i n t e r n a t i o n a l j o u r n a l o f g r e e n h o u s e g a s c o n t r o l 1 ( 2 0 0 7 ) 6 9 – 7 4

‘‘The Schweinrich structure’’, a potential site for industrialscale CO2 storage and a test case for safety assessment inGermany

Eric Kreft a,*, Christian Bernstone b, Robert Meyer c, Franz May c, Rob Arts a, Arie Obdama,Rickard Svensson b, Sara Eriksson b, Pierre Durst d, Irina Gaus d, Bert van der Meer a,Cees Geel a

aTNO, PO Box 80015, 3508 TA Utrecht, The NetherlandsbVattenfall Utveckling AB, SE-162, 87 Stockholm, SwedencBGR, Stilleweg 2, 30655 Hannover, GermanydBRGM, F-45060 Orleans, France

a r t i c l e i n f o

Article history:

Received 31 July 2006

Received in revised form

20 December 2006

Accepted 22 December 2006

Published on line 9 March 2007

Keywords:

CO2 storage

Aquifer storage

Geological storage

Safety assessment

FEP analysis

a b s t r a c t

The identification of risks associated with the geological storage of CO2 requires methods

that can analyse and assess potential safety hazards. This paper evaluates how perfor-

mance assessment can be used as a method for assessing the impact of CO2 storage on

health, safety and the environment (HSE) with particular respect to potential future aquifer

storage in the anticlinal structure Schweinrich in Germany. The performance assessment

was conducted under the CO2STORE European Fifth Framework project as one of the four

cases on the aquifer storage of CO2. It is known as the Schwarze Pumpe case study.

Being a case study, it is restrictive from a feasibility study point of view—i.e., the

extended identification of the key safety factors where an actual CO2 storage project would

be considered for the Schweinrich structure. The study is based on data currently available,

gathered in prior surveys, and on the use of simplified models, with CO2 leakage levels from

natural analogues being the evaluation criteria. While the results should be interpreted as

provisional, they point out clearly which additional data should be gathered in relation to

the long-term storage performance in the event that the site warrants further investigation.

# 2007 Elsevier Ltd. All rights reserved.

avai lab le at www.sc iencedi rec t .com

journal homepage: www.e lsev ier .com/ locate / i jggc

1. Introduction

Assessing and managing the risks associated with the

geological storage of CO2 is a relatively new area of research;

there is no detailed knowledge base as a frame of reference.

This therefore shifts the focus to learning from studies

conducted for other substances whose risks are similar yet

contain differences. A thorough safety assessment method

called performance assessment has been adopted on this

* Corresponding author. Current address: EBN, 3511 DX, Utrecht, TheE-mail address: eric.kreft@ebn.nl (E. Kreft).

1750-5836/$ – see front matter # 2007 Elsevier Ltd. All rights reserveddoi:10.1016/S1750-5836(07)00009-6

basis. Natural gas storage teaches us that several HSE risks

associated with subsurface storage can be managed through

the strict regulation of site selection and well construction.

Thus, it is also important to establish guidelines, in addition to

performance assessment evaluation, for inclusion in a future

safety standard for CO2 storage.

The performance has been applied to the Schwarze Pumpe

case study. The Schwarze Pumpe plant is located in

Brandenburg (Niederlausitz) 150 km southeast of Berlin and

Netherlands.

.

i n t e r n a t i o n a l j o u r n a l o f g r e e n h o u s e g a s c o n t r o l 1 ( 2 0 0 7 ) 6 9 – 7 470

operated by Vattenfall Europe Generation. The potential

Schweinrich storage site is located in the north-eastern region

of Germany, about 100 km north-west of Berlin, at a depth of

approximately 1600 m. It was selected as the most suitable

candidate in the north-eastern German basin for the under-

ground storage of 400 million tonnes of CO2, which corre-

sponds to 40 years’ production from a 1600 MW lignite-fired

power plant.

2. The geology of the structure

The Schweinrich structure covers an area of about 100 km2

and its estimated storage potential is between 500 and

840 tonnes of CO2 (Meyer et al., 2006). Its anticlinal structure

(Fig. 1) contains two main reservoirs: the shallower one in the

Lower Jurassic (Lias, Hettang) and the deeper one in the Upper

Triassic (Keuper, Contorta). The total reservoir thickness

ranges between 270 and 380 m and consists of several layers

of fine-grained, highly porous sandstone, which is overlain by

several hundred meters of thick Jurassic clay formations that

cap the storage system.

The current geological model of Schweinrich is based on

the information available from 2D seismic lines and from

exploration wells near the structure, mostly recorded and

drilled in the early 70s. No wells penetrate the anticline.

Refinement of the geological model combined with new

data as it is gathered will continue throughout the

performance assessment process. The Schweinrich struc-

ture has been used as a representative generic model that is

also valid for other potential storage sites in north-east

Germany.

3. Methodology

The purpose of the performance assessment study has been to

evaluate its suitability as a method for determining the Health,

Safety and Environmental (HSE) effects of CO2 storage. These

effects have also been evaluated from a feasibility point of

view, thereby identifying and evaluating the key safety factors

at an early stage (for further examination in a follow-up

project).

Fig. 1 – Cross-section of the Schweinrich anticline between two

storage position.

The methodology involves the:

� D

s

efinition of a basis for the assessment;

� a

nalysis of features, events and processes (FEPs)

� f

ormation of safety scenarios;

� d

evelopment of dedicated models for probabilistic simula-

tion of safety scenarios;

� e

valuation of HSE effects on safety.

The basis of the methodology is a comprehensive evalua-

tion of potential safety factors (FEPs) that may affect the future

performance of the storage site. A large number of FEPs are

evaluated and the most relevant and critical regarding safety

are selected for further evaluation. The FEPs are the building

blocks for the construction of safety scenarios, which are

simulated by numerical models. The long-term storage

performance is evaluated using probabilistic simulation in

order to cover the uncertainty related to the future impact of

the safety factors. This study compares the results from the

simulation models against the CO2 leakage levels from natural

analogues (e.g., reported in Streit and Watson, 2004).

The FEP analysis of the Schweinrich structure evaluates

potential HSE factors within the next 1000 years following CO2

injection. However, since the safety factors that are identified

may generate hazards, the simulations are run for additional

9000 years. The outcome of the safety scenarios was expressed

as the maximum concentration and maximum flux of CO2 in

the pore system in the Pleistocene sediments in the shallow

subsurface and represented in the simulation models by the

topmost subsurface layer. Since no modelling of the flow and

fate of CO2 in the shallow groundwater compartment was

conducted, no outcome was given with respect to groundwater

deterioration and mobilisation of heavy metals. This has been

planned for the next phase of the performance assessment.

4. Identified safety scenarios

Four safety scenarios were identified through FEP combina-

tions:

1. R

a

eference: no failure of the containment zone occurs. This

scenario, considered to be the most likely, reflects the CO2

lt diapers. The hatched area indicates the reservoir and

Table 1 – Number of grid blocks of different Schweinrich simulation models

# Grid blocks X # Grid blocks Y # Grid blocks Z Time per run

Simplified radial model (2D) 25 16 22 s

Simplified Cartesian model (3D) 20 21 17 20 min

Deterministic 3D model 84 40 22 2 days

i n t e r n a t i o n a l j o u r n a l o f g r e e n h o u s e g a s c o n t r o l 1 ( 2 0 0 7 ) 6 9 – 7 4 71

injection process and the flow and fate of CO2 in the

reservoir after abandonment of the site.

2. L

eaking-seal: the leaking seal scenario reflects the CO2

injection process and the flow and fate of CO2 through the

cap rock due to geochemical deterioration. The reason for

the possible release of CO2 through the cap rock might be

due to small amounts of carbonates and thin marl layers in

the shale layers that form the cap rock.

3. L

eaking-fault: the leaking fault scenario reflects the flow and

fate of CO2 through a fault system running from the cap

rock to the shallow subsurface. The interpretation of the

existing seismic lines over the Schweinrich structure are

not conclusive due to the poor data quality, but the

existence of fault systems in the Mesozoic and Caenozoic

overburden cannot be ruled out (Fig. 1). At this moment, the

constituency of such a fault system and its permeability are

simply unknown and require additional data acquisition.

4. L

eaking-well: the leaking well scenario reflects the CO2

injection process and the flow and fate of CO2 along the

well trajectory due to several events and processes. The

drilling and completion schedule of future wells are

unknown. Therefore a ‘‘generic’’ abandoned old well

safety scenario based on a previous study was applied to

the FEP evaluation (Wildenborg et al., 2005). This scenario

was chosen mainly to evaluate the differences in outcome

with the above scenarios. It must be noted, that no

abandoned wells penetrate the Schweinrich structure

and that all possible precautions can still be taken for

future injection wells making the occurrence of leaking

wells highly unlikely.

5. Model development

The safety scenarios present the possible future flow and fate

of CO2 for 10,000 years after injection. The scenarios are

represented in simplified 2D and 3D models with stochasti-

cally varied input parameters using the multi-component flow

Fig. 2 – Development of the CO2 gas satura

simulator SIMED-II (Stevenson and Pinczewski, 1995). Because

Simed II does not allow mechanical and chemical processes to

be modelled, the mechanical and chemical safety factors that

apply to the identified safety scenarios are represented by

adjustments in hydrodynamic properties (Svensson et al.,

2005). The advantage of using simplified models is their

limited run time, which allows a large number of stochastic

input combinations to be modelled (Table 1). Simplifications

relate mainly to the limited number of grid cells and the

homogeneous layer properties. The simplified models have

been calibrated to a detailed, deterministic, finely scaled

model of the Schweinrich structure over an injection period of

40 years (Fig. 2) based on the following:

1. S

tio

imilar input properties: the stochastic models should have

similar input properties, such as porosity and (relative)

permeability. For the coarser stochastic models these

properties have been averaged on the basis of the detailed

deterministic model.

2. S

imilar pressure development: the development of the

pressure of the reservoir in time needs to be similar to

the development in the deterministic model over the first 40

years.

3. S

imilar CO2 spread/distribution: the spread of the CO2 front in

the stochastic models should be similar to the spread in the

deterministic model where the CO2 front has a lateral

spread of more than 2 km after 40 years.

Buoyancy is the main cause of CO2 rising. One thousand

simulations were carried out for each safety scenario, with the

variation of the stochastic input parameters, such as perme-

ability, constrained on the basis of related studies (Hilden-

brand et al., 2004; Schlomer and Krooss, 1997). An example of

the magnitude is given in Table 2 (Svensson et al., 2005). In the

event of uncertainty about input parameters that were not

varied stochastically, worst-case values were generally

selected. Moreover, neither the CO2 dissolution in the aqueous

phase nor capillary entry pressures were taken into account. A

n in time (3D deterministic model).

Table 2 – Stochastic input parameters of the leaking fault scenario

Parameter Units Type of distribution Low High

Reservoir horizontal permeability mD Uniform 50 1000

Distance between fault and well m Uniform 50 2500

Fault vert. permeability (clay) mD Uniform 10�3 10�1

Fault vert. permeability (silt) mD 100 � fault vert. perm. (clay) 10�1 101

Fault vert. permeability (sand) mD 10,000 � fault vert. perm (clay) 101 103

i n t e r n a t i o n a l j o u r n a l o f g r e e n h o u s e g a s c o n t r o l 1 ( 2 0 0 7 ) 6 9 – 7 472

few sensitivities were run, including CO2 dissolution, in order

to evaluate the effects on the outcome of the model. Changes

in flux and concentration with respect to the case without CO2

dissolution varied between 2% and 25% in flux and between 0%

and 7% in concentration for high and low release rates,

respectively. For these reasons the outcome expressed as the

maximum flux of CO2 in the shallow subsurface Pleistocene

sediments are biased towards the worst-case scenarios.

6. Simulation results

The reference scenario and the leaking-seal scenario show no

increase of CO2 in the Pleistocene sediments over 10,000 years.

The CO2 escaping from the seal is sufficiently held up and

spread over time that it does not reach the shallow subsurface

(Fig. 3).

The leaking-fault scenario, i.e. where it is assumed that a

fault extends from the cap rock to the shallow subsurface,

shows a relatively slow migration process of CO2 along the

fault plane. An example of one of the simplified model runs is

presented in Fig. 4. Maximum CO2 fluxes vary between 0.00025

and 0.62 tonnes/(year m2) in the Pleistocene sediments (Fig. 5).

These values are comparable to leakage rates from natural CO2

accumulations in Europe and Australia (Streit and Watson,

Fig. 3 – Cross-sections of the simplified 2D flow model presenti

The CO2 injection is positioned on the left-hand side of the sec

2004). The maximum CO2 gas concentration in the shallow

subsurface Pleistocene sediments is less than 4% at a depth of

80 m, which is close to the lower limit of moderate effects on

trees and crops (Saripalli et al., 2002). The effects of the fluxes

and concentrations on the shallow subsurface ecosystem will

be investigated in a later phase.

The ranges in outcome show that further research on the

existence of the faults through the cap rock is required. Such a

study would be a priority if actual CO2 storage project were to

be considered at Schweinrich. The location of the faults can be

investigated by running a 3D seismic survey, and the fault

properties determined by conducting a special study on the

local fault permeability. These fault property values would be

needed for detailed coupled THMC models in order to restrict

the range of fluxes obtained from the stochastic models. Note

that the simulation results should be interpreted as worst-

case scenarios, especially since the presence of faults cutting

the cap rock has not yet been established.

The leaking-well scenario (Fig. 6) is the most dramatic,

with average release percentages of 60% of the total amount

of injected CO2. The release of CO2 is directly proportional to

the permeability of the well zone, which increases in time as

a result of various FEPs that apply specifically to this

scenario. It should be noted is that the study is based on the

data of an existing abandoned old well and not the quality to

ng subsurface CO2 saturation in the leaking-seal scenario.

tions.

Fig. 4 – Quadrant of the simplified 3D flow model presenting subsurface CO2 saturation in the leaking-fault scenario. The CO2

injection is positioned in the lower left corner. Note that this scenario assumes that a permeable fault from the cap rock to

shallow subsurface is present, but this cannot be confirmed at this stage. More data (i.e., seismic data) is needed to explore

the extent of the fault system.

Fig. 5 – Simulated maximum fluxes and maximum concentrations in local groundwater for the leaking-fault scenario

(assuming there is a leaking fault). Results were acquired without modelling CO2 dissolution in the aqueous phase.

Fig. 6 – Cross-sections of the simplified 2D leaking-well scenario, presenting the subsurface CO2 saturation.

i n t e r n a t i o n a l j o u r n a l o f g r e e n h o u s e g a s c o n t r o l 1 ( 2 0 0 7 ) 6 9 – 7 4 73

i n t e r n a t i o n a l j o u r n a l o f g r e e n h o u s e g a s c o n t r o l 1 ( 2 0 0 7 ) 6 9 – 7 474

be expected from a purpose-designed abandoned CO2

injection well.

Maximum fluxes in the Pleistocene vary between

15 tonnes/(year m2) and 350 tonnes/(year m2). This flux is

about 1–10 times the fluxes measured at the Cava dei Selci

near Rome, where fatal accidents were caused by natural CO2

emissions (Carapezza et al., 2003). However, this is not a

realistic scenario for Schweinrich since new wells would be

better designed and high leakage rates in the well zone

detected early enabling mitigating actions to be taken to avoid

further leakage. Most probably, the injection wells would be

placed on the lower flanks of the Schweinrich structure. As

mentioned previously, this scenario was run mainly to

evaluate the differences in outcome with the scenarios above.

7. Conclusions

This first HSE performance assessment of the Schweinrich

structure was conducted on the basis of the available existing

and limited input data prior to commercial site exploration.

The reference scenario and some worst case scenarios have

been analysed using simplified models. The outcome is biased

towards worst-case scenarios because of the uncertainty

about the input parameters and the use of simplified models.

The results are provisional, given the ongoing data-gathering

process and refinement of the geological model.

The methodology evaluation shows that the performance

assessment methodology is a powerful tool for use in safety

assessments of CO2 storage projects, one that is able to

distinguish relevant safety scenarios at an early stage. The

performance assessment clearly reveals which additional data

should be gathered in respect of long-term storage perfor-

mance if further investigation of the site should be needed.

Acknowledgements

This study is part of the European CO2STORE project. The

authors thank the European Commission and the industrial

consortium partners for funding this research.

r e f e r e n c e s

Carapezza, M.L., Badalamenti, B., Cavarra, L., Scalzo, A., 2003.Gas hazard assessment in a densely inhabited area of ColliAlbani Volcano (Cava dei Selci, Roma). J. Volcanol.Geotherm. Res. 123, 81–94.

Hildenbrand, A., Schlomer, S., Krooss, B.M., Littke, R., 2004. GasBreakthrough Experiments on Pelitic Rocks: ComparativeStudy with N2, CO2 and CH4, Geofluids. Blackwell PublishingLtd, 4-6180.

Meyer, R., May, F., Krull, P., Durst, P., Gaus, I., Kreft, E., Meer,B.v.d., Geel, K., Svensson, R., Bernstone, C., 2006. Geologicalcharacterisation of the structure Schweinrich—a suitablecandidate site for industrial CO2 storage in Germany? In:Proceedings of the 8th International Conference onGreenhouse Gas Control Technologies, 19–22 June, 2006,Trondheim.

Saripalli, K.P., Cook, E.M., Mahasenan, N., 2002. Risk and hazardassessment for projects involving the geologicalsequestration of CO2. In: Gale, J., Kaya, Y. (Eds.), Proceedingsof the 6th International Conference on Greenhouse GasTechnologies, Kyoto, October 2002.

Schlomer, S., Krooss, B.M., 1997. Experimental Characterisationof the Hydrocarbon Sealing Efficiency of Cap Rocks, Marineand Petroleum Geology, vol. 14, no. 5. Elsevier Science Ltd.,pp. 565–580.

Stevenson, M., Pinczewski, V., 1995. ‘‘SIMED—MulticomponentCoalbed Gas Simulator’’, User’s Manual Version 1.21,Australian Petroleum Cooperative Research Centre ReportNo. 022.

Streit, J.E., Watson, M.N., 2004. Estimating Rates of Potential CO2

Loss from Geological Storage Sites for Risk and UncertaintyAnalysis, GHGT-7 2004, Vancouver.

Svensson, R., Bernstone, C., Eriksson, S., Kreft, E., Arts, R.,Obdam, A., Meyer, R., 2005. Safety assessment of structureSchweinrich – part of CO2STORE case study SchwarzePumpe – Internal Report CO2STORE, 93 pp.

Wildenborg, A.F.B., Leijnse, A.L., Kreft, E., Nepveu, M.N.,Obdam, A.N.M., Orlic, B., Wipfler, E.L., van der Grift, B., vanKesteren, W., Gaus, I., Czernichowsky-Lauriol, I., Torfs, P.,Wojcik, R., 2005. Risk assessment methodology for CO2

storage: the scenario approach. In: Benson, S.M. (Ed.),Carbon Dioxide Capture and Storage in Deep GeologicFormations—Results from the CO2 Capture Project, 2.Elsevier Ltd., pp. 1293–1316.