Skip to main content
European Commission logo print header

Predicting and monitoring the long-term behavior of CO2 injected in deep geological formations

Final Report Summary - PANACEA (Predicting and monitoring the long-term behavior of CO2 injected in deep geological formations)

Executive Summary:
PANACEA addressed the problem of how to understand and predict the long-term fate of the stored CO2. Contribution to the achievement of this overall goal was defining and analyzing the behavior of the stored CO2 at different time and space scales, identifying the key factors and quantifying their impacts, identifying the key processes associated to the CO2 storage such as trapping mechanisms and the environmental impacts (regional pressure impacts, leakage, brine migration, mobilization of hazardous minerals trapped in the rock). This was done in hierarchical way, i.e. analyzing the fundamental processes (mixing, convective dissolution, evaporation, interfacial problems), analyzing the impacts of heterogeneity and uncertainty and developing new approaches for their quantification, developing anew paradigm for cost-efficient, reliable and safe monitoring of the CO2, analyzing monitoring technologies from the Hontomin and Heletz sites and conducting large scale simulations of contemporary (including model validation) and analogues sites, using HPC (High performance computing) technologies.

Project Context and Objectives:
PANACEA was structured in eight interrelated work-packages, as depicted in project flow-chart in Figure 4.1. The objectives of the eight work packages are:

WP02: CO2 sites and analogues

•Assemble comprehensive datasets for the calibration, validation and verification of conceptual and computational simulation models of the long term CO2 storage in deep seated saline aquifers.

•Construct a comprehensive database of pressure, temperature, stress directions and fluid composition of CO2 contained and failed storage analogues over geological time scales (St. Johns and Fizzy Field).

•Develop a set of methodological approaches for the determination of pressure, temperature, and stress history and fluid composition pertinent to long term CO2 storage characteristics of geological reservoirs (St. Johns, Fizzy).

•Provide large and small scale contemporary experimental data (Hontomin, Heletz, Maguelone Otway, Goldeneye, Sleipner, Snohvit, Weyburn, Miranga and In Salah).

•Evaluate possible leakage scenarios based on analogues.

•Analyze the behavior of the stored CO2 body in the upper layer of the Sleipner reservoir and attempt improving the quality of the history matching, which so far has not been very successful.

WP03: Mixing and trapping processes, reactivity

•Investigate key trapping mechanisms such as dissolution and residual trapping;
•Investigate the occurrence and magnitude of instabilities at the interface between the formation water the stored CO2;
•Investigate CO2 reactivity in carbonate reservoirs;
•Investigate the impact of the composition of the CO2 stream and of the formation fluid.

WP04: Leakage

•Assess long-term behavior of caprocks and at wells in the frame of CO2 injection.
•Produce validated operational (from lab and field experiments) tools and understanding for modelling thermo-hydro-mechanical and chemical (THMC) mechanisms occurring when scCO2 and CO2-enriched brine flow through fractured seal rock and cements and at the cement-casing and cement-caprock interfaces.
•Investigate leakage by means of laboratory analogues.
•Test and calibrate (from Hontomin high pressure test) large scale model of caprock deformation induced by fluid injection.

WP05:Spatial heterogeneity: Effective dynamics and uncertainty

•Upscaling the multiphase flow and transport dynamics, and quantification of heterogeneity-induced mixing dissolution dynamics.
•Upscaling mixing and chemical reaction rates for homogeneous and heterogeneous reactions.
•Quantifying the heterogeneity-induced uncertainty of the predicted large scale flow and transport behavior.
•Comparing various methods.

WP06: Far field impacts and impacts on fresh-water aquifers

•Estimate the evolution and impact of the large pressure plume caused by CO2 injection;
•Estimate the evolution and impact of far-field brine migration, and possible consequences to adjacent freshwater aquifers;
•Develop robust modeling approaches to analyze the above, and to analyze the effects of CO2 injection,
•possible CO2 leakage to the near-surface aquifers;
•Investigate the impact of geo-chemical changes subsequent to the presence of leaked CO2 on the mobilization of hazardous minerals, otherwise trapped in the solid matrix. role of

WP07: Prediction and validation of the long term behavior of the stored CO2

•Provide a quantitative assessment of temporal and spatial mass flows at a regional i.e. reservoir scale for the prediction of plume spreading and brine displacement);
•Provide the theoretical and process oriented basis for the simulation of small (core) and large (borehole, reservoir) scale fluid flow and mass transport;
•Evaluate and use novel computational models, suggest measures to increase efficiency of simulations of key processes;
•Investigate the effects of heterogeneity on storage efficiency and risk assessment in terms of coupled processes;
•Provide quantitative data for risk assessment;
•Provide spatial and temporal information for monitoring network design.

WP08: Cost-effective and reliable monitoring

•Evaluate the ability of the various monitoring technologies installed at Heletz and Hontomin;
•Design build, install and cement at Maguelone the new “RTSG” (Resistivity / Temperature/ Strain / Gas) down-hole monitoring instrument. Gas detection will be performed with direct measurement of CO2 from a new sensor based on Bragg’s grating technique attached to dedicated optical fiber devices;
•Perform CO2 injection experiments at Maguelone to test the new RTSG down-hole instrument installed at 20 m depth and within 10 to 20 m of the injection hole, in the direction of the pumping hole;
•Suggest MMV technologies suitable for implementation at the large scale CCS demonstration projects.

WP09: Dissemination, communication and public acceptance

•Establish a platform for ensuring efficient and reliable flow of information and knowledge within the consortium during the project lifetime;
•Develop a knowledge management system to be implemented at the initial stage of the project to facilitate and support the internal and external transfer of knowledge;
•Disseminate results, findings and deliverables within the consortium and to targeted communities of scientists, engineers, regulators and the wider, by putting in place a solid and relevant structure for communicating and disseminating the project achievements;
•Ensure adequate degree of communication and understanding at the level of the scientific community;
•Create links and relations with R&D related projects, Europe, the USA, Canada and Australia;
•Deliver clear, transparent and reliable information of the PANACEA findings in format to be adapted to different targeted audiences;
•Contribute to the effort of public outreach and acceptance of CCS.
Project Results:

In PANACEA, we carried out the work five parallel tracks, as illustrated in Figure 1.

Track 1 - Natural analogues: Most the documented CO2 injection sites are rather recent and the available data is not sufficient for validating the predictions on the long-term fate of the stored CO2. Analogues can contribute in providing this missing information. In PANACEA, we investigated a number of analogues sites, conducted laboratory analysis of samples from these sites and constructed a conceptual and computational of the Saint-John site (USA). We summarized the findings of this track in the deliverables of work-packages 2 and 7.

Track 2 – Fundamental processes: this is where most of the resources were invested in PANACEA. This included the investigation of processes occurring to CO2 during the injection and spreading phases (mixing with local brine, dissolution, convection, fingering, residual trapping, evaporation, CO2-water interfacial processes, impact of impurities in the CO2, reactivity), the potential environmental impacts resulting from the CO2 storage (pressure build-up, leakage induction, brine mobilization). We also investigated sensitivity to reservoir heterogeneity and uncertainty. The work included analysis of laboratory analogues, development of quasi-analytical and fast solutions for simplified geometries and reservoir scale simulation, including the use of simplified hydraulic models for pressure buildup prediction as well as the application of full models for history matching. Work in this track encompassed work-packages 2, 3, 4, 5, 6 and 7.

Track 3 - Leakage mitigation technologies: work consisted in developing and demonstrating at the laboratory scale chemical solutions capable of sealing possible fissures in the cement and or the caprock. Work here was concentrated in work-package 4 and key findings are summarized in the deliverables associated to this work-package.

Track 4 - Monitoring technologies: this included the development and shallow testing (at Maguelone, France) of a novel integrated monitoring platform, RSTG (Resistivity, Seismicity, Temperature and Geochemistry), which could become the next generation of monitoring technologies, capable of providing large amounts of data and information per well in non-intrusive setting (“behind the casing”). We reviewed the monitoring technologies installed at Hontomín and Heletz. However, due to delays in the beginning of the experimental sequences in both sites, no CO2 injection data was available now of the project closure. We conducted the work for this track in the frame of Work package 8 and key findings are summarized in deliverables of this work-package.

Track 5 – Communication and dissemination: activities included communication within the consortium, organization of two international brainstorming days in Trondheim (2013, targeted at the scientific community) and Paris (2014, target at the regulatory community). We exposed PANACEA to the scientific and regulatory communities, in form of large number of scientific publications, presentations at key international lCCS conferences and workshops. The work was carried out in the frame of work-package 9.

CO2 analogues

We collected samples from a number of analogue sites, including: 1) St Johns (depth and surface); 2) Bahrenborstel; 3) Wissey (Depth ≈1660m); 4) Wildcat; 5) East Brae Field: Shale (Depths of 3900m and 5200 m); 6) Boulby Mine; 7) Zechstein evaporite (Depth ≈ 1350m) and 8) Saint John. We developed an integrated methodological approach was developed for the assessment of the genesis and long-term performance of CO2 analogue reservoirs. This approach provides a unique set of instruments for the determination of key parameters to evaluate the complexity and dynamic of large geological reservoirs and to predict vulnerabilities and the reliability of CO2 storage sites over geological time-scales. We used field data and rock samples for analytical and experimental investigations, and numerical modelling to understand the fundamental mechanisms of CO2 of the long-term storage and interaction with sealing rocks. We identified the main factors leading to leakage from such reservoirs and the processes that lessen or increase the chance of leakage from a geological storage site during and after CO2 storage. This allowed an understanding of the “natural plumbing” and fluid flow in reservoirs and addressing the reduction of risk from a commercial point of view in order to increase the viability of CO2 storage. We geological parameters from a number of analogues in order to define the conditions of CO2 entrapment and mobilization in geological time spans. We then formed a comprehensive database of natural CO2 reservoirs. We selected three natural analogues - St. Johns, Fizzy Field, and Bahrenborstel - because of their high CO2 occurrence, but different CO2 mobility. Investigation methods on rock samples from those sites included fluid inclusion studies, micro-fracture analysis, cathodoluminescence studies, thin-section microscopy, Raman spectroscopy, and (isotope)-geochemical analysis. Investigated key parameters to elucidate the geological history and development of the different reservoirs included: Temperature development; stress distribution; Petrological and geochemical composition; Hydraulic parameters and fluid flow; CO2 reactivity in carbonate rocks. This allowed the preparation of a Database of natural CO2 reservoirs. With this data, we developed geological models for the three analogue sites (as well as conceptual and numerical models. The natural analogues St. Johns, Fizzy Field, and the Lower Saxony Basin (North German Basin) were studied with respect to their structural geometry (reservoir and caprocks) and characteristics (hydraulic and petro-graphical) to elucidate effective geo-tectonic processes that affected (partial) opening or closure of fault and fracture systems throughout their geological history.

3D Conceptual models of St Johns and Fizzy CO2 reservoirs

Saint John Field

St. Johns dome is located in New Mexico, USA, and provides an example for a leaky CO2 storage site, where CO2 discharge from naturally carbonated mineral springs has formed extensive travertine deposits at the surface. We constructed the model from wells logs drilled to exploit the CO2 reservoir. Gas migration and leakage to the surface appears to be controlled by a deep-reaching fault zone (Gilfillan et al., 2011). We analysed sample material from the basement (granite) and reservoir units (anhydritic rock) of the St. Johns reservoir. Cathodoluminescence microscopy was applied for mineral determination and to identify growth-zoning features. Dolomite precipitated along veins in altered granite. Fluid inclusion studies and micro-thermometry showed that older dolomite precipitated at ca. 75°C, while younger dolomite precipitated at ca. 145°C. This thermal anomaly can be explained by a hydrothermal event, probably caused by volcanic activity of the nearby Springerville Volcanic Field.

Fizzy Field
The Fizzy Field located in the UK sector of the Southern North Sea is part of a petroleum system in the Southern Permian Basin. CO2 containing hydrocarbon gasses, mainly generated in Carboniferous coal deposits, are stored in Permian Rotliegend and Triassic Bunter Sandstone reservoir rocks that are sealed by evaporites of the Permian Zechstein and mudstones of the Triassic Bunter Shale. This field, as an example for a tight reservoir, contains high CO2 concentrations (>50 mole %) that resulted from gas migration into extant structures. Most of the CO2 is present as free gas phase, while minor amounts of CO2 are stored in solid phases (dawsonite and dolomite). Aeolian sandstone from the Rotliegend group (provided by UK Geological Survey) allowed to determine the present and paleo-porosity, and the composition of fluid inclusions in dolomite and quartz by micro-thermometry, cathodoluminescence microscopy, and point counting microscopy. The cores of zoned euhedral dolomite grains contain primary low-saline H2O-NaCl fluid inclusions that were derived from marine solutions. In contrast, the rims of the dolomite crystals contain medium-saline H2O-NaCl-CaCl2 inclusions, which represent typical basement brines. Quartz cement contains primary aqueous inclusions with relatively low salinity (marine source, H2O-NaCl). CO2 was not detected in fluid inclusions, but Heinemann et al. (2013) reported that isotope ratios (13C/18O) from younger dolomite and from the free gaseous CO2 within the reservoir coincide. Homogenisation temperatures show that the youngest dolomite phase, which precipitated from CO2-oversaturated brine, occurred after basin inversion, when CO2 had entered the reservoir, coming from deeper units. The Fizzy Field model is based on published horizon maps of the Rotliegend reservoir that are based on a 3D seismic survey of the area (Ithaca & Tullow Oil relinquishment reports). The original 3D seismic data was not accessible. Well data was incorporated into the model and the thickness of the seal rocks and overburden is based on the interpretation of the Fizzy wells as well as surrounding wells. Cross sections based on seismic data from Underhill et al. (2009) and Yielding et al. (2011) were used to validate the model.

North German Basin (Bahrenborstel)

The North German Basin has abundant gas reservoirs, partly containing high levels of CO2. The study site located in the Lower Saxony Basin consists of a thick sequence of Permian through Tertiary rocks that were deposited on top of folded Variscan basement rocks. Permian dolomitic limestone form the reservoir and caprock for hydrocarbon gases. Tectonic processes and basin inversion led to re-opening of fault structures in the Late Cretaceous accompanied by fluid and gas mobilization along fractures. The studied site therefore represents a location where both scenarios, leakage by re-opening of fractures through tectonic processes and subsequent closure due to healing of fractures can be investigated. Core samples of Upper Permian dolomitic limestone (Zechstein group, Stassfurt carbonate sequence, Ca2) from the Bahrenborstel gas field were analysed to determine fluid evolution in the Lower Saxony Basin with respect to production and migration of CO2. Fluid inclusion studies, cathodoluminescence microscopy, and standard petrographic microscopy were applied to distinguish different mineral phases, microstructural features, and fluid generations in the sedimentary rock. Using microthermometry, paleo-pressure and –temperature conditions of the reservoir rock during maximum burial could be calculated. ICP-MS analysis was performed to reveal the REE distribution within fluorite of different ages. Laser-Raman spectroscopy was applied to determine the gas ratios within fluid inclusions. We compared our results with published data and developed a 1D model of the investigated area based on the simulation of different scenarios using PetroMod (Schlumberger). According to our results, CO2 was only present in traces in the reservoir rock during burial. CO2 concentrations of ca. 75 to 90% in fluid inclusions of post-diagenetic fluorite indicate a later CO2 influx into the reservoir, after maximum burial of the basin, most likely during basin inversion.

Geo-process facies analysis of experimental injection sites

Five CO2 storage sites (Otway, In Salah, Sleipner, Snohvit and Buracica) and three natural CO2 reservoirs (Miller Field, St. Johns Dome, Fizzy Field) were analysed using the geomechanical facies approach outlined by Edlmann et al. (2014) and partially developed within PANACEA. The factors that area crucial to assessing the CO2 storage security of a storage basin such as basin architecture, caprock architecture, reservoir quality, stress state, mechanical characteristics, fractures, burial depth, geothermal gradient, risk of orogenic modification, structural stability and preservation potential can all be taken into account. Using this first order assessment the storage potential of the studied sites was correctly predicted and only in areas where significant orogenic overprinting of the basins has taken place the method has limitations. This indicates that the geomechanical facies approach provides a highly suitable first order assessment of the global CO2 storage opportunities, it is not a tool for site specific suitability analysis but the first step in a complex and iterative site specific assessment procedure, where uncertainty decreases as the data requirement increases.

Of the three analogues conceptual models only one was translated into a computational one – the Saint John site. We constructed a full three-dimensional model, based on the conceptual model. We conducted simulations for 3,000 years until we reached a quasi-stable situation and obtained an estimate of the leaking CO2 fluxes though the fault (Figures 5 and 6).

CO2 injection sites

Hontomin and Heletz

We created for Hontomin and Heletz, pre-CO2 injection, comprehensive datasets from these two highly controlled and heavily monitored experiments. We integrated the collected data into the CO2 sites database templates created in the frame of PANACEA. We interpreted the collected data with emphasis on CO2 flow patterns and trapping processes, possible leakage paths, the impact of spatial heterogeneities, near and far field impacts, to develop and test different models, and to evaluate different monitoring technologies.

In Heletz we conducted simulations of CO2 injection (up to 20,000 tons in the small structure) and evaluated the mechanical impacts of the pressure buildup, using the DUMUX simulator. We estimated the large scale mechanical impacts in this small part of the Heletz Geological structure (see Figure 7).


We received from STATOIL data from Sleipner and Snohvit. Data on Miranga and Weyburn was not available to the consortium. We integrated the obtained data in the new database developed in the frame of PANACEA. This database allows recording all the administrative, geographical, groundwater, hydrological, geological and meteorological features. In addition further components specifically designed to store and manage data from CO2 injection site are included.

In Sleipner, we modelled the upper layer (L9) reservoir with the aim to improve the state of the art on the quality of the history matching of the stored CO2. We received the Sleipner layer 9 (L9) model from STATOIL in the Schlumberger Eclipse format. The model included high-resolution grid, static and dynamic properties and information on the injection rate back calculated from seismic data. Other relevant parameters where obtained from an SPE paper published by STATOIL. Previous modelling efforts failed to predict the evolution in time of the shape of the stored CO2 body. Nine different scenarios were simulated to test the result sensitivity to various common history-matching parameters. Reservoir heterogeneity was the first parameter tested and we concluded that heterogeneity has negligible effects on plume movement in L9. The reason is CO2 mobility in the high permeability reservoir where small variations in permeability and porosity do not have first order effect on CO2 spreading. Variations in injection rate and the use of linear relative permeability were also tested as history matching parameters and both were found to be of limited effects. Reducing reservoir hydrostatic pressure and subsequent correction for CO2 mass flow rate in L9 were found to have positive effects on plume movement, enhancing CO2 propagation towards the north. Considering the effects of residual components injected together with CO2 was the other parameter tested with considerable positive effect on matching CO2 plume at top Utsira (Figures 8 and 9) . The final sensitivity study that resulted in the best match was simulating the differential dissolution of CO2 and hydrocarbon gas injected together with CO2 in brine resulting in even higher concentration of methane at top Utsira. To the best of our knowledge, this study is one of the most comprehensive history-match trials performed on Utsira using L9 model determining the parameters that influence the CO2 movement in a shallow reservoir like Utsira.

In summary, from the analysis of analogues reservoirs, we concluded that CO2 leakage is controlled by:

CO2 state: 40% of gaseous reservoirs leak, only 6% of supercritical reservoirs leak.
CO2 density: All leaking reservoirs have low density <250 kg/m3 compared to non-leaking reservoirs (~550 kg/m3).
Caprock thickness: Leaking reservoir have thin (~160m) caprocks, sealing reservoirs have thick (~220m) caprock.
Faults: Faults are the preferred leakage pathways, 5 out of 6 leaking reservoir leak along faults.

The large scale modeling effort conducted at Sleipner showed that we were able to reproduce the shape of the stored CO2 body as measured by the seismic surveys. This shows that with an adequate understanding of the interactions between the geology, reservoir pressure and temperature conditions and the conditions of CO2 modelling can be reliable and produce a satisfactory history matching.
Dissolution and residual trapping

We used mathematical modelling and laboratory experiments to elucidate and understand mechanisms that act to trap CO2 in aquifers, such as structural trapping, whereby the topography, or shape, of the overlying caprock causes the CO2 to pool close to the injection site and limit its lateral spread under buoyancy. Another trapping mechanism is the dissolution of CO2 into the formation brine. Such dissolution results in a dissolved solution that is denser than both the pure CO2 and the brine, causing it to fall under its own weight and become permanently stored at the base of the reservoir. We have explored the interactions between the rate of trapping due to dissolution and fluid flow, focusing on how they depend on new dynamics introduced by the confinement by the lower boundary of the reservoir. For the structural trapping, we have developed a theoretical description of the evolution of a thin layer of CO2 injected into reservoir with a varying topography caprock. Analysis of the equations over a large class of idealized, three-dimensional topographies has revealed that different topographic structures lead to different dynamical controls on the spread of CO2. Certain structures, for example, cause the spread of CO2 to become constrained by the need to develop gradients in its hydrostatic (gravitational) pressure. In such cases, the lower surface of the CO2 maintains significant spatial structure to long times. The spread of CO2 along other kinds of structures, such as those of the dome—shaped caprock at the Sleipner site, were shown instead to become constrained by the need to flow downwards against buoyancy, developing instead a horizontal lower surface. The critical outline in parameters that distinguishes these different regimes was catalogued and the key time scale that describes the large-time transition was obtained. By applying our theoretical results to Sleipner, we found that the characteristic time scale on which the flow is influenced by variations in the topography is of the order of one hundred years, which is an order-of-magnitude larger than the current running time of the project. This suggests that topographic variation is unlikely to explain the large-scale deviations from axisymmetric flow that have been observed there. Instead, the results indicate these deviations are the result of other factors and this is fully compatible with the results of the high resolutions simulations that we have carried out.

Our analysis reveals how certain far-field properties of the geometry of the reservoir, such as the specific locations of distant faults, can have fundamental influence on the near-field flow of CO2. For example, it is found that the injected CO2 is naturally directed towards the nearest fault or, more generally, in the direction in which it is easiest for the pressurization of the flow to drive the ambient brine. Our results have shown that, in the presence of a fracture in a confined reservoir, two fundamentally different regimes of flow can emerge, depending on whether the steady-state hydrostatic head of the CO2 current below the fracture is shallower than or spans the depth of the aquifer. These theoretical predictions were tested by their comparison with data from an analogue laboratory experiment.

In summary, the topography of the caprock and the confinement of the flow between horizontal boundaries can dictate the important, leading-order flow of CO2. While such geometrical aspects have direct ramifications for the structural trapping of CO2, they also have new implications for other trapping mechanisms, such as that of dissolution, where the rate of dissolution can become limited by the need to drive the dissolved CO2 away.
Coupling between CO2 Dissolution and fingering evolution

Motivated by processes occurring during CO2 storage in a brine reservoir, we examine two-dimensional convection in a finite-depth homogeneous porous medium induced by a solute introduced at the upper boundary. Once dissolved, the solute concentration is assumed to decay via a first-order A→B chemical reaction, where species A (but not B) increases the solution density, which restricts the depth over which the solute can penetrate the domain. The substrate of the reaction is assumed to remain abundant, i.e. the concentration of minerals in the porous rock is sufficiently large for any changes to have negligible effect on reaction rates. We identified conditions for which convection will not take place. Under certain conditions (suffciently low Damkohler number), the solute penetrates to the base of the domain and solute gradients diminish to the extent that the reaction effectively becomes of order zero. It is notable that for both small and large Damkohler numbers, fully developed convection is characterised by vertical plumes emerging from, and sustained by, a thin dynamic boundary layer at the upper surface.

Nonlinear simulations reveal how at large times an active boundary layer forms near the upper surface and large plumes sink into the domain below (Figure 7). The large plumes transport solute away from the boundary; the solute concentration decays as it descends, but creates an upwelling of dilute solvent between the plumes. Small plumes also form in the boundary layer and grow before being swept horizontally across the boundary layer, driven by the recirculation of the dilute solvent towards a concentration hotspot. This acts as a negative-buoyancy source for larger plumes to grow deeper into the domain.
Water evaporation at the fluid interface

We modelled interfacial processes that can occur during the injection of supercritical CO2 into brine in a deep reservoir. The injection regime is defined by a low mobility ratio between the supercritical CO2 and brine, typically of the order 10 and a high capillary number creating a highly unstable interface between the two fluids. Interfacial mechanics dominate the flow characteristics between the immiscible displacements of brine with CO2. There is inherent instability at the interface, due to the sharp change in fluid properties such as density and viscosity, with the latter giving rise to a process known as viscous fingering. We focused on the lateral movement of the injection plume and the subsequent interface mechanics. A 2-D Hele-Shaw model was used that represents a planar view just underneath the cap rock of the aquifer. Temperature effects can cause significant changes to the interface topology of the fluids, initiating evaporation of the resident brine and Marangoni effects due to temperature gradients. Mass transfer can take place between the brine and CO2 as injection temperatures can be as high as 60-70°C, causing a significant rate of evaporation from the brine. This can alter the viscosity and density of the CO2 as well as the shape of the advancing front.

To study the interaction processes and the long term evolution of low mobility ratio flows with evaporation effects, a two-phase model has been formulated based on a classical boundary element approach. This numerical formulation allows the long term non-linear dynamics of growing viscous fingers to be explored accurately and efficiently. Fingering instabilities initiated at the interface between a low viscosity fluid and a high viscosity fluid are analysed at varying capillary numbers and mobility ratios, simulating flow conditions typically found in CO2 sequestration processes. A temperature distribution resulting from the injection of a heated inner fluid is considered, with resulting evaporation flux at the interface between the two fluids.

Simulations in low mobility ratio regimes reveal large differences in viscous fingering patterns and mechanisms compared to those predicted by previous single phase and high mobility ratio models. Most significantly, classical finger shielding between competing fingers is inhibited. Secondary and side branching fingers can posses significant momentum, allowing much greater interaction with primary fingers compared to high mobility ratio flows. Eventually, this interaction can lead to base thinning and the breaking of fingers into separate bubbles. The evaporation model reveals that with an imposed temperature field on the underlying viscous fingering pattern, mass transfer across the interface can cause fingers to coalesce. This leads to bubbles of the resident fluid pinching off into the advancing inner fluid, eventually evaporating away as they progress closer to the source. Evaporative mass transfer accelerates the growth of side branching fingers that experience higher temperatures than radially growing primary fingers. Reduction of the boundary layer between competing fingers eventually leads to their coalescence, forming a pinched bubble and a new single finger (Figure 10).
Sharp interface modeling of the interfacial processes

Interfacial instabilities can occur during the process of CO2 storage. We conducted a detailed study of the problem occurring in a Hele-Shaw cell, where the fluid flows between two thinly separated plates, giving indications of possible plume patterns that can develop during the CO2 injection. We consider both viscous and thermocapillary effects on the interactions between the fluids and the CO2 plume evolution. Viscous fingering occurs during the displacement of a high viscosity fluid by a low viscosity fluid, in which interfacial instabilities may arise and subsequently evolve to form complex interface topologies. Perturbations greater than a certain wavelength create instabilities along the fluid interface, promoting the growth of long fingers which penetrate into the more viscous fluid. Immiscible displacement is characterised by a sharp interface, across which the properties of the fluids (such as viscosity and density) vary discontinuously. We use a boundary element numerical scheme to compute the normal velocity at the interface of the two fluids (with finite mobility ratio) through the evaluation of a hypersingular integral. The boundary integral equation is solved using a Neumann convergent series with cubic B-Spline boundary discretisation, showing 6th order spatial convergence. The numerical scheme allows the long term non-linear dynamics of growing viscous fingers to be explored accurately and efficiently. The finite mobility ratio model allows investigation into the effects that a low mobility ratio and high capillary number have on the CO2 plume evolution. When the mobility ratio of the two fluids is of order 10 - 50, the fingering characteristics are vastly different to those predicted by high or infinite mobility ratio models (typically used for gas-oil injection scenarios). The near stagnation points on the bases of the fingers found in infinite mobility ratio flows were not found when using the finite mobility ratio model for low mobility ratio flows. Finger interaction was found to be much more significant, and on small wavelength perturbations could lead to base thinning and eventual finger breaking, shown in Figure 11. After breaking, the detached bubbles would continue with the velocity of the surrounding fluid (Figure 11). The numerical method allows the resolution of the immiscible lubrication layer between fingers meaning the finger breaking and coalescing mechanisms could be explored more explicitly than in previous models.

To study thermocapillary effects, we explore the dependence of interfacial tension on temperature. During CO2 injection, the supercritical CO2 is generally injected with relatively high temperature compared to that of the resident brine, initiating temperature gradients at the interface. The surface tension of the brine is a function of the temperature and hence can vary along the interface, affecting the interfacial growth and subsequent viscous fingering patterns compared to cases with a constant surface tension. Thermocapillary effects can be incorporated into the finite mobility ratio model by allowing the surface tension to vary along the interface. The surface tension gradient induces a local shear stress at the interface causing tangential movement, however in the case of CO2 injection where the injection temperature and resident brine temperatures vary by around 30°C, the shear stress will generally be very small and can be considered negligible. We instead focus on the effect of varying surface tension on the capillary pressure jump at the interface of the two fluids. Using an indirect boundary element method we can reconstruct the velocity of the fluid at every internal nodal point in the domain, allowing the solution of a convection-diffusion equation for the heat transfer between the inner and outer fluid. By assuming that the thermal diffusivity of both fluids is equal, we solve a single zone heat transfer problem, and interpolate the transient temperature field at the interface between the fluids, allowing the evaluation of the surface tension. We find that the transient temperature field induces localized surface tension gradients at the fluid interface. With a varying surface tension, and a capillary pressure jump at the interface that takes into account the meniscus in the Hele-Shaw cell, we see a significant change in the interfacial position compared to the cases with a constant surface tension (Figure 12). The pressure jump caused by the meniscus is non-constant and becomes the leading order term in the normal stress balance at the interface. This magnifies the surface tension gradient, causing areas of the interface which lie closer to the injection source (and hence at higher temperature) to be pulled towards the interface that lies further away from the source. We find the finger bases are pulled towards the finger fronts, and pinched inwards by the localized surface tension gradient, inhibiting the formation of large Saffman-Taylor fingers (red line in Figure 11). Without the perfectly wetting Hele-Shaw meniscus, we find that the varying surface tension has little effect on the interfacial movement (dashed line in Figure 12). The normal stress balance is much more sensitive to variations in the interface curvature than the small changes in surface tension due to the temperature field. To further explore this problem, the full Marangoni effect needs to be considered whereby the tangential shear stress needs to be balanced at the interface, requiring a Brinkman flow formulation including 2nd order velocity terms. The finite mobility ratio model developed in this work highlights the difference in injection regime between previously considered high mobility flows such as gas-oil displacement and that of supercritical CO2 – brine displacement. The lower mobility ratio creates a much larger swept volume of brine, and a much less convoluted front than a corresponding high mobility ratio injection. Thermo-capillarity induces localized surface tension gradients at the interface, which can help to dampen the formation of long viscous fingers, if the variation in surface tension is significantly large.
CO2 reactivity in carbonate reservoirs

We have investigated the investigation of the evolution of relevant transport parameters of sedimentary reservoir rocks upon reaction with CO2 during laboratory experiments. Different reservoir rock types (limestone, dolostone). We developed conceptual models to describe the observed phenomena and incorporated them into numerical simulation codes. We performed continuous-flow laboratory experiments, using vertical scanning interferometry (VSI), dissolution kinetics and the effect of passivation for carbonate rocks. The monitoring of the evolution of physical properties (e.g. porosity, permeability) and mineralogical composition as CO2-rich brine flows through reservoir rock samples (lab column experiments), the investigation of calcite and dolomite dissolution rates and the impact of passivation (precipitation of secondary minerals, e.g. gypsum or anhydrite) at different partial pressure of CO2 and temperatures (lab column experiments). Main findings are presented below:
Calcite dissolution rates as a function of pH
To obtain calcite dissolution rates as a function of pH to derive an expression of the calcite dissolution rate law. Flow-through experiments with calcite fragments (Iceland spar) in acid solutions (pH = 1-7) at atmospheric pCO2 were run. The exposed calcite surface was examined by VSI to study the pH effect on calcite reactivity experiments. The results from the flow-through-VSI experiments were used to derive an expression of the calcite dissolution rate law to be implemented in the reactive transport codes.

Column experiments with limestone and dolostone rocks

To evaluate the interaction of sedimentary reservoir rocks upon reaction with CO2 in flow-through columns at the laboratory scale. The objective was accomplished by performing flow-through column experiments under subcritical CO2 conditions (PTotal = 10 bar, pCO2 = 10 bar, T = 25, 40 and 60 °C) using crushed limestone and dolostone samples. The role of injected solution composition (gypsum undersaturated and gypsum equilibrated solutions) on changes in the mineralogy were investigated. In this range of P-pCO2 and T only gypsum precipitation took place and it only occurred if the injected solution was equilibrated with respect to gypsum. Comparing the two reservoir rock reactivities, limestone dissolution induced late gypsum precipitation (long induction time), in contrast to dolostone dissolution that promoted fast gypsum precipitation. A decrease in T favored limestone dissolution regardless of pCO2. However, gypsum precipitation was favored at high T and atmospheric pCO2 conditions and un-favored at high T and 10 bar of pCO2.

Flow-through limestone rock sample dissolution (role of pH and heterogeneity)

To evaluate the role of the injected pH solution into a natural limestone core sample. See D3.2 for more details. We observed that, as expected, the acid attack showed clear effects on all investigated hydraulic transport parameters. Dissolution at lower pH, while maintaining the same flow rate, leaded to faster increases in permeability and wormhole formation. The effective diffusion coefficients increased much more than the increase in porosity, suggesting increase in geometric factor. In the water retention curve, the increase in porosity affected different ranges of pore sizes depending on the effective diffusion coefficient of each sample increases by factor 3 for the sample L1 y L2 with high pH (pH = 5) and by factor 12 for sample L3 with low pH solution (pH = 3.5). The diffusion coefficient increases much more than the increase in porosity. Conversely, the proportion of small pores increases after each experiment. This indicates that some small particles move within the sample, maybe from the sample inlet to the outlet. This particle movement clearly induces the temporary permeability decrease. This means that the permeability is directly linked to the hydraulic pore diameter, and not only to the tortuosity, which decrease here. These experiments show that far from the injection well site, dissolution processes are characterized by slow mass transfer, which may include, in the case of carbonate rock, transport of fine particles, which locally clog the porous space. Then, this leads to the damage of the carbonate reservoir both in terms of connectivity of the porous space and CO2 injectivity.
The impact of the composition of the CO2 stream and of the formation fluid

We conducted a review on the impact of chemical impurities on the CO2 stream over carbon storage processes. We identified the relevant impurities that we could find at the CO2 streams and how they can change the CO2 stream properties. We also focused on the effect of these impurities on the storage processes and the trapping strategies. We observed that impurities produce modifications in the behavior of the CO2 stream during storage. From a thermodynamic point of view, SO2 and N2 provide the highest impact to the stream, with opposite effects. SO2 increases density and viscosity while, N2 produces a minor reduction of density and of viscosity, enhancing buoyancy in the injected CO2 stream. The presence of others impurities in their usual mass fractions have a much smaller effect, except for the H2S, which provides low pH to stream and it is highly reactive near the wellbore and in the caprock (as the SO2). Impact on trapping strategies are different depending on the properties variations caused by impurities. In case of SO2, dissolution and structural trapping are enhanced near the well. Moreover, in long-term, porosity decrease and injectivity is worse than in case of pure CO2 stream. On the other hand, in case of the N2 where effect is greater than others, the efficiency of CO2 dissolution and trapping in the pores as a residual phase is lower, due to less density fluid. Also, it reduces the lateral spreading of the plume, hence it reduces the formation of a residual phase.

Characterization and modelling of the effect of scCO2 and CO2-enriched brine flow through fractured seal rocks (from marl to shale)

We investigated the long-term control of the thermal, hydraulic and mechanical conditions on fracture behavior. We conducted experimental investigations into supercritical CO2 flow behavior within fractured seal rock. This, under thermal, hydraulic and mechanical conditions, typical of analogue CO2 storage sites (Fizzy Field of the Southern North Sea, and St John’s Dome, USA), which allowed identifying and understanding the key mechanical and hydraulic mechanisms that influence the flow of scCO2 through fractures under in-situ conditions. Cyclic water / scCO2 flow experiments were undertaken on fractured seal rock and reservoir rock samples, to assess the hydraulic response of the fractures and pore-network respectively to cyclical multi-phase flow. Simulation of the experimental results was carried out through development of a hybrid analytical numerical finite element model capable of simulating elastic and plastic deformation, and non-linear flow. Additionally, fracture profiling data was used to develop statistical models of the fracture surface, and fit aperture distributions. The main finding from the scCO2 fracture flow experiments are (See Figures 13 , 14, 15):

Permeability of the fractured samples to single phase supercritical CO2 decreases under increasing confining pressure, and increases under increasing fluid pressure in line with expected fracture mechanics behavior (Bandis et al., 1983).
The initial stress loading cycle results in significant inelastic fracture aperture reduction, with increasingly smaller magnitude effects for subsequent cycles. Hysteresis is significant during confining pressure stress loading, and is present during fluid pressure cycling.
Lower fracture permeability and hydraulic aperture were observed at higher flow rates under constant stress conditions, suggesting non-linear flow effects occur during fracture flow of scCO2. The occurrence of non-linear flow is likely to be a result of an increase in inertial effects due to high tortuosity.
Permeability of the fracture to scCO2 is observed to be slightly lower at high temperatures (60°C, increased from 40°C), but hydraulic conductivity is greater due to the lower viscosity of supercritical CO2 at the higher temperature. This indicates that the change in viscosity is more influential than mechanical changes resulting from the temperature increase during supercritical CO2 flow.
In addition to experimental studies, simulation of the experimental results was carried out through development of a hybrid analytical numerical finite element model capable of simulating elastic and plastic deformation, and non-linear flow. Additionally fracture profiling data was used to develop statistical models of the fracture surface, and fit aperture distributions.
The hybrid analytical numerical model was able to match the experimental results with typical parameters extremely well despite widely varying conditions of flow, fluid pressure and stress. The use of simplified effective stress models for the prediction of fluid flow through rocks can be shown to be inadequate for the in situ fluid and rock stresses under investigation. Stress corrosion was considered to be the main process contributing to the inelastic deformation of the fracture.
The experimental work and the numerical investigation has led to a significant increase in process understanding with respect to the flow of high pressure fluids through fractures in low permeability media at depth.
Residual trapping is a key mechanism by which CO2 can be stored in an immobile state in a geological formation. The aim of this work was to investigate the changes to multiphase flow characteristics of supercritical CO₂ and water during cyclic flow. Unique high pressure (10-20MPa) and temperature (40°C) equipment was used to experimentally investigate fluid behavior and interactions during six alternate scCO₂ and water cycles through two rock samples; a permeable sandstone and a fractured shale caprock.
During cyclic phases of scCO2 and then brine/water injection there is a clear trend of increasing differential pressure (decreasing permeability) over the course of six water / scCO2 injection cycles. For a permeable sandstone, the average differential pressure for the water flow doubled from 5.6psi in cycle 1 to 11.3psi in cycle 6. The fractured caprock sample showed similar results with the average differential pressure almost doubling from 217.6psi to 408.1psi. This implies that the amount of residual trapping is enhanced with cyclic flow therefore increasing the storage integrity of the aquifer over time.
The cyclic flow results indicate that injection rates of scCO2 may decline over time if there are gaps within the injection that allow water to imbibe back into the injection area.
The water relative permeability curve for both samples shows a gradual decrease in permeability with decreasing water saturation as more CO2 flows through the system. This is consistent with the flow experiment results indicating that residual trapping plays an important part in controlling multiphase fluid dynamics during cyclic flow.

We worked on the determination of the thermo-hydro-chemical alteration mechanisms of Heletz caprock occurring when CO2-enriched brine flows through fractures. This included: 1) setting the experimental equipment for reproducing leakage condition through the caprock; 2) conducting a set of experiments of flow of CO2-rich brine through fractured samples of Heletz caprock under thermal, hydraulic and chemical conditions corresponding to Heletz site conditions; 4) Identify and understand the key mechanisms that control the alteration of the rocks..

Two different caprock samples were investigated. One mainly composed of carbonate cement and taken from cores of well Heletz H18. The second one is mainly composed of quartz and clay minerals and come from the ESGNJ block. CO2-rich brine flow through experiments were realized through both fractured caprock samples and through the fracture interface of the second caprock sample and the corresponding sandstone sample from ESGNJ block.
Permeability decrease was recorded for overall experiment as the most dominant hydrodynamic feature.
During the experiments conducted on the caprock-sandstone block interface (ESGNJ), permeability decrease is recorded during experiments. Nevertheless, very few changes are recorded on the chemical processes (low outlet cation concentration). The permeability decrease can be related here to particle dragging through the fracture and clogging process.
The main chemical process observed during CO2-rich fluid injection through the Heletz caprock sample coming from H18 well is carbonate dissolution (calcite and siderite) which locally increases the fracture aperture.
Laboratory analogues and fast quasi-analytical solutions

Work in this task focused on the investigation of the possible rates of leakage in confined aquifers from localized points in the near vicinity of the injection point with the ultimate goal of comparing the possible leakage rates with observed spreading rates in the Sleipner (or In Salah) CO2 injection sites. We developed the theoretical a framework for leakage between confined aquifers, including the buoyancy and viscosity ratios of CO2 and brine, developing numerical routines for estimating leakage rates between confined aquifers and finally, we designed and built and testing of analogue experiments to benchmark leakage estimates.

The results are (see Figures 16 and 17):

The flow of CO2 injected into a confined reservoir is driven by the imposed pressure gradient and the buoyancy (or density difference) between fluids.
Leakage from confined aquifers may be driven by the buoyancy of CO2 in the leakage point (eg. a fracture) or from the pressure gradient induced by injection across the leakage point (see figure below).
In confined aquifers buoyancy driven leakage is limited by the finite extent of the aquifer. This produces a fundamental difference with previous models which predicted that all injected CO2 may ultimately leak.
The ratio of hydrostatically (ie. Buoyancy) driven leakage to injection pressure leakage may be characterized by a single parameter that we call D.
The effect of leakage on the long-term viability of storage may be assessed by the efficiency of storage, E, which is the ratio of the volume of CO2 still within the aquifer to the amount injected.
For D > 1 the long-term leakage rates are limited to a finite value and the efficiency of storage reaches a constant, while for D < 1 leakage rates increase so that the efficiency of storage E ~ t-1/2, and hence ultimately decay to zero (summarized in the figure below).
Injecting in thin, or highly confined, aquifers therefore limits long-term leakage.
In addition, the initial description of the task included an assessment of vertical migration rates at the Sleipner and In Salah sequestration sites. Data from the In Salah site was not forthcoming during the course of the project. High resolution seismic data from the Sleipner site was processed, showing the planform of spreading CO2. However, a detailed map of the volume of CO2 within each horizon at Sleipner is needed to quantitatively compare with the results of the leakage calculations. Robust estimates of the volume of stored CO2, particularly in the top layer, is currently being compiled for future comparison with the leakage models developed in this work package.

Quantification of leakage from faults in real scale reservoirs

Brine displacement is an inevitable byproduct of CO2 injection into brine reservoirs, at midscale space and time ranges, since the volume of the injected CO2 spreads, dissolves, mixes with the brine and replaces it in the formation. This term is normally used in connection with fluid flow along the layers of the formation, which (in the context of CO2 injection sites) are by and large horizontal. Leakage might occur at all spatial/temporal scales, and the term is used in general in connection with vertical flow, mainly of CO2. However, brine in the front and CO2 behind is the result of any fluid movement caused by CO2 injection, in all directions, so these phenomena cannot be split (Cihan, Birkholzer, & Zhou, 2013). We consider injection into a generic medium scale 3D open boundary horizontal domain, which includes an extending 1300m vertical fault located 50m away from the injection point, during several years. The focus is on brine moving outwards through any open boundary, followed naturally by brine-CO2 mixture. The purpose of this setting is the assessment of a type of an upper bound for leakage, since in realistic domains faults split sideways and therefore are not vertical all the way up to the upper reservoir, their permeability is variable and may contain sections with caprock permeability, and they are partially open to additional permeable horizontal layers. The questions of brine migration and CO2 leakage in such a scenario are inseparable, and enable addressing the question of the transient composition of fluid reaching the upper aquifer through the fault’s exit, as well as the ‘competition’ or division the between fluid exiting from the horizontal boundaries vs. the vertical fault. The simulator used is TOUGH2-ECO2M (Pruess, Oldenburg, & Moridis, 1999) (“ECO2M: A TOUGH2 Fluid Property Module for Mixures,” n.d.) one of a few simulators capable of calculating all phase combinations of co2 and brine, relevant for the changing phase of co2 going upwards in the fault. Therefore, considerable effort has been dedicated to verify the reliability of this study via grid independence, checking effect of horizontal boundary condition and of upper fault boundary condition.

The most important parameter predicting fault fluxes is the fault permeability. In some permeability ranges, the variation of formation permeability has no effect at all and the permeability of the fault is the only factor determining the fluxes from the fault – i.e. the vertical brine migration and CO2 leakage. The formation permeability has a very small effect on the brine (horizontal and vertical) and CO2 outward fluxes. The fault width has an effect, which supersedes the relative area factor. This observation could be possibly attributed to the wide fault south-west corner being closer to the injection point. This induces variability within the cross section of the fault. The flux (per unit area) is 1.5 larger through the wider fault.
Main results: 0.066% of the total injected CO2 mass leaks (as brine mixed with CO2) after 5 years through a fault whose area is 1m2, whose permeability is 3.e-13m2 (formation perm. 5. e-13m2). The composition of the leaking fluid is 2-phase brine and gaseous CO2. Therefore, 0.05% of total injected CO2 leaks through the fault. In the corresponding case, when both formation and fault permeability are 8.e-13m2 0.26% of the total injected CO2 mass leaks (as brine mixed with CO2) after 3 years. Therefore, 0.14% of total injected CO2 leaks through the fault. The predictions presented here compared with estimates of flux (g/m2/day) from natural analogues from the literature are much higher (5-20 times for the 5.e-13 formation/1.e-13 fault case). Apparently, natural analogues, which leak with such a rate, would have been empty in a short time period, so those, which continue to leak for a long time, are characterized with small permeability values and large aquifers. This comparison adds strength to the claim that the scenario simulated here can be viewed as an upper bound for leakage.
In conclusion, the quantitative evaluation of leakage via faults of substantial permeability show that the total leakage amounts to much less than 1% of the injected CO2 (Figures 18 and 19).

Characterization and modelling well integrity

We investigated thermo-hydro-chemical alteration mechanisms of cement caprock occurring when CO2-enriched brine flows through fractures. We set the experimental equipment for reproducing leakage condition, perform a set of experiments of flow of CO2-rich brine through fractured samples of standard Portland cement and Class G cement under hydraulic and chemical conditions corresponding to Heletz site conditions and Identify and understand the key mechanisms that control the alteration of the cement. The key results are (Figures 20 and 21):

Standard Portland cement and Class G Portland cement display similar behavior.
At the sample scale, the cement alteration by CO2 enriched brine is controlled by the renewal of the reactants and product into the fracture and by diffusion through the cement matrix.
When diffusion in the fracture is dominant, calcite fills the fracture. In this case, one fund similar behavior to that obtained using batch experiments.
In advection-dominated cases, we observe the development of a Si-Al amorphous layer near the fracture. For long exposure times, the silica rich layer becomes more and more thick and can reduce the fracture aperture. This supports the conjecture that leakage paths for CO2-saturated brine in fracture may be self-limiting.
We observe the accumulation of cations Cr, Fe rich oxides at the fracture-cement interface which would indicate that the silica layer may promote the retention of alkali and transition metal species as well as metalloids and acts as mitigating pollutant leakage from the storage reservoir.

Large scale coupled model for the caprocks

We addressed here the evaluation of the coupled thermo-hydro-mechanical (THM) processes at large scale:

Assess the coupled THM mechanisms to caprock deformation including fault reactivation;
Develop of a THM model for Heletz site, using the available site data (Figure 22);
Adapt the TOUGH-FLAC code developed to simulate THM processes to the Heletz site in collaboration with the developer, Dr. Jonny Rutqvist from the Lawrence Berkeley National Laboratory (LBNL), California.

The main achievements include:

The thermo-hydro-mechanical-chemical process was reviewed and decided to focus on hydro-mechanical and thermo-hydro-mechanical effects. Chemical effects were considered in sense of tracking CO2 and brine migration but no chemical reactions were considered. Thermal effects were included as part of the non-isothermal conditions affecting hydro-mechanical effects. A hydro-mechanical model was developed within the framework of TOUGH-FLAC (Rutqvist et al., 2002; Rutqvist, 2011) to study the interaction between mechanical deformation and fluid flow in two existing faults during CO2 injection and storage (Figure 1). The work focused on the integrity of the CO2 storage, induced by CO2 injection in a three-layer storage formation at the Heletz site. The consequences caused by potential fault reactivation, namely, the fault slip and the CO2 leakage through the caprock, were evaluated. The difference in the results obtained by considering the three-layer system as an equivalent single-layer storage formation is analyzed. Five main sources of data uncertainty were identified: the fault dip angle , the ratio SR between the horizontal and vertical stress components, the vertical offset d of the layers across the fault F2, the initial permeability k0 of the faults and the permeability of the confinement formations. In a base case study, these key parameters were set to reasonable values. Then, we conducted a sensitivity analysis to study the influence of those parameters on the obtained result.

We found that after 5 years of CO2 injection, the faults are not reactivated by an increase in pore pressure, but a small leakage through the caprock is observed at one of the faults, which can be explained by a buoyant force that causes CO2 to migrate upwards through the permeable fault. The analysis of the difference in the results obtained by considering the three-layer storage as an equivalent single-layer storage formation showed that the discrepancies in the pore pressure build-up close to the faults are small, but the discrepancies in the CO2 spread are significant. We found that the results are not very sensitive to the fault dip angle, because no fault reactivation is observed at alternative values of dip angles. However, the faults are reactivated when (1) the ratio between the horizontal and vertical stresses is equal or smaller than 0.6 (2) the permeability of the confinement formations is smaller than 10-18 m2 or (3) the initial permeability of the faults is 10-18 m2. However, the impact in the Heletz case is not significant. In the first case, the pore pressure build-up close to the faults is not enough to induce significant shear plastic strains and displacement along the faults. Results obtained in cases (2) and (3) showed that even when the total pore pressure at the storage layers is two to three times the initial pore pressure, the maximum fault slip displacement is 1.5 cm and the maximum fault permeability is increased by only two times its initial value. This is because the plastic shear strains occur mainly in a fault section that is only about 10 m in length, corresponding to the thickness of the storage formation. No CO2 leakage through the caprock was observed, because, in these two cases, the lateral and vertical extensions of the CO2 plume are more limited. On the other hand, our study shows that the offset of the layers across the fault F2 is an important parameter, because it limits the lateral extension of the CO2 plume and the CO2 leakage through the caprock increases. As a concluding remark, the present study indicates three parameters of significant interest for the fault zone hydromechanics associated with CO2 injection and storage. These are offset of storage layer across faults, permeability of confinement layers, and thickness of storage formation, that has direct relation with the extent of any potential fault reactivation. These are parameters, which have not received sufficient attention but should be included among the key parameters to be evaluated in site characterization for CO2 storage.

Spatial heterogeneity: Effective dynamics and uncertainty

Multiphase Flow Upscaling

The purpose of the work was to develop models that allow taking into consideration the impact of heterogeneity that characterizes geological formations. Strong spatial heterogeneity occurs for instance, in fractured reservoirs. We can conceptualize by means of two flow domains, a mobile fracture and a virtually immobile matrix (Figure 23). The fracture and matrix domains are coupled by continuity of mass flux and continuity of capillary pressure at their interface. We rigorously derive an upscaled multi-continuum model (an equivalent model) using a two-scale expansion in the framework of homogenization theory. The resulting multiphase multicontinuum model is capable of predicting multiphase flow in geologic media characterized by strong heterogeneity that leads to a clear distinction between mobile and immobile domains. The upscaled flow model consists of a flow equation for the saturation of displacing fluid in the fracture domain, and a capillary flow equation for saturation in the matrix. Mobile and immobile regions are coupled via capillary-driven mass flux, which is reflected in the upscaled multicontinuum model by a coupling term in the equation for the mobile saturation. By linearizing capillary counter current flow in the matrix domain, we combine this system of equations into a non-local single-equation model for the fracture saturation, which can be interpreted as a multi-rate mass-transfer (MRMT) model for immiscible displacement. We discuss this simplification and the parameterization of the upscaled model equation from local hydraulic parameters obtained from rock samples and from knowledge of the average flow properties of the fracture network. We demonstrate the performance of the model for predicting two-phase flow by considering a single fracture with imbibition into a rectangular matrix domain. The upscaled model is parameterized directly from geometry and hydraulic parameters of matrix and fracture of the reference model, which means that no parameters need to be fitted.

We compare detailed numerical simulations (DUMUX) of the full two-phase flow model and upscaled multiphase MRMT model in terms of breakthrough curves for the displaced fluid at a control plane within the medium. Both the detailed numerical simulations and the upscaled model, shown in Figure 24, show a preasymptotic square-root of time scaling and a breakoff at the characteristic time scale for filling the matrix by counter current flow.

Upscaling of Flow, Transport and Reaction Processes

Heterogeneity affects the efficiency of dissolution and precipitation reactions and their interaction with the anomalous dispersion dynamics at multiple scales, from pore to field scale. In order to gain deeper understanding of these processes and mechanisms, we studied the dynamics of flow, transport and reaction processes and their upscaling in terms of Lagrangian particle statistics using novel approaches based on correlated continuous time random walks and the quantification of the effective action of the deformation properties of the flow field on the mixing and reaction processes. Here we summarize the work done on the (i) characterization of pore scale flow and its significance for the understanding and quantification of anomalous dispersion and non-Fickian transport features, (ii) the analysis of pore scale reactions and the quantification of the impact of pore scale flow heterogeneity, (iii) the derivation of a mixing model for Darcy scale solute transport. Pore scale flow and transport were simulated using smoothed particle hydrodynamics (SPH), the Darcy scale flow and transport simulations were performed using the modeling platform H2OLAB.

Flow intermittency and anomalous pore scale dispersion

The heterogeneity of natural flows strongly affects transport, mixing, and chemical reactions, including spreading of dissolved CO2 and its effective reaction kinetics. In heterogeneous porous media, we may observe non-Gaussian velocity distributions, which can lead to a persistent non-Fickian dispersion regime that cannot be quantified by standard transport models based on the Fickian paradigm. Various stochastic models have been proposed to represent this property, with very different underlying mechanisms, such as mobile-immobile mass exchange, long-range correlated spatial motions, or heavy-tailed trapping time distributions. These different models may provide equally good fits to data, such as first passage time distributions. Yet, their implications can be dramatic when transport controlled chemical reaction processes are considered. A key challenge is to relate these upscaled flow models to the microscale flow properties. We demonstrated the existence of persistent intermittent properties of Lagrangian velocities in porous media, and we formulated a new dynamical picture of intermittency based on the remarkable spatial Markov property of Lagrangian velocities. The resulting upscaled transport model is a correlated continuous time random walk (CTRW). It provides a new dynamical picture of intermittency and an upscaled transport model, which is fully consistent with the micro-scale flow dynamics.

Anomalous kinetics of reaction fronts in porous media

We studied the irreversible bi-molecular reaction A+BC in complex pore-scale flow. Complexity is induced by the heterogeneous pore structure. We consider the evolution of the reaction for the scenario of displacement of a solution containing B by a solution containing A in a 2-dimensional medium, see Figure 25. For advection and diffusion of a material line in uniform flow, the reaction rate decreases with the square root time, or in other words, the production of C increases with the square root of time. In general, production of C scales with the volume of the mixing zone between A and B. Thus, the reaction rate is directly related to the mixing induced by the underlying flow field. We observe two distinct regimes for the evolution of the reaction. The early time regime is dominated by stretching and compression of the interface between A and B, and by the formation of fingers, see Figure 25.

In this regime, the interface length increases linearly with time, thus the stretching rate (elongation rate per elongation) evolves inversely with time. From mass conservation, we directly obtain the compression rate as the negative stretching. From this, we obtain the average thickness of the interface by considering the competition of interface compression and diffusion perpendicular to it. This simple model explains the observed strong increase of mass production in this regime, which evolves approximately with the square of time. In addition, it is a manifestation of the increase in mixing due to the "stirring" action of the complex flow field. In the coalescence regime, at times larger than the diffusion time over a typical pore throat dimension, mass production is still faster than expected for a homogeneous flow field, but slower than in the stretching-enhance mixing regime. We find here that the mixing volume between A and B can be quantified by the spreading of purely adjectively transported fluid particles. The volume of the mixing zone is given by the interface length times its width. In this regime, the interface is formed of bundles of aggregated lamellae that merge when they overlap. The individual bundle length can be approximated by the longitudinal spreading of fluid particles and thus the interface length is obtained as the number of bundles times the spreading length. The number of bundles, however, is reduced as they merge and therefore is inversely proportional to the average interface with. This reasoning gives directly the observed behavior. The reported work sheds new light on the complex reaction dynamics in heterogeneous porous media flows and allows to quantify the reaction efficiency based on the mixing dynamics of the underlying flow field.

Mechanism and regimes of mixing in porous media

We studied mixing of a dissolved substance in heterogeneous hydraulic conductivity fields, whose structural disorder varies from weak to strong. Enhanced mixing is induced by the stretching and folding action of the flow field (stirring protocol) and the action of molecular diffusion. A range of stretching regimes is observed which depends on the level of structural heterogeneity, quantified by the variance of the log-conductivity field. We develop a unified framework to quantify the overall concentration distribution to predict its shape and rate of deformation as it progresses toward uniformity in the medium. The scalar mixture is represented by a set of stretched lamellae whose rate of diffusive smoothing is locally enhanced by kinematic stretching. Overlap between the lamellae is enforced by confinement of the scalar line support within the dispersion area. Based on these elementary processes, we derive analytical expressions for the concentration distribution across lamellae, and the probability density function (PDF) of concentration values that result from the interplay between stretching, diffusion, and random overlaps. This theory holds for all field heterogeneities, residence times, and Péclet numbers. This work sheds new light on the mixing mechanisms in Darcy scale heterogeneous media and provides a large step forward towards quantifying mixing, reaction and dissolution processes in porous media.

Convective Mixing and Mixing-Induced Dissolution Reactions Porous Media

We studied the fundamental problem of mixing and chemical reactions under a Rayleigh-Bénard-type hydrodynamic instability in a two miscible fluids system. This scenario is characteristic for and describes the processes at the interface between supercritical carbon dioxide and brine during deep geological storage. The dense fluid mixture, which is generated at the CO2-brine interface, leads to the onset of a convective instability. At the same time, a fast chemical dissolution reaction produces a characteristic porosity pattern that follows the regions of maximum mixing. Contrary to intuition, the dissolution pattern does not map out the finger geometry of the unstable flow. Instead, it displays a dome-like, hierarchical structure that reflects the positions of the ascending fluid interface. We find that this behavior is caused by stagnation points along the deformed interface, which act as mixing and reaction hotspots due to a strong compression of the interfacial boundary layer. We develop a model for mixing and reaction around the stagnation points of the deformed fluids interface that captures the evolution of the global scalar dissipation and reaction rates and predicts their independence of the Rayleigh number. Figure 5.3a illustrates the mixing ratio of dissolve CO2 in brine, the reaction rate and log-reaction rate as well as the porosity pattern generated due to dissolution of calcite. Figure 5.3b shows the mixing rates for two different Rayleigh numbers obtained from direct numerical simulations and the prediction by the developed mixing and dissolution model.
Uncertainty Quantification

Dispersion variance in heterogeneous porous media

We study the uncertainty of dispersion of a dissolved substance in two and three-dimensional Darcy scale heterogeneous media motivated by the question of how well dispersion coefficients that are defined as ensemble averages, describe solute spreading in a single medium realization. In other words, we consider the self-averaging properties of dispersion coefficients in heterogeneous media. This addresses a fundamental question of the stochastic approach and provides a measure for uncertainty in estimated plume extensions, which is of central importance for risk assessment studies and for the interpretation of dispersion data from field and laboratory experiments. A related issue discussed in this work regards the definition and meaning of dispersion measures in single realizations and the related fluctuation behavior. In summary, we study dispersion in heterogeneous porous media for solutes evolving from point-like and extended source distributions in two and three dimensions. The impact of heterogeneity on the dispersion behavior is captured by a stochastic modeling approach that represents the spatially fluctuating flow velocity as a spatial random field. We focus here on the sample to sample fluctuations of the dispersion coefficients about their ensemble mean. For finite source sizes, the definition of dispersion coefficients in single realizations is not unique. We consider dispersion measures that describe the extension of the solute distribution, as well as dispersion coefficients that quantify the solute spreading relative to injection points of the partial plumes that constitute the solute distribution. While the ensemble averages of these dispersion quantities may be identical, their fluctuation behavior is found to be different. Using a perturbation approach in the fluctuations of the random flow field, we derive explicit expressions for the temporal evolution of the variances of the dispersion coefficients between realizations. Their evolution is governed by the typical dispersion time over the characteristic heterogeneity scale and the dimensions of the source distribution.

We find that the dispersion variance decreases towards zero with time in three dimensions, while in two dimensions it converges towards a finite long time value that is independent of the source dimensions. The concept of effective dispersion serves as an important estimate for the quantification of the extension of the dissolved CO2 within the heterogeneous medium and as a first estimate for the mixing and reaction potential. This study sheds light on its predictive power and provides a measure for the uncertainty attached to the use of these measures in the modeling of the fate of CO2 in heterogeneous formations.

Uncertainty Quantification

The Monte-Carlo method is a widely used and effective approach to solving systems of partial differential equations with random inputs. In this method, we generate the relevant parameter values from their probability distributions and the governing equations are solved for many such samples. This gives a set of samples of the output variables, from which one can calculate various statistical quantities of interest, such as mean values, variances and estimates of cumulative distributions functions. The classical Monte-Carlo method is very easy to implement, it can be applied to any type of problem including non-linear problems, it does not suffer from the so-called “curse of dimensionality” and it is possible to compute an estimate of the error as part of the solution process. The main difficulty with the method is its very slow rate of convergence: the error decreases as the inverse of the square root of the number of samples. For large-scale time dependent problems, this would require an impractical number of simulations in order to study the impact of heterogeneity via a brute force Monte Carlo approach. We investigated alternative methods for solving the flow and transport governing partial differential equations with random inputs, considered to be in our case the permeability field. There are some improvements to Monte Carlo simulations. We implemented and tested approaches to quantify uncertainty on the prediction of flow and transport in porous media with application to CO2 Storage sequestration; Monte Carlo (MC), Quasi Monte Carlo (QMC), Multi-Level Monte Carlo (MLMC) and Gaussian Emulator (GE), and a detail comparison between their accuracy and efficiency is reported.

Monte Carlo method (MC)

This approach is straightforward if the simulator is computationally cheap enough to be evaluated a sufficient number of times. Unfortunately, for complex simulators, the Monte Carlo approach is too computationally expensive.

Quasi Monte Carlo Method (QMC)

We use this method to approximate the solution of our problem, based on pseudo-random number sampling algorithms. The Quasi Monte Carlo (QMC) method aims to choose adequate locations of the pseudo-random numbers in a deterministic manner when filling the sampling space. It guaranteea that the points are uniformly spaced in the random space in order to obtain the maximum possible information about the quantity of interest with the minimum number of simulator runs, and therefore optimizing the MC method.

Multilevel Monte Carlo Simulation (MLMC)

To overcome the computational cost of Monte Carlo (MC) simulation, Cliffe et al. (2011) and Bath et al. (2011) employed the multilevel Multilevel Monte Carlo (MLMC) method for the uncertainty quantification in the groundwater flow simulation. The basic idea of the multilevel Monte Carlo (MLMC) technique is to find in an efficient way the expected value of a linear of nonlinear functional in achieving the same tolerance level.

Gaussian Processes Emulators (GE)

Process emulator is a statistical model used to make maximum the use of the outputs of a complicated computer-based simulation model. The overall analysis involves two models: the simulation model, or "simulator", and the statistical model, or "emulator", which notionally emulates the unknown outputs from the simulator. The Gaussian process (GP) emulator model treats the problem from the viewpoint of Bayesian statistics. In this approach, even though the output of the simulation model is fixed for any given set of inputs, the actual outputs are unknown unless the computer model is run and hence can be made the subject of a Bayesian analysis. The main element of the GP emulator model is that it models the outputs as GP on a space that is defined by the model inputs. The model includes a description of the correlation or covariance of the outputs, which enables the model to encompass the idea that differences in the output will be small if there are only small differences in the inputs.
The Bayesian methodology relies on a prior belief on the functional form of the model that is updated in the light of the training data in order to define the function posterior distribution. In general, a posterior distribution can be obtained from the prior distribution and the likelihood by applying Bayes rule. An emulator, sometimes called meta-model, only requires a small number of runs of the expensive simulator, making the process computationally cheaper than carrying out the full MC analysis. Comparative results of Gaussian and emulation and QMC approaches are presented in Figure 26.

In summary, we have approached heterogeneity by means of a mathematical developments leading to new balance equations, and by means of a highly efficient statistical approach (Gaussian Emulation).

Far field impacts and impacts on fresh-water aquifers

Impact of the stored CO2 in the pressure regime

We simulated CO2 injection and the resulting pressure buildup in a synthetic aquifer and two real aquifers. The synthetic example considered is a dipping aquifer with depth trend of porosity, permeability, salinity and temperature. A schematic for the geometry of the generic aquifer is given in Figure 27 (Left).

In the synthetic example, we examined pressure buildup at different distances from the injection well in the up-dip direction using the TOUGH2-ECO2N simulator. The injection rate was 1.0 Mt/year and the injection period was 50 years. Results showed that the pressure plume could reach a large distance (up to a hundred kilometers), while the front the injection fluid can only travel a few kilometers. Pressure plume can drive large-scale brine migration in the aquifer and can cause leakage of brine to other geological units through localized pathways. We showed that an analytical solution can be used to predict the pressure plume evolution. Because there is a depth trend for the properties of the formation material (e.g. porosity and permeability) and the fluid (e.g. salinity and viscosity), average values of the fluid and medium properties between the injection well and pressure observation points needed to be used for the analytical calculations. The two real aquifer examples considered were the South Scania site (Figure 27, Right) and the Baltic Sea Basin site (Dalders Monocline). In these examples, injection-induced pressure buildup were simulated and its sensitivity to various aquifer parameters and boundary conditions were investigated. Figure 28 shows the results for the South Scania site. We showed that the most sensitive parameter is the aquifer permeability. The boundary conditions were also shown to have significant influence on the pressure evolution. Both the near-wellbore and far-field pressure buildup can be predicted using a two-phase two-component analytical solution for pressure.
We developed a series of modelling approaches of varying complexity and evaluated their output for the three cases of aquifers. Modeling procedures included: 1) generalized multiphase modelling Injection and migration of CO2 into brine formations present a two-phase flow problem, using the simulator TOUGH2-ECO2N; 2) Vertically integrated models with a sharp interface between the CO2 and the water phases; 3) single phase hydraulic model. The latter is particularly attractive for the prediction of the pressure impact (as explained above), as single phase, hydraulic models are fast and easy to construct.

We show that semi-analytical approach based on superposition of image well solutions can reasonably well match the pressure prediction based on numerical modelling (Figure 29). Using the analytical solution, we examined the sensitivity of pressure buildup to variations in parameters representing medium and fluid properties, such as formation permeability, porosity, rock compressibility, relative permeability, viscosity, etc. It was found that permeability is the most sensitive parameter for pressure buildup. Figure 30 shows that the pressure increase exhibits a wide range of values as permeability varies.

Impact of geochemical changes resulting from the presence of CO2 on the mobilization of potentially hazardous minerals trapped in the solid matrix

In addition to the general changes in the chemistry of the brine, we focused on the issue of the possibility of mobilization of hazardous materials residing in the solid matrix. In general, it is assumed that the brine in the reservoir and the solid matrix have been in equilibrium for a very long time. However, when the rock formation contains hazardous metals, there is a risk that the chemical changes that will occur subsequently to the CO2 injection will cause these metals to dissolve in the brine and be displaced with it. We could envisage several scenarios. Brine may migrate through the fracture, eventually reaching an overlying a fresh ground water aquifer and then pollute the aquifer. If hazardous materials are present in the brine, they will also contribute to aquifer pollution. If injected CO2 reaches the fracture, or improperly sealed wells, it will migrate upward through the fracture. The critical point of CO2 is at Tcrit = 31.04°C, Pcrit = 73.82 bar. At lower (subcritical) temperatures and/or pressures CO2 can exist in two different phase states: a liquid and a gas, as well as a mixture of these states. When the CO2 escaping from the storage reservoir reaches reservoirs or aquifers at lower temperatures and pressures, it will enter the latter under subcritical conditions. Thus, the CO2 migrating through a fracture may reach an aquifer as a mixture of CO2 and brine. For the sake of simplicity, we have assumed that CO2 will reach the overlying aquifer with brine only at residual brine saturation. Since the CO2 will be under subcritical conditions, the spreading of the CO2 in the GW aquifer is examined as single phase flow model where the ground water is the only phase and CO2 is represented as a dissolved species.

We investigated the mobilization of Pb subsequent to the storage of CO2. The multiphase flow and reactive transport modeling is performed with the PFLOTRAN code. Flow processes can be modeled isothermally (as done in this case) or non-isothermally, and phase conditions may include a single (aqueous or CO2 -rich) phase, as well as two-phase mixtures. The CO2 in the reservoir is in supercritical form. First, an initial equilibrium run was conducted by the PHREEQC code with a specified mineralogical condition to establish a quasi-state composition of the brine and the host rock prior to the CO2 injection. After completing the initialization step, reactive transport simulation was conducted for the hydrogeological environment with the initial aqueous chemical composition obtained from the initial equilibrium run. The dissolution of CO2 in a rich carbonate solution may produce a significant decrease in the pH level and a dissolution of minerals comprising the solid matrix. Since a large amount of CO2 is injected into the reservoir, the brine pH level decreases significantly from the pH level of 7.3 at the beginning to pH in the range of 5 to 6.45 after 1 month. The galena mineral controls the aqueous lead concentration under acidic conditions. Thus, we expect a higher Pb concentration within the brine since the pH level is reduced as a result of the CO2 injection. The amount of Pb added to the brine is significantly low and it lies in the range between the initial concentration and 3 times higher (1.4x10-9 to 3.36x10-9 mole per liter).

Code development and evaluation of HPC software

We added CO2-relevant functionality to the CSIC object-oriented modeling platform PROOST. We also have developed a new code (CHEPROO++), capable of dealing with complex geochemical problems. In PROOST, the following elements have been implemented:

Equations of state: PROOST has been equipped with an equation of state module, which implements different state-equations, describing density, viscosity and enthalpy of mixtures of CO2 and brine in function of pressure, temperature and composition. Among the models included are the Redlich-Kwong equations, Peng-Robinson, and a model based on interpolation of a table of properties coming from the Span and Wagner equations.
PROOST is also equipped with a model for the equivalent state properties of a fluid-gas mixture of CO2 (as is present during CO2 phase change) which allows us to simulate the descent of CO2 in an injection well.
Multiphase flow equation: PROOST models multiphase flow with two-phase mass balance equations. They can be expressed in terms of pressure or saturation. Capillary effects are modeled with Van Genuchten or other retention and relative permeability curves.
Solvers: PROOST has been equipped with a rich set of solution methods for the multiphase flow equations. They can be solved fully coupled (with Newton-Raphson or modified Newton-Raphson) or with sequential iterations (Picard’s method); if using Picard’s method, the individual phase balance equations can be solved implicitly or explicitly. In addition, we have implemented domain-decomposition methods, where we solve multiphase flow using Newton-Raphson’s method in the fraction of the domain where two phases are present, while implicit solving is used in the fraction of the domain where only water is present.

The CHEPROO++ code is conceived as a portable library that can be easily coupled to different codes (e.g. conservative transport simulator) and easily modified thanks to its object oriented (OO) modularity. The OO design methodology is widely credited for reducing coupling (referring to the interdependence of different parts of a body of software) and favoring code reuse. At its core are the concepts of “class” and “object”, roughly analogue to the mathematical concepts of “set” and “element”. Within this analogy, the class is a definition of a set (such as for example the complex numbers) and of the possible operations over this set (for example multiplication). Therefore, a class definition contains both a storage structure and a collection of functions. Classes can be used to represent a wide range of concepts in the context of a geochemistry simulation code, for example “phase”, “species” or “reaction”. The reduction in coupling is achieved by two mechanisms: first, common interfaces are defined for different specialization classes (say, “aqueous phase” and “mineral phase”) so that the code that uses a class (“phase”) becomes independent of the particular specialization that is being used. Second, it is discouraged to directly access class attributes from outside the class. Instead, an interface is defined to ask for information or give it to a class using methods. That way, a code that uses a particular class becomes independent of the storage structure used by the class. Besides, this structure of the code facilitates its expansion, since developers can implement new functions and specializations (inherited classes) without knowledge of the wider program structure.

Comparison of Parallel simulators

The simulation of CO2 injection and spreading a reservoir scale is a computationally intensive task. In order to allow meaningful waiting times for a given simulation, there is a need use simulation software in the context of high performance computation (HPC) framework. We evaluated two public domain HPC or massively parallel codes, capable of simulating two-phase water-CO2 flow in reservoirs: PFLOTRAN and DUMUX. In order to conduct the evaluation we have set we have set a HPC computing structure, which included two Dell Precision R5550 Workstations with two Intel 6-core Xeon CPU X5690. In total, we had 24 cores and 48GB of RAM (24 GB per workstation or node). The workstations are connected via Mellanox InfiniBand high-performance adapters, which deliver up to 40GB/s transfer rates. This structure is expandable to an unlimited number of nodes (workstations). We tested these two models with a synthetic domain of 5km by 5Km by 20 m, made of homogeneous sandstone. We injected 2.5 Million tons of CO2 over 5 years and then had a relaxation period of 45 years. Results obtained by DUMUX and PFLOTRAN are different, even for this very simple case (as shown in Figure 31). We have no clear explanation about this discrepancy at the exception of the apparent lack of maturity of DUMUX as compared to PLFOTRAN, at least in the context of CO2 storage simulations.

PFLOTRAN is written in the FORTRAN 90 language and is well structured. DuMux is written in c++, which is a more flexible language. We were able to compile PFLOTRAN both under the Linux and Windows operating systems, while we failed to compile DuMux under Windows and succeeded with Linux. The setting of a simulation under PFLOTRAN is straightforward, as there is a clear separation between the code and the input data. On the other hand setting a simulation under DuMux requires intervention in (changing) the code and therefore a clear understanding of the code structure and the programming language. This can be seen as a major drawback, as the need to program the code for each case may result in many errors and is probably time consuming. Overall, it seems that PFLOTRAN is a more mature and ready to use simulator with very good HPC characteristics.

Cost-effective and reliable monitoring

Monitoring technologies would include, direct measurement of pressure and temperature (P/T) along the injection tube from the wellhead to the well bottom, distributed temperature sensing (DTS) via optical fiber, downhole fluid sampling, Electrical resistivity tomography (ERT), seismic monitoring (SM), neutron scattering based measurement. In Heletz, we have installed in the frame of the MUSTANG project, P/T, DTS and fluid sampling. In Hontomin and ERT was installed in addition to P/T, DTS and fluid sampling. These technologies and their installation (both downhole and above the ground) require relatively heavy investment and are associated with many risk of failure during the installation and or operation. However, the largest investment in monitoring is by far the drilling and the completion operations of the deep well. Therefore, cost-effectiveness would mostly focus on optimizing the amount of information that could be produced from a well, by installing as much as possible complementary monitoring technologies, capable of allowing to produce a more reliable and precise picture of the injected CO2 by combing their output.

Development, production and shallow testing of a multi-parameter downhole array

We designed and built a new multi-parameter downhole array for gas geological storage monitoring, installed and tested it through the clastic sequences of the Maguelone shallow experimental site (Mediterranean coastline, Gulf of Lions, France). Called RTSP (Figure 32) for simultaneous and multi-parameter monitoring (resistivity, temperature, seismic and pressure). In a later gas injection experiment from a well located 10 m away from the new downhole observatory at surface, the RTSP was further tested and a gaseous CO2 plume was detected at depth by all RTSP monitoring sensors deployed as part of PANACEA. Such a multi-parameter monitoring system should reduce the number of monitoring wells in the future, hence reduce monitoring costs as well as the possibility of CO2 leakage along monitoring holes. This prototype has been further developed in order to be deployed the deep monitoring well to be drilled at Heletz in the frame of the TRUST project.

CO2 injection experiments at shallow depth

A CO2 injection experiment in a shallow aquifer at Maguelone has been monitored with different surface and downhole techniques (seismic, electrical resistivity, pressure, temperature and water sampling). Based on previous N2 and CO2 injection experiments and monitoring in 2012-2013, a new injection hole was drilled down to 10 m depth and screened only at 8-9 m depth in front of the R2 reservoir in order to have gas injection and gas storage at the same depth. Also a new single multi-parameter (resistivity, temperature, SP, pressure and seismic measurements) observatory was drilled and equipped for this experiment. Such specific monitoring system should reduce the number of monitoring wells, hence redure possible CO2 leakage along monitoring holes as well as monitoring costs. An amount of ~48 m3 of gaseous CO2 was injected during 2 hours into a small saline coastal reservoir with impermeable boundaries located at 8-9 m depth. CO2 storage happened in the same reservoir at the base of the Late-Holocene lagoonal sediments (impermeable dark green clays) forming an impermeable seal overlying the homogeneous fine-grained Pliocene continental deposits. The gaseous CO2 plume was detected at this depth by all of the monitoring techniques deployed during the experiment.

Recommendations regarding MMV technologies

Monitoring the storage of CO2 is a necessity and a regulatory requirement. For deep geological formations, the cost of the drilling will represent the major component of the total cost of monitoring and measurement and therefore one should aim at maximizing the amount of data and information that can be safely produced from a monitoring well. Pressure and temperature information are key for the determination of the state of the CO2 and for model enhancing the reliability of the model prediction. The use of optical fiber in the context of CO2 has been limited to temperature sensing, however other types of information could be produced, provide that adequate deconvolution software and hardware are available. This would include: strain, stresses and pressure. Bragg sensors are also a very promising technology. This sensors can provide precise measurement of temperature, but could be extended for CO2 saturation. Fluid sampling in key in the context of scientifically oriented experiments, but less in the context of industrial deployment. Finally, well completion is an important part of the MMV technology. So far monitoring wells are perforated and thus could become a leakage path through the well. “Behind the casing” completions, whereby the monitoring technologies are installed in the cementation and the perforation takes place from the cementation towards the formation, could represent a much safer alternative. This, combined to the concept of integrated technologies could represent the most adequate MMV platform for CO2 storage.

Potential Impact:
Potential impact

Scientific and Regulatory

PANACEA is a scientifically driven project, with substantial contribution to the understanding of long-term fate of the stored CO2. PANACEA was exposed to the scientific community in a large number of international forums such as AGU, EGU, CO2GEONET, GHGT and TCCS and to the regulatory community such as during the information days held in Israel (for the Israel Water Authority). The key impact of SMART is beyond any doubt scientific and its findings are mostly destined to the scientific community.

At the scientific level, we can definitely claim that PANACEA has vastly contributed to the improvement of the state of the art as follows:

1. Improved understanding of the long term storage of CO2 via natural analogues;
2. Improved understanding of the fundamental processes occurring during the CO2 injection and spreading thus leading to accurate and reliable models of the storage;
3. Development of fast solutions for CO2 trapping and leakage for simple geometries, allowing a Tie-1 evaluation of the CO2 storage risks;
4. Development and testing of simplified, hydraulic, models capable of predicting pressure impacts of the CO2 storage at regional scale;
5. Development of novel methodologies for incorporating the impact of geological heterogeneity into a new mathematical framework;
6. Development of new statistical approach, which enables a highly efficient quantification of the uncertainty impacts;
7. Quantifying leakage at regional scale;
8. Developing procedures for leakage mitigation;
9. Development and testing of novel monitoring technologies.

Some of these findings have the potential to contribute beyond the scope of the scientific community:

1. Public acceptance: the good history matching results obtained for Sleipner, which is so far the largest CO2 storage site, could contribute to a greater confidence in the model predictions.
2. The fast solutions and the approximate models will enable regulators to have a better understanding of the CO2 storage, a better assimilation of the challenges and a better evaluation of the risks. CO2 storage models are notoriously known for their complexity. They construction and operation requires lots of expertise and therefore are not widely accessible. In that context, fast solutions and approximate (and far simpler) models providing reliable results on some of the key impacts of the stored CO2 would contribute to a better and wider understanding of the storage and a more rational approach with regard to its inherent risks.
3. The development of the concept of integrated monitoring technologies, capable of increasing the amount of information per monitoring well would contribute to the regulatory requirements with regard to the monitoring of the storage.

Socio-economic Impact

It could be a difficult to identify how the PANACEA findings, which are mainly of scientific nature, could have a direct and measurable socio-economic impact. The PANACEA consortium comprises mostly scientists with no a priori bias with regard to the suitability of CO2 storage, both as a mitigation technology in general or as a technology for a specific site. Therefore we could claim that our findings could have the potential to be more widely accepted by non-expert and therefore would contribute to boosting public acceptance, which undeniably a socio-economic factor.

Public Acceptance

A major point that we must address is how CCS project developers should take into account public acceptance what should be the role of CCS in the portfolio of climate change technologies from the public and stakeholders point of views. We prepared a comprehensive documental analysis reviewing public acceptance studies of the last few years. From this documental analysis, we prepared a, non-exhaustive, inventory and summary of public engagement good practices. We highlighted different factors, which can influence public perception of a project, such as the concept of trust, the level of knowledge of the public, the spatial vicinity of the public to the CCS site or the increase of awareness to the wider public on the climate change issues. Then, we conducted a state of the art of the existing CCS legal framework at global level, including the role of international laws in the CCS regulatory mechanism as well as presentation of the emission trading system in place. In countries with a high level of interest for CCS or EOR technologies, such as Australia, Canada, the United States or Europe (Norway), logically a greater policy response has been put in place. In recent years, several CCS emerging countries such as China, Malaysia or Qatar have also settled basis of a regulatory framework on CCS. Finally, we focused on the European Directive and conclusions about future prospects on CCS regulations. In particular, we identified potential links between public acceptance and regulations such as the development of a mature CCS regulatory framework including effective risk communication to engage the public and involvement of all stakeholders in risk-related decisions. We also conducted a consultation phase of several CCS project developers and stakeholders. Interview guides for project developers and other stakeholders have been developed specifically for this study and are include in the relevant deliervable. The consultation highlighted the difficulties to communicate between project developers and stakeholders. Among the reasons, the lack of confidence of the stakeholders toward industry prevented any further dialogue between both parties. Independence of CCS communicant, technological risks, financial and technical liabilities of the industrial management and transparency in the communication of the project developers are issues reflecting public expectations. Finally, address stakeholder’s perception on CCS projects and the public expectations on CCS regulations. They present the lack of confidence of European citizens that have been observed toward the decision-makers and regulatory-makers. Recommendations are made for the development of a mature regulatory framework, addressing public concerns, and the organization of a large citizen public debate, facilitated by a neutral third-party.

Communication strategy and implementation

We established a communication strategy, based on traditional communication tools (flyers, newsletter), internet site and the organization of international workshops aimed at improving the communication between scientists and more generally the CCS community, as preliminary step towards the progress in communicating with the wider public.

PANACEA website ( One of the major achievements is the PANACEA website implementation. We launched it at an early stage of the project and frequently updated it during the project lifetime. The website is first an internal organisation tool, through the intranet section, to assist the consortium partners in sharing the findings of their work as well as useful information about internal meetings. The website is also an external tool to disseminate and promote the results of PANACEA by gathering and integrating the results of the work packages WP2 to WP8 in a consistent way. The PANACEA website includes 14 sections and 5 additional sections for logged in partners. The website is evolving tool. We updated and enriched its content during the whole life of the project.

PANACEA flyer: The PANACEA flyer was created by a French professional agency (Vent Portant) to deliver clear, transparent and reliable information on the PANACEA project at a glance. It aims at being distributed in CCS conferences and other related events. It was distributed for instance during the Trondheim CCS Conference in Norway (from 4th to 6th June 2013). It can be downloaded at

Movie: we prepared at the request of EU a movie on the projects in which we are involved (including MUSTANG, PANACEA, TRUST and CO2QUEST). The movie was shown on March 31st at the Conference on impact of EU research in southern and eastern countries. The movie comprised short explanations of the research work, the findings and a short survey of the Heletz site. It will be posted on the PANACEA, MUSTANG and TRUST sites, after some optimization for file size reduction.

Scientific Dissemination

PANACEA consortium, such as oral and poster presentations in meetings and internationally recognized conferences (GHGT, AGU, EGU, UKCCS, EU-Australia events…) and scientific papers published in CCS related journals (Journal of Fluid Mechanics, International Journal of Greenhouse Gas Control, Water Resources Research...). Dissemination activities related to 2012, 2013 and 2014 summarized in the relevant PANACEA deliverables, also available for download on the PANACEA website. Overall, the PANACEA consortium was extremely prolific regarding dissemination over the three-year project duration. More than thirty papers have been published in scientific journals, and about fifteen additional papers are still expected for 2015. In addition, a total of seventy oral and poster presentations have been performed by the PANACEA consortium in various conferences and events over the three-year project duration.

Brainstorming Day – 2013 – Trondheim (Norway)


In order to improve communication and collaboration between different CCS projects, PANACEA joined forces together with four other FP7 projects (MUSTANG, ULTIMATE CO2, CO2CARE and CARBFIX) to organise an international CCS workshop in Trondheim, Norway: the “Brainstorming Day on the long-term fate of geologically stored CO2”. This event was hosted by Statoil in the Clarion Hotel & Congress Trondheim.

The event took place on June 3rd 2013, just before the Norwegian bi-annual Trondheim Carbon Capture and Storage (TCCS-7) Conference, from June 4th to 6th 2013. Grouping of the Brainstorming day with this internationally known CCS event was a great opportunity to reach a wide scientific and industrial public.
The Brainstorming Day gathered 49 attendees. The targeted audience was a technical and scientific public from universities, research institutes and industry. Initially, only partners involved in the five related projects (MUSTANG, ULTIMATECO2, CO2CARE, CARBFIX and PANACEA) were invited. Then, the registration has been opened to external participants. In particular, members of the CO2FIELDLAB project ( SITECHAR project ( and CCS Network ( were invited. With regard to the organisation, the Brainstorming Day was made of six one-hour sessions including two phases:

- The first phase of each session was aimed at defining a number of key issues related to the topic, during approximately a quarter of hour. Experts were in charge of presenting a state of the art, gaps and challenges and a series of key questions to be addressed.
- The second phase of each session was a forty-five-minutes open discussion on these topics, with a guidance of an appointed panel of experts.

Communication activities for promotion

To promote efficiently this event, multiple communication channels have been used.

Mailing list constitution and email campaign

First, a mailing list has been created gathering members of the five consortia (ULTIMATECO2, CARBFIX, MUSTANG, CO2CARE and PANACEA). Members of this mailing list have received information about the Brainstorming Day by email.
Before the event, the aim of this mass mailing campaign was to inform the members about the organisation (date, location, program…) and to ask them for registration. Registration of the participant purpose was to check at any time that the number of attendees did not exceed the total capacity and/or to plan additional invitations.
After the event, the aim of the email campaign was to inform the members about publications related to the Brainstorming Day: book compiling the presentations, movie of the event, list of the attendees...
A new mail address need has been created for the event promotion:


A graphic chart was defined for the communication material (website, flyer, brochure…). A dedicated logo was created for Brainstorming Day by a professional communication agency (Vent Portant).


A dedicated internet page has been created for the Brainstorming Day to sum up all organisation details and allow attendees to register for the event. A specific domain name was allocated: To ensure its visibility, the registration tool was be available on each page of the website.

The website content was the following:
- Home
- Practical information
- Programme
- Speakers
- Links
- Contact


A flyer of two A4 pages was prepared for the Brainstorming Day. It aimed at summing up all the information at a glance. No mass printing was planned, the flyer has been published in pdf file on the Brainstorming Day website, available for everybody to print it individually.


A brochure of ten pages was created for the Brainstorming Day. This document aimed at summing up the programme of the event, abstracts of the sessions, information related to the five organising projects and other practical details related to the Brainstorming Day.The brochure was printed and distributed to the attendees during the event.

Press release

Before the event, a press release was written and sent to newspapers and other organisations (Carbon Capture Journal, IEA-GHG, GCCSI, CCS Network, CCSA, SCCS...). The link to the website, to the flyer, the logo and contact details were attached. Following this press release, list of publications related to the Brainstorming Day is showed in table 1:


After the Brainstorming Day, the minutes of meeting of the event were compiled in a proceedings document. The main interest of this document was to reflect the interactive discussion with the audience. Therefore, for each session, it included the presentations and a short summary of the topics addressed during the discussion part.

Movie of the event

As this event aimed to be interactive with many unplanned interventions from the audience, a movie of the event was made to capture the exchanges. Once edited, the movie will be uploaded on YouTube, published on the Brainstorming Day website and sent to the attendees. It might also be published on the website of each CCS project involved in the organisation. The movie will be a great communication tool to be sent to the press and to be published on the social media.


A major part of the communication work took place in the period of February-March 2013 with the participant database constitution, website and flyer creation. The second intense communication period took place after the event, in order to send the book, the movie and the list of attendees to all the participants. In parallel, communication activities were carried out with the press, on blogs and social networks.

Brainstorming Day –2014 – Paris (France)

Following the success of the Brainstorming Day 2013, Bureau Veritas organised in 2014 a second edition of this event to disseminate the PANACEA findings towards the CCS community. The Brainstorming Day 2014 took place on December 19th 2014, at Bureau Veritas Head Office in Neuilly-sur-Seine (Paris), France. The one-day workshop gathered successfully more than 40 participants from diverse countries such as Germany, United Kingdom, Canada, Norway, Israel, Sweden, Spain and France. The targeted audience was made of policy-makers, legislators, CCS site planners, operators and CCS associations that advice previously listed organizations. The structure of the event, still allowing long time slots dedicated to open discussions, and the variety of topics addressed (public acceptance, regulations, processes in the reservoir, pressure, leakage and associated mitigation and remediation means, prediction, natural analogues, contemporary storage, uncertainties and monitoring) were highly appreciated by the attendees. Beyond the presentation of PANACEA outcomes, one specific session of the workshop was dedicated to the UltimateCO2 project, strengthening the link between the two EU FP7-funded projects. Proceedings of the workshop are available for download on the PANACEA website (

List of Websites:
contact person: Jacob Bensabat ( or Sagi Dror (