Risk evaluations defined that a crisis
could occur and defined guidelines to manage the situation as well as proposed to build a new tunnel
for a great water flow diversion. The new tunnel was in construction when the collapse occurred.
Remedial works included 3 emergency ways to bypass the river in the tailings zone and the
overtopping risk in the main sand wall was controlled (after few days of hard work). If the collapse
would occur, the wave was calculated between 800 m3/s to 1.200 m3/s (two different sources) entering
a big city downstream of the mining area. Economic consequences included near 100 days of lack of
production, strong remedial works, fast track method for new tunnel, penalties and compensations.
Updated total costs of this emergency are estimated to be 200 to 300 MUS$. Currently, the system is
operating safely with a low risk probability (with 2 tunnels).
𝑃 = 0.001 ∙ 0.5 ∙ 1 ∙ 0.5 ∙ 1 ∙ 20 = 0.0005 (9)
That is P= 0.05%. Then, C = 30 MUS$ (only lack of production in case of flood). See analysis in Fig. 2-
A.

Case N° 2 (unidentified company). Some time ago, a mine company decided to initiate its operation
with a short time tailings dam constructed in a narrow valley close to the metallurgical plant. The
site evaluation was complete, the design was done according to all the rules, and operations used an
unknown contractor. The main risk was the high, growing rate of the lake waters during the first
period of construction (close to 3 meters per months).
Risk calculation can be difficult in the sand dam construction during the first 30 months of operations
(lower sand slope than expected / lower sand production at quality expected / others).

𝑃 = 0.01 ∙ 1 ∙ 1 ∙ 2 ∙ 3 ∙ 3 = 0.18 (10)

That is P= 18%. A consequence of overtopping failure could be a 3-5 m wave trough an agricultural
valley resulting in high environmental damage. Then C = 600 MUS$ (updated value). Mitigation
works were proposed, and the company agreed to increase the starter dam in 10 meters with an
additional wall over its crest and special rules to prevent a crisis if sand construction were to have
delays, plus strong training of operators. In this situation, the probability dropped to 2% and
consequences were reduced to 10 MUS$ (only lack of production). The real situation was that sand

9
production had a big problem (one cyclon battery failed and sand production dropped 40% for few
weeks), but operators where well-trained and initial sand wall was built in appropriate time.
Mitigation work was a successful (with only 3 MUS$ of updated cost). See the analysis in Figure 2-B.

Figure 2 (A) Case analysis 1; (B) Case analysis 2

CONCLUSION

 JRI’s method to calculate risks in tailings dam in operation allows for the evaluation of the
financial magnitude of danger (in probability and consequence), plus the impact and profit
of mitigation actions.
 This method enables the calculation of the NRV (updated value of the risk) to compare with
the NPV of the mining project, and to define mitigation works and its capex costs that could
reduce the NRV.
 JRI´s method makes it possible to identify whenever a risk is important enough in terms of
probability and consequence that it could result in mitigation work and incur major costs.

REFERENCES

Penman A. (2001). Tailings Dams: Risk of Dangerous Occurrences. Geoenvironmental Engineering:

Geoenvironmental Impact Management, pp150-156.

Danihelka P, Červeňanová E. (2007). Tailing Damps Risk Analysis and Management. UNECE
Workshop on TDS, Yerevan.

10
Construction Quality Assurance Audits: An Effective
Tailings Governance Tool
Jorge Troncoso1 and Daniel Warner2
1. BGC Ingeniería, Chile
2. Minera Candelaria, Lundin Mining, Chile

ABSTRACT
The International Council on Mining and Metals (ICMM), has determined that appropriate tailings
stewardship and management are fundamental to prevent catastrophic failures of tailings storage
facilities. Among several tailings stewardship and management tools, construction quality assurance
(CQA) audits are perhaps not used as frequently. The CQA audit program referenced in this paper
is a process by which an independent tailings facility construction expert conducts regular on-site
inspections, during construction, of the construction quality control and assurance processes
implemented by the project execution team, to identify or anticipate deviations from the design
approved by the regulator and owner.
This paper describes a successful CQA audit program implemented in the construction of a new
tailings storage facility in northern Chile. The construction phase of the project was deemed a success
by the owner. Elements of the CQA audit program that contributed to the success of the project are
identified as well as opportunities for improvement. Realized benefits from the audit program are
described based on tangible and specific examples that can be used in other projects. The authors are
of the view that the described CQA audit program is a contribution to effective tailings governance,
it adds resilience to the TSF and facilitates integration of several project stakeholders.
INTRODUCTION AND BACKGROUND

Compañía Contractual Minera Candelaria´s (CCMC) current life-of-mine (LOM) extends until 2030.
A new tailings storage facility (TSF) was needed to allow continued operation for the LOM. This new
TSF, named Los Diques, is adding 600 Mt of tailings storage capacity since commissioning in mid-
2018, sufficient for the current LOM at a throughput of 70 ktpd. The main characteristics of this new
Los Diques TSF are:
 Conventional downstream tailings dams (Main Dam, South Dam and North Dam), built
with compacted waste rock, Main Dam starter dam elevation of 773 masl, to be built in nine
phases to an ultimate elevation of 873 masl (height of approximately 160 m).
 Efficient water recovery from the TSF is a main objective. To meet this objective, the Main
Dam starter dam considers a permeable upstream slope face consisting of a protective
layer, a filter layer and a transition layer over the waste rock, and a network of finger drains
at the base of the Main Dam to collect and convey seepage to a collection system (SCS).
 Seepage collection system located downstream of the Main Dam, consisting of a drain wall
equipped with pump-back wells, a low permeability plastic concrete cutoff wall anchored
in bedrock and a grout curtain extending to low permeability rock. In addition to water
recovery from the TSF, this SCS is meant to prevent migration of tailings seepage water.
Wells are installed upstream and downstream of the cutoff wall to monitor performance
and serve as pump-back wells, if necessary
 South and North Dams are lined with an HDPE geomembrane on the upstream face and
anchored to a concrete plinth founded on bedrock. A grout curtain is installed beneath the
plinth that extends to low permeability bedrock. Monitoring wells are installed
downstream of the South and North Dams
 Tailings deposited at 50% solids concentration, mainly from spigots located over the Main
Dam and discharges located along the impoundment perimeter, maintaining the reclaim
water pond at a distance from the Main and South Dams
 Dam performance is monitored mainly with piezometers and shape array inclinometers.
The dam is located relatively close to the mine pit, therefore use of waste rock for dam construction
is convenient and cost-efficient.
The design of the TSF was approved by the environmental authority and regulators, Dirección
General de Aguas (DGA) and Sevicio Nacional de Geología y Minería (Sernageomin). Within the
permit conditions, the regulator, Sernageomin, imposed that CCMC assign a permanent “Engineer
of Record (EoR)” during construction and operation, external to CCMC, “whose role is to ensure that
the approved design of the TSF is satisfied in all aspects” (translation from Spanish). CCMC assigned
this EoR role to the design engineer that also fulfilled the role of Engineer of Record (EoR) according
to Lundin Mining Corporation’s (LMC) corporate tailings management technical standard. Potential
deviations from the original design were approved by this EoR when appropriate. Within these roles,
this engineer performed construction quality assurance (CQA) duties by signing off and approving
construction works performed by contractors and the owner (CCMC) according to approved issued-
for-construction (IFC) drawings and specifications, documenting construction activities, reviewing
and collecting construction records in a timely manner so that they werr complete and representative
of as-built conditions.
In addition to these roles, the operation (CCMC) and corporate (LMC) requested independent CQA
audits. The purpose of these audits was to identify gaps and provide recommendations to CCMC
with respect to the construction quality assurance performed by the EoR. These audits focused on
the quality control and assurance; they did not include technical review of design changes. The
reviews of design changes were limited to spot checks on completeness and consistency, and whether
the design change was documented well enough to justify the change without modifying the design
intent approved by the regulators. The CQA auditor did not conduct independent peer or technical
reviews since this was considered the responsibility of the design engineer and EoR.
Within the context of this project, the main activities of the CQA auditor consisted of inspecting the
practices, organization, methods, procedures, documentation, resources and personnel capabilities
of the project quality assurance team, with the understanding that the responsibility for construction
quality assurance remained with the EoR.
Some additional project characteristics that enhanced the benefits of an independent audit:
 The engineer that completed the original design that was permitted was changed for the
detailed design, construction and operation phases. The new engineer endorsed the
permitted design but identified opportunities for improvement, prompting design changes
that would require proper justification with the regulator.
 CCMC, the owner and operating mine, self-performed a significant and critical amount of
the earthworks for dam construction. Waste rock from the mine was used for dam
construction and the mine performed all handling and compaction of the dam fill materials.
The incentives for mine production may conflict with the incentives for dam quality
construction.
These project characteristics are not uncommon; the independent CQA audit assisted in resolving
some of the gaps arising from the conflicts suggested above.
Finally, it is worth indicating that the requirement for the independent CQA audit was identified
early in construction and was supported from the beginning by all incumbents: project execution
team (LMC Capital Projects), CCMC mine operation (entity that would assume ownership of the
project after it was commissioned and LMC corporate (responsible for tailings governance and
stewardship). The EoR was also supportive of the audit from the beginning.

PROJECT ORGANIZATION
The design and construction of the starter dam and initial phases of the project was performed under
the owner with a one team approach focusing the operations and projects groups on one common
goal. Upon commissioning the project was transferred to the mill operation, CCMC. Tailings
operations and subsequent phases of construction are expected to be performed by CCMC.
The project organization during design and construction of these initial phases, until commissioning,
was led by the owner´s Project Manager under LMC Capital Projects. Quality control and assurance
was led by an owner´s team, under the Project Manager. The responsibilities of the owner´s quality
control and assurance team included the following:
 Implementation of a quality plan for the project, prepared by the EoR and approved by the
owner
 Conduct inspections and audits of the quality management systems of the contractors
 Inspect quality of construction materials and construction activities, according to standards
and approved specifications
 Testing, documenting and approving materials produced by the materials processing
plant, owned and operated by the project execution team (the materials processing plant
was subsequently transferred to CCMC after commissioning)
 Tracking resolution of construction non-conformance reports
 Administer turn over packages and transfer of these to the operation (CCMC)
Quality control was the responsibility of each contractor, including the owner that self-performed the
majority of the dam construction earthworks. Quality assurance was the responsibility of the EoR´s
on-site team. The owner´s construction quality control and assurance team administered the contract
of a laboratory to perform most of the earthworks quality control and assurance tests.
The EoR´s on-site team included a chief engineer, responsible for ensuring that the project was built
according to approved design, site supervisors and engineering and design support staff. The
laboratory administered by the owner reported technically through the EoR.
The EoR also held an off-site team, including a design engineer responsible for ensuring that the
designs met standards of practice and develop or check design changes while ensuring that these met
approved design intent. This design engineer also provided technical engineering analyses and
design support to the construction team.

CONSTRUCTION QUALITY ASSURANCE AUDIT PROCESS

The construction quality assurance audits consisted of the following main activities:
1. An initial activity consisting of an in-depth study and understanding of the design basis and
criteria (including understanding of reasoning behind relevant design features) and design
intent approved by the regulator and owner CCMC, including: sectorial permit for
construction issued by the regulator, construction quality assurance plan prepared by the
design engineer, and IFC drawings and specifications. The main purpose of this activity was
to define the basis for the audits. Deviations, lack of information or inconsistencies with
respect to this basis were identified as findings during the audits.
2. Develop understanding of the project organization, particularly with respect to roles and
responsibilities of the EoR, owner´s project management structure and owner´s quality
control and assurance staff. As the project progressed, the relevance of other roles emerged.
Specifically, the role of the person responsible for understanding permits associated with the
 new facility became relevant as the project headed into commissioning. The main purpose
of this activity was to identify gaps or conflicts in the project organization that could hinder
quality or the design intent.
3. Review of change management and non-conformance resolution processes. The main
purpose of this activity was to identify gaps or conflicts in the processes that could hinder
construction quality or lead to changes that did not meet design intent.
4. Develop understanding of the construction schedule. The main purpose of this activity was
to become aware of critical paths where construction quality issues, if not detected and
addressed in a timely manner, could result in impacts to schedule. The placement and
compaction of waste rock for dam fill (namely Material Zone 3 in the project), self-performed
by the owner, was identified early on as a critical path activity that required attention from a
quality assurance perspective.
5. Site visits and audits were held approximately every 3 months. The main purpose of the
audits was to detect findings related to quality assurance deficiencies and provide
recommendations to resolve each finding. The general agenda for each site visit included:
a. Review of actions and progress made by the project execution team, including the
EoR and contractors, to address findings of previous audits
b. Review of latest construction schedule
c. Presentation by site team of audits or recommendations made by the regulator
d. Site team presents their assessment of situations that could pose a threat to
construction quality
e. Inspection of the work fronts by the auditor together with the owner´s quality
control and assurance team as well as the EoR.
f. Initial and spot check inspections of the laboratory on site, including: review of test
procedures, documentation, qualifications of technicians, equipment calibration
certifications and quality controls.
g. Review of dam foundation preparation records
h. Auditor conducted spot checks of construction documentation, generally including:
i. Daily reports of the CQA team under the EoR
ii. Raw construction quality control data, such as: Lugeon tests performed on
natural and grouted bedrock, grout injection plots, etc
iii. List and status of non-conformances, including supporting information on
how non-conformances have or are being resolved.
iv. Completed and signed forms that certified that work had been completed
according to IFC specifications and drawings. These were checked for
completeness and consistency, and that they included or referenced
supporting information such as laboratory test results, surveys, material
certificates, etc.
v. Technical documentation prepared by the EoR to assess construction
quality data. This documentation was checked by the auditor for
completeness and consistency, so that the information and data was
supportive of the conclusions reached by the EoR in the document.
 vi. Technical documentation prepared to support design changes. The
information was checked technically only for outstanding flaws. As such,
the auditor did not conduct a detailed review of calculations and analyses
since the EoR is ultimately responsible for the results.
vii. Data obtained from geotechnical instrumentation installed during
construction.
viii. Drafts of as-built records and drawings as they were generated.
ix. Turn over packages from the contractors as they were submitted.
x. Construction records reports and certifications prepared by the construction
management team and the EoR. The main purpose of these documents was
to serve the requirement by the regulator to obtain the permit to begin
operating the commissioned facility
i. Auditor meets with site personnel, as needed, to clarify situations or request back
up information prior to developing the list of findings from the audit
j. Close-out meeting was held on site (immediately at the end of each site visit) where
the auditor presented a preliminary assessment and list of findings resulting from
the audit. Each finding was accompanied with a practical recommendation that the
project execution team, EoR, CCMC or the contractors could use as guidance to
develop action items to resolve each finding
k. Auditor reported the results of each audit. Each audit finding was explained and a
recommendation on how to address the finding was included. This
recommendation was provided as a guidance but not mandatory. Based on the
information provided during the site visit, previous audit findings were identified
in the report as: addressed (the finding was considered closed and would not appear
in the next audit report), not addressed (no action was taken to address the finding
and thus would be carried to the next audit), dropped (the finding was no longer
applicable and would not appear in the next audit report), in progress (the finding
was partially addressed and was therefore carried to the next audit report)

The construction quality assurance audits for Los Diques were conducted between January 2017 and
May 2018. Construction of the initial phases of the Los Diques TSF were completed in June 2018 and
tailings were being deposited within the impoundment at design throughput rate by September 2018.

CONTRIBUTIONS OF THE CONSTRUCTION QUALITY ASSURANCE AUDITS TO

THE SUCCESS OF THE PROJECT
Construction of the initial phases of the Los Diques TSF is considered a success. The facility was
constructed and commissioned on schedule and on budget while adhering to high technical and
quality standards defined by the design engineers and approved by the regulators and owner (LMC).
Permission from the regulator to begin operating this new Los Diques TSF was obtained on schedule.
Specific contributions of the CQA audits to this success are described as follows:
 Early identification of gaps and potential conflicts in the project organization that could
hinder quality assurance. Particularly, the independence of the EoR was reinforced by
allowing the EoR to reject a design change promoted by the project execution team
 Reinforced the role of the EoR to validate the quality of the results of the on-site laboratory,
administered by the owner´s project execution team.
 Recommended implementation of formal quality control processes and protocols for the
earthworks activities that were owner self-performed (OSP). As a result, the project
execution team requested a quality control supervisor within the OSP. Quality of
earthworks performed by the OSP was assured and documented, including supporting
information such as in-situ density tests, type and quality of waste rock used in dam
construction, control surveys, number of passes of compaction equipment, lift thicknesses,
etc. Adequate density control of the dam was demonstrated and remains documented
 Early review of dam foundation preparation records and approvals by the EoR, thus
ensuring that records were complete and consistent before fill placement.
 Early identification and closing of gaps in quality control testing of in-place material.
 Early identification of anomalies or singularities in quality control data as it was generated
in the field, and prior to formal reporting, that could potentially lead to a non-compliance,
deviation from the original design or questioning or refusal from the regulator.
Specifically, slightly higher hydraulic conductivities in specific locations of the natural
bedrock and within the grout curtains were detected in early stages of construction. These
singularities were addressed without delaying construction or commissioning of the
facility.
 Early identification of construction deficiencies. It is worth noting that construction
deficiencies were few and minor throughout the project, highlighting proactiveness of the
project quality assurance team.
 Waste rock placement and compaction tests for dam construction (test fills) were
conducted in advance to allow sufficient time for analysis, to establish acceptable quality
control parameters and to assist in planning of quality control procedures. Density control
of the dam is well supported with applicable test fills. The auditor assisted the EoR in the
interpretation of the compaction test results.
 Waste rock used in dam construction was classified according to the type of rock emanating
from the mine and specifically if it is acid generating or not. To date, no acid generating
material has been placed within the dam phases 0 and 1. The characteristics of the waste
rock material used in dam construction are well documented based on information
generated by the mine and specific geochemical testing.
 Timely review and recommendations provided to the design engineer to complement
analyses in support of increases of compacted waste rock lift thicknesses, that would allow
a faster rate of construction while maintaining construction quality and ensuring design
intent.
 Timely review and recommendations provided to the design engineer to complement
analyses in support of design changes. The complementary analyses were recommended
based on standards of practice and to demonstrate that the proposed design changes met
design intent.
 Early identification of proposed design changes that were not likely to meet design intent.
  Provided review and recommendations for the construction records reports and
certifications provided to the regulator to obtain permission to operate the facility.
 Facilitated connection between the project execution team, CCMC and LMC, by providing
awareness and common understanding of the standard required to be met according to
LMC corporate requirements.
 Helped facilitate transition from construction to operation based on experience of the
auditor in the commissioning of new tailings facilities and as an independent tailings
stewardship reviewer for LMC´s operations globally.

CQA AUDIT IMPROVEMENTS

After completion of the initial construction of the Los Diques project it was recognized that, even
though there were issues of concern and findings that were detected early in the construction
schedule, there were a few that were not given the proper attention or closed in a timely manner.
Accordingly, it is suggested that a rating of the findings detected and updated in each audit would
have helped in assigning importance and priorities to the running list of findings that had not been
closed. The authors suggest that each finding be assigned a rating (1 to 3 for example, with 1 being
of highest consequence or risk to the project and, thus, provide guidance to the project team to assign
attention and resources to close the finding immediately, and a rating of 3 indicating that attention
can be postponed for 3 months, until the next audit). These finding ratings should be assigned jointly
by the construction manager and the CQA auditor.

IMPLEMENTATION OF CONSTRUCTION QUALITY ASSURANCE AUDITS IN

OTHER PROJECTS
The success of the project and the contributions of the CQA audits to this success suggest that a
similar CQA audit process, as described above, is worth considering for other similar projects.
Based on results to date of the construction quality assurance performed on Los Diques, the authors
are of the view that a successful CQA audit program can be implemented in other tailings dam
construction projects by adopting the following main aspects of the audits implemented at Los
Diques:
1. Alignment and support of the CQA audit process and auditor by the project execution team,
operator (who will eventually own the facility delivered by the project execution team) and
corporate. Early buy-in and alignment with the entity responsible for delivering construction
quality assurance is also necessary.
2. Mandating and allowing in-depth understanding, by the auditor, of the design intent, basis
and criteria and of IFC documents.
3. Ensuring independence and experience of the auditor.
4. Allowing audits to be performed regularly and frequently, as required to anticipate and react
to audit findings in a timely manner. Findings should be prioritized so that they are
 addressed in a timely manner. Regular and frequent audits will also allow adjustments of
findings according to the dynamics of the dam construction process.
5. Audits should be viewed by project participants as a helpful tool to achieve project objectives,
specifically with respect to achieving the quality standard defined and approved by the
regulator and project owner.
6. Audit results need to be accompanied by practical recommendations to resolve each finding.
Recommendations need to be consistent, to the extent practical, with relevant construction
project goals such as schedule and budget.
7. Scope and exclusions for the audit need to be clearly defined and understood by project
participants.
8. Knowledge and understanding, by the auditor, of the owner´s corporate standards for
tailings management to ensure that standards are met, to facilitate connection among project
participants and to facilitate critical decisions that may result from the audit findings.

CLOSING
A construction quality assurance audit is shown to be an effective tailings governance tool since it
allows an independent level of inspection of the construction quality control and assurance processes,
identifying areas of deficiencies and opportunities for improvement that other entities, that are
involved in the day to day and that may have other interests over strict adherence to quality and the
design intent, may not visualize. Strict adherence to quality and the design intent during construction
adds elements of resilience to the TSF and is an additional step towards preventing catastrophic
failures.
ICMM has recently announced the development of an international standard for tailings facilities
which is expected to include, among other items, a “system for credible, independent reviews of
tailings facilities.” ICMM or regulators to not specifically identify construction quality assurance
audits but the authors are of the view that these audits performed by an independent tailings dam
construction expert can be considered to partially fulfill this new ICMM standard.
Independence of the auditor as well as audits being performed regularly and frequently, to allow
early detection of issues and implementation of recommendations in a timely manner, is key. Good
communication between the construction management team and the auditor, to jointly prioritize
findings is also important.
The authors propose that independent quality assurance audits may also be implemented during the
operation of the TSF by adding adherence to relevant operating parameters, defined by the design
engineer and approved by the regulator, to the scope of the audit.

REFERENCES

ICMM. (December, 2016). Position statement on preventing catastrophic failure of tailings storage
facilities.
ICMM. (February, 2019). ICMM commits to create an international standard for tailings dams. Press
release.
Approach and Considerations for OMS Manual
Development
Ivette De Jesús1and Paul Bedell2
1. Quebrada Blanca Fase 2, Teck Resources, Chile
2. Golder Associates, Canada

ABSTRACT
This paper describes the recommended approach and key considerations for the development of an
operation, maintenance, and surveillance (OMS) manual for tailings management facilities based on
the Mining Association of Canada’s (MAC) guidelines and a case study. The OMS manual constitutes
a foundational component for tailings management systems as recommended by recognized
international good practice. Early development of an OMS manual ensures timely compliance during
the operation as well as more focused training and onboarding for personnel in addition to having
operations align with the design intent of the tailings management facility.
Typically, a consultant, usually the engineer of record, leads the development of the OMS manual.
This, in turn, is then handed over the Owner for completion and use. This approach typically leads
to Owner´s personnel not internalizing the document or making proper use of it. The new revision
(2019) of the guideline for Developing an Operation, Maintenance and Surveillance Manual for
Tailings and Water Management Facilities, issued by the MAC, advocates for Owner-led
development; this approach was adopted as part of the operational readiness activities for the
Quebrada Blanca Fase 2 Project (QB2). This paper discusses recommendations for approach (such as
using Last planner® approach for the development planning) and key considerations to bear in mind
based on the lessons learned through the development of the QB2 OMS manual. A practical overview
of the OMS manual development for QB2 will serve as the case study discussed in this paper.

There is no full article associated with this abstract.

1
Engineer of Record Role and Experience in Chilean
Tailings Storage Facility
Francisco Vera, Marcelo Arenas, Alejandra Neira and Esteban Barría
Wood, Chile

ABSTRACT
A Tailings Storage Facility (TSF) is a system that is built, operated and managed on a daily basis and
therefore experiences change constantly. This scenario involves professionals with a focus on the
operation, design and safety of the structure, who sometimes interact, work in parallel or in separate
lines. This requires a professional able to integrate all the skills and corporate standards
(characteristic of the owner), national and international guides, as well as best available practices
(BAP) and best available techniques (BAT) observed in the different TSFs of Chile and abroad, with
the objective of certifying the TSF safety. It is important that the professional/the company is
independent of the designer, operator and owner of the facility.
In addition, the design, the operation and the construction of a TSF are sources of a large amount of
valuable information, which requires document management systems, that are not affected by
organizational changes, modifications to the design, and completion of growth phases, among others,
allowing to have detailed historical record.
In this context, the role of the Engineer of Record (EOR) arose from the need that in an operation as
large as a TSF, which presents a certain risk level to the community, the environment as well as the
continuity of the owner’s business, the EOR is required to preserve three key aspects for the TSF’s
safety: knowledge base, coordination or consistency and responsibility. An EOR must be present
during life cycle of TSF, from the conception of the design until the closure phase.
A description of the experience gained by Wood during the last several years in the implementation
of the Engineer of Record system on several TSFs in Chile is presented in this article, indicating
limitations, the identified critical variables, and the challenges presented by this new role in the
management, construction and operation of TSF.

1
1.0 INTRODUCTION

TSFs are works in constant design, construction and operation, the useful life of which is equivalent
to, and in many cases greater than, human timeframes. Due to these characteristics, TSFs require
constant design and operational adjustments that ensure the safety of the works and as well as their
performance throughout its life, including closure.
In this context, over the last decade, the position of Engineer or Record (EoR) has been incorporated
within the governance of TSFs in mining companies The EoR advises the owner on matters of design,
construction, operation and closure. However, in general, each particular contract defines the role
and responsibilities of the EoR in the operations, therefore there is currently an important discussion
to be had on this issue.
The objective of this article is to discuss the role of the EoR in TSFs developed by Wood, through the
revision of the guidelines, standards and current regulations of the EoR services, going into more
depth in the identified critical variables and the difficulties faces in the development of the services.

2.0 ROLE OF THE ENGINEER OF RECORD

The EoR is a concept applied in the USA and Canada for civil works construction. The EoR is a single
person who is responsible to ascertain that the final design (including any changes made during
construction) meets applicable design standards, criteria and guidelines. However, historically the
concept of EoR has not been utilised in the design and operation of TSFs, mainly because the concept
of the EOR is not completely compatible with works that are in constant growth could have changes
in its operational philosophy, designer and even owner over the course of the work’s useful life.
Following the tailing dam foundation failure occurred at the Mount Polley Mine in British Columbia,
Canada in 2014 and the failure of the tailing dam at the Samarco mine in 2015 (Brazil), mining
companies have made significant efforts to incorporate a standardized structure for tailings
management into their corporate management. This structure must be oriented towards complying
with environmental legislation and minimising risks inherent to design, construction and operation
of Tailings Storage Facilities.
Within this context, the Canadian Dam Association (CDA) introduced the Engineer of Record (EoR)
concept for tailings dam in 2014. This role corresponds to an experienced and competent Engineer
(or company) who is responsible for ensuring that TSFs are safe. Also, EoRs must ensure that design,
construction, operation and closure is in accordance with current legislation and the best engineering
practices.
The third version of A Guide to the Management of Tailing Facilities from the Mining Association of
Canada (MAC) (2) introduces the EoR concepts as follows:

2
“Engineer-of-Record: The Owner, in assuring that a tailings facility is safe, has the responsibility to
identify and retain an EoR, who provides technical direction on behalf of the Owner. The EoR verifies
whether the tailings facility (or components thereof) has been:
 Designed in accordance with performance objectives and indicators, applicable guidelines,
standards and legal requirements;
 Constructed, and is performing, throughout the life cycle, in accordance with the design
intent, performance objectives and indicators, applicable guidelines, standards and legal
requirements.
For tailings facilities that include retention structures/dams, the EoR is responsible for Dam Safety
Reviews and associated reports. The EoR should also participate in the facility’s risk assessments and
be available to Independent Reviewers.”
In 2018, the Geoprofesional Business Association (GBA)(2) published guidelines called Proposed Best
Practices for the Engineer of Record (EoR) for Tailings Dams incorporating recommendations from a
panel of expert professionals in geotechnical engineering to define the roles and responsibilities of
the EoR for TSFs focusing on improving the safety of the TSFs and their social-environmental
sustainability. This guide takes the definition of EoR proposed by MAC, cited previously, as a starting
point, incorporating the operational component:
 The TSF is performed and is operated during its life cycle in accordance with its design
intention, objectives and performance indicators, guidelines, standards and regulatory
requirements.
As well as expanding and complementing the definition of the EoR, the GBA guideline contributes
to the discussion of important principles which are not explicit in the CDA definition of the EoR, the
most important of which are the following:
 As far as possible, maintain an EoR throughout the design, construction and operative
phases, as well as the closure and post-closure phases. This provides an important benefit
for the owner of the operation, and it is common industry practice.
 The role of the EoR is more effective when it is supported by an experienced and committed
team. When it is adequately executed, the focus of the team provides long term
sustainability and a platform for continuous improvement.
Following the failure of Mount Polley, the government of the province of British Columbia (Canada),
established in 2017 in its Health, Safety and Reclamation code, the obligation to have an EoR
responsible for each TSF.
In Chile, the role of the EoR is not yet included in legislation, however, the most important mining
companies (national and international) are considering the EoR as a fundamental part of their
operation.

3
3.0 FOCUS OF THE ENGINEER OF RECORD SERVICE

The scope of the work related with EoR services varies, depending on the requirements of the clients,
their operation and governance. In general, three types of scopes can be identified:
1. EoR during the design, then during construction and in parallel during the operation.
2. EoR during the design and the construction, including the Construction Quality Assurance
(CQA).
3. EoR during the operation, which includes specialised assistance to the operation.
The main scopes for the EoR service during the operation and the specialised consulting are the
following:
 Review of existing information, designs, explorations, construction and operation records.
Identifying missing information and studies.
 Monthly inspections on the TSF, with the issuance of a report that includes a summary of
the findings and the analysis of monitoring and surveillance data.
 Revision of engineering designs.
 Warning the operation team in the event of problems related to poor operational practices,
escalating the alert in the measure of its seriousness.
 Carrying out extraordinary inspections to verify the status of the works after abnormal
events (earthquakes, extraordinary rain events, etc.).
 Supporting the operation/mining team in the preparation of technical documents such as;
reports, presentations or others for government, environmental and sectorial
organisations, etc.
 Participation in independent review panels, in interviews with the Dam Safety Review,
workshops related to the safety of the TSF, risk assessments, among others.
 Monitoring and verification of regulatory compliance and current authorisations in the
TSF.

4.0 KEY COMPONENTS OF THE OPERATION

During the development of the EoR service, key components of the operation have been identified,
which are as follows:
 Operation manuals. This document is priority for the adequate operation of the TSF, as it
corresponds to a log of the operation. As it is a dynamic document, in constant review and
updating, it must be clear and complete in accordance with the activities that are required
for the correct operation. Operation manual should permit that any operator can perform
the operational tasks of the TSF, even when they do not have knowledge of the history of
the TSF, as these tasks should be clearly described in this document. When there are
modifications to the functioning or management of the TSF this document should be
updated. Also, the manual must include all the operational controls, periodic supervisions,
records, etc.

4
  Emergency manual. As well as the previous document, this manual must be systematically
updated, as risks are identified, incorporating new controls that require actions in the event
of deviation from the established thresholds for the safe operation, similarly when new
conditions are identified, climate changes or changes in the governance of the TSF.
 Geotechnical instrumentation for the monitoring of the TSFs. For the management of all
the TSF engineering projects, geotechnical instrumentation that permits the monitoring at
the start, during the construction, during the operation and after the closure of the facility
should be installed. However, it is very frequent to observe instrumentation in poor
condition or that simply does not utilise its real potential for the operation, either due to an
insufficient frequency to obtain data, due to an individual rather than integral
instrumentation philosophy (consideration of the installed instrumentation set), or simply
due to a misunderstanding of the records. The monitoring instrumentation allows
assessment of the condition of the TSF, when addressed and analysed with time, allows
the identification and correction of potential failures or deviations from the design.
However, these systems require periodic maintenance, analysis of the variations observed
in the records, integration of the different measurement systems and understanding of the
performance when facing changes which are part of the operational phases that the TSFs
experience.
 Monitoring thresholds. Although this point can be associated with geotechnical
monitoring and even with the emergency and/or operations manuals, it is considered a
very important and a fundamental point for the safety of the TSF. Defined thresholds
permit a simple assessment of the conditions of the TSF by means of instrumentation,
manual or automated measurements or simply visual supervision. Furthermore,
thresholds permit confirmation of compliance with the considerations and criteria
established during the design of the TSFs, and rapidly identify deviations. In practice, it
has been observed that the thresholds defined in the design are normally maintained,
however, the thresholds must be reviewed, analysed, and updated in a regularly during
the operation.
 Tailings Deposition Plan. This is one of the key factors identified during the service, as it
sets out the expected growth to be measured and allows the projection of key parameters
for the operation and for the regulatory entities of these works, such as freeboard, pond
volume, tailings volume and mass, and seepage among others. The deposition plan should
be reviewed regularly to reflect the real TSF geometry. This is achieved by means of a
dynamic process in which the input variables are recalibrated over time.
 Bathymetric/topographies. Associated with the tailings deposition plan, the execution of
bathymetric studies allows determination of the condition of the TSF basin in for
comparison with its deposition plan, therefore it is a fundamental tool for operational
control and the identification of deviations from what was planned for the growth. It is
important to establish a review frequency which permits verification and/or actions to be
taken in time in case of events of non-compliance with the established plans.

In some of the EoR services, the update of critical documents or variables is included as part of the
EoR service or within the specialised assistance to the operation.

5
5.0 DIFFICULTIES IN THE IMPLEMENTATION OF EOR SERVICE

This section lists and briefly describes the main difficulties that have arisen during the development
of the EoR service in the operation in different TSFs:
 Deficient document management system. Due to the large quantity of documents
generated during the different stages of the life cycle of the TSFs, it is vital to have a system
which permits the studies generated to be stored in an orderly and clear manner. This
system must not be affected by, for example, changes in the organisational chart of the
personnel in charge of the operation/construction of the TSF or by changes in the contractor
companies that support the operation/construction.
 Misunderstanding of the service. Since the figure or role of the EoR is relatively new in
Chile, in general in the industry there is a misunderstanding of the objective and the scope
of the EOR’s service. This often generates confusion and a mismatch with the client’s
expectations.
 Discontinuity between design and construction stages. Frequently there is a lack of
document control with the records of the quality controls performed during the
construction and with the as-built drawings of the works. This generates uncertainties
related to the modifications to the design made during construction, generating potential
risks for the safety and operation of the TSFs.
 Rotation of personnel in charge of the operation. It has been observed that in general the
personnel in charge of the TSFs do not remain in their positions for long periods. This
personnel rotation means that the operators have poor specialisation, especially at sites
where the TSF operation teams are composed of few professionals. TSFs require many
disciplines which makes the communications between the personnel in charge of the TSFs
and the EoR more difficult.
 Responsibilities of the personnel in charge of the TSFs. Although progress has been
made in recent years, it is still observed in some tasks that the roles and responsibilities of
the professionals in charge have not been clearly defined. At the same time, the
professionals have not been empowered to make decisions on the management of the TSFs.
Sometimes, the problem is not the lack of clarity of responsibilities, but the number of
people involved in decision making. This slows down the response times needed to a
continuous and dynamic operation, making the decision processes bureaucratic.
 Low Operational Budget for TSFs. The recognition of the importance of TSFs has grown
recently. This is due to recent catastrophes that have affected these facilities which have
revealed the importance of control in all their stages. However, frequently they are still
considered the "garbage dumps" of mining operations, so any attempt to improve their
standard, or implement works that involves capital expenditure, attracts the lowest budget
priority of the owners of the mine. This situation has been changing, but it is still a difficulty
for the EoRs to achieve improvements in the standard of these facilities, since their
contributions are related to the security of the TSF and do not have direct monetary return.
 Deficient communications between different work areas. Deficient communication has
often been observed between the different operational and corporate areas of the TSF, or
between the project area and the area of operation. This makes the EoRs tasks difficult

6
 because they do not have all the information quickly, or their requirements may be
delegated to another area.

6.0 FUTURE CHALLENGES

A list of future challenges is presented below, which in the opinion of the consultant are relevant for
the adequate management of the EoR:
 Definition of the EoR in national regulations. Due to the fact that this role is already being
considered by the main mining companies in the country, it is advisable to regulate
definitions and responsibilities by means of national regulatory guidelines, which would
facilitate EOR implementation and understanding by both operators and collaborators in
the operation.
 Availability and control of predictive models. The EoR must be presented with all the
analyses carried out for the modelling of the deposit, such as tailings deposition sequence,
physical stability, management of water resources, among others. These models must be
reviewed periodically to identify their deviations with time as well as generate action plans
to correct them. For control, the EoR must have information on the operational variables of
the deposit, such as: topographic and bathymetric information, mass distribution of the
discharges, operational freeboards, management of water resources, etc.
 Knowledge of Future projects. The EoR must be in complete knowledge of all the projects
being developed that have repercussions in the general operation of the TSFs. Therefore, it
is recommended that meetings be held between the EoR and the operations and projects
teams of the TSFs.
 Availability of Instrumentation Records. As mentioned, given the vital importance of the
EoR in monitoring the performance of the TSF, it is necessary that the EoR always has
access to all the instrumentation records of the TSF. Considering the technological progress
and the tools available, the goal should be that these systems are monitored in real time.
An ambitious idea would be to develop a national integrated on-line system that monitors
all the TSFs of the country.
 Safety Culture. Given the magnitude of TSFs, the numbers of TSFs present in our country,
as well as the catastrophic events occurred lately, it is vital to create a transversal safety
culture across the TSF operation departments at a national level.

7.0 ACKNOWLEDGEMENTS

We would like to thank Wood and all of the consultants that assisted us in the development of this
publication and we hope that we will continue to develop the systems and the tools that allow us to
feel secure of these colossal works.

8.0 NOMENCLATURE
CDA Canadian Dam Association

7
CQA Construction Quality Assurance
EOR Engineer of Record
TSF Tailings Storage Facility
BAP Best Available Practices
BAT Best Available Techniques

9.0 REFERENCES

(1) Mining Association of Canada (2017) A Guide to the Management of Tailing Facilities, Third Edition,
https://mining.ca/sites/default/files/documents/MAC-Guide-to-the-Management-of-Tailings-Facilities-
2017.pdf

(2) Geoprofessional Business Association (2018) Proposed best practices for the engineer of record (EOR)
for tailings dams.

8
Management of Tailings Storage Facility Operational
Data: Lessons Learned
Esteban Barría, Gian Armijo, Marcelo Arriagada and Alejandra Neira
Wood, Chile

ABSTRACT
A Tailings Storage Facility (TSF) is a complex structure, and given their constant growth, is in
continuous evolution, varying both the internal (basin) and external configuration (dams and other
additional works). There are multiple factors that can influence the changes of a TSF, these include:
technical, economic, environmental, regulatory framework and commitment.
With the aim of maintaining records of the evolution throughout the life cycle of the facility, it is
necessary to establish early on the variables to be monitored, which depend on characteristics of each
TSF and the changes to the original design made during the operation. The development of this
process of defining, implementing, recording and analysing data is called Operational Data
Management.
Operational Data Management generates a complete record of the relevant parameters with useful
information for the performance analysis of the TSF, allowing to detect/move ahead critical situations
and deviations regarding correct operation.
This paper presents the main deficiencies detected by Wood during operational data controls of TSFs,
highlighting the lessons learned, with the aim to be applied to future projects.

1
INTRODUCTION

Tailings is the by-product of mining processes, it must be collected and stored into complex
engineered structures called a Tailings Storage Facility (TSF). The life cycle of the TSF consists of
project conception, planning, design, initial construction, operation and ongoing construction, close
and post-closure.
The last major failures occurred, such as Mount Polley Mine in British Columbia (2014), Samarco
Mina in Minas Gerais Brasil (2015) and the Brumadinho dam of Vale Mine in Minas Gerais Brasil
(2019) have increased the commitment of mining companies to the responsibly managed of TSF
including environmental and safety criteria.
The management of tailings facility includes the prevention of impacts to human health and safety,
the environment, and infrastructure. Mining companies are accountable and responsible for
managing tailings during the full life cycle.
This paper is focused on the operational data management during the operations and ongoing
construction phases of a TSF. The operational data recording and its analysis are useful to evaluate
the TSF performance and identify any deviation on time to mitigate the risks. The aim of this paper
is to present the main deficiencies identified by Wood during the operational data management.

MANAGEMENT OF TAILING STORAGE FACILITY

The purpose of a tailings management system is to provide a framework for the management of
tailings facilities, to assist owners in the development of systems that include environmental and
safety criteria, and to improve the consistency of application of reasonable and prudent engineering
and management principles to tailings facilities.
Responsible tailings management considers the prevention of impacts to human health and safety,
the environment, and infrastructure. Responsible management must be applied in a permanent
manner in all phases of the tailings facilities life cycle (see figure N°1). A comprehensive risk
assessment, both physical and chemical, must be carried out associated with the tailings facilities and
the implementation of the appropriate controls to effectively manage the risk.

2
 Policy and commitment

ACT PLAN

Phases of Tailings
Management Facility Life Cycle Planning
review for
continual Project Conception and
improvement Planning

Design

Initial Construction

Operations and ongoing

construction

Change in life of mine or

clousure plan

Temporary or permanent
closure
Performance Implementing the
evaluation tailings
management
Post-closure framework

CHECK DO

Figure 1 Elements of the tailings management framework (Source: The Mining Association of Canada,
October 2017)

As is presented in the Figure 1, appropriate tailings management system implementation will help
owners to achieve and demonstrate:
1. A commitment to develop, implement, review and maintain a tailings policy;
2. Objective setting for planning and strategic activities related to performance and risk
management of tailings facilities.
3. Continual improvement of a tailings management system.
4. The internal controls and procedure are appropriate and disclosure.
5. There is an appropriate governance to decision-making.
6. That verifiable, clearly defined and updated critical controls and procedures are in place
to manage risk.
This paper will present learned lessons of the operational controls monitored during the operations
and ongoing construction phase.

3
OPERATIONAL DATA OF TSF

The purpose of the operational data analysis is evaluating the performance of TSF through a
continuing control that allows to avoid and mitigate the risk of failure.

Critical operational data

Critical operational data are defined as the variables that can directly affect the safety of the TSF if
they surpass the established allowable limits (alert and alarm levels). For each of these variables there
must be defined alert and alarm thresholds. Complementary, thresholds should trigger action plans
applicable to emergency situations, which should be incorporated into the Emergency Preparedness
and Response Plans (EPRPs).
Recorded operational data must be analyzed and compared with the different aspects of tailings
management, such as; the tailings deposition plan, water balance model, physical stability models,
deformations analysis, among others.
The operational data control will be analyzed in two areas: Tailings Impoundment and Tailings Dam.

Tailings Impoundment
The main operational controls in the tailings impoundment are to verify the tailings deposition plan
and the water management.
For the Tailings Deposition Plan, the following data must be recorded:
1. Tailings mass per discharge point.
2. Tailings production rate.
3. Beach and Under water deposition slopes.
4. Concentration in weight of tailings pulp.
For water management inside the impoundment, the following data must be recorded:
1. Records of decanting pond in parameters as: level, volume, area and depth in the zone of
water reclaim system.
2. Hydraulic and operational freeboard.
3. Records from water reclaim system.
4. Available volume of water, between the spillway level and the water level pond, to consider
volume for flood regulation and for dilution. Additionally, spillway level must be under the
contact between the tailings and dam and over the tailings level at the spillway.

Tailings Dam

In the Tailings Dam the geometry and the materials characteristic must be measured, recorded and
analyzed.
The geometry of the dam is measured and controlled through the following data:

4
  Topography, with registration of main dimensions: wall crest level, global and local slope,
 Production and placement rates of materials (cyclone tailings or mine waste),
 Deformation measurements,
 Piezometric levels.
The materials characteristics of the dam are measured and controlled through the followings data:
 Strength properties of the materials,
 Compaction and granulometry tests of each of the layers placed in the dam,
 Potentially Acid Generating (PAG).

Control and monitoring of critical operational data

Control and monitoring of critical operational data must be clearly described in the Operation,
Maintenance and Surveillance Manual (OMS Manual) as well as the procedures related with those
activities. These documents should define the measurement frequency, alert and alarm levels, the
tracking monitoring records and analysis of the data.
The team involved in the control and monitoring of the operational data must know the aim of each
measurement and how it must be made.
On the other hand, the methodology proposed for the control and monitoring must provide a flow
of information that allows an effective evaluation of the TSF global behavior, in order to detect
anomalous situations on time. This with the purpose of implement actions to mitigate the risks in the
TSF.
The controls can be regular or exceptional, the first are described in the OMS Manual and the second
are defined in the Emergency Preparedness and Response Plan (EPRPs).

Regular Control

Routine controls of the operation data should be established in each TSF, these data should be
analyzed periodically and compared with the design criteria and models.
For example, geotechnical instrumentation data should be analyzed monthly with respect to defined
thresholds from a stability analyzes of the TSF. For this, the construction and operations information
of each instrument is required.

Exceptional Control

The TSF must have an inspection plan for exceptional events which should consider a review of the
performance of the critical operational data during and after the event. Some events that fall into this
category are: (i) seismic events of great magnitude that require an immediate revision of the
geotechnical instrumentation; (ii) increase of flows in water reclaim systems; and (iii) increase in pore
pressure levels.

5
Historical Records

It should be established a structure that will be in charge of maintaining the historical records of the
operational data. In this way the records will remain in time regardless of the changes that may occur
in the staff.
A historical record will allow a better analysis of the different situations that may occur during the
cycle of life of TSF, providing a valuable point of comparison.
The historical record must keep a complete registry of each operational data, as well as information
on installation protocols in the case of instrumentation.

LESSONS LEARNED

Wood has analyzed the operational data in all the projects in which has involves as Engineer of
Record of the TSF. Based on this experience, common problems have been identified in several
projects related to the management of operational data.
Table 1 presents a description of the problems identified, the findings, the impact on operations and
the corrective and preventive actions proposal by the authors.
The main deficiencies detected during operational data management of TSF are resumed in the
following 4 lessons learned:
 Lack of definition and management of the governance (see items 1 and 3 in table 1).
 Lack of an appropriate document control system (see item 8 in table 1).
 Deficiency in definition communication channels and training programs (see items 2, 6, 9
and 10 in table 1).
 Deficient external and internal audit programs (see items 4 and 5 in table 1).

6
Table N°1 Description of the problems identified in operational data management ½
Ítem Problems Findings Impact in the Operations Data Management Corrective and Preventive Actions
In some cases it have not been defined the Delay in activities related to the management of
Designate a responsible person of the operations
of roles, responsibilities and operational data
Definition of roles, data management.
accountabilities of the professionals Increase of time in making decisions and
1 responsibilities and The responsible person must have accountability
involved in the TFS operations, or its are uncertainty in solving problems
accountabilities and level authority in order to make decisions
defined but it have not been understanding Slow response in situations of high priority or
related with the operations data management.
clearly by the team. urgency
Delay in the transmission of information
Its is nos clear as and who must be informed of
relevant information, thus generating loss of
information Define a clear communications process which
Definition of
Lack of clear definition of internal Delay in communication of internal and external established the internal and external
2 communication
communication channels information, this related to recommendations and / communications channels.
channels
or solutions that can avoid situations of risk or
emergencies
Slow response in situations of high priority or
urgency
The measurements and recording of
operations data are not reviewed, with the
Problem to analyze the records when those are
aim to identify mistakes or deficiencies in Define a Responsible Person(s) of operations data
Procedure for deficiencies
the procedures. control and its reviewing and identifying of
3 operations data Delays in deadlines for operational data analysis
The team involved in operations data corrective actions when it be necessary.
reviewing Lack of registration and/or corrective actions when
control are not awareness of the procedures Monthly reviews of operations data.
altered result is identified.
and the importance of this process in the
risk assessment.

In some TFS it have not been carried out Carry out an annual internal audits plan
The lack of audits does not allow a continuous
Insufficiente of internal audits, or it have carried out Follow up and verify compliance of corrective
4 improvement in tailings management or a learning
internal audit without sufficient dissemination or actions
team involved
following up of the findings Dissemination of findings and corrective measures

Carry out an annual internal audits plan

The lack of audits does not allow a continuous
Insufficiente of In some TFS it have not been carried out Follow up and verify compliance of corrective
5 improvement in tailings management or a learning
external audit external audits. actions
team and companies involved
Dissemination of findings and corrective measures

7
Table N°1 Description of the problems identified in operational data management 2/2
Ítem Problems Findings Impact in the Operations Data Management Corrective and Preventive Actions
Define and communicate the threshold levels for
When there are not a methodology to identify
Definitions of criticals control such as water level recording by the
Alert and alarm thresholds for the when operations data are over the values allowed
6 threshold for critical piezometers
operations data control are not defined (alert and alarm). the TFS do not have a tracking of
control Include the threshold levels in the OMS Manual
corrective actions to avoid and minimized risk.
and EPRPs.

Some TFS do not have enough reviewes or Operations control describes in OMS Manual are
updated of the Operation, Maintenance outdated according the ongoing
Update of OMS It is recommended that the OMS Manual and
7 and Surveillance manual (OMS) and operations/construction. Therefore, the critical
manual and EPRPs EPRPs should be updated every 1 year
Emergency Preparedness and Response control and its performance criteria defined in the
Plans (EPRPs). document could have change.

Difficulty to have access to document of Create and implement a system of Document

There are not a Document Control process constructions such as TOP folder, drawings as-built, Control clear from the beginning of the project in
or it is deficient. etc. order to ensure the recording during all the cycle of
Deficient Document
8 Difficulty to find or have access to records Loss of important information (document, life.
Control
related to the TFS operations, constructions, drawings, protocols etc.) for the TFS safe Implement a procedure of Document Control to
designs, etc. management. ensure that document information will be created,
Obsolete version have not been updated. maintained, retained and archived.

The team involved in the operation have

not training in the use of the procedures Operations data erroneous, then loss of time and Transmit procedures and training to those are
Comunication of related with operations data control. rework. involved in different activities within the operation
9
the Procedure Therefore, they have not awareness of the It is more difficult to identify problems during the Generate a culture of correct execution of tasks in
importance of operational controls, critical operations. the operations team.
controls and risk management.
A training program must be developed.
Professionals involved in the operation of
General knowledge The training should included risk management,
TFS have not knowledge about the Inadequate operational data control, then a
10 of the operation of critical control, OMS Manual and EPRPs in order to
objective of operational data control and delayed in identifications of risk.
the TFS understand the importance of operations data
associated risks that these may indicate.
control to the TFS safety.
Geotechnical instrumentation could fail,
then it is necessary make a reviewing of its Recording of illogical instrumentation data could Implement a quality assurance system during the
Review of the
status in order to descartar erroneous produce a erroneous analysis. installation of instrumentations.
11 instrumentation
measurement. Difficulty of analyzing operational data, due to lack Develop a periodic reviewing of the
status
Installation protocols are not complete with of information in the records. instrumentation
all the information or not archived.

8
NOMENCLATURE

TSF Tailings Storage Facility

EPRPs Emergency Preparedness and Response Plans
OMS Manual Operation, Maintenance and Surveillance Manual
EoR Engineer of Record
PAG Potentially Acid Generating

REFERENCES

The Mining Association of Canada, (October 2017), A Guide to the Management of Tailings Facilities,
Third Edition.

9
A Risk-Based Approach to Support the
Implementation of the New Guidelines Related to
Tailings Management
Michel Julien1, Pascal Lavoie1, Édouard Masengo1 and Thomas Lépine1
Agnico Eagle Mines, Canada

ABSTRACT
Agnico Eagle Mines Limited (AEM) is a Canadian gold producing company that operates 9 mines in
Canada, Mexico and Finland, and is also managing a series of closed sites mainly in Canada. The
mines produce tailings in different forms: conventional slurry, thickened tailings and filtered tailings.
AEM also operates complex infrastructures somewhat analogous in terms of risks like Heap Leach
Facilities, Rockfill Storage Facilities and Water Management Infrastructures. AEM has decided to
apply the same rigor in the management and design, the construction, the maintenance and
monitoring of all these infrastructures, and this throughout the whole life-cycle of these facilities. A
key part of this journey toward self-improvement in the management of these infrastructures was to
adopt some robust internal systems and to follow Best Available/Applicable Practices.
Implementation of the Mining Association of Canada (MAC) Towards Sustainable Mining® (TSM®)
3rd Edition (2017) is considered such a best practice. The changes of the latest 3rd Edition require a
certain strengthening of the governance of the management of its tailings facilities. Among the
different changes is the need to have more clarity on roles and responsibilities, the inclusion of a more
formal risk-based approach, etc. This paper presents some of the adjustments needed to adjust the
implementation of the Guide to the unique conditions of the company, its organizational structure,
the internal practice and the interpretation of the roles.

1
INTRODUCTION

Agnico Eagle Mines Limited (AEM) is a Canadian gold producing company that operates 9 mines in
Canada, Mexico and Finland, and is also managing a series of closed sites mainly in Canada. The
company operates mines under very diverse regulations, climates and environment. The mines are
also different by their orebody, operating conditions and mineralogy. The mines produce tailings
under different forms: conventional slurry, thickened tailings and filtered tailings. These tailings are
stored in Tailings Storage Facilities (TSF) of different types and design. Some have been designed a
few years ago, while others have been evolving over several decades. AEM also operates other
complex infrastructures somewhat analogous in terms of risks like Heap Leach Facilities (HLF),
Waste Rock Storage Facilities (WRSF) and Water Management Infrastructures (WMI). AEM has
decided to treat them in an analogous manner. While, the following discussion focuses on TSF, it
should be assumed it applies as well to HLF, WRSF and WMI.
Agnico Eagle Mines Limited is an active member of the Mining Association of Canada (MAC). MAC
presented its 3rd Edition of the Guide to the Management of Tailings Facilities in 2017 (MAC, 2017).
The first Edition of MAC’s Tailings Guide (the Guide) was released in 1998. In 2004, MAC introduced
the Towards Sustainable Mining® (TSM®), and the Guide was integrated in the TSM as the Tailings
Management Protocol through a series of performance indicators. A guide for the preparation of
Operation, Maintenance and Surveillance Manuals (OMS) was also developed in parallel (MAC,
2011) and updated in 2019. Application of TSM is mandatory for MAC members for their operations
in Canada. It involves 3 interesting characteristics: performance assessment by site; the disclosure of
the results and the involvement of a Community of Interest's panel (COI) involving a multi-
stakeholder group to review the process. The 3rd Edition followed the 2014 tailings dam failure at
the Mount Polley Mine in British Columbia. The 3rd Edition retains the second Edition’s strong
emphasis on management systems. However, it provides more clarity on roles and responsibilities
and risk management. In particular, it involves:
 A solid and flexible governance model with respect to tailings management. Specifically, it
provides more clarity on roles and responsibilities by identifying a certain number of
functions with their specific responsibilities:

2
 o The Owner (The Owner’s Board of Directors or Governance Level);
o The Accountable Executive Officer (AEO);
o The Responsible Person (RP);
o The Engineer of Record (EoR);
o The Independent Reviewer (IR); and,
o The Design Engineer (DE).
 A strong emphasis on Risk-Based Approach (RBA);
 A focus on Best Available Technology (BAT) and Best Available/Applicable Practices (BAP);
and,
 The identification of critical controls and their implementation as performance monitoring
indicators.

One the most significant contribution of the Guide remains its recognition of the need to adapt itself
to a broad range of organizational structures, mining activities, regulatory environments, etc. To
achieve this adaptability, the Guide has therefore an intrinsic part of flexibility for the key definitions
allowing a certain modulation of their application to respect the unique characteristics of the different
organizations and sites.
The successful implementation of such management system over the full life cycle to final closure of
projects rests on the close cooperation and interaction between all the different stakeholders involved
at all phases of the project. This is particularly true for TSF and similar infrastructures since their
design, construction and operation can spread over several decades. This close cooperation and
interaction should occur as part of the normal operation (e.g. the careful tailings deposition planning
and annual water balance analysis) but also during more difficult circumstances when deviations
from the design intent and anomalies during operation are observed.
Consequently, AEM embarked in a process to integrate the Guide intent and requirements to its own
reality, and, is still going through this journey. This paper presents some parts of the approach being
followed to implement the recommendations of the Guide in order to adapt it to its uniqueness. In
no means, this paper is intended to be a recommended path forward for other companies or
organizations. Each of them will have to find its own way.

THE GOVERNANCE

A fundamental part of the changes the Guide brings, is to provide more clarity on the whole
discussion of governance associated with tailings management, and, other mine waste and water
management infrastructures; referred as high-risk infrastructures. A significant shift can be observed
in the way this issue has been tackled in the last 5 to 10 years. Traditionally, the mining industry has
treated these activities as peripheral (not a core business), typically deferred to external consultants
or contractors. As mentioned by Julien (2016 A, B), this activity is now viewed as a core activity of
the mining industry to the same extent as excavating rock and processing it are. Many recent
examples have shown this activity can have a huge impact on an organization in case of performance

3
issues or major problems. This realization has forced some companies to assert more explicitly their
ownership in the outcomes of this activity by increasing in-house capabilities, having more
involvement in the design activities, and, by improving the linkage with the production; essentially
to act as an Owner. One the fundamental contribution of the MAC Guide has therefore been to
provide more clarity on the governance aspects associated with tailings management in order to
support this function of ownership. In 2016, the International Council on Mining and Metals (ICMM,
2016), produced the Tailings Management Guidelines and Recommendations for Improvement
Report focused on the key aspects of good practice. The report looked at a range of existing guidance
in various countries. Their report concluded that «The Canadian guidelines (MAC and CDA) when
taken together represent the most comprehensive of the national frameworks. Member companies
that adopt the Canadian guidelines guidance would be rated as adequately complying with good
practice».
As emphasized by MAC (2017), ultimately, the accountability for decisions related to tailings
management rests with the Owner’s Board of Directors or Governance Level. The Board of Directors
or Governance Level is expected to designate an Accountable Executive Officer (AEO) for tailings
management. The AEO needs to be aware of the key outcomes of tailings facilities risk assessments
and how these risks are managed; has accountability and responsibility for putting in place an
appropriate management structure; provides assurance to the Owner and its communities of interest
that tailings facilities are managed responsibly.
Therefore, it is important, and as recognized by the Guide, at the planning phase of a new TSF or
analogous infrastructures, that the Owner is made aware of the design process and possible disposal
options. It should also be informed of the potential risks and consequences in the event of non-
performance, and the all-inclusive capital and operating costs for each options to allow the
opportunity to proceed after careful evaluation to the best judgement choice. A TSF management
system should then be tailored to the chosen process and tailings disposal options. The required
specifics of the management system will have therefore to be adapted for the various possible options
and the understanding of potential risk and consequences that would be associated with each option.

RISK-BASED APPROACH

The 3rd Edition of the Tailings Guide introduces explicitly the notion of risk-based approach as an
important building block of having a well-functioning tailings management system. This area is
particularly sensitive and should be looked at very carefully to make it truly meaningful. In order to
embark on a meaningful risk-based approach, a higher level of sophistication is needed beyond
classical risk analysis approach (IEC 2009) when dealing with high consequence infrastructures, and
in particular if their design, construction and operation may have spanned over several decades.
Classical risk analysis tools (where risk is expressed as a product of the probability of occurrence
times the consequence to provide an associated coding for the level of risk) are found to be
particularly challenging for tailings management systems where the frequency of occurrences can be

4
low to very low and consequences high to very high. This aspect has been pointed out in several
publications (Caldwell et al, 2015 and Oboni et al, 2016).
At the end of the day, it is important to remember that these tools serve two main purposes: a mean
to separate cases so the focus can be on priorities, and, as a communication and awareness tool.
Following several discussions, it was felt that classical risk analysis tools can make the messaging
more complex for this type of risks given their low probability and their high consequences. These
methods are nevertheless useful in a first evaluation but something more robust or easily
communicable is needed, particularly with senior management and operators who may not have the
same technical background to appreciate the subtleties of this field.
The implementation of a proper management system during actual mine TSF operation, using a risk-
based approach needs to be first considered not at the mine operation stage but much earlier, at the
outset of evaluation and selection of an appropriate design and construction methodology for the
TSF. Specifically, several TSF design options should be evaluated and determined if suitable for the
site climatic and seismic conditions, tailings grind and variations of over time due to ore changes,
daily tailings tonnage to be processed, surrounding environmental situation and the use of the
territory. The options should be carefully evaluated from a corporate risk tolerance standpoint. A
robust low risk option would not likely be the most economical, whereas the most economical option
may put pressure on the execution level given its exposure to inherent uncertainties.
Beyond the issue of the challenges of risk evaluation for these high consequence events, the additional
challenge with a classical risk-based approach comes from the very strong influence of highly
technical matters and the link with the often-complex life cycle of these facilities where they can be
progressively designed and built while being operated. With respect to the technical aspects, the
discussion was found to be recently strongly dominated by complex considerations associated with
static liquefaction, post-peak and post-liquefaction behaviour, residual strength parameters, tailings
dam classification, dam breach analysis, etc., making the communication of these risks even more
complicated.
As an example, the discussion on dam breach analysis is particularly interesting because it rests on a
knowledge base strongly influenced by some empirical relationships developed from actual totally
different cases that may not be linked with the specific actual site conditions. The outcome of these
analyses, if taken out of context, may generate inaccurate, if not, emotional responses. Some of the
publications in this field (e.g. Rico et al, 2006) are good examples. It is thus important that these
discussions be brought back to actual sites conditions and rest on similar field-case examples. The
discussion on static liquefaction or liquefaction under seismic loadings (Been, 2016) is another area
that occupies a lot of the discussion space and is extremely challenging and even troubling for non-
specialists.
Another challenge with the implementation of a risk-based approach is the linkage that tends to occur
between design criteria and the consequences analysis. The Dam Classification (Table 2-1, CDA 2013)
is a good example to illustrate this challenge. On one hand, there is merit to adjust the design criteria

5
to the risk level, and, to have the industry converging toward a common state of practice. The
challenge then becomes the Owner’s tolerance to the risk, and in particular to the consequences of a
non-performance of these facilities. For example, if a company was to classify a structure based on
the CDA Dam Classification, most likely it would classify it as a Very High or Extreme (e.g, from an
Owner perspective 1, 10, 100 fatalities are similar and are all unacceptable). Specifically, this table
has been found to be difficult in application and even potentially misleading in the development of
a foundation for a risk-based approach. In the author’s opinion, design criteria development should
rest more on the standard practice rather than some subjective classification of some metrics assessing
potential consequences. In parallel, the Owner should go through its own risk tolerance exercise
along the lines of Caldwell et al (2015) and Oboni et al (2016) to clearly identify its own risk tolerance
level.
The usage of simple frameworks to illustrate these risks is therefore warranted to raise awareness
and to focus on priorities. In this context, it was decided to push forward the internal usage of a
simple tool along the lines of Silva et al (2008). Figure 1 is particularly enlightening by showing some
relationships between Factor of Safety (FoS) and Annual Probability of Failure (APF) for different
categories of projects that have been back-analyzed. These relationships are based on actual
engineering projects and developed through quantified expert judgement. In our opinion, such
relationships are powerful tools to communicate actual risks in a language that non-experts can
understand. Besides broadening the discussion on the notion of FoS by including the notions of
uncertainty (Herzal et al, 2017), such approach introduces the notion of the interaction between
design, construction and operation, so fundamental in mine waste management. Such an approach
has the potential to become a valuable communication tool for the associated risks of mining
infrastructures. It also serves as a mean to prioritize mitigation works. For example, using such an
approach, an infrastructure with higher FoS but higher APF may be considered in priority for
mitigation works and considered as requiring more stringent monitoring as compared to an
infrastructure with lower FoS but lower APF. This aspect is also discussed in Herzal et al (2017) in
the context of the FoS and its inherent notion of uncertainty traditionally built into it.

6
 Figure 1 Factor of safety versus annual probability of failure (after Silva, Lambe and Marr, 2008)

The methodology presented in Silva, Lambe and Marr (2008) introduces the notion of Level of
Engineering and link it to the notion of consequences. The authors are therefore linking the Level of
Engineering to the type of facility or its consequences: Best Level of Engineering to facilities with
high failure consequences; Above Average to ordinary facilities; Average to unimportant or
temporary facilities with low failure consequences; and, Poor to little or no engineering. While useful,
this linkage can be considered as debatable when applied to mining projects.
Further, the methodology was developed for embankment in soils and focuses strongly on
engineering aspects. The rating is obtained by assessing the Level of Engineering during the design
(investigation, testing and analysis), construction and operation and monitoring to obtain an overall
rating (from 1 to 4; 1 being the best and 4 the worst) and then to associate a certain FoS to an APF (see
Fig. 1). The integration in the assessment of the performance of a facility in a more holistic way was
found particularly attractive in the mine waste management practice, since performance is often
linked to design but also to construction and operation. This is particularly true if a facility is
designed and constructed while operated like it is often the case.
In reviewing the original methodology, it was then felt that it could be extended by generalizing it
by including the notion of management systems while respecting its original intent and conclusions.
This generalization rests on 4 considerations:

7
 1. If a project has been done using a strong management system, its outcome is most likely
susceptible to converge toward a higher level of engineering;
2. If a project is done in a collaborative way (e.g. good and continuous interactions between the
DE, RP, EoR and IR), the same outcome could possibly be achieved;
3. If a project is subject to a formal and thorough review process, the likelihood of identifying
problematic technical issues is higher, and, the project will converge naturally toward a
higher level of engineering;
4. Finally, if a project like those in mine waste management is allowed to integrate a retroactive
process at all stages of its life cycle and use the information obtained from the other stages,
the project will also converge toward a higher level of engineering.
A modified Table 1 inspired of Silva and al (2008) was therefore developed using the same
fundamental considerations but aimed to express the Level of Engineering as a more general notion:
Level of Practice for Design-Construction-Operation (DCO) based on the current state of practice and
for the whole life cycle. The intent of this classification is to assess infrastructures for which the
design, construction and operation may span over many decades, and, therefore with a continuously
evolving state of practice with the eyes of today’s practice. For example, an infrastructure may have
been designed, constructed and operated with a state of practice that was totally acceptable 30 or 40
years ago but if looked through our current practice of today could be found to exhibit a weaker state
of practice. One of the main challenges facing the industry is this continuously evolving state of
practice. It was therefore decided to assess the different infrastructures with today’s eyes and
disregard if they were or not designed, constructed or operated with a high level of practice in the
past.
A High Level of Practice for DCO would therefore require that throughout the whole life cycle of an
infrastructure, the same level of practice, as applied today, would have been applied, and, it would
have to be of a high level on all aspects. A Good Level of Practice for DCO would require that
throughout most of the life cycle of an infrastructure, a high level of practice has been applied but
not necessarily at all time, or, its level of practice was not of the highest level on all aspects at all time.
Similarly, an Average Level for DCO would suggest that an infrastructure that may have undergone
a variable level of practice over the years or would show some clear gaps in terms of the practice as
it is today. Finally, a Low Level for DCO would be characterized by clear gaps in terms of the practice
as it is today.
The main benefit of such an approach is to look at infrastructures more objectively and accept they
have often undergone in many cases a varying state of practice over the years. This aspect is
particularly relevant for older infrastructures that may have been designed, built and operated with
limited information in the past. Despite the fact that they may be managed with today highest level
of practice, the gaps of the past remain and need to be addressed. This methodology was used at 2
of our sites and allowed us to rapidly identify and justify focus areas.
While the integration of the design, construction and operation is a major contribution of the original
methodology, it was felt its scoring system was too biased toward design for the weight given to the

8
different components: the design accounted for 60% (investigation: 20%, testing: 20%, analysis and
documentation: 20%); while construction (20%) and operation and maintenance (20%) had a
somewhat lower weight given their importance. In particular, it was felt that it diminished the
importance of operation and maintenance so critical in mining applications. This aspect is even more
relevant when a facility is constructed and operated in parallel like in the case of upstream raises
construction. Nevertheless, to protect the overall conclusions of the method, it was decided to lower
the weight of testing from 20% to 10% and to increase operation and monitoring to 30%. Actually,
the trend in the practice is more and more on relying on actual field measurements like those obtained
from the CPTus, SCPTus and Field Vane Tests (FVTs), which allows to easily build a robust and
reliable database for the foundations and the stored tailings while still supporting it with laboratory
analysis. Therefore, this adjustment was considered quite reasonable. It is clear for other
infrastructures where CPTus, SCPTus and FVT cannot be used, such as RSF, alternate relevant
assessment methods, such as geophysics, may be considered.
As mentioned, Table 1 presents the different descriptors and generalizes them by putting emphasis
on the importance on management systems, collaboration, review process, a more integrated design
and other best practices. It was therefore suggested, while respecting the intent of the authors in
reconstructing their Table 1, there was merit for our internal intent to actualize it by taking into
account the context of the current practice in the mine waste management. The analysis for a certain
site has allowed us to identify infrastructures with higher values of APF and to selectively develop
appropriate targeted mitigation measures (Note: the Level of Practice for DCO = score of
investigation + score of testing + score of analysis and documentation + score of construction + score
of operation and maintenance). It is important to highlight that this methodology only applies for
static factor of safety and for conditions where peak strength parameters are used. This assumption
is considered reasonable in the context that the knowledge base and experience for these conditions
are actually solid and well established. Furthermore, it is believed that based on recent experience
that there is merit to further expand the discussion space on more classical analysis using static
conditions and peak strength parameters. Too much debate is currently occurring on complex failure
mechanisms dominated by empirical or very approximate, if not unreliable strength parameters or
methodologies.
While any risk-based approach includes a part that may be considered subjective or dependent on
the persons applying it, the proposed approach is found to be providing quite realistic assessment in
the context of built-in redundancy with the review and collaborative process such a management
system promotes. Specifically, if such a process can be integrated in a thorough on-going formal
review process, both internal and external, it will naturally reduce the risk of bias.
Finally, it should be remembered that the ultimate intent of such a method is primarily internal and
to find a realistic range of APF for an infrastructure to express its criticality, rather than a
deterministic number like the FOS. Furthermore, using APF is seen a much more palatable concept
to non-experts than deterministic and sometimes misleading notion like the factor of safety. This
rationale is ultimately to help the Owner to exercise its ownership. It is doubtful that such method

9
could be used for permitting given the state of the practice and how guidelines are currently written
in many jurisdictions which are strongly driven by a deterministic view of the factor of safety.

10
 Table 1 Level of Practice for Design-Construction-Operation (DCO) based on current state of practice for the whole life cycle

Level of Practice Design Construction Operation and monitoring

for Design-
Construction- Investigation Testing Analysis and documentation
Operation (DCO)

I - Level of 1.1 Baseline and background information 2.1 Laboratory test 3.1 Site selection study: Very detailed, consultative 4.1 Project management: in place 5.1 Governance: Owner with strong and
about site: Very detailed program development: and reviewed (all life cycle) and robust to allow works to clear governance for the facility (clear
practice for
High collaboration and proceed (for all life cycle) roles and responsibilities) for all life
DCO: High 1.2 Investigation program development: 3.2 Design Basis Memorandum: (DBM)
review through all life cycle
Highly connected with analysis, development : Detailed, up to date with state of 4.2 Supervision program:
cycle practice, collaborative and reviewed (for all life Dedicated qualified personnel on 5.2 Management system (e.g. TSM):
construction and operation phases with
cycle) a full-time basis with clear roles implemented, audited, results showing
high collaboration and review (for all life 2.2 Laboratory test results
and responsibilities (for all life strong commitment for all life cycle
cycle) supporting investigation 3.3 Analysis: Extensive, detailed, involving
cycle)
and analysis: Complete different failure modes with sensitivity analysis, 5.3 Operation, Maintenance and
1.3 Soil profile, stratigraphy, and geology:
and support in situ use knowledge obtained from investigation, 4.3 QA/QC program: detailed and Surveillance Manual: up to date, audited,
High level of certainty of foundations and
construction and operation phases, and, reviewed. agreed testing program and implemented
assessments and analysis
site conditions (all life cycle)
through all life cycle 4.4 Monitoring during 5.4 Monitoring: program and system in
1.4 Strength characteristics of foundations: 3.4 Risk analysis: detailed risk analysis to identify construction: On-going to adapt place with data acquisition system and
2.3 QA/QC program: Solid
High level of certainty (particularly for key risks and mitigations. Strong focus on failure construction rate to observed proper real-time analysis and follow-up
program to verify
weak and soft zones) and relying heavily mode analysis and determination of possible conditions
quality/validity of results 5.5 Disclosure: On-going and
on in situ information consequences.
4.5 Construction reports: As-built communication on performance of
(for all life cycle)
1.5 Tailings behaviour: Very detailed 3.5 Consultation: Proper and regular consultation reports and drawings to capture facility
2.4 Tailings: Detailed process with different stakeholders the history of the construction (for
information on in situ characteristics and 5.6 Expansion plan: In place to allow
characterization (physical all life cycle)
strength 3.6 Review process: Solid and formalized internal proper planning several years in advance
and chemical)
and external process (for all life cycle) 4.6 Construction meetings:
1.6 Groundwater conditions and 5.7 Performance review to calibrate
2.5 Geochemistry: Detailed Frequent and involving Owner,
piezometry: High level of certainty 3.7 Hydrological and hydrogeological analysis: original models and adjust it : Periodic or
characterization on all designer, Engineer of Record and
through time based on robust and reliable Done to evaluate short- and long-term performance before any phase of construction
materials Contractor. Design changes
field data of facility
agreed by all parties. 5.8 Emergency preparedness program: In
2.6 Reporting: Very 3.8 Costs: Detailed costs estimates updated place and tested
1.7 hydrology and climate: Highly robust 4.7 Review process: Formal
complete and reviewed - progressively
models based on field measurements and internal and external process (for 5.9 Formal review process (internal and
same quality over all life
calibration 3.9 Integration of other project phases to adapt all life cycle) external): Operational, collaborative and
cycle analysis: Done on a systematic basis through all open
1.8 Construction materials characteristics 4.8 Designer: Highly involved in
life cycle
(physical and chemical) and quantities: the construction works (for all life 5.10 Risk-based approach: Risk registry
Solid information to assess costs within 3.10 Information management: Very robust. cycle) up to date and includes consequences
Information and history protected analysis of failure modes
15%

0.2 0.1 0.2 0.2 0.3

11
II - Level of 1.1 Detailed but may include certain 2.1 Evidences of 3.1 Detailed, consultative and reviewed (for 4.1 Project management in 5.1 Owner with clarity on roles and
practice for uncertainties collaboration and review most of life cycle) place and good coordination, or responsibilities, or satisfactory for
through most of life cycle satisfactory for most of life most of life cycle
DCO: Good 1.2 Done with some linkage with analysis, 3.2 Good document, up to date, collaborative
cycle
construction and operation phase, good 2.2 Good support to and reviewed. 5.2 Implemented and audited for most
collaboration and reviewed (for most of analysis and to in situ 4.2 Supervision personnel of life cycle
3.3 Detailed, involving many failure modes,
the life cycle) measurements through present on site on quasi-
good retroaction from investigation, 5.3 Up to date, audited, and
most of life cycle continuous basis, or
1.3 Good to high level of certainty of construction and operation, and reviewed (for implemented
satisfactory for most of life
foundations and site - some gaps may exist 2.3 Good program for most most of life cycle)
cycle 5.4 Program and system in place and
of life cycle
1.4 Good level of certainty (particularly for 3.4 Risk analysis completed to identify key follow-up on a non-continuous basis
4.3 Solid program in place.
weak and soft zones) and relying on in situ 2.4 Adequate risks and mitigations. Failure modes analysis
5.5 Done beyond compliance
information - some gaps may exist characterization program done. 4.4 Monitoring on at least
requirements
for both physical and intermittent basis.
1.5 Reliable information on in situ 3.5 Good consultation process with different
chemical properties but 5.6 Long term vision developed but
characteristics and strength stakeholders 4.5 Reporting focusing on as-
may present some gaps remain high level
built drawings and reports, or
1.6 Good level of certainty and based on 3.6 Internal and external review process in
2.5 Good characterization considered satisfactory for most 5.7 Not done systematically
robust field program through time place and working well (for most of life cycle)
on key materials of life cycle 5.8 Program in place but not tested
1.7 Models based on adequate field 3.7 Done to evaluate short- and long-term
2.6 Good quality and 4.6 Meetings on site periodic 5.9 Operational, collaborative and
measurements and calibration performance of facility.
reviewed for most of life and design changes not fully open
1.8 Reliable information to assess cycle 3.8 Good costs estimate using solid communicated and reviewed.
construction costs within 30% assumptions but may lack details. 5.10 Risk registry up to date and a
4.7 Internal and external living document
3.9 Done on a frequent basis (for most of life process in place satisfactory for
cycle) most of life cycle
3.10 Good to solid information management 4.8 Involved in the construction
system. works on a part-time basis.

0.4 0.2 0.4 0.4 0.6

12
III - Level of 1.1 Considered incomplete to understand 2.1 Some collaboration and 3.1 Limited in scope and focus on compliance 4.1 Limited project 5.1 Roles and responsibilities not fully
practice for fully the site limited review, or requirements or satisfactory for only part of life management in place or defined, or satisfactory for only part of
satisfactory for only parts cycle satisfactory for only part of life life cycle
DCO: Average 1.2 Done with limited linkage with other
of life cycle cycle
phases but limited collaboration and 3.2 Document set at beginning of the process, 5.2 In the process of implementation,
limited review, or, satisfactory for only 2.2 Support primarily based on compliance requirements, limited 4.2 Supervision personnel or, satisfactory for only part of life
part of life cycle) analysis or satisfactory for collaboration and review. present on site on periodic cycle
only part of life cycle basis, or satisfactory for only
1.3 Average to good level of certainty of 3.3 Fairly complete but is focusing on meeting 5.3 Exist but is not up to date and not
part of life cycle
foundations and site (may include 2.3 Average to good compliance requirements, limited retroaction fully followed
uncertainties in parts of life cycle) program or satisfactory for from investigation, construction and operation, 4.3 Minimum program in place.
5.4 Program and system exist but is
only part of life cycle and limited review
1.4 Average to good level of certainty 4.4 Monitoring on occasional lacking follow-up
(particularly for weak and soft zones) and 2.4 Limited 3.4 Risk analysis completed to identify key basis.
5.5 Done to meet compliance
relying on limited in situ information characterization program risks and mitigations. Failure modes analysis
4.5 Reporting limited to site requirements
for both physical and done.
1.5 Information on in situ characteristics memos, or satisfactory for only
chemical properties 5.6 Long term vision exists but strategy
and strength considered limited 3.5 Done in the intent of meeting compliance part of life cycle
mainly reactive with little flexibility
2.5 Some characterization requirements.
1.6 Approximate information based on 4.6 Meetings rare and design
on some materials 5.7 Done when needed
limited field measurements through time 3.6 Review process limited in scope. changes not communicated and
2.6 Average to good reviewed. 5.8 Program exists but not up to date
1.7 Moderately robust models based on 3.7 Focus to meet compliance requirements.
quality and limited and not tested
limited measurement and relying on 4.7 Minimal internal and
review, or satisfactory for 3.8 Approximate costs estimate.
information from stations away from site external process in place, or 5.9 Exist but no evidence that it is done
only part of life cycle 3.9 Done occasionally, or satisfactory for only in a collaborative way
satisfactory for only part of life
1.8 Approximate information to assess
part of life cycle cycle
construction costs within 50% 5.10 High level risk analysis
3.10 Information management system 4.8 Involvement limited.
moderately organized.

0.6 0.3 0.6 0.6 0.9

13
IV - Level of 1.1 Limited and incomplete or containing 2.1 no evidence of 3.1 Pre-defined site, no- consultation and no 4.1 Limited project 5.1 Lack of clarity on roles and
practice for significant gaps consultation and review review. management structure responsibilities

DCO: Low 1.2 Limited, non-collaborative and not 2.2 Results limited and not 3.2 Limited information on design criteria and 4.2 Limited supervision on site 5.2 No formal system in place
connected to other phases validated focus is on compliance.
4.3 Limited QA/QC 5.3 No clear OMS in place
1.3 Low level of certainty of foundations 2.3 Weak QA/QC program 3.3 Limited in scope and intended to meet information collected
5.4 Limited monitoring program in
and site (for most of the life cycle) compliance requirements.
2.4 Not done and rely on 4.4 Weak to non-existent place
1.4 Low level of certainty particularly for experience 3.4 Very high-level risk assessment.
4.5 Limited information, report 5.5 Minimum disclosure to meet
weak zones
2.5 Limited geochemical 3.5 Limited consultation to extent of and document control compliance
1.5 No in situ information. Characteristics analysis done compliance requirements.
4.6 Limited meetings and 5.6 No formal expansion plan
and strength assumed from literature and
2.6 Reporting: Incomplete 3.6 Weak review process. reviews
experience 5.7 Limited to compliance
and not reviewed
3.7 Limited analysis. 4.7 Limited to non-existent requirements
1.6 Low level of certainty through time
3.8 Approximate cost estimates. 4.8 Not involved 5.8 Not available
1.7 Models from derived from stations
from site, literature and not calibrated 3.9 No retro-action with other project phases. 5.9 Weak or non-existent

1.8 Limited knowledge of borrow sources 3.10 Weak information management. 5.10 Limited in scope and general
and quality - high uncertainty on
construction costs

0.8 0.4 0.8 0.8 1.2

14
EXAMPLES

To show the proposed approach, two hypothetical cases are considered:

 Dam A at site X: The original Dam A has been constructed 30 years ago. It was constructed
using limited geotechnical information (3 boreholes) on silty clay/clayey silt foundations
characterized with 6 FVTs. The Dam was raised at least 6 times sometimes with tight QA/QC
and sometimes not. The estimated static factor of safety is 1.6 based on available information
and it meets with the regulatory requirements and recommendations of the state of practice
(CDA guidelines). Some monitoring instruments have been installed. One of the
inclinometers has been showing small but constant movement in the silty clay/clayey silt
layer over the last few years. An Independent Review Board (IRB) has been put in place 2
years ago. The tailings management governance of the site is now robust but this site was
acquired some years ago and it is clear that over the years the Level of Practice for the DCO
has varied substantially. Estimated level of practice: 2.5. Estimated Annual Probability of
Failure of about 5.0*10-4.
 Dam B site Y: The Dam B at site Y has been constructed 5 years ago with an acceptable (at
that time) static factor of safety of 1.4. While Dam B may have a lower factor of safety, its
foundation has been extensively investigated using CPTus, SCPTus, FVTs, and conventional
geotechnical drilling. The investigation has allowed identifying a layer of silty clay/clayey
silt fairly narrow and thick at one location. Systematic investigations using several
methodologies have been conducted to characterize this weak zone in the foundations with
a fairly high degree of certainty. Samples were collected and sent to the laboratory for further
testing. The design was done considering the available data set obtained from the extensive
geotechnical criteria and in a conservative manner. At the time of the construction, Dam B
construction was monitored very closely. The DE was closely involved during the
construction phase. The construction has benefited from solid documentation presented to
an IRB periodically. Given the uncertainties with the foundations, close monitoring has been
put in place and a drainage system has been installed to promote consolidation over time.
Estimated level of practice: 1.5. Estimated Annual Probability of Failure of about 1.0*10-4.

It is interesting to note that with such an analysis, Dam A of site X has a higher annual probability of
failure (5 times higher) than Dam B of site Y despite have a having a higher factor of safety. Using
such a tool, it becomes easier to build a case that Dam A needs more attention that Dam B through a
tighter monitoring program and possible additional mitigation measures. With such a simple tool,
it is possible to envision to evaluate the entire dam’s portfolio and focus the company’s actions on
priorities. It is interesting to mention that such analysis does not take into account explicitly possible
consequences of a failure. This aspect can be evaluated at the phase of the development of the
monitoring and possible mitigation measures.

15
CONCLUSION

The introduction of the 3rd Edition of the Tailings Guide of the Mining Association of Canada is seen
as an important step toward the continuation of the application of best practices in the tailings, mine
waste and water management. The Guide was written in the context of diverse companies and
organizations having their own unique organizational structures, resources and systems in place, and
therefore, its implementation requires some adjustments and adaptability. The Guide is very helpful
in clarifying governance aspects linked to tailings management, and in particular, the different roles
and responsibilities by providing more consistency.
One important aspect of the revised Guide is the formal introduction of a risk-based approach. As
mentioned, there is a clear need to use methodologies that can facilitate the communication of risks
associated with the non–performance, low probability of failure and high consequences events such
as tailings dam failures. This paper presents an approach derived of a similar approach developed
by Silva, Lambe and Marr (2008) that integrates more explicitly management systems. Such a method
is seen as a powerful tool to support best practice. Using this method, a dam with a higher static
factor of safety but a Level of Practice for Design, Construction and Operation (DCO) considered
lower than another dam with a lower static factor of safety, could be prioritized for more extensive
monitoring and mitigation measures. The importance of pushing to the next level the risk-based
approach in the context of the challenges associated with the description and communication of these
low probability and high consequence risks by the introduction of a risk-based approach integrating
the different phases of the life cycle on these high-risk infrastructures is seen as attractive, especially
in its ability to facilitate internal communication with non-experts. It is believed the integration of
such a tool into the management system will allow risks to be approached in a more holistic and
realistic manner which will ultimately contribute in reducing the long-term risks of these facilities.

REFERENCES
Caldwell J., Oboni F., Oboni C. (2015) Tailings Facility Failures in 2014 and an Update on Failure
Statistics, Tailings and Mine Waste 2015, Vancouver, Canada, October 25-28.

Julien, M. (2016 A) Presentation: Management requirements for water and mineral wastes. State of
Practice in Water and Mineral Waste Management Workshop Environmental and Social
Responsibility Society (ESRS) and Tailings Working Group (MAC) 2016 CIM Convention, May 1,
2016, Vancouver Convention Centre.

Julien, M. (2016 B) Presentation: Governance aspects with mine waste management facilities. World
Mining Congress, Rio de Janeiro, Brazil, October 2016.

Oboni, F., Oboni, C. (2016) A systemic look at tailings dams failure process, Tailings and Mine Waste
2016, Keystone, Colorado, USA, October 2-5, 2016

16
Morgenstern, N.R. (2018) Geotechnical risk, regulation, and public policy. The Sixth Victor de Mello
Lecture. Soils and Rocks, São Paulo, 41(2): 107-129 May-August, 2018.

Julien, M., E. Masengo, P. Lavoie, T. Lepine (2019) An operational perspective in the implementation
of the new guidelines related to tailings management. International Conference on Large Dam
(ICOLD 2019), Ottawa, Canada.

Duncan, M. (2000) Factors of safety and reliability in geotechnical engineering. J. of Geotechnical and
Geoenvironmental Engineering, ASCE, April 2000, 307-316.

Rico M., Benito G. and Díez-Herrero A. (2006) Floods from tailings dam failures. J. of Hazardous
Materials 154 (2008) 79–87.

Silva, F., Lambe W.T. and Marr W.A. (2008) Probability and risk of slope failure. J. of Geotechnical
and Geoenvironmental Engineering, ASCE 134 (12) 1691-1699.

Been, K. (2016) Characterizing Mine Tailings for Geotechnical Design. Australian Geomechanics. Vol.
51, No.4. December 2016.

Herzal J., Ashley, M. and Thorp J. (2017) Factor of Safety? - Do we use it correctly? ANCOLD 2017.

Australian Government, Department of resources, Energy and Tourism, (2008) Leading practice
sustainable development program for the mining industry. Risk assessment and management. ISBN
0 642 72599 3, Canberra, Australia.

IEC (2009) Risk management – Risk assessment techniques. STANDARD IEC/ISO 31010 Edition 1.0
2009-11 31010

CDA (2013) Dam Safety Guidelines. Table 2-1. Dam Classification.

Mining Association of Canada (AMC) (2011), Developing an Operation, Maintenance and

Surveillance Manual for Tailings and Water Management Facilities (the OMS Guide).

ICMM (2016) Tailings Management Guidelines and Recommendations for Improvement Report of
the International Council on Mining and Metals.

GBA (2016) Proposed Best Practices for the Engineer of Record (EoR) for Tailings Dams. GBA Tailings
EOR Task Force document.

Mining Association of Canada (MAC) (2017) A Guide to the Management of Tailings Facilities third
Edition

Alberta Chamber of Resources (2017) Roles and Accountabilities for Dam Safety Management in
Alberta. Dam Integrity Advisory Committee, Engineer of Record Subcommittee.

17
Los Diques Tailings Storage Facility: Construction
Quality Assurance
David Muñoz1 and Víctor Soto2
1. Minera Candelaria, Lundin Mining, Chile
2. Wood, Chile

ABSTRACT
Recent catastrophic failures of tailings dams highlight the relevance of a proper governance
framework for tailings storage facilities (TSF’s) to ensure the implementation and verification of
controls, in order to minimize risk over the full life of the facility. In this regard, construction is
recognized as a progressive process during the life cycle of most TSF’s, requiring robust management
of controls to assure tailings dam safety.
This article presents the construction quality assurance (CQA) developed during the first phase of
Los Diques TSF1, focusing on the role of the Engineer of Record onsite (EoR onsite) as a part of the
governance framework established by the Owner. It also addresses its relevance for the effective CQA
of an Owner self-perform (OSP) tailings rockfill dam, using mine waste rock directly from the
Candelaria open pit.
In effect, the OSP construction allowed to increase the efficiency and to reduce the footprint of the
project. In addition, the role of the EoR onsite described in this article, proved to be a key part of the
success of this construction scheme, fulfilling the approved design requirements and permit
commitments related to the tailings dam.
Los Diques downstream constructed rockfill dam design considers a maximum final height of 160 m
(first construction phase of 60 m height) and a total tailings storage capacity of approximately 600 Mt
(up to 75 ktpd of whole tailings deposited with a solid concentration by weight about 50%).

1Los Diques is the new TSF incorporated to the Candelaria Copper Mining Complex, located near
Copiapó in northern Chile

1
INTRODUCTION

The Candelaria Copper Mining Complex comprises two adjacent copper mining operations,
Candelaria (CCMC) and Ojos del Salado (CCMO), that produce copper concentrates from an open pit
and three underground mines. The complex is owned by Lundin Mining (80%) and Sumitomo (20%)
with a current life-of-mine (LOM) extending until 2040. The whole complex is located near Copiapó
in the Atacama Region, northern Chile, at about 900 m a.s.l. The copper ore is processed by a
concentrator plant located at CCMC, with a capacity of 75 ktpd and the Pedro Aguirre Cerda (PAC)
flotation concentrator plant with a capacity of 3.8 ktpd,
The CCMC Tailings Storage Facility (TSF), operating since 1994, is about to complete its storage
capacity during the second quarter of 2019, being replaced by the new Los Diques TSF, designed for
a total tailings storage capacity of approximately 600 Mt. Figure 1 presents a plan view of the project.

Candelaria
N
TSF

North
Waste dump

Candelaria
Open Pit

Concentrator
Plant

Nantoco
Waste dump

New TSF
~1 km Los Diques

Figure 1 CCMC plan view

The environmental impact study for the Los Diques TSF was approved in 2015 by the Chilean
environmental authorities. Key sectorial permits for Los Diques including those from the Servicio
Nacional de Geología y Minería (Sernageomin, the mining regulator) and the Dirección General de Aguas
(DGA, the water stewardship regulator) were granted in 2016 allowing the tailings dam construction
activities to commence.

2
Permit conditions included the Sernageomin's requirement of assigning an external permanent
Engineer of Record (EoR) during construction and operation for ensuring that the approved design
of the TSF is fulfilled, following the best practices for the mining industry.
The EoR role was implemented comprehensively in the project, within the Governance framework
established by CCMC, which follows the Owner’s corporate guidelines. This framework is aligned
with the global mining industry commitments about tailings dam safety during the whole life cycle
(International Council of Mining & Metals, ICMM, 2016).
This article describes the Construction Quality Assurance (CQA) for the TSF, focusing on the EoR
role that includes onsite support to the CQA within the broad view of the CCMC Governance
Framework. It also focuses on the benefits of this implementation considering an Owner Self-Perform
(OSP) construction of the tailings dam, using waste mine rock transported and placed at the dam by
layers, from the open pit by the mining trucks of the operation.

TSF DESIGN AND CONSTRUCTION – OWNER SELF-PERFORM TAILINGS DAM

Los Diques TSF is designed to handle approximately 600 Mt of tailings storage capacity with a
throughput up to 75 ktpd of whole tailings deposited with a solid concentration by weight of about
50%. The total tailings impoundment final area is about 6 km2. Tailings discharges are located both
on the dam itself and at the perimeter to allow the water pond to be kept away from the main dam.
The tailings dam is a compacted rockfill dam, founded on competent alluvial soils and rock,
constructed in stages and following the downstream construction method (final maximum height
160 m and total rockfill mass of 70 Mt). Figure 2 presents the final projected dam configuration.
The rockfill material is mine waste rock from the open pit, being high strength rocks. The tailings
dam is divided into 3 sections (North Dam, Main Dam and South Dam). The upstream face of the
rockfill dam includes engineered granular filter materials and a HDPE geomembrane with specific
configuration for each dam section. Specifically, North and South sections include grout curtains
connected to the HDPE geomembrane through a plinth at the upstream dam trace. The dam has a
robust basal drainage system that discharges to a general seepage collection system (SCS)
downstream of the dam, that includes further grout curtain with a drainage system for water
recovery using pumps in wells. Monitoring wells for hydrogeological monitoring are included.
Dam performance has been continuously monitored since the first construction stages, by
conventional automated geotechnical instrumentation which includes fiber optics piezometers,
shape array inclinometers, accelerometers and regular topographic monitoring, using drones and
radar for survey and mapping.

3
 0m 500 m

Figure 2 Plan view of the projected Tailings Dam (Final) – Los Diques TSF

The first dam construction stage developed (with the CQA previously described) considers a
maximum height of 60 m. This stage included hauling mine waste rock from the open pit between 65
ktpd (average) and 150 ktpd (maximum throughput) for the OSP tailings rockfill dam.
OSP tailings dam construction considers mining trucks from the mine for transporting and placing
mine waste rocks in compacted layers at the dam, instead of being transported to conventional waste
rock dumps. It also allows for increasing the efficiency of the operation by means of using the
installed capacity of the operation and reducing the footprint.
Footprint reduction is possible as the ongoing hauling and mine waste management activities are
being taken advantage of, being redirected for dam construction, with no requirements of additional
actions for dam buildup. Moreover, this strategy has reduced hauling distances, as the TSF is located
closer to the mine than most of the existing waste dumps.
OSP tailings dam construction considers CCMC mining team basically acting as a major earthworks
contractor, managing massive excavations and fills; and all extraction, handling and compaction for
rockfill construction, involving a high standard engineering structure.
Works by the Owner (CCMC) included managing an aggregate plant, assembled within the Los
Diques TSF Basin, operated to produce and supply the upstream dam face engineered granular filter
materials required. The equipment used by the mining team as a major earthwork contractor includes
bulldozers, articulated trucks, excavators, backhoe, payloaders, graders and compactors, among
others.

4
Therefore, the aforementioned change of activities, from waste dump generation to tailings dam
construction, including processing engineered materials (filter materials), has brought about a
significant improvement in mining operation. Other contributing factors have been the team’s
experience including the EoR onsite activities, good communication, detailed planning and team
work, As a result, no items are left unattended and all risks are being identified, keeping in mind the
final objectives of Operational Continuity for Candelaria, in particular:
 Incident Free (Zero Harm)
 Meeting Permits requirements and commitments to the Authorities
 Full Compliance of Project Specifications & Intention of Design
 Keeping On Schedule & Under Budget

These challenges were faced with a governance framework providing the management and
responsibilities structure for a successfully achievement of these objectives – including the EoR onsite
for enhancing CQA in this context.

GOVERNANCE FRAMEWORK – THE ROLE OF THE EOR ONSITE

Based on extensive discussions with professionals and regulators in the mining industry, the
Canadian Dam Association (CDA) Mining Dams Committee prepared in December 2016 a revision
to Section 3.1.2 of Technical Bulletin ‘Application of Dam Safety Guidelines to Mining Dams’, to cover
“Key Positions Related to Dam Safety”, including excellent definition of the role of the Owner and
EoR.
Following the CDA aforementioned bulletin “The Owner is ultimately responsible for the safety of the
dam” and as a part of their responsibilities described we found “ Understanding the risks created by the
dam; Implementing a management system framework, including policies, plans, procedures and audits
necessary for complying with the requirements of the applicable dam safety statutes, regulations and good
practice; Communicating effectively with the Engineer of Record regarding potential changes to the dam,
operations, maintenance, surveillance or instrumentation (This is important so the Engineer of Record can
confirm that a tailings storage facility or dam has been designed, constructed and is being operated in
accordance with the design intent, applicable guidelines, standards and regulations.)".
Respect to the role of the EoR, we can find the following “The Engineer of Record is defined herein as the
professional engineer responsible for assuring that the dam is safe, in that it is designed, constructed, and
operated in accordance with the current state of practice and applicable regulations, statutes, guidelines, codes,
and standards. The Engineer of Record is an integral part of the risk management system for mining dams, and
provides design continuity and ongoing technical support to the Owner with respect to dam safety issues over
the life of the dam.”
The latest revision (draft) of Section 3.1 of the Mining Dams Bulletin (CDA, October 2018), also
provides relevant guidelines about these roles into a Dam Safety Management System Framework,
aligned with the Management of Tailings Facilities Guide, The Mining Association of Canada (MAC),
2019; which is cited later in this article.
CCMC is committed to TSF safety and the tailings dam construction is under an effective
management system framework, which relies on a well-established Governance Framework, where

5
the role of EoR is fulfilled and incorporated intensively in the CQA of the TSF, with emphasis on the
OSP activity of dam construction with the role of the EoR onsite (part of the general EoR
organization).
The general Governance Framework for the CQA and CQC (Construction Quality Control) is
described schematically in Figure 3.
EoR onsite is an agent of the general EoR, fully agreed with a risk-based approach (MAC, 2019), in
which the identification of potential risks at the project is well understood and updated periodically
throughout the life cycle, in particular during construction, and well managed within the governance
framework and effectively communicated to the operators (OSP) onsite.
Following the Management of Tailings Facilities Guide, MAC, 2019, the EoR “verifies whether the
tailings facility (or components thereof) has been: 1) Designed in accordance with performance objectives and
indicators, applicable guidelines, standards and regulatory requirements; and 2) Constructed, and is
performing, throughout the life cycle, in accordance with the design intent, performance objectives and
indicators, applicable guidelines, standards and regulatory requirements”. So, the EoR onsite role
implemented, intent to merge the design and construction activities.

Figure 3 Governance framework for the CQA and CQC – Los Diques TSF

CCMC assigned this role of EoR to Amec Foster Wheeler, a Wood company. Wood has been deeply
involved in both the design and improvements to the original design of the project, which have
attained clearance from the Chilean Authority. The EoR team has performed on site construction
quality assurance duties (EoR onsite) by signing off and approving construction works, performed
both by contractors and the Owner as Owner Self perform (OSP), according to approved issued-for-
construction drawings and specifications. This synergy and the EoR onsite proved to benefit the
CQA, especially regarding the dynamism that the OSP construction required.

6
The EoR has also documented construction activities, reviewing and collecting construction records
in a timely manner, to assure it was complete and representative of as-built conditions. This EoR
scope of work included an off-site design Engineer responsible for ensuring the designs met
standards of practice, verifying design changes while ensuring that they meet approved design
intent. The design Engineer also provided technical engineering analyses and design support for the
construction team.
To ensure quality construction CCMC has its own Quality Assurance/Quality Construction (QAQC)
Team. This team reports to the QAQC Superintendent and its objective is to ensure full compliance
of all Projects procedures, technical specifications, quality plan, and inspection plan. In particular,
the team is formed by the following professionals: QA Manager, QA Electromechanical Supervisor
(A/B shift), QA Civil and Geotechnical Supervisor (A/B shift), and a TOP/Punch List Coordinator.
Under the QAQC team command is the EoR team, which also provides quality assurance support.
The EoR team is formed by the following professionals EoR Manager, EoR administration staff, two
senior Geotechnical specialists (A/B shift), two Civil Coordinators (A/B shift), a Geologist, an
Environment/Permits Engineer, QA Civil Supervisors (A/B Day/Night shifts), QA Geotechnical
Supervisors (A/B Day/Night shifts), Draftsmen, and a senior Geologist and Hydro Geologist on
demand.
The QAQC team also leads a laboratory and a surveyor’s team. The laboratory works as a CCMC
external subcontract, and its staff includes: two class B laboratory Technicians (A/B Day/Night shifts),
6 class C laboratory Technicians (A/B Day/Night shifts), 17 laboratory Assistants (A/B Day/Night
shifts), a Civil field Inspector (A/B shift), a Mechanical/Piping field inspector (A/B shift), and
laboratory administration staff. Lastly, the surveyor’s team is formed by: an Administrator, a
Surveyor, two Assistants and a Draftsman.
The role of external independent audits for design and CQA and CQC is led by BGC Ingeniería
Limitada (BGC), providing valuable third-party review and audit field visits. Other external audits
are performed bythe Owner Corporate Independent Tailings Stewardship Review (coordinated by
Lundin Mining) and by the regulator (Sernageomin) by means of unannounced periodic audit visits
(a total of 7 times).

CONSTRUCTION QUALITY ASSURANCE OF THE OSP TAILINGS DAM

As part of activities:
 Field Inspections (24/7): this task includes coordination with contractor Supervisor and
laboratory staff for daily sampling and testing requirements, keeping of construction
records and QAQC activities in various time periods, contractor audits and supervision of
QC activities, and submittal of No Conformity Reports (NCR, detailed below).
 Office work: this task centers mainly in document control and includes Turn Over Packages
(TOP) administration, technical workshops, KOM with contractors to define QAQC scope

7
 and documentation, verification and validation of Red Line Drawings, compiling,
distribution and answer to NCR’s, and coordination and response to all technical external
audits.
 Field laboratory administration: this activity includes directing and coordinating sample
collection and field and laboratory testing, as well as coordination of field technical
inspection (Civil and Mechanical).
 Crew surveying field inspection: directing and coordinating field surveying activities and
inspections.
 Continuous communication with general EoR during the construction activities.

Aside from the quality assurance activities described above, the EoR onsite performs its own CQA
activities, providing support for the QAQC team. These activities are: on site and test supervision
24/7, test and construction protocols revision, test result acquirement and analysis, and NCR
submittal, as well as any other QA service CCMC request. Through all these actions the QAQC team
and the EoR are capable to keep a constant and on time control for the TSF Los Diques construction
activities, detecting any deviation and taking the necessary steps to assure construction quality. The
main tool to undertake this task is shown in the NCR described in Figure 4.

Figure 4 NCR flow chart – Los Diques TSF

The QAQC team implement audits on all contractors, according to the Audit Program previously
presented to them. The frequency varies between one per month to one every two months, depending
on contract duration and criticality. In general, two methods are used for auditing:

8
  Procedures Extensive for Certified Contractors (ISO 9001 2008)
 Procedures Summarized for Non-Certified Contractors

Generally, audits are carried out by a team including QA, Document Control and EoR (if applies):

 External Geotechnical Audit – BGC – every 3 months

 Lundin Mining Independent Tailings Stewardship Review – once per year
 Sernageomin Audit [Government Mining Department; Safety (x3) / Environmental (x2) /
Tailings (x2)] – unannounced
 DGA Audit [Government Water Management] – unannounced

External CQA audits are performed periodically to oversee the quality assurance service provided
by the EoR, to identify gaps and give recommendations to CCMC.

CONCLUSION
The Governance framework applied to the TSF management by CCMC follows the best practice of
industry to ensure tailings dam safety, incorporating well-established Owner’s corporate guidelines,
and for the specific construction first stage, an Owner’s CQA management team reporting to the
Owner’s CQA-CQC Superintendent. The Owner’s team is well supported by the EoR team (design
and analysis at the office and CQA on site, by the EoR onsite), as well as external audits hired by the
Owner and corporate review.
The CQA management framework has been functional to the specific project challenges during
construction. The OSP tailings dam construction implies increasing efficiency and reducing footprint.
However, it also imposes several challenges requiring comprehensive and dynamic CQA on field
and further training of the mining crew “as a major earthworks contractor”. The successful
management of this challenges has been facilitated and assured by the EoR onsite, being part of the
general EoR team with a deep understanding of the main drivers of the project and permit
commitments. This role and tools like NCR, within specific flowchart of management, has proven to
be a key part to accomplish a successful assurance of the construction quality, involving good records
of CQC.

REFERENCES

International Council of Mining & Metals, ICMM. (December 2016). Position statement on preventing
catastrophic failure of tailings storage facilities.

Canadian Dam Association, CDA (December 2016). Key Positions Related to Dam Safety including
the Owner and Engineer of Record. Section 3.1.2 of Technical Bulletin: Application of Dam Safety
Guidelines to Mining Dams issued in 2014.

9
Canadian Dam Association, CDA (October 2018). Revision to Section 3.1 of the Technical Bulletin:
Application of Dam Safety Guidelines to Mining Dams issued in 2014.

The Mining Association of Canada, MAC (2019). Management of Tailings Facilities Guide Version
3.1.

10
Main Challenges in Tailings Dams Risk Management
Luis Valenzuela
Independent Geotechnical Consultant, Chile

ABSTRACT
A paper published in 2015 by Bowker & Chambers showed that the total number of tailings dam
failures was declining since the 80’s. However, and at the same time, there was an increasing
tendency in the number of failures with serious and very serious consequences. The authors, based
on information collected until 2012, concluded that there were no indications how this tendency
could be modified in the near future and that a certain number of serious failures should be expected
in subsequent years. Unfortunately, their forecast was right, and the mining industry had suffered
more than 10 important tailings dams’ failures during the last 7 years, some of them with very serious
consequences. The most impacting cases are Mount Polley in Canada in 2014, Fundao dam in Brazil
in 2015 and quite recently, in January 2019, Brumadinho dam also in Brazil. All three of them with
serious consequences.
These failures happened despite the effort taken forward by the industry and international
institutions to deal with dam safety by the improvement of design and operation guidelines as well
as the recommendation of important changes in owner’s organization and governance for tailings
management.
In this paper, different factors that should be considered in order to improve the tailings deposits and
dams risk management systems are discussed. These factors refer to different aspects such as
engineering design; construction and operational procedures; governance and human factors;
supervision and monitoring; land use planning and closure of tailings deposits. Special emphasis is
placed on the discussion of the importance of risk assessment and risk management as fundamental
tools to control and to diminish the risk of failure of these structures.

1
INTRODUCTION

The mining industry has a very poor record of tailings dam’s safety as shown in Figure 1. For the
1800 – 2000 period, the number of tailings dam failures represents 40% of the total number of failures
of dams of any type. This percentage reached 70% between 1960 and 2000, in which period the mining
industry experienced a major increase in production (Valenzuela, 2015). Although statistics have
shown a decrease in the number of failures since the 90s, the fact is that the number of tailings dams’
failures with serious and very serious consequences has increased as it can be observed in Figure 2
(Bowker & Chambers, 2015). Based on their analysis, these authors concluded that two to five tailings
dams’ failures – some of them with serious to very serious consequences – could occur annually in
the subsequent decades, an estimate that unfortunately has been confirmed in the recent past years.

Figure 1 Statistics on dam failures according to ICOLD (1995, 2001).

2
 Figure 2 Increase of failures with serious consequences (Bowker & Chambers, 2015).

Even before the latter article was completed, the collapse of Mount Polley tailings dam occurred in
2014 in Canada, even though this tailings deposit was following the latest recommendations of the
CDA (Canadian Dam Association). After this event, in November 2015, the catastrophic failure of
Fundao tailings dam in Brazil happened, causing 19 fatalities and vast environmental damage after
releasing 45 Mm3 of tailings. These two disasters brought worldwide attention to the mining industry
and to regulatory agencies. Consequently, new initiatives were put in place promoting improved
governance policies and more comprehensive emergency response plans, the ICMM (International
Council for Mining and Metallurgy) proposed new recommendations for the management of tailings
dams. Unfortunately, the general failure trend was confirmed again by the recent event on 25 January
2019 in Brazil, when tailings dam N°1 of “Corrego do Feijao” – a mine located in the Brumadinho
area in the state of Minas Gerais in Brazil - failed spontaneously releasing more than 12 Mm3. As an
immediate consequence of the dam breach, over 200 people were declared dead and about 100 more
missing. The flood wave advanced at a speed of more than 120 km/h, reaching the Parapeba river
after 7 km, extending the environmental impacts downstream. The impact of these failures has been
significant for the mining industry, governments and communities.
The statistical probability of failure of tailings dams have been estimated to be of the order of 1:700
to 1:750 (Davies et al, 2000) or between 1:1.000 and 2:10.000 in a more recent work (Oboni & Oboni,
2013). The serious failures that have occurred in the last 5 years have most probably increased the
historical failure frequency. This probability of failure is much higher than the one estimated for
conventional dams for water impoundment, estimated in 1:10.000 (Peck, 1980) that, after 1990, could
have diminish even to lower values such as 1:100.000 (Oboni & Oboni, 2013) although the recent
serious failure of the Oroville dam spillway, owned by the California Department of Water Resources
- the highest water dam in the USA – that has called the attention on some vulnerabilities that could
be present in dams designed, operated and supervised by competent institutions such the U.S.
Federal Energy Regulatory Commission, FERC (Morgenstern, 2018).

Table 1 Statistics of tailings dam’s failure in different periods (Oboni & Oboni, 2013)

Even more dramatic is the trend in the statistics of fatalities produced by tailings dam´s failures as
indicated in Figure 3. While the total number of failures is decreasing, the consequences of the failures
are becoming much larger than before with an increasing number of fatalities. This could be caused
by larger deposits and dams built in the last 30 to 40 years, using, in many cases design and

3
construction techniques that previously were considered adequate for smaller deposits. Another
important conclusion is the location of many of these failed dams, upstream from areas occupied by
towns and communities, thus highlighting the importance of proper land use planning.

300?

268

200+
171

1962 1965 1985 2019

Hougudu, China El Cobre, Chile Stava, Italia Brumadinho, Brazil

* Mir Mine Sgorigrad, Bulgaria (1966). 188 – 488 ? fatalities

Figure 3 Number of fatalities produced by tailings dam´s failures are increasing with time

The situation presented by the frequent tailings dam’s failures, some of them with serious
consequences, is not sustainable and clearly not acceptable. There is no ethical justification for the
difference in the probability of failure of water impoundment dams for irrigation or energy
generation purposes and the probability of failure for dams built for the mining industry.
Consequently, something must be done urgently by the mining industry to change this situation.
Many mining companies are implementing initiatives in order to have high-standard dams –
probably not at the same level of water impoundment dams because of the potential problems
associated with a structure that it is built to stand for a long period of time – but at a level that could
guarantee an acceptable low probability of failure. The problem is that not all mining companies are
doing what is needed to reach that acceptable low probability of failure and consequently, failures
will continue to occur, impacting the mining industry negatively.
In the following paragraphs, the analysis of the different factors that should be considered to
diminish the risk of failures will be commented upon taking into consideration the most prominent
failure cases of recent years. Most of these factors or requirements are valid for any type of tailings
deposit but are particularly important for conventional or slurry tailings deposits that correspond to
a significant percentage of the existing tailings deposits, many of which will continue operating for
several years or decades. New deposits in the future could use different technologies that are
presently under development as will be mentioned later in this paper.

4
BASIC REQUIREMENTS TO GUARANTEE THE SAFETY OF TAILINGS DAMS

Looking for basic requirements to reach a high standard on tailings dams’ safety, i.e. with an
acceptable low probability of failure, the following conditions should be fulfilled:
 Adequate site selection considering the potential consequences of a failure.
 Proper engineering design following not only local legal requirements but also the best
practice and recommendations from technical organizations such ICOLD (International
Commission on Large Dams); MAC (Mining Association of Canada); U.S. Bureau of
Reclamation; ICMM and CDA.
 Proper construction and operational procedures according to engineering design through
the entire life of the TFS.
 Comprehensive, robust and redundant monitoring systems, ideally on-line.
 Proper closure design and adequate post-closure supervision and monitoring.
(Abandonment or “walk away” concepts are difficult to guarantee adequate safety.)
 Adequate governance and stewardship structure, with appropriate assignment of roles
and responsibilities.
 Risk management procedures and organization to allow risk-based information to
support management and operational decisions.
Most of these requirements are discussed and commented in several of the design and operational
guidelines prepared by the institutions and agencies mentioned above. Surprisingly, the operational
and economic aspects are not the only aspects that explain the relatively high frequency of tailings
dam’s failure. Professor Morgenstern in a recent lecture pointed out that many of the recorded
failures “arise from deficiencies in engineering practice associated with the spectrum of activities
embraced by design, construction, quality control, quality assurance and related matters”
(Morgenstern, 2018).
In this paper, the two last requirements – governance and risk management - will be discussed in
more detail, especially risk assessment concepts and procedures.

ADEQUATE GOVERNANCE AND STEWARDSHIP STRUCTURE

In 2001, ICOLD recognized the main causes that could explain the high number of tailings dams’
failures and ICMM (2016) confirmed the ICOLD appraisal adding some important comments
regarding the role of tailings management in many of the failures. ICOLD and ICMM both mentioned
the following main causes (ICMM comments shown in italics letters):
 Lack of control of the water balance (leading to overtopping).
 Lack of construction control (primarily quality).
 A general lack of understanding of the features that control safe operations; and

5
  Lack of responsibility and ownership by operators (probably meaning poor governance,
management and assurance).
In addition to the aspects mentioned in ICOLD 2001 and ICMM 2016 reports, the fact that tailings
dams are built and operated for a substantial period of time has been important contributory factor
in explaining the high number of dam’s failures. This characteristic represents a special challenge for
engineering, tailings management and governance. The relatively frequent turnover in operational
personnel results in a lack of continuity and probably also a lack of adequate supervision of design
compliance during the life of the mine. In this situation, there is a risk of misinterpreting the original
design as well as of a reducing or interrupting already in-place control schemes (Valenzuela, 2015).
The cyclic nature of mining economy often leads to periods of temporary low prices of mining
commodities that could bring a tendency to apply cost reductions that might affect the adequate
construction, operation and maintenance of the tailings facilities or could promote the deferring of
necessary planned investments of some ancillary structures (Bowker & Chambers, 2015).
Basic requirements for adequate governance in tailings management according to ICOLD and ICMM
are:
 Tailings Management should be positioned at a high level in the organization and should
not depend directly on the production line management structure.
 Tailings Manager should be a technically competent person with knowledge and experience
in geotechnical and hydraulic engineering.
 Independent Engineer of Records to guarantee continuity in design and operation.
 Periodic Dam Safety Review by independent engineering companies.
 Independent Review Panel reporting to CEO and company Board of Directors.
Figure 4 shows an organization chart according to the present trend in tailings governance in the
industry. The tailings governance problem has been discussed recently in more detail by Valenzuela
(2017 and 2018).

6
 Figure 4 Organization chart showing the present trend in tailings governance (Valenzuela, 2017)

These basic requirements that have been already applied in many of the major mining companies in
Australia, Canada, Chile and the USA, and is starting to be applied in other mining countries, should
be complemented with other structural elements, also indicated in Figure 4, such as:
 Special Technical Committee as a sub-committee of the company Board of Directors
 Risk Management procedures supporting risk-based decisions and governance.

TOLERABLE RISK LIMITS ON DAMS FAILURE

It is clear that the present situation of relatively high frequency of tailings dam´s failures in
comparison with a lower frequency of failures of water impoundment dams, as shown by the
statistics mentioned above - dramatically confirmed with the recent failures of Mount Polley, Fundao
and Brumadinho – is not sustainable and there is an urgent need for the mining industry to improve the
risk management of tailings deposits and dams. The natural question that could arise is the following:
Is the fulfilment of the engineering and operational requirements of local regulations and
recommendations by recognized international organizations enough to guarantee the safety of
tailings dams?

7
The first answer to this question is that there is the additional need for adequate governance and
stewardship structures, with an appropriate assignment of roles and responsibilities, as already
commented. The second answer is associated with the need for appropriate risk management
procedures.
Despite the relatively recent improvement in governance and organization in many mining
companies, major tailings dams’ failures have continued occurring with high costs for the mining
industry and the society, including in some cases several fatalities. One of the main costs mining is
paying, among other costs, is the loss of reputation in the eyes of the affected societies. Society expects
that safety of dams be as high as possible, above all if human life and environmental damage are
compromised in a potential failure of tailings dams. In several countries the society has already
defined tolerable risk limits for the probability of failure of dams and other geotechnical structures
(landslides) when human life is in risk.

Figure 5 Societal tolerable risk limits for dams (ANCOLD, 2003) and landslides (Wong et al, 1997). Both
references cited by Zhang et al. (2016)

Figure 5 shows graphs with annual frequency of events causing fatalities and the tolerable risk limits
for landslides in Hong Kong, which is the first of its kind where when quantitative risk and tolerable
risk limits were included in a state´s regulation (Wong, 2017 and Morgenstern, 2018). The figure also
shows the same type of graph for water dams in Australia (ANCOLD, 2003), which is the first
application of this type of standard for the tolerable risk limits for dams.
Figure 6 shows new and more strict tolerable risk limits for dam failure as defined by the New South
Wales Dam Safety Committee in Australia (New South Wales Dam Safety Committee, 2006) after
discussions with the community.

8
 (*) SBA: Standard-based design according to ANCOLD (2004).

Figure 6 New South Wales Dam Safety Committee societal tolerable risk limits for (a) existing and (b) new
dam failures (reproduced from NSW DSC, 2006)

It should be noted that limits like the tolerable risk limits for loss of life could be defined for significant
environmental damage. It should be recognized that to determine the level of societal or society’s
tolerable risks limits is not a trivial exercise since it requires an ample discussion with the community,
regulators and other stakeholders. It is something that must be done in each societal setting or
alternatively to prescribe in law the level proposed by other countries or institutions. There is no
specific definition for tolerable risks limits for tailings dams. It is reasonable to assume that society
expects that tailings dam´s safety should have similar tolerable limits as the ones defined for water
dams and also demand that the probability of failure to be “as low as reasonably practicable”, i.e. the
ALARP concept. As a general reference, the red circles in Figure 5 indicate the statistical frequency
of tailings dam’s failure as shown in Table 1.
The definition of tolerable risk limits for the mining industry and regulators, in those cases where no
risk of fatalities or of major environmental damage are expected, is still relatively complex, although
in these cases higher tolerable annual probability of failure could be accepted in comparison with
societal tolerable limits. Nevertheless, just the effort to try to define the level of tolerable risk is a very
valuable task that allows each country to identify many of the factors that influence the acceptability
or the rejection of a risk level in a project or operation under study. In the tolerable risk approach,
there is one possible damage that could be underestimated. That is, the loss of reputation of the
mining industry within the society. This is damage that could have a high value for the industry.

9
RISK ANALYSIS, RISK ASSESSMENT AND RISK MANAGEMENT

Defining a tolerable risk limit is one of the main parts of risk management and probably the most
difficult step, unless reference values from other sources are adopted. Also, in order to implement
proper risk management of tailings dams, it is necessary to define the consequences of the potential
failure of the dam as well as the probability of failure of the dam for different failure modes.
The consequences of a potential failure are generally identified through a dam breach analysis and
an economic evaluation of loss of production, construction of a new facility, indemnities,
compensations and penalties plus other more subjective issues such as loss of reputation or loss of
social license to operate, issues to which are not always easy to assign a monetary value. The analysis
of the risk of fatalities is done separately from the number of fatalities since the estimation of
indemnities is already included in the economic risk analysis. A similar treatment could be adopted
when significant environmental damages are involved (economic and strategic approaches).
Regarding the determination of the probability of failure – which is also not always an easy task – it
is necessary to identify and select possible failure modes and to assign a probability of failure for each
mode based on quantitative probabilistic analysis, an event tree or a fault tree analysis. Quite
commonly, a non-quantitative hazard analysis is done based on subjective probabilities of basic
events.

This judgement of experts has been

an important source of information
in risk analysis, even though these
subjective probabilities are often
affected by biases and heuristics
(Fischoff et al., 1977 cited by
Taguchi, 2014).
Figure 7 shows a scheme of the risk
management framework for dam
safety. The result would be
confronted with the tolerable risk
limits at the level of expected
consequences.
It is not uncommon that probabilistic
concepts are combined with
deterministic issues. One example is
when evaluating earthquake
conditions and a low probability
earthquake (EQ) is considered, such
as 1:10.000 or even a “maximum
Figure 7 Risk management framework for dam safety (Bureau credible EQ” together with a
of Reclamation, 2015). deterministic stability or
deformation analysis instead of a
probabilistic stability or deformation

10
analysis. A similar situation is presented when analysing extreme floods in order to verify the
potential overtopping risk.
When a deterministic stability analysis is used – which is the most common approach – it is assumed
that with an adequate factor of safety FS (for instance higher than 1.5), an acceptable low level of
probability of failure is obtained. This is not necessarily true since there is no direct relationship
between a deterministic factor of safety and the probability of failure.
Silva et al (2008) made a statistical analysis reviewing historic slope failure rates of earth structures
producing major accidents. In their study they analysed 75 projects covering over 4 decades including
zoned and homogeneous earth embankments, tailings embankments, natural and cut slopes, and
several earth retaining structures. The authors used an iterative process to obtain the probability of
failure of structures by adopting 2 data points as reference points for the curves of the relationship
illustrated in Figure 8: a FS of 1.5 and probability of failure 10-4 for conservatively designed structures
and FS of 1.0 and failure probability of 0.5 for normally distributed uncertainty on the factor of safety.

The authors identified four

categories of structures based on
engineering, construction,
operation and monitoring level,
ranging from I (Best) represented
by facilities designed, built and
operated with state-of-the-art
engineering; and IV (Poor)
corresponding to facilities with
little or no engineering. From
Figure 8, a FS of 1.5 matches with
a probability of failure of 10-4 for a
structure type II that corresponds
to facilities designed, built and
operated using standard
Figure 8 Annual probability of failure vs. factor of safety in earth engineering practices with many
structures (Silva et al., 2008) ordinary facilities falling in this
category, while a FS of 1.5
corresponds to a probability of failure of 10 for structures type III without site-specific design and
-2

sub-standard construction and operation.

Figure 8 serves the purpose of conceptually illustrating the differences between the probabilistic
approach and deterministic practice.
It should be noted that risk analysis also could have some problems if not applied correctly. Vick
(2017) recognizes the problem of “normalization of deviance” in which robust and even sophisticated
risk assessment methods could be neutralized, mainly due to human factors according to the author:
“Repeated deviations from intended performance that became accepted as normal”
“Deviations were rationalized”

11
 “Warning signs and near-misses were ignored”
“Accepted deviations allowed failure triggers to go unrecognized”.
Nevertheless, risk analysis and risk assessments are the only tools that can be applied in order to
verify the probability of failure of an earth structure and if the risk level associated with the expected
consequences is within the societal or individual tolerable risk limits.

CONCLUSIONS

 The history of tailings dam´s failures in the last decades has affected the mining industry
globally, not only in terms of costs – sometimes significantly impacting the value of a
company – but also in terms of loss of reputation and loss of credibility with regulators and
the community.
 These tailings dam´s failures have resulted in increasing difficulties in the authorization
process for new projects or expansion of existing mines. The mining industry is also facing
increasing opposition from communities thus putting the social license to operate at risk.
 There is a sense of urgency in the need for actions from the mining industry in order to
adopt risk management procedures to diminish the risk of failures of tailings dams.
 Most probably, society will tend to demand from the mining industry similar tolerable risk
limits to those already defined for water dams and expect that these represent risks “as low
as reasonably practicable”. To guarantee the fulfilment of this condition, complete risk
analysis and risk assessment of tailings dams must be carried out, at least for dams for
which failure could produce serious consequences.
 Although risk reduction measures could involve a large scale change in organizational
values and culture, to achieve effectiveness in dam safety, risk management must be
applied. The mining industry has a good example in the “Zero Accidents” policy that had
significant success in reducing the frequency of accidents and fatalities in the mining
industry. Now is the time to define a “Zero Tailings Dam Failure” policy, supported by the
mining industry and encouraged by regulators.
 Risk analysis and risk assessment must be applied in parallel with adequate governance
and mine organizational structure as well as all the recommended basic requirements
necessary to guarantee dam safety in terms of engineering, construction, operation and
closure with proper control and supervision.

ACKNOWLEDGEMENTS

The author acknowledges the collaboration of Engineers José Campaña, Raúl Norambuena and Jaime
Urquidi, through their comments and support in the bibliographic research.

12
REFERENCES

Bowker, L.N. & Chambers, D.M. (2015) The Risk, Public Liability & Economics of Tailings Storage Facility
Failures published by Bowker Associates Science & Research in The Public Interest and Centre for
Science in Public Participation.

Canadian Dam Association (2014) Application of Dam Safety Guidelines to Mining Dams, CDA Technical
Bulletin. www.cda.ca.

Davies, M.P., Martin T.E. & Lighthall, P. (2000) Mine Tailings: When Things Go Wrong, Tailings Dams
2000, Association of State Dam Safety Officials, US Committee on Large Dams, pp 261-273.

International Council on Mining and Metals (2016) Review of Tailings Management Guidelines and
recommendations for Improvement, ICMM Report. www.icmm.com.

International Commission on Large Dams (1995) Dam Failure Statistics, ICOLD Bulletin 99.

International Commission on Large Dams (2001) Tailings Dams Risk of Dangerous Occurrences, ICOLD
Bulletin 121.

Mining Association of Canada (2017) A Guide to the Management of Tailings Facilities, 3rd edition, TSM
Protocol, a Tool for Assessing Tailings Management Performance, MAC June 2017, http://mining.ca/

Morgenstern, N.R. (2018) Geotechnical Risk, Regulation, and Public Policy, The Sixth Victor de Mello
Lecture, Salvador, Brazil, www.abms.com.br

NSW – DSC (2006) Risk Management Policy Framework for Dam Safety, New South Wales Government
- Dam Safety Committee.

Oboni, C., Oboni, F. (2013) Factual and Foreseeable Reliability of Tailings Dams and Nuclear Reactors: a
Societal Acceptability Perspective, Tailings and Mine Waste 2013, Banff, AB, November 6 to 9.

Peck, R.B. (1980) Where Has All the Judgement Gone? The 5th Laurits Bjerrum Memorial Lecture,
Canadian Geotechnical Journal, 17-4, p. 584-590.

Silva, F., Lambe, T.W., Marr, W.A. (2008) Probability and Risk of Slopes Failure, Journal Geotechnical
and Geo-environmental Engineering, ASCE, December 2008, p. 1691 – 1699.

Taguchi, G. (2014) Fault tree analysis of slurry and dewatered tailings management – a framework, Thesis
University of British Columbia
(https://open.library.ubc.ca/collections/ubctheses/24/items/1.0103393).

Valenzuela, L. (2015) Tailings Dams and Hydraulic Fills, The 2015 Casagrande Lecture. XV Pan-
American Conference of the International Society of Soil Mechanics and Geotechnical Engineering,
Buenos Aires, Argentina.

Valenzuela, L. (2017) Tailings Dam Safety and Governance, Power point presentation at the 4th
International Seminar on Tailings Management, Gecamin.com/tailings.

13
Valenzuela, L. (2018) Human Factors and Governance in Tailings Dam Safety, Minexcellence, 3rd
International Seminar on Operational Excellence in Mining, Santiago Chile, www.gecamin.com

Valenzuela, L. (2019) Improving Risk Management in Tailings Deposits and Dams- Challenges and
Proposals, Power point presentation at CESCO Week, Santiago Chile, April 2019, www.cesco.com

Vick, S.G. (2017) Dam Safety Risk – From Deviance to Diligence, Geo-Risk 2017 GSP 282, p. 19 -30.

Wong, H.N., Ho, K.K.S., Chan Y.C. (1997) Assessment of Consequences of Landslides, Proceedings of the
International Workshop on Landslides Risk Assessment, Honolulu, Hawaii, USA, p. 111-149.

Wong, H.N. (2017) Forty Years of Slope Engineering in Hong Kong, HKIE Geotechnical Division Annual
Seminar, Hong Kong Institution of Engineering, p. 1-10.

Zhang, L., Peng, M., Chang, D., Xu, Y. (2016) Dam Failure Mechanisms and Risk Assessment. Published
by John Wiley and Sons Inc. United States. ISBN 10: 1118558510 / ISBN 13: 9781118558515

14
Tailings is focused on being a
forum where professionals can
learn and analyze the recent
innovations related to mining
tailings management. The
seminar gathers prominent
executives, professionals and
academics in one place.

O R G A N I Z E D BY