Citation: Vafaie, A., Soler, J.M., Cama, J., Kivi, I.R., Vilarrasa, V., 2024.

A MATLAB approach for developing digital rock models

of heterogeneous limestones for reactive transport modeling. Geologica Acta, 22.3, 1-14.
DOI: 10.1344/GeologicaActa2024.22.3

A MATLAB approach for developing digital rock models of

heterogeneous limestones for reactive transport modeling

Atefeh Vafaie1 Josep M. Soler2 Jordi Cama2 Iman R. Kivi1 Victor Vilarrasa3

1
Department of Earth Science and Engineering, Imperial College London
Exhibition Rd, South Kensington, London SW7 2AZ, United Kingdom. E-mail: [email protected]
Kivi E-mail: [email protected]

2
Institute of Environmental Assessment and Water Research (IDAEA), CSIC
Jordi Girona, 18-26, 08034 Barcelona, Catalonia, Spain. Soler E-mail: [email protected].
Cama E-mail: [email protected].

3
Global Change Research Group (GCRG), IMEDEA, CSIC-UIB
Miquel Marquès, 21, 07190 Esporles, Illes Balears, Spain. E-mail: [email protected]

ABSTRACT

Porosity is a key parameter controlling the physico-chemical behavior of porous rocks. Digital rock physics offers
a unique technique for imaging the inherently heterogeneous rock microstructure at fine spatial resolutions and its
computational reconstruction, through which a better understanding and prediction of the rock behavior can be
achieved. In this study, we propose a simple but accurate method to build a 3D porosity map of centimeter-scale
carbonate rock cores from X-ray Micro Computed Tomography (XMCT) imaging data. The method consists of
3 main steps: i) decomposition of 3D volumetric XMCT data into sub-volumes, ii) processing of equidistributed
2D cross-section images in each sub-volume and iii) 2D slice-by-slice calculation of porosity and its assembly
to reconstruct a 3D continuum porosity map over the whole core domain using a MATLAB code. The proposed
approach significantly conserves the required memory to manipulate large image datasets. The digitized porosity
representations are used to build 3D permeability maps of the cores by applying an explicit permeability-porosity
relationship. The permeability maps are used as input for numerical simulation of the rock response to the
percolation of reactive fluids through which the general validity of the approach is verified. The developed digital
rock model paves the way for an improved understanding of reactive transport in carbonate rocks.

KEYWORDS Computed tomography. Digital rock physics. Rock heterogeneity. Porosity reconstruction. Reactive transport modeling.

INTRODUCTION 2022; Taron and Elsworth, 2009). The rock behavior under
diverse loading conditions is controlled by its inherently
Understanding the rock response to Electrical, heterogeneous microstructure (Ju et al., 2014; Li and
Thermal, Hydraulic, Mechanical, and Chemical (ETHMC) Benson, 2015). Importantly, the density, distribution, and
loading paths is of paramount importance to a variety of shapes of the pores and grains largely control macroscopic
geoscience and engineering applications (Rutqvist et al., rock properties (Eberli et al., 2003; Wan and Xu, 2014).

 A. Vafaie, J.M. Soler, J. Cama, I.R. Kivi, V. Vilarrasa, 2024 CC BY-SA

Geologica Acta, 22.3, 1-14 (2024) 1

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

Therefore, a thorough characterization of these properties Here we propose a simple yet computationally efficient
requires an accurate assessment of relevant microstructural and accurate MATLAB-based method to construct a 3D
features (Ju et al., 2014; Wan and Xu, 2014). To fulfill this representation of porosity distribution in heterogeneous
requirement, several imaging techniques, including but not carbonate rocks. The code is publicly accessible at https://
limited to, Scanning Electron Microscopy (SEM), Focused github.com/AVafaie/Porosity-reconstruction-from-CT.git and
Ion Beam (FIB), and X-ray Micro Computed Tomography http://hdl.handle.net/10261/284837. The proposed method
(XMCT) have been developed (Ju et al., 2014; Zhang et includes: i) decomposition of the 3D volumetric image
al., 2016). The XMCT technique, although with a relatively of the rock into equal-size sub-domains, ii) processing of
lower spatial resolution compared to SEM and FIB, provides individual, equidistributed 2D images and calculation of
invaluable 3D information on the rock microstructure (Ju et porosity along each sub-domain and iii) assembly of slice
al., 2014; Rabbani et al., 2016, 2019; Wan and Xu, 2014; porosity distribution and mapping onto a 3D continuum rock
Zhang et al., 2016). domain that is defined by arbitrary, user-defined grid sizes
and can be used directly in numerical codes. The proposed
Digital rock physics, i.e. imaging and computing, emerged decomposition of the rock domain and 2D slice-by-slice
in the 1990s as a powerful tool for visualizing the internal image processing significantly minimizes computational
structure of rocks and inferring effective rock properties costs, particularly, in terms of the required memory to
from 3D realizations of the rock microstructure (Fredrich et handle data, while maintaining accuracy. We also propose a
al., 1995; Spanne et al., 1994). Since then, this technology technique to account for sub-CT (Computed Tomography)
has been increasingly employed as a complementary resolution porosity distribution in the rock. As an example,
instrument to conventional core experiments not only we used the obtained porosity distribution of a limestone core
for the nondestructive characterization of geomaterials sample, i.e. Pont du Gard Limestone (Vafaie et al., 2023a)
but also for developing a deep understanding of relevant to build 3D permeability maps as inputs for CrunchFlow
physical processes, mainly involving fluid flow and rock (Steefel and Molins, 2016), an open-access continuum-scale
deformation (Andrä et al., 2013). The latter is achieved by Reactive Transport (RT) code. The proposed procedure can
numerically simulating physical processes in digital rocks, be applied to other problems that require the magnitude and
featuring a representative distribution of relevant properties, distribution of porosity over porous geomaterials.
and making quantitative comparisons with experimental
measurements (Jackson et al., 2018; Rabbani and Jamshidi,
2014). The simulations are commonly carried out at the METHODS
continuum scale, where the numerical grid cells are based on
the concept of Representative Elementary Volumes (REV) Image acquisition
comprising a large enough assembly of pores and grains
with equivalent properties (Hao et al., 2013; Jackson et al., The proposed method relies on discretizing the full core
2020; Smith et al., 2017). The size of the grids is assigned XMCT image into smaller sub-volumes along the axis of
primarily based on computational costs (in particular for the rock sample (z-direction). The discretization enables
3D models where the computational time and memory the parallelization of subsequent data processing on sub-
requirements may become exceedingly high for fine grid volumes, which significantly reduces the computational
cells) and constraints on the applicability of physical cost. To this end, the image dataset is opened using the
models. Accordingly, digital rock physics applies statistical open-access myVGL 2022.4 (data viewer from Volume
or homogenization approaches to derive representative Graphics Software, see www.volumegraphics.com), and
distribution maps of rock properties from microstructural the 3D volumetric image is discretized into an arbitrary
data. These techniques have been successfully used to number of rectangular cuboid layers (i.e. sub-volumes)
evaluate a variety of ETHMC properties of rocks (Durán et with the same size along the core axis. Equal-size 2D
al., 2019; Ettemeyer et al., 2020; Hao et al., 2013; Smith grayscale cross-section images in which every pixel has
et al., 2017; Vanorio and Mavko, 2011; Wu et al., 2020). an intensity value between 0 and 255 (0 for black and
Among them, porosity is a primary parameter that can 255 for white) are then exported out of each sub-volume
be derived from the analysis of XMCT data and is also a using an equal lengthwise interval. This means that each
key input variable to predict other physical rock properties sub-volume comprises a sequence of equal-size 2D cross-
(Eberli et al., 2003). section images separated from each other by a particular
distance. To avoid errors in porosity calculations associated
As digital rock physics gains momentum in geosciences and with small-scale heterogeneities along the core axis, this
engineering applications, the need for appropriate approaches distance should be ideally kept at the minimum value equal
to construct 3D image-based maps of rock properties to the voxel resolution. Excluding the non-rock portions at
becomes a priority. Such approaches are computationally the bottom and top sides of the images, the remaining cross-
demanding as they require analyzing large imaging datasets. section images are exported for each cuboid sub-volume.

Geologica Acta, 22.3, 1-14 (2024) 2

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

Extracted images are first cropped to squares fitting the one hand, a simple approach is to set the threshold value
the circular core cross-sections. The images are then in a way the measured total porosity equals independent
segmented (binarized) by applying a global threshold reference measurements (e.g. from helium or mercury
(Otsu thresholding technique, Otsu, 1979; see Porosity porosimetry). Nevertheless, this method is well known to
distribution of a Pont Du Gard Limestone sample Section) result in over-segmentation and unrealistic representation
to differentiate the pore network from the solid matrix. The of the rock structure (Berg et al., 2018). On the other
output is a stack of 2D binary images (Fig.1). Accordingly, hand, the binarization scheme could be optimized through
pixels on the grayscale image having intensities smaller than the comparison of a dry scan (air-saturated) image with
the threshold value are generally characterized as pores and that of water-saturated rock, as this technique eliminates
converted to black pixels on binary images (elements with the need for knowing the CT number of the solid matrix,
value 0 in image matrices), while the remaining ones above which is principally unknown (Akin and Kovscek, 2003;
the threshold represent the solid rock matrix and converted Pini and Madonna, 2016). The CT number of a material
to white pixels on binary images (elements with value 1 (nCT Mat) is defined in the Hounsfield Unit (HU) (Withjack,
in image matrices). The dark air-filled pixels of the region 1988)
surrounding the core with resembling intensities to pores
turn to black pixels after segmentation. This region is then (1)
manually converted to solid material (white area) before
exporting images to serve as inputs for MATLAB code. where µMat is the linear X-ray beam attenuation
coefficient of the material and µRef is the coefficient of a
The choice of threshold values is however uncertain and reference material, i.e. usually water. Note that CT numbers
thus intrinsically prone to user perception. Appropriate are projected onto the grayscale level (0 for black and 255
criteria should be applied to calibrate segmentation. On for white) in CT images for illustration purposes.

A B
Impor t 2D image sequenc e > Cropping & Segmentat io n >

2.5
Save binar y i mage sequence
Pixel frequency

2
2.5 mm

C 1.5 2

2.5 mm

FIGURE 1. A) A 2D XMCT grayscale cross-section image of the Pont Du Gard Limestone, B) flow diagram of adjusting segmentation threshold using
the ImageJ software and C) extracted 2D binary images.

Geologica Acta, 22.3, 1-14 (2024) 3

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

We propose here a simple segmentation procedure Domain discretization

because more accurate measurements on fluid–saturated
samples are not available in this study. This approach To build a 3D porosity representation over the core
includes the application of several automatic algorithms domain using the obtained 2D binarized images, the domain
(e.g. Renyi Entropy (Kapur et al., 1985), Yen thresholding is first discretized into grid cells. The number of grid cells
(Yen et al., 1995), Moments thresholding (Tsai, 1985)) that is selected based on the problem of interest. The larger the
statistically find optimal threshold values from intensity number (or the smaller the size) of the grid cells, the closer
histograms (indicating how many pixels of each image the 3D representation to the actual rock microstructure, but
share the same gray level). We select the best thresholding the higher the computational cost. An arbitrary number of
method by looking for visual similarities in sharp structural cells on the horizontal plane (Cell_X_No and Cell_Y_No)
features between grayscale and binary images at several can be assigned, but the number of cells along the core
points along the core. The same algorithm is applied to length (Cell_Z_No) equals that of horizontal segments
all images for the sake of consistency. Pores with sizes for which binary images were acquired. In this case, we
less than XMCT image resolution may still result in a have used a total number of 20 horizontal segments along
significant underestimation of the rock porosity, which the core. Yet, there is no restriction on defining more/less
is compensated by randomly distributing small pores horizontal segments based on the problem of interest, the
in the rock matrix (explained in detail in the following size of the sample, and the CT resolution. This information
sections). All image processing steps are performed in together with binary 2D images is given as inputs to the
an automated manner using the open-access Java-based code.
ImageJ software (Schneider et al., 2012) (Fig. 1). Note that
image segmentation could have also been performed using Porosity map reconstruction
MATLAB or other suitable programs (Abdulrahman and
Varol, 2020; Rabbani and Ayatollahi, 2015). All binarized The code first reads the 2D binary images between
images from the inlet to the outlet of the core are saved in the inlet and the outlet of the core but separately and
8-bit .tif format (Fig. 2), stored in a specified folder, and consecutively for the sub-volumes along the core sample.
subsequently used as input files for the MATLAB code These images are matrices (named IMG in the code) of
(https://github.com/AVafaie/Porosity-reconstruction-from- the same size, in which the number of rows and columns
CT.git and http://hdl.handle.net/10261/284837). is equal to the number of pixels in the Y and X directions

B
B input
C
C
CS=1
CS=2
A CS=3
HS=1

HS=1
HS=2
⋮
L = length

HS=3
CS=N

⋮
D
X
HS=M Y
Z
output

FIGURE 2. A) Photograph of the core of Pont Du Gard Limestone used in this study that comprises 1400×1400×2300voxels with a resolution of
20μm, B) Illustration of the Horizontal Segments (HS) along the core length, C) 2D binary Cross-Section (CS) images in one segment and D)
Magnified pixel-based illustration of a pore in a cross-section.

Geologica Acta, 22.3, 1-14 (2024) 4

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

(not necessarily the same), respectively. Each element section pixels (i.e. the sum of Section_Domain elements).
within these matrices is either 0 or 1, referring respectively The number of pore (0) pixels is obtained by subtracting
to black (pores) and white pixels (solid part) (Fig. 3). The the sum of solid part (1) pixels in the IMG matrix from
obtained 3D array of binary images in this step is the basis the total number of pixels in each image. Subsequently, the
of the 3D porosity reconstruction algorithm. average XMCT porosity of each sub-volume (Porosity_
zgrid_total in the code) and the whole core (Porosity_total
Figure 4 illustrates the entire workflow of the proposed in the code) are derived by calculating the arithmetic mean
algorithm, which includes six steps: of the porosity of all images in domains of interest. Note
that the calculated porosity accounts for both connected
1. Locating the core cross-section in the image and isolated pores.

Given that 2D binary images are squared, a white 3. Modifying XMCT-based porosity (optional)
marginal area surrounds the circular core cross-section of
the core. Image processing starts with diagnosing the image The average dry XMCT porosity calculated in the
pixels located on the core (named Section_Domain matrix previous step may be considerably smaller than the
in the code) from those pixels belonging to the marginal core effective porosity (Por_Eff in the code) measured
area. Assuming that the core cross-section is perfectly by other methods. This difference is due to the XMCT
circular, the image pixels with a (X, Y) coordinate belong resolution that limits the diagnosis of relatively small
to the rock only if (X-X0)2+(Y-Y0)2≤ r2, where (X0,Y0) and pores (Mayo et al., 2015; Ramandi et al., 2016). To use
r are the coordinates of the center and radius of the core, the 3D porosity maps in ETHMC numerical simulations,
respectively. The rock pixels are set to 1 in the Section_ the effective porosity distribution should be first adopted.
Domain matrix while the pixels of the white marginal area One should notice that such effective porosity maps
are set to 0. accounting for sub-resolution porosity can be directly
obtained by processing XMCT images of the rock
2. Calculating the average XMCT porosity of the core saturated with a high salinity brine referred to as doped
specimen fluid (Lin et al., 2016). The high CT contrast of the fluid
with the background mineral(s) enables resolving the
The porosity of 2D cross-section images (named effect of tiny pores on the average porosity of the rock.
Porosity_slice_total in the code) is calculated by dividing the However, this requires integrating a flow-through setup
number of pore pixels by the total number of the core cross- into a XMCT system, which is not frequently available

A
A BB
1 2 3 ⋯ ⋯ Cell_X_No 1 ⋯ Cell_X_No
1 1 1 1 1 ⋯ 1 1 1 1
2 ⋯ ⋯
1 1 1 1 1 1 1 1
3 1 ⋮ ⋱ ⋮ ⋮ ⋱ ⋮
⋮ 1 1
⋯
1 1 1 1
⋯
1 1
1 1 1 1 1 1 1 1
⋱ ⋰
1 1 0 0
⋯
1 0 0 0
⋮ ⋮ ⋮ ⋱ ⋮ ⋮
1 1 0 0
⋯
1 1 0 0

⋮ 1 1 1 1
⋰
⋯
⋱
1 1 1 1
Cell_Y_No

⋯ ⋯
Cell_Y_No

1 1 1 1 1 1 1 1
⋮ ⋱ ⋮ ⋮ ⋱ ⋮
1 1 1 1 1 1 1 1
⋯ ⋯
1 1 1 1 1 1 1 1

Pixel_No×Pixel_No

FIGURE 3. A) Illustration of a discretized 2D binary image in X and Y directions and B) the corresponding (0, 1) matrix representation of the image
with the size of Pixel_No×Pixel_No.

Geologica Acta, 22.3, 1-14 (2024) 5

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

as it is the case in this study. Thus, we here take into in the code) with the size of Cell_Y_No × Cell_X_No is
account sub-resolution porosity by adding the difference created for each cross-section image (Fig. 3). Each cell in
between the effective porosity and the average XMCT the Cell_Phi matrix is correlated with a number of pixels in
porosity, i.e. Porosity_Eff - Porosity_total, to the porosity the IMG matrix based on the relative size and coordinates
of each cross-section image. Accordingly, the equivalent of pixels in cells.
number of additional pore elements (Delta_Pixel in the
code) is randomly distributed on the solid portion of 5. Calculating the porosity of grid cells
the core cross-section in 2D binary images. The solid
portion of the core cross-sections is identified by dot The porosity of each grid cell is calculated using the
multiplying (element by element multiplication) IMG same logic as in step 2, i.e. by counting the number of the
and Section_Domain matrices, which returns a matrix of pore (0) elements and dividing it by the total number of
the same dimension with elements equal to 1 denoting elements occupying each cell. It is worth noting that the
the solid portion of the rock. zero-porosity grid cells (white marginal area) surrounding
the core form a non-reactive, impervious zone in our RT
4. Correlating pixel and grid cell domains simulations and help to impose appropriate boundary
conditions. However, these elements are excluded for
Regardless of the implementation of the porosity illustration purposes.
modification step, the workflow continues with correlating
the pixel domain with the grid cell domain to construct 6. Visualizing and exporting the 3D porosity map
the grid-based porosity map across the core as these two
domains have different dimensions. A matrix (Cell_Phi Following the calculation of porosity in grid cells of
each sub-volume, an output 2D array (Porosity_2D in the
code) is defined to store (X, Y) coordinates of the centers of
the cells and their corresponding porosities. These matrices
are then used to plot grid-based porosity distribution maps
in each segment. The higher the Cell_Y_No and Cell_X_
No, the more precise the porosity representation map. The
compilation of porosity distribution matrices for 2D cross-
sections (Porosity_2D in the code) is used to construct the
3D porosity map of the core. Several options are provided
to create different 2D and 3D porosity maps. 3D arrays of
all initial and adjusted binary images, porosity distribution
on the grid domain, and grid information are exported for
further analyses.

RESULTS

Porosity distribution of a Pont Du Gard Limestone

sample

The described image processing approach for 3D

XMCT data is applied to a dry Pont Du Gard Limestone
core (diameter of 25mm and length of 44mm). The
grayscale image of the rock volume, previously
reconstructed from the X-ray CT radiographs, comprises
1400×1400×2300voxels with a resolution of 20μm and is
represented by a dataset of 8-bit unsigned integers with an
approximate size of 4.5GB. The dataset is discretized into
20 sub-volumes along the core axis, and 108 cross-section
images are exported at a section-to-section distance of
20μm (i.e. the voxel resolution) for each sub-volume.
FIGURE 4. Workflow of the algorithm proposed to reconstruct 3D Cropping images to squares fitting the cylindrical core
maps of porosity and associated parameters. The algorithm steps are
differentiated with numbers located on the left side of the computations yields cross-sections containing 1250×1250pixels, i.e.
section. Pixel_No=1250.

Geologica Acta, 22.3, 1-14 (2024) 6

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

As for the segmentation, the Otsu thresholding which the pore and the air-filled surrounding region pixels
technique (Otsu, 1979) returns reliable realizations of the are black and solid matrix pixels are white (Fig. 5B). The
rock structure. This technique selects appropriate thresholds black surrounding region is then manually converted to
from gray-level histograms by minimizing the intra-class white (solid matrix) before exporting binarized images (not
intensity variance or equivalently by maximizing the inter- shown in Fig. 5) to be used in MATLAB code as explained
class intensity variance (σ2) as in the image acquisition section.

(2) The core domain is uniformly discretized into 20×20×20

grid cells, corresponding to a grid width of 1.25mm in the
where σ2 is the inter-class variance and µ is the average X and Y directions and a length of nearly 2.2mm in the
value of intensities weighted by w as the number of pixels Z direction. Each grid cell thus contains 62×62×108voxels
occupying the corresponding intensity range. Indices p at the original CT resolution (20μm). The variation of
and s point to pore (intensities smaller than the threshold) computed XMCT porosity along the length of the limestone
and solid (intensities larger than the threshold) classes, sample is shown in Figure 6, highlighting the heterogeneous
respectively. nature of the rock structure. The average porosity over the
grid cells satisfactorily captures the microstructural patterns
As a representative example, Figure 5 depicts cross- rendered by the high-resolution cross-section images. The
section images of the core sample (Fig. 5A, B), the initial CT porosity (blue curves in Fig. 6) varies between
corresponding gray-level histogram and the evolving inter- 0.1 to 0.16 with an average of 0.13, suggesting that the
class variance (Fig. 5C). Two sharp peaks and a wide valley imaging approach underestimates the bulk rock porosity
in between are recognized in the histogram. Note that the (0.24) by 0.11 units. This underestimation originates from
first peak includes both the pore region and the air-filled the resolution of the CT imaging technique by which a
dark region surrounding the core. In particular, it is widely considerable portion of relatively small pores (<20μm)
recognized that the Otsu approach performs well when the cannot be diagnosed. The observed porosity layering pattern
histogram features a bimodal distribution with sharp peaks is reproduced by the adjusted porosity (brown curves in Fig.
(Kittler and Illingworth, 1985). The inter-class variance 6) owing to the adopted porosity updating strategy of adding
peaks at 109 (Fig. 5C) indicate that intensities larger than uniformly the porosity difference of 0.11 to all cross-section
this threshold serve as the solid rock mineral, whereas the images. It should be noted that comparison with the effective
smaller gray levels point to either pores or the surrounding rock porosity is subject to the assumption that the volume
air-filled region. After applying segmentation, the 2D fraction of isolated pores in the Pont Du Gard Limestone is
grayscale images are transformed into binarized images in negligible.

A C

FIGURE 5. Pont du Gard Limestone core: A) 8-bit gray-level cross-section image from the XMCT data, B) corresponding binary image with a size of
1.49MB and C) the segmentation procedure based on the Otsu method.

Geologica Acta, 22.3, 1-14 (2024) 7

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

for sub-resolution porosity, i.e. pores <20μm (subplots

on the bottom row). The grid-based porosity distribution
maps with 20×20 grid cells satisfactorily reproduce the
structural features observed in binary images, especially
the distribution of large pores that could contribute to
flow localization. Further refinement of the grid (40×40)
significantly increases the accuracy of the porosity
distribution map. Grid refinement, however, increases the
computational cost for numerical simulations built on the
extracted porosity maps.

The heterogeneous distribution of porosity over

FIGURE 6. Workflow of the algorithm proposed to reconstruct 3D the examined core can be also inferred from the 3D
maps of porosity and associated parameters. The algorithm steps porosity map constructed by image processing and
are differentiated with numbers located on the left side of the the corresponding frequency distribution (Fig. 8). The
computations section.
extracted porosity map yields a mean value of 0.217 with
a standard deviation of 0.102. The mean porosity of the
digital rock at a large grid size is smaller than the effective
Figure 7 illustrates the porosity adjustment procedure rock porosity (0.238). This is due to the square-edged
for a single image with general validity for the whole shape of the grid domain that deviates from the cylindrical
sample. Adding the pores randomly at pixel level results core shape and adds to the real bulk rock volume. Mean
in uniform porosity enhancement across the rock slice digital rock porosities of 0.227, 0.231, 0.233, and 0.235
(compare the subplots in the two rows of Fig. 7). While were calculated for 40×40, 60×60, 80×80, and 100×100
image processing initially classifies the rock as non-porous grid schemes perpendicular to the core axis with always
in different possible grid cells (subplots on the top row), 20 grids along the core axis, respectively. The larger the
adjusted images point to a porous background accounting number of grid cells, the closer the grid domain shape to

A B C

D E F

20 × 20 40 × 40
X
2.5 mm
Y

FIGURE 7. 2D illustrations of: A) a raw binary cross-section image, B) a reconstructed grid-based porosity map with a size of 20×20 cells and C) a
reconstructed grid-based porosity map with a size of 40×40 cells. Similar plots D, E and F) provided for the cross-section after adjusting the XMCT
porosity to effective porosity from bulk porosimetry.

Geologica Acta, 22.3, 1-14 (2024) 8

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

a cylinder and the more accurate the calculated mean grid in porous rock media (Steefel et al., 2015). It should be
porosity. noted that our numerical grid size (1.25×1.25×2.2mm3) is
much larger than the largest pore throat size measured by
Code performance the MICP test to be ≈100µm (Vafaie et al., 2023a). Thus,
numerical cells well represent the average properties of the
The described image processing steps were implemented involved microstructural features and allow considering
in MATLAB version 2010b and run on a quad-core Intel their effect on flow, transport, and reaction processes, a gap
Core i7-4790 (up to 3.6GHz) with a maximum of 8GB in conventional simulation workflows (Hao et al., 2013;
RAM. The total runtime was 211s, divided by 57.7s for Smith et al., 2017; Wenck et al., 2021). CrunchFlow uses
importing and calculating porosity for all images, 130.2s a single-column format .txt file for initial rock properties
for adjusting the calculated porosity to the effective porosity in each direction (3 files in 3D problems). The input files
obtained from mercury porosimetry, 6.4s for building consist of one value per line, specifying a rock property
porosity maps on the grid cell domain of 20×20 size, and (e.g. porosity, permeability) in one single grid cell (see
16.8s for calculating permeability maps (as described in Steefel and Molins, 2016, for a detailed explanation of the
the application to reactive transport modeling section) and file formats).
exporting data for further analyses. MATLAB used a total
memory of less than 750MB to perform all calculations. We use the 3D grid-based porosity map to build the
The code can be further optimized to conserve more single-column-format permeability files (PermX, PermY,
memory and make processing larger datasets affordable and PermZ in the code) for the studied limestone core. The
with low-power computational resources. initial permeability of each grid cell from the corresponding
porosity was calculated using the Kozeny-Carman cubic
law, which is expressed as
APPLICATION TO REACTIVE TRANSPORT
MODELING
(3)
In this section, we discuss the application of the
developed digital rock model to reactive transport modeling
of an experiment, in which an acidic solution (CO2-saturated where ki and ϕi are the initial permeability and porosity
water) percolated through a Pont du Gard Limestone core. of the grid cell i in the core domain, and keff and ϕeff are the
The experimental results were reproduced by 3D Darcy- effective porosity and permeability of the core, respectively,
scale reactive transport modeling using the CrunchFlow measured by appropriate laboratory techniques and
code (Dávila et al., 2021; Steefel and Molins, 2016; Steefel supplied as input values to the code (Fig. 4). It should be
et al., 2015). This code can consider the spatial distribution noted that the Kozeny-Carman relationship is supposed to
of porous rock properties (i.e. porosity, permeability work well for intact rocks when strong modifications in the
and tortuosity) when solving fluid flow, solute transport porosity and pore structure do not occur. To demonstrate
and multicomponent equilibrium and kinetic reactions the validity of this relationship for our rock sample, we

A B Z
X
Y

FIGURE 8. Heterogeneous distribution of porosity over the examined core: A) histogram showing the grid-based distribution of total porosity obtained
from image processing and B) 3D porosity map with the size of 20×20×20 grid cells for Pont Du Gard Limestone core.

Geologica Acta, 22.3, 1-14 (2024) 9

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

have used the generated permeability map to reproduce to calculate the evolution of permeability with porosity
the initial bulk permeability of the core. Interestingly, the changes (Hommel et al., 2018)
higher or lower n values cause deviations of the estimated
bulk permeability from the experimentally measured value. (4)
Indeed, using the same keff (measured experimentally for
the bulk rock permeability), but different n exponents in
the porosity-permeability relationship (Eq. 3) generates where ϕ and k are porosity and permeability for each
different realizations of the initial permeability distribution grid cell, respectively, and ϕi and ki are initial porosity
and the numerically calculated bulk flow across the rock. and permeability values, respectively. Note that the
initial rock heterogeneity is exclusively considered in the
The extracted permeability maps in the CrunchFlow permeability map. Thus, ϕi is constant and equals the
code were used to assess the influence of rock heterogeneity effective rock porosity in all cells. The exponent n is an
on the reactive flow of CO2-saturated water in the limestone empirical parameter against which the model is calibrated
core during a percolation experiment (see Vafaie et al., to reproduce the structure of the formed wormhole.
2023a for detailed information). In this experiment, the
CO2-saturated water with a pH of ≈3.13 ([CO2 (aq)]= By setting n=15 (Eq. 4), simulations satisfactorily
1.03mol/kg_H2O) was injected into the core for 28 days at reproduce the composition of the solution at the outlet
a constant flow rate of 0.15mL/min and a constant pressure of the rock core (not shown here) and both the porosity
and temperature (100bar and 60°C, Vafaie et al., 2023a). enhancement in the reacted limestone and the structure and
XMCT observations after injection revealed a significant orientation of the created wormhole (Fig. 9C). The large
increase in bulk rock porosity (∆ϕ ≈ 13.8%) due to calcite n value, which increases the sensitivity of permeability
dissolution (Fig. 9A, B), leading to wormhole formation in to porosity changes, is necessary to reproduce the strong
response to the initial heterogeneity of the rock (Vafaie et feedback between flow and mineral dissolution resulting
al., 2023a, b). in fluid channeling and wormhole creation. The obtained
exponent value is consistent with reported values for
In the numerical simulation, the initial and boundary carbonate rocks used to simulate the injection of acidic
conditions in numerical simulations are considered identical fluids into these rocks (Hao et al., 2019; Hao et al., 2013;
to those of the experiment: i) no-flow lateral boundaries, ii) Smith et al., 2017; Voltolini and Ajo-Franklin, 2019).
constant pressure at the outlet and iii) constant flow rate at
the inlet under constant temperature conditions during the Figure 10 shows histograms of permeability distribution
experiment. Rate laws and reaction rate constants (inputs across the core sample before and after the percolation
for simulations) are those given by Dávila et al. (2015). experiment. It is observed that upon interaction with CO2-
A power-law porosity-permeability relationship, similar saturated water, permeability largely increases (by orders
format to the Kozeny-Carman cubic law (Eq. 3), is used of magnitudes) corresponding to the large n value of 15 in

Inflow Inflow Inflow Porosity, ∅ [-]

A B C
n = 15

Outflow Outflow Outflow

XMCT Simulation
FIGURE 9. Dissolution pattern in the Pont du Gard Limestone core after percolation with CO2-saturated water: A) grayscale XMCT cross-section of the
reacted core representing pore space distribution (black areas) featuring a wormhole structure, B) porosity map of the same cross-section, and C)
RT simulation of the porosity distribution in the same cross-section capturing wormhole formation.

Geologica Acta, 22.3, 1-14 (2024) 10

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

Equation 4. Permeability values in grid cells of the altered CONCLUSIONS

rock span a wide yet generally reasonable range. There is
only a small number of cells near the core inlet (nearly 40 In this study, we have developed a MATLAB-based
cells out of the whole rock domain) in which permeability approach to quantify and visualize the distribution of
reaches a very high value in the order of 10-5m2 as a result porosity in limestones (or other similar rocks) from XMCT
of the large porosity increase that they experienced. We can images. The method has been applied to a core from
approximate the permeability of these cells in the wormhole the Pont Du Gard Limestone. The code implements the
region using the Poiseuille–Hagenbach equation for a tube image processing procedure at a low computational cost,
(Sutera and Skalak, 1993). If the tube diameter equals which is affordable by standard desktop computers. A
the cell width (=1.25mm), the Poiseuille–Hagenbach good correspondence is obtained between measurements
equation yields a permeability of 5×10-8m2 for each cell, at the voxel scale and model grid domains in terms of
which is comparable to those predicted using a power- both bulk rock porosity and spatial pore distribution. The
law relationship with n=15 in our model. This difference finer the grid size, the more accurate the continuum-scale
could be even smaller for a planar array of high-porosity porosity maps. Overall, the digital rock realizations allow
cells forming the wormhole cross section as the Poiseuille– the study of the heterogeneous structure of the Pont du
Hagenbach permeability scales with the square of wormhole Gard Limestone sample reacted with CO2-rich water. We
diameter. Moreover, the effect of the above-mentioned very demonstrate the applicability and validity of the proposed
high-permeability cells on the overall simulation results is approach by parameterizing a continuum 3D Darcy-scale
minimal as these cells constitute a very small portion of the reactive transport model of CO2-rich water injection into
total numerical cells within the rock. a core sample to reproduce the dissolution pattern formed
in the experiment. The developed methodology effectively
Overall, the ability of the model to capture the brings together digital rock physics and Darcy-scale reactive
rock response to the percolation of the acidic solution transport modeling. This combination should enable
demonstrates that although the model is simple compared improvements in our understanding and upscaling of
to other modeling approaches, e.g. the use of the Navier- geochemical interactions between fluids and rocks, which
Stokes equation or a combination of the Navier-Stocks is fundamental for the optimization and risk assessment of
equation with the Darcy equation (Panga et al., 2005; Liu geo-energy applications.
and Liu, 2016; De Paulo Ferreira et al., 2020), it yields
a reliable predictive performance. Nevertheless, a better
match with experimental data may be achieved by refining ACKNOWLEDGMENTS
the mesh and analysis of the parameters controlling the flow
and reaction front advance in the studied limestone. The This work has received funding from the Catalan Government
constructed initial permeability maps exhibit the small- through project 2021 SGR 00308 and Shell through the InFUSE
scale heterogeneities of the limestone core leading to the Prosperity Partnership. I.R.K. acknowledges support by the
localization of flow and reaction. The model can properly PCI2021-122077-2B project (www.easygeocarbon.com) funded
capture the structure of the dissolution pattern indicating by MCIN/AEI/10.13039/501100011033 and the European Union
the general validity of the proposed MATLAB code to NextGenerationEU/PRTR. I.R.K. also acknowledges funding
obtain the initial porosity and permeability maps. from the Engineering and Physical Sciences Research Council
through the UKRI Postdoc Guarantee Award THMC4CCS
[Grant number EP/X026019/1]. V.V. acknowledges funding from
the Spanish Ministry of Science and Innovation through the
project HydroPore (PID2019-106887GB-C32). IDAEA-CSIC
is a Centre of Excellence Severo Ochoa (Spanish Ministry of
Science and Innovation, Project CEX2018-000794-S funded by
Frequency

MCIN/AEI/10.13039/501100011033). This research has been

carried out within the framework of the activities of the Spanish
Government through the “Maria de Maeztu Centre of Excellence”
accreditation to IMEDEA (CSIC-UIB) (CEX2021-001198).

REFERENCES
Permeability (m2)
Abdulrahman, A., Varol, S., 2020. A review of image
FIGURE 10. Histograms of the grid-based permeability distribution in
the intact (blue color bins) and altered limestone core (apricot color segmentation using MATLAB environment. Beirut
bins). (Lebanon), 2020 8th International Symposium on Digital

Geologica Acta, 22.3, 1-14 (2024) 11

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

Forensics and Security (ISDFS), IEEE, 1-5. DOI: 10.1109/ Hao, Y., Smith, M.M., Carroll, S.A., 2019. Multiscale modeling of
ISDFS49300.2020.9116191 CO2-induced carbonate dissolution: From core to meter scale.
Akin, S., Kovscek, A.R., 2003. Computed tomography in International Journal of Greenhouse Gas Control, 88, 272-
petroleum engineering research. London, The Geological 289. DOI: https://doi.org/10.1016/j.ijggc.2019.06.007
Society, 215(1, Special Publications), 23-38. DOI: https://doi. Hommel, J., Coltman, E., Class, H., 2018. Porosity–permeability
org/10.1144/GSL.SP.2003.215.01.03 relations for evolving pore space: a review with a focus on
Andrä, H., Combaret, N., Dvorkin, J., Glatt, E., Han, J., Kabel, M., (bio-) geochemically altered porous media. Transport in
Keehm, Y., Krzikalla, F., Lee, M., Madonna, C., Marsh, M., Porous Media, 124(2), 589-629. DOI: https://doi.org/10.1007/
2013. Digital rock physics benchmarks—Part I: Imaging and s11242-018-1086-2
segmentation. Computers & Geosciences, 50, 25-32. DOI: Jackson, S.J., Agada, S., Reynolds, C.A., Krevor, S., 2018.
https://doi.org/10.1016/j.cageo.2012.09.005 Characterizing drainage multiphase flow in heterogeneous
Berg, S., Saxena, N., Shaik, M., Pradhan, C., 2018. Generation of sandstones. Water Resources Research, 54(4), 3139-3161.
ground truth images to validate micro-CT image-processing DOI: https://doi.org/10.1029/2019WR026396
pipelines. The Leading Edge, 37(6), 412-420. DOI: https:// Jackson, S.J., Lin, Q., Krevor, S., 2020. Representative elementary
doi.org/10.1190/tle37060412.1 volumes, hysteresis, and heterogeneity in multiphase
Dávila, G., Luquot, L., Soler, J.M., Cama, J., 2015. 2D reactive flow from the pore to continuum scale. Water Resources
transport modeling of the interaction between a marl Research, 56(6), e2019WR026396. DOI: https://doi.
and a CO2-rich sulfate solution under supercritical CO2 org/10.1029/2019WR026396
conditions. International Journal of Greenhouse Gas Ju, Y., Zheng, J., Epstein, M., Sudak, L., Wang, J., Zhao, X., 2014.
Control, 54, 145-159. DOI: https://doi.org/10.1016/j. 3D numerical reconstruction of well-connected porous
ijggc.2016.08.033 structure of rock using fractal algorithms. Computer Methods
Dávila, G., Cama, J., Chaparro, M.C., Lothenbach, B., Schmitt, D.R., in Applied Mechanics and Engineering, 279, 212-226. DOI:
Soler, J.M., 2021. Interaction between CO2-rich acidic water, https://doi.org/10.1016/j.cma.2014.06.035
hydrated Portland cement and sedimentary rocks: Column Kapur J.N., Sahoo, P.K., Wong, A.K.C., 1985. A new method for gray-
experiments and reactive transport modeling. Chemical level picture thresholding using the entropy of the histogram.
Geology, 572, 120122. DOI: https://doi.org/10.1016/j. Graphical Models and Image Processing, 29(3), 273-285.
chemgeo.2021.120122 DOI: https://doi.org/10.1016/0734-189X(85)90125-2
De Paulo Ferreira, L., Surmas, R., Tonietto, S.N., Pereira da Silva, Kittler, J., Illingworth, J., 1985. On threshold selection using
M.A., Pires Peçanha, R., 2020. Modeling reactive flow on clustering criteria. IEEE transactions on systems, man,
carbonates with realistic porosity and permeability fields. and cybernetics, SMC-15 (5), 652-655. DOI: 10.1109/
Advances in water resources, 139, 103564. DOI: https://doi. TSMC.1985.6313443
org/10.1016/j.advwatres.2020.103564 Li, B., Benson, S., 2015. Influence of small-scale heterogeneity on
Durán, E.L., Adam, L., Wallis, I.C., Barnhoorn, A., 2019. upward CO2 plume migration in storage aquifers. Advances in
Mineral alteration and fracture influence on the elastic Water Resources, 83, 389-404. DOI: https://doi.org/10.1016/j.
properties of volcaniclastic rocks. Journal of Geophysical advwatres.2015.07.010
Research: Solid Earth, 124(5), 4576-4600. DOI: https://doi. Lin, Q., Al-Khulaifi, Y., Blunt, M.J., Bijeljic, B., 2016. Quantification
org/10.1029/2018JB016617 of sub-resolution porosity in carbonate rocks by applying
Eberli, G. P., Baechle, G. T., Anselmetti, F. S., Incze, M. L., 2003. high-salinity contrast brine using x-ray microtomography
Factors controlling elastic properties in carbonate sediments differential imaging. Advances in Water Resources, 96, 306-
and rocks. The Leading Edge, 22(7), 654-660. DOI: https:// 322. DOI: 10.1016/j.advwatres.2016.08.002
doi.org/10.1190/1.1599691 Liu, N., Liu, M., 2016. Simulation and analysis of wormhole
Ettemeyer, F., Lechner, P., Hofmann, T., Andrä, H., Schneider, propagation by VES acid in carbonate acidizing. Journal of
M., Grund, D., Volk, W., Günther, D., 2020. Digital sand Petroleum Science and Engineering, 138, 57-65. DOI: https://
core physics: Predicting physical properties of sand cores by doi.org/10.1016/j.petrol.2015.12.011
simulations on digital microstructures. International Journal Mayo, S., Josh, M., Nesterets, Y., Esteban, L., Pervukhina, M.,
of Solids and Structures, 188, 155-168. DOI: https://doi. Clennell, M.B., Maksimenko, A., Hall, C., 2015. Quantitative
org/10.1016/j.ijsolstr.2019.09.014 micro-porosity characterization using synchrotron micro-CT
Fredrich, J.T., Menéndez, B., Wong, T.F., 1995. Imaging the pore and xenon K-edge subtraction in sandstones, carbonates,
structure of geomaterials. Science, 268(5208), 276-279. DOI: shales and coal. Fuel, 154, 167-173. DOI: https://doi.
10.1126/science.268.5208.276 org/10.1016/j.fuel.2015.03.046
Hao, Y., Smith, M., Sholokhova, Y., Carroll, S., 2013. CO2- Otsu, N., 1979. A threshold selection method from gray-
induced dissolution of low permeability carbonates. Part level histograms. IEEE transactions on systems, man, and
II: Numerical modeling of experiments. Advances in water cybernetics, 9(1), 62-66. DOI: 10.1109/TSMC.1979.4310076
resources, 62, 388-408. DOI: https://doi.org/10.1016/j. Panga, M.K., Ziauddin, M., Balakotaiah, V., 2005. Two‐scale
advwatres.2013.09.009 continuum model for simulation of wormholes in carbonate

Geologica Acta, 22.3, 1-14 (2024) 12

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

acidization. AIChE journal, 51(12), 3231-3248. DOI: https:// Steefel, C. I., Appelo, C. A. J., Arora, B., Jacques, D., Kalbacher,
doi.org/10.1002/aic.10574 T., Kolditz, O., Lagneau, V., Lichtner, P. C., Mayer, K. U.,
Pini, R., Madonna, C., 2016. Moving across scales: a quantitative Meeussen, J. C. L., Molins, S., 2015. Reactive transport codes
assessment of X-ray CT to measure the porosity of rocks. for subsurface environmental simulation. Computational
Journal of Porous Materials, 23, 325-338. DOI: https://doi. Geosciences, 19, 445-478. DOI: https://doi.org/10.1007/
org/10.1007/s10934-015-0085-8 s10596-014-9443-x
Rabbani, A., Jamshidi, S., 2014. Specific surface and porosity Sutera, S. P., Skalak, R., 1993. The history of Poiseuille’s law.
relationship for sandstones for prediction of permeability. Annual review of fluid mechanics, 25(1), 1-20.
International Journal of Rock Mechanics and Mining Taron, J., Elsworth, D., 2009. Thermal–hydrologic–mechanical–
Sciences, 71, 25-32. DOI: https://doi.org/10.1016/j. chemical processes in the evolution of engineered geothermal
ijrmms.2014.06.013 reservoirs. International Journal of Rock Mechanics
Rabbani, A., Ayatollahi, S., 2015. Comparing three image and Mining Sciences, 46(5), 855-864. DOI: https://doi.
processing algorithms to estimate the grain-size distribution org/10.1016/j.ijrmms.2009.01.007
of porous rocks from binary 2D images and sensitivity analysis Tsai, W., 1985. Moment-preserving thresholding: a new approach.
of the grain overlapping degree. Special Topics & Reviews in Computer Vision, Graphics, and Image Processing, 29(3), 377-
Porous Media: An International Journal, 6(1), 71-89. DOI: 393. DOI: https://doi.org/10.1016/0734-189X(85)90133-1
10.1615/SpecialTopicsRevPorousMedia.v6.i1.60 Vafaie, A., Soler, J.m M., Cama, J., Kivi, I. R., Vilarrasa, V., 2022.
Rabbani, A., Ayatollahi, S., Kharrat, R., Dashti, N., 2016. [CODE] A MATLAB code for digitized reconstruction of
Estimation of 3D pore network coordination number of the rock porosity distribution from Computed Tomography
rocks from watershed segmentation of a single 2D image. (CT) images. Digital CSIC Library. Last accessed: 05/2024.
Advances in Water Resources, 94, 264-277. DOI: https://doi. Website: http://hdl.handle.net/10261/284837
org/10.1016/j.advwatres.2016.05.020 Vafaie, A., Cama, J., Soler, J. M., Grgic, D., Vilarrasa, V., 2023a.
Rabbani, A., Mostaghimi, P., Armstrong, R.T., 2019. Pore Chemo-hydro-mechanical effects of CO2 injection into
network extraction using geometrical domain decomposition. a permeable limestone. International Journal of Coal
Advances in Water Resources, 123, 70-83. DOI: https://doi. Geology, 278, 104359. DOI: https://doi.org/10.1016/j.
org/10.1016/j.advwatres.2018.11.003 coal.2023.104359
Ramandi, H.L., Mostaghimi, P., Armstrong, R.T., Saadatfar, Vafaie, A., Cama, J., Soler, J. M., Kivi, I. R., Vilarrasa, V., 2023b.
M., Pinczewski, W.V., 2016. Porosity and permeability Chemo-hydro-mechanical effects of CO2 injection on
characterization of coal: a micro-computed tomography study. reservoir and seal rocks: A review on laboratory experiments.
International Journal of Coal Geology, 154, 57-68. DOI: Renewable and Sustainable Energy Reviews, 178, 113270.
https://doi.org/10.1016/j.coal.2015.10.001 DOI: https://doi.org/10.1016/j.rser.2023.113270
Rutqvist, J., Wu, Y. S., Tsang, C. F., Bodvarsson, G., 2002. A Vanorio, T., Mavko, G., 2011. Laboratory measurements of the
modeling approach for analysis of coupled multiphase fluid acoustic and transport properties of carbonate rocks and their
flow, heat transfer, and deformation in fractured porous link with the amount of microcrystalline matrix. Geophysics,
rock. International Journal of Rock Mechanics and Mining 76(4), E105-E115. DOI: https://doi.org/10.1190/1.3580632
Sciences, 39(4), 429-442. DOI: https://doi.org/10.1016/ Voltolini, M., Ajo-Franklin, J., 2019. The effect of CO2-induced
S1365-1609(02)00022-9 dissolution on flow properties in Indiana limestone: An in
Schneider, C.A., Rasband, W.S., Eliceiri, K.W., 2012. NIH Image situ synchrotron X-ray micro-tomography study. International
to ImageJ: 25 years of image analysis. Nature Methods, 9(7), Journal of Greenhouse Gas Control, 82, 38-47. DOI: https://
671-675. DOI: https://doi.org/10.1038/nmeth.2089 doi.org/10.1016/j.ijggc.2018.12.013
Smith, M.M., Hao, Y., Carroll, S.A., 2017. Development and Volume Graphics GmbH (2016). A New, More Compatible
calibration of a reactive transport model for carbonate Software Generation: VGStudio 3.0, VGMetrology 3.0, and
reservoir porosity and permeability changes based on CO2 myVGL 3.0. e-Journal of Nondestructive Testing. https://
core-flood experiments. International Journal of Greenhouse www.ndt.net/?id=20203.
Gas Control, 57, 73-88. DOI: https://doi.org/10.1016/j. Wan, K., Xu, Q., 2014. Local porosity distribution of cement
ijggc.2016.12.004 paste characterized by X-ray micro-tomography. Science
Spanne, P., Thovert, J. F., Jacquin, C. J., Lindquist, W. B., Jones, K. W., China Technological Sciences, 57, 953-961. DOI: https://doi.
Adler, P. M., 1994. Synchrotron computed microtomography org/10.1007/s11431-014-5513-5
of porous media: topology and transports. Physical review Wenck, N., Jackson, S.J., Manoorkar, S., Muggeridge, A.,
letters, 73(14), 2001. DOI: https://doi.org/10.1103/ Krevor, S., 2021. Simulating core floods in heterogeneous
PhysRevLett.73.2001 sandstone and carbonate rocks. Water Resources
Steefel, C. I., Molins, S., 2016. CrunchFlow. Software for modeling Research, 57(9), e2021WR030581. DOI: https://doi.
multicomponent reactive flow and transport. User’s manual org/10.1029/2021WR030581
(2016). Berkeley, Lawrence Berkeley National Laboratory, Withjack, E. M., 1988. Computed tomography for rock-
91pp. property determination and fluid-flow visualization. SPE

Geologica Acta, 22.3, 1-14 (2024) 13

DOI: 10.1344/GeologicaActa2024.22.3
A . Va f a i e e t a l . Construct digital rock models of limestones in MATLAB

formation evaluation, 3(04), 696-704. DOI: https://doi. Yen, J.C., Chang, F.J., Chang, S., 1995. A new criterion for
org/10.2118/16951-PA automatic multilevel thresholding. IEEE Transactions on
Wu, Y., Lin, C., Yan, W., Liu, Q., Zhao, P., Ren, L., 2020. Pore- Image Processing, 4(3), 370-378. DOI: 10.1109/83.366472
scale simulations of electrical and elastic properties of shale Zhang, T., Du, Y., Huang, T., Yang, J., Lu, F., Li, X., 2016.
samples based on multicomponent and multiscale digital Reconstruction of porous media using ISOMAP-based MPS.
rocks. Marine and Petroleum Geology, 117, 104369. DOI: Stochastic environmental research and risk assessment, 30,
https://doi.org/10.1016/j.marpetgeo.2020.104369 395-412. DOI: https://doi.org/10.1007/s00477-015-1142-1

Manuscript received June 2023;

revision accepted May 2024;
published Online June 2024.

Geologica Acta, 22.3, 1-14 (2024) 14

DOI: 10.1344/GeologicaActa2024.22.3