Numerical Investigation of Evolution of Reservoir Characteristics and Geochemical Reactions of Compressed Air Energy Storage in Aquifers

Abstract

Compressed air energy storage in aquifers presents a promising approach for large-scale energy storage, yet its implementation is complicated by geochemical reactions, such as pyrite oxidation, which can impact reservoir integrity and operational efficiency. This study numerically investigates the evolution of reservoir characteristics and geochemical processes during CAESA operations to address these challenges. Using the TOUGHREACT simulator, we developed one-dimensional and two-dimensional reactive transport models based on the Pittsfield aquifer field test parameters to simulate coupled thermal-hydrological–chemical processes under varying injection rates, temperatures, reservoir depths, and operational cycles. The results demonstrate that higher injection rates induce greater near-well pressure buildup and extended thermal zones, while deeper reservoirs exhibit abrupt declines in pressure and gas saturation due to formation constraints. Geochemical analyses reveal that pyrite oxidation dominates, leading to oxygen depletion, groundwater acidification (pH reduction), and secondary mineral precipitation, such as goethite and hematite. These findings underscore the critical interplay between operational parameters and geochemical reactions, highlighting the need for optimized design to ensure long-term stability and efficiency of aquifer-based energy storage systems.

1. Introduction

To address the regulation and storage of traditional and renewable energy in electricity supply and demand, the importance and necessity of electricity energy storage (EES) systems have been increasingly recognized. Among various storage technologies, pumped hydro energy storage (PHES) and compressed air energy storage (CAES) are the most practical for large-scale and long-duration applications [1,2,3,4]. PHES provides high storage capacity, proven reliability, and long operational life, but is limited by site-specific requirements, environmental impacts, and low energy density [5]. CAES offers an alternative by storing compressed air in geological formations and releasing it through turbines during peak demand, and global assessments indicate it is a promising, scalable energy storage technology with broad regional potential [6,7].

The concept of CAES was proposed in the 1940s [8], and after several decades of research and development, the technology has been successfully applied in practical power systems. This underground long-term and large-scale gas storage is generally implemented in two principal types of geological formation: caverns (such as salt caverns and hard rock caverns) and porous media (such as depleted hydrocarbon reservoirs and aquifers) [4,9,10]. The most prominent commercial applications of CAES are the Huntorf plant in Germany and the McIntosh plant in the USA, both of which utilize underground salt caverns for air storage and have demonstrated sustained operational reliability in grid-scale energy management [11]. However, the limited availability of caverns and depleted reservoirs constrains the deployment of CAES, whereas aquifers—widely distributed and proven effective for gas storage—offer a promising alternative that could substantially alleviate these geological restrictions [12,13,14].

Compressed air energy storage in aquifers (CAESA) operates by injecting compressed air into deep porous formations saturated with water, where the air displaces water and forms a storage bubble contained by the surrounding hydrostatic pressure and impermeable caprock. During discharge, the stored air is withdrawn and expanded through turbines to generate electricity, while the porosity and permeability of aquifers enable large-scale storage capacity [13,15]. To date, the feasibility of CAESA has been demonstrated through the previous field tests and comprehensive numerical simulations. The Pittsfield CAESA field test in the USA confirmed that air can be injected to form a large gas bubble within the aquifer and withdrawn at rates suitable for energy recovery [16]. Kushnir et al. [17] presented a theoretical investigation of compressible gas flow in aquifer reservoirs for CAES systems. Simplified expressions for periodic air pressure distribution were derived and applied to determine well pressure and air–water interface stability. The findings provide essential insights for evaluating reservoir behavior and optimizing CAES design in aquifer formations. Oldenburg and Pan [9] simulated 100 daily cycles in a depleted hydrocarbon reservoir and demonstrated that CAES in porous media can achieve efficient energy storage while maintaining reservoir integrity, thereby confirming its technical feasibility for practical application. Guo [18] conducted numerical simulations and demonstrated that integrating aquifer thermal energy storage with CAESA maintained high recovery efficiency while enhancing storage capacity, with system performance being most sensitive to reservoir boundary permeability. Li et al. [19,20,21,22] demonstrated through coupled wellbore–aquifer CAESA models and field-scale simulations that aquifer heterogeneity, layering, and wellbore design critically affect air–water flow, pressure stability, and energy recovery, providing key guidance for the design, site evaluation, and large-scale implementation of CAESA systems. Although CAESA has not yet been commercially implemented, the aforementioned tests and simulation studies indicate that aquifer reservoirs represent a viable and promising medium for large-scale CAES [17,19].

This study focuses on pyrite oxidation in CAESA systems, as pyrite (FeS₂, iron disulfide) is the most abundant and widely distributed sulfide mineral in the Earth’s crust and plays a critical role in geochemistry and environmental processes [23,24]. Pyrite oxidation has been widely investigated, primarily in the context of mining. It generates sulfuric acid, which alters primary minerals, produces secondary assemblages, mobilizes metals, and causes acid mine drainage, acid sulfate soils [25,26,27,28,29], and aquifer contamination [30], while also providing critical insights for understanding its broader geochemical implications in subsurface energy applications. In CAESA systems, the injection of air containing oxygen into deep aquifer reservoirs may cause complex geochemical reactions with potential implications for groundwater quality, while also giving rise to technical challenges such as metal corrosion in wellbores and structural damage to rock formations [31]. Reports from the Pittsfield CAES site indicated that oxygen depletion in stored air was primarily caused by reactions with sulfide and clay minerals [32]. Laboratory investigations further demonstrated that pyrite, marcasite, illite, and smectite deplete oxygen through aqueous oxidation, with pH exerting a strong influence on mineral reactivity [33]. Aghababaei and Sedae [34] demonstrated that oxygen from air injection in a CAES system induces mineral transformations (pyrite oxidation, calcite dissolution, anhydrite precipitation). Guo et al. [15] also highlighted that oxygen depletion linked to seasonal operations in CAESA could be severe, significantly reducing energy efficiency. Because of insufficient oxygen in unreplenished storage air, diabatic CAES operation may be impaired by limited combustion for heating the expanding gas [35]. These findings reveal the impact of pyrite reactivity and oxygen consumption in CAESA systems.

Based on the analysis of previous literature, we found that these studies primarily focused on the hydraulic and thermal behaviors of CAESA systems, as well as operational strategies, with the majority being model-based investigations. Due to constraints in experimental scale, cost, and technology, field tests for CAESA have been scarce. Furthermore, the coupled geochemical processes induced by air injection—particularly pyrite oxidation and its impacts on oxygen consumption, mineral transformations, and groundwater quality—have received less attention. These processes not only affected the long-term chemical stability of the reservoir but also directly influenced the operational efficiency of diabatic CAES facilities by altering the oxygen content in the stored air. Currently, the limited existing research is almost based on the Pittsfield field CAESA test from the 1980s, and our study was also predicated on site-specific data from Pittsfield.

Compared with previous numerical studies, we have conducted a systematic geochemistry analysis of the study area. In this study, we developed a series of one-dimensional and two-dimensional reactive transport models using the TOUGHREACT simulator to quantitatively investigate the evolution of reservoir characteristics and geochemical reactions during compressed air injection and storage in aquifers. Special attention was devoted to elucidating the mechanisms and spatial–temporal evolution of pyrite oxidation and its associated secondary mineral formation under various operational conditions. The results aimed to advance the understanding of oxygen–mineral interactions in CAESA systems and provided theoretical guidance for the design, optimization, and sustainable operation of aquifer-based compressed air energy storage projects.

2. Model Setup

2.1. Background

The Pittsfield aquifer test field is near Pittsfield, IL, USA, ~145 km northwest of St. Louis, within the dome of the Pittsfield doubly plunging anticline. Topographically, it hosts a shallow, confined aquifer where the St. Peter sandstone reservoir lies at ~200 m depth, capped by low-permeability rocks like Joachim dolomite. Stratigraphically, layers include Pleistocene glacial drift/alluvium, Mississippian cherty limestone, Silurian shale/limestone, and Ordovician Maquoketa, Galena, Platteville, Joachim, and St. Peter formations—St. Peter sandstone is the main reservoir (18% porosity, ~750 md permeability). A CAES in aquifer field test, led by PNL and partners, aimed to validate compressed air storage in aquifers. It covered air injection, storage, cycling, and assessment of thermal effects, water coning, etc. [16]. Figure 1 presents a conceptual diagram of CAESA. Specifically, the investigation focused exclusively on the subsurface reservoir domain, while surface infrastructure such as compressors, motor–generators, and turbines was not considered.

The St. Peter Sandstone, a permeable quartz-dominated formation, underlies the impervious Galena–Platteville–Joachim carbonate cap rock complex. Core analyses confirmed that the cap rock exhibited sufficiently low permeability to confine compressed air throughout the duration of the field test, ensuring reservoir integrity. The St. Peter Sandstone, with a total thickness of at least 25 m, was subdivided into three stratigraphic sub-layers: the Green Layer (3 m), White Layer (5.5 m), and Gray Layer (16.5 m) [36]. This study was based on the stratigraphic conditions and parameters of the Pittsfield aquifer field test. Since the primary focus was on geochemical reactions, the model neglected variations in surface topography, and the St. Peter Sandstone was simplified as a horizontal formation.

2.2. Numerical Method

TOUGHREACT V3.32, developed by the Lawrence Berkeley National Laboratory (LBNL) under the U.S. Department of Energy, is a numerical simulation code designed to model coupled non-isothermal multiphase flow, heat transport, and geochemical reactions in porous and fractured geological media. Written in FORTRAN with OpenMP-based parallelization, it effectively handles large-scale simulations involving complex thermal–hydrological–chemical (THC) processes. The software integrates reactive transport mechanisms such as mineral dissolution and precipitation, gas–liquid–rock interactions, ion exchange, and redox reactions, allowing users to investigate the dynamic evolution of subsurface environments. TOUGHREACT supports one-, two-, and three-dimensional modeling and employs a comprehensive thermodynamic database that enables simulations over a broad range of temperatures, pressures, and ionic strengths. It also provides multiple “Equation-of-State” modules (e.g., EOS1, EOS2, EOS3, ECO2N) to represent different fluid systems, including water, air, CO₂, and brine mixtures. The program has been extensively applied in fields such as CO₂ geological storage, geothermal energy development, nuclear waste disposal, acid mine drainage remediation, and groundwater geochemical evolution. By coupling multiphase flow with reactive chemistry, TOUGHREACT provides a robust and versatile computational framework for analyzing complex interactions among thermal, hydrological, and chemical processes in natural and engineered subsurface systems.

Table 1. Governing equations for fluid and heat flow, and chemical transport of TOUGHREACT.

Component	Equation
	Mass/Heat Term	Flux Term	Source and Sink Term
Water	$M_w=\varphi (S_l{\rho }_l{X}_{\mathit{wl}}+S_g{\rho }_g{X}_{\mathit{wg}})$	$F_w={\mathit{u}}_l{\rho }_l{X}_{\mathit{wl}}+{{\mathit{u}}_g\rho }_g{X}_{\mathit{wg}}$	$q_w=q_{\mathit{wl}}+q_{\mathit{wg}}$
Air	$M_c=\varphi (S_l{\rho }_l{X}_{\mathit{cl}}+S_g{\rho }_g{X}_{\mathit{cg}})$	$F_c={\mathit{u}}_l{\rho }_l{X}_{\mathit{cl}}+{{\mathit{u}}_g\rho }_g{X}_{\mathit{cg}}$	$q_c=q_{\mathit{cl}}+q_{\mathit{cg}}+q_{\mathit{cr}}$
Heat	$M_h=\varphi \left(S_l{\rho }_l{U}_l+S_g{\rho }_g{U}_g\right)+(1-\varphi ){\rho }_s{U}_s$	$F_h={\displaystyle \sum _{\beta =l, g}}{h}_{\beta }{\rho }_{\beta }{\mathit{u}}_{\beta }-\lambda \mathsf{\nabla }T$	$q_h$
Chemical component j	$M_j=\varphi S_l{C}_{\mathit{jl}}$	$F_j={\mathit{u}}_l{C}_{\mathit{jl}}-(\tau \varphi S_l{D}_l)\mathsf{\nabla }{C}_{\mathit{jl}}$	$q_j=q_{\mathit{jl}}+q_{\mathit{jg}}+q_{\mathit{js}}$

lists the governing equations for fluid and heat flow, and chemical transport of TOUGHREACT [37].

The TOUGHREACT simulator calculates the rates of mineral dissolution and precipitation based on the Lasaga equation [37]:

$${\mathrm{r}}_{\mathrm{n}}=\pm \mathrm{k}{\mathrm{A}}_{\mathrm{n}}{\left|1-{\Omega }_{\mathrm{n}}^{\theta }\right|}^{\eta }$$

(1)

where ${\mathrm{r}}_{\mathrm{n}}$ is a kinetic rate of mineral n (mol/s), A_n is the specific reactive surface area (m²/kg·H₂O), and Ω_n is the kinetic mineral saturation ratio. The empirical parameters θ and η are determined from experiments or usually taken as 1. The temperature dependence of the rate constant k (mol·m⁻²·s⁻¹) is expressed using Arrhenius equation [37]:

$$\mathrm{k}={\mathrm{k}}_{25}\exp \left[{\displaystyle \frac{-{\mathrm{E}}_{\mathrm{a}}}{\mathrm{R}}}\left({\displaystyle \frac{1}{\mathrm{T}}}-{\displaystyle \frac{1}{298.15}}\right)\right]$$

(2)

where E_a (J/mol) is the activation energy, k₂₅ (mol·m⁻²·s⁻¹) is the rate constant at 25 °C, R (8.314 J·mol⁻¹·K⁻¹) is the universal gas constant, and T (K) is the absolute temperature. The kinetic rate constant k in Equation (2) primarily represents the reaction mechanism investigated under neutral pH conditions in pure water. However, mineral dissolution and precipitation processes are frequently catalyzed by H⁺ (acid mechanism) and OH⁻ (base mechanism). For certain minerals, the kinetic rate constant k incorporates the combined effects of all three mechanisms [37]:

$$\begin{gathered} \mathrm{k}={\mathrm{k}}_{25}^{\mathrm{nu}}\exp \left[{\displaystyle \frac{{-\mathrm{E}}_{\mathrm{a}}^{\mathrm{nu}}}{\mathrm{R}}}\left({\displaystyle \frac{1}{\mathrm{T}}}-{\displaystyle \frac{1}{298.15}}\right)\right] + {\mathrm{k}}_{25}^{\mathrm{H}}\exp \left[{\displaystyle \frac{{-\mathrm{E}}_{\mathrm{a}}^{\mathrm{H}}}{\mathrm{R}}}\left({\displaystyle \frac{1}{\mathrm{T}}}-{\displaystyle \frac{1}{298.15}}\right)\right]{\mathrm{a}}_{\mathrm{H}}^{{\mathrm{n}}_{\mathrm{H}}} \\ + {\mathrm{k}}_{25}^{\mathrm{OH}}\exp \left[{\displaystyle \frac{-{\mathrm{E}}_{\mathrm{a}}^{\mathrm{OH}}}{\mathrm{R}}}\left({\displaystyle \frac{1}{\mathrm{T}}}-{\displaystyle \frac{1}{298.15}}\right)\right]{\mathrm{a}}_{\mathrm{OH}}^{{\mathrm{n}}_{\mathrm{OH}}} \end{gathered}$$

(3)

where superscripts or subscripts nu, H, and OH indicate neutral, acid and base mechanisms, respectively; a is the activity of the species; and n is power term.

Table 2. Kinetic rate parameters for primary and secondary minerals [38].

Mineral		Kinetic Rate Parameters
		Neutral Mechanism		Acid Mechanism			Base Mechanism
		k25	Ea	k25	Ea	n (H+)	k25	Ea	n (H+)
Primary minerals	Calcite	Equilibrium mineral
	Anhydrite	6.45 × 10−4	14.3
	Quartz	1.02 × 10−14	87.7
	Illite	1.66 × 10−13	35.0	1.05 × 10−11	23.6	0.34	3.02 × 10−17	58.9	−0.40
	K-feldspar	3.89 × 10−13	38.0	8.71 × 10−11	51.7	0.50	6.31 × 10−22	94.1	−0.82
	Kaolinite	6.92 × 10−14	22.2	4.90 × 10−12	65.9	0.78	8.91 × 10−18	17.9	−0.47
	Pyrite	6.46 × 10−13	56.9
Secondary minerals	Goethite	1.15 × 10−8	86.5
	Hematite	2.51 × 10−15	66.2	4.07 × 10−10	66.2	1.00

presents the kinetic rate parameters of the primary and secondary minerals involved in this study [38].

2.3. Reservoir Batch Model

In this study, a series of one-dimensional (1D) and two-dimensional (2D) reactive transport models were developed using the TOUGHREACT simulator to investigate the coupled THC processes within the St. Peter Sandstone formation. Due to limitations imposed by simulators and experimental data in the Pittsfield field test, all of the models were predicted on the following assumptions:

• The overlying and underlying strata in the study area were assumed to be impermeable, while large-volume grids were employed for the lateral boundaries to simulate open boundaries.
• The TOUGHREACT simulator was primarily used to investigate Darcy-scale or pore-scale fluid migration and chemical reactions, while neglecting non-Darcy flow effects and microbial processes involved in CAESA.
• It was difficult to characterize the reservoir in detail due to the physical parameters of the St. Peter Sandstone formation in the Pittsfield field. Inter-stratal physical parameters (e.g., permeability and porosity) were derived from site-specific data. However, within a single stratum, we assumed it to be a homogeneous anisotropic stratum, and the mineral parameters within the stratum were assumed to be consistent.
• The oxidation reaction of pyrite and its products were hypothesized to be complete reactions, and possible incomplete reactions and their products were ignored.

The corresponding modeling cases are summarized in

Table 3. Classification of simulation cases and corresponding research purposes for air injection-related numerical experiments.

Dimension	Case	Purpose
1D	Case 1-1 Case 1-2 Case 1-3	Effects of different air injection rates
	Case 1-1 Case 1-4 Case 1-5 Case 1-6	Effects of different air injection temperatures
	Case 1-1 Case 1-7 Case 1-8	Effects at different formation depths, which represent aquifers of different scales
2D	Case 2-1 Case 2-2 Case 2-3	Influence of different CAESA cycles
	Case 2-4 Case 2-5	Analysis of the geochemical reactions

.

The data in

Table 4. Reservoir parameters (modified from Yang et al. [39]).

Parameter	Value	Parameter	Value
Grain density	2600 kg/m3	Heat conductivity	2.16 W/(m·°C)
Grain specific heat	920 J/(kg·°C)	Pore compressibility	2.10 × 10−9 Pa−1
Relative Permeability	van Genuchten-Mualem model
λ	0.457	Sls	1.0
Slr	0.2	Sgr	0.1
Capillary pressure	van Genuchten model
λ	0.457	P0	1.19 × 103 Pa
Slr	0.15	Sls	1.0
Pmax	1.0 × 106 Pa
St. Peter Sandstone	Porosity	kx = ky (m2)	kz (m2)
Green Layer	0.17	1.81 × 10−13	7.60 × 10−14
White Layer	0.16	4.03 × 10−13	6.62 × 10−13
Gray Layer	0.16	8.70 × 10−13	7.27 × 10−13

, derived from the literature by Yang et al. [39], is reliable because the study had used a numerical model (TOUGH3/EOS3) based on the Pittsfield aquifer field test, analyzing parameters such as pressure, gas saturation, and temperature. Its model had demonstrated that simulation results aligned well with observational data, which validated the empirical basis and consistency of the findings. The 1D simplified model was constructed based on the petrophysical parameters of the Gray Layer within the St. Peter Sandstone. The reservoir was assumed to extend infinitely in the horizontal direction, with a total simulated length of 1000 m and a thickness of 50 m. A radial grid was employed, discretized into 60 non-uniform cells. The gas production/injection well was positioned at the first grid block, representing a well radius of 0.1 m. This model setup allows for an efficient representation of near-wellbore transport processes while maintaining computational stability. The 2D model was designed to better capture the spatial heterogeneity of the St. Peter Sandstone formation. The sandstone layer was bounded above and below by low-permeability confining rocks, ensuring vertical containment of the injected gas. The reservoir parameters are shown in Table 4. The 2D computational domain covered an area of 1000 m × 25 m, discretized into a 79 × 8 grid to provide refined spatial resolution in both radial and vertical directions. Mesh generation has been tested multiple times and could effectively balance result accuracy with computational efficiency. The gas injection/production well was located at the center of the Green Layer, also with a well radius of 0.1 m. The mineral composition of the rock, as shown in

Table 5. Initial mineral composition of the rock.

Mineral	Formula	Volume Fraction (%)
Calcite	CaCO3	3.0
Anhydrite	CaSO4	1.0
Quartz	SiO2	33.0
Illite	K0.6–0.85(Al, Mg)2(Si, Al)4O10(OH)2	28.0
K-feldspar	KAlSi3O8	5.0
Kaolinite	Al2Si2O5(OH)2	10.0
Pyrite	FeS2	20.0

, includes the seven most common minerals found in the formation. The initial geochemical conditions of the reservoir fluid, presented in

Table 6. Initial geochemical conditions of the reservoir fluid.

Species	Equilibrium Concentration (mol/kg·H2O)	Species	Equilibrium Concentration (mol/kg·H2O)
H+	6.36 × 10−8	SiO2(aq)	1.21 × 10−4
Ca2+	9.04 × 10−2	HCO3−	1.93 × 10−4
Mg2+	5.93 × 10−6	SO42−	6.72 × 10−3
K+	4.84 × 10−3	AlO2−	2.31 × 10−10
Fe2+	4.09 × 10−6	O2(aq)	3.48 × 10−71
pH	7.32	H2O	1.0 (default value)

, are the results obtained after the initial equilibrium calculation in TOUGHREACT.

Table 7. Initial and boundary conditions.

Case	Initial Condition	Boundary Condition
Case 1-1	P = 2 MPa, T = 21 °C (200 m depth)	Air Injection: T = 20 °C with a fixed specified enthalpy, rate = 1 kg/s
Case 1-2	P = 2 MPa, T = 21 °C	Air Injection: T = 20 °C with a fixed specified enthalpy, rate = 5 kg/s
Case 1-3	P = 2 MPa, T = 21 °C	Air Injection: T = 20 °C with a fixed specified enthalpy, rate = 10 kg/s
Case 1-4	P = 2 MPa, T = 21 °C	Air Injection: T = 40 °C with a fixed specified enthalpy, rate = 1 kg/s
Case 1-5	P = 2 MPa, T = 21 °C	Air Injection: T = 60 °C with a fixed specified enthalpy, rate = 1 kg/s
Case 1-6	P = 2 MPa, T = 21 °C	Air Injection: T = 80 °C with a fixed specified enthalpy, rate = 1 kg/s
Case 1-7	P = 5 MPa, T = 30 °C (500 m depth)	Air Injection: T = 20 °C with a fixed specified enthalpy, rate = 1 kg/s
Case 1-8	P = 10 MPa, T = 45 °C (1000 m depth)	Air Injection: T = 20 °C with a fixed specified enthalpy, rate = 1 kg/s
Case 2-1	Atmosphere pressure: 1.01 × 105 Pa Temperature: 15 °C Hydrostatic gradient: 9.8 kPa/m Geothermal gradient: 30 °C/km	Daily cycle:  0–10 h injection, rate = 5 kg/s  16–18 h withdrawal, rate = −3 kg/s
Case 2-2		Weekly cycle:  Weekend: 0–10 h injection per day, rate = 5 kg/s  Weekday: 16–18 h withdrawal per day, rate = −3 kg/s
Case 2-3		Monthly cycle:  0–15 days: 0–10 h injection per day, rate = 5 kg/s  16–18 days: shut-in  19–28 days: 16–18 h withdrawal per day, rate = −3 kg/s
Case 2-4		Air Injection: T = 20 °C with a fixed specified enthalpy, rate = 5 kg/s
Case 2-5		Air Injection: T = 20 °C with a fixed specified enthalpy, rate = 5 kg/s, without minerals

lists the initial and boundary conditions for all the simulation cases.

The data were considered reliable because Yang et al. conducted a three-dimensional numerical simulation using TOUGH3/EOS3 based on the Pittsfield aquifer field test. They analyzed key parameters, including pressure, gas saturation, and temperature, and found that the simulation results agreed well with the observed data, particularly the wellhead pressure after the first 30 days of air injection. The study was thus grounded in real field measurements, employed a validated numerical model, and demonstrated consistency between simulated and observed results, supporting the credibility of the data used.

3. Results and Discussion

3.1. Evolution of Aquifer Reservoir Characteristics

To investigate the effects of different air injection rates, injection temperatures, storage depths, and operational cycles in CAESA, a series of numerical simulations were conducted to evaluate the reservoir characteristics of the aquifer.

3.1.1. Effects of Different Air Injection Rates

Case 1-1, Case 1-2, and Case 1-3 interpreted the evolution of pressure, temperature, and gas saturation (Sg) under three compressed air injection rates (1 kg/s, 5 kg/s, and 10 kg/s, respectively) after 10 years of simulation. As shown in Figure 2a, the pressure distribution along the radial direction from the injection well (distance = 0 m) revealed a distinct dependence on the air injection rate. Near the wellbore, the pressure increased markedly with injection rate: ~2.3 MPa for 1 kg/s, ~2.8 MPa for 5 kg/s, and ~3.3 MPa for 10 kg/s. This trend was attributed to the fact that higher injection rates introduced a greater mass of fluid per unit time, thereby resulting in more conspicuous pressure buildup near the wellbore. The rate of pressure decline was steepest in the near-well region (e.g., within 100 m) and slowed at larger distances. As the distance increased, the pressure decreased monotonically for all cases, indicating fluid flow-induced pressure dissipation within the porous medium. By 900–1000 m, all cases converged to 2.0 MPa, signifying a quasi-steady pressure state far from the well.

Figure 2b depicts temperature versus distance. Near the wellbore, the temperature of all cases stabilized at approximately 210 °C after 10 years of simulation. With increasing distance from the wellbore, the thermal response to injection rate followed the order of 10 kg/s > 5 kg/s > 1 kg/s. Specifically, for the 1 kg/s case, the temperature increase extended to about 100 m; for 5 kg/s, to approximately 150 m; and for 10 kg/s, to nearly 200 m. This behavior indicated that higher injection rates enhanced convective heat transfer, thereby facilitating the transport of thermal energy over a larger spatial scale.

Figure 2c illustrates the variation in Sg with distance from the wellbore. Initially, no free gas phase existed in the reservoir. After 10 years, Sg approached 1.0 for all injection cases in the near-well region, reflecting that the injected fluid was predominantly composed of compressed air. For the injection rate of 1 kg/s, Sg was 1.0 within approximately 30 m from the wellbore; for 5 kg/s, Sg remained at 1.0 within about 80 m; and for 10 kg/s, Sg maintained 1.0 up to roughly 100 m. This result revealed a distinct development of air bubbles, depicting the evolving interface between the gas and aqueous phases. Moreover, the extent of the air bubble increased with higher air injection rates.

3.1.2. Effects of Different Air Injection Temperatures

Case 1-1, Case 1-4, Case 1-5 and Case 1-6 interpreted the evolution of pressure, temperature, and gas saturation under four compressed air injection temperatures (20 °C, 40 °C, 60 °C, and 80 °C, respectively), with a 1.0 kg/s injection rate after 10 years of simulation.

As shown in Figure 3a, the radial pressure distribution from the injection well exhibited a discernible dependence on the compressed air injection temperature after 10 years of operation. Within the near-well region (approximately 0–50 m), pressure varied slightly with temperature yet remained close to 2.30 MPa across all cases. As illustrated in Figure 3b, higher injection temperatures resulted in correspondingly elevated formation temperatures near the wellbore, reflecting the influence of thermal energy carried by the injected air. As shown in Figure 3c, Sg near the wellbore (distance < 50 m) approached 1.0 for all cases, indicating the injected fluid is mainly in the gas phase. Overall, the effects of injection temperature on both pressure and gas saturation were minor, consistent with the trends reported by Guo et al. [18], suggesting that temperature variations exerted limited influence on reservoir pressurization and phase distribution under long-term injection conditions.

3.1.3. Effects of Initial Temperature and Pressure at Different Formation Depths

To analyze the 10-year simulation results of CAESA at three different reservoir depths (Case 1-1: 200 m, Case 1-7: 500 m, and Case 1-8: 1000 m), we examined the trends of pressure, temperature, and gas saturation with distance from the injection well with a 1.0 kg/s injection rate.

As illustrated in Figure 4a, the initial pressure and temperature at different reservoir depths were determined from the corresponding pressure and temperature gradients. Although the depths at 200 m and 500 m differed, their pressure distributions exhibited a consistent declining trend with radial distance from the injection well, suggesting a similar hydraulic response under these conditions. However, at a depth of 1000 m, a distinct pressure drop was observed at approximately 300 m from the wellbore. As shown in Figure 4b, the temperatures near the wellbore (within 50 m) for all three depth cases stabilized at approximately 210 °C. This result indicated that the thermal responses induced by the same air injection rate were generally consistent across different reservoir depths. At depths of 200 m and 500 m, the temperature distributions within 100 m of the wellbore showed similar trends. In contrast, at a depth of 1000 m, an abrupt temperature drop was observed at approximately 30 m from the wellbore. In Figure 4c, the variations in Sg at reservoir depths of 200 m and 500 m exhibited consistent trends, with the air bubble formed at 200 m being slightly larger than that formed at 500 m. By comparison, at a depth of 1000 m, Sg showed an abrupt decline at a distance of approximately 10 m from the wellbore, and the resulting air bubble was significantly smaller than those observed at the shallower depths. The abrupt decreases in pressure, temperature, and gas saturation were likely attributed to the relatively low injection pressure associated with the 1 kg/s air injection rate, which might have been insufficient to overcome the formation pressure and sustain continuous air flow, thereby constraining convective heat transfer.

3.1.4. Influence of Different CAESA Cycles

Case 2-1, Case 2-2, and Case 2-3 simulated the operation of the CAESA system under daily, weekly, and monthly cycles, respectively. The temporal variations in pressure, temperature, and gas saturation were analyzed to assess the influence of cyclic frequency on reservoir behavior and system stability.

As shown in Figure 5a, the evolution of pressure over 28 days at a distance of 0.25 m from the injection well exhibited distinct patterns under the three operational cycles. The initial pressure was 2.1 MPa. In the daily cycle, pressure fluctuated periodically in response to alternating air injection and withdrawal. During the first few days, it oscillated around 8 MPa and gradually increased, reaching a relatively stable range of 10–11 MPa after approximately six days. By the end of the 28-day simulation, pressure remained around 10–10.5 MPa, indicating a well-stabilized near-wellbore condition. In the weekly cycle, pressure increased rapidly to about 10 MPa during the initial two-day injection, followed by cyclic variations that stabilized within 5.5–7.0 MPa. Under the monthly cycle, pressure rose steadily to 10–11 MPa during the first 15 days of injection, then decreased after a three-day shut-in and subsequent 10-day withdrawal, ultimately declining to approximately 6.4 MPa. These observations suggested that shorter cyclic intervals, such as daily operation, sustained higher average pressures due to frequent reinjection and shorter depressurization periods, whereas longer cycles allowed greater dissipation and a more pronounced decline between injection stages.

Figure 5b presents the corresponding temperature evolution at the same location. In the daily cycle, the temperature increased rapidly from an initial 21 °C to approximately 140 °C after 5 days, continuing to rise gradually and reaching about 200 °C at the end of the simulation. The weekly cycle showed a sharp temperature rise at the beginning of each injection, followed by conspicuous weekly oscillations between 65 °C and 140 °C. In the monthly cycle, the temperature increased to nearly 180 °C during the 15-day injection phase and subsequently decreased during the withdrawal period, reaching 122 °C after 28 days. These results indicated that the daily cycle maintained higher overall temperatures because its shorter period minimized thermal losses, whereas the weekly and monthly cycles provided sufficient time for conductive heat dissipation and thermal equilibration with the surrounding formation.

The distribution of Sg (Figure 5c) revealed consistent cyclic responses across the three cases. Initially, all cases had Sg = 0. In the daily cycle, Sg rose rapidly, reaching 1.0 after 10 days and remaining nearly constant thereafter. Under the weekly cycle, Sg exhibited moderate oscillations in step with the periodic variations in pressure and temperature. In the monthly cycle, Sg gradually increased to approximately 1.0 after 8 days and then declined during the withdrawal, decreasing to about 0.8 at the end of the simulation. These results implied that the daily cycle facilitated a more stable and sustained gas-phase distribution. Overall, the comparison demonstrated that high-frequency cycling (daily cycle) enhanced system responsiveness [39,40], whereas low-frequency operation resulted in greater thermodynamic and hydraulic variability within the storage formation.

Based on the simulation outcomes, higher injection rates generated larger gas bubbles within the reservoir, enabling greater compressed-air storage. Injection temperature exerted only a minor influence on gas bubble development but implying that the aquifer effectively supported thermal energy storage. Increasing reservoir depth reduced the extent of the gas bubble, as deeper formations required higher injection pressures to overcome stratigraphic constraints. In addition, higher operational frequencies, such as daily cycling, promoted more stable gas-phase distributions and facilitated steadier system performance.

For practical CAESA deployment, these parameters influence project quality from multiple dimensions and thus require integrated assessment during system design. Adjusting a single factor offered limited improvements. The influencing parameters considered in actual projects are even more, and a comprehensive evaluation criterion needs to be established for measurement.

3.2. Evolution of Geochemical Reactions

Through the 2D models of Case 2-4, and Case 2-5, a comprehensive analysis of the geochemical reactions occurring within the reservoir was conducted.

3.2.1. Dissolution of Oxygen in Injected Air

Figure 6a presents the spatial distribution of dissolved oxygen after 28 days of simulation. Horizontally, the air injection well was located at the “Distance = 0 m” position within the Green Layer, with distance increasing outward from the wellbore. In the near-well region, the dissolved oxygen concentration was markedly elevated, indicated by the red and orange zones on the color scale corresponding to the highest concentration levels. Within approximately 100 m from the injection well in the Green Layer, the region was dominated by red shading, representing concentrations close to 6.24 × 10⁻² mol/kg·H₂O. With increasing horizontal distance (100–800 m), the color gradually transitioned from yellow (4.46 × 10⁻² mol/kg·H₂O) to green (3.57 × 10⁻² mol/kg·H₂O) and light blue (8.91 × 10⁻³ mol/kg·H₂O), indicating a continuous decline in concentration. In the far-field zone (distance > 800 m), the reservoir was characterized by dark blue shading, suggesting that the dissolved oxygen concentration approached zero. Vertically, the high-concentration zones (red/orange) were primarily concentrated near the wellbore (0–100 m) within the Green Layer. In contrast, the White Layer and Gray Layer, which possessed higher permeability, exhibited broader distributions of dissolved oxygen, with the high-concentration regions extending to approximately 400 m. However, in the deeper portion of the Gray Layer, the concentration decreased due to limited diffusive transport.

Figure 6b illustrates the spatial distribution at Time = 365 days, which shows a marked change in spatial patterns compared with the results at 28 days. The most notable feature was that the high-concentration bands (red/orange zones) that had dominated the near-wellbore region at 28 days had almost disappeared. Instead, the entire reservoir was predominantly characterized by green and blue, indicating that the concentration across different layers and radial distances had substantially decreased. In contrast, the concentration slightly increased at distances greater than 700 m from the wellbore, suggesting that over a long-term simulation period, dissolved oxygen had gradually migrated toward the reservoir boundaries.

At the early stage of CAESA operation (28 days), a distinct high-concentration plume of dissolved oxygen developed near the wellbore due to oxygen release and diffusion from the injected compressed air, reflecting the coupled effects of diffusion and convection. After one year (365 days), this plume had largely dissipated, and the overall concentration markedly declined, mainly as a result of geochemical consumption through reactions with aquifer minerals (discussed in Section 3.2.3). The evolution of spatial distribution underscored the dynamic interaction between the storage system and the groundwater environment. These results highlighted that CAESA operations induced both short-term redox perturbations and long-term geochemical transformations, emphasizing the need for continuous monitoring of key indicators such as dissolved oxygen.

3.2.2. pH of Solution

Figure 7a illustrates the spatial distribution of pH at a depth of 200 m after 28 days of CAESA operation. In the near-well region of the Green Layer (distance < 10 m), low pH values were observed, with the dark blue and light blue zones corresponding to pH levels of approximately 4.29–4.67. This low-pH zone vertically extended throughout the Green Layer and partially penetrated the overlying White Layer. With increasing horizontal distance and depth, the pH gradually increased: within the Green, White, and upper Gray Layers, the color transitioned from cyan (~5.05) to green (~5.81) and finally to yellow (~6.57). Beyond 10 m in the White and Gray Layers, the pH continued to rise, approaching its initial level. These results indicated that at the early simulation stage, the low-pH anomaly was primarily confined to the near-well region. Figure 7b showed the pH distribution at the same depth after 365 days of simulation, revealing significant changes compared with the 28-day results. The low-pH (blue/light-blue) zone expanded horizontally to within 20 m of the wellbore and vertically to a depth of approximately 220 m. Adjacent to this region (20–60 m), the pH ranged from cyan (~5.05) to yellow (~6.57), and further outward (60–100 m), the pH transitioned from yellow (~6.57) to orange (~6.95).

The most striking difference from the 28-day results was the previously high-pH (red/orange) regions transformed into low (blue) and moderately low (green) pH zones widely distributed near the wellbore across all three layers. This transformation was driven by long-term geochemical processes. Over a year, the injected air interacted with aquifer minerals through dissolution, precipitation, and ion exchange reactions (a detailed discussion is provided in Section 3.2.3). These reactions consumed alkaline constituents or released acidic species, thereby reducing the pH of the groundwater in the aquifer.

3.2.3. Mineral Dissolution or Precipitation

Figure 8a shows the spatial distribution of the relative variation in pyrite abundance after 28 days of simulation in the CAESA system (relative to Time = 0 of mineral abundance in vol%, with positive values indicating mineral precipitation and negative values indicating dissolution). The majority of the reservoir was dominated by deep blue zones, suggesting that pyrite abundance remained nearly unchanged across most regions. Noticeable variations occurred only in the vicinity of the wellbore (horizontal distance: 0–10 m; vertical depth: 200–210 m). Within the Green Layer surrounding the wellbore (0–10 m), the orange and yellow zones corresponded to pyrite abundance changes of approximately −1.22 × 10⁻³ to −1.05 × 10⁻³ vol%. In the White Layer, a red zone appeared within 0–5 m, representing the largest change in pyrite abundance, up to about −1.40 × 10⁻³ vol%. After 365 days (Figure 8b), the spatial distribution changed substantially. The red and orange regions (up to −1.40 × 10⁻³ vol%) covered a much broader area in the Green Layer compared with the 28-day result. In the White and Gray Layers, significant variations were also observed within 40 m of the wellbore. The radial and vertical expansion of high-variation zones indicated that prolonged geochemical reactions progressively influenced a wider spatial extent of the reservoir.

The spatiotemporal evolution of pyrite abundance suggested that pyrite dissolution was driven by oxidation reactions, which consumed dissolved oxygen and released H⁺, thereby decreasing pH. The pyrite oxidation process [41,42,43] can be represented as:

FeS₂(s)_(pyrite) + 7/2O₂(aq) + H₂O → Fe²⁺ + 2SO₄²⁻ + 2H⁺

(4)

This reaction produced ferrous iron (Fe²⁺) and sulfate while generating acidity, consistent with the observed decreases in both dissolved oxygen and pH in the near-well region. Under persistent oxidizing conditions, the released Fe²⁺ was further oxidized to ferric iron (Fe³⁺), as represented by the following reaction:

Fe²⁺ + 1/4O₂(aq) + H⁺ → Fe³⁺ + 1/2H₂O

(5)

Over time, these oxidation-induced alterations, driven by pyrite dissolution, reflected the strong coupling between fluid transport and redox geochemistry, and the release of acidity and secondary ions may further modify mineral–water equilibria and promote secondary mineral precipitation.

Figure 9a illustrates the spatial distribution of relative goethite abundance changes after 28 days of CAESA simulation. The reservoir was predominantly deep blue, indicating that goethite variations were negligible (≈0 vol%) in most regions. Significant changes were localized near the wellbore (0–5 m horizontal distance). In the White Layer (depth interval adjacent to the well), orange and red areas corresponded to goethite changes of 1.06 × 10⁻³ to 1.22 × 10⁻³ vol%. In contrast, the Gray Layer (depth > 210 m) exhibited minimal variation beyond the wellbore, with changes below 1 × 10⁻⁴ vol%, indicating that mineral abundance remained stable. After 365 days (Figure 9b), both the spatial extent and magnitude of goethite variation increased markedly. High-change regions (red, orange, yellow) expanded significantly, demonstrating that the geochemical influence propagated from localized zones near the well to broader reservoir areas over time. Temporal evolution analysis revealed that goethite precipitation closely followed pyrite dissolution patterns, indicating that pyrite dissolution was the dominant mechanism driving goethite formation:

Figure 9c,d show relative hematite abundance changes after 28 and 365 days, respectively. Spatial patterns remained broadly similar over time, with minor increases in magnitude. Most of the reservoir remained deep blue, reflecting minimal hematite variation. Changes were restricted to regions near the wellbore (0–20 m), extending downward but limited horizontally. This indicated that hematite-related geochemical reactions progressed very slowly. These results were consistent with the experimental observations of Müller and Regenspurg [44], who proposed that ferric ions might precipitate as goethite or hematite. The geochemical reactions involved were as follows:

Fe³⁺ + 3H₂O → Fe(OH)₃ + 3H⁺

(6)

Fe(OH)₃ → FeOOH_goethite + H₂O

(7)

2Fe(OH)₃ →Fe₂O_3(hematite) + 3H₂O

(8)

Insoluble ferric hydroxide existed in an amorphous state and gradually transformed into more stable mineral phases. Goethite was the most stable over a pH range of 2–6 [45] and, consequently, was the predominant secondary mineral within the formation. In contrast, hematite formed only through the dehydration of ferric hydroxide at temperatures approaching 100 °C [35]. Such conditions were typically reached only in the vicinity of the wellbore, which explained why, as shown in Figure 9c,d, hematite was primarily localized near the wellbore.

Figure 10 illustrates the temporal evolution of mineral abundance relative to the initial state for the other six minerals—calcite, anhydrite, quartz, kaolinite, illite, and K-feldspar—at a location 73.5 m from the wellbore over a 365-day simulation. At the beginning, all minerals had an abundance change of zero. As compressed air was injected and oxygen dissolved into the formation water, various minerals underwent dissolution or precipitation. For instance, calcite gradually dissolved throughout the simulation, reaching a relative decrease of approximately −1.40 × 10⁻⁴ vol% after 365 days, the largest decline among all minerals, indicating that calcite participated most actively in the geochemical reactions. K-feldspar and quartz exhibited similar trends of dissolution, with final abundance changes of about −5.36 × 10⁻⁶ vol% and −6.49 × 10⁻⁸ vol%, respectively, reflecting much slower reaction rates compared with calcite. The abundance of kaolinite decreased during the first 250 days, implying dissolution, but showed a slight increase thereafter, suggesting that chemical conditions during the later stage favored its precipitation. Anhydrite also underwent minor dissolution, though the change (−1.85 × 10⁻¹² vol%) was negligible. In contrast, illite was the only mineral whose abundance increased, indicating precipitation; its relative change peaked at approximately 3.97 × 10⁻⁵ vol% around 240 days and then stabilized. These variations reflected the distinct reactivities and kinetic behaviors of different minerals and highlighted the complex fluid–mineral interactions occurring within the reservoir. Such findings provided critical insights into the long-term reservoir evolution and the mechanisms governing how porosity–permeability characteristics responded to fluid–rock interactions.

Figure 11 illustrates the spatial evolution of porosity change and the corresponding permeability ratios (K_z/K_z0) at 28 days and 365 days. In the early injection stage, the dissolution of reactive minerals—primarily carbonates and minor silicates—temporarily increased pore volume near the wellbore, producing porosity enhancements on the order of 10⁻³ and permeability ratios approaching 1.10. These increases were most pronounced in the Green Layer, consistent with its higher proportion of reactive mineral phases and greater exposure to oxygen-rich injected air. As the simulation proceeded, precipitation processes gradually counterbalanced a part of the dissolution effects. Secondary mineral formation—such as iron oxyhydroxides from pyrite oxidation—attenuated the change in porosity. By 365 days, the combined effects of reduced dissolution rates and localized precipitation resulted in porosity changes falling below ~1.5 × 10⁻³ across most of the reservoir, while permeability returned to values close to the initial state (1.00–1.03).

Overall, the interaction between dissolution and subsequent precipitation governed the structural changes. Dissolution dominated the initial response, enhancing pore connectivity near the well, whereas precipitation exerted a stabilizing influence over longer time-scales. The long-term impact of mineral reactions on reservoir hydraulic properties was therefore limited, confirming that reactive processes primarily influenced short-term behavior rather than inducing persistent large-scale changes.

3.2.4. Oxygen Depletion

Figure 12 illustrates the temporal evolution of the oxygen content difference (ΔO₂ = n_{Case 2–5} − n_{Case 2–4}) between two cases, which exhibited a distinct three-stage evolution. During the initial rapid growth stage (0–10 days), ΔO₂ increased sharply from 0 mol to approximately 8.0 × 10⁴ mol. In the subsequent peak and rapid decline stage (10–90 days), ΔO₂ reached its maximum of about 1.8 × 10⁵ mol at day 21, followed by a continuous decrease to ~6.0 × 10⁴ mol by day 90. In the late steady stage (90–360 days), the decline rate slowed markedly, and ΔO₂ stabilized at around 3.0 × 10⁴ mol after 240 days with minimal changes thereafter. This temporal evolution reflected the regulatory role of mineral-mediated geochemical processes on oxygen migration and transformation: the early sharp increase was likely driven by oxidation reactions of reactive minerals, the subsequent decline resulted from the progressive depletion of reactive mineral phases, and the final stabilization indicated the establishment of a dynamic equilibrium between oxygen supply and consumption. This depletion was most likely caused by the oxidation of minerals, particularly pyrite—within the reservoir rock matrix [15,35]. Overall, the evolution of ΔO₂ quantitatively revealed the short-term perturbation and long-term regulation of oxygen behavior by minerals, offering key insights into the time-dependent characteristics of geochemical processes in mineral-bearing systems.

4. Conclusions

This study numerically investigated the evolution of reservoir characteristics and geochemical reactions during CAESA using coupled THC simulations. The results provided quantitative insights into the interplay between pressure–temperature evolution, gas saturation development, and mineralogical transformations within the storage formation. The major findings are summarized as follows:

(1) Reservoir pressure, temperature, and gas saturation evolution were strongly controlled by injection parameters and reservoir depth. Higher air injection rates induced greater near-well pressure buildup and larger air bubble zones, while elevated injection temperatures primarily enhanced the thermal field near the wellbore. Differences in initial pressure and temperature with depth affected local flow behavior, with the deepest (1000 m) reservoir showing abrupt declines in pressure, temperature, and gas saturation due to insufficient injection pressure to overcome the native formation stress. These findings demonstrated that the hydraulic and thermal responses of the aquifer were highly sensitive to operational parameters, which should be optimized to maintain storage efficiency and stability.

(2) Operational cycle frequency significantly influenced thermodynamic stability and pressure–temperature fluctuations. Daily operation maintained higher average pressure and temperature owing to more frequent air reinjection and shorter relaxation intervals, minimizing energy loss. In contrast, weekly and monthly cycles caused greater pressure and temperature dissipation, leading to larger fluctuations in gas saturation. This indicated that shorter cycles promoted dynamic equilibrium and thermal stability near the well, whereas longer cycles favored pressure decay and heat loss, which could reduce system efficiency over extended operations.

(3) Geochemical reactions were dominated by pyrite oxidation and subsequent secondary mineral formation. Oxygen introduced by the injected air dissolved in groundwater and reacted with pyrite, generating Fe²⁺, sulfate, and H⁺, which lowered pH and induced acidic conditions near the well. Under sustained oxidizing conditions, Fe²⁺ was further oxidized to Fe³⁺, precipitating as goethite and, to a lesser extent, hematite. Calcite, quartz, anhydrite, kaolinite, and K-feldspar experienced gradual dissolution, while illite exhibited precipitation. These reactions collectively reflected complex fluid–mineral interactions and demonstrated the strong coupling between redox chemistry and reactive transport in the reservoir. The interaction between dissolution and precipitation resulted in only transient and localized structural adjustments, indicating that mineral reactions had limited long-term influence on reservoir hydraulic properties.

(4) Oxygen depletion and mineral oxidation jointly regulated the long-term geochemical evolution and storage performance. The temporal evolution of oxygen content difference (ΔO₂) between mineral-bearing and mineral-free systems exhibited a three-stage pattern—rapid increase, peak decline, and stabilization—representing initial oxidation, progressive depletion of reactive minerals, and establishment of oxygen equilibrium. Oxygen consumption caused by mineral oxidation, particularly pyrite, was identified as a critical mechanism reducing the available oxygen for combustion during air withdrawal, potentially affecting the energy efficiency of CAESA.

In summary, this research provides comprehensive insights into the complex interactions between operational parameters and geochemical processes in CAESA systems. The results establish a scientific basis for optimizing storage design and operational strategies to ensure environmental sustainability and long-term viability of aquifer-based compressed air energy storage projects. However, the absence of laboratory or field validation limits the direct transferability of the results to site-specific design or environmental assessment. To strengthen the predictive capability of CAESA modeling, future work should incorporate laboratory experiments under CAESA-relevant pressure–temperature conditions, comparisons with available field datasets, and systematic sensitivity or uncertainty analyses to evaluate the robustness of key outputs such as pH, oxygen consumption, and mineral transformations. These efforts will be essential to translate the numerical insights presented here into validated frameworks for practical engineering applications and environmental risk assessment.