See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/344960124

Can machine learning reveal sedimentological patterns in river deposits?

Article in Geological Society London Special Publications · July 2019

DOI: 10.1144/SP488-2018-84

CITATIONS READS

8 318

3 authors:

Vasily Demyanov Arnold Jan Reesink

Heriot-Watt University Lancing College
144 PUBLICATIONS 2,083 CITATIONS 35 PUBLICATIONS 774 CITATIONS

SEE PROFILE SEE PROFILE

Daniel Arnold
Heriot-Watt University
86 PUBLICATIONS 586 CITATIONS

SEE PROFILE

All content following this page was uploaded by Vasily Demyanov on 28 June 2021.

The user has requested enhancement of the downloaded file.

 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

Microbial carbonates: a sampling and measurement challenge

for petrophysics addressed by capturing the
bioarchitectural components
PATRICK CORBETT1*, FELIPE YUJI HAYASHI2, MICHAEL SAAD ALVES2,
ZEYUN JIANG1, HAITAO WANG1, VASILY DEMYANOV1, ALESSANDRA MACHADO2,
LEONARDO BORGHI2 & NARENDRA SRIVASTAVA3
1
Heriot – Watt University, Edinburgh, EH14 4AS, UK
2
Universidade Federal do Rio de Janeiro, Rio de Janeiro, 2194-1916, Brazil
3
Universidade Federal do Rio Grande de Norte, Natal, 59.072-970, Brazil
*Corresponding author (e-mail: [email protected])

Abstract: Ancient and modern stromatolites are potentially a challenge for petrophysicists when
characterizing biosediments of microbial origin. Because of the heterogeneity, sometimes very
cemented and lacking porosity, sometimes highly porous, these widely differing states can be
used to develop techniques that can have wider application to addressing the representative elemen-
tary volume (REV – single or multiple REVs) challenge in microbial carbonates. Effective media
properties – like porosity – need to be defined on REV scales and the challenge is that this scale is
often close to or significantly larger than the traditional core plugs on which properties are tradition-
ally measured. A combination of outcrop images, image analysis techniques, micro-computed tom-
ography (CT) and modelling have been used to capture the porosity (or in some cases, precursor
porosity) architecture and provide a framework for estimating petrophysical property sensitivities
in a range of situations that can be subjected to further calibration by measurements in relevant
microbial reservoir rocks. This work will help guide the sampling approach along with the interpret-
ation and use of petrophysical measurements from microbial carbonates. The bioarchitectural com-
ponent, when controlling porosity in microbial carbonates, presents a significant challenge as the
REV scale is often much larger than core plugs, requiring careful screening of existing data and
measurement and additional geostatistical model-based approaches (with further calibration).

Microbial carbonates (microbialites) have recently which could potentially provide microbialite reser-
become systems of renewed interest for the oil and voir analogue samples.
gas industry after the discovery of more than 30 Biosediments of microbial nature are complex
billion barrels of oil equivalent (BOE) in 11 Pre- cyanobacterial colonial structures variously includ-
Salt accumulations of the Santos Basin, offshore ing biofilms, mats, mounds, oncolites, stromatolites,
Brazil (with reserves of more than 8–12 billion thrombolites, dendrolites, leiolites, tufas and traver-
BOE; Teixeira 2013). Pre-Salt reservoirs, which tines, which can vary in size from a few centimetres
were producing more than 300 000 barrels of oil (oncoids, spherulites and small single stromatolite
per day from just 17 wells, are expected to provide columns) to large structures several metres across.
almost half of the Brazilian oil production of A bioherm is a biosedimentary structure whose
4.2 million bbls/year in 2020 (Hayashi 2013). minimum width is less than or equal to its maxi-
These reservoirs include carbonate intervals of mum thickness, and a biostrome is a stratiform bio-
possible microbial origin (Le Ber et al. 2013; sedimentary structure whose minimum width is
Virgone et al. 2013); however, in some Pre-Salt greater than its maximum thickness. These struc-
lacustrine carbonates, actual microbial structures tures present petrophysical challenges as traditional
may be rare (Wright 2013). petrophysical measurement volumes (probe, plug,
In Brazil, there are a number of microbialite- whole core and log investigation volumes) may or
bearing intervals in the stratigraphic column rang- may not be appropriate statistical support volumes
ing in age from the Proterozoic (e.g. Irecê Basin; to satisfy the conditions for representative elemen-
Pereira 2012) to the Recent lagoonal stromatolites tary volumes (REVs), over and above the normal
on the coast of Rio de Janeiro state (e.g. Lagoa sampling complexity in carbonate pore systems
Salgada, 3000–3500 years BP; Iespa et al. 2011), (Corbett et al. 1999). The methods could be applied

From: Bosence, D. W. J., Gibbons, K. A., Le Heron, D. P., Morgan, W. A., Pritchard, T. & Vining, B. A. (eds)
Microbial Carbonates in Space and Time: Implications for Global Exploration and Production. Geological Society,
London, Special Publications, 418, http://dx.doi.org/10.1144/SP418.9
# 2015 The Geological Society of London. For permissions: http://www.geolsoc.org.uk/permissions.
Publishing disclaimer: www.geolsoc.org.uk/pub_ethics
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

P. CORBETT ET AL.

Fig. 1. Large colonial bioherm from the Neoproterozoic of the Irecê Basin, Bahia, Brazil identifying a complex
association of internal stromatolite columns at a scale larger than log measurements. How would you determine the
effective petrophysical property of such a structure? Scale bar ¼ 1 m.

to abiotic carbonates where there might be similar coastal area (Região dos Lagos, Rio de Janeiro,
sample volume issues (e.g. travertines). Brazil) to address the general aspects of petrophysi-
In this paper, we will present progress towards cal sampling in these systems. The location of these
a more complete porosity description and character- sites is shown in Fig. 4.
ization for microbial carbonates (with potentially In the bioherms from both of these sites there is
wider applications to other petrophysical mea- development of both primary and secondary poros-
surements), which considers the role of micro-CT ity. Primary structure is often preserved, in some
and geostatistics to build nested models for model- form, and will exert control on the REV needed
ling the various petrophysical parameters that are for statistically valid pore characterization.
required for reservoir description and characteriz- Primary and secondary processes lead to many
ation (permeability, kV/kH, relative permeabil- types of porosity development, both intra- and inter-
ity, capillary pressure, formation factor, resistivity granular in nature. As well as the grain-scale poros-
index, NMR response, etc.). We show that petrophy- ity, there is also primary ‘biostructural’ porosity
sical analysis in biosediments requires a com- (shelter, cavity, boring) as well as later diagenetic
plex mixture of measurement and modelling, and development of vugular or secondary porosity to
propose that petrophysical description of biosedi- consider. Describing porosity is a fundamental chal-
ments requires special attention to the develop- lenge as the many porosity types typically present
ment of new workflows, including the techniques will each have their own connectivity issues to
described here. consider with respect to understanding, modelling
We consider for discussion bioherms with large and predicting permeability and/or electrical
columnar (Conophytoida, Kussielida and Gymno- conductivity.
solenida; Fig. 1) stromatolites and small, digiform
(Jurussania sp.; Fig. 2) columnar stromatolites
from the Neoproterozoic Una Group, Irecê Basin Bioarchitecture in carbonates
(Pereira 2012), and stromatolite– thrombolite bios-
tromes (Lagoa Vermelha; Fig. 3) from some of Carbonate structures are often biological in origin.
the present-day lagoons in the Rio de Janeiro state We have adapted the classic reef architectural

Fig. 2. The digiform stromatolite (Jurusania sp.) from the Neoproterozoic of the Irecê Basin, Bahia, Brazil. Right,
oblique view; centre, miro-CT view (solid green; pores red); left, plan view. Note the complex structure at the borehole
(between core plug and wireline log) scale. How would you select representative core plugs for calibration with logs?
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

THE SAMPLING AND MEASUREMENT CHALLENGE

Fig. 3. Stromatolite– thrombolite structure from the Recent Lagoa Salgada, North Fluminese state, Brazil. Porosity is
variously developed throughout this structure as a result of the depositional and growth characteristics of the
stromatolites and other organisms. Note the challenge presented for plugging this sample – where should you take a
sample(s) if you want to measure a representative porosity? The right-hand image shows a 3D representation of the
porosity generated by image analysis of a CT scan.

concept (from Ginsburg and Lowenstam, 1958, in In these diagrams no scale is implied as boring
Moore 2011) to one that can be used to represent a (for instance) can be micro-scale (by algae) or
more general bioarchitectural model (Fig. 5a). macro-scale (by gastropods). Similarly the frame-
There are four main components: work can be centimetres or decametres.
As we are primarily interested in stromatolites in
frame – reef builders such as corals and, as in the
this paper, the primary porosity controls first need
case of this paper, stromatolites;
to be defined for stromatolite systems. In the Pro-
cement – cementing organisms and early diagenetic
terozoic examples, where there are few (if any)
cements;
bio-eroders, this mechanism might need to be dis-
bioerosion – erosion by burrowing and boring
counted, so clearly not all processes will necessarily
organisms;
occur in every case and the model does not intend to
detrital fill – sedimentary detritus, bioclastic frag-
imply that.
ments, ooids, coquinas, sands and muds.

Bioarchitectural porosity development

in carbonates
There are many ways in which porosity can be
created (and occluded) in carbonates and this diver-
sity should be born in mind when approaching
the characterization of a microbial sediment. Car-
bonate pore systems have been classified by
several authors with Choquette & Pray (1970) and
Lucia (1995, 1999) being two of the schemes that
we follow here and apply to the general bioarchitec-
tural model (Fig. 5b).
Matrix permeability in carbonates results from
connectivity of the matrix pore system. From an
understanding of porosity development, in terms
of architecture and interrelationships, a predictive
model for permeability might follow (Fig. 5c) to
show regions of connected porosity (sometimes
helped perhaps by fractures) to define pods of poten-
tial reservoir property rock (with similar porosity
Fig. 4. Location map showing location of the and permeability). Porosity needs to be connected
Neo-Proterozoic, Irece Basin and the Recent, Lagoa to contribute to permeability and microbial porosity
Salgada in Brazil. may be either isolated or part of a connected system.
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

P. CORBETT ET AL.

Actual measurement of permeability in carbonates issues that might ultimately help in the modelling
is subject to fundamental statistical support and of permeability. We do not address permeability
stationarity issues (Corbett et al. 1999) and it is measurement here, as characterizing and measur-
important to first understand the controls on porosity ing porosity in microbialites first requires careful
distribution before attempting to identify appropri- attention.
ate support volumes on which to perform per- From the conceptual porosity diagram (Fig. 5b)
meability measurements. In this respect, this paper and a possible permeability realization of the
focuses on measuring and modelling porosity distri- model, it can be seen that the bio-sedimentary archi-
bution as a constraint on porosity modelling and tecture will be likely to impact the distribution of
providing understanding of porosity connectivity porosity either by textural variations within the

Fig. 5. (a) A representation of bioarchitecture in carbonate rocks (after Ginsburg and Lowenstam, 1958, in
Moore 2011). (b) Porosity types in a bio-architectural model presented in (a) (using terminology from Choquette &
Pray 1970, and Lucia 1995, 1999). Lucia’s grain size-based definition overlays on Choquette and Pray’s fabric
definitions.
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

THE SAMPLING AND MEASUREMENT CHALLENGE

structures or by control of the distribution of sur- (exo-) polymeric substances (EPS) with mineral
rounding sediment. Any technique for assessing precipitation and grain trapping’ (Riding 2011a).
the porosity distribution in a biosediment should The mixture of microbial growth, mineral precipi-
consider both intra- and inter-bioarchitectural tation and grains defines these structures and their
porosity. In this paper we consider primarily the fundamental textural properties, which play a
intra-bioarchitectural porosity and future work will major role in determining future petrophysical prop-
consider the inter-bioarchitectural porosity. How- erties of bioarchitectural carbonates.
ever, the examples described below and the tech- The biological and chemical processes that
niques used will provide insights and directions to drive the growth of microbialites are very com-
address aspects of inter-bioarchitectural porosity and plex. Whilst earlier models suggest a simple ‘trap-
bio-architectural porosity distributions as a whole. ping and binding’ model (Reid et al. 2000), more
Intra-bioarchitectural porosity will control capillary recent work has also emphasized the importance
pressure, saturation distributions and oil recoveries, of biochemical controls. One lamina might result
and could potentially (if pervasive enough in the from five biosedimentary processes (Vasconcelos
reservoir) have a similar effect to the lamina-scale et al. 2013):
features in shallow marine sandstones (Corbett & (1) oxygenic photosynthesis;
Jensen 1993; Ringrose et al. 1993) and diatomites (2) respiration;
(Akhimona et al. 2013). (3) phototrophic sulphide oxidation;
(4) phototrophic sulphate reduction;
Microbial sediments (5) anaerobic methane oxidation.
These processes are in turn controlled by the pri-
Microbial sediments are defined as ‘organosedimen- mary energetic drivers of (Visscher & Stoltz 2005):
tary deposits that have accreted as a result of a
benthic microbial community trapping and bind- (1) source of energy (photo or chemo);
ing detrital sediment and/or forming the locus of (2) electron donor (litho/organo);
mineral precipitation’ (Burne & Moore 1987). This (3) carbon source (auto /hetero).
definition was further refined as the importance of The variety of primary lamination structure is there-
the microbial growth, mineral precipitation and fore potentially very variable. The exact mechan-
grains became better understood to include ‘pro- ism(s) that preserve this primary lamination as a
duced by the interaction of microbial growth with stable mineral in ancient rocks is not straightfor-
metabolism, cell surface properties and extracellular ward. Night and day (diurnal) growth variations

Fig. 5. (Continued) (c) Potential permeability distribution in a bioarchitectural system modified by open fractures.
Note that the pods of permeability are defined by those porosity systems that connect. Where microbial porosity is
developed it may not connect to the ‘main’ permeability conduits. Note that no scale is implied as the processes can
occur at many scales and the model should be taken to imply that every process is present at all times.
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

P. CORBETT ET AL.

Fig. 6. Definition of microbial sediment types (Riding 2011a, b). The structure, individual bioherms or collective
biostromes will provide contrasting pore types from the microbial structure to the intercalated matrix.

can be seen in modern stromatolites, but a single a fundamental control on inter- and intra-
lamination might take several years to accumulate bioarchitectural porosities:
under laboratory conditions (Vasconcelos et al.
2013). What ultimately controls the petrophysi- dendrolite – dendrite or shrub-like benthic micro-
cal properties is the nature of the preserved and bial deposits;
altered lamination, and this end product is what thrombolite – clotted benthic microbial deposits;
can be directly measured. stromatolite – laminated benthic microbial deposits;
leiolite – structureless benthic microbial deposits.

Stromatolite classification We can see that stromatolite is one form of micro-

bial structure, but it is of interest as it is the form
Riding (2008, 2011a, b) defined four types of micro- that combines a defined external form with a comp-
bial structures based on internal fabric and exter- lex laminated internal structure. This means that it
nal structure (Fig. 6), and these will also provide is a ‘typical’ bioarchitectural structure that lends

Fig. 7. Definitions of stomatolitic lamina types. Left, laminated fabric type defined by mechanism (adapted from Riding
2011a, b) showing characteristics of Lagoa Salgada stromatolite. Right, laminated fabric type defined by grain size and
crystal growth (adapted from Riding 2008) showing characteristics of Irece Basin stromatolites.
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

THE SAMPLING AND MEASUREMENT CHALLENGE

Fig. 8. Laminated nature of stromatolites at the centimetric (subcentimetric) scale with textural variations more
common in modern forms (Lagoa Salgada). Lower left, planar laminated form. Lower right, crinkly laminated form.

itself to this study as a representative of the class of Kussiella

microbial sediments where both intra- and inter-
bioarchitectural porosities require characterization. Kussiella is a well-developed colonial stromatolite
with characteristic internal irregular concave-down
structure (Figs 1 & 9–11 top) that is well exposed
Stromatolite textures in pavements (‘Lajedos’ in Portuguese) in Bahia
province of Brazil, west of the town of Morro de
The mixture of microbial growth, mineral precipi- Chapeu. A small example of this structure was
tation and grains that is fundamental to the defini- photographed in plan view (Fig. 12). This example
tion of microbial sediment has been used to define (0.6 × 1.8 m) is a smaller bioherm relative to most
the variation in texture represented as a series of of the structures exposed at this location, which
ternary diagrams, showing the lamina-building might reach more than three times this size. It was
mechanism (Fig. 7 right, Riding 2011a) and grain
size and cement (Fig. 7 left, Riding 2008) with
emphasis on what is visible from the geological
record. It is the laminated nature of these sedi-
ments that shows up well in thin sections (Fig. 8)
and in weathered outcrops (Fig. 9) but might not be
immediately visible in the core (Fig. 10). The exact
origin of these laminae is open to interpretation in
each case as there is usually no remnant of the soft
organic material or information on the precise
nature of the local (bio-)chemical gradients. These
intra-stromatolitic textures are all at a very fine
(subcentimetre) scale where capillary forces play an
important role in reservoir performance (Corbett &
Jensen 1993; Ringrose et al. 1993; Akhimona
et al. 2013).

Proterozoic stromatolite examples

The Proterozoic of the Irece Basin contains some
very well-preserved microbial structures occurring
at various scales in single biostromal beds (Fig.
11). As well as large domal stromatolites exposed
in three dimensions (more than 1 m by several
metres extending over hundreds of metres), there
are smaller interdigitating stromatolite colonies
(over tens of metres) where intra- and inter-
stromatolite textures can be examined in close Fig. 9. Weathered outcrop (Kussiella from Irece Basin)
proximity. These most ancient sediments are often on the side of a pit showing laminated nature of the
very well cemented with little porosity, but this is internal Stromatolite structure. Coloured bands are due
not always the case, as will be seen below. to varying water levels within the pit.
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

P. CORBETT ET AL.

Fig. 10. Core (Kussiella from Irece Basin); note that the lamination is clear at the scale of the thin-section (Fig. 8) and
weathered outcrop (Fig. 9) but not so clearly visible in core. Note the presence of stylilites and a vug, the latter
apparantly unrelated to the lamination, showing that other mechanisms for porosity development are also potentially
present in this system.

not possible to photograph the larger structures in datasets to be handled later using geostatistical soft-
entirety from the ground, so this example was ware. The maximum consideration of the quality of
selected for the purpose of this study. the original image before image capture greatly
The colour image was converted to black and benefits this kind of work, depending on what
white (Fig. 13) and turned into a binary dataset for level of detail is desired (Stockham 1972).
statistical analysis. Mathematical morphology is a Variography can be used to capture various
key resource in image analysis to improve the over- aspects of the spatial correlation structure (Fig.
all quality of the source samples, enhancing features 14). The orientation of the structure is captured by
of interest and balancing different colour patterns variograms across and along the structure (Fig. 14,
(water-level marks, shadows cast by strong sunlight, upper). In the carefully selected cross-sectional
etc.) throughout outcrops digitally captured in the view the vertical and horizontal anisotropy can be
field. Thus, images had to be treated carefully to demonstrated (Fig. 14, middle). At the larger scale,
avoid loss of resolution and additional artefacts analysis of different regions (moving window) in
(Serra 1982). this stromatolite structure captures some of the
In order to define an approach for treating these spatial variability of the colony, by the change in
samples, different types of morphological opera- anisotropy angle (represented by the variogram
tors were tried (Beucher 1999) for this work; ellipsoid rotation) and the multiscale character of
closing and opening proved the most successful. the correlation (showing periodicity and anisotropy
All field digital images were converted from RGB variation with the range; see Fig. 14, lower). In each
to 8-bit (greyscale) binary images using ImageJ segment the proportions of dark and light lami-
1.46r – a freeware pre- and post-processing solu- nations are approximately 50%. These variograms
tion. After that, a wide range of erosion and dilation and the proportions might be used to build a geo-
algorithms were applied to filter relevant infor- statistical model to represent this spatial structure.
mation from these images, avoiding very erratic The correlation lengths can be used to suggest
values owing to noise and capture limitations. appropriate sample volumes where the properties
Then these images were outputted as. TEXT-based are stationary. Attempts to reproduce aspects of the
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

THE SAMPLING AND MEASUREMENT CHALLENGE

Fig. 11. Various well-preserved microbial sediments from the Proterozoic Irece Basin. Top, Kussiella to the left
side of the photograph – note the man’s legs for scale; middle, Jurusania (hand specimen, 10 cm across); bottom,
thrombolite from Irece Basin (note that edge of field notebook is close to the size of a 1 5 inch core plug and this
sample volume should not pose as many sampling challenges as the microbial material in the top two images).
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

P. CORBETT ET AL.

Fig. 12. Kussiella stromatolite colony prior to image processing.

Kusssiela structure using the variograms in pixel the Kussiella ‘heads’, further work is need to nest
models (Kanevski & Maignan 2004; Remy et al. these individual stromatolite head models inside
2009) were not successful, so a multipoint technique the bioherm colonial structure.
was used. Full details of this work can be found in
Alves (2015). Jurusania
Realizations generated by (semi-)variogram-
based algorithms are not capable of modelling Jurusania is a colonial stromatolite with many small
complex, connected and curvilinear spatial varia- finger-sized columns and characteristic concave-
tion (like the stromatolite bioarchitecture). In order down internal structure (Figs 2 & 17) that is found
to overcome these limitations, multiple-point geo- to the NW of Morro de Chapeu in Bahia state in
statistics was introduced together with the con- the Irece Basin.
cept of a ‘training image’ and its repeated-pattern This digitate stromatolite (possibly more strictly
information (Guardiano & Srivastava 1993). The classified as a dendrolite) structure also lends
challenging features of these bioherms, therefore, itself to 3D analysis (Fig. 11, middle) in outcrops
require one to work with a multiple-point pattern- within an old sulphate mine. Discrete examples
based approach using the training images in of the stromatolite columns can be extracted and
Figure 14. these lend themselves to micro-CT analysis (Figs
The algorithm FILTERSIM (Zhang et al. 2006; 2 & 17). The micro-CT was acquired using a micro-
Wu 2007) was selected owing to its pattern-based tomography SKYSCAN (model 1173 High Energy)
approach, which is more capable of reproducing in the Geology Department of the Universidade
non-linear features like the lobe-shaped tops of the Federal do Rio de Janeiro. The voxel size is
specimens from the Kussiella group. All simulations 29.02842 mm. The micro-CT image clearly demon-
were unconstrained, unconditional and using two strates internal porosity (i.e. Fig 2; the red porosity
categorical training images for the plane view and is not only developed on exposed internal struc-
profile (Figs 15 & 16). Whilst these models cap- ture but fully enclosed) and this is constrained by
ture the local horizontal and vertical variation of the primary laminated nature of the stromatolite.

Fig. 13. Kussiella stromatolite colony after image processing. The binary image lends itself to statistical
structural analysis.
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

THE SAMPLING AND MEASUREMENT CHALLENGE

Fig. 14. Variography showing different aspects of the Kussiella stromatolite. (a) Plan view of section of the colony
(refer to Appendix 1). (b) Variogram showing varying correlation lengths along (below) and across (above). (c)
Cross-sectional view (carefully selected from Fig. 6). (d) Short vertical and long horizontal correlation length. (e)
Variography for various regions (from image in Fig. 13) showing the predominant orientation of the structure in each
segment. This variographic analysis provides a two-point statistical description for geostatistical model building in three
dimensions. The sill on the variograms suggests stationarity and samples at least twice the correlation length (c. 10 cm
vertically and 30 cm or more laterally) might give representative porosities – if the porosity variation follows the pattern
of the black and white image.
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

P. CORBETT ET AL.

Fig. 15. Two unconditional realizations performed with the FILTERSIM algorithm (Zhang et al. 2006) on a categorical
training image with two classes, search template of 51 × 41×1 and inner patch dimensions of 37 × 27 × 1. All
simulations have a grid size of 480 × 474 pixels, and were sampled from the plane view of Kussiella colony (Fig. 14a),
with a range of 4 m2 (2 × 2 m).

The lamination contains crandallite (Al-Apatite). In environments. Often highly variable and containing
some cases this is the material that dissolves to multiple pore systems, the analysis of the porosity
create porosity, but other factors also play a role in systems is critical to the understanding of these stro-
porosity creation, such as micro-fissures, micritiza- matolites and, understanding how the petrophysical
tion, recrystalization and dolomitization. properties relate to the various originating processes
Pore network approaches utilize micro-CT will help guide the sampling, interpretation and
images and these techniques can be adopted to modelling. These porosity systems are strongly con-
build models at larger scales (Fig. 18). Pore network trolled by bioarchitecture and allow us to investigate
models have been validated elsewhere in simpler the 3D REV challenge.
systems (Jiang et al. 2007, 2013; Knackstedt et al.
2007; Ryazanov et al. 2009) and validation in com- Lagoa Salgada
plex microbial systems presents challenges because
of the scale of the REV and the shape and arrange- Samples of present day and recent stromatolites
ment of pores. An aspect of future investigation from the coastal lagoons of Brazil (Lagoa Vermilha,
and research is to develop strategies for collecting near Araruama, and Lagoa Salgada, Paraiba do Sol
appropriate samples and take measurements to test delta, near Campos dos Goytcazes, Rio de Janeiro
the validity of simulations at this scale. state) have been examined by several authors as
potential analogues for ancient stromatolites
(Vasconcelos et al. 2006; Iespa et al. 2009, 2011;
Modern example dos Reis Neto et al. 2011), including laboratory
studies (Vasconcelos et al. 2013). In this study, an
There are also excellent modern analogues of stro- example stromatolite from Lagoa Salgada (Fig. 3)
matolite structures in Brazil’s coastal lagoonal was subjected to CT imaging (Fig. 19) and an

Fig. 16. One unconditional realization performed with the FILTERSIM algorithm (Zhang et al. 2006) on a categorical
training image with two classes, search template of 51 × 41×1 and inner patch dimensions of 37 × 27 × 1. All
simulations have a grid size of 720 × 172 pixels and were sampled from the profile view of Kussiella colony (Fig. 14b),
with a range of 5 m2 (1 × 5 m).
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

THE SAMPLING AND MEASUREMENT CHALLENGE

Fig. 17. Segmentation of an individual Jurusania stromatolite showing pore (blue) and solid (green) structure and
extracted pore network (porosity 15.5%, one subnetwork of slit-shaped pore structures).

examination of the intra-stromatolite porosity was compared with thin sections from the same
development was performed. sample. The porosity was very unpredictable when
X-ray CT is a non-destructive technique that taking the entire sample into consideration, but for
allows the visualization, and most importantly, subregions defined by the texture, the porous
quantification of the internal structure and volume space was more predictable (Fig. 20). Details of
of objects, determined mainly by variations in den- the approach can be found in Hayashi (2014).
sity and atomic composition (Mees et al. 2003). For To analyse the porosity structure of the sample,
this project we used a medical CT at the ‘Hospital Avizo Fire software was used. The program allowed
Beneficencia Portuguesa’ in Rio de Janeiro. The us to manipulate the images, subdividing the image
device’s technical specifications are not known, but to focus the different subregions identified. We were
the voxel size is 1 mm. The purpose of the CT is to able to identify, quantify and filter the pores depend-
give an overall notion of the stromatolite sample’s ing on volume, as well as segmenting them from the
porous space (Figs 2 & 19–21). The CT image rest of the rock, in order to understand how porosity

Fig. 18. A pore network of the pore system imaged in Figures 14 and 15. Note that the concavity of the pores has been
ignored and they are represented as simple discs – which might be appropriate for simple connectivity studies. Analysis
of the pore network allows large models of the pore structure to be built and these can be used for generation of
petrophysical properties at various scales.
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

P. CORBETT ET AL.

Fig. 19. CT volume of a modern stromatolite from Lagoa Salgada. Specimen is 8 cm square.

varies throughout the sample. Consideration of the (3) a dying phase where salinity becomes very
variability within the stromatolite (Fig. 22) shows high and growth periods less common.
that the porosity is very non-stationary (Corbett
et al. 1999), with the development of three distinct We can identify many bioarchitectural types of por-
layers with different REV volumes. osity, largely of a primary nature:
Medical CT of the large 8 × 8 × 4 cm volume † intra-stromatolite porosity;
yielded a rather low resolution. Micro-CT of a † inter-stromatolite porosity;
small 1 × 1 × 3 cm volume resolved the pore struc- † bioerosional; and
ture much more effectively (Fig. 21). In the sample † detrital porosity.
studied in this work (and in many of the other
samples of stromatolites from the modern The three-fold division into ‘garret’, ‘piano nobile’
lagoons), we can identify a same three-fold layering and ‘cellar’ pore systems (Fig. 21) defines the poros-
and this has developed as a result of evolving pri- ity type layering, with the cellar porosity being the
mary depositional controls; most stable areally (and therefore probably the
most important contributor to effective per-
(1) the initial creation of the lagoon with access to meability). The REV has been determined for the
open marine water influences; cellar, piano nobile and garret, and the porosity cal-
(2) a rapid growth phase of the stromatolites culated for the total stromatolite (Fig. 22). The cellar
under optimal conditions; and is the only material that has an REV close to that of

Fig. 20. Distribution of porosity in subregions of the CT volume showing a high degree of valiablity (and a lack of
statistical stationary).
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

THE SAMPLING AND MEASUREMENT CHALLENGE

Fig. 21. High-resolution micro-CT image nested within the low-resolution CT image. With higher resolution one
can distinguish a three-layer variation in pore type as a result of the evolution of the bioarchitecture in response to the
water depth and dynamics in the lagoon. At the top is ‘garret’ – rather ‘miserable’ – porosity; in the centre is the ‘piano
nobile’ porosity, the largest and most flashy; and at the base is the ‘cellar’ porosity, underground and by far the best
connected. Each of these porosity types will have different effects on the connectivity and permeability of the system.

a 1 inch core plug. This suggests that the cellar carbonate (,1000 mm3, Vik et al. 2013). This
porosity could be measured effectively by a 1 inch factor illustrates why microbial carbonates require
plug. It is probably a coincidence that the cellar special attention to porosity structure derived from
porosity (15%) is quite close to the total stromato- a variety of depositional fabrics. This example has
lite porosity (18%). The REV in microbial carbon- had limited diagenesis, which would probably
ates is demonstrated here to be potentially much change the REV, but it is expected that primary
larger (.5000 mm3) than has been identified in a bio-architectural textures would retain an influence
detailed REV study of a macroporous (even vuggy) on the porosity REV.

Fig. 22. Variation of porosity for different volumes extracted from the CT volumes. (a) As the volume increases, the
variability in porosity estimates decreases. The reduction in variability occurs where the representative elementary
volume is reached. (b) The REV in the cellar is reached close to the volume of a 1 inch core plug. Garret porosity is also
close to the volume of a 1 inch core plug but the REV for the piano nobile is a much larger volume. Note that in this case
the porosity for the total stromatolite (18%) is close to the porosity of the cellar (15%).
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

P. CORBETT ET AL.

Conclusions Alves, M. S. Y. 2015. Geological modelling of biosedi-

mentary structure in microbialites of the Irecê Basin,
Porosity measurements and porosity distributions BA, Brazil. Unpublished thesis, Universidade Federal
in microbial sediments are complex because of do Rio de Janeiro.
pore development as function of the distribution of Beucher, S. 1999. Mathematical Morphology and
primary frame, bioerosion, detrital fill and cement Geology: A Review of Some Applications. University
of Liège, Belgium, Geovision ‘99.
elements. The early controlling factors can influ-
Burne, R. V. & Moore, L. S. 1987. Microbialites: orga-
ence porosity distribution through to the present nosedimentary deposits of benthic microbial commu-
day. The preservation and connection between these nities. Palaios, 2, 241– 254.
elements will have a strong influence on the per- Choquette, P. W. & Pray, L. C. 1970. Geologic nome-
meability of a microbial carbonate reservoir. clature and classification of porosity in sedimentary
Modern and ancient stromatolites are examples carbonates. AAPG Bulletin, 54, 207 –250.
of microbial sediments characterized by complex Corbett, P. W. M. & Jensen, J. L. 1993. An application
primary textural and micro-structural control on of probe permeametry to the prediction of two-phase
porosity development. Some ancient examples flow performance in laminated sandstones (Lower
Brent Group, North Sea). Marine and Petroleum
show that the primary characteristics of a discon-
Geology, 10, 335 –346.
tinuous lamination structure in stromatolites can Corbett, P. W. M., Anggraeni, S. & Bowen, D. 1999.
still retain an influence on porosity development The use of the probe permeameter in carbonates –
after more than 1200 myr. The pores in ancient addressing the problems of permeability support and
examples have been shown to be slit-like. stationarity. The Log Analyst, 40, 316– 326.
Microbial sediments by their nature require the Guardiano, F. & Srivastava, R. M. 1993. Multivariate
mapping and quantification of the bioarchitectural- geostatistics: beyond bivariate moments. In: Soares, A.
related variability for model building to estimate (ed.) Geostatistics Troia 92. Kluwer Academic,
petrophysical properties at the appropriate statisti- Dordrecht, 1, 133–144.
Hayashi, F. Y. 2014. Evaluation of the representative
cal support scale. The scale that these microbial
elementary volume in a recent stromatolite (Lagoa
structures occur at is often larger than conventional Salgada, RJ). Unpublished thesis, Universidade
core plugs or even whole core sampling allow. Federal do Rio de Janeiro (in Portuguese).
In these circumstances modelling becomes more Hayashi, M. Y. 2013. Pre-Salt: challenges and opportu-
useful than measurements that will be prone to nities (A possible dream). Interesse Nacional, July/
artefacts. Sept, 34–40 (in Portuguese).
Petrophysical analysis in microbial sediments, Iespa, A. A. C., Iespa-Damazio, C. M. & Borghi, L.
increasingly based on 3D networks, requires an 2009. Microstratigraphy of the stromatolite, thrombo-
understanding of bioarchitectural controls and mod- lite and oncoid Holocene Complex of Lagoa Salgada,
Rio de Janeiro State, Brazil. Revista de Gologia (For-
elling techniques that capture the structure of these
teleza), 22, 7 –13 (in Portuguese).
components at various scales. Iespa, A. A. C., Borghi, L. & Damazio Iespa, C. M. 2011.
O plexo estromatólito-trombólito-oncoide, Lagoa
Salgada, RJ, Brasil. In: Carvalho, I. S. et al. (eds)
The work mentioned in Brazil was carried out in associ-
Paleontologia: cenários da vida. Interciência, Rio de
ation with the ongoing Sergipe-Alagoas Carbonate Lab-
Janeiro, 57–68 (in Portuguese).
oratory Project, funded by BG Group Carbonate
Jiang, Z., Wu, K., Couples, G. D., van Dijke, M. I. J. &
Technology Hub at Universidade Federal do Rio de
Sorbie, K. S. 2007. Efficient extraction of networks
Janeiro (UFRJ) and registered as ‘Análise geológica sedi-
from 3d porous media. Water Resources Research,
mentaria de sucessôes carbonáticas cretácias em uma bacia
43, W12S03, http://dx.doi.org/10.1029/2006WR
sedimentaria brasileira’ (Fundação Coppetec project IGEO
005780
15.981; UFRJ/BG Brasil/ANP) funded by BG Brasil
Jiang, Z., van Dijke, M. I. J., Sorbie, K. S. & Couples,
under the Autoridade Nacional do Petroleo ANP R&D
G. D. 2013. Representation of multi-scale heterogen-
levy as ‘Compromisso de Investimentos com Pesquisa e
eity via multi-scale pore networks. Water Resources
Desenvolvimento’. P. Corbett was also funded by BG
Research, 49, 5437–5449, http://dx.doi.org/10.1002/
Group for a two-year assignment in UFRJ in connection
wrcr.20304
with this project. L. Braga is thanked for useful discussions
Kanevski, M. F. & Maignan, M. 2004. Analysis and
on mathematical morphology and geostatistics.
Modelling of Spatial Environmental Data. Ecole Poly-
technique Federal de Lausanne Press, Switzerland.
Knackstedt, M. A., Arns, C. H. et al. 2007. Archie’s
References exponents in complex lithologies derived from 3D
digital core analysis. In: Transactions, SPWLA 48th
Akhimona, N., Corbett, P. W. M. & Geiger, S. 2013. Annual Logging Symposium, Austin, Texas, paper no.
Numerical modelling of a low-permeability, highly UU 1 –16.
compactible waterflooded reservoir. In: SPE 163590, Le Ber, E., Le Heron, D. P., Vining, B. A. & Kamona, F.
SPE Reservoir Simulation Symposium held in The 2013. Neoproterozoic Microbialites in a Frontier
Woodlands, Texas, 18– 20 February 2013. Basin; The Rasthof Formation (Cryogenian). In:
 Downloaded from http://sp.lyellcollection.org/ at Heriot-Watt University on May 27, 2015

THE SAMPLING AND MEASUREMENT CHALLENGE

Proceedings of Microbial Carbonates in Space and of oil layers during water invasion. Transport in Porous
Time Workshop. Geological Society, London, 19– 20 Media, 80, 79–99, http://dx.doi.org/10.1007/s11242-
June 2013. 009-9345-x
Lucia, F. J. 1995. Rock fabric/petrophysical classification Serra, J. 1982. Image Analysis and Mathematical Mor-
of carbonate pore space for reservoir characterisation. phology. Academic Press, London, I.
AAPG Bulletin, 79, 1275– 1300. Stockham, T. G., Jr 1972. Image processing in the
Lucia, F. J. 1999. Carbonate Reservoir Characterisation. context of a visual model. IEEE Proceedings, 60,
Springer, New York. 828– 842.
Mees, F., Swennen, R., van Geet, M. & Jacobs, P. 2003. Teixeira, A. A. 2013. The exploration and production of
Applications of X-Ray Computed Tomography in the petroleum in Brazil, 15 years since the opening up.
Geosciences. Geological Society, London, Special Interesse Nacional, July/Sept, 34– 40 (in Portuguese).
Publications, 215. Vasconcelos, C., Warthmann, R., McKenzie, J. A.,
Moore, C. H. 2011. Carbonate Diagenesis and Porosity. Visscher, P. T., Bittermann, A. G. & van Lith, Y.
Developments in Sedimentology, 46. Elsevier, 2006. Lithifying microbial mats in Lagoa Vermelha,
Amsterdam. Brazil: modern Precambrian Relics? Sedimentary
Pereira, C. P. 2012. Carbonates of the Irecê Basin. Asso- Geology, 185, 175– 183.
ciação Brasileira de Geólogos do Petróleo, Rio de Vasconcelos, C., Dittrich, M. & Mackenzie, J. A.
Janeiro (field course guide, June 2012). 2013. Evidence of microbiocenoensis in the forma-
Reid, R. P., Visscher, P. T. et al. 2000. The role of tion of laminae in modern stromatolites. Facies, 60,
microbes in accretion, lamination and early lithifica- 3– 13, May, http://dx.doi.org/10.1007/s10347-013-
tion of modern marine stromatolites. Nature, 406, 0371-3
989–992. Vik, B., Bastesen, E. & Skauge, A. 2013. Evaluation of
dos Reis Neto, J. M., Fiori, A. P. et al. 2011. X-ray com- representative elementary volume for a vuggy carbonate
puted microtomography integrated with petropraphy in rock – part 1: porosity, permeability, and dispersivity.
a three-dimensional study of porosity in rocks. Revista Journal Petroleum Science and Engineering, 112,
Brasileira de Geosciencas, 41, 498 –508. 36–47, http://dx.doi.org/10.1016/j.petrol.2013.03.029i
Remy, N., Boucher, A. & Wu, J. 2009. Applied Geosta- Virgone, A., Broucke, O. et al. 2013. Continental car-
tistics with SGeMS, A User’s Guide. Cambridge Uni- bonates reservoirs: the importance of analogues to
versity Press, Cambridge. understand Pre-Salt discoveries. In: Proceedings of
Riding, R. 2008. Abiogenic, microbial and hybrid authi- Microbial Carbonates in Space and Time Workshop,
genic carbonate crusts: components of pre-Cambrian Abstract, Geological Society, London, 19–20 June
stromatolites. Geologia Croatica, 61, 73–103. 2013.
Riding, R. 2011a. Microbialites, stromatolites, and Visscher, P. T. & Stoltz, J. F. 2005. Microbial mats as
thrombolites. In: Reitner, J. & Thiel, V. (eds) Ency- bioreactors, populations, processes and products.
clopedia of Geobiology. Encyclopedia of Earth Palaeogeography, Palaeoclimatology, Palaeoecology,
Science Series, Springer, Heidelberg, 635– 654. 219, 87–100.
Riding, R. 2011b. The nature of stromatolites, 3500 Wright, V. P. 2013. To be or not to be, Microbial: Does it
million years of history and a century of research. In: Matter? In: Proceedings of Microbial Carbonates in
Reitner, J., Queric, N.-V. et al. (eds) Advances in Space and Time Workshop, Abstract, Geological
Stromatolite Geobiology. Lecture Notes in Earth Society, London, 19–20 June 2013.
Sciences, Springer, Heidelberg, 131, http://dx.doi. Wu, J. 2007. Non-stationary Multiple-point Geostatisti-
org/10.1007/978-3-642-10415-2_3 cal Simulations with Region Concept. Department of
Ringrose, P. S., Sorbie, K. S., Corbett, P. W. M. & Geological and Environmental Sciences, Stanford Uni-
Jensen, J. L. 1993. Immiscible flow behaviour in lami- versity, Stanford, CA (Proceedings of the Stanford
nated and cross-bedded sandstones. Journal of Pet- Center for Reservoir Forecasting, 20).
roleum Science and Engineering, 9, 103– 124. Zhang, T., Switzer, P. & Journell, A. 2006. Filter-
Ryazanov, A. V., van Dijke, M. I. J. & Sorbie, K. S. based classification of training image patterns for
2009. Two-phase pore-network modelling: existence spatial simulation. Mathematical Geology, 38, 63–80.

View publication stats