- Email: [email protected]
Monitoring of CO2 injected at Sleipner using time-lapse seismic data
Energy 29 (2004) 1383–1392 www.elsevier.com/locate/energy
Monitoring of CO2 injected at Sleipner using time-lapse seismic data R. Arts a,, O. Eiken b, A. Chadwick c, P. Zweigel d, L. van der Meer a, B. Zinszner e a
c
Netherlands Institute of Applied Geoscience TNO—National Geological Survey, PO Box 80015, 3508 TA Utrecht, Netherlands b Statoil, R&D Centre, NO-7005 Trondheim, Norway Britsh Geological Survey, Kingsley Dunham Centre, Keyworth, Nottingham, Nottinghamshire NG12 5GG, UK d SINTEF Petroleum Research, S.P. Andersens vei 15B, 7465 Trondheim, Norway e Institut Francais du Pe´trole, 1 et 4 Av. de Bois-Pre´au, F-92852, Rueil-Malmaison, France
Abstract Since October 1996, Statoil and its Sleipner partners have injected CO2 into a saline aquifer, the Utsira Sand, at a depth of approximately 1000 m. The aquifer has a thickness of more than 200 m near the injection site and is sealed by thick shales. A multi-institutional research project SACS (Saline Aquifer CO2 Storage) was formed to predict and monitor the migration of the injected CO2. To this end two time-lapse seismic surveys over the injection area have been acquired, one in October 1999, after 2.35 million tonnes of CO2 had been injected, and the second in October 2001, after approximately 4.26 million tonnes of CO2 had been injected. Comparison with the baseline seismic survey of 1994 prior to injection provides insights into the migration of the CO2. In this paper the results of the seismic interpretation will be shown, supported by synthetic seismic modelling and reservoir flow simulation of the migrating CO2 at the two different time-steps. # 2004 Elsevier Ltd. All rights reserved.
1. Introduction CO2 is injected near the base of the Utsira Sand at a depth of 1012 m below sea level. With the injected CO2 in a supercritical or liquid state the main mechanism driving the flow of the CO2 is gravity, the CO2 rising buoyantly until it reaches a shale layer [1,2]. Beneath each intrareservoir shale the CO2 accumulates with high saturations which follow the structural relief.
Corresponding author. Tel.: +31-30-256-4638; fax: +31-30-256-4605. E-mail address: [email protected] (R. Arts).
0360-5442/$ - see front matter # 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.energy.2004.03.072
1384
R. Arts et al. / Energy 29 (2004) 1383–1392
The overall effect of the CO2 on the seismic signal is significant [3]. By 1999 the CO2 appeared to have reached the top of the reservoir. At several depth levels within the Utsira Sand a large increase in reflectivity has been observed on the time-lapse seismic data caused by individual CO2 accumulations under thin, intra-reservoir, shale layers. The presence of these thin shale layers acting as (at least temporary) boundaries has been demonstrated on well data, but their effect on the 1994 (pre-injection) seismic signal is too small for a reliable lateral interpretation [3]. With CO2 captured underneath, the location of these shale layers can be identified on the seismic data as amplitude anomalies, despite the thicknesses of the accumulations being below what is generally considered to be the limit of seismic resolution. This is mainly caused by the high compressibility of the CO2 and by the constructive tuning effects of the top and bottom reflections at the CO2 accumulations. The effect of the density is less important since the CO2 is in a dense-fluid supercritical state at the reservoir conditions and not in a gaseous phase. The thicknesses of the accumulations can be estimated quantitatively using the seismic amplitude information and assuming a tuning relationship [4,5]. The seismic interpretation has been supported by synthetic seismic modelling in order to understand the seismic response below the thin intra-reservoir shale layers. Beneath the CO2 plume, a ‘‘velocity push-down effect’’ can be observed on the seismic data due to seismic waves travelling more slowly through CO2-saturated rock than through watersaturated rock [4,5]. In the first section of this paper the effect of CO2 in the Utsira Sand on seismic velocities is explained using the so-called Gassmann model [6,7]. The consequences for the seismic data are illustrated on synthetic seismic models. The results of the seismic interpretation of the CO2 accumulations in 1999 and 2001 are discussed. The seismic interpretation has been used to construct a reservoir flow simulation model. With this model the injection process has been simulated and the distribution of the CO2 at the times of the monitoring seismic surveys has been ‘predicted’. These distributions have been used to generate synthetic seismics that are compared directly to the seismic field data. The satisfactory match increases our confidence in a correct understanding of the fluid flow in the reservoir. 2. Gassmann modelling Seismic velocities were modelled as a function of CO2 saturation using the Gassmann relationships [6] which enable the elastic properties of a porous medium saturated with a fluid to be derived from the known properties of the same medium saturated with a different fluid. A very practical overview of the use and implementation of Gassmann’s theory is given in Ref. [7]. As a basic assumption in this analysis, homogeneous mixtures of brine and CO2 with respect to the seismic wavelength are assumed. In reality this condition is likely to be only approximately fulfilled. The densities and compressibilities of the saturating fluids, the rock matrix and the porosity of the rock are assumed to be known (Fig. 1). In the Sleipner case the 100% water-saturated P- and S-wave velocities are known from well logs. The main constituent of the rock matrix is quartz with a known density and compressibility. The natural pressure and temperature at the injection point are estimated from regional v well data measurements at 10 MPa and 36 C, respectively. Under these conditions CO2 is in a supercritical state. In practice this means that it has the density of a fluid, but the compressibility of a gas. The most likely values for the density of the CO2 are between 600 and 700 kg m3.
R. Arts et al. / Energy 29 (2004) 1383–1392
1385
Fig. 1. P- and S-wave velocities of the Utsira Sand as a function of water–CO2 saturation using Gassmann’s model.
The bulk modulus K of the CO2 is considered likely to have a value not greater than 0.0675 GPa under these conditions. This implies generally a fairly constant P-wave velocity < 1450 m s1 for the Utsira Sand for CO2 saturations in the range of 20–100% compared to a P-wave velocity of 2050 m s1 for fully water-saturated Utsira sand. Fig. 1 shows the modelling results for the velocities as a function of water–CO2 saturation for three different bulk moduli. The relevant parameters for the Gassmann calculations are reported on the side. 3. Seismic modelling In order to perform seismic modelling, a (zero-phase) wavelet was estimated from the seismic data [8]. Using the estimated elastic parameters for the 100% water-saturated sandstone and for the (100%) CO2-saturated sandstone, a simplified impedance model was created in order to predict the seismic response of the injected CO2. The elastic properties of the shale layers have been estimated from sonic and density logs (Vp ¼ 2270 m s1, Vs ¼ 850 m s1 and q ¼ 2100 kg m3). Fig. 2 shows the synthetic seismic response for CO2 accumulating under a thin intra-reservoir shale layer. A range of thicknesses of the CO2 plume was modelled in order to investigate the influence on the seismic signal. Two dominant effects determine the seismic response: . The negative seismic impedance contrast between the shale and the sandstone becomes more negative (larger in absolute value) when CO2 is present. . The seismic response is a composite wavelet caused by interference from sequences of watersaturated sand and shale, CO2 saturated sand and water-saturated sand again. The first effect leads to stronger negative seismic amplitudes as for a classical ‘‘bright spot’’. The second effect (tuning) can lead to destructive or constructive interference, depending on the thickness of the CO2 layer, as evident from the seismic modelling. As the thickness of the CO2
1386
R. Arts et al. / Energy 29 (2004) 1383–1392
Fig. 2. Simplified model of a variable thickness (0–8 m) of CO2 beneath a thin shale layer (2 m) and the corresponding synthetic seismic response.
column increases, a gradual increase of the (negative) amplitude is observed. Maximum constructive interference corresponds to a CO2 thickness of about 8 m, the so-called ‘tuning thickness’. 4. Seismic interpretation As expected from seismic modelling, introducing CO2 into the Utsira Sand has a dramatic effect on the reflectivity. This is illustrated in Fig. 3 showing the seismic inline (of the 1994, 1999 and 2001 survey) through the injection point. A number of strong negative reflections (black
Fig. 3. An inline through the injection area for the 1994, 1999 and the 2001 surveys.
R. Arts et al. / Energy 29 (2004) 1383–1392
1387
peaks) are observed both on the 1999- and the 2001 time-lapse surveys. The consistency between the CO2 levels of both vintages is striking. In general, the 2001 CO2 levels have a larger lateral extent and have been ‘‘pushed down’’ slightly more with respect to the 1999 CO2 levels. This can be easily explained considering that more injected CO2 causes more pushdown. A more detailed analysis on the interpretation can be found in Refs. [5,8]. By 1999 the CO2 had reached the top of the reservoir. Nine separate stratigraphic levels with CO2 accumulations below shale layers have been interpreted, with the deepest level referred to as level 1 and the shallowest as level 9 [8]. Note that all but one of the reflections of the thin shale layers on the 1994 baseline data without CO2 present are below the limit of seismic resolution. Only with CO2 underneath them the seismic signal is ‘amplified’ above seismic resolution and the shale layers become detectable. The tuning effect as described in the previous section is illustrated in Fig. 4. For this figure, locations have been selected in the 1999 and 2001 survey, where only a single shale layer with CO2 trapped underneath is present. As expected these locations are concentrated towards the
Fig. 4. Demonstration of the tuning relation derived from synthetic seismics compared to the observed relation in the seismic data of 1999 and 2001 originating from locations where only a single shale layer is present.
1388
R. Arts et al. / Energy 29 (2004) 1383–1392
outer limits of the total CO2 accumulations. Note that the only criterion is the presence of a single shale layer, but not necessarily the same layer everywhere. For these selected locations the ‘‘pushdown’’ in time (which for a single layer is linearly related to the thickness of the CO2 accumulation) has been plotted against the seismic reflection amplitude. The data appear to follow a tuning relation as expected from the synthetic seismic modelling. A prominent vertical feature, which can be clearly distinguished in Fig. 3, is characterized by localized pushdown and much decreased reflection amplitudes. This is interpreted as a ‘‘chimney’’ of CO2, situated approximately above the injection point and forming a major vertical migration path which conducts CO2 almost directly to the top of the reservoir. Note that Fig. 4 suffers from some ‘‘large pushdown, low amplitude’’ anomalies not consistent with the tuning relation. These anomalies correspond to interpretations within the chimney where most of the seismic amplitude information is destroyed. 5. Reservoir flow simulation In order to help understand the CO2 flow in the reservoir and to predict any future flow, a full-field 3D reservoir flow simulation model has been constructed based on the seismic interpretation [2]. This was not a straightforward task, taking into account that the intra-reservoir shale layers could only be interpreted on the time-lapse surveys with CO2 present below them. Moreover, these interpretations were in two–way traveltime and had to be time-depth converted for the flow simulation model. With the large velocity variations occurring in the reservoir due to the presence of CO2, an accurate time–depth conversion is not simple and an
Fig. 5. Result of a reservoir flow simulation at the time of the first time-lapse seismic survey in October 1999 (after 3 years of injection) after 2.35 million tonnes of CO2 injected [2].
R. Arts et al. / Energy 29 (2004) 1383–1392
1389
accurate depth model of each individual shale layer has not been obtained so far. To circumvent this problem, these intra-shale layers have been assumed parallel to the upper shale layer just below the top of the reservoir. The total simulation flow model covers an area of 15.7 km2. The Utsira Sand was represented by a total of 59 layers. The shale layers were represented in the model by transmissibility modifiers between the appropriate layers. The result of the simulation after 3 years of injection, at the time of the first time-lapse seismic survey in October 1999, is shown in Fig. 5. The resulting saturation model has been transformed into a 3D acoustic impedance model using the Gassmann relations in order to perform a forward seismic modelling. A cross-section, representing crossline 3123 in 2001, is shown in Fig. 6. The leftmost image represents the acoustic impedance model in depth, the middle image is the synthetic seismic data and the rightmost image corresponds to the seismic field data both in two-way traveltime. Taking into account the assumptions made to construct the model, the synthetic seismic data mimic the field data fairly well. The pushdown effect is clearly demonstrated on the synthetic seismic data. Fig. 7 shows the lateral extent of the pushdown anomalies for the 1999 seismic data, for the 2001 seismic data and for the synthetic data (based on the simulation for 2001). The field data show a more north outh elongated shape, whereas the synthetic data follow the domal shape of the upper shale layer (level 8) topography.
Fig. 6. Cross-section from west to east of the acoustic impedance model derived from the 2001 reservoir simulation model in depth (left), the synthetic seismic data (middle) and the corresponding 2001 seismic data both in two-way traveltime.
1390
R. Arts et al. / Energy 29 (2004) 1383–1392
Fig. 7. Structure top of the upper shale layer (level 8) with superimposed outlines of the pushdown anomalies as observed in the 1999 dataset (light grey), in the 2001 dataset (dark grey) and in the 2001 synthetic seismic data (black).
This effect can also be observed on the synthetic data shown in Fig. 6. The lateral extent of the CO2 anomaly on the east–west synthetic data of crossline 3123 is larger than on the field data.
6. Conclusion Time-lapse seismic surveying appears a highly suitable geophysical technique for monitoring CO2 injection into a saline aquifer. The effects of the CO2 on the seismic data are large both in terms of seismic amplitude and in observed velocity pushdown effects. In addition to straightforward mapping of changes in the time-lapse seismic data, quantitative approaches to mapping the supercritical CO2 saturations have been carried out, supported by seismic modelling. A more detailed analysis can be found in Refs. [4,5].
R. Arts et al. / Energy 29 (2004) 1383–1392
1391
The interpreted seismic data have been used for a full 3D simulation model. The construction of such a model was problematical for two main reasons: . The intra-reservoir shale layers could only be interpreted on the time-lapse surveys when CO2 was present below them. . The conversion from time to depth of these intra-reservoir shale layers is not sufficiently accurate due to the large velocity variations caused by the presence of CO2. As a consequence, an assumption on the depth structure of the intra-reservoir shale layers has been made, assuming them to be parallel to the upper shale layer, the only one that has been interpreted on the baseline seismic data. The saturation model resulting form the reservoir simulation has been transformed into an acoustic impedance model using the Gassmann model, which in turn has been used to create synthetic seismics. Taking into account the assumptions made to construct the model, the synthetic seismic data match the field data fairly well. Acknowledgements The SACS project was funded by the European Community (Thermie program), by the industry partners, i.e.: Statoil, BP, Norsk Hydro, ExxonMobil, TotalFinaElf and Vattenfall and by national governments. The R&D partners are TNO-NITG (TNO-Netherlands Institute of Applied Geoscience—National Geological Survey), SINTEF Petroleum Research, BGS (British Geological Survey), BRGM (Bureau de Recherches Ge´ologiques et Minie`res), GEUS (Geological Survey of Denmark and Greenland) and IFP (Institut Francais du Pe´trole). The IEA-Greenhouse Gas Programme is involved in the dissemination of the results. References [1] Lindeberg E, Zweigel P, Bergmo P, Ghaderi A, Lothe A. Prediction of CO2 dispersal pattern improved by geology and reservoir simulation and verified by time lapse seismic. In: Williams DJ, et al., editors. Greenhouse gas control technologies. Collingwood, Australia: CSIRO Publishing; 2001, p. 372–7. [2] Meer LGHvander, Arts RJ, Peterson L. Prediction of migration of CO2 injected into a saline aquifer: reservoir history matching to a 4D seismic image with a compositional gas/water model. In: Williams DJ, et al., editors. Greenhouse gas control technologies. Collingwood, Australia: CSIRO Publishing; 2001, p. 378–84. [3] Eiken O, Brevik I, Arts R, Lindeberg E, Fagervik K. Seismic monitoring of CO2 injected into a marine aquifer. Society of Exploration Geophysicists (SEG), Calgary 2000 International Conference and 70th Annual Meeting, Calgary, Paper RC-8.2. 2000, p. 4. [4] Arts R, Elsayed R, van der Meer L, Eiken O, Ostmo S, Chadwick A, et al. Estimation of the mass of injected CO2 at Sleipner using time-lapse seismic data. 64th European Association of Geoscientists and Engineers (EAGE) Meeting, Florence, Paper H016. 2002, p. 4. [5] Chadwick RA, Arts R, Eiken O, Kirby GA, Lindeberg E, Zweigel P. 4D seismic imaging of an injected CO2 plume at the Seipner Field, central North Sea. In: Davies R, Cartwright J, Stewart S, Underhill J, Lappin M, editors. 3D seismic data: application to the exploration of sedimentary basins, vol. 29. London: Memoir of the Geological Society; 2004, p. 305–14. [6] Gassmann F. Uber die Elastizitat poroser Medien. Vier, der Natur, Gesellschaft in Zurich 1951;96:1–23 in German.
1392
R. Arts et al. / Energy 29 (2004) 1383–1392
[7] Mavko G, Mukerji T, Dvorkin J. The rock physics handbook. Cambridge: Cambridge University press; 1998 pp. 168-170. [8] Arts R, Eiken O, Chadwick A, Zweigel P, van der Meer L, Kirby G, Seismic monitoring at the Sleipner underground CO2 storage site (North Sea). In: Baines SJ, Worden RH, editors Geological storage of CO2 for emissions reduction. Geological Society, London, Special Publication [in press].
Monitoring of CO2 injected at Sleipner using time-lapse seismic data R. Arts a,, O. Eiken b, A. Chadwick c, P. Zweigel d, L. van der Meer a, B. Zinszner e a
c
Netherlands Institute of Applied Geoscience TNO—National Geological Survey, PO Box 80015, 3508 TA Utrecht, Netherlands b Statoil, R&D Centre, NO-7005 Trondheim, Norway Britsh Geological Survey, Kingsley Dunham Centre, Keyworth, Nottingham, Nottinghamshire NG12 5GG, UK d SINTEF Petroleum Research, S.P. Andersens vei 15B, 7465 Trondheim, Norway e Institut Francais du Pe´trole, 1 et 4 Av. de Bois-Pre´au, F-92852, Rueil-Malmaison, France
Abstract Since October 1996, Statoil and its Sleipner partners have injected CO2 into a saline aquifer, the Utsira Sand, at a depth of approximately 1000 m. The aquifer has a thickness of more than 200 m near the injection site and is sealed by thick shales. A multi-institutional research project SACS (Saline Aquifer CO2 Storage) was formed to predict and monitor the migration of the injected CO2. To this end two time-lapse seismic surveys over the injection area have been acquired, one in October 1999, after 2.35 million tonnes of CO2 had been injected, and the second in October 2001, after approximately 4.26 million tonnes of CO2 had been injected. Comparison with the baseline seismic survey of 1994 prior to injection provides insights into the migration of the CO2. In this paper the results of the seismic interpretation will be shown, supported by synthetic seismic modelling and reservoir flow simulation of the migrating CO2 at the two different time-steps. # 2004 Elsevier Ltd. All rights reserved.
1. Introduction CO2 is injected near the base of the Utsira Sand at a depth of 1012 m below sea level. With the injected CO2 in a supercritical or liquid state the main mechanism driving the flow of the CO2 is gravity, the CO2 rising buoyantly until it reaches a shale layer [1,2]. Beneath each intrareservoir shale the CO2 accumulates with high saturations which follow the structural relief.
Corresponding author. Tel.: +31-30-256-4638; fax: +31-30-256-4605. E-mail address: [email protected] (R. Arts).
0360-5442/$ - see front matter # 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.energy.2004.03.072
1384
R. Arts et al. / Energy 29 (2004) 1383–1392
The overall effect of the CO2 on the seismic signal is significant [3]. By 1999 the CO2 appeared to have reached the top of the reservoir. At several depth levels within the Utsira Sand a large increase in reflectivity has been observed on the time-lapse seismic data caused by individual CO2 accumulations under thin, intra-reservoir, shale layers. The presence of these thin shale layers acting as (at least temporary) boundaries has been demonstrated on well data, but their effect on the 1994 (pre-injection) seismic signal is too small for a reliable lateral interpretation [3]. With CO2 captured underneath, the location of these shale layers can be identified on the seismic data as amplitude anomalies, despite the thicknesses of the accumulations being below what is generally considered to be the limit of seismic resolution. This is mainly caused by the high compressibility of the CO2 and by the constructive tuning effects of the top and bottom reflections at the CO2 accumulations. The effect of the density is less important since the CO2 is in a dense-fluid supercritical state at the reservoir conditions and not in a gaseous phase. The thicknesses of the accumulations can be estimated quantitatively using the seismic amplitude information and assuming a tuning relationship [4,5]. The seismic interpretation has been supported by synthetic seismic modelling in order to understand the seismic response below the thin intra-reservoir shale layers. Beneath the CO2 plume, a ‘‘velocity push-down effect’’ can be observed on the seismic data due to seismic waves travelling more slowly through CO2-saturated rock than through watersaturated rock [4,5]. In the first section of this paper the effect of CO2 in the Utsira Sand on seismic velocities is explained using the so-called Gassmann model [6,7]. The consequences for the seismic data are illustrated on synthetic seismic models. The results of the seismic interpretation of the CO2 accumulations in 1999 and 2001 are discussed. The seismic interpretation has been used to construct a reservoir flow simulation model. With this model the injection process has been simulated and the distribution of the CO2 at the times of the monitoring seismic surveys has been ‘predicted’. These distributions have been used to generate synthetic seismics that are compared directly to the seismic field data. The satisfactory match increases our confidence in a correct understanding of the fluid flow in the reservoir. 2. Gassmann modelling Seismic velocities were modelled as a function of CO2 saturation using the Gassmann relationships [6] which enable the elastic properties of a porous medium saturated with a fluid to be derived from the known properties of the same medium saturated with a different fluid. A very practical overview of the use and implementation of Gassmann’s theory is given in Ref. [7]. As a basic assumption in this analysis, homogeneous mixtures of brine and CO2 with respect to the seismic wavelength are assumed. In reality this condition is likely to be only approximately fulfilled. The densities and compressibilities of the saturating fluids, the rock matrix and the porosity of the rock are assumed to be known (Fig. 1). In the Sleipner case the 100% water-saturated P- and S-wave velocities are known from well logs. The main constituent of the rock matrix is quartz with a known density and compressibility. The natural pressure and temperature at the injection point are estimated from regional v well data measurements at 10 MPa and 36 C, respectively. Under these conditions CO2 is in a supercritical state. In practice this means that it has the density of a fluid, but the compressibility of a gas. The most likely values for the density of the CO2 are between 600 and 700 kg m3.
R. Arts et al. / Energy 29 (2004) 1383–1392
1385
Fig. 1. P- and S-wave velocities of the Utsira Sand as a function of water–CO2 saturation using Gassmann’s model.
The bulk modulus K of the CO2 is considered likely to have a value not greater than 0.0675 GPa under these conditions. This implies generally a fairly constant P-wave velocity < 1450 m s1 for the Utsira Sand for CO2 saturations in the range of 20–100% compared to a P-wave velocity of 2050 m s1 for fully water-saturated Utsira sand. Fig. 1 shows the modelling results for the velocities as a function of water–CO2 saturation for three different bulk moduli. The relevant parameters for the Gassmann calculations are reported on the side. 3. Seismic modelling In order to perform seismic modelling, a (zero-phase) wavelet was estimated from the seismic data [8]. Using the estimated elastic parameters for the 100% water-saturated sandstone and for the (100%) CO2-saturated sandstone, a simplified impedance model was created in order to predict the seismic response of the injected CO2. The elastic properties of the shale layers have been estimated from sonic and density logs (Vp ¼ 2270 m s1, Vs ¼ 850 m s1 and q ¼ 2100 kg m3). Fig. 2 shows the synthetic seismic response for CO2 accumulating under a thin intra-reservoir shale layer. A range of thicknesses of the CO2 plume was modelled in order to investigate the influence on the seismic signal. Two dominant effects determine the seismic response: . The negative seismic impedance contrast between the shale and the sandstone becomes more negative (larger in absolute value) when CO2 is present. . The seismic response is a composite wavelet caused by interference from sequences of watersaturated sand and shale, CO2 saturated sand and water-saturated sand again. The first effect leads to stronger negative seismic amplitudes as for a classical ‘‘bright spot’’. The second effect (tuning) can lead to destructive or constructive interference, depending on the thickness of the CO2 layer, as evident from the seismic modelling. As the thickness of the CO2
1386
R. Arts et al. / Energy 29 (2004) 1383–1392
Fig. 2. Simplified model of a variable thickness (0–8 m) of CO2 beneath a thin shale layer (2 m) and the corresponding synthetic seismic response.
column increases, a gradual increase of the (negative) amplitude is observed. Maximum constructive interference corresponds to a CO2 thickness of about 8 m, the so-called ‘tuning thickness’. 4. Seismic interpretation As expected from seismic modelling, introducing CO2 into the Utsira Sand has a dramatic effect on the reflectivity. This is illustrated in Fig. 3 showing the seismic inline (of the 1994, 1999 and 2001 survey) through the injection point. A number of strong negative reflections (black
Fig. 3. An inline through the injection area for the 1994, 1999 and the 2001 surveys.
R. Arts et al. / Energy 29 (2004) 1383–1392
1387
peaks) are observed both on the 1999- and the 2001 time-lapse surveys. The consistency between the CO2 levels of both vintages is striking. In general, the 2001 CO2 levels have a larger lateral extent and have been ‘‘pushed down’’ slightly more with respect to the 1999 CO2 levels. This can be easily explained considering that more injected CO2 causes more pushdown. A more detailed analysis on the interpretation can be found in Refs. [5,8]. By 1999 the CO2 had reached the top of the reservoir. Nine separate stratigraphic levels with CO2 accumulations below shale layers have been interpreted, with the deepest level referred to as level 1 and the shallowest as level 9 [8]. Note that all but one of the reflections of the thin shale layers on the 1994 baseline data without CO2 present are below the limit of seismic resolution. Only with CO2 underneath them the seismic signal is ‘amplified’ above seismic resolution and the shale layers become detectable. The tuning effect as described in the previous section is illustrated in Fig. 4. For this figure, locations have been selected in the 1999 and 2001 survey, where only a single shale layer with CO2 trapped underneath is present. As expected these locations are concentrated towards the
Fig. 4. Demonstration of the tuning relation derived from synthetic seismics compared to the observed relation in the seismic data of 1999 and 2001 originating from locations where only a single shale layer is present.
1388
R. Arts et al. / Energy 29 (2004) 1383–1392
outer limits of the total CO2 accumulations. Note that the only criterion is the presence of a single shale layer, but not necessarily the same layer everywhere. For these selected locations the ‘‘pushdown’’ in time (which for a single layer is linearly related to the thickness of the CO2 accumulation) has been plotted against the seismic reflection amplitude. The data appear to follow a tuning relation as expected from the synthetic seismic modelling. A prominent vertical feature, which can be clearly distinguished in Fig. 3, is characterized by localized pushdown and much decreased reflection amplitudes. This is interpreted as a ‘‘chimney’’ of CO2, situated approximately above the injection point and forming a major vertical migration path which conducts CO2 almost directly to the top of the reservoir. Note that Fig. 4 suffers from some ‘‘large pushdown, low amplitude’’ anomalies not consistent with the tuning relation. These anomalies correspond to interpretations within the chimney where most of the seismic amplitude information is destroyed. 5. Reservoir flow simulation In order to help understand the CO2 flow in the reservoir and to predict any future flow, a full-field 3D reservoir flow simulation model has been constructed based on the seismic interpretation [2]. This was not a straightforward task, taking into account that the intra-reservoir shale layers could only be interpreted on the time-lapse surveys with CO2 present below them. Moreover, these interpretations were in two–way traveltime and had to be time-depth converted for the flow simulation model. With the large velocity variations occurring in the reservoir due to the presence of CO2, an accurate time–depth conversion is not simple and an
Fig. 5. Result of a reservoir flow simulation at the time of the first time-lapse seismic survey in October 1999 (after 3 years of injection) after 2.35 million tonnes of CO2 injected [2].
R. Arts et al. / Energy 29 (2004) 1383–1392
1389
accurate depth model of each individual shale layer has not been obtained so far. To circumvent this problem, these intra-shale layers have been assumed parallel to the upper shale layer just below the top of the reservoir. The total simulation flow model covers an area of 15.7 km2. The Utsira Sand was represented by a total of 59 layers. The shale layers were represented in the model by transmissibility modifiers between the appropriate layers. The result of the simulation after 3 years of injection, at the time of the first time-lapse seismic survey in October 1999, is shown in Fig. 5. The resulting saturation model has been transformed into a 3D acoustic impedance model using the Gassmann relations in order to perform a forward seismic modelling. A cross-section, representing crossline 3123 in 2001, is shown in Fig. 6. The leftmost image represents the acoustic impedance model in depth, the middle image is the synthetic seismic data and the rightmost image corresponds to the seismic field data both in two-way traveltime. Taking into account the assumptions made to construct the model, the synthetic seismic data mimic the field data fairly well. The pushdown effect is clearly demonstrated on the synthetic seismic data. Fig. 7 shows the lateral extent of the pushdown anomalies for the 1999 seismic data, for the 2001 seismic data and for the synthetic data (based on the simulation for 2001). The field data show a more north outh elongated shape, whereas the synthetic data follow the domal shape of the upper shale layer (level 8) topography.
Fig. 6. Cross-section from west to east of the acoustic impedance model derived from the 2001 reservoir simulation model in depth (left), the synthetic seismic data (middle) and the corresponding 2001 seismic data both in two-way traveltime.
1390
R. Arts et al. / Energy 29 (2004) 1383–1392
Fig. 7. Structure top of the upper shale layer (level 8) with superimposed outlines of the pushdown anomalies as observed in the 1999 dataset (light grey), in the 2001 dataset (dark grey) and in the 2001 synthetic seismic data (black).
This effect can also be observed on the synthetic data shown in Fig. 6. The lateral extent of the CO2 anomaly on the east–west synthetic data of crossline 3123 is larger than on the field data.
6. Conclusion Time-lapse seismic surveying appears a highly suitable geophysical technique for monitoring CO2 injection into a saline aquifer. The effects of the CO2 on the seismic data are large both in terms of seismic amplitude and in observed velocity pushdown effects. In addition to straightforward mapping of changes in the time-lapse seismic data, quantitative approaches to mapping the supercritical CO2 saturations have been carried out, supported by seismic modelling. A more detailed analysis can be found in Refs. [4,5].
R. Arts et al. / Energy 29 (2004) 1383–1392
1391
The interpreted seismic data have been used for a full 3D simulation model. The construction of such a model was problematical for two main reasons: . The intra-reservoir shale layers could only be interpreted on the time-lapse surveys when CO2 was present below them. . The conversion from time to depth of these intra-reservoir shale layers is not sufficiently accurate due to the large velocity variations caused by the presence of CO2. As a consequence, an assumption on the depth structure of the intra-reservoir shale layers has been made, assuming them to be parallel to the upper shale layer, the only one that has been interpreted on the baseline seismic data. The saturation model resulting form the reservoir simulation has been transformed into an acoustic impedance model using the Gassmann model, which in turn has been used to create synthetic seismics. Taking into account the assumptions made to construct the model, the synthetic seismic data match the field data fairly well. Acknowledgements The SACS project was funded by the European Community (Thermie program), by the industry partners, i.e.: Statoil, BP, Norsk Hydro, ExxonMobil, TotalFinaElf and Vattenfall and by national governments. The R&D partners are TNO-NITG (TNO-Netherlands Institute of Applied Geoscience—National Geological Survey), SINTEF Petroleum Research, BGS (British Geological Survey), BRGM (Bureau de Recherches Ge´ologiques et Minie`res), GEUS (Geological Survey of Denmark and Greenland) and IFP (Institut Francais du Pe´trole). The IEA-Greenhouse Gas Programme is involved in the dissemination of the results. References [1] Lindeberg E, Zweigel P, Bergmo P, Ghaderi A, Lothe A. Prediction of CO2 dispersal pattern improved by geology and reservoir simulation and verified by time lapse seismic. In: Williams DJ, et al., editors. Greenhouse gas control technologies. Collingwood, Australia: CSIRO Publishing; 2001, p. 372–7. [2] Meer LGHvander, Arts RJ, Peterson L. Prediction of migration of CO2 injected into a saline aquifer: reservoir history matching to a 4D seismic image with a compositional gas/water model. In: Williams DJ, et al., editors. Greenhouse gas control technologies. Collingwood, Australia: CSIRO Publishing; 2001, p. 378–84. [3] Eiken O, Brevik I, Arts R, Lindeberg E, Fagervik K. Seismic monitoring of CO2 injected into a marine aquifer. Society of Exploration Geophysicists (SEG), Calgary 2000 International Conference and 70th Annual Meeting, Calgary, Paper RC-8.2. 2000, p. 4. [4] Arts R, Elsayed R, van der Meer L, Eiken O, Ostmo S, Chadwick A, et al. Estimation of the mass of injected CO2 at Sleipner using time-lapse seismic data. 64th European Association of Geoscientists and Engineers (EAGE) Meeting, Florence, Paper H016. 2002, p. 4. [5] Chadwick RA, Arts R, Eiken O, Kirby GA, Lindeberg E, Zweigel P. 4D seismic imaging of an injected CO2 plume at the Seipner Field, central North Sea. In: Davies R, Cartwright J, Stewart S, Underhill J, Lappin M, editors. 3D seismic data: application to the exploration of sedimentary basins, vol. 29. London: Memoir of the Geological Society; 2004, p. 305–14. [6] Gassmann F. Uber die Elastizitat poroser Medien. Vier, der Natur, Gesellschaft in Zurich 1951;96:1–23 in German.
1392
R. Arts et al. / Energy 29 (2004) 1383–1392
[7] Mavko G, Mukerji T, Dvorkin J. The rock physics handbook. Cambridge: Cambridge University press; 1998 pp. 168-170. [8] Arts R, Eiken O, Chadwick A, Zweigel P, van der Meer L, Kirby G, Seismic monitoring at the Sleipner underground CO2 storage site (North Sea). In: Baines SJ, Worden RH, editors Geological storage of CO2 for emissions reduction. Geological Society, London, Special Publication [in press].