Energy Procedia 40 ( 2013 ) 418 – 427

1876-6102 © 2013 The Authors. Published by Elsevier Ltd.Selection and peer-review under responsibility of the GFZ German Research Centre for Geosciencesdoi: 10.1016/j.egypro.2013.08.048

ScienceDirect

European Geosciences Union General Assembly 2013, EGU

Division Energy, Resources & the Environment, ERE

A dynamic flow simulation code intercomparison based on the revised static model of the Ketzin pilot site

Thomas Kempkaa,*, Holger Classb, Uwe-Jens Görkec, Ben Nordena, Olaf Kolditzc,d, Michael Kühna, Lena Walterb, Wenqing Wangc, Björn Zehnerc

aGFZ German Research Centre for Geosciences, Potsdam, Germany bInstitute for Modeling Hydraulic and Environmental Systems, Universität Stuttgart, Stuttgart, Germany

cHelmholtz Centre for Environmental Research - UFZ, Leipzig, Germany dTechnical University Dresden, Faculty of Forest, Geo and Hydro Science, Dresden, Germany

Abstract

The available static geological model of the Stuttgart Formation at the Ketzin pilot site was revised based on the re-interpretation of the available 3D seismic data. Using this model three independent modelling groups initiated an intercomparison study using the standard industrial (ECLIPSE 100) and scientific dynamic flow simulations codes (TOUGH2-MP/ECO2N, DuMuX and OpenGeoSys) to employ their strategies for matching of downhole pressure and CO2 arrival times. The current results demonstrate that the introduction of distinct near- and far-well areas with different permeability tensors is required to achieve a reasonable match with the data observed at the Ketzin pilot site. © 2013 The Authors. Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the GFZ German Research Centre for Geosciences Keywords: CO2 storage, numerical modelling, Ketzin pilot site, TOUGH2, ECLIPSE, DuMuX, OpenGeoSys

1. Introduction

CO2 injection at the Ketzin pilot site located in Eastern Germany (State of Brandenburg) about 25 km west of Berlin is undertaken since June 2008 with a scheduled total amount of about 70,000 t CO2 to be injected into the saline aquifer represented by the Stuttgart Formation at a depth of 630 m to 650 m until

* Corresponding author. Tel.: +49-331-288-1865; fax: +49-331-288-1529. E-mail address: kempka@gfz-potsdam.de.

Available online at www.sciencedirect.com

© 2013 The Authors. Published by Elsevier Ltd.Selection and peer-review under responsibility of the GFZ German Research Centre for Geosciences

Open access under CC BY-NC-ND license.

Open access under CC BY-NC-ND license.

Thomas Kempka et al. / Energy Procedia 40 ( 2013 ) 418 – 427 419

the end of August 2013 [1-2]. The Stuttgart Formation is of fluvial origin composed of sandstone channels of high permeability embedded in floodplain facies of low permeability and a high heterogeneity of facies distribution, porosity and permeability [3]. The geological model of the Stuttgart Formation [4] was revised within the scope of the present study, and served as the basis to match CO2 arrival times in the monitoring wells and reservoir pressure to the observations. A code intercomparison between industrial (ECLIPSE) and scientific dynamic flow simulations codes (TOUGH2-MP/ECO2N, OpenGeoSys and DuMuX) was carried out to review the simulator capabilities by means of representing a complex heterogeneous reservoir. The simulation results achieved by the three participating modelling groups using four different numerical simulators are discussed within the scope of this manuscript.

2. Revision of the static geological model

The Stuttgart Formation in Ketzin consists of siltstones and mudstones deposited on a flood plain, in which sandstones of the channel facies are incised [3]. In distinct areas of the German Basin, a south to southwest-oriented palaeocurrent with transport from northern and eastern Europe across the German Keuper basin was observed [5]. However, flow directions for the northern part of Germany cannot be determined due to absent outcrops. In addition, the lateral extension of the channel belts, formed by amalgamation of individual fluvial channels, is highly variable. At Ketzin, the Stuttgart Formation is about 72 m thick [3] and the lateral extension of the channel belts may account from about 600 m to 2,500 m [4]. The encountered sandstones show the same homogenous grain-size pattern as observed basin-wide, indicating rapid transport and deposition (e.g. [6]). As site-specific data is limited in terms of spatial distribution and the seismic 3D data available for the Ketzin site is not able to resolve the internal structure of the Stuttgart Formation, the presented reservoir model follows an integrated geological concept taking into account the basin-wide observed characteristics of the Stuttgart Formation as well as site-specific point and spatial data. In general, the construction of the reservoir model is based on the geological concept discussed in more detail by Norden and Frykman [4] using a geostatistical approach to describe the facies distribution and the reservoir architecture of the formation. This model was updated using a revised interpretation of the seismic data and further adapted based on the available monitoring data.

In the current model, top and bottom of the reservoir were revised based on the re-interpretation of the available 3D seismic data. In addition, the fault system at the top of the Ketzin anticline structure was integrated into the reservoir model. The horizontal and vertical resolution of the model was chosen in order to allow the reproduction of the spatial CO2 plume migration and the prediction of its future development. Therefore, the 72 m thick reservoir formation was subdivided into the three zones a, b, and c (from top to bottom). The uppermost zone (zone a), where the main reservoir sand is located, shows a thickness of 24 m and was discretized by 0.5 m in the vertical direction. Zone b has a thickness of 12 m and a vertical discretization of 1 m, while zone c with a thickness of 36 m was discretized with 3 m. For all zones, the horizontal discretization of the geological model amounts to 5 m x 5 m.

The conceptual facies model of all zones focusses on the two major facies types, represented by floodplain facies and channel facies. Due to the sparse borehole data available from the Ketzin area, the further interpretation of the sandstone as channel depositions or levee-crevasse deposits is difficult. From literature, no levee and crevasse sequences were encountered in Central Germany [7], whereas in southern Germany, levee-crevasse deposits were described by Ricken et al. [8]. We still assume that levee-crevasse deposits are not present or of subordinated occurrence. The characteristics of the facies modelling, which was performed using the Petrel software package [9], are shown in Tab. 1. The channel properties were derived from borehole analysis (CO2 Ktzi 201/2007) and estimated based on the results of a revised spectral decomposition (SD) of the 3D seismic as initially performed by Kazemeini et al.

420 Thomas Kempka et al. / Energy Procedia 40 ( 2013 ) 418 – 427

[10]. The results of the revised SD were also used as an additional input for the facies modelling, serving as a probability map for the occurrence of the channel facies. In order to reproduce the monitoring results, especially the 3D repeat seismic [11-12] further artificial boreholes were introduced to the static modelling to guide the stochastic simulations. After several possible realizations of the facies geometry were established, the distribution of petrophysical properties within the facies needs to be addressed. This was modelled using a sequential Gaussian simulation. Based on the petrophysical core and log data available from the site [13], variograms of total porosity and effective porosity for the channel and floodplain facies, respectively, were established and extended by literature data [14]. In a first step, total porosity was simulated. Based on the seismic SD trend-maps of the total porosity distribution were established, allowing a co-Kriging of this parameter for the channel facies. Then, effective porosity was modelled using the established variograms and co-Kriging with the results of the total porosity simulation for the channel facies. In a last step, permeability was calculated using the determined poro-perm relationship for the different facies environment [4]:

a) Channel facies: PERMC = 9 x 106 x PHIT6 and b) Floodplain facies: PERMF = 9 x 1010 x PHIT14.5.

The parameters and values used in the petrophysical modelling are summarized in Table 2.

Table 1. Input parameters used in facies modelling

a) For zones a and b b) For zone c

Min. Mean Max. Min. Mean Max.

N/G ratio - 0.38 - - 0.28 -

Channel direction 345 355 370 0 45 60

Channel amplitude 400 1,200 2,500 400 1,200 2,500

Channel wave length 4,000 5,000 8,000 4,000 5,000 8,000

Channel width 100 480 1,600 100 480 1,600

Channel thickness 1 4 8 1 3 10

Table 2. Parameter and values used for petrophysical modelling of all zones. Exception: For zone c, an anisotropy azimuth of 15 was used (asterisk in the table)

Min. Max. Vert. Comments

Total porosity channel facies 0.08 0.30 Allowed output range, collocated co-Kriging with channel

trend map (correlation coefficient 0.8)

Effective porosity channel facies 0.01 0.26 Collocated co-Kriging with total porosity (correlation

coefficient 0.8)

Total porosity floodplain facies 400 2,500 Allowed output range

Effective porosity floodplain facies 0.01 0.24 Collocated co-Kriging with total porosity (correlation

coefficient 0.8)

Anisotropy range Anisotropy azimuth

250 m 345* 400 m 1 Variogram input

Thomas Kempka et al. / Energy Procedia 40 ( 2013 ) 418 – 427 421

3. Dynamic flow modelling

The involved work groups from GFZ German Research Centre for Geosciences, University of Stuttgart and UFZ Environmental Research Centre Leipzig used four different numerical simulators anddifferent strategies to calibrate their models to the Ketzin pilot site monitoring data. All simulations werebased on the same revised static geological model discussed above.

3.1. GFZ German Research Centre for Geosciences simulation results using TOUGH2/ECO2N and ECLIPSE

The GFZ Hydrogeology Section work group used the industrial standard simulator ECLIPSE 100[15][Schlumberger, 2009] and the scientific open source simulator TOUGH2-MP/ECO2N [16-17]. Modelimplementation and parameterization are discussed in detail in Kempka et al. [18] and Kempka and Kühn [19], while long-term CO2 trapping mechanisms and pressure effects are addressed by Kempka et al. [20] and Klein et al. [21]. Pressure matching was undertaken using a near-field (near-well) and a far-field area determined by the size of the CO2 plume after a simulation time of 400 days. Different permeabilitytensors (determined by permeability multipliers in the main axial directions) were applied at the near andfar-field present in the reservoir model. Fig. 1 illustrates the downhole pressure monitored at the CO2 Ktzi 201/2007 injection well [22] and the pressures simulated with the TOUGH2 and ECLIPSEsimulators. The simulation results are in good agreement with the observed data with a deviation of lessthan 3 %. Differences occurring between the TOUGH2 and ECLIPSE simulator are discussed in Kempkaand Kühn [19] in detail.

Fig. 1. CO2 Ktzi 201/2007 injection well downhole pressure simulated with the TOUGH2 and ECLIPSE simulators is in goodagreement with the observed data

Fig. 2 shows the comparison of simulated downhole pressure in the CO2 Ktzi 202/2007 observation well with the observed data which was recorded for about one year during the Ketzin pilot site operation.A slight deviation (less than 1 bar) can be observed towards the end of the simulation runs potentiallyassociated with the implemented boundary conditions [18].

The CO2 mass balance for about four years of simulation time is plotted in Fig. 3. The amount of gaseous and dissolved CO2 shows low deviations during the injection period for the simulations carried out using TOUGH2 and ECLIPSE. The free gas phase calculated by DuMuX is about 10 % higher after four years.

422 Thomas Kempka et al. / Energy Procedia 40 ( 2013 ) 418 – 427

Fig. 2. CO2 Ktzi 202/2007 monitoring well downhole pressure simulated with the TOUGH2 and ECLIPSE simulators is in excellent agreement with the observed data

Fig. 3. CO2 mass balance calculated with the ECLIPSE, TOUGH2 and DuMuX (discussed in the following paragraph) simulators for about four years of simulation time

After applying two different permeability tensors to the geological model, simulated CO2 arrival timesas discussed in Kempka and Kühn [19] show deviations of 6 % (ECLIPSE) to 11 % (TOUGH2) for theCO2 arrival in the first observation well (CO2 Ktzi 200/2007) and 6 % (ECLIPSE) to 15 % (TOUGH2)for the second one (CO2 Ktzi 202/2007). This variation is accounted to the well implementation availablein ECLIPSE (well flow equations) and the work around applied in the TOUGH2 simulations to account for CO2 injection into the Stuttgart Formation in the numerical model [19].

3.2. University of Stuttgart simulation results using DuMuMM X

The work group from Stuttgart University applied the numerical simulation environment DuMuX

(www.dumux.org; [23]). DuMuX is an in-house product, specifically designed for flow and transport of multiple fluid phases through porous media and based on DUNE (Distributed and Unified NumericsEnvironment, www.dune-project.org). The results of this study are obtained with the isothermal CO2module including compositional effects. DuMuX offers implementations of different spatial discretization methods. For the present study, both the Box method and a cell-centred finite volume scheme were tested.ff

Thomas Kempka et al. / Energy Procedia 40 ( 2013 ) 418 – 427 423

In the following two different approaches carried out by the Stuttgart University work group are briefly discussed. The first one uses the Box method on a tetrahedron mesh and follows the same strategyas applied by the GFZ work group, i.e. fitting two permeability tensors, thereby distinguishing between anear- and a far-field. The second approach uses a hexahedron mesh, in fact a very similar mesh asKempka and Kühn [19] applied, and the cell-centred finite-volume method. Another history matchingstrategy is pursued here, where an inverse model is applied to match less parameters than before to thedata during the first 50 days, and a subsequent extrapolation to the full injection period is applied.

The tetrahedron mesh required a mapping of the permeability and porosity values provided on a hexahedral grid in the reservoir model. The tetrahedron grid has a circular refinement region around theinjection well and observation wells with a coarser resolution in the far-field. The hexahedron mesh isalso refined around the injection well. Both meshes are shown in Fig. 4.

Fig. 4. Tetrahedron (left) and hexahedron mesh (right) with the mapped permeability fields

The discretization length in the hexahedron mesh is 5 m in the near-well region in horizontal directionand 0.5 m in vertical direction in the upper part of the formation, increasing to 2 m in the middle and 5 m in the lower regions of the formation. In the tetrahedron mesh, the grid resolution is below 2 m in the near-well region. However, due to grid quality requirements, the horizontal and vertical discretization lengths are linked to each other. Thus, the vertical resolution of the tetrahedrons cannot always resolvethe strongly channelled structure of the geological model as defined by the 0.5 m resolved (in the vertical)hexahedrons. This is even more severe in the far-field, where the vertical heterogeneity cannot berepresented well by the coarse tetrahedrons. This issue of gridding and mapping of geological data needs to be kept in mind when comparing the different history matching results. The pressure match obtainedwith the tetrahedron approach using the same six multipliers as the GFZ work group is shown in Fig. 5.Tab. 3 lists the different sets of fitted permeability multipliers.

Table 3. Different sets of fitted permeability multipliers

Simulation 1 Simulation 2 Simulation 3

Region Near-well Far-field Near-well Far-field Near-well Far-field

-direction 0.34 0.34 0.272 0.85 0.136 1.02

r y-direction 0.07 0.07 0.056 0.175 0.028 0.21

Mulitplier z-direction 0.07 0.07 0.056 0.175 0.028 0.21

Arrival at CO2 Ktzi 200/2007 (days) 25.5 26.92 27.7

CO2 Ktzi 202/2007 (days) 502.46 562.24 765.82

424 Thomas Kempka et al. / Energy Procedia 40 ( 2013 ) 418 – 427

Fig. 5. Pressure at the injection well CO2 Ktzi 201/2007 (left) obtained with the tetrahedron mesh and fitting of six permeability multipliers

The reduction of permeability leads to a significant overestimation of the arrival times, in particular at the CO2 Ktzi 202/2007 well. This discrepancy is assumed to occur due to the applied mapping of the geological model to the tetrahedron grid (inverse distance method). Therefore, the second approach with the hexahedron mesh is undertaken using the cell-centred finite-volume method. This second approach involves (i) estimating three parameters by inverse modelling of the first 50 days (ii) extrapolating the matched model to the full injection period and further adapting the model parameters to match the pressure measurements, and (iii) insert a local geological feature according to the characteristics in the reservoir for matching the second arrival time. This procedure is described in detail in Walter [24].

Fig. 6 displays the matched pressure curves for the second approach. The left figure shows that with the global multipliers, the extrapolated best-fits from the 50 days inversion lead to a significant deviation after roughly 400 days. Distinguishing between near- and far-field, the pressure can be matched with very similar accuracy as presented above by the GFZ work group. The arrival time may be matched either by implementation of a local hydraulic barrier in the upper part of the formation between the injection well and the second observation well (as described in [24]) or by anisotropic permeability multipliers as applied by the GFZ work group.

Fig. 6. Pressure match for different sets of estimated parameters (left and right) obtained with the hexahedron grid

Thomas Kempka et al. / Energy Procedia 40 ( 2013 ) 418 – 427 425

3.3. UFZ Environmental Research Centre Leipzig simulation results using OpenGeoSys

OGS (OpenGeoSys) is an open-source platform for numerical simulation of coupled thermo-hydro-mechanical / chemical (THM / C) processes in porous and fractured media (www.opengeosys.org). The object oriented finite element code is designed for applications in geomechanics, catchment hydrology and energy research [25].

A fine finite element mesh is necessary to cover the details of the geometry and the material properties of the Stuttgart Formation at the Ketzin pilot site. This leads to a large number of unknowns of the equations to be solved. In the present study, we employ PETSc routines and data structures to parallelize the finite element computing. The current parallel finite method takes the overlapping domain decomposition approach. As a prerequisite of this approach, the finite element mesh has to be partitioned into subdomains, which is done by using METIS before starting a simulation. With the current parallel finite element scheme, a parallel finite element method simulation is conducted with the procedures of: 1) partitioning mesh by using METIS, 2) starting the parallel finite element program, 3) reading subdomain mesh or partition generated in step 1 in the parallel manner such that the memory usage of the mesh is distributed to computer nodes, 4) parallel assembling of the matrix and the vector of the subdomain, 5) assigning boundary conditions locally, and 6) solving the linear equations with the solver provided in PETSc package. For the domain discretization, we use the mesh generated by the University of Stuttgart group (cf. Fig. 4, left) which is based on the GFZ work group hexahedron grid. The mesh consists of 4,043,119 tetrahedral elements and 707,713 nodes. Porosity and permeability data interpolated to that mesh were provided by the University of Stuttgart work group.

The hydraulic steady state of pressure is used as the initial condition. CO2 injection into the Stuttgart Formation is simulated with a maximum time step size of one day for 1,596 days. In the OGS model, the injection well is represented by a straight line treated as a Neumann boundary condition, where the injected CO2 mass is uniformly distributed. To account for CO2 migration and track the injection pressure history, the diagonal entries of the permeability tensor are multiplied with different sets of constant factors. By applying a global permeability scaling factor of 0.09, the simulated injection pressure variation at a specified observation point is close to the observed data as plotted in Fig. 7.

Fig. 7. Downhole pressure at a specified observation point plotted against the data observed at the CO2 Ktzi 201/2007 well

426 Thomas Kempka et al. / Energy Procedia 40 ( 2013 ) 418 – 427

4. Conclusions

A revision of the static geological model of the Stuttgart Formation at the Ketzin pilot site was carried out in the present study based on a re-interpretation of the available 3D seismic data and integration of the fault system present at the top of the Ketzin anticline. Matching the recorded pressure data at the injection well and CO2 arrival times at both observation wells of the Ketzin pilot site requires the introduction of distinct near-well and far-field permeability tensors. Taking this into account TOUGH2, ECLIPSE and DuMuX simulation results show a good to excellent agreement of simulated and observed pressures. Regarding the arrival times, the model reacts very sensitive to geological features which can change locally severely by using different mesh resolution. Thus, a matching of the arrival time, i.e. a point information, is associated with huge uncertainty regarding the underlying geological model. DuMuX simulations demonstrate that application of tetrahedral elements can make it impossible to maintain the porosity and permeability distribution given by the initial hexahedral grid. OpenGeoSys simulations again emphasize the required introduction of near- and far-field, since the simulations carried out demonstrate that a reasonable pressure match may not be achieved by application of a single permeability tensor.

Our intercomparison study demonstrates that a further revision of the static geological model of the Stuttgart Formation at the Ketzin pilot site is required in addition to an assessment of uncertainty of the observed arrival times and permeability anisotropy. As a consequence, we aim to integrate the results of a new tracer test started in January 2013 into our numerical models as these become available, since the obviously present permeability anisotropy in the Stuttgart Formation, which is also supported by the 3D seismic interpretation, may have a relevant impact on predictive simulations of CO2 plume migration and reservoir pressure development. In addition to that, further efforts are currently invested into an integrated analysis of 3D seismic data recorded and dynamic flow simulation results.

Acknowledgements

We appreciate the financial support within the scope of the CO2MAN project (GEOTECHNOLOGIEN R&D program) funded by the German Ministry of Education and Research (grant 03G0760) and the industrial partners VNG, Vattenfall, RWE, Statoil, Dillinger Hüttenwerke, Saarstahl and OMV. This is publication no. GEOTECH-2102.

References

[1] Martens S, Kempka T, Liebscher A, Lüth S, Möller F, Myrttinen A et al. -operating on-shore CO2 storage site at Ketzin, Germany: a progress report after three years of injection. Environ. Earth Sci. 2012;67(2):323-334, doi: 10.1007/s12665-012-1672-5.

[2] Würdemann H, Möller F, Kühn M, Heidug W, Christensen NP, Borm G et al. CO2SINK From site characterisation and risk assessment to monitoring and verification: One year of operational experience with the field laboratory for CO2 storage at Ketzin, Germany. Int J Greenh Gas Control 2010;4(6):938-951.

[3] Förster A, Schöner R, Förster HJ, Norden B, Blaschke AW, Luckert J et al. Reservoir characterization of a CO2 storage aquifer: The Upper Triassic Stuttgart Formation in the Northeast German Basin. Marine and Petroleum Geology 2010;27(10):2156-2172.

Thomas Kempka et al. / Energy Procedia 40 ( 2013 ) 418 – 427 427

[4] Norden B, Frykman P. Geological modelling of the Triassic Stuttgart Formation at the Ketzin CO2 storage site, Germany. Int J Greenh Gas Control 2013 (in press), doi:10.1016/j.ijggc.2013.04.019.

[5] Beutler G, Häusser I. Über den Schilfsandstein in der DDR. Zeitschrift für Geologische Wissenschaften 1982;10:511 525 (in German).

[6] Aigner T, Bachmann GH. Sequence-stratigraphic framework of the German Triassic. Sedimentary Geology 1992;80:115135.

[7] Shukla UK, Bachmann GH, Singh IB. Facies architecture of the Stuttgart Formation (Schilfsandstein, Upper Triassic), central Germany, and its comparison with modern Ganga system, India. Palaeography, Palaeoclimatology, Palaeoecology 2010;297:110 128.

[8] Ricken W, Aigner T, Jacobsen B. Levee-crevasse deposits from the German Schilfsandstein. Neues Jahrbuch für Geologie und Paläontologie Mh. 1988;77 94.

[9] Schlumberger.Petrel Seismic-to-Evaluation Software, 2011, Version 2011. [10] Kazemeini SH, Juhlin C, Zinck-Jørgensen K, Norden B. Application of the continuous wavelet transform on seismic data

for mapping of channel deposits and gas detection at the CO2SINK site Ketzin, Germany. Geophysical Prospecting 2009;57(1):111123.

[11] Juhlin C, Bergmann P, Giese R, Götz J, Ivanova A, Juhojuntti N et al. Preliminary results from 3D repeat seismics at the CO2SINK injection site, Ketzin, Germany. 72nd EAGE Conference & Exhibition incorporating SPE EUROPEC 2010, Barcelona, Spain, 14 17 June 2010, P201.

[12] Lüth S, Bergmann P, Cosma C, Enescu N, Giese R, Götz J et al. Time-lapse seismic surface and down-hole measurements for monitoring CO2 storage in the CO2SINK project (Ketzin, Germany), Energy Procedia 2011;4:3435 3442.

[13] Norden B, Förster A, Vu-Hoang D, Marcelis F, Springer N, Le Nir I. Lithological and petrophysical core-log interpretation in CO2SINK, the European onshore research storage and verification project. SPE Reservoir Evaluation & Engineering 2010;179192.

[14] Wolfgramm M, Rauppach K, Seibt P. Reservoirgeologische Charakterisierung mesozoischer Sandsteine im Norddeutschen Becken auf Basis petrophysikalischer und petrographischer Daten. Z. geol. Wiss. 2008;4(5):249 266 (in German).

[15] Schlumberger. Eclipse Reservoir Engineering Software, 2009, Version 2009.2. [16] Pruess K ECO2N: A TOUGH2 Fluid Property Module for Mixtures of Water, NaCl, and CO2. Report LBNL-57952 2005,

Lawrence Berkeley National Laboratory, Berkeley, California. [17] Zhang K, Wu YS, Pruess K. -MP A Massively Parallel Version of the TOUGH2 Code. Report

LBNL-315E 2008, Earth Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California. [18] Kempka T, Kühn M, Class H, Frykman P, Kopp A, Nielsen CM et al. Modelling of CO2 arrival time at Ketzin Part I. Int J

Greenh Gas Control 2010;4(6):1007-1015, doi:10.1016/j.ijggc.2010.07.005. [19] Kempka T, Kühn M. Numerical simulations of CO2 arrival times and reservoir pressure coincide with observations from the

Ketzin pilot site, Germany. Environ. Earth Sci. 2013 (in press), doi:10.1007/s12665-013-2614-6. [20] Kempka T, Klein E, De Lucia M, Kühn M. Assessment of the contribution of CO2 trapping mechanisms at the Ketzin pilot

site (Germany) by coupled numerical modelling. Energy Procedia 2013 (in press), doi:10.1016/j.egypro.2013.06.460. [21] Klein E, De Lucia M, Kempka T, Kühn M. Evaluation of long-term mineral trapping at the Ketzin pilot site for CO2

storage: An integrative approach using geochemical modelling and reservoir simulation. Int J Greenh Gas Control 2013 (in press), doi:10.1016/j.ijggc.2013.05.014.

[22] Möller F, Liebscher A, Martens S, Schmidt-Hattenberger C, Kühn M. Yearly operational datasets of the CO2 storage pilot site Ketzin, Germany. Scientific Technical Report: Data 2012, 12/06, doi: 0.2312/GFZ.b103-12066 (online only).

[23] Flemisch B, Darcis M, Erbertseder K, Faigle B, Lauser A, Mosthaf K, et al. Dumux: DUNE for multi-phase, component, flow and transport in porous media. Advances in Water Resources 2011;34(9):1102-1112.

[24] Walter L. Uncertainty studies and risk assessment for CO2 storage in geological formations, Doctoral dissertation 2013, Universität Stuttgart, Germany.

[25] Kolditz O, Bauer S, Bilke L, Böttcher N, Delfs JO, Fischer T et al. OpenGeoSys: an open-source initiative for numerical simulation of thermo-hydro-mechanical/chemical (THM/C) processes in porous media, Environ. Earth Sci. 2012;67(2):589-599.