Laboratory calibration of the seismo-acoustic response of CO2 saturated sandstones

Ultrasonic experiments were undertaken on CO₂ flooded sandstone core samples, both synthetic sandstones and core plugs from the CRC1 CO₂ injection well in the Otway Basin, Victoria, South Eastern. Australia. The aim of these laboratory tests was to investigate the effects of CO₂ as a pore fluid on the seismo-acoustic response of the sandstone and ultimately to provide an indication of the sensitivity of time-lapse seismic imaging of the eventual CO₂/CH₄ plume in the Otway, Waarre C formation.The synthetic sandstones were manufactured using both a proprietary calcium in situ precipitation (CIPS) process and a silica cementing technique. Samples were tested in a computer controlled triaxial pressure cell where pore pressures can be controlled independently of the confining pressures. The pressure cell is equipped with ultrasonic transducers housed in the loading platens. Consequently, effective pressures equivalent to those expected in the reservoir can be applied while ultrasonic testing is undertaken. Both compressional, P and shear waves, S were recorded via a digital oscilloscope at a range of effective pressure steps. Pore pressures were varied from 4 MPa to 17 MPa to represent both the gaseous and liquid phase regions of the CO₂ phase diagram. Similar experiments were conducted on core plugs from the Waarre C reservoir horizon obtained from the CRC1 injection well, but with an intervening brine-saturated step and in some cases with a CO₂/CH₄ mix of 80%/20% molar fraction which is representative of the field situation. However, the pore pressure in these experiments was held at 4 MPa. Finally, acoustic impedances and reflection coefficients were calculated for the reservoir using Gassmann theory and the implications for imaging the CO₂ plume is discussed.     

Model prediction on the rise of pCO2 in uniform flows by leakage of CO2 purposefully stored under the seabed

A numerical study was conducted to predict pCO₂ change in the ocean on a continental shelf by the leakage of CO₂, which is originally stored in the aquifer under the seabed, in the case that a large fault connects the CO₂ reservoir and the seabed by an earthquake or other diastrophism. The leakage rate was set to be 6.025 × 10⁻⁴ kg/m²/sec from 2 m × 100 m fault band, which corresponds to 3800 t-CO₂/year, referring to the monitored seepage rate from an existing EOR field. The target space in this study was limited to the ocean above the seabed, the depth of which was 200 or 500 m. The computational domain was idealistically rectangular with the seabed fault-band perpendicular to the uniform flow. The CO₂ takes a form of bubbles or droplets, depending on the depth of water, and their behaviour and dissolution were numerically simulated during their rise in seawater flow. The advection–diffusion of dissolved CO₂ was also simulated. As a result, it was suggested that the leaked CO₂ droplets/bubbles all dissolve in the seawater before spouting up to the atmosphere, and that the increase in pCO₂ in the seawater was smaller than 500 μ atm.     

A model for the CO2 capture potential

Global warming is a result of increasing anthropogenic CO₂ emissions, and the consequences will be dramatic climate changes if no action is taken. One of the main global challenges in the years to come is therefore to reduce the CO₂ emissions.Increasing energy efficiency and a transition to renewable energy as the major energy source can reduce CO₂ emissions, but such measures can only lead to significant emission reductions in the long-term. Carbon capture and storage (CCS) is a promising technological option for reducing CO₂ emissions on a shorter time scale.A model to calculate the CO₂ capture potential has been developed, and it is estimated that 25 billion tonnes CO₂ can be captured and stored within the EU by 2050. Globally, 236 billion tonnes CO₂ can be captured and stored by 2050. The calculations indicate that wide implementation of CCS can reduce CO₂ emissions by 54% in the EU and 33% globally in 2050 compared to emission levels today.Such a reduction in emissions is not sufficient to stabilize the climate. Therefore, the strategy to achieve the necessary CO₂ emissions reductions must be a combination of (1) increasing energy efficiency, (2) switching from fossil fuel to renewable energy sources, and (3) wide implementation of CCS.     

CO2SINK—From site characterisation and risk assessment to monitoring and verification: One year of operational experience with the field laboratory for CO2 storage at Ketzin, Germany

The CO₂SINK pilot project at Ketzin is aimed at a better understanding of geological CO₂ storage operation in a saline aquifer. The reservoir consists of fluvial deposits with average permeability ranging between 50 and 100 mDarcy. The main focus of CO₂SINK is developing and testing of monitoring and verification technologies. All wells, one for injection and two for observation, are equipped with smart casings (sensors behind casing, facing the rocks) containing a Distributed Temperature Sensing (DTS) and electrodes for Electrical Resistivity Tomography (ERT). The in-hole Gas Membrane Sensors (GMS) observed the arrival of tracers and CO₂ with high temporal resolution. Geophysical monitoring includes Moving Source Profiling (MSP), Vertical Seismic Profiling (VSP), crosshole, star and 4-D seismic experiments. Numerical models are benchmarked via the monitoring results indicating a sufficient match between observation and prediction, at least for the arrival of CO₂ at the first observation well. Downhole samples of brine showed changes in the fluid composition and biocenosis. First monitoring results indicate anisotropic flow of CO₂ coinciding with the “on-time” arrival of CO₂ at observation well one (Ktzi 200) and the later arrival at observation well two (Ktzi 202). A risk assessment was performed prior to the start of injection. After one year of operations about 18,000 t of CO₂ were injected safely.     

Embodied energy and CO2 in UK dimension stone

A process based life cycle assessment of dimension stone production in the UK has been carried out according to PAS 2050. From a survey of eight production operations, on a cradle-to-site basis for UK destinations the carbon footprint of sandstone is 77 kgCO₂e/tonne, that of granite is 107 kgCO₂e/tonne and that of slate is 251 kgCO₂e/tonne. These values are considerably higher for stone imported from abroad due to the impact of transport. Reducing the reliance on imported stone will contribute to emissions reduction targets as well as furthering the goals of sustainable development.     

Study on the failure mechanism of potassium-based sorbent for CO2 capture and the improving measure

With thermogravimetric apparatus (TGA), X-ray diffraction (XRD) and barium sulfate gravimetric methods, the carbonation reactivities of K₂CO₃ and K₂CO₃/Al₂O₃ in the simulated flue gases with SO₂ are investigated and the reaction equations are inferred. Results show that there are KHCO₃ and K₂SO₃ generated. The generation K₂SO₃ reduces the utilization ratio of the sorbent. H₂O may accelerates the sulfation reaction of AR K₂CO₃ as K₄H₂(CO₃)₃·1.5H₂O is generated in the reaction among K₂CO₃, SO₂ and H₂O. K₂SO₃ is directly generated from sulfation reaction of K₂CO₃/Al₂O₃, because there are K₂CO₃·1.5H₂O and K₂SO₃ generated in the reaction among K₂CO₃/Al₂O₃, SO₂ and H₂O. K₂CO₃·1.5H₂O does not react with SO₂, and K₂CO₃·1.5H₂O/Al₂O₃ reacts with SO₂ slowly. Compare with the reaction process without H₂O pretreatment, the reaction rates of KAl30 increased after H₂O pretreatment and the failure ratio is about a half of that without H₂O pretreatment. So, K₂CO₃/Al₂O₃ shows good carbonation and anti-sulfation characteristic after H₂O pretreatment.     

Scenario simulations of CO2 injection feasibility, plume migration and storage in a saline aquifer, Scania, Sweden

Deep saline aquifers have large capacity for geological CO₂ storage, but are generally not as well characterized as petroleum reservoirs. We here aim at quantifying effects of uncertain hydraulic parameters and uncertain stratigraphy on CO₂ injectivity and migration, and provide a first feasibility study of pilot-scale CO₂ injection into a multilayered saline aquifer system in southwest Scania, Sweden. Four main scenarios are developed, corresponding to different possible interpretations of available site data. Simulation results show that, on the one hand, stratigraphic uncertainty (presence/absence of a thin mudstone/claystone layer above the target storage formation) leads to large differences in predicted CO₂ storage in the target formation at the end of the test (ranging between 11% and 98% of injected CO₂ remaining), whereas other parameter uncertainty (in formation and cap rock permeabilities) has small impact. On the other hand, the latter has large impact on predicted injectivity, on which stratigraphic uncertainty has small impact. Salt precipitation at the border of the target storage formation affects CO₂ injectivity for all considered scenarios and injection rates. At low injection rates, salt is deposited also within the formation, considerably reducing its availability for CO₂ storage.     

Methodology for CO2 chain analysis

The paper presents a methodology for CO₂ chain analysis with particular focus on the impact of technology development on the total system economy. The methodology includes the whole CO₂ chain; CO₂ source, CO₂ capture, transport and storage in aquifers or in oil reservoirs for enhanced oil recovery. It aims at supporting the identification of feasible solutions and assisting the selection of the most cost-effective options for carbon capture and storage. To demonstrate the applicability of the methodology a case study has been carried out to illustrate the possible impact of technology improvements and market development. The case study confirms that the CO₂-quota price to a large extent influence the project economy and dominates over potential technology improvements. To be economic feasible, the studied chains injecting the CO₂ in oil reservoirs for increased oil production require a CO₂-quota price in the range of 20–27 €/tonne CO₂, depending on the technology breakthrough. For the chains based on CO₂ storage in saline aquifers, the corresponding CO₂-quota price varies up to about 40 €/tonne CO₂.     

Pipeline design for a least-cost router application for CO2 transport in the CO2 sequestration cycle

CO₂ capture and geological storage (CCS) is considered as a viable option to mitigate greenhouse gas emissions during the transition phase towards the use of clean and renewable energy. This paper concentrates on the transport of CO₂ between source (CO₂ capture at plants) and sink (geological storage reservoirs). In the cost estimation of CO₂ transport, the pipeline diameter plays an important role. In this respect, the paper reviews equations that were used in several reports on CO₂ pipeline transport. As some parameters are not taken into account in these equations, alternative formulas are proposed which calculate the proper inner diameter size based on flow rate, pressure drop per unit length, CO₂ density, CO₂ viscosity, pipeline material roughness and topographic height differences (the Darcy–Weisbach solution) and, in addition, on the amount and type of bends (the Manning solution). Comparison between calculated diameters using the reviewed and the proposed equations demonstrate the important influence of elevation difference (which is not considered in the reviewed equations) and pipeline material roughness-related factor on the calculated diameter. Concerning the latter, it is suggested that a Darcy–Weisbach roughness height of 0.045 mm better corresponds to a Manning factor of 0.009 than higher Manning values previously proposed in literature. Comparison with the actual diameter of the Weyburn pipeline confirms the accuracy of the proposed equations. Comparison with other existing CO₂ pipelines (without pressure information) indicate that the pipelines are designed for lower pressure gradients than 25 Pa/m or for (future) higher flow rates. The proposed Manning equation is implemented in an economic least-cost route planner in order to obtain the best economic solution for pipeline trajectory and corresponding diameter.     

Analytical solution to evaluate salt precipitation during CO2 injection in saline aquifers

Carbon dioxide sequestration in deep saline aquifers is a means of reducing anthropogenic atmospheric emissions of CO₂. Among various mechanisms, CO₂ can be trapped in saline aquifers by dissolution in the formation water. Vaporization of water occurs along with the dissolution of CO₂. Vaporization can cause salt precipitation, which reduces porosity and impairs permeability of the reservoir in the vicinity of the wellbore, and can lead to reduction in injectivity. The amount of salt precipitation and the region in which it occurs may be important in CO₂ storage operations if salt precipitation significantly reduces injectivity. Here we develop an analytical model, as a simple and efficient tool to predict the amount of salt precipitation over time and space. This model is particularly useful at high injection velocities, when viscous forces dominate.First, we develop a model which treats the vaporization of water and dissolution of CO₂ in radial geometry. Next, the model is used to predict salt precipitation. The combined model is then extended to evaluate the effect of salt precipitation on permeability in terms of a time-dependent skin factor. Finally, the analytical model is corroborated by application to a specific problem with an available numerical solution, where a close agreement between the solutions is observed. We use the results to examine the effect of assumptions and approximations made in the development of the analytical solution. For cases studied, salt saturation was a few percent. The loss in injectivity depends on the degree of reduction of formation permeability with increased salt saturation. For permeability-reduction models considered in this work, the loss in injectivity was not severe. However, one limitation of the model is that it neglects capillary and gravity forces, and these forces might increase salt precipitation at the bottom of formation particularly when injection rate is low.     

Dynamis CO2 quality recommendations

In the carbon capture and storage (CCS) chain, transport and storage set different requirements for the composition of the gas stream mainly containing carbon dioxide (CO₂). Currently, there is a lack of standards to define the required quality for CO₂ pipelines. This study investigates and recommends likely maximum allowable concentrations of impurities in the CO₂ for safe transportation in pipelines. The focus is on CO₂ streams from pre-combustion processes. Among the issues addressed are safety and toxicity limits, compression work, hydrate formation, corrosion and free water formation, including the cross-effect of H₂S and H₂O and of H₂O and CH₄.     

Designing a supercritical steam cycle to integrate the energy requirements of CO2 amine scrubbing

Absorption by chemical solvents combined with CO₂ long-term storage appears to offer interesting and commercial applicable CO₂ capture technology. However one of the main disadvantages is related to the large quantities of heat required to regenerate the amine solvent that means an important power plant efficiency penalty. Different studies have analyzed alternatives to reduce the heat duty on the reboiler and the thermal integration requirements on existing power cycles. In these studies integration principles have been well set up, but there is a lack of information about how to achieve an integrated design and the thermal balances of the modified cycle flowsheet. This paper proposes and provides details about a set of modifications of a supercritical steam cycle to overcome the energy requirements through energetic integration with the aim of reducing the efficiency and power output penalty associated with CO₂ capture process. Modifications include a new designed low-pressure heater flowsheet to take advantage of the CO₂ compression cooling for postcombustion systems and integration of amine reboiler into a steam cycle. It has been carried out several simulations in order to obtain power plant performance depending on sorbent regeneration requirements.     

Numerical parametric study on CO2 capture by indirect thermal swing adsorption

Post-combustion CO₂ capture remains one of the most-challenging issue to lower CO₂ emissions of existing power plants or heavy industry installations because of strong economy and energy efficiency aspects. The major issue comes from CO₂ dilution (4% for NGCC and 14% for PC) and the high flow rates to be treated. Furthermore, CO₂ purity has to be higher than 95% with recovery at 90%, to match the transportation/injection requirements.The MEA absorption process remains the reference today but its energy consumption (about 3 MJ/kg_CO2) and the amine consumption are still challenging drawbacks.The interest of CO₂ capture by indirect TSA (Temperature Swing Adsorption) was demonstrated experimentally in a previous work. The aim of this paper is to present the results of a numerical parametric study. Two main parameters are explored: the desorption temperature (100–200 °C) and the purge flow rate (0.1–0.5 Ndm³ min⁻¹). Four performance indicators are evaluated: CO₂ purity, recovery, productivity and specific energy consumption.Results show that purity above 95% can be achieved. Keeping the 95% target, it is possible to achieve recovery at 81% with productivity at 57.7 g_CO2/kg_ads h and a specific energy consumption of 3.23 MJ/kg_CO2, which is less than for the reference MEA process.Comparison with other adsorption processes exhibits that this process has good potential especially since some improvements are still expected from further research.     

Viscosities, thermal conductivities and diffusion coefficients of CO2 mixtures: Review of experimental data and theoretical models

Accurate experimental data on the thermo-physical properties of CO₂-mixtures are pre-requisites for development of more accurate models and hence, more precise design of CO₂ capture and storage (CCS) processes. A literature survey was conducted on both the available experimental data and the theoretical models associated with the transport properties of CO₂-mixtures within the operation windows of CCS. Gaps were identified between the available knowledge and requirements of the system design and operation. For the experimental gas-phase measurements, there are no available data about any transport properties of CO₂/H₂S, CO₂/COS and CO₂/NH₃; and except for CO₂/H₂O(/NaCl) and CO₂/amine/H₂O mixtures, there are no available measurements regarding the transport properties of any liquid-phase mixtures. In the prediction of gas-phase viscosities using Chapman–Enskog theory, deviations are typically <2% at atmospheric pressure and moderate temperatures. The deviations increase with increasing temperatures and pressures. Using both the Rigorous Kinetic Theory (RKT) and empirical models in the prediction of gas-phase thermal conductivities, typical deviations are 2.2–9%. Comparison of popular empirical models for estimation of gas-phase diffusion coefficients with newer experimental data for CO₂/H₂O shows deviations of up to 20%. For many mixtures relevant for CCS, the diffusion coefficient models based on the RKT show predictions within the experimental uncertainty. Typical reported deviations of the CO₂/H₂O system using empirical models are below 3% for the viscosity and the thermal conductivity and between 5 and 20% for the diffusion coefficients. The research community knows little about the effect of other impurities in liquid CO₂ than water, and this is an important area to focus in future work.     

Modelling of CO2 arrival time at Ketzin – Part I

The injection of CO₂ at the Ketzin storage site and the chemical detection of its arrival in the observation well allowed testing of different numerical simulation codes. ECLIPSE 100 (E100, black-oil simulator), ECLIPSE 300 (E300, compositional CO₂STORE) and MUFTE-UG were used for predictive modelling applying a constant injection rate of 1 kg s^?1 CO₂ and for a history match applying the actual variable injection rate which ranged from 0 to 0.7 kg s^?1 and averaged 0.23 kg s^?1. The geological model applied, is based on all available geophysical and geological information and has been the same for all programs.The results of the constant injection regime show a good agreement among the programs with a discrepancy of 21–33% for the CO₂ arrival times. However, it is determined from the comparison of the cumulative mass of CO₂ at the time of CO₂ arrival that the injection regime is an important factor for the accurate prediction of CO₂ migration within a saline aquifer. Comparing the actual variable injection regime with the simulations applying a constant injection rate the results are relatively inaccurate.Regarding the actual variable injection regime, which was evaluated using all three simulators, the computational results show a good agreement with the data actually measured at the first observation well. Here, the calculated arrival times exceeded the actual ones by 8.1% (E100), 9.2% (E300) and 17.7% (MUFTE-UG).It can be concluded that irrespective of the deviations of the simulations, due to combinations of different codes and slight differences in input parameters, all three programs are well equipped to give a reliable estimate of the arrival of CO₂. Deviations in the results mainly occur due to different input data and grid size choices done by the different modelling teams working independently of each other. Deviations of the simulations results compared to the actual CO₂ arrival time result from uncertainties in the implementation of the geological model, which was set up based on well log data and analogue studies.     

CO2 transportation for carbon capture and storage: Sublimation of carbon dioxide from a dry ice bank

Climate change is being caused by greenhouse gases such as carbon dioxide (CO₂). Carbon capture and storage (CCS) is of interest to the scientific community as one way of achieving significant global reductions of atmospheric CO₂ emissions in the medium term. CO₂ would be captured from large stationary sources such as power plants and transported via pipelines under high pressure conditions to underground storage. If a downward leakage from a surface transportation system module occurs, the CO₂ would undergo a large temperature reduction and form a bank of “dry ice” on the ground surface; the sublimation of the gas from this bank represents an area source term for subsequent atmospheric dispersion, with an emission rate dependent on the energy balance at the bank surface. Gaseous CO₂ is denser than air and tends to remain close to the surface; it is an asphyxiant, a cerebral vasodilator and at high concentrations causes rapid circulatory insufficiency leading to coma and death. Hence a subliming bank of dry ice represents safety hazard. A model is presented for evaluating the energy balance and sublimation rate at the surface of a solid frozen CO₂ bank under different environmental conditions. The results suggest that subliming gas behaves as a proper dense gas (i.e. it remains close to the ground surface) only for low ambient wind speeds.     

A method for quick assessment of CO2 storage capacity in closed and semi-closed saline formations

Saline aquifers of high permeability bounded by overlying/underlying seals may be surrounded laterally by low-permeability zones, possibly caused by natural heterogeneity and/or faulting. Carbon dioxide (CO₂) injection into and storage in such “closed” systems with impervious seals, or “semi-closed” systems with non-ideal (low permeability) seals, is different from that in “open” systems, from which the displaced brine can easily escape laterally. In closed or semi-closed systems, the pressure buildup caused by continuous industrial-scale CO₂ injection may have a limiting effect on CO₂ storage capacity, because geomechanical damage caused by overpressure needs to be avoided. In this research, a simple analytical method was developed for the quick assessment of the CO₂ storage capacity in such closed and semi-closed systems. This quick-assessment method is based on the fact that native brine (of an equivalent volume) displaced by the cumulative injected CO₂ occupies additional pore volume within the storage formation and the seals, provided by pore and brine compressibility in response to pressure buildup. With non-ideal seals, brine may also leak through the seals into overlying/underlying formations. The quick-assessment method calculates these brine displacement contributions in response to an estimated average pressure buildup in the storage reservoir. The CO₂ storage capacity and the transient domain-averaged pressure buildup estimated through the quick-assessment method were compared with the “true” values obtained using detailed numerical simulations of CO₂ and brine transport in a two-dimensional radial system. The good agreement indicates that the proposed method can produce reasonable approximations for storage–formation–seal systems of various geometric and hydrogeological properties.     

The impact of heterogeneity on the distribution of CO2: Numerical simulation of CO2 storage at Ketzin

Numerical modelling of multiphase flow is an essential tool to ensure the viability of long-term and safe CO₂ storage in geological formations. Uncertainties arising from the heterogeneity of the formation and lack of knowledge of formation properties need to be assessed in order to create a model that can reproduce the data available from monitoring. In this study, we investigated the impact of unknown spatial variability in the petrophysical properties within a sandy channel facies of a fluviatile storage formation using stochastic methods in a Monte Carlo approach. The stochastic method has been applied to the Ketzin test site (CO₂SINK), and demonstrates that the deterministic homogeneous model satisfactorily predicts the first CO₂ arrival time at the Ketzin site. The equivalent permeability was adjusted to the injection pressure and is in good agreement with the hydraulic test. It has been shown that with increasing small-scale heterogeneity, the sharpness of the CO₂ front decreases and a greater volume of the reservoir is affected, which is also seen in an increased amount of dissolved CO₂. Increased anisotropy creates fingering effects, which result in higher probabilities for earlier arrival times. Generally, injectivity decreases with increasing heterogeneity.     

Comparison of several packings for CO2 chemical absorption in a packed column

CO₂ capture and storage has gained widespread attention as an option for reducing greenhouse gas emissions. Chemical absorption and stripping of CO₂ with hot potassium carbonate (K₂CO₃) solutions has been used in the past, however potassium carbonate solutions have a low CO₂ absorption efficiency. Various techniques can be used to improve the absorption efficiency of this system with one option being the addition of promoters to the solvent and another option being an improvement in the mass transfer efficiency of the equipment. This study has focused on improving the efficiency of the packed column by replacing traditional packings with newer types of packing which have been shown to have enhanced mass transfer performance. Three different packings (Super Mini Rings (SMRs), Pall Rings and Mellapak) have been studied under atmospheric conditions in a laboratory scale column for CO₂ absorption using a 30 wt% K₂CO₃ solution. It was found that SMR packing resulted in a mass transfer coefficient approximately 20% and 30% higher than that of Mellapak and Pall Rings, respectively. Therefore, the height of packed column with SMR packing would be substantially lower than with Pall Rings or Mellapak. Meanwhile, the pressure drop using SMR was comparable to other packings while the gas flooding velocity was higher when the liquid load was above 25 kg m⁻² s⁻¹. Correlations for predicting flooding gas velocities and pressure drop were fitted to the experimental data, allowing the relevant parameters to be estimated for use in later design.     

Integration of counter-current relative permeability in the simulation of CO2 injection into saline aquifers

Carbon dioxide (CO₂) injection into saline aquifers is one of the promising options to sequester large amounts of CO₂ in geological formations. During as well as after injection of CO₂ into an aquifer, CO₂ migrates towards the top of the formation due to density differences between the formation brine and the injected CO₂. The time scales of CO₂ migration towards the top of an aquifer and the fraction of CO₂ that is trapped as residual gas depends strongly on the driving forces that are acting on the injected CO₂.When CO₂ migrates to the top of an aquifer, brine may be displaced downwards in a counter-current flow setting particularly during the injection period. A majority of the published work on counter-current flow settings have reported significant reductions in the associated relative permeability functions as compared to co-current measurements. However, this phenomenon has not yet been considered in the simulation of CO₂ storage into saline aquifers.In this paper we study the impact of changes in mobility for the two-phase brine/CO₂ system as a result of transitions between co- and counter-current flow settings. We have included this effect in a simulator and studied the impact of the related mobility reduction on the saturation distribution and residual saturation of CO₂ in aquifers over relevant time scales. We demonstrate that the reduction in relative permeability in the vertical direction changes the plume migration pattern and has an impact on the amount of gas that is trapped as a function of time. This is to our best knowledge the first attempt to integrate counter-current relative permeability into the simulation of injection and subsequent migration of CO₂ in aquifers. The results and analysis presented in this paper are directly relevant to all ongoing activities related to the design of large-scale CO₂ storage in saline aquifers.