Fault-seal analysis for CO2 storage: an example from the Troll area, Norwegian Continental Shelf

Peter Bretan, Graham Yielding, Odd Magne Mathiassen and Tove Thorsnes
Petroleum Geoscience, 17, 181-192, 27 April 2011, https://doi.org/10.1144/1354-079310-025

Abstract

ABSTRACT Fault seal plays a critical part in hydrocarbon traps, and the same will be true for CO2 storage. The standard workflow for prediction of capillary seal of hydrocarbons can be readily adapted to prediction of CO2 seal since the fluid properties of CO2 at reservoir temperatures and pressures are within the range shown by hydrocarbons. The workflow is applied in a feasibility study into the proposed CO2 storage in the Johansen Formation of the Troll Field.

Computation of Shale Gouge Ratio (SGR) over the fault surfaces, in combination with juxtaposition diagrams, was used to estimate the sealing potential of faults cutting the Johansen Formation. SGR values were converted to potential CO2 column heights that might be trapped at each fault. Column heights are generally less than 100 m at each fault, allowing a cross-fault migration route from the Johansen Sand via the Statfjord Formation, Cook Formation and Brent Group.

Analysis of in-situ stresses suggests that the faults in the Troll Field are not close to failure and therefore up-dip leakage of CO2 is unlikely. Extremely large CO2 columns (>300 m) would be required to change this stress stability, and across-fault capillary leakage would occur first.

KEYWORDS

INTRODUCTION

The sealing or non-sealing behaviour of faults within hydrocarbon-bearing reservoirs is a crucial aspect in assessing the overall structural integrity, and economic viability of a fault-bounded trap. Faults can form impermeable barriers to the migration of hydrocarbons over geological timescales. Alternatively, they may act as ‘baffles’ that impede or even change the flow direction of hydrocarbons over production timescales, owing to abrupt changes in reservoir pressure during injection and/or depletion (Yielding et al. 1999). The same will also be true for the injection and storage of CO2. In order to assess whether the proposed storage site for CO2 is geologically stable over long time periods, it is vitally important to assess the likely sealing/non-sealing behaviour of the faults to progressively increasing CO2 fluid pressures. Two aspects of fault behaviour are paramount in CO2 storage. First, will the fault act as a sealing lateral barrier thus permitting CO2 to accumulate within the trap, and if so what would be the likely height of the CO2 that the fault could support without leaking? Secondly, will the increasing pressure generated by the CO2 column trigger fault instability and reactivation, leading to the potential loss of CO2 by migration up the fault?

Given the critical role played by faults in affecting fluid flow within a reservoir, it is surprising that there is little published work that specifically addresses the across-fault leakage potential of faults in CO2 reservoirs. Current work on seal capacity and fault behaviour within CO2 reservoirs is generally focused on three key aspects. First, seal capacity and CO2 column heights are evaluated in terms of top seals and cap rocks using theoretical relationships between pore throat size, wettability and interfacial tension of CO2–water–rock systems (e.g. Daniel & Kaldi 2008). Secondly, the seal integrity of faults is investigated in terms of fault stability, reactivation and vertical leakage (up the faults) induced by elevated formation pressures (e.g. Streit & Hillis 2004; Chiaramonte et al. 2008; Rutqvist et al. 2007). In this type of analysis, the faults are analysed as 3D fault model surfaces without any variable fault-zone properties. Detailed fieldwork on faults that are shown to be leaking to CO2 (e.g. Shipton et al. 2004) can provide additional input for these essentially geomechanical models. Finally, the impact of faults on the rate of CO2 migration within a reservoir is addressed using Discrete Fracture Models and fluid flow simulations. Faults within simulation models are generally treated as simple flow barriers having constant low permeability (e.g. Chadwick et al. 2009; Idling & Ringrose 2009) or as open conduits and/or closed barriers to CO2 migration (e.g. Chang & Bryant 2008).

In this contribution we will show how the methodology for assessing fault seal potential in hydrocarbon–water systems can be used to address fault seal risk within a proposed CO2 storage site in an offshore oil and gas field in the North Sea. The results of this study can be used to help refine the fault-seal methodology for CO2–brine systems.

REVIEW OF FAULT-SEAL METHODOLOGY

A methodology for predicting fault-seal behaviour in mixed clastic sequences in areas of low differential stress has been well documented in recent years (e.g. Bouvier et al. 1989; Jev et al. 1993; Childs et al. 1997; Fristad et al. 1997; Fulljames et al. 1997; Yielding et al. 1997; Knipe et al. 1998; Yielding 2002; Bretan et al. 2003). The basis for the methodology is that most fault seals in petroleum reservoirs are capillary (or membrane) seals (Jennings 1987; Watts 1987). In capillary seals, surface tension forces between hydrocarbon and water hold back the hydrocarbon phase and prevent it from entering the water-wet seal lithology. Leakage of hydrocarbons through a water-wet fault zone is by capillary action and takes place when the difference in pressure between the water and hydrocarbon phases (buoyancy pressure) exceeds the pressure required for hydrocarbons to enter and pass through the largest interconnected pore throat pathway across the seal (displacement or capillary threshold pressure). In this context, capillary threshold pressure is controlled by the pore throat size within a fault zone. The smaller the pore throat size, the higher the capillary threshold pressure required for the seal to fail and the greater the hydrocarbon column that can potentially be supported prior to failure.

The principal steps for assessing fault-seal in typical hydrocarbon exploration settings are briefly described below.

The first step is to use seismic-scale interpretations to construct a geologically realistic 3D structural model of the faulted horizons. Limitations in vertical seismic resolution place a major constraint on the level of stratigraphic detail that can be picked on seismic data. However, since seals can only be as strong as their weakest point, the thinnest layers that may lead to a potentially ‘weak’ juxtaposition (reservoir-on-reservoir) must also be included in the model. The missing stratigraphic detail between the interpreted seismic horizons can be interpolated either by using a geological layer model for the intervening stratigraphic relationships, or by the direct mapping of Vshale curves onto the fault. Both approaches enable the construction of fault-plane diagrams (Allan 1989) to help identify the distribution of juxtaposed reservoirs and possible migration paths between fault blocks (Gibson & Bentham 2003).

The next step is to apply predictive algorithms to characterize the different parts of the fault surface according to an estimate of the shale (or clay) content of the fault-zone rock, as it is the fine grained phyllosilicate particles that are primarily responsible for reducing the pore throat size of the fault rock (e.g. Yielding et al. 1997; Wibberley et al. 2008). Two types of predictive algorithms are commonly used to predict the clay content of a fault zone, namely Smear Factor and Gouge Ratio algorithms. Smear Factor algorithms include the Clay Smear Potential (Bouvier et al. 1989; Jev et al.1993; Fulljames et al. 1997) and the Shale Smear Factor algorithm (Lindsay et al. 1993). These algorithms attempt to model the morphology of shale smears along fault planes. Calibration studies using the Clay Smear Potential algorithm have been described (e.g. Bouvier et al. 1989; Jev et al. 1993; Fulljames et al. 1997) but are more qualitative compared to calibrations using the SGR algorithm (see below). As a result, we concentrate here on the Gouge Ratio method.

The Gouge Ratio algorithms include the Shale Gouge Ratio (or SGR) of Freeman et al. (1998) and Yielding et al. (1997). In this algorithm it is assumed that sand and shale material in the wall rock are incorporated into the fault gouge in the same proportions (ratio) as they occur in the wall rocks of the slipped interval, hence SGR is considered as a predictor of up-scaled fault-rock composition ( Fig. 1). The amount of shale material in the wall rock, referred to as the volumetric clay fraction (Vclay), is typically derived from well data such as gamma-ray and neutron-density logs. One approach (e.g. Lyon et al. 2005) is to assign a 100% shale value to the maximum average gamma-ray value; a 0% shale value to the average minimum gamma-ray value and to assume a linear (or non-linear) relationship between increasing gamma-ray value and shale volume. Bretan et al. (2003) and Bretan & Yielding (2005) show how different estimates for the shale volume can lead to significant uncertainties in estimating the capillary threshold pressures of fault zones.

Fig. 1
Fig. 1

Schematic diagram showing definition of Shale Gouge Ratio, after Yielding et al. (1997). At any point on the fault surface, the SGR is equal to the net shale (or clay) content of the interval (t) that has slipped past that point.

A high SGR value is expected to correspond to more phyllosilicates in the fault zone (e.g. clay smear), and therefore to higher capillary threshold pressure and lower permeability. Field studies (e.g. Foxford et al. 1998; van der Zee & Urai 2005; see also Yielding 2002; Wibberley et al. 2008) have shown that there is a general correlation between the measured clay content of a fault zone and the calculated SGR value, higher SGR values being derived for fault zones containing a higher observed clay content. In many basins and in particular the Brent Province (Yielding 2002), it is observed that SGR>15–20% corresponds to faults that are sealing to hydrocarbons. In the faulted Jurassic reservoirs of the northern North Sea, SGR values <15–20% correspond to fault rock dominated by disaggregation zones, with negligible sealing capacity (Fossen et al. 2007).

The final step in the workflow is to calibrate the estimate of the fault-rock composition with observed hydrocarbon accumulations and then use the calibration to estimate the potential column heights that might be trapped at the fault. SGR can be quantitatively calibrated by compiling a database of the pressure differences held back at faults in cases where reservoirs on both sides of the fault are drilled (Yielding et al. 1997; Bretan et al. 2003). Plotting the SGR values against in-situ pressure data onto one calibration diagram reveals a general trend of increasing SGR value supporting increasing across-fault pressure difference ( Fig. 2) (Yielding 2002; Bretan et al. 2003). The empirical equation defining these seal-failure envelopes is Embedded Image (1) where C is 0.5, 0.25, 0 for increasing burial depths (see Bretan et al. 2003 for more details). These seal-failure envelopes are similar to trends in measured capillary threshold pressures of actual fault rocks (Sperrevik et al. 2002).

Fig. 2
Fig. 2

Empirical approach to fault-seal calibration (after Yielding 2002; Bretan et al. 2003) showing a global compilation of across-fault pressure differences and their relationship to Shale Gouge Ratio (SGR) at the same point on the fault surface. Data points are colour-coded by burial depth (blue, < 3.0 km; red, 3.0–3.5 km; green, > 3.5 km). Dashed lines represent the maximum across-fault pressure that a specific SGR could support without leaking (the seal envelopes).

Assuming the seal-failure envelopes represent fault-rock threshold pressure Pc (i.e. Pc = AFPD), then potential maximum hydrocarbon column heights trapped by the fault are given by: Embedded Image (2) where ρw and ρh are the densities of water and hydrocarbon and g is the acceleration due to gravity.

The ‘critical’ (first-to-fail) leak point on a fault is found by mapping these potential column heights over the fault plane surface, taking account of the 3D geometry of the reservoir sand on the trap side of the faults where it is juxtaposed against other sands on the far side of the faults (Bretan et al. 2003).

An additional step can be included in the fault-seal workflow to assess the risk that a fault which is sealing in all other respects might leak because of fault reactivation. In such cases, the seal/leak behaviour of the fault is principally controlled by the stress state and fault rock mechanical properties (e.g. Ferrill et al. 1999; Mildren et al. 2005). When a fault is critically stressed and close to failure, dilatant micro-fracturing may provide fluid pathways to allow buoyant fluid to migrate up the fault zone to higher reservoirs. Assessing whether a fault is critically stressed requires knowledge of the in-situ stress states, pore-pressure, orientation of the fault with respect to the principal stress axes, and the geomechanical strength of the fault rock. This type of approach is often termed Geomechanical Analysis and has been used over recent years to investigate the mechanical stability of faults within CO2 reservoirs (e.g. Streit & Hillis 2004) and to estimate CO2 column heights that specific faults could support without undergoing reactivation (e.g. Chiaramonte et al. 2008).

FAULT SEAL METHODOLOGY APPLIED TO CO2 STORAGE

In this section we demonstrate the application of the proven fault-seal methodology in the context of CO2 storage using an example from the Norwegian sector of the North Sea. The large Mongstad oil terminal and refinery is located near Bergen, Norway. It is estimated that approximately 2.2 million tonnes of CO2 will be produced per year from the terminal and its new Combined Heat & Power Station (Gassnova 2007). As the emissions permit from the Norwegian Ministry of Environment requires the development of a full-scale Carbon Capture and Storage (CCS) project in parallel with construction and operation of the terminal, a key concern is the sequestration of the large volume of CO2 produced by the terminal. One solution under consideration would be to transport the CO2 by pipeline for storage in the super-giant Troll oil and gas field ( Fig. 3).

Fig. 3
Fig. 3

General location map showing the Mongstad refinery and the Troll field.

The Troll field is located about 60 km offshore Norway (Fig. 3) on the northwestern edge of the Horda platform. The regional geological settings and reservoir geology of the Troll field is described elsewhere (e.g. Bolle 1992). Troll consists of three large-volume eastward-tilted fault blocks that increase in crestal depth from east to west. Three communicating hydrocarbon provinces coincide with the fault blocks of which the Troll West Gas province forms the basis of this study. Hydrocarbon production from Troll West is from the Middle to Upper Jurassic Fensfjord and Sognefjord Formations of the Viking Group ( Fig. 4).

Fig. 4
Fig. 4

Jurassic lithostratigraphy of the Troll field showing the main producing reservoirs in Troll West Gas Province (Middle to Upper Jurassic Fensfjord and Sognefjord Formations) circled in red and the proposed CO2 repository (Lower Jurassic Johansen Formation) circled in blue (from Fraser et al. 2003).

In terms of CO2 storage, attention is focused on the saline aquifers below the producing reservoirs, such as the 80–120m thick Lower Jurassic Johansen Formation which is c. 500m below the Fensfjord producing reservoir. The Johansen Formation is located at a depth of 2000–2500 m below sea level and is stratigraphically isolated from the producing reservoirs by the thick shale sequences of the Dunlin Group (Fig. 4). Well data indicates that the Johansen Formation was not involved in the hydrocarbon migration and trapping within Troll nor is there any data to show that the formation is being affected by production. The Johansen Formation consists of high porosity (average 25%) sandstone, the upper Johansen Sand, and the lower Johansen Shale (Fig. 4). The high porosity combined with the overlying sealing lithologies of the Dunlin Formation make the Johansen Formation a suitable candidate for storage of CO2 (e.g. Bergmo et al. 2009; Eigestad et al. 2009).

Of concern are the large normal faults, defining the Troll fault block structure, that cut through the entire Jurassic reservoir sequence and which might provide migration pathways for the injected CO2 from the deeper Johansen Formation to the shallower Fensfjord and Sognefjord hydrocarbon-producing reservoirs ( Fig. 5). In 2008, a detailed study was carried out for the Norwegian Petroleum Directorate (NPD) to investigate the sealing properties and potential for reactivation of these large trap-bounding faults at the level of the Johansen Formation. The analysis is based on a geological dataset provided by NPD comprising fault & horizon interpretations on 3D seismic data and well paths, well picks & Vclay log curves. The current study did not take into account transient pore pressure variations arising from CO2 injection and/or plume migration within the Johansen Formation, which can only be assessed using fluid flow reservoir simulations (e.g. Bergmo et al. 2009; Eigestad et al. 2009).

Fig. 5
Fig. 5

SW–NE seismic section across the Troll Field on the Horda Platform (after Fraser et al. 2002).

Methods for estimating the height of a trapped hydrocarbon column relate the size (radii) of pore throats within a seal rock, the interfacial tension between water and hydrocarbons, and the pressure differentials due to buoyancy forces. These deterministic methods require an estimate for the size of the pore throats within fault-zones and the interfacial tension of oil to water at reservoir conditions (Jennings 1987; O'Connor 2000). The value of interfacial tension for CO2 –brine in the Johansen Sand will depend upon pore pressure and pore-water salinity. In their regional study covering the Norwegian sector of the North Sea, Grollimund et al. (2001) show that the pore pressure is close to hydrostatic in the areas surrounding the Troll field. At a depth of 2000 m below sea level the pore pressure derived from RFT data is c. 20MPa (= 200 bars). Moss et al. (2003) report salinity values ranging from 425 000–86 000 ppm NaCl from reservoirs on the Horda Platform. At pore-pressures of 200 bar and water salinity values of c. 1M NaCl, Chalbaud et al. (2006) show that CO2 –brine interfacial tension values are c. 28 mN m–1, which is very similar to typical oil–brine values (oil industry default value of 30 mN m–1).

To estimate potential CO2 column heights, the density of CO2 at reservoir temperatures and pressures is required. An estimate can be derived from the bottom-hole temperature from wells within the study area. In well 31/2–1, located in the central part of the study area ( Fig. 6) a bottom-hole temperature of 65 °C at 2433 m depth (TVDSS) implies a geothermal gradient of 30.8 °C km–1. This geothermal gradient, in turn, implies a supercritical CO2 density of 0.67 g cm–3 at the planned storage depths of 2200 m (Chadwick et al. 2004). A brine density of 1.035 g cm–3 was used, which is consistent with the measured salinity values and expected temperature/pressure conditions within the Johansen Sand.

Fig. 6
Fig. 6

Structure map of the Top Johansen Formation showing the faults illustrated in this study. Well locations shown as grey circles with common well names (e.g. 31/2–1) and block numbers (e.g. 31/2) in bold text.

Given the overall similarity between the CO2–brine and oil–brine data, the calibrated SGR analysis for hydrocarbon–water can be applied to a CO2–brine system. Fault rock threshold pressures estimated from calibrated SGR values can be converted to trappable CO2 column height using equations (1) and (2) for a given density contrast between CO2 and brine.

There is some remaining uncertainty in the contact angle for the CO2–water–rock system, which is assumed to be zero in the above approach. If the contact angle ranges up to 60° at high pressure (100 bar), then the predicted threshold pressure (Pc) and column heights would be reduced by a factor of × 0.5 (Daniel & Kaldi 2008).

Fault-seal and potential CO2 column heights in Troll West Gas province

Figure 6 is a structure map of the Top Johansen Formation within the main southern fault block of the Troll West Gas province. Faulting consists of numerous intra-block faults trending north–south in the southern part of the block to dominantly NW–SE near the crest. The intra-block faults form an ‘open’ fault pattern in map view in that there are few if any fault–fault intersections that would act as barriers to CO2 migration within the block. The block boundaries are defined by large faults, downthrowing to the west and SW.

The proposed CO2 injection point lies to the south, beyond the southern edge of the present interpretation area shown in Figure 6. The precise flow path into, and migration within the fault block is not known. In this analysis, we have assumed that the injected CO2 would migrate northwards within the Johansen Sand moving up structural dip towards the crest of the structure with flow along and around the intra-block faults. We have also assumed a fill-to-spill migration model with the majority of the injected CO2 accumulating at or near the crest of the structure although a small column may be preserved in a fault-bounded intra-block high located in the SE part of the block (Fig. 6). For the purpose of this contribution, the analysis has focused on five large faults near the crest of the structure. For each fault, two perspective views are presented, the first an Allan diagram showing stratigraphic juxtapositions for the Johansen Sand, and the second showing calculated SGR along the Johansen Sand ( Figs 7–10).

Fig. 7
Fig. 7

Perspective fault plane diagram of Fault_1 viewed looking towards the west, showing (a) Allan Diagram: Top Johansen Formation in the upthrown side juxtaposed against Fensfjord-Krossfjord (blue), Heather (purple), Brent Group (yellow) and Dunlin Group (grey) in the downthrown side; (b) computed SGR values at the Top Johansen Formation in the upthrown side of Fault_1. Vertical exaggeration × 5.

The western boundary to the fault block is Fault_1, downthrowing to the west (Fig. 6). Northward-directed flow of CO2 in the Johansen Sand would approach this fault on the upthrown side. The upthrown Johansen Sand is juxtaposed against downthrown Brent Group and (locally) Fensfjord Formation reservoirs (Fig. 7a). All along the upthrown Johansen Sand the SGR is high, greater than 25% and locally 35% in the south (Fig. 7b), which would correspond to trapped CO2 column heights in excess of 100 m (and locally 200 m in the south). Therefore, across-fault leakage to the Fensfjord or Brent Formation is not expected until at least 100 m of CO2 column accumulates at the block crest.

A significant intra-block fault is the NW–SE trending Fault_2 (see Fig. 6), which is downthrown to the SW. Northward-flowing CO2 in the Johansen Sand would approach on the downthrown side of the fault. The downthrown Johansen is juxtaposed against itself near the fault tips and against Statfjord Formation at depth ( Fig. 8a). Buoyant fluid would preferentially migrate along the highs, and therefore would encounter the self-juxtapositions. The self-juxtaposition of Johansen Sand shows low SGR (<15%) and is thus expected to leak CO2. Across-fault leakage would be into the upthrown Johansen Sand at the crest of the Troll West Gas fault-block. The critical part of the Troll West fault block is its crest. Faults controlling fluid trapping are the WNW-trending Fault_3 and the NNE-trending Fault_4. At Fault_3, the Johansen Sand on the south side is downthrown, and therefore is juxtaposed against stratigraphically deeper levels across the fault. Fault_4 downthrows to the west, therefore the primary interest is the fault-zone sealing potential at the level of the upthrown Johansen Sand.

Fig. 8
Fig. 8

Perspective fault plane diagram of Fault_2 viewed looking towards the NE, showing (a) Allan Diagram: Johansen Sand in the downthrown side juxtaposed against Johansen Sand (red), Johansen Shale (grey) and Statfjord Formation (yellow); (b) computed SGR values at the Johansen Sand in the downthrown side of Fault_2. Vertical exaggeration × 5.

Fault_3 ( Fig. 9) forms the northern boundary to the fault-block crest, separating it from the upthrown block containing well 31/2–3. The downthrown (south side) Johansen Sand is juxtaposed almost entirely against the upthrown Statfjord Formation. Where the fault displacement is slightly reduced there is juxtaposition against upthrown Johansen Shale. The SGR calculations for this juxtaposition area give values mostly higher than 25%. However an SGR minimum of c. 23% occurs at the very apex of the crest.

Fig. 9
Fig. 9

Perspective fault plane diagram of Fault_3 viewed looking towards the NNE, showing (a) Allan Diagram: Johansen Sand in the downthrown side juxtaposed against Johansen Shale (grey), and Statfjord Formation (orange); (b) computed SGR values at the Johansen Sand in the downthrown side of Fault_3. Vertical exaggeration × 5.

Fault_4 forms the northwestern boundary to the fault-block crest, separating it from a downthrown terrace which broadens northwards ( Fig. 10). The upthrown Johansen Sand is juxtaposed almost entirely against Dunlin Group along this fault, with just a little Cook Formation being involved at depths below 2095 m. A juxtaposition seal is expected against the Drake shales. The SGR calculations for this juxtaposition area give values in the 25–35% range, and therefore fault seal is expected even if some thin sands are present in the Dunlin.

Fig. 10
Fig. 10

Perspective fault plane diagram of Fault_4 viewed looking towards the west showing (a) Allan Diagram: Johansen Sand in the upthrown side juxtaposed against Dunlin Group (grey) and the Cook Formation (green); (b) computed SGR values at the Johansen Sand in the upthrown side of Fault_4. Vertical exaggeration × 5.

Using the SGR calibration presented in the previous section, SGR has been converted to fault-rock threshold pressure, and then to maximum trappable CO2 column height. Adding the calculated column height to the in-situ depth at each calculation point then gives the maximum contact depth which each part of the fault can support, shown in more detail for Fault_3 in Figure 11. The shallowest of these values defines the critical leak point on the fault surface. The critical point on Fault_3 is at the junction with Fault_4. This Johansen–Statfjord overlap supports CO2 fill down to c. 2120 m. Then CO2 would leak across the fault into the Statfjord Formation on the upthrown side. Figure 12a illustrates the extent of the trap when the Top Johansen is filled down to 2120 m. With this degree of CO2 fill, the apex of the trap is expected to leak (arrow) across the c. east–west Fault_3 into the Statfjord Formation of the block to the NE. At the Statfjord Formation, 2120 m is very close to the crest of the fault block (Fig. 12b). One possibility is continued fill of the Johansen Formation to the south and further fill of the Statfjord Formation to the north, on both sides of Fault_3. The next question would be whether the other faults bounding the upthrown Statfjord Formation can retain further CO2 fill. In the block surrounded by these 3 faults, there is extensive Statfjord–Cook juxtaposition on Fault_5. A multi-fault trap analysis using the SGR/column-height calculations suggests that the weak point is the crest of the Statfjord Formation on western fault Fault_4 (arrow in Fig. 12b). Maximum supportable column height here would be only 50 m, giving a CO2 contact depth of 2110 m, which is in fact shallower than the 2120 m contact that had driven fluid across Fault_3 into this block. So the Statfjord Formation here provides a ‘short-circuit’ from the Johansen Sand in the south to the Cook Formation in the west ( Fig. 13).

Fig. 11
Fig. 11

Detailed view of Fault_3 showing the critical juxtaposition between the Johansen Sand (downthrown) against the Statfjord Formation (upthrown). The juxtaposition is colour-coded by the maximum supported CO2–brine contact depth (calculated as fault depth + column height, see text). At the critical leak point of 2056 m, the fault could only support a CO2 column of 64 m before leaking CO2 into the Statfjord Formation. Vertical exaggeration × 5.

Fig. 12
Fig. 12

Map views of the Top Johansen (a) and Top Statfjord horizon surfaces (b), colour-coded to show fill down to 2120 m (dashed contour), which is the maximum fill predicted for the southern block before it leaks through Fault_3 from Johansen to Statfjord formations. Bold arrows show the expected leak points out of the Johansen (a) and Statfjord (b) Formations through Fault_3 and Fault_4 respectively.

Fig. 13
Fig. 13

Seismic profile (a) illustrating possible CO2 migration from the Johansen Formation in the Troll West Gas Province to the productive Fensfjord Formation in the Troll West Oil Province. Trace of seismic line shown in map (inset). The critical area (white circle) shown in detail (b) illustrates communication across Fault_3 from Johansen Sand to Statfjord Formation in horst, then across Fault_4 to Cook Formation.

Once migrating fluid has passed the fault intersection area, it can migrate up dip to the NW in the Cook Formation. Further analysis of the faults to the NW show similar across-fault pathways to the Brent Group and ultimately to the Fensfjord Formation of the Troll West Oil Province.

Fig. 14
Fig. 14

Synopsis of the inferred CO2 migration (red arrows) and predicted fill in each block on the route; closure at ‘1’ and ‘2’ is in Johansen Sand (J), at ‘3’ it is in the Statfjord Formation (S), at ‘4’ is in the Cook Formation (C), and ‘5’ is in the Brent Group (Br). Troll West Oil Province (TWOP) is coloured by the Sognefjord/Fensfjord hydrocarbon fill (Fen). Colours show the fill required at each closure to cause spill/leak: c. 75 million m3 each (50 million tonnes CO2). Predicted CO2–brine contact depths in metres.

Figure 14 provides a synopsis of the inferred CO2 migration routes based on the preceding analyses. The important steps are:

  1. From the south, CO2 migration in the Johansen Sand can migrate around or leak across the intra-block faults to the local structural closure (crest 2060 m). This closure can fill to c. 2160 m then spill northwestwards.

  2. From the south, CO2 migration in the Johansen Sand can leak across Fault_2 to reach the apex of the main southern Troll West Gas Province fault block. At this apex, fill can occur in the Johansen down to c. 2120 m. Then across-fault leakage can occur across Fault_3 into the upthrown Statfjord Formation.

  3. The CO2 column in the Statfjord Formation can then immediately leak westwards across Fault_4 into the downthrown Cook Formation.

  4. Migration northwestwards in the Cook Formation will continue to the apex of the next fault-block (1900 m). When this block has filled to c. 1950 m then CO2 may leak across the faults from the Cook Formation into the downthrown Brent Group and then ultimately into the Fensfjord Formation in the Troll West Oil Province.

  5. For each accumulation on this migration route, simple estimates of trapped CO2 volumes are about 75 million m3, equivalent to about 50 million tonnes of CO2, corresponding to over 22 years CO2 production from the Mongstad refinery.

Stress-driven fault leakage

The previous section described fault-seal analysis based on the capillary strength of fault-zone rock, i.e. how much buoyancy pressure can the rock support before fluid invades the connected pore-space and flows across the fault zone. An alternative mechanism for CO2 leakage from a fault-bound trap is CO2 flow up the fault, which is principally controlled by the stress state (Streit & Hillis 2004).

In-situ stress data for this study comes from a Statoil report (Skomedal & Raaen 1999) on stress measurements in well 31/6-A-21, located to the SE of the study area. Vertical stress (Sv) was determined from density measurements through the overburden, and minimum horizontal stress (SHmin) was determined from closure of hydraulic fractures. Pre-production pore-pressure in the area is hydrostatic. These measurements do not include the value of the maximum horizontal stress (SHmax), nor the orientations of the maximum and minimum stresses. Additional information is available in the public-domain World Stress Project (Heidbach et al. 2008), and data from the Horda Platform. The nearest data-point to the Troll Field is at the NW edge of the field – one ‘C-quality’ observation point based on a single normal faulting earthquake focal mechanism, with SHmax oriented at an azimuth of 013°. Further north (c. 45km) in Quad 35 (Vega South / Fram) and west (c. 55kms) in Quad 30 (Oseberg), a number of ‘A-quality’ data-points based on well data (breakouts, fractures, etc) indicate a strike-slip regime with SHmax oriented c. 081°. Wiprut & Zoback (2002) report that a strike-slip regime is common in many northern North Sea fields, with SHmax typically around 1.3–1.4 times the vertical stress (Sv). Frictional rock strength places limits on the ratio of principal effective stress values, with (S1-pp)/(S3-pp) <3.1, where S1 and S3 are maximum and minimum principal stresses and pp is pore pressure.

The available stress observations are therefore ambiguous and permit two different possibilities. The nearest (lower quality) observation implies a present-day normal faulting stress regime, whereas the regional data (high quality) imply a present-day strike-slip faulting stress regime. Both of these possibilities have been used to perform geomechanical modelling of the stress states of the Troll fault network ( Table 1). In the normal faulting regime, SHmax has orientation 013° and is assumed to be just a little less than Sv (to be as consistent as possible with the regional strike-slip observations). In the strike-slip regime, SHmax has orientation 081° and is assumed to be as large as possible without exceeding its frictional maximum; this gives an end-member model with maximum possible stresses.

View this table:
Table 1

Stress data used for the normal (a) and for the strike-slip (b) stress regimes. For both cases, the pore pressure, vertical stress and minimum horizontal stresses were defined from stress data from well 31/6-A-21. In the normal stress regime, the maximum horizontal stress (SHmax) is assumed to be just a little less than the vertical stress, with an orientation of 013 degrees. In the strike-slip regime, maximum horizontal stress is at a frictional maximum, with an orientation of 081 degrees

The results of the geomechanical analyses for the two stress regimes are shown explicitly for the mapped fault planes in Figure 15. In these map views, the stress tensor has been resolved onto the faults to compute the shear and normal stresses, and the faults are colour-coded by the resultant value of slip tendency (ratio of shear stress to normal stress, Ferrill et al. 1999). A slip tendency of 0.6 corresponds to the frictional strength of a cohesionless rock surface, and can be considered an approximate estimate of the stress state that would induce slip (and therefore cause the fault to act as a fluid conduit). In the normal faulting stress regime, none of the faults are close to failure (slip tendency 0.6), but note that the east–west trending faults have higher slip tendency values when the strike-slip regime is used (Fig. 15b).

Fig. 15
Fig. 15

Comparison of Slip Tendency values (ratio of shear stress to normal stress) on all mapped faults in the study area, for the normal (a) and strike-slip regimes (b). Slip tendency is well below the failure threshold of 0.6 for the normal regime whereas some east–west faults are closer to failure using the strike-slip regime.

One way to further quantify the possibility of slip is to use a different stress attribute called fracture stability. This is the critical pressure perturbation required to induce failure on a particular fault orientation in the present day stress regime (e.g. Mildren et al. 2005). The pore-pressure increase might occur as a result of a new buoyant fluid column such as CO2. The critical column height corresponding to the pore-pressure increase P depends on the fluid densities: Embedded Image (3) where ρw and ρh are the densities of water and buoyant fluid and g is the acceleration due to gravity.

Fig. 16
Fig. 16

Computed Fracture Stability on all fault planes, displayed as equivalent CO2 column heights. At most of the faults, several kilometres of column would be required to induce failure; even on the more favourably oriented east–west faults a CO2 column greater than 300 m in height is required.

Figure 16 shows the calculated fracture stabilities (using the strike-slip stress regime) converted to an equivalent CO2 column height. This shows the predicted amount of trapped CO2 column that would be required to induce fault slip and therefore to cause up-dip fault leakage out of the Johansen reservoir. On most of the faults, CO2 column heights in excess of 1000 m are needed to reach failure stress. On some of the east–west structures, required column heights are smaller because these faults are more favourably oriented for slip in the strike-slip stress regime. However, all of these faults require more than 300 m of CO2 column before reaching failure. The across-fault capillary leakages described earlier in this contribution are much weaker than this, and would fail first. Therefore, up-fault CO2 leakage is very unlikely to happen because even with an extreme (high-stress) in-situ stress field, very large pore-pressure perturbations are needed to induce slip and the capillary strength of the fault rocks is sufficiently weak that across-fault migration will occur first.

CONCLUSIONS

  • A detailed fault framework has been built from interpretation of 3D seismic data over the western Troll Field.

  • Allan diagrams of all faults have been combined with fault-seal prediction, to identify those faults critical to fluid migration through the area.

  • Established fault-seal calibration for hydrocarbons has been adapted for CO2 storage, taking account of CO2 fluid properties.

  • A CO2 column of c. 60 m (c. 50 million tonnes, corresponding to over 22 years CO2 production from the Mongstad refinery) could be trapped in the Johansen Sand by faults at the apex of the Troll West Gas fault block.

  • A migration route then occurs across several faults to the NW, each trapping c. 60 m of column. Ultimately, migration is into the Fensfjord reservoir of Troll West Oil province.

  • Analysis of in-situ stress suggests that the extra CO2 pore-pressure is unlikely to cause direct leakage up the fault planes.

Acknowledgments

The authors are grateful to John Underhill and to two anonymous reviewers for their thoughtful and constructive comments. The Norwegian Petroleum Directorate (NPD) is thanked for its approval to publish the results of the study.

REFERENCES

  1. ↵
    1. Allan U.S.
    1989. Model for hydrocarbon migration and entrapment within faulted structures. AAPG Bulletin, 73, 803–811.
    OpenUrlAbstract
  2. ↵
    1. Bergmo Per E.S.,
    2. Lindeberg E.,
    3. Riis F.,
    4. Johansen W.T.
    2009. Exploring geological storage sites for CO2 from Norwegian gas power plants: Johansen Formation. Energy Procedia, 1, 2945–2952.
    OpenUrlCrossRefWeb of Science
  3. ↵
    1. Bolle L.
    1992. Troll Field: Norway's Giant Offshore Gas Field. In: Halbouty M.T. (ed.) Giant Oil and Gas Fields of the Decade 1978–1988. AAPG Memoir, 54, 447–458.
    OpenUrl
  4. ↵
    1. Bouvier J.D.,
    2. Kaars-Sijpesteijn C.H.,
    3. Kluesner D.F.,
    4. Onyejekwe C.C.,
    5. Van Der Pal R.C.
    1989. Three-dimensional seismic interpretation and fault sealing investigations, Nun River Field, Nigeria. AAPG Bulletin, 73, 1397–1414.
    OpenUrlAbstract
  5. ↵
    1. Bretan P.,
    2. Yielding G.
    2005. Using buoyancy pressure profiles to assess uncertainty in fault seal calibration. In: Boult P., Kaldi J. (eds) Evaluating Fault and Cap Rock Seals. AAPG Hedberg Series, Tulsa, 2, 151–162.
    OpenUrl
  6. ↵
    1. Bretan P.,
    2. Yielding G.,
    3. Jones H.
    2003. Using calibrated shale gouge ratio to estimate hydrocarbon column heights. AAPG Bulletin, 87, 397–413.
    OpenUrlAbstract/FREE Full Text
  7. ↵
    1. Chadwick R.A.,
    2. Holloway S.,
    3. Brook M.S.,
    4. Kirkby G.A.
    2004. The case for underground CO2 sequestration in northern Europe. In: Baines S.J., Worden R.H. (eds) 2004. Geological Storage of Carbon Dioxide. Geological Society, London, Special Publications, 233, 17–28.
    OpenUrlAbstract/FREE Full Text
  8. ↵
    1. Chadwick R.A.,
    2. Noy D.J.,
    3. Holloway S.
    2009. Flow processes and pressure evolution in aquifers during the injection of supercritical CO2 as a greenhouse gas mitigation measure. Petroleum Geoscience, 15, 59–73.
    OpenUrlAbstract/FREE Full Text
  9. ↵
    1. Chalbaud C.,
    2. Robin M.,
    3. Egermann P.
    2006. Interfacial tension data and correlations of brine/CO2 systems under reservoir conditions. SPE 102918.
  10. ↵
    1. Chang K.,
    2. Bryant S.L.
    2008. The effect of faults on dynamics of CO2 plumes. Presented at the 9th International Conference on Greenhouse Gas Control Technologies (GHGT-9), Washington, D.C. 16–20 November 2008. GCCC Digital Publication Series #08–03b
  11. ↵
    1. Chiaramonte L.,
    2. Zoback M.D.,
    3. Friedmann J.,
    4. Stamp V.
    2008. Seal integrity and feasibility of CO2 sequestration in the Teapot Dome EOR pilot: geomechanical site characterization. Environmental Geology, 54, 1667–1675.
    OpenUrlCrossRefWeb of Science
  12. ↵
    1. Childs C.,
    2. Watterson J.,
    3. Walsh J.J.
    1997. Complexity in fault zone structure and implications for fault seal prediction. In: Møller-Pedersen P., Koestler A.G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF) Special Publication, Elsevier, Singapore, 7, 61–72.
    OpenUrlCrossRef
  13. ↵
    1. Daniel R.F.,
    2. Kaldi J.G.
    2008. Evaluating seal capacity of caprocks and intraformational barriers for the geosequestration of CO2. PESA Eastern Australian Basins Symposium III, 475–484.
  14. ↵
    1. Eigestad G.T.,
    2. Dahle H.K.,
    3. Hellevang B.,
    4. Riis F.,
    5. Johansen W.T.,
    6. Øian E.
    2009. Geological modelling and simulation of CO2 injection in the Johansen Formation. Computational Geosciences, 13, 435–450.
    OpenUrlCrossRefWeb of Science
  15. ↵
    1. Ferrill D.A.,
    2. Winterle J.,
    3. Wittmeyer G.,
    4. Sims D.,
    5. Colton S.,
    6. Armstrong A.,
    7. Morris A.P.
    1999. Stressed rock strains groundwater at Yucca Mountain Nevada. GSA Today, 9, 1–8.
    OpenUrl
  16. ↵
    1. Fossen H.,
    2. Schultz R.A.,
    3. Shipton Z.K.,
    4. Mair K.
    2007. Deformation bands in sandstone: a review. Journal of the Geological Society, London, 164, 755–769.
    OpenUrlAbstract/FREE Full Text
  17. ↵
    1. Foxford K.A.,
    2. Walsh J.J.,
    3. Watterson J.,
    4. Garden I.R.,
    5. Guscott S.C.,
    6. Burley S.D.
    1998. Structure and content of the Moab Fault Zone, Utah, USA, and its implications for fault seal prediction. In: Jones G., Fisher Q.J., Knipe R.J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 283–297.
    OpenUrlAbstract/FREE Full Text
  18. ↵
    1. Fraser S.I.,
    2. Robinson A.M.,
    3. et al.
    2003. Upper Jurassic. In: Evans D., Graham C., Armour A., Bathurst P. (eds) The Millenium Atlas: Petroleum Geology of the Central and Northern North Sea. Geological Society, London, 2003, 157–190.
  19. ↵
    1. Freeman B.,
    2. Yielding G.,
    3. Needham D.T.,
    4. Badley M.E.
    1998. Fault seal prediction: the gouge ratio method. In: Coward M.P., Daltaban T.S., Johnson H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 19–25.
    OpenUrlAbstract/FREE Full Text
  20. ↵
    1. Fristad T.,
    2. Groth A.,
    3. Yielding G.,
    4. Freeman B.
    1997. Quantitative fault seal prediction: a case study from Oseberg Syd. In: Møller-Pedersen P., Koestler A.G. (eds). Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF) Special Publication, Elsevier, Singapore, 7, 107–124.
    OpenUrlCrossRef
  21. ↵
    1. Fulljames J.R.,
    2. Zijerveld L.J.J.,
    3. Franssen R.C.M.W.
    1997. Fault seal processes: systematic analyses of fault seals over geological and production timescales. In: Møller-Pedersen P., Koestler A.G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF) Special Publication, Elsevier, Singapore, 7, 51–59.
    OpenUrlCrossRef
  22. ↵
    1. Gassnova
    2007. Beslutningsgrunnlag knyttet til transport og deponering av CO2 fra Kårstø og Mongstad. Report no 06/177 (In Norwegian). (available online at www.gassnova.no)
  23. ↵
    1. Gibson R.G.,
    2. Bentham P.A.
    2003. Use of fault-seal analysis in understanding petroleum migration in a complexly faulted anticlinal trap, Columbus Basin, offshore Trinidad. AAPG Bulletin, 87, 465–478.
    OpenUrlAbstract/FREE Full Text
  24. ↵
    1. Grollimund B.,
    2. Zoback M.D.,
    3. Wiprut D.J.,
    4. Arnesen L.
    2001. Stress orientation, pore pressure and least principal stress in the Norwegian sector of the North Sea. Petroleum Geoscience, 7, 173–180.
    OpenUrlAbstract/FREE Full Text
  25. ↵
    1. Heidbach O.,
    2. Tingay M.,
    3. Barth A.,
    4. Reiecker J.,
    5. Kurfeb D.,
    6. Muller B.
    2008. The 2008 release of the World Stress Map (available online at www.world-stress-map.org).
    1. Iding M.,
    2. Ringrose P.
    2009. Evaluating the impact of fractures on the long-term performance of the In Salah CO2 storage site. Energy Procedia, 1, 2021–2028.
    OpenUrlCrossRefWeb of Science
  27. ↵
    1. Jennings J.B.
    1987. Capillary pressure techniques: application to exploration and development geology. AAPG Bulletin, 71, 1196–1209.
    OpenUrlAbstract
  28. ↵
    1. Jev B.I.,
    2. Kaars-Sijpesteijn C.H.,
    3. Peters M.P.A.M,
    4. Watts N.L.,
    5. Wilkie J.T.
    1993. Akaso Field, Nigeria: Use of integrated 3D seismic, fault slicing, clay smearing, and RFT pressure data on fault trapping and dynamic leakage. AAPG Bulletin, 77, 1389–1404.
    OpenUrlAbstract
  29. ↵
    1. Knipe R.J.,
    2. Jones G.,
    3. Fisher Q.J.
    1998. Faulting, fault sealing and fluid flow in hydrocarbon reservoirs: An introduction. In: Jones G., Fisher Q.J., Knipe R.J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, vii–xxi.
    OpenUrlAbstract/FREE Full Text
  30. ↵
    1. Lindsay N.G.,
    2. Murphy F.C.,
    3. Walsh J.J.,
    4. Watterson J.
    1993. Outcrop studies of shale smear on fault surfaces. In: Flint S.S., Bryant I.D. (eds) The Geological Modelling of Hydrocarbon Reservoirs and Outcrop Analogues. International Association of Sedimentologists, Special Publication, 15, 113–123.
    OpenUrl
  31. ↵
    1. Lyon P.J.,
    2. Boult P.,
    3. Hillis R.R.,
    4. Mildren S.
    2005. Sealing by Shale gouge and subsequent sea breach by reactivation: A case study of the Zema Prospect, Otway basin. In: Boult P., Kaldi J. (eds) Evaluating Fault and Cap Rock Seals. AAPG Hedberg Series, 2, 179–197.
    OpenUrl
  32. ↵
    1. Mildren S.D.,
    2. Hillis R.R.,
    3. Dewhurst D.N.,
    4. Lyon P.J.,
    5. Meyer J.J.,
    6. Boult P.J.
    2005. FAST: a new technique for geomechanical assessment of the risk of reactivation-related breach of fault seals. In: Boult P., Kaldi J. (eds) Evaluating Fault and Cap Rock Seals. AAPG Hedberg Series, 2, 73–85.
    OpenUrl
  33. ↵
    1. Moss B.,
    2. Barson D.,
    3. Rakhit K.,
    4. Dennis H.,
    5. Swarbrick R.
    2003. Formation pore pressures and formation waters. In: Evans D. (eds) The Millennium Atlas: Petroleum Geology of the Central and Northern North Sea. Geological Society, London, 317–329.
  34. ↵
    1. O'Connor S.J.
    2000. Hydrocarbon-water interfacial tension values at reservoir conditions: Inconsistencies in the technical literature and the impact on maximum oil and gas column height calculations. AAPG Bulletin, 84, 1537–1541.
    OpenUrlAbstract/FREE Full Text
    1. Rutqvist J.,
    2. Birkholzer J.,
    3. Cappa F.,
    4. Tsang C.F.
    2007. Estimating maximum sustainable injection pressure during geological sequestration of CO2 using coupled fluid flow and geomechanical fault-slip analysis. Energy Conversion and Management, 48, 1798–1807.
    OpenUrlCrossRefWeb of Science
  36. ↵
    1. Shipton Z.K.,
    2. Evans J.P.,
    3. Kirschner D.,
    4. Kolesar P.T.,
    5. Williams A.P.,
    6. Heath J.
    2004. Analysis of CO2 leakage through ‘low-permeability’ faults from natural reservoirs in the Colorado Plateau, east–central Utah. In: Baines S.J., Worden R.H. (eds) Geological Storage of Carbon Dioxide. Geological Society, London, Special Publications, 233, 43–58.
    OpenUrlAbstract/FREE Full Text
  37. ↵
    1. Skomedal E.,
    2. Raaen A.M.
    1999. Troll East Gas Well 31.6-A-21 Stress Measurement. Statoil report, document no. 99A0200000451.
  38. ↵
    1. Sperrevik S.,
    2. Gillespie P.A.,
    3. Fisher Q.J.,
    4. Halvorsen T.,
    5. Knipe R.J.
    2002. Emprirical estimation of fault rock properties. In: Koestler A.G., Hunsdale R. (eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society (NPF) Special Publication, Elsevier, Amsterdam, 11, 109–125.
    OpenUrlCrossRef
  39. ↵
    1. Streit J.E.,
    2. Hillis R.R.
    2004. Estimating fault stability and sustainable fluid pressures for underground storage of CO2 in porous rock. Energy, 29, 1445–1456.
    OpenUrlCrossRef
  40. ↵
    1. Van der Zee W.,
    2. Urai J.L.
    2005. Processes of fault evolution in a siliciclastic sequence: a case study from Miri, Sarawak, Malaysia. Journal of Structural Geology, 27, 2281–2300.
    OpenUrlCrossRefWeb of Science
  41. ↵
    1. Watts N.L.
    1987. Theoretical aspects of cap-rock and fault seals for single- and two phase hydrocarbon columns. Marine and Petroleum Geology, 4, 274–307.
    OpenUrlCrossRefWeb of Science
  42. ↵
    1. Wibberley C.,
    2. Yielding G.,
    3. Toro G. D.
    2008. Recent advances in the understanding of fault zone internal structure: a review. In: Wibberley C., Kurz W., Imber J., Holdsworth R.E., Collettini C. (eds) The Internal Structure of Fault Zones: Implications for Mechanical and Fluid-Flow Properties. Geological Society, London, Special Publications, 299, 5–33.
    OpenUrlAbstract/FREE Full Text
  43. ↵
    1. Wiprut D.,
    2. Zoback M.D.
    2002. Fault reactivation, leakage potential and hydrocarbon column heights in the northern North Sea. In: Koestler A.G., Hunsdale R. (eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society (NPF) Special Publication, Elsevier, Amsterdam, 11, 203–219.
    OpenUrl
  44. ↵
    1. Yielding G.
    2002. Shale Gouge Ratio – calibration by geohistory. In: Koestler A.G., Hunsdale R. (eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society (NPF) Special Publication, Elsevier, Amsterdam, 11, 1–15.
    OpenUrlCrossRef
  45. ↵
    1. Yielding G.,
    2. Freeman B.,
    3. Needham T.
    1997. Quantitative fault seal prediction. AAPG Bulletin, 81, 897–917.
    OpenUrlAbstract
  46. ↵
    1. Yielding G.,
    2. Øverland J.A.,
    3. Byberg G.
    1999. Charcterization of fault zones for reservoir modeling: an example from the Gullfaks field, northern North Sea. AAPG Bulletin, 83, 925–951.
    OpenUrlAbstract
Back to top
Alerts
Citation tools
Permissions
View PDF
Email to
Print
Download PPT
Bookmark this article