An integrated geological and GIS-based method to assess caprock risk in mature basins proposed for carbon capture and storage

Subsurface injection of carbon dioxide (CO 2 ) is a technique to enhance oil recovery and so the economic value of depleting fields. It complements carbon capture and storage, which is a key technology to mitigate greenhouse gas emissions. In this work, an integrated method developed by the British Geological Survey and Cardiff University uses high-resolution 3D seismic and borehole data from the Jæren High to analyse potential seal breaches and fluid flow paths in a frontier area of the North Sea, ultimately assessing the risk of a possible carbon capture and storage site. We integrate the spatial analysis of subsurface fluid flow features with borehole and geochemical data to model the burial and thermal history of potential storage sites, estimating the timing of fluid expulsion. On seismic data, fluid pipes connect reservoir intervals of different ages. Spatial analysis reveals clustering of fluid flow features above strata grounded onto deep reservoirs intervals. Our integrated method shows that gas matured from Dinantian coal and migrated up-dip during the Triassic-Jurassic into the lower sandstone reservoir of the Rotliegend Group. The containing seal rock was breached once sufficiently large volumes of gas generated high overpressures in the reservoir. Some of these fluid flow features may still be active conduits, as indicated by bright amplitude anomalies within the pipes. This study shows how integrated analyses may enhance our understanding of fluid-flow pathways, de-risking prospective sites for carbon capture and storage. The method proposed in this work is particularly important to assess the suitability of area with trapped gas pockets and understand tertiary migration in areas proposed for geological storage of CO 2.


Introduction
Enhanced hydrocarbon recovery techniques are critical for extracting maximum volumes of hydrocarbons from mature fields. They involve thermal, chemical and gas-injection techniques, one of which comprises the injection of CO 2 -enriched solutions in reservoirs approaching depletion (Thomas, 2008). Carbon Capture and Storage (CCS) is itself regarded as an important technique to mitigate CO 2 emissions. This forms a positive feedback in the energy industry since mature basins such as the North Sea contain viable targets for CCS, at the same time benefiting from a vast database of seismic, well and dynamic production data.
In the North Sea, depleted gas fields have been assessed for combined enhanced oil and gas recovery and CCS, e.g. the Bunter Sandstone in the Southern North Sea (Agada et al., 2017;Williams et al., 2013), and the P18-4 Gas Field offshore The Netherlands (Arts et al., 2012).
The Utsira Sands, located within the overburden of the Sleipner Gas Field (Norwegian North Sea) have been successful in storing CO 2 for more than 20 years, while natural gas is produced (Eiken et al., 2011). Depleted hydrocarbon fields are particularly important for the implementation of CCS; they can successfully store large volumes of CO 2 as demonstrated by their containment of hydrocarbons over geological timescales. In addition, potential CO 2 storage sites may occur adjacently to known producing fields. One of such cases is the regionally extensive Permian Rotliegend Group, which is a saline aquifer in areas where hydrocarbons are not found in economic volumes (Wilkinson et al., 2013). Notwithstanding the latter example, the injection of CO 2 for combined enhanced hydrocarbon recovery and CCS requires a detailed understanding of geological risks and uncertainties prior to the implementation of such techniques, namely the recognition of potential fluid flow paths through caprocks and the sealing capacity in target storage areas. Previous studies predicting the migration of injected CO 2 plumes have indicated that CO 2 can migrate for large distances. Uncontained CO 2 plumes can reach a large number of nearby wells and create risks in terms of well integrity and gas management in many an oil and gas field (Li et al., 2018).
The study area is located on the Jaeren High basement structure; this is a relatively underexplored part of the Norwegian Central North Sea and extends into the UK sector, in which wells target Paleocene successions near large salt diapirs on the eastern margin of the East Central Graben (Fig. 1). In contrast, wells in the Norwegian North Sea penetrate Triassic strata and the Upper Permian Zechstein Group (Fig. 2). Two oil fields have so far been discovered on the Jaeren High; the Gaupe Field to the north of the study area and the Brynhild Field to the south (Fig. 1c). Both contain Upper Jurassic reservoirs above broad, developed salt walls. In the study area, pipes and associated depressions are imaged on 3D seismic reflection data within the Mesozoic succession of the Jaeren High. These features are likely associated with fluid flow and are rooted in deep Paleozoic strata. Hence, our setting contrasts with salt injection features (Clark et al., 1999) and other documented pipes in the North Sea, which typically originate from Cenozoic strata (e.g. Berndt et al., 2003), or are located above salt diapirs (Salisbury, 1990;de Mahiques et al., 2017).
Pipes are narrow, vertical structures that cross-cut seismic reflections. They consist of high-or low-amplitude seismic anomalies with a columnar geometry in three dimensions (Løseth et al., 2011;Cartwright and Santamarina, 2015). Pipes represent a highly focused fluid-flow path from source to seabed (or paleo-seabed), and indicate that seal integrity has been lost and the seal has been breached (Cartwright et al., 2007).
Throughout this work, the term 'pipe' follows the definition above. In contrast, fluid flow features at the upper terminations of pipes are termed depressions due to their specific morphology; the interpretation of these features is made following the study of their morphology and formation mechanisms. High-amplitude anomalies (bright spots) associated with pipes and depressions may relate to the presence of gas, overpressured strata, local lithological variations and cemented intervals (Løseth et al., 2009). Hence, the aims of this paper are to: a) Present a novel method to analyse the spatial distribution of buried depressions and pipes by integrating geographic information systems (GIS), seismic and well modelling, using the example of the Jaeren High; b) Determine the timing of fluid flow in the Mesozoic succession of the Jaeren High; c) Relate the presence of fluid flow features to the local and regional geological settings of the UK and Norwegian North Sea; d) Discuss the implications of seal bypass and associated risks to CCS based on the new method presented here.
Several approaches have been proposed in the literature to characterise areas for carbon capture and storage (CCS), including analytical, semi-analytical, experimental, numerical, and a combination of two or more methods (Crow et al., 2009;Le Guen et al., 2011;Nogues et al., 2012). Geldof et al. (2014) utilised a GIS method for the first time, using 3D seismic data at the seabed level, to study shallow gas migration and subsea geohazards. However, none of the published methods have integrated algorithms embedded in GIS with seismic and petroleum systems data as used in this paper.

Geological framework
The stratigraphic framework of the Jaeren High is summarised in Fig. 2 and Table 1.

Paleozoic evolution
The study area is located on the Jaeren High, a tilted fault block forming the eastern edge of the East Central Graben (Fig. 1). The Central Graben comprises one of three rift arms of Triassic-Jurassic age, centred 50 km northwest of the Jaeren High (Zanella and Coward, 2003). During the Carboniferous widespread peat, fluvial-deltaic sand Structure was generated from maps provided by the Norwegian Petroleum Directorate. c) Two-way-time elevation surface of the Top Rotliegendes. Key wells and fields are labelled. Well names in boxes are used in the burial history models in Fig. 11. Seismic line C-C' is highlighted in Fig. 5. d) Seismic example of a pipe terminating in a pockmark. e) Seismic example of a pipe terminating in a high-amplitude anomaly. A colour bar is included whereby red and yellow seismic reflections are positive (hard), whilst blue are negative (soft). and mud were deposited across a broad plain extending south from the Caledonian Highlands of Scotland and Norway towards the Variscan foreland basin (Glennie and Underhill, 1998). The peat deposits converted to coal during subsequent burial. The Variscan Orogeny followed, which originated the E-trending Northern and Southern Permian Basins filled by desert sand and muds of the Rotliegend Group (Glennie, 1998). During the late Permian-earliest Triassic, extension and subsidence of the Permian basins outpaced sedimentation, resulting in sealevel rise and deposition of the evaporitic Zechstein Group (Underhill, 2003) (Fig. 2).

Mesozoic evolution of the Jaeren High
The region returned to a continental setting in the Triassic, a period characterised by active extension (rifting) in semi-arid, intra-continental basins, and halokinesis in areas with thick salt deposits (Goldsmith et al., 2003). This process is described by Penge et al. (1993) and Alves and Elliott (2014) as rift-raft tectonics. Subsiding Triassic mini-basins (pods) and adjacent salt walls formed as a result of rift-rafting, some of which are 2.3 km high. A positive feedback mechanism ensued until Triassic pods became grounded on the Permian sandstones, forming salt welds. In some cases, the pods were inverted into turtle anticline structures (Karlo et al., 2014). Importantly, grounding of Triassic pods over the Permian strata may have occurred as early as the Middle-Late Triassic on the Jaeren High (Smith et al., 1993).
During the Jurassic and Cretaceous, a mantle hot-spot known as the 'North Sea Dome' was present in the region (Underhill and Partington, 1993). This thermal dome caused regional uplift of an area of 2400 km 2 across the Jaeren High, promoting important erosion down to Lower Jurassic and Upper Triassic strata (Erratt et al., 1999). Subsequent collapse of the thermal dome marked the onset of Late Jurassic deposition, commencing at the furthest reaches of the grabens and younging towards the centre (Rattey and Hayward, 1993).

Late Jurassic to Recent evolution
By Late Jurassic time, rifting and transgression outpaced sediment supply, resulting in the deposition of shallow marine sands and deepmarine shales. This occurred initially in isolated, supra-salt minibasins; then homogeneously across the deepest parts of the Jaeren High (Høiland et al., 1993). Consequently, Upper Jurassic units on the Jaeren High range in thickness from 0 to 200 m, whilst in the adjacent East Central Graben reach up to 1000 m (Fraser et al., 2002).
As active rifting propagated towards the North Atlantic, thermal subsidence resulted in the drowning of the North Sea rift and subsequent deposition of marine sequences of marls, limestones and carbonates during the Early Cretaceous (Cromer Knoll Group; Copestake et al., 2003). In the study area, buried pipes and pockmarks occur in Upper Jurassic-Lower Cretaceous strata, at a depth of 2.5-4 km, revealing episodes of fluid flow during the Mesozoic. Pockmarks were first discovered by King and MacLean (1970) as shallow, crater-like depressions, formed as subsurface water or gas leaked to the seabed, throwing fine sediments into suspension that were dispersed by bottom currents. They have since been widely recorded by multibeam echosounders, sidescan sonar and seismic data (e.g. Hovland and Judd, 1988).
Clastic units were reintroduced into the basin in the Paleocene (Ahmadi et al., 2003), leading to the deposition of > 2 km of strata and the (deep) burial of older rift-related sequences. The detailed seismicstratigraphy of the study area is summarised in Appendix A.

Database development
The interpreted 3D seismic reflection data covers an area of 2400 km 2 across the Jaeren High. It was provided by PGS (Petroleum Geo-Services) and processed in zero phase, SEG (positive) standard polarity, such that an increase in acoustic impedance with depth manifests as a peak, which is coloured red, whilst a trough is negative and blue (Fig. 2). Bin spacing is 50 x 50 m and the vertical sampling rate is 4 ms. Tuning thickness and lateral resolution of the Mesozoic section approach 50 m, using equations defined by Badley (1985). The main focus of this study is the Mesozoic section, from 2 to 4.3 s two-way travel time (TWT).

Seismic interpretation
Schlumberger's Petrel ® was used to interpret the seismic and well data. The adopted methodology is summarised in Fig. 3. Six horizons (H1 to H6) were mapped on seismic data to highlight the pod-interpod structure of the Mesozoic succession and the associated distribution of buried pockmarks (Figs. 2 and 6). Five wells were tied to the seismic data to correlate horizons and constrain the local stratigraphy, of which three wells (6/3-2, 7/1-1 and 7/7-2) are shown in Fig. 4. Isopach maps were generated from interpreted surfaces to highlight lateral thickness changes and the distribution of salt walls and welds.

Seismic attributes
Volume and surface attributes in Petrel were used to constrain and visualise fluid flow features in the 3D seismic volume. Variance (or coherence) is a measure of the waveform similarity, recording how similar a given trace is to its neighbouring trace. Similar traces are mapped as low variance values, while discontinuities have high variance values (Brown, 2011). The variance attribute highlights irregularities such as faults, depressions, chaotic pipe interiors and salt walls. The surface attribute 'Maximum Magnitude' extracts the maximum positive or negative seismic amplitude value between two horizons or a time window (Omosanya and Alves, 2013;Marfurt and Alves, 2014).
Maximum Magnitude maps were extracted between the interpreted Mesozoic surfaces to determine whether any high-amplitude anomalies were prevalent -possibly representing fluid pockets or cemented units -and if there is any spatial significance to them (Fig. 5). However, not all the high-amplitude anomalies in the Triassic pods were extracted initially, as they have similar or lower amplitudes to high-amplitude strata in the Skagerrak Formation where this latter is present immediately (down to 150 ms TWT) below horizon H4 (Top Cromer Knoll Group; Figs. 5 and 8). Equally, the extent of any high-amplitude anomalies observed immediately below the pockmarks cannot be mapped clearly if they are similar in amplitude to the more continuous seismic reflections in the Skagerrak Formation. To overcome these limitations, the anomalies within 150 ms TWT below horizon H4 (Top Cromer Knoll Group) were mapped manually on Petrel ® , and their presence was assigned to the attribute tables of depressions that coincide with seismic anomalies (Table 2). Secondly, the window for Maximum Magnitude extraction was reduced to between 150 ms TWT Fig. 3. Flow chart summarising the methodology used in this study. Three different types of software were used; Petrel® for seismic interpretation (blue boxes); ArcGIS® for mapping pockmarks and anomalies (green boxes), as well as spatial and morphological analysis of mapped features, and Petromod® for the maturation modelling of key wells (red boxes).
below the Top Cromer Knoll Group (horizon H4), and 50 ms TWT above the Top Salt horizon H2, thus reducing any spurious results due to human error when mapping the Top Salt Horizon (Fig. 5). The resulting Maximum Magnitude surface was then exported to ArcGIS and the high-amplitude anomalies were identified using the semi-automated mapping tool developed by the British Geological Survey, described below.

ArcGIS semi-automated mapping and characterisation
Three seismic surfaces containing circular depressions were exported from Petrel into ArcGIS ® version 10.1: the Top Mandal Formation (H3), Top Cromer Knoll Group (H4) and Intra-Chalk (H5) (Fig. 4). An ArcGIS-based Toolbox developed by the British Geological Survey -the BGS Seabed Mapping Toolbox (Gafeira et al., 2012(Gafeira et al., , 2018 -was used to semi-automatically map the confined depressions within each horizon and extract morphological characteristics that include Vertical Relief, Width, Length and Width:Length Ratios, and distribution of depressions. The depressions were mapped automatically by the Toolbox; however, manual interpretation is required to quality-control the automatic mapping. The Toolbox was first used for delineating pockmarks in multibeam echosounder data (Gafeira et al., 2012), and once in 3D seismic data (Geldof et al., 2014), but only at the seabed and 150 ms TWT below the seabed.

Fig. 4.
Well correlation panel using gamma ray logs, from the northwest (6/3-2) to southeast (7/7-2), through salt walls and a Triassic pod. The seismic horizons mapped in this work are named H1 to H6. The Toolbox mapping thresholds described in Gafeira et al. (2012) (Minimum Vertical Relief, Minimum Width and Minimum Width:Length Ratio) were adapted for this dataset. The Minimum Vertical Relief was defined as 25 m, half of the vertical resolution. The horizontal resolution is 50 m; as a geometric shape must be defined by at least 2 pixels, the Minimum Width was chosen to be 100 m, whilst the Minimum Width:Length ratio was 0.2, as used in Gafeira et al. (2012).
The morphological characteristics are used to compare the mapped depressions with the published data and to interpret possible mechanisms for their formation. High-amplitude anomalies in Triassic basins (pods) were also delineated by polygons in ArcGIS ® using this mapping Toolbox, although further manual interpretation was required to compare them with the depressions as some high amplitudes are due to lithological variations. The width of both depressions and acoustic anomalies was defined using the ArcGIS minimum bounding geometry function. This function computes a rectangle defining the smallest area enclosing the delineated depressions and acoustic anomalies, which is used to infer pockmark eccentricity and ellipse axes lengths (Geldof et al., 2014). The Width and Width:Length Ratio of the acoustic anomalies, as well as the pipe height, were assigned to the ArcGIS attribute table describing the depressions.
In order to obtain the morphological characteristics of depressions, interpreted surfaces had to be converted from two-way time to the depth domain. Interval velocities were obtained from wells from the nearby Brynhild and Gaupe fields (Fig. 1) and used to depth-convert the interpreted seismic volume in Petrel ® .
Spatial analysis was conducted visually by overlaying the delineated depressions and anomalies above the salt isopach surface. Statistical data was computed based on the BGS Seabed Mapping Toolbox by plotting box plots, calculating percentages of depressions in different settings, and plotting data on scatter graphs.

Geochemical analysis and basin modelling
Schlumberger's Petromod ® was used to generate simplified, 1D burial history plots. Model inputs such as sub-surface lithologies, thicknesses, ages and bottom-hole temperatures were derived from well reports from the Norwegian Petroleum Directorate database. The magnitude and timing of tectonic uplift events, as well as a palaeo-sea level curve, were obtained from the literature (e.g. Ziegler, 1992;Underhill and Partington, 1993;Monaghan et al., 2017). The thermal history used was taken from Frederiksen et al. (2001). Although the thermal history in this latter work is characteristic of strata further to the south of the Jaeren High, it was considered as representing the regional heat flow values for the study area. Model parameters are summarised in Appendix B.
The thickness of strata not penetrated by wells, such as Zechstein, Rotliegendes and Carboniferous strata, are estimated from the depthconverted seismic volume and by using interval velocities from exploration wells 7/3-1 and 6/3-2. The interpretations in Milton-Worssell et al. (2010) also provided valid constraints on the thickness of Paleozoic strata. Pepper and Corvit (1995) ran a series of experiments to redefine the hydrocarbon generation windows for different organofacies, from marine shale source rocks to terrestrial, humic coal source rocks. They concluded that for humic coals such as the Westphalian in the Southern North Sea, the gas generation window ranges from 175°C to 220°C to generate between 10% and 90% of the potential gas in Paleozoic source rocks. The oil generation window for a marine shale source such as the Kimmeridge Clay Formation (UK equivalent of the Mandal Formation in the Norwegian North Sea) ranges from 105°C to 145°C, and its gas generation window varies from 145°C to 210°C. These isotherm thresholds for the two source rocks are labelled on the resulting burial   history models and were used to interpret the timing of source rock maturation on the Jaeren High. Geochemical well reports from the NPD website were investigated to summarise the hydrocarbon findings of each well.

Results
The results of our mapping are illustrated in Fig. 6, which highlights the key features of each interpreted horizon. Rotliegendes strata dip to the southeast and contain an array of faults that is oriented N-S and E-W. Velocity pull-ups in the basement correspond to areas of thick salt ( Fig. 6a and 6b). The Upper Jurassic Mandal Formation is absent updip, towards the west, whilst the overlying Cretaceous sediments are present across the study area and shallow to the northeast. This suggests tilting of the basement and erosion or non-deposition of the Mandal Formation during the Late Jurassic, with subsequent subsidence of the Jaeren High in the Cretaceous. Pockmarks are observed at the Top of the Mandal Formation, at the Top of the Cromer Knoll Group, and inside the Chalk, whilst being absent at Top Chalk level, a character suggesting that fluid venting had ended by this time (Fig. 6f). The non-uniform structure of the Top Chalk horizon implies that salt movement continued to occur until, at least, the Late Cretaceous.
The results of semi-automatic mapping and characterisation of the depressions in three surfaces -Top Mandal Formation (H3), Top Cromer Knoll Group (H4) and Intra-Chalk -as well as high-amplitude anomalies in Triassic strata, are summarised as follows.

Spatial statistics
A total of 196 depressions were identified across the three mapped surfaces, of which 18.4% occur at Intra-Chalk level, 57.7% in the Cromer Knoll Group and 24% in the Mandal Formation. Superimposing the delineated depressions onto the salt isopach map shows that depressions are solely distributed above the Triassic pods, not above the salt walls (Fig. 7). This suggests that the salt welds are tertiary migration windows, i.e. fluids have not migrated along and above the salt walls. Depressions appear to be concentrated in the centre and up-dip to the northwest of the Jaeren High.
High-amplitude anomalies are found throughout the study area within the Triassic pods, immediately below depressions, and in pipes.  Table 2 summarises the relationship among pipes, depressions and seismic anomalies. No amplitude anomalies were found above the depressions in the Chalk Group, revealing that gas has not migrated up to this strata and there are no gas pockets in the Chalk Group. The depth of the anomalies is variable and does not appear to correlate with any specific stratigraphic layer. Fig. 8 shows examples of the different anomalies and depressions on the seismic data. In particular, Fig. 8a shows that the Top Zechstein salt (horizon H2) was deformed (inflected up) vertically, which could be a result of fluid below the salt breaching the seal and migrating upwards with relatively high pressures.
Of the 295 mapped amplitude anomalies, 55.5% are not shown to be connected to a depression in younger strata (e.g., Fig. 8b). Discrete amplitude anomalies are prevalent in the study area (74.6%), while the remaining 25.4% are stacked in pipe columns. The majority (81.8%) of depressions have an associated high-amplitude anomaly within the pipe. All these positive (hard) anomalies indicate cemented pipes, carbonate nodules, or possibly sandstone bodies within a largely mudstone succession.
Twenty-three of the anomalies that are not associated with a depression are negative (soft) -these indicate isolated gas pockets as the presence of gas in sediments reduces the average seismic velocity of the unit, and can manifest as a negative seismic reflection if the velocity is lower than the overlying unit (Andresen et al., 2011). Of the 113 depressions observed at horizon H4 (Top Cromer Knoll Group), only 12.4% are interpreted to exhibit high-amplitude anomalies directly below the depressions. These could represent cemented zones or carbonate nodules that formed just below the surface once the pipe and depression were established. Pipes range in height from 660 m to 2280 m, with 50% of the pipes ranging between 1100 m and 1500 m (Fig. 9). Blow-out pipes of similar dimensions have been recorded offshore Nigeria (Løseth et al., 2011). The three levels of depressions mapped in this work suggest that fluid flow occurred during the Late Jurassic, Early Cretaceous and in the early Late Cretaceous, with the main episode coinciding with the end of the Early Cretaceous (top Cromer Knoll Group horizon).  Table 3 summarises these plots numerically.

Morphological analysis
The Vertical Relief value describes the vertical fall from the rim of the depression to its deepest point. Vertical Relief ranges from 15 m to 178 m. The Intra-Chalk depressions have the smallest range of values, from 15 m to 71 m, with a median value of 30.5 m. These values are much smaller than those in the CKG and Mandal Formation, which have similar interquartile ranges -from 35 m to 75 m, and medians of 50 and 55 m, respectively (Fig. 9).
The width of the depressions ranges from 225 m to 842 m, with a median value of 450 m. The width of the anomalies ranges from 60 m to 575 m, with an outlier at 1170 m. The median width is 220 m, almost half of that of the depressions (Fig. 9).
Aspect ratio (width:length) is a proxy for eccentricity and ranges from 0.6 to 1.0 for all the depressions, with a median ratio of 0.88. These values reflect low eccentricity, or high sphericity. In contrast, the width:length ratios of amplitude anomalies range from 0.18 to 1.0, with a median of 0.70. This reflects the highly irregular nature of the amplitude anomalies. Due to the high sphericity of the depressions, orientation measurements are heavily biased according to the mapping algorithm function in ArcGIS, hence orientation is not deemed a reliable parameter.
Their width and vertical relief were compared for depressions coinciding vertically at different levels. The plots in Fig. 10a and c show that the width and vertical relief of depressions in the Cromer Knoll Group are always greater than those at the Intra-Chalk horizon H5, as all the points plot below the y = x line. All the Intra-Chalk depressions Table 2 Summary of the number of depressions and anomalies mapped at different seismic-stratigraphic horizons.  Fig. 10b and d).
From these results it is interpreted that the depressions at Intra-Chalk level may not be real pockmarks, and are instead 'drape features' across the underlying pockmarks at Top Cromer Knoll Group level -the Intra-Chalk unit drapes over the underlying topography and is not formed by further fluid flow from below -and likewise for the Cromer Knoll Group depressions over the Mandal Formation pockmarks. Based on the graphs and manual investigation in Petrel, we suggest that ten depressions in the Cromer Knoll Group, overlying depressions in the Mandal Formation, are drape features, whilst the remaining seven depressions are associated with stacked pockmarks that indicate multiple episodes of fluid flow through the same pipe(s).
The total number of interpreted pipes is 321; the sum of the pipes which terminate at the Mandal Formation (12.5%), the Cromer Knoll Group (32.0%) and the anomalies without shallower depressions (55.5%). The fact that over half of the interpreted pipes are not connected to palaeo-pockmarks suggests that active seal breaching and leakage could be occurring at present across the salt welds. Fig. 11 shows the burial and thermal histories from three wells: a) on the Jaeren High, b) downdip from the Cod Terrace, and c) from the East Central Graben. Heat flow increased very moderately during Triassic rifting and more significantly during the growth of the North Sea Dome in the Early Jurassic, from 60 mW/m 2 up to 90 mW/m 2 , decaying in the Cenozoic (Frederiksen et al., 2001). Therefore, despite the significant tectonic uplift associated with the North Sea Dome, Carboniferous coals on the Cod Terrace, and in the East Central Graben, entered the peak gas generation window as early as the Triassic and expelled gas throughout the Mesozoic. These rocks have been buried to depths over 6 km and are presently overmature (Fig. 11).

Maturation models
The Carboniferous on the Jaeren High is 4 to 6 km deep and is interpreted to have entered the peak gas generation window during the Oligocene, remaining there to the present day. Fig. 11 shows the dramatic cooling effect of the thick salt due to its high thermal conductivity compared with clastic sediments, as the two modelled wells in a salt wall (7/7-2) and Triassic pod (7/1-1) show very different thermal histories. As heat flow has regional sources, it is likely that the thermal model from 7/7-2 is more representative of the Jaeren High. However, it Fig. 9. Box plots summarising key morphological attributes of depressions and amplitude anomalies, from left to right: width, width:length ratio -a proxy for eccentricity, vertical relief and pipe height. Attribute ranges are consistently similar for the Mandal Formation and Cromer Knoll Group depressions, whilst those of the Intra-Chalk are generally smaller in width and vertical relief. Pipe heights vary according to the Triassic pod isopach, which increases down dip on the Jaeren High.

Table 3
The minimum, maximum and median values of three morphological parameters of the depressions and amplitude anomalies: width, width:length ratio and vertical relief.

Horizon/zone
Width ( n/a n/a n/a is also possible that some gas was generated on the Jaeren High earlier in its history, possibly during Late Jurassic-Early Cretaceous rifting. It is clear that the Upper Jurassic Mandal Formation reached the gas generation window during the Oligocene in the East Central Graben, whereas on the Cod Terrace it only reached similar conditions during the Quaternary. Whilst it could be possible for gas to migrate along the graben-bounding fault to the Jaeren High in the Cenozoic, large salt diapirs extending from the basement up to the Cenozoic units line up on the graben margin, likely preventing migration across to the Jaeren High. This is also proved by the presence of Paleogene sand plays above salt diapirs in this the UK waters, such as the Lomond and Pierce gas fields, sourced by gas from the Kimmeridge Clay/Mandal Formation buried in the East Central Graben. Therefore, it is unlikely that the gas forming the pipes was sourced from the Upper Jurassic Mandal Formation.

Geochemical evidence
Findings from key wells on the Jaeren High are summarised in Table 4. Most of the wells drilled on the Jaeren High are located above salt walls, targeting the Jurassic and Triassic sandstone units. Those that found oil were sourced from the Upper Jurassic Mandal Formation. Only three wells were drilled into the Triassic pods. The wells terminate within the upper 300 m of the Skagerrak and Smith Bank Formation, 7/ 1-1, 7/4-3 and 7/7-1, all of which were dry and with no gas shows.
Well 7/1-1 was drilled adjacently to a depression in the Cromer Knoll Group, with a high-amplitude anomaly below it, but no gas shows were recorded. The maximum methane recorded was 0.37% in the top of the Upper Cretaceous Chalk, with average values remaining well below 0.1% throughout the Chalk and into the Triassic. Cores were also taken in the Chalk, which were proven to be tight. Stylolites are common, which when fractured along these lines of weakness were seen to have very thin bands of black, shale-like impurities that do not fluoresce under UV light. The Chalk Group did not contain any oil stains or visible porosity.

A model for fluid migration and caprock breaching on the Jaeren High
Detailed investigation of fluid flow features combined with burial history modelling allows us to describe the fluid migration history of the Jaeren High, adding to the geological history described by Høiland et al. (1993). Fig. 12 summarises the history of fluid flow on the Jaeren High.
Imaged pipes are shown to be rooted in the Permian Rotliegendes and cut through the Triassic succession, across salt welds. Previous studies state that salt welds formed as early as the Middle-Late Triassic due to sediment loading and salt evacuation (Høiland et al., 1993). The geometry of Triassic pods on the Jaeren High supports this statement. Seismic reflections in the Smith Bank Formation bend down onto the Rotliegend Group and salt wall edges, forming turtle anticlines, a character indicating grounding of sediment pods (e.g. Karlo et al., 2014), whilst the younger Triassic reflections level out and become parallel towards the top of the pods (Fig. 12). A salt weld cannot be formed solely by salt evacuation, as the boundary drag forces increase as salt thickness decreases (Jackson and Cramez, 1989;Hudec and Jackson, 2007). The remaining salt can be removed by dissolution (Wagner and Jackson, 2011).
Pipes terminate in the form of depressions within the Mandal Formation (Late Jurassic) and Cromer Knoll Group (Early Cretaceous), and are interpreted to form pockmarks at this level. The build-up of excessive gas and fluid within the Rotliegend Group caused high overpressures to develop. Once the capillary entry pressure of the overlying seal was exceeded, gas and fluid migrated across salt welds and through the Triassic mudstone succession by a hydrofracturing mechanism, forming pipes that terminated at the palaeo-seafloor as those described in Moss and Cartwright (2010). Sediment was removed into the water column, leaving behind a depression known as a pockmark. Therefore, the timing of tertiary fluid migration is indicated by the level of the pockmark; the pipes terminating in pockmarks formed during the Late Jurassic-Early Cretaceous. The migration of gas could have occurred over years to centuries or more (Cathles et al., 2010). The term 'craters' may be a more apt name for these features, owing to their large sizes, as modern, recorded pockmarks are typically up to 300 m wide and range in vertical relief between 1 m and 80 m deep (Hovland and Judd, 1988;Gay et al., 2006). Large pockmarks have nonetheless been described on the Scotian Shelf as typically measuring 10-700 m in diameter and being 1-45 m deep (Hovland et al., 2002).
It is possible for pipes to form due to rapid compaction and dewatering of mudstones (Berndt et al., 2003). However, the pipes identified on the Jaeren High are rooted in Rotliegendes strata and, in most cases, generate local folds (inflections) within the salt (Fig. 8). This indicates a breach in the seal and fluid bypass through the salt welds, supporting the interpretation of gas and fluid expulsion from depth. This interpretation also relates to giant pockmarks found at the top of Lower Cretaceous Marl in the Lower Saxony Basin of the Netherlands (Strozyk et al., 2018). F. Strozyk and co-authors interpreted that gas sourced from Carboniferous coals migrated across salt welds into shallower Cretaceous strata, forming giant pockmarks.
Results show that 28% of the pipes with depressions formed during the Late Jurassic, whilst the main period of fluid flow on the Jaeren High occurred during the Early Cretaceous, as 72% of the pockmarks terminate at this level. The pipes forming stacked pockmarks may have been active throughout this time, whilst those generating discrete pockmarks were plugged and cemented at an early stage (e.g. Hovland et al., 2005;Andresen et al., 2011). Thirty-six depressions were mapped at the Intra-Chalk level. However, morphological analysis shows that the Intra-Chalk depressions are always associated with corresponding depressions at the Top Cromer Knoll Group horizon H4, which are consistently greater in width and vertical relief. This lead us to interpret the depressions at Intra-Chalk level as drape features due to differential compaction. Still, it is possible that low-level seepage occurred after the initial formation of the pockmarks. Not all the depressions show these 'drape' features, perhaps due to laterally varying lithology in the overlying Chalk Group, or due to thickness variations and different degrees of differential compaction.
Although the pipes are interpreted to be sourced from Rotliegendes strata, the gas is likely to have been expelled from deeper (Carboniferous) coals. The burial history plot for the Jaeren High shows that Carboniferous strata did not reach sufficiently high temperatures to enter the gas maturation window during the Mesozoic (Fig. 11a). However, investigation of nearby potential source kitchens showed that the Carboniferous entered the gas maturation window during the Triassic, remaining there throughout the Mesozoic (Fig. 11b and c). We interpret that gas could have been sourced to the west of the Jaeren Fig. 11. Results of the burial history modelling from four wells in the Central North Sea, locations labelled in Fig. 1. The Carboniferous strata in the East Central Graben (a) and Cod Terrace (c) became gas mature for coals in the Triassic, and overmature from the Cretaceous to present. Some gas may have been produced on the Jaeren High during the Late Jurassic (b), but more likely from the Oligocene to Recent, depending on the regional cooling impact of the salt. We interpret that long distance migration in the Mesozoic sourced the pipes at that time, with local maturation and migration occurring at present.

Table 4
Summary of the objectives, drilling location with respect to salt walls or Triassic pods, and results of the key wells on the Jaeren High. Data has been summarised from the Norwegian Petroleum Directorate web pages. No Skagerrak Fm. Abandoned as a dry well.
C. Roelofse et al. International Journal of Greenhouse Gas Control 80 (2019) 103-122 High in the East Central Graben, migrating vertically along the grabenbounding fault. Importantly, the presence of pipes and pockmarks up to 40 km away from this fault, down-dip of the Jaeren High, suggests that large volumes of gas accumulated there and putatively sourced gas to the study area from the southeast, i.e. from the Cod Terrace region in Norwegian waters. These two source kitchens, exceeding 6 km depth, are presently overmature.

Data limitations and the importance of understanding leakage risk
Morphological analyses have been undertaken to describe the pockmarks and discern their formation mechanisms. However, it became apparent that due to the resolution of the data and the scale of variation of the morphological characteristics, only a few basic parameters were viable for analysis, e.g. width, vertical relief and eccentricity. Small-scale changes in slope and roughness may give insight into whether the depressions are pockmarks, sinkholes or hypogenic collapse structures, but these variations are not resolved on the interpreted seismic volume. However, as the depressions are situated within the marine shales of the Mandal Formation, and marls of the Cromer Knoll Group, it is not possible for the depressions to be hypogenic karst structures. Furthermore, compaction and burial will have removed or distorted morphological features that would be clear at the seabed or shallow subsurface. This could impact interpretations as the depressions within the Chalk Group, considered to be 'drape' features, can actually represent active fluid flow, but are smaller due to the differing lithology of the Chalk Group compared to the Cromer Knoll Group and Mandal Formation. Xia and Wilkinson (2017) summarised the geological reasons for exploration failure in boreholes drilled on the UK Continental Shelf. They noted that 6% of the wells failed due to poor caprock sealing, 6% due to incorrect trap depth prediction, and 21% due to incorrect prediction of trap geometry and extent. This emphasizes the importance of analysing fluid migration and leakage across broad regions of sedimentary basins.
We interpret the 'hard' anomalies in this work to be cemented zones or carbonate concretions. Only 13% of the anomalies that are not connected to pockmarks are 'soft' -if these represent pockets of gas, then it is possible that leakage is occurring at present, an unfavourable setting when assessing sites for CO 2 injection. Fluid flow conduits are actively bypassing salt welds and the overlying mudstone successions, despite thicknesses of up to ∼2200 m. Thickness is one of the four seal risk factors described by Downey (1984) together with capillary entry pressures, lateral lithological variations and variable degrees of fracturing. A thicker seal unit is considered to be more effective, but is still dependent on the height of the fluid column and resulting build-up of pore-pressure. On the Jaeren High, with seal thickness ranging from 650 to 2200 m, the seal was still breached (Fig. 8).
In summary, the method presented here is important for predicting the mechanisms behind pipe formation and whether gas can be trapped in the pipe itself. It also helps discerning if the pipes are active and gas is migrating towards a shallower reservoir, or even leaking towards the surface. If depressions had been mainly found within the Chalk Group, then these could be collapse structures formed due to hypogenic fluid migration when the Chalk Group was buried, as opposed to fluids venting onto the sea floor (Masoumi et al., 2014). This also means that fluid migration could have happened relatively recently in time, and gas should be trapped within the Chalk Group. However, the depressions are found within the marls and shales, rebuking such an interpretation. Exploration wells consistently found tight limestones with very low gas concentrations in the Chalk Group, so it is unlikely that gas has accumulated at this level, and that the pipes are active at present.

Conclusions
This study shows how a novel integrated method using GIS, borehole geochemical and 3D seismic data can be used to better visualise and model fluid flow features that have previously been overlooked, or Fig. 12. Summary of the fluid migration evolution of the Jaeren High, from a) to d). a) Shows a regional schematic, modified from Hodgson et al. (1992). b) To d) focus on the Jaeren High only, marked in a grey box in a). Gas pockets (red) and cemented zones (blue) are interpreted from 'soft' and 'hard' anomalies respectively. Arrows indicate migration pathways. CGK -Cromer Knoll Group.
unresolved. Understanding the timing of tertiary fluid migration and the locations of fluid flow pathways is important for assessing and managing the risk of fluid escape for both hydrocarbon exploration and carbon capture and storage. The results of this work can be summarised as follows: a) In the study area, large vertical fluid pipes sourced in the Permian Rotliegend Group cut across 600 to 2300 m of Triassic mudstones, across salt minibasins, or pods. These pipes terminate as pockmarks in the Upper Jurassic Mandal Formation and Lower Cretaceous Cromer Knoll Group. This indicates the timeframe of vertically focused fluid expulsion and the potential for fluids to leak from these established conduits. b) As the pockmarks range from 200 to 800 m in width and 15 to 200 m in vertical extent, they reflect the transport of extremely large volumes of fluid, which were required to form features of this scale. The gas migrated into the Mesozoic strata sequence sourced from the East Central Graben to the west, and Cod Terrace to the southeast. c) Long-distance vertical and up-dip migration occurred towards the Jaeren High, until pore-fluid pressures were sufficiently high to breach the Smith Bank Mudstone Formation seal. d) Thus, high-amplitude anomalies represent both cemented pipes and potential gas pockets, whose presence indicates increased risk as a potential flow pathway. These are crucial data to determining where active leakage is occurring at present, and whether a site will be suitable for containment and geological storage of carbon dioxide.

Declarations of interest
None.
during the Early Jurassic doming of the North Sea. Upper Jurassic strata have a limited distribution, being absent above most of the Triassic pods and present directly above salt walls -the thickest section encountered is 157 m in well 6/3-2. The Mandal Formation manifests as a high-amplitude seismic reflection that is characteristic of a 'hot shale', just as the Kimmeridge Clay Formation, its UK equivalent. The Mandal Formation is also clearly marked by a sharp peak in gamma ray curves (Fig. 4). Towards the east of the study area, where the Mandal Formation is interpreted to occur above the Triassic pods, some depressions are seen, with pipes imaged below (Fig. 6c).

Horizon 4 -Top Cromer Knoll Group (Early Cretaceous)
The Base Cretaceous Unconformity lies across the Triassic pods and Upper Jurassic strata above salt walls. The Top Cromer Knoll Group is also an unconformity representing the top of either the Rødby, Sola, Tuxen or the Åsgard Formations. The Cromer Knoll Group comprises continuous, gently undulating internal seismic reflections. The top of this Group marks the transition from marls to the Chalk Group, and manifests as a bright hard seismic reflection, draping the full extent of the Triassic-Jurassic pods-interpods. This surface is shown to be covered in large circular depressions (Fig. 6d).

Horizon 5 -Intra-Chalk Group (Late Cretaceous)
This horizon is an intra-Chalk unconformity, possibly corresponding to the Top Hod Formation. It varies greatly in amplitude, but is largely continuous, draping the underlying topography. In some cases, the surface contains depressions above those seen in the Cromer Knoll Group.

Horizon 6 -Top Chalk Group (Late Cretaceous -Early Paleocene)
The Top Chalk is another bright, continuous seismic reflection, marking the transition from Cretaceous-Paleocene Chalk to Tertiary clastics. Internal seismic reflections in the Chalk Group are generally parallel to sub-parallel and of low amplitude. Some faults appear to cut the Chalk Group, whilst no depressions are interpreted inside this unit.

Top Sele Formation (Paleocene)
The Paleocene unit contains mostly low amplitude, parallel to sub-parallel internal reflections representing hemipelagic mudstones onlapping onto basin highs. The Sele Formation is cut by high-angle, high-density polygonal faults that extend into younger strata.

Hordaland and Nordland Groups (Paleocene -Recent)
The Hordaland and Nordland Groups, from Eocene to present, consist of parallel, continuous reflections, with occasional, weakly sub-parallel and onlapping seismic reflections demonstrating the largely hemipelagic nature of their strata.