Data in Brief 40 (2022) 107797

Contents lists available at ScienceDirect

Data in Brief

journal homepage: www.elsevier.com/locate/dib

Data Article

3D Dataset of binary images: A collection of

synthetically created digital rock images of
complex media
Javier E. Santos∗, Michael J. Pyrcz, Maša Prodanović
Hildebrand Department of Petroleum and Geosystems Engineering, The University of Texas at Austin, Austin, TX,
USA

a r t i c l e i n f o a b s t r a c t

Article history: Digital rock images are computational representations that

Received 27 September 2021 capture the geometrical complexity of systems present ubiq-
Revised 22 December 2021 uitously in nature. In recent years, their use has become
Accepted 3 January 2022 widespread due to the increasing availability of repositories,
Available online 6 January 2022
and open-source physics simulators and analysis tools. Here,
we present a dataset of 3D binary geometries in a standard-
Keywords:
ized format that represent a wide variety of geological and
Sphere-packs
Fractures media
engineering systems. Our data is freely available at [1].
Catalyst layers © 2022 The Author(s). Published by Elsevier Inc.
Shales This is an open access article under the CC BY license
Vuggs (http://creativecommons.org/licenses/by/4.0/)

Specifications Table
Subject Computers in Earth Sciences
Specific subject area 3D Digital Rock Images
Type of data Image
How data were acquired Computer simulations
Data format 3D arrays in HDF5 format
Parameters for data collection After carrying-out the simulation to generate each image, all the isolated pore
bodies were removed and a check was performed to assess connectivity along
the third coordinate axis
(continued on next page)

∗
Corresponding author.
E-mail address: [email protected] (J.E. Santos).
Social media: (J.E. Santos), (M.J. Pyrcz), (M. Prodanović)

https://doi.org/10.1016/j.dib.2022.107797
2352-3409/© 2022 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/)
2 J.E. Santos, M.J. Pyrcz and M. Prodanović / Data in Brief 40 (2022) 107797

Description of data collection The images are presented as binary 3D arrays of zeros and ones. The former
represent the pore-space and the later the solid phase.
Data source location Institution: The University of Texas at Austin
Data accessibility Repository name: Digital Rocks Portal (DPR) Data identification number: 374
Direct URL to data: https://www.digitalrocksportal.org/projects/374 The code to
create this dataset is part of the MPLBM-UT library [2].

Value of the Data

• The dataset includes 75 3D geometries (rock samples) representing a wide variety of deposi-
tional systems and diagenetical processes. This data has been used in a number of research
articles [3–11].
• The collection is standardized with a high-performance file format (HDF5) that integrates
its metadata (for portability), making it accessible with any programming language and
popular scientific-visualization software. All the samples were post-processed to ensure 6-
connectivity which allows flow simulations to converge faster.
• Researchers and users involved in direct simulations methods of physical processes can bene-
fit from these images for benchmarking their simulations and inferring new physical insights
with a standardized dataset. AI researchers will utilize this dataset of complex geometries to
test supervised and unsupervised classification algorithms.
• The images can be used as inputs for physics-based simulations (e.g., single- and multi-phase
flow [2,12,13], particle transport, nanoconfinement, electrical conductivity, and solid mechan-
ics) to get new physical insights (e.g., scaling relationships, surrogate modeling, etc.), geo-
metrical analysis (e.g., geostatistics, spatial statistics, etc.) [14], train supervised/unsupervised
deep learning models, on samples that are representative of different subsurface settings.

1. Data Description

Flow and transport in porous media are critical to understand the geological processes of
rock formation and applications such as the management of groundwater resources, carbon se-
questration, enhanced oil recovery, and contaminant transport. Typical geological systems are
composed of a broad spectrum of porous media with features such as porosity and permeability
varying by orders of magnitude within an individual system. An example application is to uti-
lize a digital rock as an analog to derive the relationship between porosity and permeability to
impute missing permeability data from porosity data points, which are widely available.
The term ‘digital rock’ refers to a computational model of a porous medium in 2D or 3D
whose source is either an actual image of a rock or soil (e.g., an x-ray microtomography image),
or it is based on a model (e.g., calculated via process-based simulations of sediment transporta-
tion and deposition stochastic realization from a geostatistical model). These are commonly for-
matted in a regular grid or mesh that allow their use in a numerical simulator to forecast a
transport process or property of interest. These simulations are then post-processed to obtain
a bulk property that characterizes the rock in question, and this research area is often referred
to as digital rock physics, or digital core analysis. Digital Rocks Portal [15] provides a curated
library of digital rocks representing a wide variety of domains ranging from imaged biofilms in
soil to texturally equilibrated models of salt pore spaces. Digital Rocks Portal stores both actual
images from a variety of imaging techniques and representative porous media models. Digital
Rocks Portal is organized in units called projects.
The dataset collection (Digital Rocks Portal project) described here provides 3D porous and
fractured models that systematically vary a range of their properties and can be used for bench-
marking simulation algorithms, including systematic training of deep learning algorithms pre-
dicting transport properties [10,11]. Datasets are in the form of binary images that have desig-
nated pore (void) and solid (grain) numerical cells or voxels, and are organized in 10 groups (or
 J.E. Santos, M.J. Pyrcz and M. Prodanović / Data in Brief 40 (2022) 107797 3

samples, in Digital Rocks Portal jargon). Many have already been used in previously published
work though not necessarily shared. The groups are described as follows:
Group 1 contains discretized representation of experimental measurement of a packing of
identical spheres (Sample 1, the original is stored at the portal [16]). This sample is then sys-
tematically changed to mimic diagenesis processes. One morphological sphere dilation and six
morphological erosions are numerically performed to obtain samples with higher (Sample 0)
and lower porosity (Sample 2–8) respectively, but that still exhibit a similar pore structure to
the original sample. These samples come in both 2563 and 4803 .
Group 2 contains self-similar fractures running through the middle of the volume. This group
contains five fractures with varying surface roughness representing different materials. These
five fractures are downscaled three times to create fractures with smaller aperture that still ex-
hibit the same degree of surface roughness. All the downscaled fractures have the same mean
aperture (hence, porosity) respectively.
Group 3 contains packings of randomly placed spheres. Unlike Group 1, these spheres are
allowed to overlap slightly to achieve a target porosity. These images are well-suited to rep-
resent catalyst layers [17]. This group contains samples with six porosities (10, 15, 20, 25, 30,
and 35%) with three realizations each (since the algorithm uses a random number seed to start
the simulation). Group 7 contains samples created with the same algorithm and with different
seeds.
Group 4 contains two process-based reconstructions of nanoporous organic matter in shale
[5]. These samples have a porosity of 20 and 30%.
Group 5 contains four fractured samples. Unlike Group 2, the fractures have an adjacent per-
meable matrix [3,4].
Group 6 contains two samples with a linear gradient in the porosity of their slices in the
z-direction.
Group 8 contains a realistic single fracture on a solid domain. The roughness of this sample
is heterogeneous and the aperture field is not self-affine.
Group 9 Randomly placed overlapping spheres with an embedded synthetic fracture along
the z-axis.
Group 10 Voxelized sphere packs where a certain number of solid grains were removed to
simulate a vuggy domain.
The highest and lowest porosity values (defined as number of zero-entries over image size)
in each group can be visualized in Fig. 1. A selection of samples is shown in Fig. 2.

Lowest Porosity
40 Highest Porosity
Porosity(%)

20

0
1 2 3 4 5 6 7 8 9 10
Group
Fig. 1. Lowest and highest porosity sample values in each group.
4 J.E. Santos, M.J. Pyrcz and M. Prodanović / Data in Brief 40 (2022) 107797

Fig. 2. 3D plots of dataset samples. Each subplot shows the Euclidean distance of the pore space for visualization pur-
poses. The code to reproduce this figure is shown in Listing 1.
 J.E. Santos, M.J. Pyrcz and M. Prodanović / Data in Brief 40 (2022) 107797 5

Listing 1. Python code to download and plot the binary images as shown in Fig. 2.

2. Experimental Design, Materials and Methods

All the image groups were created synthetically to represent different natural settings. The
main motivation to create this dataset was to have a diversity of benchmark samples to per-
form transport simulations and geometric analysis. All the samples are discretised as 2563 vol-
umes and some have an additional 4803 volume. All of these are presumed unit-less and can be
scaled to any physical length of interest (for instance, spheres diameters can be scaled by a rele-
vant mean grain diameater). The pore-space is represented with zeros and the solid matrix with
ones. Most of the images have been used in projects related to a wide range of earth-science
applications.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal rela-
tionships which have, or could be perceived to have, influenced the work reported in this article.

Acknowledgments

We would like to acknowledge the contributions of Dr. Yin Ying, Dr. Christopher J. Landry,
and Dr. Hasan J. Khan to this project.

References
[1] J. E. Santos, Y. Yin, M. Prodanovic, A. Gigliotti, N. Lubbers, H. Khan, 3D Collection of Binary Images, 2021a, doi:10.
17612/nxgk-k611. (http://www.digitalrocksportal.org/projects/374a).
[2] J. E. Santos, A. Bihani, A. Gigliotti, MPLBM-UT: Single- and Multi-phase simulations in PM, 2021b, 10.5281/zenodo.
4665905.
[3] M. Prodanovic, S.L. Bryant, Physics-Driven interface modeling for drainage and imbibition in fractures, SPE J. 14 (03)
(2009) 532–542, doi:10.2118/110448-PA.
6 J.E. Santos, M.J. Pyrcz and M. Prodanović / Data in Brief 40 (2022) 107797

[4] M. Prodanovic, S.L. Bryant, Z.T. Karpyn, Investigating matrix/fracture transfer via a level set method for drainage and
imbibition, SPE J. 15 (01) (2010) 125–136, doi:10.2118/116110-PA. https://onepetro.org/SJ/article/15/01/125/192097/
Investigating- Matrix- Fracture- Transfer- via- a- Level.
[5] C.J. Landry, M. Prodanović, P. Eichhubl, Direct simulation of supercritical gas flow in complex nanoporous media
and prediction of apparent permeability, Int J Coal Geol 159 (2016) 120–134, doi:10.1016/j.coal.2016.03.015. https:
//linkinghub.elsevier.com/retrieve/pii/S0166516216300659.
[6] Y. Yin, Z. Qu, C. Zhu, J. Zhang, Visualizing gas diffusion behaviors in three-dimensional nanoporous media, Energy
& Fuels 35 (3) (2021) 2075–2086, doi:10.1021/acs.energyfuels.0c03529.
[7] E. Guiltinan, J.E. Santos, Q. Kang, Residual saturation during multiphase displacement in heterogeneous fractures
with novel deep learning prediction, in: Unconventional Resources Technology Conference (URTeC), Society of
Petroleum Engineers (SPE), Austin, 2020, doi:10.15530/urtec- 2020- 3048.
[8] E.J. Guiltinan, J.E. Santos, M.B. Cardenas, D.N. Espinoza, Q. Kang, Two-phase fluid flow properties of rough fractures
with heterogeneous wettability: analysis with lattice Boltzmann simulations, Water Resour Res (2020), doi:10.1029/
2020WR027943.
[9] J. Santos, M. Prodanovic, C. Landry, H. Jo, Determining the impact of mineralogy composition for multiphase flow
through hydraulically induced fractures, in: SPE/AAPG/SEG Unconventional Resources Technology Conference 2018,
URTC 2018, 2018, doi:10.15530/urtec- 2018- 2902986.
[10] J.E. Santos, D. Xu, H. Jo, C.J. Landry, M. Prodanović, M.J. Pyrcz, PoreFlow-Net: a 3D convolutional neural network to
predict fluid flow through porous media, Adv Water Resour 138 (February) (2020) 103539, doi:10.1016/j.advwatres.
2020.103539. https://linkinghub.elsevier.com/retrieve/pii/S0309170819311145.
[11] J.E. Santos, Y. Yin, H. Jo, W. Pan, Q. Kang, H.S. Viswanathan, M. Prodanović, M.J. Pyrcz, N. Lubbers, Computation-
ally efficient multiscale neural networks applied to fluid flow in complex 3D porous media, 0123456789, Springer
Netherlands, 2021, doi:10.1007/s11242- 021- 01617- y.
[12] J.E. McClure, Z. Li, M. Berrill, T. Ramstad, The LBPM software package for simulating multiphase flow on digital
images of porous rocks, Comput. Geosci. 25 (3) (2021) 871–895, doi:10.1007/s10596- 020- 10028- 9.
[13] J. Gostick, M. Aghighi, J. Hinebaugh, T. Tranter, M.A. Hoeh, H. Day, B. Spellacy, M.H. Sharqawy, A. Bazylak, A. Burns,
W. Lehnert, A. Putz, OpenPNM: a Pore network modeling package, Computing in Science & Engineering 18 (4)
(2016) 60–74, doi:10.1109/MCSE.2016.49. http://ieeexplore.ieee.org/document/7478437/.
[14] J. Gostick, Z. Khan, T. Tranter, M. Kok, M. Agnaou, M. Sadeghi, R. Jervis, Porespy: a Python toolkit for quantitative
analysis of porous media images, Journal of Open Source Software 4 (37) (2019) 1296, doi:10.21105/joss.01296.
[15] M. Prodanovic, M. Esteva, M. Hanlon, G. Nanda, P. Agarwal, Digital Rocks Portal: a repository for porous media
images, 2015, https://www.digitalrocksportal.org/. 10.17612
[16] J. Finney, M. Prodanovicc, Finney Packing of Spheres, 2016, (http://www.digitalrocksportal.org/projects/47).
10.17612/P78G69
[17] K.J. Lange, P.-C. Sui, N. Djilali, Pore scale simulation of transport and electrochemical reactions in reconstructed
PEMFC catalyst layers, J Electrochem Soc 157 (10) (2010) B1434, doi:10.1149/1.3478207.