A unified workflow for hydrocarbon migration simulation: Embedding trap and seal assessment in seismic interpretation

Introduction

The effective exploration and management of subsurface resources requires an early and thorough evaluation of traps and seals. Whether the objective is a new hydrocarbon prospect, bypassed attic oil in a mature field, or a storage site for CO2 or hydrogen, the geoscientist must identify every plausible trap configuration and every leakage pathway as early as possible (Patruno et al., 2024). Calibration studies show that estimates of the probability of success are systematically biased and inconsistent between teams (Milkov, 2015, 2017), and a review of more than 20 years of exploration wells identifies charge and migration as the dominant cause of failure in frontier settings, precisely where the data that would constrain them are scarcest (Rudolph and Goulding, 2017). Both processes elude direct observation: migration pathways and their timing cannot be imaged or measured from seismic or well data. Of this compound risk, the workflow presented here targets the migration and retention component (could hydrocarbons reach a given trap, and would the seals hold them?); source presence, maturity and expulsion timing remain the province of basin modelling.

A more fundamental difficulty is organisational rather than physical. Trap and seal evaluation is usually performed in a petroleum-systems tool that is disconnected from the seismic interpretation, and that consumes a heavily simplified model in which faults, geobodies and much of the structural and stratigraphic complexity have been filtered out in order to build a tractable geocellular grid or pile of horizon surfaces. The result is a paradox: sophisticated physics is run on a support that has lost the very elements that control trapping and migration. Meanwhile, DHI and amplitude-versus-offset (AVO) analyses are carried out on the seismic data in a separate track. The two are reconciled only late and by hand, so that some DHIs are pursued without a credible trap, some mapped traps are never checked against the amplitude data, and the migration model quietly diverges from the interpretation.

We present a methodology that addresses these limitations. We describe a trap and seal assessment workflow that is embedded in the seismic and DHI interpretation environment and that is built, in both its map-based and its three-dimensional forms, to honour structural traps (four-way and three-way closures) and stratigraphic traps together, on the same model that was interpreted. Four ideas organise the contribution: the integration of trap and seal assessment with interpretation and DHI analysis; the ability to build and run many scenarios interactively; the adaptation of the migration physics to the data at hand; and, in three dimensions, the use of a relative geological time (RGT) model both to define a property-bearing layering that is decoupled from the property support (i.e., the grid, surface or volume on which a property is stored) and to resolve dip so that migration can run on a structured regular grid.

State of the art

Established approaches fall into three families: fill-and-spill techniques, petroleum-system models and seismic-scale migration simulation. Trap and seal assessment has long relied on map-based fill-and-spill analysis on interpreted structural surfaces, which provides rapid estimates of trap capacity and spill relationships (Abrahamsen et al., 1998). Such analysis is fast and intuitive but is confined to a single surface and does not, on its own, capture the interaction between stacked reservoirs separated by weak seals or connected by faults. At the other end of the spectrum, basin and petroleum-systems modelling simulates the full system forward through geological time, but its cost at basin scale restricts it to coarse grids that resolve neither carrier-bed heterogeneity nor fault compartmentalisation at trap scale, and it generally requires an explicit geocellular grid or a stack of surfaces that must be built from a simplified interpretation (Evenick, 2022).

A third, seismic-scale family performs a simplified migration directly in a regular grid, using the seismic cube or a sub-sampled cube as the support and an externally derived property, typically a capillary entry pressure painted throughout the cube from a function or transform of seismic attributes. Because such tools do not carry explicit stratigraphic information, they cannot represent the control that layering exerts on permeability anisotropy and on migration direction, and they cannot assign different physics to different layers. Risk and DHI frameworks have advanced in parallel, from consistent risking tables (Rose, 2001; Milkov, 2015) to integrated chance-of-success schemes that score geophysical evidence against drilling outcomes (Monigle et al., 2025), but these are score observations rather than incorporating a process-based, interpretation-consistent charge model. Finally, signal-driven global interpretation now converts a seismic cube into a dense chronostratigraphic model, the RGT volume (Pauget et al., 2009; Lacaze et al., 2011), which opens the possibility of using that model directly as the support for migration.

To our knowledge, none of the established migration approaches takes an RGT model, derived from the seismic interpretation, as input: they rely either on explicit surfaces and stratigraphic grids, in which node elevation encodes topography, or on regular grids populated with externally derived properties, which demands extra geomodelling effort and often sacrifices accuracy relative to using the interpretation products directly. Nor do they allow different physical models (buoyancy-only, pseudo-permeability, invasion percolation) to be combined across the layers of a single model. These two gaps motivate the method below.

Method

The workflow integrates automated seismic interpretation with rapid trap analysis in an interactive, scenario-based loop with four components: a seismic-to-model conversion, an interactive migration simulation, an iterative data-integration path, and a systematic comparison with seismic indicators.

Integration with interpretation and DHI

Rather than interpreting a few key horizons, a semi-automatic full-volume method converts the 3D seismic into a dense chronostratigraphic model, the RGT volume, that honours both major reflectors and subtle events (Pauget et al., 2009). Seismic attributes (Chopra and Marfurt, 2007) are projected onto the resulting closely spaced surfaces to reveal facies variations, channel geometries and pinch-outs (Lacaze et al., 2011). Because the workflow lives inside the interpretation platform, modelled traps can be compared directly with seismic anomalies and DHIs at any stage. This tight coupling makes three things systematic that are otherwise hard to achieve when interpretation and evaluation sit in separate tracks: only DHIs that coincide with a plausible modelled trap are promoted, every mapped trap is tested for an expected DHI expression, and the trap and seal model is the interpreted model itself, retaining faults, geobodies and stratigraphic detail, rather than an over-simplified surrogate. Both map-based and 3D modes identify four-way and three-way structural closures, as well as stratigraphic traps, within a single run. A second requirement is completeness of the trap inventory, addressed by a dedicated ‘trap-finding’ mode: instead of releasing hydrocarbons from a few injection points, the simulator treats every cell lying immediately beneath an impermeable seal, in the map-based mode, or every cell of a designated source layer, in the three-dimensional mode, as a virtually unlimited source. A single migration run then charges and tests, for a given top-seal and fault-seal scenario, every structural and stratigraphic trap in the area of interest, so that the complete trap inventory is recovered in one operation rather than prospect by prospect.

Rapid scenario evaluation

Because trap performance hinges on a few uncertain controls, these controls can be edited and their consequences are recomputed at once: an interpreter varies the reservoir top and base geometry, the topology of the fault network, and the capacity of the faults and of the top seal, and obtains updated migration pathways, retained columns and trapped volumes for each variant. The controls are physically explicit. A fault may be passing (fluid migrates laterally within the same carrier), leaking (fluid is transmitted to the layer above), or sealing. Because these behaviours co-exist on real fault planes (Holden et al., 2022), they are assigned per layer and per fault segment and edited interactively, down to individual grid cells, so that a single fault may seal over one portion of its extent and pass or leak over another. A top seal may fail either by capillary leakage, once the buoyancy pressure exceeds its entry pressure, or by mechanical failure once a critical column height is reached. In the latter case the engine does more than flag the breach: it simulates the leakage of the previously trapped hydrocarbons through the damaged seal and records the zones the accumulation once occupied. These paleo-accumulations correspond to what wells may encounter as shows or residual hydrocarbon traces on cores and logs. Scenarios that vary these behaviours are compared side by side (Figure 1), so that the factors controlling trap performance are identified and outcomes are framed as best, most-likely and worst cases, including the effect of fault-seal parameters on retained column height (Miocic et al., 2019).

Data-adaptive migration physics

The physical fidelity is matched to the data. At each step the invasion front advances into one admissible neighbouring cell, chosen by one of three rules of increasing sophistication. Writing $z$ for depth (increasing downward), $\Delta z$ for the depth difference between the current cell $c$ and a candidate neighbour $n$, $g$ for gravity, $k$ for a pseudo-permeability, $P_{ce}$ for the capillary entry pressure and $\Delta \rho$ for the brine-minus-hydrocarbon density contrast, the candidate is selected as follows:

(1a) buoyancy only: invade $\mathrm{argmin}{z}_n$ over $N(c)$

(1b) pseudo-permeability: invade $\mathrm{argmax}{k}_n$ over $\{n\in N(c):z_n<z_c,k_n>0\}$

(1c) invasion percolation: invade $\mathrm{argmin}(P_{ce,n}-\Delta \rho g\Delta z)$ over $F$

where $N(c)$ denotes the neighbourhood of $c$ and $F$ is the invasion front. In its simplest form (1a) migration is driven by buoyancy alone: with the density contrast and gravity effectively constant over a small neighbourhood, and the depth of the current cell fixed, this amounts to selecting the shallowest admissible (i.e., permeable and shallower) neighbouring cell. This approach is rigorously valid only in an isotropic, homogeneous material, but it needs only structural geometry and suits stochastic cataloguing at the earliest stages. Where a property that co-varies locally with permeability can be derived, for example from impedance inversion or facies classification, the pseudo-permeability rule (1b) applies: among the neighbours shallower than the current cell that carry a positive pseudo-permeability, the one of highest value is invaded. Because only the ranking of the property matters, the pseudo-permeability need not be calibrated or scaled, a decisive practical advantage when it is taken directly from a seismic attribute. At the highest fidelity, where a capillary entry pressure model is available, macroscopic invasion percolation (Carruthers, 2003) invades the neighbour of lowest effective entry pressure (1c), buoyancy lowering the threshold by the term $\Delta \rho .g.\Delta z$; this gives the most rigorous treatment of seal bypass and column-height limitation. The same three rules apply in the map-based and the three-dimensional settings; Figure 2 compares them on one map-based case.

Map-based hydrocarbon migration

In the map-based mode, a combined spill-and-seal analysis on the interpreted top-reservoir surface (after Abrahamsen et al., 1998) is coupled with the invasion-percolation algorithm to model migration beneath a sealing surface. Structural closures and spill points are identified on the surface, stratigraphic traps such as a pinch-out against a nose structure are captured as well (Figure 3), trap capacities are computed, and fill, spill and leak are simulated. Each scenario is defined by the sealing state of the faults, the entry points, and the seal and reservoir properties; for each, the analysis returns closure and spill maps, gross rock volume and explicit leakage pathways. This approach is used for both prospect and storage-site evaluation (Tortarolo et al., 2026; Abdallah et al., 2025).

Three-dimensional migration and the role of the RGT model

Where a 3D volume is available, secondary migration is resolved within the carrier volume itself, capturing vertical and lateral connectivity, migration across fault compartments, and transfer between stacked intervals, which makes it particularly well suited to modelling stacked reservoir systems. The RGT model plays two distinct roles here, and both are central to linking migration with interpretation.

First, the RGT defines a property-bearing layering that is decoupled from the property support. The domain is partitioned into layers, each being the set of cells whose RGT value falls in a given iso-value interval. Each layer is then assigned petrophysical properties and a migration physics. The key novelty is that the layering, defined by RGT iso-values, is fully independent of the grids or meshes that store the properties: a property can be provided as a constant, as a 2D map, or as a 3D volume held on a support entirely separate from that of the RGT. The RGT provides the partition; each property is honoured on its native support, so that data of very different maturity (a regional constant, a mapped trend, an inverted volume) can coexist in one model without resampling (Figure 4).

Second, the RGT resolves updip and downdip directions and quantifies dip without extracting explicit horizon surfaces or building a geocellular grid. Because a buoyant fluid migrates updip, and because in sub-horizontal settings the geometric dip is too subtle to interpret reliably, the RGT gradient acts as the tie-breaker that sets the lateral migration direction. This lets the entire simulation run on a structured regular grid, for example a seismic cube of cubic or rectangular voxels, with the depth axis aligned to gravity. The simulation therefore operates in the depth domain, and time-migrated volumes and their RGT models are converted to depth beforehand. Velocity uncertainty is consequently a first-order control on the structural geometry, and hence on migration pathways and trapped volumes; alternative depth conversions are treated as scenarios in the same rapid loop, and their impact on predicted accumulations is illustrated in Philit et al. (2025). The three physics rules are assigned per layer, and a migrating hydrocarbon body whose front crosses an interface between layers of different rule is split, and the new body migrates onward under the rule of the layer it enters. In the buoyancy and pseudo-permeability rules an external reservoir versus non-reservoir criterion, here supplied directly by the layer typing, determines where upward migration halts and an accumulation forms; in invasion percolation the capillary thresholds play that role.

The engine simulates fill and spill of structural and stratigraphic closures on the fly, together with per-column seal breach, and reports the migration pathways, accumulations, capillary leak points and spill points (Figure 5). Two thresholds are distinguished for each seal, a maximum column height before capillary leakage and a maximum column height before mechanical, or integrity failure; the smaller fails first. On a capillary breach a new body forms in the overlying layer and inherits the column; on an integrity breach the drained body is released. Faults are honoured per layer and segment, so a single fault may seal over one portion of its extent and pass or leak over another. To operate at seismic resolution, the invasion state is held in a sparse, lazily allocated structure so that memory footprint scales with the invaded sub-volume, and data are read on demand, so that grids of hundreds of millions to billions of voxels are handled within a modest memory budget.

This second role is what makes the link with interpretation practical. The RGT model is the end product of the seismic interpretation; using it as the main input to the migration means that a change of interpretation, or a new scenario, can be looped back into a migration run almost immediately. It also avoids the limitation of the property-only regular-grid seismic-scale invasion percolation tools, which are blind to the control that layering and permeability anisotropy exert on migration.

Results

The workflow has been applied across a range of settings. Three cases illustrate the approach: a study of subtle direct hydrocarbon indicators on the Goodwyn field, a dry-well post-mortem on Poseidon, and a three-dimensional stacked-trap example from the Maui field, offshore New Zealand.

The Goodwyn field, on the North West Shelf of Australia, illustrates the value of tying predicted accumulations to direct hydrocarbon indicators where those indicators are ambiguous (Figure 6). The field lies about 130 km offshore in some 130 m of water. Gas is held in pre-rift fluvio-deltaic sandstones of the Triassic Mungaroo Formation, in gently dipping reservoir units commonly 30 to 80 m thick (Young, 1998; Longley et al., 2002). Most accumulations are pre-unconformity: reservoir units and fault blocks are truncated beneath the regional Callovian breakup surface and sealed by Lower Cretaceous marine mudstones (Bishop, 1999). Because that surface cuts across tilted fault blocks, connectivity and migration screening are sensitive to how fault intersections and terminations are represented, which is precisely the detail that a simplified structural framework discards.

Since the productive interval is a retrograding delta front of interbedded thin sands and shales rather than a massive sand body, its amplitude expression is subtle and can mislead: a flat spot, for example, cannot develop within a shale unit. Running the migration on the interpreted, fault-resolved model and overlaying the predicted accumulations and pathways on the seismic therefore allows the interpreter to judge which of the subtle bright spots and flat spots coincide with a charged, sealed trap and which do not, so that DHI analysis is concentrated where it is physically justified.

Figure 1 compares two connectivity scenarios across the Goodwyn, Dockrell and Echo-Yodel reservoirs. In the first, the intervening faults seal and the reservoirs remain compartmentalised: the modelled Goodwyn accumulation covers 120 km² with a maximum column height of 130 m. In the second, the faults pass and Dockrell is charged from Goodwyn: the charged area grows to 135 km² while the maximum column height falls to 125 m, because part of the accumulation now resides in the Dockrell compartment and the connected system spills from Dockrell into Yodel. Since Goodwyn occupies the updip position in the connected system, the two scenarios differ by only 1% in trapped gross rock volume: the connection redistributes the accumulation laterally rather than draining it, so the volume retained at Goodwyn is only weakly sensitive to the sealing state of the intervening faults. Both scenarios were built and run in a single session on the interpreted, fault-resolved model, without rebuilding a structural framework between variants.

In the Poseidon area of the Browse Basin (North West Shelf, Australia) the workflow was used for a dry-well post-mortem on the Plover Formation (Figure 7). The area lies in the Caswell Sub-basin, approximately 480 km north of Broome, where the reservoirs are fluvio-deltaic to marginal-marine sandstones of the Lower to Middle Jurassic Plover Formation, overlain by Jurassic to Lower Cretaceous marine claystones (Blevin et al., 1998; Struckmeyer et al., 1998). Poseidon-1 encountered a 317 m gross gas-bearing Plover interval comprising three sands that total 228 m, without penetrating the gas-water contact; because the well sits more than 100 m below the mapped crest, the structure implies a gross gas column in excess of 430 m. Closures are defined by northeast-striking normal faults, and the sealing behaviour of those faults is the principal control on trap definition and the focus of the post-mortem that follows.

The prospect drilled by Poseidon-North-1 had appeared promising, yet the well found only traces of hydrocarbons. Multiple scenarios were generated by altering the fault framework and varying seal capacity. Calibrated against the gas column at Poseidon-1, these scenarios indicate a high caprock capacity and effective sealing of the main normal faults; on that basis, the dry result is best explained by a spill point towards the south-west into the Torosa field, combined with a passing rather than sealing fault between the well and the main lead. The alternative scenario, in which that fault seals, predicts an accumulation at the well. Discriminating between the two therefore constrains both the outline of the field and the extent and risking of the surrounding prospects.

Because the migration model, the fault framework and the seismic data share a single domain, the seismic expression of the fault separating Poseidon from Poseidon North can be inspected directly where the two scenarios diverge. Along most of its length this fault juxtaposes hanging-wall and footwall intervals of continuous, sub-parallel reflectors. Approaching its intersection with the main NE-SW normal fault, reflector continuity degrades and the pattern becomes chaotic over an along-strike distance of 3000 to 5000 m, with variance rising from 0-0.1 in the undisturbed hanging wall and footwall to 0.4-0.5 within the disrupted zone (Figure 7). This disruption is consistent with a damage zone of increased fracture density developed at the fault intersection, and it coincides with the segment that the migration scenarios require to be passing rather than sealing. The observation is corroborative rather than diagnostic, but it can be made in the same session and on the same model as the migration run, and it allows the passing behaviour to be assigned to a specific fault segment rather than to the fault as a whole.

The three-dimensional mode, with its RGT-defined layering and per-layer physics, is exercised on the stacked reservoirs of the Maui field, offshore New Zealand (Figures 4 and 8). The field lies in the offshore Taranaki Basin, some 40 km from the Taranaki coast in about 110 m of water, and is New Zealand’s largest hydrocarbon accumulation at approximately 3.4 Tcf. Gas and condensate are held in stacked Paleocene to Eocene sandstones of the Kapuni Group, deposited in coastal-plain to fluvio-marine settings and sealed by marine shales of the Turi Formation (King and Thrasher, 1996; Higgs et al., 2012). Reservoir quality is high: porosities in the Mangahewa C Sands reach 27% and permeabilities approach 2000 mD. The trap is a broad, low-relief, fault-segmented anticline, and the productive interval comprises several discrete sand units separated by intraformational shales. Each unit carries its own fluid contact, which makes the field a natural test of a layered migration model.

The RGT model partitions the Kapuni Group into three layers, one for each sand level (the C, D and F Sands), so that each sand unit and its overlying intraformational shale becomes a separate migration domain with its own properties and physics, without extracting horizons or building a geocellular grid. A single run charges the three sand levels independently and reproduces a discrete accumulation in each, with distinct contacts and distinct spill points, together with the transfer of hydrocarbons between units where the capacity of the intervening seals is exceeded. Migration proceeds updip within each carrier, and the C Sands, the thickest and highest-quality reservoir interval of the stack, accumulate the largest trapped volume, 79% of the total predicted for the stack (Figure 8).

Figure 8 compares the predicted accumulations with the amplitude data in section and in map view, in the same domain and without transfer between applications. In section, the modelled contacts fall within ±5 m of the flat spots and downdip amplitude terminations picked on the same lines. In map view, the amplitude extracted on the RGT-derived surface bounding each sand defines anomalies whose outlines follow the predicted accumulation footprints. Because the layering derives from the interpretation itself, the comparison can be repeated unit by unit through the stack and each predicted accumulation attributed to the sand that hosts it; a simulation run on a property painted into a regular grid would resolve the stack as a single carrier and could not make that attribution.

Discussion

The central benefit of the approach is consistency: the migration is run on the same faults, geobodies and stratigraphic detail that the interpreter sees, and the amplitude evidence and the trap inventory are cross-checked systematically rather than opportunistically. Predicted accumulations can be forward-modelled and compared with observed DHIs, bright and dim spots, flat spots, polarity reversals and AVO effects, so that the relevance of an indicator is validated while the set of viable hypotheses on trap, seal and charge is narrowed. The multi-scenario design turns this into an explicit risk matrix, in which the conditions for each trap to succeed or fail are catalogued in advance. The same bookkeeping makes the workflow a natural post-mortem tool: its predictions can be confronted with the full range of well outcomes (dry wells, discoveries and wells that encountered only shows or residual traces), and can offer a coherent explanation for each.

Because the physics is matched to the data, the same formulation applies from a buoyancy-only screen over a frontier survey to a capillary-controlled study in a well-characterised field. In three dimensions, decoupling the RGT layering from the property support removes a long-standing constraint: properties can be stored on whatever grid, surface or volume is natural to each dataset, while the layering that governs anisotropy and migration comes from the interpretation. Running directly on the structured regular grid, with dip resolved from the RGT rather than from extracted horizons, keeps the interpretation-to-migration loop fast.

The approach is complementary to full petroleum-systems modelling rather than a replacement for it: it does not model source-rock kinetics or thermal history, and it targets secondary migration and trap fill at play-to-block scale, fast enough for stochastic cataloguing. Its domain of validity should be kept in mind. The simulation considers a single mobile phase and is quasi-static; invasion percolation is treated macroscopically rather than at the pore scale; and migration is simulated on the present-day geometry, so the timing of charge relative to trap formation is not modelled. Structural restoration, like the derivation of per-domain fault attributes, lies outside the scope of the workflow. These are deliberate choices in the service of speed, adaptability and direct comparability with the geophysical observations.

Conclusion

We have described a trap and seal assessment workflow embedded in the seismic and DHI interpretation environment, spanning a fast map-based mode and a full three-dimensional mode in which the relative geological time model defines the property-bearing layering and resolves dip. The case studies show what this integration delivers in practice: on Goodwyn, running the migration on the interpreted, fault-resolved model focused the DHI analysis on the anomalies that a charged, sealed trap can explain; on Poseidon, scenario testing of fault behaviour and seal capacity offered a coherent explanation for a dry well; and on Maui the accumulation predicted in each sand of the stack was cross-checked against the amplitude data in both map and section view, sharpening the characterisation of the individual reservoir units. Natural extensions include automated Monte-Carlo propagation of parameter uncertainty into chance-of-success estimates, saturation models to convert accumulation geometries into volumetric distributions, systematic forward modelling of the seismic expression of predicted accumulations for quantitative comparison with DHIs, and the derivation of fault behaviour from juxtaposition and membrane-seal analysis (e.g. Holden et al., 2022) so that the sealing, passing or leaking state is computed rather than assigned.

Acknowledgements

We would like to thank Geoscience Australia and New Zealand Petroleum & Minerals for providing data used in this study, and Petronas through Malaysia Petroleum Management for its support of related work.

References

• Abdallah, H., Souche, L. and Etchebes, M. [2025] A multi-scenario approach for evaluating carbon storage sites: example of the P18 field, offshore Netherlands. World CCUS Conference 2025, Extended Abstracts.
• Abrahamsen, P., Hauge, R., Heggland, K. and Mostad, P. [1998] Uncertain cap rock geometry, spill point, and gross rock volume. SPE Annual Technical Conference and Exhibition, Expanded Abstract, SPE-49286-MS.
• Bishop, M.G. [1999] Total petroleum systems of the Northwest Shelf, Australia: the Dingo-Mungaroo/Barrow and the Locker-Mungaroo/Barrow. U.S. Geological Survey Open-File Report 99-50-E.
• Blevin, J.E., Struckmeyer, H.I.M., Cathro, D.L., Totterdell, J.M., Boreham, C.J., Romine, K.K., Loutit, T.S. and Sayers, J. [1998] Tectonostratigraphic framework and petroleum systems of the Browse Basin, North West Shelf. In: Purcell, P.G. and Purcell, R.R. (Eds.) The Sedimentary Basins of Western Australia 2, 369-395.
• Carruthers, D.J. [2003] Modeling of secondary petroleum migration using invasion percolation techniques. In: Duppenbecker, S. and Marzi, R. (Eds.) Multidimensional basin modeling. AAPG/Datapages Discovery Series, 7, 21-37.
• Chopra, S. and Marfurt, K.J. [2007] Seismic attributes for prospect identification and reservoir characterization. SEG Geophysical Developments Series, 11.
• Evenick, J. [2022] Assessing hydrocarbon charge access via map-based fetch analysis: Going beyond petroleum systems event charts. Geosystems and Geoenvironment, 1(2).
• Higgs, K.E., King, P.R., Raine, J.I., Sykes, R., Browne, G.H., Crouch, E.M. and Baur, J.R. [2012] Sequence stratigraphy and controls on reservoir sandstone distribution in an Eocene marginal marine-coastal plain fairway, Taranaki Basin, New Zealand. Marine and Petroleum Geology, 32, 110-137.
• Holden, N., Osmond, J.L., Mulrooney, M.J., Braathen, A., Skurtveit, E. and Sundal, A. [2022] Structural characterization and across-fault seal assessment of the Aurora CO2 storage site, northern North Sea. Petroleum Geoscience, 28(4).
• King, P.R. and Thrasher, G.P. [1996] Cretaceous-Cenozoic geology and petroleum systems of the Taranaki Basin, New Zealand. Institute of Geological and Nuclear Sciences Monograph, 13.
• Lacaze, S., Pauget, F., Lopez, M., Gay, A. and Mangue, M. [2011] Seismic stratigraphic interpretation from a geological model -a North Sea case study. SEG Technical Program, Expanded Abstracts.
• Longley, I.M., Buessenschuett, C., Clydsdale, L., Cubitt, C.J., Davis, R.C., Johnson, M.K., Marshall, N.M., Murray, A.P., Somerville, R., Spry, T.B. and Thompson, N.B. [2002] The North West Shelf of Australia -a Woodside perspective. In: Keep, M. and Moss, S.J. (Eds.) The Sedimentary Basins of Western Australia 3, 27-88.
• Milkov, A.V. [2015] Risk tables for less biased and more consistent estimation of probability of geological success (PoS) for segments with conventional oil and gas prospective resources. Earth-Science Reviews, 150, 453-476.
• Milkov, A.V. [2017] Integrate instead of ignoring: base rate neglect as a common fallacy of petroleum explorers. AAPG Bulletin, 101(12), 1905-1916.
• Miocic, J.M., Johnson, G. and Bond, C.E. [2019] Uncertainty in fault seal parameters: implications for CO2 column height retention and storage capacity. Solid Earth, 10(3), 951-967.
• Monigle, P.W., Hedayati, T.S. and Goulding, F.J. [2025] Integrated and improved direct hydrocarbon indicators: a step forward in petroleum risk discrimination. AAPG Bulletin, 109(5), 617-636.
• Patruno, S., Green, A., Corcoran, J., Caldarella, D., Scisciani, V. and Przywara, M. [2024] A conceptual CO2 fill-and-spill mega-fairway in the UK Southern North Sea: A new approach to identify and optimise large-scale underground carbon storage (CCS). International Journal of Greenhouse Gas Control, 133.
• Pauget, F., Lacaze, S. and Valding, T. [2009] A global approach in seismic interpretation based on cost function minimization. SEG Technical Program, Expanded Abstracts, 2592-2596.
• Philit S., Souche L., and Collet M. [2025] Stochastic Prospect Assessment Workflow – Example of the Poseidon Case Study. 86th EAGE Annual Conference and Exhibition, Extended Abstracts.
• Rose, P.R. [2001] Risk analysis and management of petroleum exploration ventures. AAPG Methods in Exploration, 12.
• Rudolph, K.W. and Goulding, F.J. [2017] Benchmarking exploration predictions and performance using 20+ yr of drilling results: one company’s experience. AAPG Bulletin, 101(2), 161-176.
• Struckmeyer, H.I.M., Blevin, J.E., Sayers, J., Totterdell, J.M., Baxter, K. and Cathro, D.L. [1998] Structural evolution of the Browse Basin, North West Shelf: new concepts from deep-seismic data. In: Purcell, P.G. and Purcell, R.R. (Eds.) The Sedimentary Basins of Western Australia 2, 345-367.
• Tortarolo, L., Abdallah, H. and Souche, L. [2026] From seismic to simulation: a versatile multi-scenario workflow for subsurface trap evaluation. Sixth EAGE Digitalization Conference and Exhibition, Extended Abstracts.
• Young, A. [1998] Development of the Goodwyn gas field, Dampier Sub-basin. In: Purcell, P.G. and Purcell, R.R. (Eds.) The Sedimentary Basins of Western Australia 2, 535-540.