Fault seal behaviour in Permian Rotliegend reservoir sequences: case studies from the Dutch Southern North Sea

Abstract Permian Rotliegend reservoir rocks are generally characterized by high net/gross (N/G) ratios, and faults in such sand-dominated lithologies are typically not considered likely to seal. Nevertheless, many examples of membrane sealing are present in Rotliegend gas fields in the Southern Permian Basin. This manuscript reviews examples of membrane sealing in the Dutch Rotliegend; it presents an extensive dataset of petrophysical properties of Rotliegend fault rocks and analyses two case studies using commonly used workflows. Fault (membrane) seal studies have been carried out on two Rotliegend fields to test the level of confidence and uncertainty of prediction of ‘across fault pressure differences’ (AFPD) based on existing SGR-based algorithms. From the field studies it is concluded that observable small AFPDs are present and that these are likely pre-production AFPDs due to exploration-time scale trapping and retention of hydrocarbons. Two shale gouge ratio (SGR)-based empirical algorithms have been used here to estimate AFPDs in lower N/G reservoir intervals with the aim of predicting membrane seal behaviour, and these results are compared to field data. It is concluded the selected SGR-based tools predict AFPD for Upper Rotliegend lower N/G reservoir rocks with reasonable results. Nonetheless, the core sample datasets show a much wider range of permeability and capillary entry pressure than predicted by the selected SGR transforms. This highlights the potential to modify existing workflows for application to faults in high N/G lithologies. Data sharing and collaboration between industry and academics is encouraged, so that in the long run workflows can be developed specifically for faults in high N/G lithologies.

Most gas-producing fields in The Netherlands are situated in reservoirs formed by mixed aeolian and fluvial deposits of the Permian Upper Rotliegend Group. Since the discovery and development of the giant Groningen gas field (in 1959), the Rotliegend gas play is now considered a very mature play (EBN 2017). Ongoing production of many fields towards and into the tail end of their production life has made apparent that discrepancies between the static and dynamic volumes are observed. These discrepancies are related to the presence of vertical baffles within those fields, presenting restrictions to pressure communication and therefore a certain level of compartmentalization, and are in most cases explained by the presence of (partly) sealing faults (van Hulten 1996(van Hulten , 2010van Wijhe et al. 1980) or fault systems (Corona 2005;Geiss et al. 2009). Examples of gas field compartmentalization and the presence of fault sealing have been described in various publications ( Fig. 1; Table 1). The publications outlined in Table 1 clearly identify a high level of variation in fault styles and geometries, trends, and history. From the above references, it is concluded that, at a regional scale, fault sealing causes and mechanisms are understood, but that it remains difficult to predict fault seal on a smaller scale. Understanding the sealing behaviour of faults remains very important because many of the remaining economically attractive exploration prospects in the Dutch onshore and offshore area depend upon structural closure defined by a spill point related to possible fault sealing. Additionally, it is equally important to understand the level of compartmentalization on a field production timescale to allow identification of the economically most attractive appraisal and development scheme (e.g. the number and position of wells to be drilled).
The term 'fault seal' covers a range of situations in which flow across a fault is absent, or hampered, including those situations where:  Table 1. K. VAN  (1) low-permeability rock is juxtaposed against higher permeability rock at the fault face ( juxtaposition sealing); (2) situations where faults support large hydrocarbon columns over geological time; and (3) those situations where faults act as minor or major production baffles. The most common usage of the term 'membrane seal' refers to those situations where fault sealing relies on capillary processes.
An overview will be presented here of Southern North Sea fields where the likely presence of membrane sealing has been confirmed by data collected in exploration and production wells ( Fig. 1): for example, in the form of observed unexpected free water level depth differences across faults or the lack of dynamic pressure support across fault zones.
Then follows a summary and review of data from detailed core analyses carried out on a selection of fault and host rock samples carefully selected from well-core material from Upper Rotliegend intervals, and for which data have been used for calibrating existing predictive property transformation functions.
Thirdly, two case studies of Rotliegend fields from the Dutch offshore are presented where available data strongly suggest the presence of membrane sealing across major faults within the field. These case studies are: (1) the L12B-C Field (operated by Neptune Energy); and (2) an anonymized field in the Southern North Sea area, here referred to as the SNS-A Field. Data from these fields have been used to validate two commonly used fault rock property transformation functions, those of Sperrevik et al. (2002) and Bretan et al. (2003).

Stratigraphy and palaeogeography
The Rotliegend gas play (Figs 2 & 3) is a textbook example of the superposition of three key components of hydrocarbon plays: (1) prolific Late Carboniferous coal-rich source rocks for gas; (2) laterally very extensive sheets of thick sandstones forming the Slochteren Sandstone reservoir; and (3) the thick and continuous evaporitic Zechstein presenting an almost perfect top seal (de Jager & Geluk 2007). The most prospective area for hydrocarbons is located in an east-west-orientated fairway that stretches from the offshore UK across The Netherlands and Germany into Poland. Along the southern edge of the Southern Permian Basin, this fairway is formed by the presence of a mixed fluvial and aeolian facies belt (Gast et al. 2010). Within The Netherlands, towards the north, the Rotliegend rapidly thickens. The centre of the Rotliegend Basin was formed by an east-west-trending axis located across the northern part of the Dutch offshore, and where the largest total thickness of the Rotliegend has been attained in access of more than 1.5 km. The northern boundary to the Southern Permian Basin is formed by an aligned series of highs, including the Mid-North Sea High and the Ringkøbing-Fyn High (Pharaoh et al. 2010).
Stratigraphically, the Upper Rotliegend (Fig. 2) can be subdivided into at least two genetically linked depositional cycles: (1) a lower cycle bound by the top of the 'transgressive' Ameland Member at the top, and the base of the Lower Slochteren at the base, the latter concurrent with the Base Permian Unconformity; and (2) an upper cycle bounded by the base of the transgressive Copper Shale at the top, and the top of the transgressive Ameland Member at the base, including the Upper Slochteren and Ten Boer members, and their time-equivalent deposits (van Adrichem Boogaert & Kouwe 1993-1997George & Berry 1997;van Ojik et al. 2011). The sand content within each of the claystone members shows a gradual increase towards the basin margin and, conversely, the sandstone members demonstrate an increase in shale content in a basin-centreward direction. The centre of the Southern Permian Basin is characterized by the deposition of thick series of claystones intercalated with halite beds deposited within the Silverpit Lake, although this was more likely not to be one single lake but a system of interlinked smaller perennial saline ponds. Overall, the Rotliegend shows a pattern of increasing expansion of the Silverpit Lake from old to young, and regressive patterns of gradual back-stepping depositional systems, causing the sand-prone deposits of the Lower Slochteren Member to be present further towards the north compared to the sand-prone deposits of the Upper Slochteren Member. Figure 3 shows the present-day distribution of the Upper Rotliegend deposits. V shale , net/gross (N/G), burial depth and high-porosity contours have been obtained by convergent interpolation and contouring of data from the exploration and appraisal wells available.

Structural setting and burial history
An understanding of the fault and fracture systems present in the Upper Rotliegend rocks, their relationships to fault rock and surrounding host rock properties, and consequently sealing potential, requires an understanding of the regional tectonic evolution of the area of interest. We present a high-level overview here: for more detail, the reader is referred to Ziegler (1990), Leveille et al. (1997), Corona (2005), ), de Jager & Geluk (2007 and Ligtenberg et al. (2011), and references therein.
An overview of typical seismically observable fault patterns in the Rotliegend is presented in Figure 4, which exemplifies the various phases of fault  Geiss 2008). These fault systems in Rotliegend deposits have gone through several phases of reactivation (Ligtenberg et al. 2011). Regional subsidence Fig. 3. Simplified map of the present-day distribution of sediments of the Upper Rotliegend Group in the central Dutch offshore and north onshore area. Included here are oil and gas discoveries in the Rotliegend, and all released wells which have completely penetrated the Rotliegend. Numbers refer to fields/areas collected in Table 1 where membrane seals are identified. The dashed black line represents the 100 m isochore contour (thinner towards the south, thicker to the north), the stippled grey lines represent iso-V shale contours (average V sh . 0.75 towards the north, eastwest stretching belt with an average V sh of between 0.25 and 0.75, and an average V sh , 0.25 towards the south). The green filled polygon represents the area where the top of the Rotliegend is buried at present-day depth shallower than 3 km, and the yellow filled polygon represents the area where the average porosity is larger than c. 15%. due to the ongoing relaxation of a weak and thin lithosphere resulted in the formation of the very large Southern Permian Basin (van den Belt 2007), and provision of the accommodation space in which the sediments of the Upper Rotliegend system were deposited (van Ojik et al. 2011). Continuing mild extension during the Permian occurred in a roughly east-west direction.
Transtensional stresses during the opening of the proto-Atlantic Ocean (Early Cimmerian event, c. 230 Ma) were orientated in a roughly ENE-WSW direction causing a mild reorganization of the structural configuration of the Southern Permian Basin. Zechstein salt locally started to move in response to extensional events, showing an increasing decoupling of the mechanical response between the over-and under-burden of the Zechstein.
Ongoing break-up of the Pangaean supercontinent, associated with thermal uplift of the Mid North Sea High during the Early Jurassic is referred to as the Mid-Cimmerian event (c. 175 Ma). This is a generally NE-SW-orientated transtensional phase, and in which the main structural features of the Dutch subsurface formed are amplified in the development of large-scale graben systems such as, for example, the Broad Fourteens Basin and the Vlieland Graben. Pulses of repetitive extension related to the ongoing opening of the Atlantic Ocean and Overview of the regional tectonic kinematic history in time (horizontal axis), and approximate depth of burial and temperature (at base Rotliegend) at the time of deformation (vertical axis). Red arrows represent the directions of compressional events; green arrows represent extensional events (modified after Ligtenberg et al. 2011). break-up of Laurentia caused roughly east-westorientated extension of major graben systems, a phase referred to as the Late Cimmerian event (c. 145 Ma). Decoupling of the structural response to tectonic activity between the Zechstein overburden v. under-burden, and the absence of Zechstein salt in the southern part of the Dutch onshore and offshore, caused a strong contrast of structural styles in response to the Late Cimmerian event. In northern parts of the Southern North Sea, strongly continuous north-south-trending faults in trend with the Dutch Central Graben can be observed at Rotliegend level, whereas in the South, where Zechstein salt is absent, the pre-existing structural grain was reactivated.
Closure of the Tethys Ocean, and the collision of the African and Eurasian plates, finally caused a series of compressional pulses during the Late Cretaceous and Early Tertiary. This lead to the inversion of existing Upper Jurassic and Early Cretaceous basins (de Jager & Geluk 2007). Seismic expressions of this compressional event are the local presence of reverse faults, (over) thrusting and pop-up features along major NW-SE trends. Conjugate NE-SW-trending faults can be observed in which transfer movements accommodate some of the oblique inversion along the NW-SE trends. These NE-SW/ NNE-SSW fault trends, referred to as 'Dekeyser' faults (Dekeyser 1990), are linear and semicontinuous over large distances (50-100 km long) with only limited lateral offset. They show, however, apparent small throws close to or below seismic resolution. In places, it is a major difficulty to accurately map and/or image these faults (Geiss et al. 2009). Several field studies have demonstrated that these Dekeyser faults act as sealing faults over production time with an 'across fault pressure difference' (AFPD) in excess of 200 bar.
As described by Dekeyser (1990), these features appear to be parallel and regularly spaced (2-3 km), with occasionally rather continuous collapse zones. Geiss (2008) and Geiss et al. (2009) describe two geometries of these lineaments as seen on seismic: (1) as single, subvertical, fault planes sometimes with large throw; and (2) as two opposing fault planes creating narrow graben systems, also referred to as thin-skinned graben (Vendeville & Jackson 1992) or skinny graben (Leveille et al. 1997), depending on their width (down to one seismic trace, i.e. less than 25 m). Analysis of throw profiles shows a strong variation on a hectometre scale, suggesting a more complex segmented structure at scales beyond the seismic resolution. Similar complex fault geometry observations are documented by Corona (2005), and also by Leveille et al. (1997) who claimed that evaporites may have infiltrated from the overlying Zechstein and therewith contributed to the sealing potential of these fault systems.
More detailed structural analysis of the Cleaver Bank High area (which in the north partly overlaps the area covered in Oudmayer & de Jager 1993, Schroot & De Haan 2003and Chen 2015 supports that NNE-SSW and NE-SW fault trends show anomalously high length-to-throw ratios, and authors explain this by repetitive reactivation of much older Variscan fault structures during the Mesozoic and Cenozoic.

Field examples of membrane seals at Rotliegend level
Compartmentalization of Rotliegend gas accumulations was identified soon after the first Rotliegend gas fields came into production (van Hulten 1996). Figures 1 and 3 and Table 1 document examples of membrane seals at the Rotliegend level. Evidence for the presence of these membrane seals is provided by the presence of the free water level (FWL) depth and formation pressure differences across faults (at pristine conditions). For some of these fields there are no public data available to further follow-up and better understand the importance of membrane sealing. Some more multidisciplinary integrated and robust field (and fault) reviews were carried out where fields where studied in light of all data available, including petrophysical analysis of fault rock data (see Table 1) subsequently used in material balance and dynamic production history matching calculations. A few proprietary field reviews are available to authors where empirical shale gouge ratio (SGR)-based functions were used to predict fault capillary entry pressures and expected hydrocarbon column heights (see Table 1). The SGR function is based on the average host rock clay content which passes the calculation point on the fault. The SGR estimate of the fault rock clay content is then used as a basis for fault rock property assessment. In the following, it will be questioned whether using the SGR algorithm is a valid assumption for faults in high N/G Permian Rotliegend reservoir rocks, in particular in view of the fault displacement processes.

Fault seal prediction
In the hydrocarbon industry, fault seal studies play an important role in the evaluation of hydrocarbon traps to understand cross-fault flow and retention capacity, not just over geological timescales (most relevant for exploration) but also over production timescales in relation to field compartmentalization and differential depletion. Several fault seal analysis techniques have been developed in last decades with subsequent minor modifications since. The most common methodology consists of: (1) the construction of a discrete fault and horizon framework model, based on seismic interpretation, and their related horizon-fault and fault-fault intersection lines; (2) careful geometrical analysis to make a distinction between areas of juxtaposition sealing (where reservoir rock is juxtaposed against nonreservoir rock at the fault face) and potential leak windows at areas of reservoir-reservoir juxtaposition (Allan 1989;Knipe 1997); and (3) a prediction of the height of the hydrocarbon column that can be maintained by the fault seal through the process of membrane sealing (Bretan 2017). Ideally, fault properties such as permeabilities and entry pressures based on core data are used for these predictions, but such data are typically not available due to the lack of core material. In areas and intervals with a higher variability in N/G ratios, the maximum hydrocarbon column height is typically estimated based on the fault displacement and clay content of the host rock using the SGR algorithm (Bretan et al. 2003) or comparable type of transformation (Lindsay et al. 1993;Yielding et al. 1997;Yielding 2002;Freeman et al. 2010).
The SGR algorithm is a very useful approach for conditions with lower N/G rocks where shale-rich rocks will be incorporated into the fault zone. In the case of higher N/G rocks, different displacement processes will take place including grain reorientation, crushing, dissolution and cementation of quartz or other diagenesis, and which will affect fault properties, which is beyond the application of SGR. Stress and temperature in relation to geohistory play a very significant role here but relevant data for the evolution of fault rock properties with temperature and pressure needed to study and understand this better are limited.
Upper Rotliegend reservoir sediments of the Southern Permian Basin are typically high N/G rocks, and fault characterization based on visual inspection of core material in the form of slabs and chips suggests that these faults are indeed dominated by the presence of deformation bands and cataclasites formed under complex structural conditions (Mauthe 2003;Fisher et al. 2005;Ligtenberg et al. 2011;Busch et al. 2015). It should be noted that these observations are generally made on relatively small-scale structures (centimetres of displacement), which are not necessarily representative of the properties of seismic-scale faults. Seismic-scale faults interact with a greater amount of stratigraphy, so they are more likely to intersect sparse shale beds than centimetre-scale fractures. Shipton et al. (2006) show from outcrop data that shale beds are not well mixed in large fault zones, and can therefore play a disproportionately small role in clay smearing and fault seal behaviour. The dominance of cataclasis in the dataset, the examples of observed AFPDs for faults with reservoir-reservoir juxtaposition of high N/G rock and the poorly constrained property prediction at low SGR values pose the question of whether current shale gouge or clay smear functions should be used to predict sealing capacity of fault and fracture systems hosted in the Upper Rotliegend. The goal of this paper is to share some observations with respect to the limitations of output of certain publicly available transformations, and the need to better understand the evolution of faults and fractures in relation to their surrounding host rock properties and pressure and temperature history.

Fault rocks
The petrophysical properties of fault rocks (e.g. permeability, capillary entry pressure) are fundamental factors controlling the ability of fault rock to sustain pressure communication across the fault. These petrophysical properties depend on a large range of subsurface processes including variations in sediment composition, stress and temperature history during the complex geological history of the Southern North Sea Basin (Fisher & Knipe 1998).

Cataclasis and deformation bands
Subsurface data across various scales (including seismic data, borehole image logs and well cores) have revealed that faults in the area of interest are frequently composed of or associated with a zone of larger and smaller faults and fractures referred to as fault damage zones (e.g. Frikken 1996;Fisher & Knipe 1998;Ligtenberg et al. 2011;Busch et al. 2015) with inherent vertical and horizontal permeability variations. One of the intrinsic problems with these subsurface data is the limitation of the integration of observations and data into robust concepts across the various scales. Up-and downscaling of fault rock properties (notably permeability) would ideally allow confident prediction of transmissibility multiplier ranges used in dynamic modelling for production history matching and forecasting. This would, however, require a representative set of wells drilling through a seismically resolvable fault zone whilst acquiring necessary data across that fault zone such as core, image logs, wireline data; data that have very limited availability in the public domain.
For the current purpose of understanding permeability and fluid flow through a fault zone, representing a series of deformation bands in porous rock, the subdivision provided by Fossen et al. (2007) is considered most useful. Their classification is based on the dominant deformation mechanism, allowing the identification of four principal types and using terminology that will be used in current paper. These four types are: (1) Disaggregation bands, which form in a granular flow process in which grain rolling, boundary sliding and minor breaking occur.
(2) Phyllosilicate bands, where clay minerals promote grain boundary sliding. (3) Cataclastic bands, which occur when grains fracture and break (Aydin 1978). (4) Solution and cementation bands, where dissolution and cementation occur along a deformation band. Different deformation mechanisms produce bands with different petrophysical properties, such as permeability and threshold pressure, which are relevant parameters when modelling membrane seal behaviour.
On a core scale, the various fracture types observed include deformation bands (including cataclastics), cemented fractures, shale smears, phyllosilicate framework faults and open fractures, albeit the most common types in the Upper Rotliegend are cataclastic and cemented fractures (Ligtenberg et al. 2011). Detailed core laboratory analysis of fractures in cores (see further in this paper) has shown that both cemented and cataclastic fractures have the potential to hold significant pressure differences.
Deformation bands experience strain hardening, and therefore they can only accumulate very limited offsets (centimetres at most). Progressive deformation is first accommodated by the formation of multiple deformation bands (e.g. Antonellini et al. 1994;Shipton & Cowie 2001 after which localization of deformation leads to clustered zones of deformation bands. Subsequent deformation of these zones leads to the development of slip surfaces and the formation of cataclastic fault cores of centimetreto metre-scale thickness. This architecture, which consists of a fault core consisting of cataclasites, clustered deformation bands and slip surfaces surrounded by a damage zone with deformation bands, is representative of many seismic-scale faults (.20 m offset) in porous sandstone. Both the damage zone and the fault core can therefore act as a barrier or baffle to across-fault flow, while, at the same time, a well-developed slip surface can act as a pathway for along-fault fluid flow (Shipton et al. 2002).
Diagenesis and cementation. Cataclastic bands are the dominant fault rock reported within core from the Rotliegend (Leveille et al. 1997;Mauthe 2003;Fisher et al. 2005;). Grain-fracturing induced porosity collapse and alongside enhanced quartz cementation has resulted in these cataclastic faults having lower permeabilities and increased threshold pressures compared to the surrounding host reservoir rock. It should be noted that other diagenetic minerals such as anhydrite, barite and carbonates are encountered within Rotliegend faults. Grain-size reduction caused by shearing facilitated pervasive quartz cementation, promoted by the large grain surface area and the availability of reactive fractured surfaces; see Knipe et al. (1997), Fisher & Knipe (1998),  and Lander & Laubach (2014) for more details. Quartz solution and reprecipitation may start at temperatures of around 70°C, and typically accelerates where deformation takes place at temperatures greater than 90°C (Walderhaug 1996;Leveille et al. 1997).

Microscopic-scale measurements of fracture properties
Core analysis data and images are available from fault samples (Table 2) from wells across the UK and Dutch offshore, and Dutch and German onshore, including measurements of fracture and host rock permeability and mercury injection capillary pressures (Table 3). Some of these data have been published earlier (Leveille et al. 1997;Mauthe 2003). Use of data measured by Fisher et al. (2005) and Fisher (2006) in this paper has been approved by operators NAM and Total (and partners) (Appendix A), and includes measurements of porosity, permeability, mercury injection pressure on both host rock and fault rock core samples, clay content estimates based on XRD analysis, and various thin section and SEM imagery. Data used by  was unfortunately not publicly available, and values have been estimated from figures in that paper.
Permeability. Based on the available petrophysical data from selected samples (Fig. 5), the fault rock permeability at core-plug scale varies from 0.001 to 0.05 mD (geometric mean + one standard , and this range of permeability will depend on the level of intensity of cataclasis and cementation. Permeabilities measured in fault rock specimens here are two-four orders of magnitude lower than the host rock (Fig. 6). Empirical transformations for fault seal prediction used within the context of this study (Sperrevik et al. 2002;Bretan et al. 2003) are typically based on host rock clay content and fault rock permeability or injection threshold pressure relations, albeit these empirical relationships are based on core rock and field data from the Brent province. These Brent rocks are much younger (Middle Jurassic), and have been deposited in a shallow-marine, marginalmarine and non-marine environment, and, hence, have a different petrographical composition and are typically lower N/G rocks with higher amounts of clay minerals. In addition, they have been subject to a different burial history than the Rotliegend. Output of SGR-based algorithms, such as those by Sperrevik et al. (2002) and Bretan et al. (2003), under more poorly constrained conditions of low SGR values should therefore be used with care to predict membrane seal capacity in Rotliegend rocks.
A cross-plot of fault rock permeability against clay content for Rotliegend fault rocks is provided in Figure 7, several SGR-based transformations are included as lines. The spread in permeability of cataclastic samples is likely to be associated with the level of cataclasis and deformation in those samples, but requires further study to allow firm statements. The limited number of samples within the cemented samples appear to group together.
Transformations based on SGR, such as those provided by Manzocchi et al. (1999), Sperrevik et al. (2002), Bretan et al. (2003) and Jolley et al. (2007), typically aim to predict fault permeability, which is then subsequently transferred to transmissibility multipliers for dynamic simulation. This transfer is based on parameters such as fault throw and Table 3. Statistical overview of core analysis data available from samples from wells in the UK and Dutch offshore, and Dutch and German onshore areas (Fisher et al. 2005;Fisher 2006 (Dekeyser 1990). Fig. 6. Pairs of fault rock v. host rock permeability for selected core samples from the Rotliegend in British, German and Dutch onshore and offshore wells (Leveille et al. 1997;Mauthe 2003;Fisher et al. 2005;Fisher 2006). Fig. 7. Cross-plot of host rock clay content v. fault rock permeability for rock samples collected from Rotliegend core (Fisher et al. 2005;Fisher 2006). Symbol colour is indicative of the type of fault rock; the symbol size is proportional to present-day burial depth. Included are several common public transformations between host rock clay content and fault rock permeability (Manzocchi et al. 1999;Bretan et al. 2003;Jolley et al. 2007;Fisher 2015).
fault width with inherent uncertainty difficult to quantify given the current limited resolution of seismic data. In addition, reorganization and/or amplification of fault and fracture networks due to other tectonic events, such as the Late Cretaceous inversion event, are not taken into account here. Validation of these fault seal models can then be carried out through dynamic history matching and (flowing) material balance calculations.
Mercury injection threshold pressure. The mercury injection results from the cataclastic faults collected over the last decades and available for the current evaluation show a considerable range between 4.24 and 40.8 bar (geometric mean + one standard deviation), but reflects samples with varying intensity of cataclasis and deformation. Injection threshold pressures corrected for in situ conditions (Adams 2016) are plotted against modelled maximum burial depth (Nelskamp et al. 2014) at which rocks have been buried during their geological history (Fig. 8). Inclusion of AFPD data from lower N/G rocks from the two case studies presented within this paper (L12b-C and SNS-A) plot out of trend and are difficult to reconcile with the available petrophysical core rock measurements, which are dominated by cataclastic faults from higher N/G conditions. Available injection threshold pressure data from core rock material (Fisher et al. 2005;Fisher 2006) plotted against average host rock clay content of those samples with SGR-based empirical relationships provided by Sperrevik et al. (2002) and Bretan et al. (2003) in backdrop ( Fig. 9) yield no correlation at all. This once again supports that output from SGR-based algorithms are probably invalid for fault seal predictions in high N/G Rotliegend reservoir rock which are likely to be dominated by cataclastic faults.

Validity of output of transformations and functions
At the Rotliegend Field scale, membrane seal calculations are frequently made for those situations where there is gas fill on both sides of a fault but the FWL is different on both sides. Underschultz (2007) described three fundamentally different pressure patterns for this situation. In this paper we will Fig. 8. Cross-plot of in situ (gas-brine system) capillary entry pressures (in bar) v. reconstructed maximum burial depth of fault rock samples from core after Fisher et al. (2005) and (Fisher 2006). Included are two data points from fields where AFPDs (in bar) have been estimated based on well pressure data (see the text for the explanation).P t is the threshold pressure.
focus on his Case 9 (discontinuous gas phase and different FWLs on both sides of a fault, but a constant water pressure gradient), as there are no Rotliegend fields in The Netherlands with tilted FWLs, and only one or two fields with active aquifer support. This situation (Case 9) is caused when the aquifer is hydraulically connected around or through the faults below the FWL (Fig. 9).
For the assignment of properties to dynamic grid-cell boundaries that represent fault planes, and subsequent translation into grid-cell transmissibility multipliers, two different types of properties may be numerically estimated: fault permeability and threshold pressure. As previously explained, the modelled capillary entry pressure of the fault plane (and, consequently, the maximum gas column height) equates to the amount of AFPD (at virgin conditions) which can be relatively easy validated in the presence of reliable pressure data collected in wells drilled on either side of a fault (Fig. 10). For the current project, it has been decided to focus on and validate the algorithms established by Sperrevik et al. (2002) and Bretan et al. (2003) since these two algorithms are available in the most commonly used fault seal evaluation software, and to our experience they are the most frequently used algorithms.
Predicting the maximum gas column height at either side of the fault is usually based on the transformation of the shale content of the fault zone, expressed as SGR to the injection threshold pressure. At least three different relationships have been published in the literature (Yielding et al. 2010;Bretan 2017) and are based on: (1) the empirical relationship between the clay content of the host rock, the amount of fault throw, burial depth and the AFPD (at the same reference depth level on either side of the fault) (Bretan et al. 2003); (2) the empirical relationship between the clay content and the threshold pressure derived from laboratory-based injection tests on fault rock samples extracted from core (Sperrevik et al. 2002); and (3) the empirical relationship between the clay content, fault throw and buoyancy pressure (Yielding et al. 2010).
Two case studies of Rotliegend fields are included in this paper for which the empirical relationships between shale content, fault throw and AFPD based on the Sperrevik et al. (2002) and Bretan et al. (2003) functions have been calibrated and validated against actual well and field data. Based on those two functions, and the average distribution of Rotliegend reservoir properties required for those functions (V f . 5%, SGR , 25%, Z f . 2500 m, Z max , 4500 m), it is expected that the capillary entry pressures and therefore the AFPD ranges may vary between 2 and 16 bar (Sperrevik et al. 2002), and 0 and 4 bar (Bretan et al. 2003) (where SGR is the shale gouge ratio, V f is the shale volume, Z f is burial depth at which fault structural deformation occurred and Z max is the maximum burial depth).

Case studies
Data from the two Rotliegend gas fields presented here have been studied in more detail and compared to outcomes of two SGR-based empirical relationships between the clay content of the host rock and the AFPD, notably those by Sperrevik et al. (2002) and Bretan et al. (2003). These two fields are L12b-C (operated by Neptune Energy) and SNS-A (anonymized), which are both located in the Dutch offshore area (Fig. 4). These fields have been selected based on the availability of sufficient data, including wells positioned on either side of a (partially) sealing fault, relevant well data including wireline logs (gamma-ray, sonic and density logs), formation test pressure data and historical production data. Both fields are covered by 3D seismic of good imaging quality. The L12b-C top reservoir is buried to a present-day depth of c. 3 km and the top reservoir of SNS-A to c. 4.5 km, allowing the comparison of results of, in particular, Sperrevik et al. (2002)'s function, which strongly depends on the maximum burial depth and the reconstructed depth at the time of deformation.

Case study 1 (L12b-C)
The L12b-C Field, operated by Neptune Energy, is located c. 5 km from the coastline. The trap is a combined fault-dip closure, with several fault blocks of Upper Rotliegend sandstone reservoir below a thick sequence of Zechstein evaporites. The field was discovered in 1979 with exploration well L12-3 drilled by NAM into the northern fault block of the field. Appraisal well L15-4 was drilled in the  Fig. 10. Conceptual diagram for a Rotliegend fault seal: (a) cross-section with a discontinuous gas phase and different FWLs on both sides of a fault, but at a uniform water pressure gradient; and (b) corresponding pressuredepth plot with a uniform hydrostatic pressure gradient for the aquifer and different gas pressure gradients for wells A and B, with P t = threshold pressure equals buoyancy pressure. Modified after Underschultz (2007).
Middle/Southern domains within the same structural closure, and the pressure data acquired suggest the presence of a different, deeper FWL to that of the northern block. In L12b-C, the reservoir sequence is formed by the presence of an 'upper' (Slochteren B) and 'lower' (Slochteren D) sand-prone unit sandwiched between more clayey units (Slochteren A, C and E) in the Upper Slochteren Member (Fig. 11) within the gas column. A considerable amount of thorough research has been carried out by current and previous operators to understand the dynamic behaviour of the field, which is partly captured by Weijermans et al. (2016).
The field was taken into production after drilling the L15-FA-106 well (abbreviated to the A106 well here) in the northern compartment close to the subsurface location of the original L12-3 discovery well in 2000. The northern compartment is now (as of 2018) depleted to a reservoir pressure of c. 50 bar. In 2014, a second producer (L15-A-108A, abbreviated to A108A here) was drilled into the central part of the field, encountering significant pressure depletion of up to 50 bar (Fig. 12), which can only be attributed to pressure depletion in the northern compartment.
The current case study comprises a high-level cross-check of available data against SGR-based algorithms to verify an alternative scenario in which a membrane seal is introduced between the northern and middle segment.
Available wireline gamma-ray data have been translated into a V sh curve and, combined with lithology interpretations from cuttings descriptions, this enables a discrete subdivision of rock into three classes: 'sand' (V sh ≤ 0.4), 'silt' (0.4 , V sh , 0.5) and 'shale' (V sh ≥ 0.5). Reservoir sections are predominantly composed of rocks with low V sh values, but a significant amount of shale is present within the complete section. It is possible that these shales may be taken up in fault zones; hence, an SGR-based approach might work here.
The juxtaposition triangle plot of the critical fault between the northern and middle segments (Fig. 13) is based on the lithological subdivision of well L12-3, and suggests that at fault throws between 0 and 10 m the sandy Slochteren B, and between 0 and 20 m the Slochteren D, units are self-juxtaposed. At fault throws of between c. 20 and 50 m the Slochteren D in the hanging-wall block is juxtaposed against Slochteren B in the footwall block. At fault throws between c. 10 and 30 m, and between 50 and 60 m there are mainly juxtapositions of good sandy reservoir rock against silty rock with worse reservoir quality; and, hence, limited likely acrossflow capacity at the fault face. At fault throws larger than c. 60 m there is no relevant reservoir rock selfjuxtaposed across the fault and therefore the fault will act as a juxtaposition seal for a cross-fault seal. Vertical seismic resolution is around 20-30 m and, hence, introduces a significant uncertainty to the amount of throw.
Cells of a 3D grid (width:length:height of cells, c. 50 × 50 × 1 m) were populated with a discrete lithology based on the extrapolation of upscaled V shale properties at the intersection of wells with the 3D grid, by using a lithology subdivision as above. This lithology grid was used to identify a juxtaposition property at each of the fault faces available in the 3D grid, resulting in six different lithology juxtaposition combinations. Figure 14 shows a view towards the north at the fault face of the eastwest-orientated fault dividing the northern and the Central Domain, with nearby wells A108A (in front of the fault), and L12-3 and A106 (behind the fault). Analysis of the 3D seismic indicates that across a significant part of the fault the Slochteren B and D are self-juxtaposed above the FWL, but also that roughly between the A108 and L12-3 well an area with significant fault throw exists in which the Slochteren B ('upper Sand') in the hanging-wall block in the south is juxtaposed against the Slochteren D ('lower Sand').
Several formation pressure data points have been acquired in the various wells within the L12b-C Field, both at virgin conditions, and at a time of significant pressure depletion in the Northern Domain, allowing the interpretation (within reasonable uncertainty limits) of the gas pressures and gradients within the field (Fig. 15). Due to the absence of reliable pressure data from the aquifer, the hydrostatic pressure gradient has been interpreted based on data from nearby fields (L12b-A and L12b-B). This in turn has allowed for the interpretation of FWLs, and consequently, based on wireline log evaluation and special core analysis data, the gas saturation profiles and depth of the gas-water contacts (GWCs) (see Weijermans et al. 2016 for a more detailed interpretation). Based on the presence of different FWLs on both sides of the dividing fault between the Northern and Middle domains, an AFPD of c. 4 bar can be reconstructed.
Based on the conversion of Hg-injection threshold pressure data of Rotliegend fault rock material explained earlier (after Fisher et al. 2005), an AFPD of less than 3 bar (at in situ conditions for the gas-water system) would be expected under the current conditions (at a relatively shallower presentday burial depth of c. 3 km). The observed AFPD of 4 bar slightly exceeds that depth trend. It is worth noting that measured permeability data from fault core material from deformation-band-dominated faults in Utah is lower than deformation bands from the same fault's damage zone (Shipton et al. 2002), due to more intense cataclasis in the core alongside local grain-contact dissolution of quartz. It may be reasonable to assume that similar  Previous interpretations of the FWLs and the level of expected pressure equilibrium across the field prior to drilling the L15b-A108A well were explained by the operator through a model in which the fault between the Northern and Central domains was fully closed (Weijermans et al. 2016). In this model, after initially sharing the same (palaeo-) FWL across the field, the northern and Central/Southern domains became isolated due to Fig. 12. NW-SE cross-section over the L12b-C Field illustrating the structural compartmentalization, due to faults, into Northern, Central and Southern domains, the position of the four wells drilled into the field, and the depth of the FWL across the field based on well observations. The position of the dividing fault between the Southern and Central domain in relation to well L15-4 is subject to interpretation: based on seismic data, several structural framework scenarios can be identified here, putting the well either in the Central Domain or the Southern Domain.   strike-slip movement probably of Late Jurassic age (c. 150 Ma) and the resulting cataclasis at the dividing east-west-trending fault(s). Structural tilting and/or seal breaching in the northern compartment then caused different FWLs and possibly the presence of gas composition variations across the field.
In addition, depth differences in the GWC and, consequently, the transition zone between the FWL and GWC have been explained by Weijermans et al. (2016) to reflect strong permeability variations across the field, deteriorating from north to south. Formation pressure data collected in well B108B, however, proved not only the presence of different FWLs, but also a significant amount of pressure depletion due to production in the Northern Domain. These data strongly resemble a situation described earlier by Underschultz (2007) in his Case 9, a model in which discontinuous gas phases and different FWLs on both sides of a fault are present, but at a constant hydrostatic pressure gradient. Gas phases in the L12b-C Field are in pressure equilibrium due to the presence of a membrane seal acting as a valve with capillary entry pressure of c. 4 bar. Seismic imaging quality of the fault between the central and southern compartment is rather limited, and an alternative concept can be presented in which the reservoir section of the L15-4 appraisal well has been drilled north of that fault, in the Central Domain. In this alternative concept, the fault between the Central and Southern domains is trending in an almost east-west direction, structurally very similar to the sealing fault between the Northern and Central domains, suggesting that these two faults may have gone through a similar geological history; hence, exhibiting comparable fault rock properties and, thus, sealing potential. Provided that reservoir rock self-juxtaposition is present across that fault within the gas column, it is plausible to suggest that any gas within the Southern Domain is in pressure communication with the Central Domain, albeit across a membrane seal that potentially causes different gas phases and FWLs on both sides of that fault, similar to the situation encountered in the northern part of the field.
Empirical functions to estimate the seal failure envelopes relating the SGR to the fault-zone capillary entry pressure as a function of burial depth (Bretan et al. 2003) and depth of deformation (Sperrevik et al. 2002) have been compared against the observed field data (Fig. 16).
It appears that both functions plausibly predict capillary entry pressure levels within the expected uncertainty ranges, although under base-case conditions the function by Bretan et al. (2003) slightly underestimates the threshold pressure (minimum capillary entry pressure of 2 bar v. AFPD of c. 4 bar) and the function by Sperrevik et al. (2002) slightly overestimates it (minimum capillary entry pressure of c. 5 bar v. AFPD of c. 4 bar). It should be noted that here only the uncertainty ranges due to variations in clay content have been included. The function by Sperrevik et al. (2002) relates seal failure to laboratory-based Hg-injection entry pressure measurements and thus includes a conversion Fig. 16. Depth-pressure cross-plot including an estimation of the capillary entry pressure profiles of one of the pillars at the east-west-trending bounding fault between the northern and central compartments. The depth interval represents the approximate area of self-juxtaposition of the lower reservoir unit B. Entry pressures profiles are based on Sperrevik et al. (2002) (red curve at the right-hand side) and Bretan et al. (2003) (green curve at the left-hand side) including their inherited uncertainty ranges. The vertical black line represents the AFPD measured between wells on either side of the fault, with the hatched area representing the uncertainty range of these AFPD measurements.
to gas-water subsurface conditions including the interfacial tensions of air-mercury and gas-water. The gas-water interfacial tension, however, is not accurately measured here and may range between 40 and 60 dyne/cm, thus introducing an additional uncertainty.
From this case study it is concluded that within the L12b Field a membrane seal could be present between the northern and middle segment. SGR-based algorithms of Sperrevik et al. (2002) and Bretan et al. (2003), which could work here in view of a lower variation in fault rock permeability at higher clay contents within the section, predict AFPD reliably compared to measured AFPD.

Case study 2 (SNS-A)
Despite anonymization of this case study due to data confidentiality, it has been added here as it offers possibly a view on membrane seal capacity at greater burial depth (between c. 4500 and 4600 m), and is associated with larger formation pressures and temperatures compared to the previous case study. The 'SNS-A' Field is located c. 60 km NW of the Dutch coastline. The trap is formed by a combined three-way dip and fault-closed structure, and comprises several compartments. The relevant compartments here are referred to as SNS-A-BX and SNS-A-BY. The reservoir sequence of SNS-A is provided by mixed fluvial and aeolian sandy deposits of the Lower Slochteren Member, overlain by sealing claystones of the Silverpit Formation, and evaporitic sequences of the Zechstein Group. The SNS-A gas field was initially appraised with well WA, drilled into block BX.
Appraisal well WB was drilled 1 year later in the southern part of the BY block, followed 4 years later by production well WC. The WC well was temporarily suspended due to technical problems and re-entered 1 year later to be completed as production well WD into the northern part of the BY block (see the schematic base map of the field in Fig. 17). Both blocks have been taken into production, albeit block Y 6 years later after block BX. The producing sequence is formed primarily by a c. 60 m sequence of mixed fluvial and aeolian sandstone (Fig. 18), informally classified as Slochteren Alpha, within an c. 125 m-thick, fining-upward sequence of sandstones and siltstones classified as the Lower Slochteren Member (van Adrichem Boogaert & Kouwe 1993-1997. Sandstone beds encountered within the underlying Carboniferous Limburg Group are occasionally situated within the gas leg and contribute to the in-place volumes. Possible contributions to flow and production have not been investigated by the authors. The Slochteren Alpha is a high N/G sandstone but significant shale intercalations are present based on inspection of core images and wireline data. It is expected that deformation bands are primarily present in the form of cataclasites and with subordinate amounts of phyllosilicate-rich deformation bands around those shale intervals. The presence of shale could lead to clay smearing into the fault zone, which has not been investigated in detail here.
The field is fully covered by 3D seismic data with sufficient sub-salt imaging quality to identify principal faults and horizons. The structural framework is characterized by the presence of a conjugate set of NW-SE-and NE-SW-trending faults. The BX and BY compartments are primarily fault-closed structures with dip closure towards the north. A NE-SW-trending fault forms the boundary between the BX and BY segments, and has been the focus of attention in the current study.
The fault offset is largest at the centre of the fault (maximum 200 m), and tapers to small offsets towards both the NE and SW tips (minimum observable offsets within the 3D seismic are c. 30 m). With an average reservoir thickness of c. 30 m, selfjuxtaposition of the Slochteren Alpha reservoir unit across the fault can be observed at both tips of the fault, although only the self-juxtaposed area in the SW is elevated above the FWLs. The impact of seismic resolution has not been investigated here. Based on the assumption of the presence of a hydrostatic pressure gradient similar to pressure gradients measured in the nearby exploration wells and the presence of pre-production pressure data in well WA has allowed a FWL in the BX segment to be interpreted at c. 4730 m TVDSS (true vertical depth below sea level) (Fig. 19). Formation pressure data representative of the BY compartment has been collected in the WB well albeit 1 year after starting production in the adjoining BX segment, allowing the observation of a pressure difference between the wells of c. 4 bar (+1 bar). Several interpretations   19. Formation pressure data plot for the SNS-A Field, including: pressure points from well WA in the southern BX block, and WB and WC in the Northern BY block, and an AFPD of c. 4 bar. In addition, formation pressure data have been acquired later after several years of production from the BX block in well WC located in the northern BY block.
to explain the observed AFPD are presented here. First of all the AFPD could be caused due to pressure communication either across the fault or through the aquifer around the fault and therefore pressure depletion (c. 4 bar) in the BY segment, at which point in time the level of pressure depletion in the BX segment was slightly more than 36 bar. It is expected that the pressure transfer between the two compartments should be visible in the pressure and flow data for both segments, which, however, is not likely to be the case during the first years of production. In addition, very little vertical variation in depletion at the start of production is observed which would be expected in the presence of vertical reservoir heterogeneity, such as demonstrated to be present based on the pressure data collected later in well WC.
An alternative interpretation explains the AFPD at the bounding fault due to membrane sealing between compartments in virgin, pre-production conditions. Well WC was drilled in the BY segment 6 years after the start of production in the neighbouring BX block (then depleted with c. 270 bar), clearly demonstrating (differential) a pressure depletion of c. 45-55 bar across several reservoir layers and supporting the hypothesis of the presence of a semipermeable fault with (limited) pressure communication across that fault. Capillary entry pressure levels of this fault have been calculated based on functions by Sperrevik et al. (2002) and Bretan et al. (2003), and compared against actual AFPD measured in wells (Fig. 20). It is likely that SGR-based predictions may only be valid for limited fault-face areas where intercalated shale layers have been ripped up and shale particles incorporated into the deformation bands.
Based on these assumptions, the capillary entry pressure profile estimated by the Bretan et al. (2003) function is in good agreement with the measured AFPD. The pressure profile predicted by the Sperrevik et al. (2002) function significantly overestimates the membrane seal potential of the fault. A sensitivity analysis carried out indicates the strong dependence of this latter function on primary variations and uncertainty in the maximum burial depth. In addition, few data with a present-day burial depth deeper than 4 km were available to Sperrevik et al. (2002) for their analyses; hence, their functions are less calibrated and output may not be suitable for the depth domain of the current SNS-A Field.

Discussion and Conclusions
Tasks of a fault seal analysis workflow As explained earlier, common tasks within a methodology to identify the presence of a membrane seal Bretan 2017) consist of at least: (1) the construction of a discrete fault and horizon framework, and the identification of several populations and generations of faults and their mutual relationships; (2) careful geometrical analysis to make a distinction between areas of juxtaposition sealing and areas of membrane sealing; and (3) a prediction of the pressure difference (and associated hydrocarbon column height) that can be maintained through the process of membrane sealing. Within the current case studies we have focused particularly on the last task with the aim of predicting AFPDs using two SGR-based algorithms. This has been done under conditions where there is actual well-data control in the fields selected to compare with the predicted AFPD to reservoir pressure data collected in those wells.
The first task of constructing a fault and horizon framework normally incorporates a detailed structural interpretation of the 3D seismic data available with the aim of understanding relationships between the various fault generations through geological time and space. Within the Rotliegend, this is a far from trivial task due to the inherited complex tectonic history, the repetitive reactivation of faults and fault systems in the presence of strong vertical geomechanical heterogeneity and discontinuity, and the limitation of insufficient data availability or lack of resolution to resolve in detail structural deformation mechanisms (i.e. fault movement directions, and amount of throw, timing, internal fault fabric, etc). Several field studies have demonstrated that certain fault generations (Dekeyser lineaments) may act as sealing faults over geological time with AFPDs in excess of 200 bar, and which cannot yet be satisfactorily explained. The amount of net displacement across these Dekeyser lineaments will be relatively small, but with very significant amounts of displacement within the small (skinny) graben on the two opposing fault faces of these lineaments, cancelling out on a slightly larger scale. On a seismic scale, these lineaments are at or below resolution and, hence, are generally mapped as one single event with small offset. Fault throw within the graben system is therefore not captured within the fault interpretation, consequently leading to a possible misjudgement of the amount of juxtaposition sealing, or an underestimate of fault throw, and consequently any sealing properties that are modelled as an SGR-based function of throw. SGR-based functions are therefore not applicable for predicting the level of fault (membrane) sealing associated with these Dekeyser lineaments if offset and crossfault juxtaposition is unknown.
The second task involves a careful geometrical analysis to identify areas of juxtaposition v. membrane sealing: for example, with Allan diagrams including the impact of (sub)seismic vertical and horizontal resolution. This process helps to clarify and quantify the vertical and horizontal distribution of juxtaposition seal as a function of reservoir depth and thickness v. fault throw, which is not necessarily a linear relationship.
A third task embraces a prediction of the expected level of membrane sealing and associated pressure difference that is maintained across the fault during geological time. SGR-based functions evaluated here are established and calibrated against more clayprone shallow-marine sediments of Middle Jurassic age from the Brent area (Central North Sea). They attempt to evaluate the cumulative effect of processes incorporating shale into the fault core (i.e. the bulk effect of processes such as shale abrasion, formation of disaggregation and phyllosilicate deformation bands). It is expected that these processes will only play a role in Rotliegend rocks in the proximity of substantially thick shale layers and the mixing of clay into the fault zone; hence, SGR-based algorithms should be used under these conditions only.
Cataclastic deformation bands are most likely to be the dominant fault rock present within high N/G Rotliegend (sandstone) rock sequences. In outcrop studies elsewhere it has been established that the various deformation mechanisms producing cataclastic bands in analogue rock types are the cause of different internal types of fault fabric and variations in pore throat size and distribution. This will in turn cause significant variations in petrophysical rock properties such as permeability and injection threshold pressure, which are key parameters in the prediction of membrane seal behaviour.
Entry pressure data recorded from Rotliegend (cataclastic) fractures contained within sandstones is difficult to reconcile with AFPDs based on measured well-pressure data. This leads to the conclusion that, indeed, the selected SGR-based algorithms should not be used under conditions where shale material is absent and cataclastic bands are the primary type of deformation.

Uncertainties and sensitivities (precision/ accuracy) of data and transformations
It has been demonstrated that many deformation bands show reductions in permeability (e.g. Shipton et al. 2002;Tueckmantel et al. 2010), some by as much as several orders of magnitude (Fossen et al. 2007). In single-and multi-phase fluid systems, other factors are likely to play an important role as well, but nevertheless host rock and fault permeability appear to be the most important parameters with a practical effect on across-fault fluid flow. As a consequence, many industry workflows for estimating the fluid-flow properties of faults are based on empirical relationships between (fault) permeability and other parameters such as clay content, porosity or permeability of the host rock. Before accepting an estimate of the fault permeability and fluid-flow properties, however, one should be aware of the level of accuracy and precision with respect to the input and output parameters of the various empirical functions used for estimating those fluid-flow properties.
In a rather simple workflow, such as described by Sperrevik et al. (2002), many different conditions may already influence the level of sensitivity and/ or uncertainty (or accuracy and precision) of permeability measurements and estimates. These conditions may be related to (and not necessarily restricted to), for example: (1) in situ, small-scale natural variations of (relative) permeability (including the variability of the permeability within fault zones and deformation bands, the impact of clay content, relative permeability in the presence of multiphase fluid conditions, in situ temperature and stress, etc.); (2) conditions of laboratory measurements and their corrections: stress and temperature, type of infiltration fluid, sample integrity, clay content, etc.; and (3) the correction and transformation of measured permeability (such as the Klinkenberg correction for slippage of gas along pore walls, corrections for stress release when taking core to surface, scale dependency, estimates based on porositypermeability functions, etc.). Estimates of the levels of accuracy and precision of parameters influencing the permeability measurements and predictions may reach a cumulative absolute order of magnitude on a logarithmic scale of c. 10-15 times, and, hence, should be treated with significant care.
Other uncertainties and lack of precision are associated with the amount of seismic vertical and horizontal resolution, fault throw, stratigraphic and sedimentary anisotropy, and discontinuity, as well as fault activity and its timing.
A sensitivity analysis has been carried out on the two transformations used here to identify which input parameters cause uncertainty variations in modelled injection threshold pressure (Sperrevik et al. 2002) or AFPD (Bretan et al. 2003). In Bretan et al. (2003)'s function, the uncertainty of AFPD under average Rotliegend reservoir conditions of burial depth (3500 m), average V f (0.1) or SGR (10%) is defined primarily by the uncertainty of the estimated SGR. Uncertainty of the injection threshold pressure in Sperrevik et al. (2002)'s function is primarily dominated by uncertainty in the estimates of the maximum burial depth (77%) and the burial depth at which structural deformation occurred (21%), followed by uncertainty in the surface tension of the gas-water system at reservoir conditions (1%) and shale volume estimates (1%). For the latter, it means that surface tension data and shale volume estimates do not contribute significantly to the variations in the outcomes.
There are several Rotliegend fields in the Southern Permian Basin in which across-fault variations in reservoir pressure and free water level (FWL) depths have been observed at (close to) virgin conditions. These AFPDs can be explained by the presence of a semi-permeable fault between wells and/or field compartments acting as a valve. Under certain conditions, small pressure differences between wells measured after the field has been taken into production do not necessarily reflect depletion, but may still be interpreted as a result of membrane sealing either pre-production over geological time, synproduction or a combination of all of the above.
Two case studies of membrane sealing in fields with a Permian Upper Rotliegend reservoir have been carried out and results are presented here to validate two selected empirical SGR-based functions predicting capillary entry pressures and therefore AFPDS at virgin (pre-production) conditions. It appears that, within an uncertainty range, both functions tested (Sperrevik et al. 2002 andBretan et al. 2003) plausibly predict expected capillary entry pressures in low N/G reservoir intervals, although some under-/overestimates are observed in relation to the maximum burial depth over geological time.
In both case studies performed there were some indications that the AFPD was different measured on a geological timescale at virgin conditions (0-15 bar) against that measured on a production timescale (100-200 bar). These observations will require much more evaluation to understand better the dynamic behaviour of the host rock and faults in terms of permeability and capillary entry pressure as a function of (production) time.
The case studies presented here and in the literature have demonstrated the occurrence of the fault seal in the high N/G reservoir rocks of the Dutch Rotliegend. There are no published workflows for predicting fault seal over geological time in these lithologies and only a small number of SGR-based approaches are available in key industry software. This paper aims to highlight a clear lack in knowledge, and to act as a call to arms for academics and industry to develop, test and or publish more data to refine and improve tools for faults in high N/G host rock. The approach taken in this paper should ideally be applied across multiple faults, and data pooled from multiple sites to robustly test the hypothesis suggested from these two case studies (Lunn et al. 2008).
Although SGR is usually not recommended for high N/G host rocks, this is not necessarily true for SGR-based transforms. Figure 12 shows the permeability and entry pressure dataset from Sperrevik et al. (2002). Most of the data are for rocks with a low clay content (0-30%) and containing multiple cataclasites. It shows that cataclastic fault rocks may have permeabilities ranging from 1 × 10 −4 to 1 × 10 2 mD and Hg-air threshold pressures of 5-3000 psi. The Sperrevik et al. (2002) function reduces the uncertainty somewhat by including burial depth as a parameter, but this still leaves uncertainties of two-five orders of magnitude for permeability and up to two orders of magnitude for threshold pressure. None of the transforms incorporate this uncertainty, but they return values near the centres of these ranges. Predictions by the selected SGR-based transforms therefore tend to produce reasonable first estimates for faults in high N/G rocks, albeit hampered by several orders of magnitude uncertainty. There is strong potential to develop workflows optimized for fault sealing in rocks with high N/G and this would benefit from more subsurface data released to the public, such as production flow and pressure data, gas composition data, special core analysis data, etc. The above will require multivariate analysis of existing datasets and outcrop studies to derive predictive parameters to minimize the uncertainty in the prediction.
Acknowledgements The authors wish to acknowledge permission from EBN to publish this paper. Opinions expressed within this paper do not necessarily reflect EBN's opinions. We thank Neptune Energy Netherlands and partners for sharing data and information of the L12b-C Field, and meaningful discussions with them. Further, we would like to thank Total E&P Nederland, NAM and partners for providing petrophysical core analysis data of fault samples taken from Rotliegend core from wells across Dutch onshore and offshore. The following Table A1 presents results of petrographical and petrophysical property investigations on samples from Dutch offshore wells (Fisher et al. 2005) carried out for Total E&P Netherlands (TEPNL). Table A2 presents results of petrographical and petrophysical property investigations on samples from Dutch onshore and offshore wells (Fisher et al. 2005) carried out for NAM/Shell. In and out refer to deformation features and undeformed sediment, respectively.