Three-Dimensional Pore Networks in Miocene Stevens Sandstone of California: Implications for CO₂ Geologic Storage

Abstract

The Miocene Stevens Sandstone in the San Joaquin Basin of California is increasingly recognized as a promising candidate for CO₂ geological storage due to the enormous storage capacity, proven sealing, and existing infrastructure. In this study, computed microtomography imaging and pore network modeling were employed to investigate the influence of pore geometry and wettability on the CO₂ injectivity and residual trapping. Image analysis revealed that a significant fraction of the cement and matrix consists of microporous regions. The microporosity can substantially increase the overall pore space, yet its contribution to permeability remains modest, particularly in samples with low permeability. The intrinsic heterogeneity of turbidite reservoirs further complicates the reservoir properties among different layers. Two-phase flow simulations under varying wettability conditions (water-wet, weak water-wet, and neutral-wet) demonstrated that the CO₂ injection is predominantly controlled by macropores. CO₂ invades microporous regions only after these larger pores are filled. The presence of microporosity leads to a decrease in both initial and residual CO₂ saturations, with the magnitude of the reduction being influenced by wettability. Neutral-wet scenarios exhibit higher CO₂ mobility and thus lower residual trapping than water-wet scenarios. The results imply that heterogeneity in pore geometry and cement distribution across different layers can result in stratified CO₂ flow pathways, complicating efforts to predict injection performance. Overall, the Stevens Sandstone shows considerable promise for CO₂ geologic storage, but effective implementation will require detailed characterization of the pore structure as well as the integration of reactive fluid flow to account for potential mineral dissolution and fines migration.

1. Introduction

CO₂ geological sequestration is a promising technique for reducing the anthropologic CO₂ emissions to the atmosphere (e.g., − ). Potential underground storage options include sedimentary rock formations (including saline aquifers and depleted oil and gas reservoirs), unmineable coal seams, organic-rich shales, basalts, and peridotites. , Among these, CO₂ geological storage in depleted oil and gas reservoirs is the most mature approach, offering significant storage capacities, existing infrastructure, detailed geologic data, and proven trapping mechanisms. , Sandstones, long considered important hydrocarbon reservoirs, also serve as sinks for the storage of CO₂ in the subsurface. In this process, supercritical CO₂ is injected into pore spaces of sedimentary rocks filled with formation fluids and trapped by capillary forces, structural traps, dissolution into formation fluids, and mineralization driven by CO₂–brine–rock reactions. ,,,− Residual trapping of fluids in a reservoir refers to the process by which micrometer-scale CO₂ bubbles become immobilized by capillary forces within the intricate pore structure of the reservoir rocks. ,,, Consequently, residual trapping provides an additional layer of security in areas with numerous preexisting penetrations (e.g., oil and gas wells, water disposal wells), such as depleted oil and gas fields.

CO₂ injection into sedimentary rock formations began in the 1970s as a method for enhanced oil recovery (EOR). Although early projects often involved EOR, the International Energy Agency estimates that up to 30 Gt CO₂ could be permanently stored in suitable formations. Currently, there are 5 commercial-scale CO₂ sequestration projects in sedimentary rock formations, including the Sleipner Project and the Snøhvit Project offshore Norway, the Decatur Project in Illinois (United States), the Quest Project in Alberta (Canada), and the Gorgon Project in Barrow Island (Australia). ,− In California, depleted hydrocarbon reservoirs may be more promising than saline aquifers due to the proven trapping structures.

Over the past decade, numerous national and regional studies have been conducted to assess California’s CO₂ geologic storage potential across oil and gas fields, saline aquifers, and existing underground gas storage sites. ,− Among these, Elk Hills Oil Field was estimated to have the highest storage capacity, ranging from 135.2 Mt CO₂ (low estimate) to 453.9 Mt CO₂ (high estimate). The reservoir properties of Elk Hills Oil Field and adjacent North and South Coles Levee oil fields have been studied extensively using well logs and routine core analysis. − The primary reservoir formation, the Stevens Sandstone, lies between 5500 and 9500 feet in depth (Figure ). In addition to Elk Hills Oil Field, Stevens Sandstone and its stratigraphic equivalent Santa Margarita Formation is the primary reservoir formation for some of the largest oil fields in San Joaquin Basin, including Midway-Sunset, Kern River, Elk Hills, and Buena Vista oil fields, as well as smaller ones such as North and South Coles Levee oil fields (Figure ). , To date, two pilot projects have successfully injected CO₂ into Stevens Sandstone, one at North Coles Levee Oil Field in 1981 and another at Elk Hills Oil Field in 2005. , Although both were originally designed for CO₂ EOR, each project demonstrated excellent injectivity and highlighted the formation’s strong potential for long-term CO₂ storage. ,

1.

Geologic cross section of the Southern San Joaquin Basin in California illustrating the multiple reservoir formations in shades of yellow (primarily sandstone). The Stevens Sandstone and its equivalent Santa Margarita Sandstone in the middle is one of the largest potential reservoirs for CO₂ storage (Reproduced with permission from).

Stevens Sandstone was deposited as turbidite channel sands on deep-sea fans on the eastern and southeastern margin of the San Joaquin Basin within the whole Miocene Monterey Formation. ,,,, In a study that summarized 1133 core analyses from 37 wells in Stevens Sandstone, the average porosity was reported ranging between 16.1 and 19.5%, while the average permeability ranged between 6 and 130 md. In certain layers, the porosity could be as high as 25%, while permeability could be close to 1000 md (Figure ). The effective porosity was primarily calculated from well log data, assuming an ineffective shale volume and shale porosity. ,, The average pore diameter is 10 μm measured by confocal microscopy and scanning electron microscopy. The permeability of Stevens Sandstone varies significantly due to the uneven distribution of chert cement and clay content. Amott tests were applied on sandstone core samples from the Monterey Formation, and they reported mixed-wet behavior of the sandstones. ,

2.

Porosity and permeability of the samples analyzed, in comparison with previous measurements from 1133 core analyses of Stevens Sandstone with samples from 37 wells in the San Joaquin Basin. Background reservoir properties were reproduced with permission from ref. (Copyright 1996 Elsevier).

Recent studies have shown that permeability heterogeneity within potential reservoirs strongly influences the injectivity of CO₂ (e.g., − ). The permeability of a reservoir depends on geometric properties of the pore system, including effective porosity, pore and pore throat size distributions, and pore connectivity. ,, The heterogeneous distribution of chert cement, clay minerals, and clay-sized particles further complicates the pore structure by creating micropores that hinder fluid flow. , Meanwhile, wettability, whether the rock surface preferentially interacts with water, CO₂, or other fluids, governs the capillary forces that control fluid displacement and residual trapping. Sandstone reservoirs are normally considered water-wet. Yet recent studies found minerals’ wettability alternation due to extended CO₂ exposure (or “aging”), from strongly water-wet to less water-wet with increasing contact angles. , Therefore, a quantitative representation of the pore structure is critical for evaluating the efficiency of CO₂ sequestration.

This study aims to characterize the three-dimensional (3D) variability of pore networks in Stevens Sandstone and determine how these variations affect CO₂ injection and residual trapping. Six representative sandstone core samples were selected for X-ray computed microtomography (micro-CT) analysis. Micro-CT is a nondestructive imaging technique that captures the spatial distribution of the linear or mass attenuation coefficient, which correlates with the sample’s density and atomic composition. , It allows three-dimensional visualization and reconstruction of pore structures at micrometer-scale resolution while preserving the integrity of the sample. The residual CO₂ trapping of a CO₂/brine system in six sandstone samples was examined across a range of microporosity levels and wettability conditions. A comprehensive set of pore structure characterizations for each sample allowed a direct linkage between pore structures and fluid flow properties. This research provides a useful analogy for CO₂ storage in sandstone reservoirs worldwide, particularly those deposited in turbidite systems.

2. Samples and Methods

The Miocene Monterey Formation contains complex lithologies including diatomite, porcelanite, siliceous shale, chert, clay-rich shale, dolomite, and interbedded thick turbidite sandstones. For the purpose of CO₂ geologic storage, sandstone is preferred over all of the other lithologies since the pore space is better connected and thus easier to inject. This work specially focuses on the sandstones. The samples used were carefully selected from one well in Elk Hills Oil Field across a span of over 500 ft to capture the wide ranges of porosity and permeability (Figure ). One-inch-long core plugs were cut from well-preserved cores. Core plugs were first cleaned using a Dean–Stark apparatus with toluene to remove residual brine. This was followed by Soxhlet extraction using methylene chloride and methanol azeotrope. Methylene chloride was used to extract the remaining hydrocarbons, while methanol helped dissolve any deposited salts. This robust two-step cleaning procedure ensured that the majority of the pore space was accessible for subsequent analyses. The cleaned samples were then cut into approximately 5 mm cubes and scanned with micro-CT.

2.1. Micro-CT Scanning

Micro-CT scanning was conducted at the Molecular Imaging Center at Keck Medicine of the University of Southern California. Samples were scanned with a Phoenix Nanotom M nanofocus X-ray CT system manufactured by Waygate Technologies. The detector was dxr-5001, the timing was 1250.361 ms, the X-ray voltage was 100 kV, the current was 90 μA, and the filter was 0.1 mm Cu. The raw scan was taken at a voxel size of 2.8–3.3 μm. Each sample took approximately 1 h to complete. Nanotom native reconstruction software was used for the 3D reconstruction. Inline-median filtering was applied to enhance contrast.

2.2. Image Processing

Image processing and the following analysis were conducted using PerGeos software from Thermo Fisher Scientific. To avoid open fractures and imaging artifacts, the raw scans were subsampled to 600³ voxels for the following analysis. Resized data volumes were filtered with Nonlocal Means Filter to reduce noise and enhance the contrast while maintaining the boundaries between mineral grains, cement, and pores. Compared with other image filters, such as Gaussian filter, anisotropic filter, and frequency domain filters, the Nonlocal Means Filter shows a better preservation of structure in a digital image and its method noise is more manageable.

After filtering, reconstructed volumes were segmented into three phases (pores, cement, and mineral grains) based on the grayscale value of voxels (Figure ). These grayscale values reflect X-ray attenuation, which is governed by atomic number and density. , The cement phase contains both matrix and cement as well as microporosity that lies below the voxel resolution and appears in grayscale values between minerals and pore space (Figure ). Each sample exhibits varying amounts of microporosity associated with cement or clay minerals, but because these features are below the scanner’s resolution, we did not directly measure them. Instead, we tested a range of assumed microporosity values to determine their impact on storage capacity, injectivity, and fluid trapping. A quantitative analysis of each phase was then conducted by digitally counting the voxels in pores, cement, and minerals and dividing by the total number of voxels in the filtered volume.

3.

3D display of the filtered volume and segmentation result. In the segmentation result, red represents mineral grains, light blue represents porosity, and dark blue represents cement phases.

4.

Cement phase and microporosity in it. Left image: segmented microtomography data illustrating porosity (light blue) and cement (dark blue) in a 3D display. Right image: SEM image illustrating the cement phase in detail including clay minerals such as kaolinite (Ko) and illite smectite mixed layer (I/S). Both microtomography and SEM analysis are from sample no. 2 (7078 ft).

2.3. Pore Network Model, Permeability, and Fluid-Flow Simulations

An idealized pore network model (PNM) was extracted from segmentation results using the method proposed by references and. A medial-axis algorithm guided by a distance map is used to reduce the pore space to a single voxel skeleton. Each segment of this skeleton is classified as a pore or a throat based on the relationship between its length and an extreme radius. Pores are identified when the extreme radius exceeds the segment length, whereas throats are those with shorter lengths. Geometric properties such as pore radius, pore volume, and throat equivalent hydraulic radius are computed.

Absolute permeability simulations were carried out based on the connected pore space of PNM. The flow of water was calculated by the Navier–Stokes equation using a finite volume solver in PerGeos software.

A two-phase fluid-flow simulation was conducted based on the PNM extracted. In this research, brine was set as the wetting phase, and supercritical CO₂ was set as the nonwetting phase. Initially, the PNM model was filled with a wetting phase fluid. CO₂ injection displaces wetting phase fluid, which is the drainage process. After CO₂ injection, the wetting phase fluid comes back into the CO₂ plume, which is the imbibition process.

The two-phase flow simulation utilizes quasi-steady-state displacement sequences of two phases to simulate the flow within the pore-network model domain. The displacements are computed based on the threshold capillary pressure, which is determined using the energy balance and the Young–Laplace equation. Once the pore fluid reaches a steady state, relative permeabilities for each phase are computed by applying a mass balance to each of the pores in the connected phase. Details of the model used to simulate the two-phase flow can be found in refs. −

The interfacial tension between wetting and nonwetting phase fluids is set at 0.035 N/m following National Institute of Standards and Technology (NIST) guidelines and suggestions made by previous research. The densities of water and CO₂ are set at 1029 and 600 kg m^–3, respectively, under the reservoir temperature and pressure of 120 °C and 27.5 MPa. The water density is elevated due to the high salinity of the groundwater.

The wettability of reservoir rocks is affected by several factors including salinity of the brine, surface mineralogy, pressure, temperature, surface roughness, and presence of organic matter. ,, Quartz and other silicas are generally considered water-wet; calcite is usually strongly or weakly water-wet; while clay minerals present a wide spectrum of wettability. Several methods have been used to alter the surface wettability, including increasing salinity, extended exposure to the nonwetting phase (often referred to as “aging”), increasing pressure, and decreasing temperature (e.g., − ). To test the influence of wettability on two-phase flow, three wettability conditions were tested, including water-wet, weak water-wet, and neutral-wet. The CO₂–brine contact angle on the reservoir rock surface has been measured in several studies. In this work, three groups of results were selected. For the water-wet condition, results from both Want et al. and Farokhpoor et al. indicate the contact angle being 0–20°. , For the weak water-wet condition, Haeri et al. measured 298 sandstone samples from diverse geologic backgrounds and summarized 40°–60° of contact angles being weakly water-wet. For the neutral-wet condition, contact angles were set as 80–100° following Yang et al. results based on measurements under reservoir pressure–temperature conditions using brine and core samples from Weyburn oil field.

The relationship between initial saturation and residual saturation of the nonwetting phase has been extensively studied, which is often referred to as the initial-residual (IR) curve or trapping model (e.g., − ). In this work, two trapping models were selected to interpret the trapping behavior. The Land model is widely accepted for water-wet porous media (eq ).

$$S_{{\mathrm{CO}}_2,r}=\frac{S_{{\mathrm{CO}}_2,i}}{1+C·S_{{\mathrm{CO}}_2,i}}$$ 1

where S _CO2,i is the initial saturation of CO₂, S _CO2,r is the residual saturation, and C is the Land trapping coefficient (C ≥ 0). When C0, it implies that all nonwetting phase is trapped, while as C increases, less nonwetting phase is trapped. In addition, the Spiteri model is selected since it fits better for intermediate and mixed-wet systems (eqs and ).

$$S_{{\mathrm{CO}}_2,r}=\alpha S_{{\mathrm{CO}}_2,i}-\beta ·S_{{\mathrm{CO}}_2,i}^2$$ 2

$$0\le \alpha \le 1,\beta \ge 0$$ 3

where α and β are parameters that define the initial slope and the curvature of the trapping model.

2.4. Routine Core Analysis and Scanning Electron Microscopy

The petrophysical properties, including porosity and permeability, were measured using standard core analysis techniques on cylindrical core plugs extracted from the whole core. Prior to testing, each plug was cleaned with organic solvents and oven-dried to remove residual fluids. Porosity was calculated using helium pycnometry by measuring the grain volume and dividing by the plug’s bulk volume. Permeability measurements were performed using a steady-state or pulse-decay apparatus. Klinkenberg correction was applied to correct any gas-slippage effect. Scanning electron microscopy (SEM) analysis was also carried out to visualize the microporosity.

3. Results

3.1. Data and Simulation Model Validations

To validate the PNM and simulations, the relative permeabilities of the wetting phase (brine) and nonwetting phase (supercritical CO₂) were simulated for both drainage and imbibition, and then compared with published laboratory data (special core analysis). For the drainage process (Figure ), Benson et al. compiled measurements from several studies, the majority of which were conducted under supercritical CO₂ conditions. Gao et al. conducted special core analyses on Berea Sandstone samples under two different pressure–temperature conditions (12.4 MPa (1800 psi) and 30 MPa (4350 psi), both at 77 °C) (Figure ). The simulated relative permeabilities for both brine and CO₂ agree well with the previously reported measurements. Figure shows the results for imbibition, where experimental data for supercritical CO₂ relative permeability remain limited. Gao et al. provided such data at 12.4 MPa (1800 psi) and 77 °C using Berea Sandstone. The simulated results also match these laboratory measurements closely, further supporting the model’s validity.

5.

Simulated drainage CO₂ and water relative permeability data for the Stevens Sandstone (sample #6) in comparison with published data. Data is also plotted from Gao et al. work on Berea Sandstone. Circles represent data collected under 12.4 MPa (1800 psi), while triangles are for data collected under 30 MPa (4350 psi). The image background is cited from Benson et al. in which all of the available CO₂–brine drainage relative permeability curves were summarized. Blue shaded curves are brine saturation while red shaded curves are CO₂ saturation. ,

6.

Simulated imbibition CO₂ and water relative permeability data for the Stevens Sandstone (sample #6) in comparison with lab-measured data from Gao et al. Gao et al. data was collected under 12.4 MPa (1800 psi) using Berea Sandstone.

3.2. Porosity, Cement, and Permeability

The results of porosity, cement, and permeability obtained from the 3D micro-CT scans are summarized in Table . In each volume, each phase is defined by the ratio of that phase’s voxels to the total voxel count. The measured porosity includes isolated pores and connected pores, ranging from 8.65 to 13.26%, while the simulated absolute permeability ranges from 0.21 to 313.95 md. From a petrological perspective, the “cement phase” in this research includes a matrix, microporosity, and cement. Microporosity, which forms through detrital or authigenic clay minerals and clay-sized mineral grains, generally occurs below the resolution of micro-CT, making accurate segmentation unfeasible. To address this limitation, a sensitivity test was conducted to assess the microporosity’s contribution to the pore network. Four scenarios of microporosity percentage were examined, including 0, 10, 25, and 50%. The results reveal that while including microporosity markedly increases total connected porosity, it has only a minor effect on permeability (Table ). Consequently, microporosity can enhance CO₂ storage but offers negligible benefits for improving CO₂ injection rates. Results in Figure and Table are based on 50% of the cement phase being porous. All six tested samples contained a cement phase ranging from 6.78 to 12.33% (Table ).

1. Volume Fraction of the Porosity and Cement Phases of Each Sample .

Sample # and depth (ft)	porosity RCA (%)	permeability RCA (md)	porosity CT (%)	cement CT (%)	porosity and MP (%)	permeability CT (md)
#1 6767	16.6	3.77	8.65	9.13	13.215	1.32
#2 6822	19.3	271	12.96	8.97	17.445	313.95
#3 7078	17.8	7.06	13.26	8.83	17.675	10.82
#4 7157	21.6	17.4	10.94	12.33	17.105	20.38
#5 7214	16.5	38	10.93	8.18	15.02	35.79
#6 7277	15.3	0.551	11.25	6.78	14.64	0.21

^aMicroporosity (MP) is calculated assuming 50% of the cement phase is porous.

2. Connected Porosity and Permeability Based on Different Microporosity Percentages.

	connected porosity				permeability
sample	0 MP	10 MP	25 MP	50 MP	0 MP	10 MP	25 MP	50 MP
#1	4.44	5.76	7.78	11.14	1.26	1.26	1.27	1.32
#2	12.76	13.64	14.97	17.18	310.34	310.49	311.37	313.95
#3	8.06	8.74	9.77	10.47	10.29	10.32	10.47	10.82
#4	10.54	11.78	13.64	16.74	19.07	19.19	19.56	20.38
#5	10.45	10.86	11.48	12.50	35.94	35.49	35.56	35.79
#6	0.00	0.00	0.00	12.64	0.00	0.00	0.00	0.21

7.

Comparisons of porosity and permeability results from routine core analysis and micro-CT analysis.

No obvious correlation between the porosity and permeability, such as that proposed by the Kozeny–Carman equation, was observed in either the routine core analysis (RCA) or the micro-CT results. This discrepancy likely arises because classic correlations assume relatively simple pore geometries such as pipe conduits, whereas the pore systems in these samples exhibit considerably more complex geometry and connectivity.

Both the porosity and the total porosity (porosity plus microporosity) are lower than those measured by RCA. This discrepancy in porosity arises primarily from the voxel size (2.8–3.3 μm in this research) and spatial resolution of the micro-CT technique, which cannot adequately capture pores or gains smaller than the voxel size.

3.3. Pore Network Model and the Size Distribution of Pores and Pore Throats

The filtered 3D scan data and extracted pore network from each sample are shown in Figure . Pore space is represented by spheres (pore bodies) and cylinders (pore throats), with radii proportional to pores’ cross-sectional radii and color coded according to pore-body radius from 0 to 60 μm. Each scanned volume spans approximately 7.76 mm³. Pore size distribution reveals important differences among layers of rocks. Samples #1 and #5 show similar pore size distribution, with 85% of the pores smaller than 14 and 11 μm, respectively. Samples #2 and #3 also show similar pore size distribution, with 85% of the pore smaller than 17 and 13 μm. Sample #4 shows larger pore sizes with 85% of the pores’ cutoff at 19 μm. Sample #6 has the smallest pore sizes, with 85% of the pores smaller than 6 μm. The pore throat sizes show more similarities than differences between samples. Most of the pore throats reside within a size range of 3–8 μm, with one exception #4, of which the throat sizes are larger and more evenly distributed.

8.

Pore and throat size distributions.

Figure illustrates the 3D rendering of all the 6 samples and their pore network models. The porosity, along with pore (spheres) and throat (blue sticks) size distributions, can be directly visualized in these 3D data sets, revealing evident differences among the samples, and layers of rocks.

9.

Idealized pore network model of each sample, in which blue sticks represent the pore throat while spheres represent the pore body. Pore bodies are color coded according to their sizes from 0 (purple) to red (60 μm).

Sample #1 contains large but sparsely distributed pores (yellow and orange spheres). Many of these pores are disconnected, leading to isolated pore bodies and partially connected small networks. In contrast, samples #2–#5 display well-connected pore networks. Particularly, sample #2 contains many large size (60 μm) pores (red spheres), whereas samples #3 and #4 have moderately sized (30 μm) pores, and sample #5 hosts smaller but more densely distributed pores. Sample #6 presents the smallest pores, many of which function as throats rather than bodies, with a substantial portion remaining unconnected. Furthermore, these pores occur predominantly on one side of the sample, rendering the opposite side effectively nonporous.

These observations are corroborated by the porosity profiles shown in Figure . Sample #1 exhibits uniformly low porosity throughout the Z-direction, whereas sample #6 shows low porosity at one end and high porosity at the other. By contrast, samples #2–#5 have more evenly distributed porosity values.

10.

Porosity profiles of the Z-direction of the three-dimensional data sets.

3.4. Two-Phase Flow Simulations

Results from the PNM simulations of the drainage-imbibition scenarios for CO₂/brine in Stevens Sandstone are shown in Figure and Table , including initial CO₂ saturation at the end of primary drainage and residual CO₂ saturation after imbibition. Three wettability conditions with four microporosity levels were simulated. For samples #1–#5, the initial CO₂ saturation ranges from 0.5089 to 0.9520 at no microporosity, from 0.3899 to 0.9057 at 10% of microporosity, from 0.2886 to 0.8570 at 25% of microporosity, and from 0.2014 to 0.7865 at 50% of microporosity. For sample #6, there was no connected pore network until the cement phase reached 50% microporosity when the initial CO₂ saturation was 0.4287. The residual CO₂ saturation results were significantly influenced by the wettability of the pore network. Residual CO₂ saturation ranges from 0.1729 to 0.6424. Overall, water-wet and intermediate-wet samples both show higher residual CO₂ saturation compared with neutral-wet samples. With increasing microporosity, decreasing trends in the initial and residual CO₂ saturations were observed (Figure ). Land and Spiteri models were fitted to the initial and residual saturations obtained by simulations (Figure and Table ).

11.

Initial and residual CO₂ saturation resulting from PNM simulations of drainage and imbibition for a CO₂/brine system in each sample. Simulations were carried out under wet water (A), weak wet water (B), and neutral wet water (C) with different levels of microporosity contributions (0, 10, 25, and 50%) to the whole pore network.

3. Results from Two-Phase Flow Simulations Using PNM, Based on Different Microporosity (MP) Levels and Wettability Conditions.

samples	Si_Water-Wet	Sr_Water-Wet	Si_Weak Water-Wet	Sr_Weak Water-Wet	Si_Neutral Wet	Sr_Neutral Wet
#1_0MP	0.5089	0.4368	0.5089	0.4627	0.5089	0.5088
#2_0MP	0.925	0.5319	0.925	0.5868	0.925	0.4142
#3_0MP	0.7911	0.6381	0.791	0.6176	0.791	0.5546
#4_0MP	0.7781	0.6027	0.7721	0.5893	0.7781	0.6424
#5_0MP	0.952	0.6318	0.9519	0.5943	0.952	0.4049
#6_0MP	0	0	0	0	0	0
#1_10MP	0.3899	0.3347	0.3899	0.3447	0.3899	0.3898
#2_10MP	0.8651	0.4974	0.8651	0.5487	0.8651	0.3874
#3_10MP	0.7295	0.5884	0.7294	0.5695	0.7294	0.509
#4_10MP	0.5393	0.5393	0.6909	0.5175	0.6909	0.5706
#5_10MP	0.9057	0.6038	0.905	0.5769	0.9057	0.3894
#6_10MP	0	0	0	0	0	0
#1_25MP	0.2886	0.2477	0.2886	0.2552	0.2886	0.2886
#2_25MP	0.7885	0.4534	0.7884	0.5001	0.7885	0.3531
#3_25MP	0.6532	0.5269	0.6532	0.5096	0.6532	0.4579
#4_25MP	0.6013	0.4658	0.5967	0.4554	0.5967	0.435
#5_25MP	0.857	0.5713	0.8563	0.5458	0.857	0.3684
#6_25MP	0	0	0	0	0	0
#1_50MP	0.2014	0.1792	0.2014	0.1718	0.2014	0.1514
#2_50MP	0.687	0.5018	0.687	0.4358	0.687	0.3076
#3_50MP	0.5562	0.4487	0.5562	0.4339	0.5562	0.3899
#4_50MP	0.49	0.3795	0.4862	0.3542	0.4867	0.3285
#5_50MP	0.7865	0.5243	0.7859	0.5009	0.7865	0.3381
#6_50MP	0.4287	0.3829	0.4287	0.3571	0.4287	0.3087

12.

Initial and residual CO₂ saturations assuming 50% of the microporosity contribution.

4. Fitted Coefficients for Theoretical Trapping Models.

model	coefficients
land	C = 0.55
Spiteri 1	α = 0.85; β = 0.27
Spiteri 2	α = 0.99; β = 0.69

4. Discussion

Using 3D image analysis of micro-CT scanned core samples and pore network modeling, this study significantly advanced our understanding of CO₂ injection and residual trapping in Stevens Sandstone in the San Joaquin Basin of California. Key physical properties of the reservoirs and fluid behavior were characterized under a variety of microporosity and wettability conditions, including porosity, permeability, pore and throat size distributions, and initial and residual CO₂ saturations.

4.1. Porosity and Cement Phases Derived from Micro-CT Scans

Compared with other bulk measurement methods such as helium porosimeters, this approach cannot detect microporosity below the voxel resolution, which is approximately 3 μm in this research. Previous works ,− have similarly noted the presence of microporosity in sandstones and carbonates that remains unresolved by micro-CT. The majority of the microporosity resides in the cement phase, primarily hosted by the clay-sized particles. We did not assign a grayscale value to microporosity or designate a fixed fraction of cement as microporosity due to the lack of an accurate method. However, this limitation likely does not affect subsequent modeling and simulation because CO₂, as a nonwetting phase, preferentially flows through the largest accessible pores, which is well characterized in the micro-CT data.

The porosity and permeability show no obvious correlation, whereas pore and throat size distributions have a greater influence on the permeability. For instance, sample #1 exhibits sparse and partially connected pore networks, which yield low permeability, while sample #6 contains the smallest pores and throats, leading to the lowest permeability overall. In contrast, samples #2 through #5 are more permeable; among them, sample #2 hosts the largest, well-connected pores, resulting in the highest permeability (Figure ).

Sample #6 contains the smallest volume of cement (6.78%), while sample #4 contains the highest cement volume (12.33%) (Table ). The effect of the cement phase on the CO₂ injectivity is complex, depending on the microporosity, mineral composition, and spatial distribution of the cement. Micropores associated with the cement phase could potentially enhance permeability by providing additional flow paths, and they might also increase capillary trapping due to their narrow pore geometry. However, injected CO₂ dissolves in brine to form carbonic acid. The acid promotes the dissolution of carbonates, feldspars, iron oxides, and clay minerals. , These reactions reshape the pore network, altering permeability and porosity, and mobilize fine particles (fines) that can block pore throats and restrict flow. , Core-flooding experiments with CO₂-saturated brine have documented permeability declines of 15–30% because of fines migration. Subsequent flow-through and batch reaction studies in multiple sandstone reservoirs confirmed the mobilization of clays, quartz grains, and cement, and their pore-clogging effects under reservoir pressure–temperature conditions. − Owing to the spatial and temporal limitations of laboratory experiments, numerical simulations have become essential for evaluating fines-migration impacts on reservoir properties. , Recent advances in reactive-transport modeling provide additional insight and several codes, including CrunchFlow, TOUGHREACT, and PHREEQC, are widely used to simulate coupled geochemical and flow processes in porous media.

The significant heterogeneity in petrophysical properties in Stevens Sandstone has been noticed by a number of previous works. − Many of the Stevens Sandstones that exhibit very low permeability (<1 md) actually have relatively high porosity, sometimes as high as 20–25%. In these rock layers, permeability is significantly reduced by the cement and matrix (primarily clay minerals and clay-sized minerals) that block pore throats and exponentially increase the surface area within the pore network. Krystinik noticed that clay volumes occupying less than 20% of the pore space could cause severe permeability loss. Microporosity, in this case, is the intricate network of micropores and interconnected authigenic clays that enclose, partition, and fill the larger pore spaces.

Previous research had found that the reservoir quality of Stevens Sandstone has been considered strongly tied to the compartments marked by sequence stratigraphic stacking patterns. , Retrogradational compartments are generally better reservoirs compared with the underlying aggradational and/or progradational compartments with higher porosity and permeability. The better reservoir quality is attributed to the increased reworking of nearshore sand bodies in conjunction with a lower influx of mud during base-level rise. In addition, the retrogradational compartments were sourced from clean sand that had gone through reworking in nearshore environments, while aggradational and progradational compartments were sourced from clay-rich sand from fluvial systems. Since the majority of the cement phase observed in this research is contributed by clay or clay-sized particles, the retrogradational compartments are more likely to be of better injectivity.

4.2. Effect of Microporosity

The effect of microporosity was investigated by performing a sensitivity analysis in which 0, 10, 25, or 50% of the measured cement phase was assigned to each of the six sandstone samples. The results indicate that microporosity can account for a significant fraction of the total pore volume (average 40.76%), thereby enhancing the overall CO₂ storage capacity of the reservoir (Table ). However, its impact on permeability was generally modest, except in the least permeable sample (sample #6, 0.21 md) (Table ). This suggests that while microporosity can increase the pore space available for fluid storage, it does not substantially improve injectivity in most cases. The small pore-throat size associated with microporosity contributes minimally to fluid transport, as flow tends to be dominated by larger macropores.

To further assess the influence of microporosity on CO₂ injection and trapping, two-phase flow simulations (CO₂ as the nonwetting phase, and brine as the wetting phase) were performed under varying degrees of water-wetness and microporosity. Under both strongly water-wet and weakly water-wet conditions, a higher microporosity led to lower initial and residual CO₂ saturations. The reduction in initial CO₂ saturation was generally more pronounced than the reduction in residual saturation. In the neutral-wet scenario, increasing microporosity also decreased both initial and residual CO₂ saturations but the reduction in residual saturation was more pronounced than under water-wet conditions, reflecting the higher fluid mobility in a less water-wet system.

Overall, these findings highlight that microporosity’s primary effect is to increase total storage volume rather than to facilitate flow. Because smaller pore throats dominate microporous regions, CO₂ injection initially favors macropores before gradually invading microporous regions as capillary forces increase. This leads to a progressive decrease in the initial CO₂ saturation as microporosity increases. Once CO₂ enters the micropores, however, it tends to become physically disconnected from the flow network, forming ganglia and increasing the potential for trapping. As a result, the reduction in residual CO₂ saturation is generally less dramatic than the drop in initial saturation under strongly and weakly water-wet conditions; in a neutral-wet system, enhanced fluid mobility contributes to a more substantial decrease in the residual CO₂ saturation.

Based on the comparison between porosity and permeability values measured by routine core analysis and micro-CT (Table and Figure ), the realistic estimation of microporosity is 50% of the cement phase being microporous.

4.3. Estimated Initial-Residual Saturation Curves

Two theoretical trapping models were fitted to the initial-residual CO₂ saturations simulated from PNM assuming 50% microporosity contribution, including the Land and Spiteri models (Figure ). The fitted coefficients are listed in Table . Land model provided a better fit for the water-wet simulations, with a Land trapping coefficient of 0.55. In contrast, the weak water-wet simulations, while following a similar trend to the strongly water-wet case, exhibited lower residual trapping and were better fitted by the Spiteri model with a trapping coefficient of α = 0.85 and β = 0.27. Under neutral-wet conditions, the Spiteri model (α = 0.99; β = 0.69) was able to capture the nonmonotonic relationship between residual saturation and initial saturation, indicating that changes in wettability strongly influence residual trapping behavior (Figure and Table ).

From a two-phase flow perspective, which fits into the case for most of the CO₂ geologic storage in saline aquifers, when the system is strongly water-wet, capillary forces favor the disconnection of CO₂ into isolated ganglia upon brine imbibition, leading to higher residual CO₂ saturations. As the wettability shifts toward a neutral state, CO₂ retains greater mobility, making it less likely to be trapped by capillary forces, thus lowering residual saturations. This effect becomes more pronounced at higher initial saturations, which often coincide with higher permeability (Figure ). In such scenarios, larger and better-connected pores drain first and are less prone to retaining disconnected CO₂ during subsequent imbibition. Consequently, although microporosity can contribute to the overall storage capacity, the dominant wettability effects and pore connectivity at the macropore scale largely determine the ultimate distribution of residual CO₂ in these rock layers.

4.4. Heterogeneity, Turbidite, and Implications for Field-Scale CO₂ Storage

The inherent heterogeneity of Stevens Sandstone, or turbidite reservoirs more broadly, has been reported by previous research. The reservoir’s physical properties, such as pore size, throat size, and porosity distribution, differ among layers, leading to considerable variations in permeability. In the high-permeability samples, initial CO₂ saturations can reach 0.7865 or 0.6870, with corresponding residual CO₂ saturations of 0.5243 and 0.5018 (Table ). For low and medium permeability layers, initial CO₂ saturation appears to be the primary control of residual trapping capacity. However, in high-permeability layers, wettability becomes the dominant factor (Figure ). Under strongly or weakly water-wet conditions, more CO₂ is immobilized through residual trapping, while in a neutral-wet system, the CO₂ bubbles exhibit higher mobility.

The results here have two main implications. On the storage capacity side, published field-scale models often estimate reservoir pore space using homogeneous reservoir properties (e.g., ,, ). However, the microscopic structural variability in different layers can produce wide-ranging initial CO₂ saturation (0.2014–0.7857, Table ), thus resulting in layers of CO₂ flow rather than evenly distribution of CO₂ (Figure ). On the trapping side, structural trapping is likely the principal mechanism. Yet for heavily drilled regions such as depleted oil and gas fields, residual trapping offers critical additional security against leakage through preexisting wellbores. Previous research has found that heterogeneities in reservoir properties can enhance residual CO₂ trapping. ,, Layers like sample #6 essentially act as internal barriers that block buoyancy-driven flow, dispersing migrating CO₂ and improving overall containment.

13.

Illustration of the layered flow of CO₂ resulted from substantial heterogeneity in reservoir properties.

In summary, these results demonstrate the strong potential of the Stevens Sandstone in the Central Valley of California for CO₂ geologic storage. However, detailed characterization of the reservoir heterogeneities is essential to accurately quantify both the storage capacity and the fluid mobility in such complex systems.

5. Conclusions

In this study, micro-CT analysis from 6 representative sandstone samples was carried out to evaluate the fundamental controls on CO₂ injectivity and residual trapping in Stevens Sandstone of the San Joaquin Basin of California, USA. Three-dimensional pore networks were extracted from processed data sets, along with measurement of pore and throat size, and porosity distributions. Permeability and two-phase flow were simulated based on PNM. The impacts of microporosity and wettability variations on CO₂ injection and trapping were quantified. Limited by the spatial resolution of micro-CT, this research took a fractional approach to estimate the contribution of microporosity to the entire pore network. Other analytical techniques such as nuclear magnetic resonance (NMR) and direct observation methods such as focused-ion-beam SEM (FIB-SEM) or broad-ion-beam SEM (BIB-SEM) have shown great detection power for the nanometer scale features (e.g., − ). Yet it is still challenging to achieve high resolution/magnification and large-scale observation simultaneously. Based on this work, the following conclusions can be drawn.

• 1.The results confirm that microporosity can contribute substantially to total pore volume (up to 56%, sample #4), enhancing overall storage capacity, while exerting less influence on permeability.
• 2.Wettability is the key factor that determines CO₂ mobility, especially in high-permeability samples. Under strongly or weakly water-wet conditions, capillary forces favor greater CO₂ trapping, whereas a neutral-wet system leads to more mobile CO₂ and lower residual saturations.
• 3.The inherent heterogeneity of turbidite reservoirs, such as Stevens Sandstone, underscores the importance of characterizing fine-scale heterogeneity, particularly the distribution of cement and clay-size particles, when evaluating CO₂ storage in sandstone reservoirs. The variability of reservoir properties and pore structure observed in this study resulted in a wide range of initial CO₂ saturations and residual CO₂ saturations. The injected CO₂ flow will likely be stratified in favor of the high-permeability layers.
• 4.Overall, the Stevens Sandstone shows significant potential for geologic CO₂ storage in California. However, a detailed reservoir characterization is required to accurately predict both injectivity and trapping efficiency at field scale.

The author declares no competing financial interest.