2-D numerical modeling of CO 2 -water-caprock interactions at a potential CO 2 storage site in Turkey

Transcription

1 Geochemical Journal, Vol. 47, pp. 499 to 511, 213 doi:1.2343/geochemj D numerical modeling of CO 2 -water-caprock interactions at a potential CO 2 storage site in Turkey CHANTSALMAA DALKHAA* and ENDER OKANDAN Department of Petroleum and Natural Gas Engineering, Middle East Technical University, 6531 Ankara, Turkey (Received January 5, 213; Accepted July 5, 213) Caprock integrity is an important subject, as it is closely related to the leakage of injected CO 2 into geological formations. Thus, it is necessary to evaluate how injected CO 2 affects caprocks during CO 2 storage projects. This study assessed CO 2 migration into a caprock from the injection reservoir and its interaction with the rock minerals of the caprock and brine. The caprock is the Sayindere Formation, which is a regionally extensive caprock in oil fields located in the southeastern part of Turkey including the Caylarbasi oil field, which has been previously identified and modeled as a potential CO 2 storage site. The upper part of the Sayindere Formation consists of clay, and the lower part consists of clayey limestone. In this work, we performed 2-D radial modeling of CO 2 injection at an annual rate of 1 million tons into the reservoir for 5 years and investigated its impact on the overlying caprock. The simulation was continued for further 1 years without CO 2 injection to evaluate the caprock evolution over longer time periods. Calcite dissolution was observed in the reservoir as well as the caprock. However, no significant increase in the porosity and permeability associated with calcite dissolution was observed in either formation. Minor diffusion of the injected CO 2 into the lowermost layer of the caprock was observed over the simulation years. After 1 years of post-injection, 7% of the injected CO 2 was trapped in the reservoir under the caprock as a plume in a free phase, and the rest was dissolved in the reservoir brine. Sensitivity analyses were performed to determine how permeability anisotropy and reactant surface area of minerals affect the numerical results. Minor variations in the spatial distributions of CO 2 saturation were observed when the permeability anisotropy was changed. No significant variation in the numerical results was observed with changes in the surface area of minerals. The simulation results suggest that the Sayindere formation is a very good caprock for potential CO 2 storage. This work provides information for future decision-making and for development of CO 2 storage demonstration projects in Turkey. Keywords: CO 2 storage, caprock integrity, CO 2 -water-rock interactions, reactive transport modeling INTRODUCTION Carbon dioxide capture and storage (CCS) is one of the proposed means to reduce CO 2 emissions into the atmosphere. CCS consists of the separation of CO 2 from industrial and energy-related sources, transport of this CO 2 to a storage location, and long-term isolation from the atmosphere (IPCC, 25). CCS has been tested at large-scale pilot sites for more than ten years. An initial test of CO 2 injection into a saline aquifer has been successfully in operation since 1996 in the off-shore Sleipner Field in the North Sea (Baklid et al., 1996). At Weyburn (Saskatchewan, Canada), CO 2 has been injected into a depleted oilfield since 2 (White et al., 24). *Corresponding author ( cdalkhaa@ucalgary.ca) *Present address: Applied Geochemistry Group, Department of Geoscience, University of Calgary, Calgary, Alberta, Canada T2N 1N4. Copyright 213 by The Geochemical Society of Japan. Mathematical models and numerical simulations are necessary for risk assessment, monitoring, and assessment of the feasibility and financial issues associated with CO 2 storage. There are a number of mathematical and numerical models applied to specific problems related to CO 2 storage such as storage potential and capacity, leakage, groundwater quality, and CO 2 -trapping mechanisms in geological formations. Gaus et al. (28) provided a detailed overview of geochemical and transport modeling approaches for CO 2 storage and discussed their progress over the last 1 years. Numerical modeling of CO 2 storage is usually divided into 3 categories; hydrodynamic, batch geochemical, and reactive transport modeling. The Eclipse code developed by Schlumberger is widely used for pure flow or hydrodynamic modeling of CO 2 injection in the oil and gas industry. Batch geochemical modeling simulates geochemical fluid-rock interactions occurring in a formation. Since no-flow conditions are considered during batch geochemical simulation, batch modeling is ideal for modeling static laboratory experiments (e.g., Holubnyak et al., 211). Not accounting for 499

2 flow in numerical modeling does not reflect reality, since hydrodynamics play an important role in different CO 2 - trapping mechanisms during CO 2 injection into geological formations. The code TOUGHREACT (Xu et al., 23) has been widely used in recent numerical modeling studies of CO 2 storage (Zhang et al., 29; Xu et al., 21; Shevalier et al., 211; Zheng et al., 29, Apps et al., 211). In the TOUGHREACT code, a variety of equilibrium chemical reactions such as aqueous and surface complexation, gas dissolution/degassing, and cation exchange are considered. Changes in porosity and permeability associated with dissolution and precipitation processes can be also considered in the TOUGHREACT code. Comparison and benchmark studies have been conducted to test the quality of the model outcomes and to assess the application of numerical codes for modeling of CO 2 storage in geological formations (Pruess et al. (24) and Class et al. (29)). These authors concluded that it is currently impossible to assess every deviation in model predictions since this requires a thorough analysis of the simplicity and assumptions in all the model implementations. Turkey is one of the countries that are taking steps to respond to the threat of climate change. Turkey acceded to the United Nations Framework Convention on Climate Change (UNFCCC) in May 24. A report entitled Climate Change and Inventory for Turkey was prepared and submitted to UNFCCC in January 27 as the First National Communication on Climate Change (FNCCC). In 199, Turkey s total greenhouse gas emission was estimated as 17.1 million tons, of which million tons were due to CO 2. By 27, the total greenhouse gas emissions had doubled, reaching 34.5 million tons, with CO 2 emissions of million tons (Gunay and Ubay, 27). Therefore, carbon dioxide capture and storage is being explored as a viable option to reduce anthropogenic CO 2 emissions in Turkey. The deep saline aquifers in the Thrace region of Central Anatolia and the southeastern part of Turkey, and the caverns of the soda mines, and depleted oil and gas fields located in the southeastern part of Turkey have been screened as possible storage sites (Okandan et al., 211). The salt caverns of soda mines were considered not suitable, as their capacity for CO 2 storage is not large or deep enough for injection of supercritical CO 2. The natural CO 2 field, Dodan, has been suggested as another possible storage site for CO 2 (Okandan et al., 211). The CO 2 from this field has been the source for the Bati Raman CO 2 -EOR project that has been operational since 1986 (Issever et al., 1993). The Dodan field has already produced 7 billion standard m 3 of CO 2 for the Bati Raman CO 2 -EOR project. The total CO 2 reservoir reserve is estimated to be about 1 billion standard m 3. Okandan et al. (211) modeled CO 2 injection into the Caylarbasi Oil Field, a partially depleted oil field in the southeastern part of Turkey. The injection of CO 2 into geological formations initiates extra processes that affect the chemistry of brines and may also impact the mineralogy of the storage formation and its caprock. There are many publications describing CO 2 -water-reservoir rock interactions (e.g., Zhang et al., 29; Xu et al., 21; Holubnyak et al., 211; Shevalier et al., 211; Apps et al., 211), but there are relatively few studies on caprock evolution under CO 2 injection (e.g., Gherardi et al., 27). One of the risks associated with CCS are the effects caprockof acidic CO 2 - rich fluids resulting from CO 2 injection on the caprock. If the caprock is damaged, the injected CO 2 may migrate into overlying formations, and potentially into shallow groundwater. The caprock at the Caylarbasi oil field is the Sayindere formation; known as an impermeable rock whose upper part is composed of clay and lower part is composed of clayey limestone. The objective of this study was to investigate the geochemical evolution of the Sayindere caprock in Turkey induced by CO 2 injection into the reservoir and its impact on the caprock integrity using the reactive transport code TOUGHREACT. GEOCHEMICAL REACTIONS After injection, some supercritical CO 2 will initially remain in free phase and be subject to structural or residual trapping in the reservoir. Supercritical CO 2 will, however, also undergo a number of chemical processes such as dissolution in the brine. During dissolution of CO 2 in the brine, the supercritical CO 2 phase and aqueous CO 2 phase are assumed to be in equilibrium: CO 2 (sc) CO 2 (aq) (Reaction 1) where subscripts sc and aq denote supercritical and aqueous CO 2, respectively. This process results in solubility trapping of injected CO 2. The aqueous CO 2 phase, CO 2 (aq), reacts with water to form carbonic acid (H 2 CO 3 ). This acid dissociates into a proton (H + ) and bicarbonate ion (HCO 3 ), resulting in a decrease in ph and acidification of the brine. CO 2 (aq) + H 2 O H 2 CO 3 (Reaction 2) H 2 CO 3 H + + HCO 3. (Reaction 3) Furthermore, carbonic acid may react with any reactant minerals present in a formation, such as calcite and albite, resulting in the release of cations (e.g., Ca 2+, Na + ). H 2 CO 3 + CaCO 3 Ca HCO 3 (Reaction 4) 5 C. Dalkhaa and E. Okandan

3 NaAlSi 3 O 8 + 4H + + 4H 2 O Na + + Al H 4 SiO 4. (Reaction 5) Dissolution of primary minerals may increase the aqueous concentration of cations such as Ca 2+, Mg 2+, and Fe 2+, which can lead to secondary precipitation of carbonate minerals such as dawsonite, siderite, and ankerite. This process results in CO 2 mineral trapping, the most stable long-term form of geological carbon sequestration. Na + + Al 3+ + HCO 3 + 2H 2 O NaAl(CO 3 )(OH) 2 + 3H +. (Reaction 6) Numerical modeling is an important tool that can give insights into the likelihood of the occurrence of these geochemical reactions that are induced by CO 2 injection in complex geological environments during CO 2 storage projects and beyond. NUMERICAL SIMULATION Numerical simulator, Petrasim/TOUGHREACT Petrasim is an interactive pre-processor and postprocessor for the TOUGH family of codes. It helps users to develop models faster and view results rapidly. TOUGHREACT is applicable to one-, two-, or threedimensional geologic media with physical and chemical heterogeneity and can be applied to a wide range of subsurface conditions. Temperature and pressure ranges are controlled by the range of the chemical thermodynamic database applied and the range of the EOS module employed. Because the equilibrium constants are generally not as sensitive to pressure as to temperature, the temperature dependence of equilibrium constants is taken into account by TOUGHREACT but the pressure dependence is not. In TOUGHREACT, the major processes for fluid and heat flow are: (1) fluid flow in both liquid and gas phases occurs under pressure, viscous, and gravity forces; (2) interactions between flowing phases are represented by characteristic curves (relative permeability and capillary pressure); (3) heat flow by conduction and convection, and (4) diffusion of water vapor and gas. Mineral dissolution and precipitation can proceed either due to local equilibrium or to kinetic conditions. Kineticallycontrolled reactions between aqueous species are not considered. Changes in porosity during simulations are calculated from changes in mineral volume fractions. However, deformation of porous media and fluid pressure effects due to porosity changes are neglected. Deformation of pores in the formation would be caused mostly by geomechanical stress induced by CO 2 injection. However, it is very difficult to describe this phenomenon using mathematical equations, and thus, it is not captured in the code. It is very unlikely that deformation of the study formation will occur due to induced pressure from injected CO 2. The increased pressure is not expected to exceed the original reservoir pressure of ~2 Mbar. Therefore, the risk of pore deformation and resulting fluid pressure effects would be minimal in this modeling work. Heat effects from chemical reactions are neglected, as are changes in thermophysical properties of fluid phases such as viscosity, surface tension, and density due to changes in chemical composition (Xu et al., 24a). TOUGHREACT uses a sequential iteration approach. The fluid velocities and phase saturations are used for chemical diffusion- and advection-transport simulations, after solution of the flow equations. Transport by advection and molecular diffusion are considered for both the aqueous and gaseous species. The chemical transport is solved on a component basis. The resulting concentrations obtained from the transport part of the model are substituted into the chemical reaction model. The system of mixed equilibrium-kinetic chemical reaction equations is solved on a grid-block basis by Newton-Raphson iteration. The chemical transport and reactions are iteratively solved until convergence (Xu et al., 24a). The fluid property module ECO2N was used in the TOUGHREACT model. It covers thermodynamic conditions of a temperature range from ambient to 1 C, a pressure range from atmospheric to 6 bars, and salinity from zero to fully saturated. These parameter ranges should be adequate for most conditions encountered during geological disposal of CO 2. Thermophysical properties are accurately calculated for gaseous as well as for liquid CO 2, but no distinction between gaseous and liquid CO 2 is made in the treatment of flow, and no phase change between liquid and gaseous CO 2 is considered in the ECO2 module (Pruess, 25). TOUGHREACT ECO2N fluid property module does not account for an oil phase. The primary focus of the study was to investigate the CO 2, brine, and mineral interaction in a caprock. The caprock does not contain any oil, but only brine. The reservoir, however, contains oil. It is expected that an oil phase would have an impact on the flow of the injected CO 2 in the reservoir and some of the CO 2 will dissolve in such an oil phase. However, it is expected to have little impact on the reactivity of the brine and mineralogy with CO 2 in the caprock and the reservoir. The extended Debye-Huckel equation of activity coefficient calculation for aqueous species was used (Helgeson and Kirkham, 1974). Descriptions of the caprock The Sayindere caprock is a regionally extensive caprock in oil fields located in the southeastern part of Turkey including the Caylarbasi oil field, which has been previously considered and modeled as a potential CO 2 storage site (Okandan et al., 211). 2-D numerical modeling of CO 2 -water-caprock interactions at a potential CO 2 storage site in Turkey 51

4 Table 1. Mineralogy of the caprock and reservoir Mineral Mass fraction Sayindere caprock Calcite.76 Quartz.227 Kaolinite 13 Karabogaz reservoir Calcite Secondary minerals Magnesite Siderite Dolomite Hematite Table 2. Reservoir formation water analysis from Caylarbasi-1 (TPAO) Concentration (ppm) Sodium 644 Calcium 12.5 Magnesium Iron.38 Sulfate 14.4 Chloride Bicarbonate ph 7.23 This caprock is generally known to be quite homogeneous. Its upper part consists of clay and its lower part consists of clayey limestone (pers. comm. with N. Sengunduz). Thin section analysis and XRD analysis of a core from the Sayindere caprock were carried out to identify the caprock mineralogy at the Middle East Technical University, Ankara, Turkey. The core was found to be composed of calcite (76%) and quartz (22.7%) and a small amount of kaolinite (1.3%) (Table 1). From the mineral composition, it was clear that this core representing the Sayindere formation was taken from the lower part of the formation, since it is mostly made up of calcite. Formation water analyses for the Sayindere caprock of Caylarbasi field were not available. However, this problem can usually be addressed by assuming that the caprock formation water is similar to the reservoir formation water (Xu et al., 25). The analysis of formation water obtained from depths between m in the reservoir through the Caylarbasi-1 well was provided by Turkish Petroleum Cooperation (TPAO) (Table 2). The redox species were arbitrarily set in order to represent the subsurface fluid and maintain the various redox states. Before commencement of reactive transport modeling using the code TOUGHREACT, batch modeling without CO 2 injection was performed for 1 years to equilibrate the formation water (Table 2) with the primary minerals present in each formation. This resulted in the initial formation water chemistries, which are summarized in Table 3, for the caprock and reservoir formations prior to CO 2 injection. Simulated model A simple 2-D radially symmetric model was developed to simulate the interaction between CO 2, brine, and Sayindere caprock minerals when CO 2 was injected into the reservoir. The oil phase was assumed not to exist in the model, since ECO2N deals with the H 2 ONaClCO 2 system. The CO 2 was injected at the reservoir conditions of 75 bars pressure and a temperature of 8 C. Table 3. Initial water chemistry used in the TOUGHREACT simulations, after equilibration with minerals Species Concentration (mol/kg) Caprock Reservoir AlO E-7 1.6E-1 Ca E E-4 Cl Fe E-6 2.8E-6 HCO Mg Na O 2 (aq) 1.12E E-19 SiO 2 (aq) 8.8E-4 1.6E-1 2 SO 4 3E-4 3E-4 ph The model boundary is 1 km in the radial direction (x-direction) and 12 m in depth (z-direction). The bottom 4 m represents the reservoir and the upper 8 m represents the Sayindere caprock. The model is divided into 1 cells in the radial direction with non-uniform spacing increasing from the injection point. The first cell has a radius of 1 m. The outermost cells were set up to be numerically very large so that the model behaves as if it acts infinitely. Several sensitivity analyses with respect to the volume of the model were performed for pressure evolution. With a closed boundary, the pressure build-up was observed to be unrealistically high during the injection. Since there are no known faults in the Sayindere formation and only a few minor faults exist in the reservoir (Okandan et al., 29), it was considered reasonable to assume that the model behaved as if it were infinite. Figure 1 illustrates the schematic representation of the model used in this work. No mineralogical analysis was carried out over the reservoir, which is known to be comprised of pure limestone. Thus, its mineralogy was assumed to be pure 1% 52 C. Dalkhaa and E. Okandan

5 8 m Caprock 12 m 4 m Reservoir CO 2 Injection zone (1 cells in radial direction) 1 m Fig. 1. Scheme of the vertical cross section of the geometrical 2-D radial model. Table 4. Physical properties of the caprock and reservoir defined in the model Sayindere caprock Porosity (fraction) 5 Permeability (md).1 Residual liquid saturation (fraction).125 Residual gas saturation (fraction) 5 Karabogaz reservoir Porosity (fraction).2 Permeability (md) 1 Residual liquid saturation (fraction).125 Residual gas saturation (fraction) 5 calcite in the model (Table 1). To obtain the initial formation water for the Caylarbasi reservoir, the measured water (Table 2) was equilibrated with calcite prior to the injection. The Sayindere caprock and the reservoir formation properties assumed for modeling are given in Table 4. The porosity of the caprock is 5. The permeability is assumed to be.1 md (pers. comm. with N. Sengunduz). The porosity of the reservoir is.2, and the permeability is 1 md (Okandan et al., 29). The porosity was homogenous and isotropic in the model. However, the permeability anisotropy was considered, as it is often used to express the degree of heterogeneity in the formation. The ratio of permeability anisotropy (ratio of vertical to horizontal permeability) is often expected to be less than unity due to the fact that layering in deep subsurface environments is generally horizontal or nearly horizontal. Actual measured data of this parameter for the formations was not available, so a value of 1 was initially assumed in the modeling. A sensitivity simulation of a Table 5. Mineral grain size and surface area Mineral Vol. frac. Grain size (m) Surface area (g/cm 2 ) Calcite Hematite Kaolinite Magnesite Quartz Siderite Dolomite different value for permeability anisotropy ratio was performed to understand its effect on the numerical results (see Subsection Permeability anisotropy ). The Van Genuchten (198) model was used to calculate the capillary and relative permeability curves for both formations. In the model, the minerals magnesite (MgCO 3 ), siderite (FeCO 3 ), dolomite (CaMg(CO 3 ) 2 ), and hematite (Fe 2 O 3 ) were defined as secondary phases, since they are the likely products of possible geochemical reactions induced by CO 2 injection based on the fact that the formation water consisted of major divalent ions such as Mg 2+, Fe 2+, and Ca 2+ (Table 2). Table 5 shows the grain size and the surface area of each mineral considered in the simulation. The evolution of surface area in natural geologic formations is very complex, particularly for multi-mineral systems, and is not quantitatively understood at present (Xu et al., 24b). The surface areas of the minerals from the work of Xu et al. (21) were used for this study and are given in Table 5. A sensitivity analysis for the kinetic reaction rate was performed, decreasing the surface area by an order of magnitude (see Subsection Reactive surface area ). Table 6 gives the kinetic rate constant at room tem- 2-D numerical modeling of CO 2 -water-caprock interactions at a potential CO 2 storage site in Turkey 53

6 Table 6. Kinetic parameters for mineral dissolution and precipitation (Palandri and Kharaka, 24) Mineral Parameters for kinetic rate law Neutral mechanism Acid mechanism Base mechanism k 25 (mol/m 2 /s) E a (kj/mol) k 25 (mol/m 2 /s) E a (kj/mol) n (H + ) k 25 (mol/m 2 /s) E a (kj/mol) n (H + ) Hematite 2.514E E Kaolinite 6.918E E E Magnesite 4.571E E Quartz 23E Siderite E E Dolomite E E Aragonite 4.571E E Anhydrite 6.457E perature, k 25 and the activation energy, E a. These parameters are needed for the calculation of kinetic rate and are taken from Palandri and Kharaka (24), who compiled and fitted experimental data reported by many investigators. A detailed list of original data sources is given in Palandri and Kharaka (24). For all minerals, it is assumed that the dissolution rate equals the precipitation rate (Xu et al., 24b). Calcite is considered to be in equilibrium as its kinetics is relatively fast. The temperature dependence of the reaction rate constant is expressed via an Arrhenius equation. CO 2 injection CO 2 Injection simulation scenario CO 2 was injected into the bottom 1 m of the reservoir at a rate of 1 Million tons/year for 5 years. A 2-D radially symmetric model was developed in which the effect of gravity was taken into account. Post-injection period of 1 years Simulation of the CO 2 injection process was stopped after 5 years of injection. Thereafter, the simulation was continued for 1 years without CO 2 injection in order to investigate the evolution of the caprock mineralogy and of the water chemistry for a longer time period and to determine the long-term effect of CO 2 injection on the porosity and permeability of the Sayindere caprock. RESULTS AND DISCUSSION The simulation results are presented in 2-D plots as a function of depth and radial distance up to 1, m at discrete time intervals of 1 and 5 years of CO 2 injection simulation and 5 and 1 years of post-injection simulation. Farther than 1, m in radial distance, no significant changes were observed in the model. CO 2 saturation Figures 2ad shows the spatial and temporal saturation of the injected supercritical CO 2 after 1 and 5 years of injection and 5 and 1 years of post-injection. The injected CO 2 migrated towards the Sayindere caprock from the injection point in the reservoir due to buoyancy and started forming a CO 2 plume beneath this caprock after 1 year of injection (Fig. 2a). The top of the CO 2 plume reached radial extents of.5 and 3.5 km, respectively, after 1 and 5 years of injection (Figs. 2a and b). The maximum saturation of the injected CO 2 was around the injection point after 1 year of injection. After CO 2 injection ceased, the top of the CO 2 plume reached radial extents of 6. and 7. km, respectively, at 5 and 1 years post-injection. It was trapped below the caprock and mainly existed in the upper 25 and 15 m of the reservoir. The maximum CO 2 saturation was.6 at 5 and 1 years of post-injection (Figs. 2c and d). Dissolved CO 2 mass fraction Figures 2eh shows the spatial and temporal distribution of mass fraction of dissolved CO 2, XCO 2 (aq) (mass of CO 2 dissolved by mass of aqueous phase) at the temporal intervals of after 1 and 5 years injection and 5 and 1 years of post-injection after injection. The initial mass fraction of dissolved CO 2 in the formation brine before the injection was negligible. Dissolution of CO 2 into brine is assumed to be an equilibrium process, and thus, an instantaneous one. The dissolution of supercritical CO 2 in the brine produced a maximum dissolved CO 2 mass fraction of ~5 after 1 year of injection, which then remained constant over the remaining simulation period. An increase in the spatial extent of XCO 2 (aq) in the brine was observed due to mixing and convection of the CO 2 saturated and unsaturated brines over the simulation times. ph and HCO 3 concentration Figures 3ah shows the areal and temporal distribution of the ph of the formation water and the HCO 3 concentration in the brine 1 and 5 years after injection began and 5 and 1 years after injection ceased. Prior 54 C. Dalkhaa and E. Okandan

7 Z(m) Z (m) Z (m) Z (m) S g XCO 2 (aq) a) 1 year of injection e) 1 year of injection b) 5 years of injection f) 5 years of injection c) 5 years of post-injection g) 5 years of post-injection d) h) 1 years of post-injection 1 years of post-injection Fig. 2. Spatial distribution of the saturation of supercritical CO 2 (fraction) and dissolved CO 2 mass fraction over the simulation time. to the injection, the initial ph of the caprock and reservoir brine were found to have neutral values of 7.4 and 7.8, respectively, from the simulation result of equilibration between the rock minerals and formation water for each formation. Following CO 2 injection, a decrease in ph was observed around the injection point due to its dissolution in the brine. After 1 year of injection, the ph dropped to 4.5 from 7.8 in the CO 2 affected part of the reservoir (Fig. 3a). No significant ph drops were observed in the caprock after 1 year of injection. However, the ph 2-D numerical modeling of CO 2 -water-caprock interactions at a potential CO 2 storage site in Turkey 55

8 Z Z b) 5 years of injection 8. f) 5 years of injection c) 5 years of post-injection 8. g) 5 years of post-injection Z Z ph HCO - 1 year of injection e) year of injection d) 1 years of post-injection h) a) 1 years of post-injection Fig. 3. Spatial distribution of the brine ph and HCO 3 concentration (mol/kg) over the simulation time. 1.2 in the caprock brine dropped to ~5. in the bottom 1, 2, and 3 m of the caprock after and 5 years of injection and 5 and 1 of post-injection, respectively (Figs. 3b, c and d). In the reservoir, the ph was still ~4.5 and its areal extent increased to ~4, 7, and 8 km after 5 years of injection, and 5, and 1 years of postinjection, respectively. Following the injection, as a result of both CO 2 dissolution in the brine and calcite dissolution, the HCO 3 concentration increased to 1.2 mol/kg from its initial value of 15 mol/kg and its areal extent in the reservoir increased over the simulation time (Figs. 3eh). The HCO 3 56 C. Dalkhaa and E. Okandan

9 Z c) 5 years of post-injection g) 2 Z Z Z Calcite Ca a) 2+ 1 year of injection e) 1 year of injection b) 5 years of injection f) 5 years of injection 2 d) 1 years of post-injection years of post-injection h) 2 1 years of post-injection Fig. 4. Spatial distribution of calcite abundance (change in volume fraction) and Ca 2+ concentration (mol/kg) over the simulation time concentration was observed to be very low, almost reaching a zero value, around the injection point during the injection process (Figs. 3e and f). This is most likely due to formation of a dehydrated region where brine was being completely displaced by CO 2 injection in that area due to the high injection rate. However, at 5 and 1 years post-injection, the brine displacement disappeared since the brine was brought back due to the capillary force as CO 2 migrated to the upper part of the reservoir (Figs. 3g and h). 2-D numerical modeling of CO 2 -water-caprock interactions at a potential CO 2 storage site in Turkey 57

10 An increase in the HCO 3 concentration in the caprock brine was also observed, reaching a value of ~ mol/kg from its initial value of 15 mol/kg at the 5 and 1 years after injection ceased. Calcite and Ca 2+ concentration Figures 4ah shows the calcite abundance and concentration of Ca 2+ in the brine after 1 and 5 years of injection and 5 and 1 years after injection ceased in the 2-D model. The calcite dissolution was mainly observed in the reservoir. The maximum absolute change in the volume fraction of calcite was -1e-4 and -2e-4 in the reservoir after 5 years of injection and 1 years after injection ceased, respectively (Figs. 4b and d). Minor calcite dissolution was also observed in the lower part of the caprock, as the brine became acidic (ph of ~5.) due to the intrusion of the injected CO 2 into the bottom of the reservoir. Due to calcite dissolution, an increase in the concentration of Ca 2+ in the formation water was observed in the CO 2 affected areas of both the caprock and the reservoir. The concentration of Ca 2+ in the brine in these areas was increased to 2 mol/kg from its initial value of 3e- 4 mol/kg. The Ca 2+ concentration was observed to be very low, almost reaching a zero value around the injection point, during the injection process (Figs. 4e and f). This is again due to formation of a dehydrated region explained in Subsection Simulated model. The dehydrated region, however, had disappeared 5 and 1 years after injection ceased as the brine was brought back due to the capillary force as CO 2 migrated to the upper part of the reservoir due to buoyancy (Figs. 4g and h). The fate of the injected CO 2 No supercritical CO 2 diffusion into the caprock was observed after 1 year of injection. However, after 5 years of injection, about 2 Mton of the injected CO 2 had diffused into the caprock. This is less than 5% of the total injected CO 2. The amount of CO 2 that had diffused into the caprock 5 and 1 years after injection ceased was 2 and 35 Mton, respectively. These values are less than 1% of the 5 Mton injected in total. Furthermore, the injected supercritical CO 2 only diffused into the lowermost 2.5 m of the caprock during these periods. As the CO 2 was dissolved in the brine, it caused a ph drop in more extended area in the caprock as a result of convective mixing in the brine. The majority of the injected CO 2 was trapped below the caprock as a free phase and dissolved in the brine via solubility trapping in the reservoir during the simulation period Mton of the injected CO 2 was in the supercritical phase and 15.4 Mton was dissolved in the reservoir brine 1 years after injection ceased. No injected CO 2 was stored via mineral trapping because the reservoir is a carbonate rock and calcite was dissolved over the simulation time. Porosity and permeability change No significant change in the volume fractions of the other primary and secondary minerals was observed. Calcite dissolution was observed in the model. However, no associated changes in the porosity and permeability as a result of the calcite dissolution were observed in the reservoir or the caprock. A porosity change is calculated from volume change as the result of precipitation and dissolution of rocks in the code TOUGHREACT. A permeability change related to porosity change is calculated using a simplified cubic law (Xu et al., 24a). Increases in porosity and permeability in the caprock formation could potentially increase the extent of possible existing fracture zones or faults in the rock, and this might create leakage paths for the injected CO 2 to migrate up to the overlying formation from the storage reservoir. SENSITIVITY SIMULATION Permeability anisotropy Permeability anisotropy ratio (ratio of vertical to horizontal permeability) is a crucial parameter as it provides information about the flow behavior within a formation. A sensitivity simulation for a different value of this parameter was carried out to observe its effect on the numerical results. Figure 5 shows the spatial and temporal distribution of the saturation of the injected CO 2 with a permeability anisotropy ratio of 1 in parts ad and of.1 in parts eh. The spatial distributions of CO 2 saturation for both cases were similar. With a lower permeability anisotropy ratio; i.e., higher vertical permeability, the injected CO 2 reached the caprock a bit faster, and thus, the front of the CO 2 plume had traveled a longer distance in the radial direction. For instance, in the case of a permeability anisotropy ratio of 1, the CO 2 plume after 5 years of injection and 1 years after injection ceased was 3.5 km and 7. km long, respectively, in the radial direction (Figs. 5b and d); whereas in the case of.1 of permeability anisotropy ratio, the CO 2 plume after 5 years of injection and 1 years after injection ceased was 4 km and 7.5 km long, respectively, in the radial direction (Figs. 5f and h). The maximum saturation 1 years after injection ceased was.6 in both cases (Figs. 5d and h). Reactive surface area The surface area of a mineral is an important parameter for kinetic reactions as the kinetic rate is the product of the reactive surface area and the rate constant. There are large uncertainties in natural variability and the meas- 58 C. Dalkhaa and E. Okandan

11 Z(m) Z (m) Z (m) Z (m) a) 1 year of injection b) 5 years of injection f) 5 years of injection.6.2 c) 5 years of post-injection Fig. 5. Spatial distribution of the saturation of supercritical CO 2 (fraction) with permeability anisotropy values of 1 (ad) and.1 (eh). e) 1 year of injection.6.2 g) 5 years of post-injection d) h) 1 years of post-injection 1 years of post-njection urement of the reactive surface area. The surface areas from the work of Xu et al. (21) were used for this modeling study (Table 5). They were based on the work of Sonnenthal et al. (25) and had already been reduced by two orders of magnitude from the surface roughnessbased surface area. Interaction with the minerals is generally expected to occur only at selective sites of the mineral surface, and the actual reactive surface area could be between one and three orders of magnitude less than the surface roughness based surface area (Lasaga, 1995; Zerai 2-D numerical modeling of CO 2 -water-caprock interactions at a potential CO 2 storage site in Turkey 59

12 et al., 26). Therefore, we decreased the surface area given in Table 5 by a further order of magnitude and performed a sensitivity simulation to evaluate how the numerical results are affected by the uncertainty. However, there were no relevant changes observed in the geochemical behavior of the system. Gherardi et al. (27) also performed the same kind of sensitivity simulation and concluded that they did not observe any significant variation. CONCLUSION A 2-D modeling and simulation study of CO 2 injection into a reservoir was performed, and its impact on the caprock was investigated. The results of the simulation of a CO 2 injection process showed dissolution of calcite in both the storage reservoir and the caprock. However, no significant increase in the porosity and permeability due to calcite dissolution was observed in either formation. Minor diffusion of the injected CO 2 into the bottom layer of the caprock was observed over the simulation years. 7% of the injected CO 2 was stored via hydrodynamical trapping under the caprock as a free phase plume, and the rest was dissolved in the reservoir brine via solubility trapping. Sensitivity analyses were performed to determine how the permeability anisotropy and reactant surface area of minerals affect the numerical results. Minor variations in the spatial distributions of CO 2 saturation were observed when the permeability anisotropy was changed. No significant variation in the numerical results was observed when the surface area of minerals was changed. This simulation study suggests that the Sayindere caprock is a very good impermeable sealing rock for potential CO 2 storage. Acknowledgments The authors would like to thank the Turkish Petroleum Corporation (TPAO) for providing the cores from the Sayindere formation and water analysis of the Caylarbasi reservoir formation. The staff of the water laboratory of the Petroleum Research Center and the plasma laboratory of the Central Laboratory, METU, are thanked for their analysis and support. Further, the Departments of Geological Engineering and Metallurgical Engineering, METU, deserve special thanks for the mineralogical analyses. One of the authors, Chantsalmaa Dalkhaa would like to thank The Scientific and Technological Research Council of Turkey (TUBITAK) for awarding her the PhD Fellowship. The valuable review of the manuscript by Prof. Dr. Bernhard Mayer from the Applied Geochemistry Group at the Department of Geoscience at the University of Calgary, Calgary, Alberta, Canada is greatly appreciated. REFERENCES Apps, J. A., Zheng, L., Spycher, N., Birkholzer, J. T., Kharaka, Y., Thordsen, J., Kakouros, E. and Trautz, R. (211) Transient changes in shallow groundwater chemistry during the MSU ZERT CO 2 injection experiment. Energia Procedia 4, Baklid, A., Korbol, K. and Owren, G. (1996) Sleipner vest CO 2 disposal, CO 2 injection into swallow underground aquifer. SPE 366. Class, H., Edigbo, A., Helmig, R., Dahle, H. K., Nordbotten, J. M., Celia, M. A., Audigane, P., Darcis, M., Ennis-King, J., Fan, Y., Flemisch, B., Gasda, S. E., Jin, M., Krug, S., Labregere, D., Naderi, B. A., Pawar, R. J., Sbai, A., Thomas, S. G., Trenty, L. and Wei, L. (29) A Benchmark study on problems related to CO 2 storage in geologic formations: Summary and discussion of the results. Computat. Geosci. 13(4) Gaus, I., Audigane, P., Andre, L., Lions, J., Jacquemet, N., Durst, P., Czernichowski-Lauriol, I. and Azaroual, M. (28) Geochemical and solute transport modeling for CO 2 storage, What to expect from it? International Journal of Greenhouse Gas Control 2, Gherardi, F., Xu, T. and Pruess, K. (27) Numerical modeling of self-limiting and self-enhancing caprock alteration induced by CO 2 storage in a depleted gas reservoir. Chem. Geol. 244, Gunay, A. and Ubay, B. (27) First National Communication of Turkey on Climate Change. Helgeson, H. C. and Kirkham, D. H. (1974) Theoretical prediction of the thermodynamic behavior of aqueous electrolytes at high pressures and temperatures: II. Debye-Huckel parameters for activity coefficients and relative partial molal properties. Amer. J. Sci. 274, Holubnyak, Y. H., Hawthorne, S. B., Mibeck, B. A. F., Miller, D. J., Bremer, J. M., Sorensen, J. A., Steadman, E. N. and Harju, J. A. (211) Modeling CO 2 H 2 Swaterrock interactions at Williston Basin reservoir conditions. Energia Procedia 4, IPCC (25) IPCC Special Report on Carbon Dioxide Capture and Storage (Metz, B., Davidson, O., de Coninck, H. C., Loos, M. and Meyer, L. A., eds.), prepared by Working Group III of the Intergovernmental Panel on Climate Change, Cambridge Univ. Press, Cambridge, U.K./New York, U.S.A. Issever, K., Pamir, A. N. and Tirek, A. (1993) Performance of a heavy-oil field under CO 2 injection, Bari Raman, Turkey. SPE 2883-PA. Lasaga, A. C. (1995) Fundamental approaches in describing mineral dissolution and precipitation rates. Chemical Weathering Rates of Silicates Minerals; An Overview (White, A. F. and Brantley, S. L., eds.), Reviews in Mineralogy and Geochemistry 31, 2386, BookCrafters, Chelsea, MI. Okandan, E., Karakece, Y., Çetin, Ç., Topkaya, I., Parlaktuna, M., Akιn, S., Bulbul, S., Dalkhaa, C., Anbar, S., Cetinkaya, C., Ermis, I., Yιlmaz, M., Ustun, V., Yapan, K., Erten, A. T., Demiralιn, Y. and Akalan, E. (29) Assessment of CO 2 emissions from industrial sites, potential for underground storage in Turkey and modeling of storage in an oil field, TUBITAK, KAMAG Project 16G11. Okandan, E., Karakece, Y., Topkaya, I., Parlaktuna, M., Akin, S., Bulbul, S., Dalkhaa, C., Anbar, S., Cetinkaya, C., Ermis, I., Yilmaz, M., Ustun, V., Yapan, K., Erten, A. T., Demiralin, Y. and Akalin, E. (211) Assessment of CO 2 storage poten- 51 C. Dalkhaa and E. Okandan

13 tial in Turkey, modeling and a prefeasibility study for injection into an oil field. Energy Procedia 4, Palandri, J. and Kharaka, Y. K. (24) A Compilation of rate parameters of water-mineral interaction kinetics for geochemical modeling. U.S. Geological Survey Open File Report Pruess, K. (25) ECO2N: A TOUGH2 fluid property module for mixtures of water, NaCl and CO 2. Lawrence Berkeley National Laboratory Report LBNL-57592, Berkeley, California. Pruess, K., García, J., Kovscek, T., Oldenburg, C., Rutqvist, J., Steefel, C., and Xu, T. (24) Code intercomparison builds confidence in numerical simulation models for geologic disposal of CO 2. Energy 29, Shevalier, M., Nightingale, M., Mayer, B. and Hutcheon, I. (211) TOUGHREACT modeling of the fate of CO 2 injected into a H 2 S containing saline aquifer: example of the Wabamum Area Sequestration project (WASP). Energy Procedia 4, Sonnenthal, E., Ito, A., Spycher, N., Yui, M., Apps, J., Sugita, Y., Conrad, M. and Kawakami, S. (25) Approaches to modeling coupled thermal, hydrological, and chemical processes in the Drift Scale Heater Test at Yucca Mountain. Int. J. Rock Mech. Min. 42, Van Genuchten, M. Th. (198) A Closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44, White, D. J., Burrowes, G., Davis, T., Hajnal, Z., Hirsche, K., Hutcheon, I., Majer, E., Rostron, B. and Whittaker, S. (24) Greenhouse gas sequestration in abandoned oil reservoirs: The International Energy Agency Weyburn pilot project. GSA Today 14(7), 41. Xu, T., Apps, J. and Pruess, K. (23) Reactive geochemical transport simulation to study mineral trapping for CO 2 disposal in deep arenaceous formations. J. Geophys. Res. 18(B2), 271, doi:1.129/22jb1979. Xu, T., Sonnenthal, E., Spucher, N. and Pruess, K. (24a) TOUGHREACT User s Guide: A Simulation Program for Non-isothermal Multiphase Reactive Geochemical Transport in Variably Saturated Geologic Media. Lawrence Berkeley National Laboratory Report LBNL-5546, Berkeley, California. Xu, T., Apps, J. A. and Pruess, K. (24b) Numerical simulations of CO 2 disposal by mineral trapping in deep aquifers. Appl. Geochem. 19, Xu, T., Apps, J. A. and Pruess, K. (25) Mineral sequestration of carbon dioxide in a sandstone-shale system. Chem. Geol. 217, Xu, T., Kharaka, Y. K., Doughty, C., Freifels, B. M. and Daley, T. M. (21) Reactive transport modeling to study changes in water chemistry induced by CO 2 injection at the Frio-I Brine Pilo. Chem. Geol. 271, Zhang, W., Li, Y., Xu, T., Cheng, H., Zheng, Y. and Xiong, P. (29). Long-term variations of CO 2 trapped in different mechanisms in deep saline formations: A case study of the Songliao Basin, China. International Journal of Greenhouse Gas Control 3, S161S18. Zheng, L., Apps, J. A., Zhang, Y., Xu, T. and Birkholzer, J. T. (29) On mobilization of lead and arsenic in groundwater in response to CO 2 leakage from deep geological storage, Chem. Geol. 27, Zerai, B., Saylor, B. Z. and Matisoff, G. (26) Computer simulation of CO 2 trapped through mineral precipitation in the Rose Run Sandstone, Ohio. Appl. Geochem. 21, D numerical modeling of CO 2 -water-caprock interactions at a potential CO 2 storage site in Turkey 511