About This Item

Share This Item

The AAPG/Datapages Combined Publications Database

AAPG Special Volumes

Abstract

AAPG/Datapages Discovery Series No. 7: Multidimensional Basin Modeling - Abstracts, p. 315–387.

AAPG/Datapages Discovery Series No. 7: Multidimensional Basin Modeling, edited by S. Duppenbecker and R. Marzi, 2003

 

Abstract

Where Did We Come From? A 1996-1997 Hydrocarbon System Technology Benchmark Study

Roger Marzi

Advanced Power Technologies, Inc., 1250 24th Street NW, Washington, DC 20037.

Basin modeling and its application are evolving rapidly as is the way it is pursued in different companies. During the recent wave of reorganization we have seen modeling teams gain in importance and lose in importance. However, integration is recognized as the most important advance possible at this time. It appears to be true that integration of geological and to some extend geophysical data has found a home in basin assessment teams, play modeling teams, integrated basin analysis teams or hydrocarbon system modeling teams. While the names and approaches might be slightly different the overall goal is the same. Instead of pushing sub-disciplines further the companies realize more is to be gained by integration of existing technology and develop only where gaps are obvious. A few years ago one major oil company sponsored a benchmark to see how far the industry as a whole moved towards the goal of integration and embracing basin modeling technology. Some of the data collected in this benchmark will be presented and the stage will be set for where we were three years ago. It will be up to the individual companies to decide where in the spectrum of basin modeling technology application they stood three years ago and where they stand now. While the presentation will not identify individual companies, this is the first time this data is presented to a wider audience.

Some key results highlight the relative importance of modeling as indicated by funding levels, the timing of application and the link to other technologies. It appears that most companies spend about 2% of their exploration budget on Hydrocarbon system technology. Most applications seem to occur late in the exploration cycle which might indicate a potential problem. When comparing RampD effort with application effort a dangerous trend becomes obvious. Even before major problems are solved the focus is shifting away from research. The data indicates that the modeling part of Hydrocarbon system work is rarely outsourced while all data collection functions are outsourced to one level or another. Other data presented shows perceived levels of integration, support for modeling software and the distribution of RampD moneys.

The data clearly show that integration of the different disciplines and the timing of the application are the frontiers at this time and it appears that less technical and more organizational or even personal boundaries are in the way of making our field of interest the most valuable of all.

2-D Compositional Basin Modeling and Organic Geochemistry Studies in Recocircncavo and Camamu Basins, Brazil: An Integrated Approach to Understand Petroleum Compositional Changes and Migration Phenomena

Henrique Luiz de Barros Penteado1, Ricardo Peacuterez Bedregal1, Maacutercio Rocha Mello1, Jean-Luc Rudkiewicz2, Franccediloise Behar2, and Feacutelix Thadeu Teixeira Gonccedilalves3

1Petrobras RampD Center, Cid. Universitaacuteria, Quadra 7, Ilha do Fundatildeo, Rio de Janeiro, Brazil; 2Institut Franccedilais du Peacutetrole, 1-4 Ave. du Bois-Preacuteau, Rueil Malmaison, France; 3Independent Consultant, Rio de Janeiro, Brazil

The Recocircncavo and Camamu basins in Northeastern Brazil are two rift basins which were formed during the Late Jurassic and Early Cretaceous as a result of the process of continental breakup between South America and Africa. Whereas the Camamu Basin, located in the offshore area, evolved into a typical passive margin basin with post-rift sediments ranging from Late Cretaceous to Tertiary, most of the sediments of the Recocircncavo Basin are of Early Cretaceous age (Figueiredo et al., 1994). Nevertheless, vitrinite reflectance, apatite fission track and porosity data indicate that more than 1000 m of post-rift sediments might have been deposited in the Recocircncavo Basin, of which only a few hundred meters are preserved nowadays.

The Lower Cretaceous lacustrine shales containing type I organic matter of the Candeias Formation (Recocircncavo Basin; Mello et al., 1994), and Morro do Barro and Rio de Contas formations (Camamu Basin; Gonccedilalves et al., 1997) constitute the main source rocks. Many reservoir levels are found throughout the syn-rift section, but the pre-rift Sergi sandstones, which underlie the source rocks, are the most important ones in the two areas.

The main goals of this study were to investigate the compositional changes petroleum undergoes before and after expulsion by integrating organic geochemical data with basin modeling techniques. To achieve this, the role of secondary cracking of NSO's and aromatics and of the retention effects of source rocks on heavy compounds was examined with the Temispack software. As a preliminary step to this study, modeling of the timing of petroleum generation and expulsion as well as of the migration routes from source rocks to traps in the Recocircncavo and Camamu basins was essential. One dip cross-section in each basin has been selected to be modeled.

A detailed geochemical study has been undertaken in several source rock samples in Recocircncavo Basin to determine their petroleum potential and their degree of maturity. A systematic comparison of Rock-Eval data from source rock samples and their respective isolated kerogens indicated a strong mineral matrix retention effect in the former which had previously led to an underestimation of petroleum potentials. Hydrogen indices of immature kerogens were found to vary between 450 and 850 mg/g TOC according to source rock location and richness. Kinetic data on primary cracking of kerogens of both basins are characteristic of type I organic matter, with one main reaction responsible for more than 90% of petroleum generation. Following the methodology developed by Behar et al. (1997), compositional kinetic data were acquired for the source rocks and were used as input data in the basin modeling study. It is worth mentioning that the C15+ fraction constitutes more than 75% of the products of the main primary cracking reaction.

Furthermore, petroleum extraction has been performed on samples covering a wide range of maturity to investigate the compositional changes of petroleum in source rocks. Liquid and gas chromatographic data indicated a clear trend of enrichment in saturates from the immature stage up to peak petroleum generation which was accompanied by a decrease in the relative abundance in aromatics and NSO's. Whereas C15+ extracts of immature samples are typically composed of 33% saturates, 11% aromatics and 56% NSO's, extract compositions at the depth of peak petroleum generation (around 2500–2600 m) resemble those of reservoired oils (72% saturates, 14% aromatics and 14% NSO's). The increase in saturates in both relative and absolute amounts has been attributed not only to production by primary cracking but also to an important secondary cracking of NSO's and aromatics within the source rocks. Below the depth of peak petroleum expulsion, the decrease in saturates was associated with their preferential expulsion from source rocks. Additionally, the amounts of organic extracts, together with porosity data, were used to estimate oil saturations, which varied from 3 to 15% at the depth of peak petroleum expulsion.

Vitrinite reflectance, kerogen transformation ratios based on the analysis of Rock-Eval pyrolysis natural series and temperature data were used to calibrate the thermal history of the model, which was constrained by estimations of crustal stretching factors. Given the uncertainties regarding the thicknesses of post-rift sediments later eroded in the Tertiary, a series of post-rift scenarios was tested, including lateral variations in thickness and different rates of erosion. In the Recocircncavo Basin, a maximum of 1200 m of an Aptian-Cenomanian post-rift section in the central portion of the basin (Dom Joatildeo High) allowed the best calibration of thermal and maturity parameters. For the Lower Cretaceous source rocks in both basins, geochemical modeling suggested that petroleum generation started at the end of the rift phase and extended into the Late Cretaceous.

Secondary migration was modeled by assigning porosity, permeability and capillary pressure data to each lithologic type, including those attributed to the zones of syn-rift normal faults. Migration modeling was constrained by the location of known petroleum accumulations and estimated residual oil saturations. Two important migration systems have been identified. In the first one, petroleum generated in source rocks migrated downwards to reservoirs of the pre-rift section and updip through permeable faults until structural traps where the Sergi sandstones are the main reservoirs. In the second migration system, petroleum migrates updip along the source rocks and accumulates in the intercalated turbidites, with some loss through the overlying rocks.

The compositional changes during petroleum generation, expulsion and migration were modeled by coupling of primary and secondary cracking, and retention phenomena. Secondary cracking parameters were constrained on the one hand by laboratory results on specific classes of compounds (Behar et al., 1992), and on the other hand by the parameters of the main primary cracking reaction for type I kerogens. Retention factors for NSO's ranging from 25 to 75% were tested. The Temispack software was therefore used as a means to estimate the values of these variables which would allow a good match of modeled results with observed geochemical data. An excellent calibration of both extract and oil compositions was obtained with secondary cracking parameters of NSO's and aromatics very close to those of the main primary cracking reaction (activation energy around 1 kcal/mole higher for the same frequency factor) and a 50% retention factor of NSO's within the source rocks. Secondary cracking of oil accumulated in reservoirs was found to be negligible.

REFERENCES CITED

Behar, F., S. Kressman, J.-L. Rudkiewicz, and M. Vandenbroucke, 1992, Experimental simulation in a confined system and kinetic modeling of kerogen and oil cracking, in C. B. Eckardt, J. R. Maxwell, S. R. Larter, and D. A. C. Manning, eds, Advances in Organic Geochemistry 1991, Organic Geochemistry, v. 19, p. 173–189.

Behar, F., M. Vandenbroucke, Y. Tang, F. Marquis, and J. Espitalieacute, 1997, Thermal cracking of kerogen in open and closed systems: Determination of kinetic parameters and stoichiometric coefficients for oil and gas generation: Organic Geochemistry, v. 26, p. 321–339.

Figueiredo, A. M. F., J. A. E. Braga, H. M. C. Zabalaga, J. J. Oliveira, G. A. Aguiar, O. B. Silva, L. F. Mato, L. M. F. Daniel, L. P. Magnavita, and C. H. L. Bruhn, 1994, Recocircncavo Basin, Brazil: A prolific intracontinental rift basin, in S. M. Landon, ed., Interior rift basins: AAPG Memoir 59, p. 157–203.

Gonccedilalves, F. T. T., R. P. Bedregal, L. F. C. Coutinho, and M. R. Mello, 1997, Petroleum system of the Camamu-Almada Basin: A 2-D compositional modeling approach, in M. R. Mello and B. Katz, eds, Extended Abstracts of the AAPG/ABGP Hedberg Research Symposium on Petroleum Systems of the South Atlantic Margin.

Mello M. R., E. A. M. Koutsoukos, W. Mohriak, and G. Bacoccoli, 1994, Selected petroleum systems in Brazil, in L. B. Magoon and W. G. Dow, eds., The petroleum system—From source to trap: AAPG Memoir 60, p. 499–512.

On the Use of Fault Seal Methodologies in Hydrocarbon Migration Modeling

Sylta1, Borge1, Childs2, Walsh2, and Manzocchi2

1Sintef Petroleum Research; 2Fault Analysis Group, University of Liverpool

In multiple reservoir hydrocarbon systems geometrical connections across faults may cause oil and gas to migrate into different carriers. In simple geologic systems migration across faults from e.g. a younger carrier into an older carrier may be easily traced by a skilled explorationist using e.g. Allan mapping. This mapping shows where sand-sand juxtaposition occurs along a single fault. However, for large and complex geologic systems that now can be mapped using 3-D seismic data this exercise becomes extremely labour-intensive, and is therefore generally not performed. Using a 3-D secondary migration simulator it is possible to automatically determine where geometric leak points are most likely to occur, thus adding more information about migration pathways to the risking process.

A more accurate determination of four-way closures results from the above methodology, but many geologic systems also contain sealing fault surfaces which prevent migration between juxtaposed sands. Predicting fault seal behaviour can be challenging, and drilling prospects that require fault seal to be economic is therefore often considered to involve too much risk. One reason for this is the lack of accurate predictive techniques to quantify fault seal potentials. The lack of suitable techniques prevents systematic a posteriori testing of fault seal prediction. Due to the complex and heterogeneous properties of faults (planes) and the disturbed zones around faults, caused by fractures, complex clay smearing and diagenetic processes, one may conclude that forward predictive models cannot be constructed for migration in faults and fractures. Such a viewpoint is however over-pessimistic and we suggest that simple models can provide important constraints on the decision-making processes prior to drilling such prospects.

Borge and Sylta (1998) have shown that pressure modeling can reproduce the pressure distribution in a sedimentary basin, e.g. the North Viking Graben in the North Sea, using a simple description for the behaviour of fault transmissibilities versus fault throw and burial depth. Their model included a first-order description of the fault transmissibilities, neglecting the capillary properties of the fault rocks because only single phase migration was considered. Using a pressure compartment methodology that has later been expanded to a full 3-D solution, they were able however to reproduce sharp lateral pressure transition zones that are commonly observed in basins and are attributed to the presence of low permeability faults. Thus, it is suggested that faults can be treated with a fair amount of accuracy in pressure modeling applications.

Proposed mechanisms for generating fault seal include incorporation of clay and shale into the fault zone. Empirical data from known sealing faults suggest that the sealing potential of a fault can be related to the proportion of shale within the faulted sequence. The percentage shale within the part of a sequence that has moved past a point on the fault surface is termed the Shale Gouge Ratio (SGR). Published calibrations of SGR against fault sealing potential exist (Fristad et al. 1997). We incorporate the effects of fault seal by the shale smearing mechanism into a ray tracing migration modeling system. Fault seal is included by modifying the grid defining the top carrier elevation so that grid cells within fault polygons are depressed by an amount equal to the potential hydrocarbon column height supported by the fault. A fault offsetting the carrier bed is treated as a "curtain" that hangs beneath the top of the carrier. The ray tracing method employed will then locate the highest across-fault spill point.

Evaluation of potential hydrocarbon column heights is based on SGR values calculated for a representative sequence containing the carrier interval. The SGR values are calculated using a sequence/throw juxtaposition diagram for a notional fault with throw values up to that of the sequence thickness. The SGR values are then transformed to potential hydrocarbon column heights either by a direct relationship or via a relationship between SGR and fault rock breakthrough pressures (Gibson 1998).

The method of estimating fault seal capacity from the breakthrough pressure for a fault at a particular throw value depends on whether the carrier interval is entirely or partially offset. When the fault throw is less than the thickness of the carrier interval i.e. the carrier is self-juxtaposed, oil leakage may occur across the fault. Minimum breakthrough pressures of the fault rock separating the up-thrown and down-thrown carrier interval determine the column height that can be supported by the fault. The minimum buoyancy pressure required to exceed the fault zone breakthrough pressure therefore determines the oil column height. This case is represented in the carrier bed topography by a grid cell that is depressed by an amount equal to this minimum column height.

For fault throws that exceed the thickness of the carrier interval, oil leakage may occur along the fault surface. Fault zone breakthrough pressures along the portion of the fault that connects the down-thrown and up-thrown carrier interval then govern the height of the trapped hydrocarbon column. Within the carrier bed grid, a fault polygon separating up-thrown and down-thrown reservoir is represented as a grid of elevations of potential depths of hydrocarbon column heights. The result of migrating along the modified elevation grid corresponds to deriving a path of minimum resistance to flow across a grid of fault surface breakthrough pressures and is equivalent to modeling migration within the fault zone.

With reference to selected examples we show that leakage from the weakest point of faults results in very focused hydrocarbon migration. Focusing of gas spillage may thus allow oil accumulations to be preserved in the "shadow" of faults, a fact which is consistent with some cases in nature. We suggest that the inclusion of fault sealing properties in migration modeling and sensitivity analysis of their affect on migration should reduce risk in the drilling of future fault seal prospects.

REFERENCES CITED

Borge, H., and Oslash. Sylta, 1998, 3-D Modeling of Fault Bounded Pressure Compartments in the North Viking Graben: Energy exploration amp exploitation, v. 16, no. 4, p. 301–323.

Fristad, T., A. Groth, G. Yielding, and B. Freeman, 1997, Quantitative fault seal prediction—A case study from Oseberg Syd, in P. Moslashller-Pedersen and A. G. Koestler, eds., Hydrocarbon Seals—Importance for Exploration amp Production 7: Trondheim, Norway, Norwegian Petroleum Society, p. 107–124.

Gibson, R. G., 1998, Physical character and fluid-flow properties of sandstone-derived fault zones, in M. P. Coward, T. S. Daltaban, and H. Johnson, eds., Structural Geology in Reservoir Characterisation 127: London, Geological Society, p. 83–97.

A New 2-D Basin Modeling Tool for HC Potential Evaluation in Faulted Area: Applications to the Congo Offshore and to the Bolivian Sub Andean Zone

F. Schneider, H. Devoitine, I. Faille, F. Flauraut, F. Willien, and S. Wolf

Institut Franccedilais du Peacutetrole, 1 amp 4 avenue de Bois-Preacuteau 92852 Rueil-Malmaison Cedex-France

Integrated basin modeling provides a strategy for optimizing exploration in frontier areas and evaluating new plays within well-explored basins. Ideally, a basin simulator should span the entire process of source-rock burial, hydrocarbon generation, expulsion, migration into a potential trap, and assessment of trap integrity throughout the evolution of a basin.

Today's state-of-the-art sedimentary basin models are able to handle relatively simple geometries resulting from deposition, erosion and vertical compaction using porosity-effective stress laws. Two main areas, however, are insufficiently or not at all treated: (1) Basin geometry is often not precise enough. Most basins are cut by faults with significant offsets. In addition, basin geometry may be affected by creeping salt or mud which in turn may lead to the formation of diapirs. (2) Fluid flow and convective heat transfer do not handle the permeability evolution of faults correctly. A fault can be pervious or act as a seal, and it can change its behavior through time.

The objective was to build an advanced prototype of 2-D sedimentary basin simulator that can handle non-vertical faults and salt tectonics. The simulator includes basin-scale compaction, overpressure generation, water flow, and temperature evolution as well as the genesis and migration of hydrocarbons through geologic time.

The prototype allows us, as all basin modeling tools do, to edit the observed present-day section with all its complexity. Then a backward process which includes a restoration module and a backstripping module allows us to determine the input data for the forward calculator. The forward simulator computes all of what is generally calculated in a basin model (temperature, maturity, fluid pressure, hydrocarbon saturations hellip) accounting for faults (geometry and property) and creeping.

This prototype called Ceres 2-D has been used to study a petroleum system in the Bolivian foothills where faults are supposed to be the most important carriers beds for hydrocarbon migration. It has also been used to study the impact of faulting on maturity of the organic matter and hydrocarbon migration in the Congo offshore.

Transport Modeling of Secondary Migration Using Extensible Invasion Percolation Techniques

Daniel J. Carruthers and Mike de Lind van Wijngaarden

The Permedia Research Group Inc., 53 Fourth Avenue, Ottawa, Ontario, Canada K1S 2L1

At the time scales associated with secondary migration, the trajectories and characteristics of the hydrocarbon flows are overwhelmingly controlled by the balance between the gravity and the capillary forces (i.e. the Bond number) (Carruthers and Ringrose, 1998). At these low flow rates, viscous forces are negligible, and can effectively be ignored. From a practical standpoint, this means that techniques used for modeling oil flows at production time scales introduce "computational overkill," and may in fact be entirely inappropriate.

Offering a potential solution are invasion percolation (IP) techniques, which have been shown to accurately reproduce drainage conditions for capillary-dominated flows (e.g. Frette et al., 1992; Meakin et al., 1992; Meakin et al., 1994). However, these models have been historically limited to length-scales of a few tens of centimetres, since traditional IP algorithms contain two fundamental flaws when larger scaled models are considered: (1) the assumption that the non-wetting phase invades via a single front, and (2) the algorithm considers the invading phase to be continuous.

Techniques have been developed which remove these limitations, resulting in migration models that can simulate flows ranging from the core-scale to the basin-scale (Carruthers, 1998; Carruthers and Ringrose, 1998). Of particular importance to the basin modeler/explorationist is the speed of these models. Models containing tens of millions of gridcells can be evaluated in a matter of minutes, whereas traditional methods are unable to handle models of this resolution.

These new techniques have proven to be highly extensible. Specifically, they have been shown to:

  1. Capture the uncertainty inherent in the variables

  2. Be equally applicable to oil and gas migration

  3. Enable the concurrent modeling of oil and gas flows

  4. Accurately represent accumulated pressure heads necessary for breaching capillary baffles

  5. Enable the modeling of migration through highly detailed fracture models

  6. Track backbones of high fluid fluxes, critical for accurate geochemical modeling

  7. Simulate migration along multiple, concurrent fronts

  8. Greatly facilitate the risking of migration pathways and accumulations through their ability to combine the results from hundreds of equally probable realisations

  9. Enable the modeling of migration in dynamic carrier and basin configurations.

This work will demonstrate these extensions through a series of examples at different length scales. The results will be compared with those generated using traditional migration modeling tools.

REFERENCES CITED

Carruthers, D., 1998, Transport modeling of secondary oil migration using gradient-driven, invasion percolation techniques: Ph.D. thesis, Department of Petroleum Engineering, Heriot-Watt University, Edinburgh, UK.

Carruthers, D., and P. Ringrose, 1998, Secondary oil migration: Oil-rock contact volumes, flow behaviour and rates, in J. Parnell, ed., Dating and duration of fluid flow and fluid-rock interaction: London, Geological Society, Special Publication 144, p. 205–220.

Frette, V., J. Feder, T. Jossang, and P. Meakin, 1992, Buoyancy-driven fluid migration in porous media: Physical Review Letters, v. 68, p. 3164–3167.

Meakin, P., J. Feder, V. Frette, and T. Jossang, 1992, Invasion percolation in a destabilising gradient: Physical Review A, v. 46, p. 3357–3368.

Meakin, P., G. Wagner, V. Frette, T. Jossang, J. Feder, and Birovljev, 1994, Gradient-driven migration in porous media: Experiments and simulations, North Sea Oil and Gas Reservoirs: Kluwer Academic Press, p. 297–305.

The Inevitability of Seal Formation and FailuremdashA Hydro-Geomechanical View of the Physical World

G. D. Couples1, H. Lewis1, N. Bicanic2, and C. J. Pearce2

1Department of Petroleum Engineering, Heriot-Watt University, Edinburgh EH14 4AS, Scotland; 2Department of Civil Engineering, Glasgow University, Glasgow G12 8QQ, Scotland.

Recent studies of mudrocks reveal a complex layered heterogeneity of primary particle types (and, hence, composition) in otherwise monotonous-looking sequences. Other studies serve to relate such variations in material type to variations in material properties (for example, oedometer tests on differentwe can conclude that sealing capacity has been achieved (for such a two-phase system, the critical parameter is the capillary entry pressure, which is directly linked with the pore-throat size distribution, and hence to the same phenomena as is the permeability reduction). Since both of these types of evidence abound in basins, we conclude that the formation of flow retarders/barriers (seals) is essentially inevitable in typical basins. However, we also know that fluid pressures do not build indefinitely, but are somehow limited. Thus, seals have only a certain capacity, and any increment of excess fluid pressure, or a change in the rock-framework mechanics, can be causally associated with the failure of the seal.

A valuable tool for understanding seal formation is Poro-Visco-Plasticity (PVP)—a material description that is tightly linked with the concept of yielding. PVP draws on the extensive literature describing the laboratory testing of rocks and soils that has been under way for many decades. PVP explicitly includes porosity as a state variable for materials. It predicts modes of failure (dilational/compactional/constant-volume) in terms of three parameters: mean effective stress, "stress difference" (expressed as an invariant of the stress tensor), and porosity (or void ratio). PVP correctly predicts the yielding/stop-yielding behaviour of materials subjected to fluid-pressure changes in both normally and over-pressured cases. It also incorporates knowledge concerning the effects of deformation (loading) rates. Alterations of physical properties (eg conductivity, permeability, sonic velocity) that may be affected by deformation (especially by changes in porosity) are treated as attached "functions," enabling PVP to describe the evolution of the material properties of basin rocks as they are deformed.

Although PVP is useful for predicting seal failure (due either to fluid-pressure increase or to rock deformation), a more complete appreciation of this process, and of the consequent re-distribution of fluids, is gained through the use of discontinuum approaches. HYDRO-DDA is a coupled simulator for fractured/fracturing rocks and their contained pore fluids. DDA is a type of discrete element model in which a fully implicit scheme and a sophisticated contact-detection approach allow blocky materials to undergo continuing fracturing and block motions. In the context of seal failure, DDA can be used to predict rock failure (fracture), and the evolution of the pattern of fractures as the deformation state changes in consequence of fluid movement through the newly formed fracture network. HYDRO simulates fluid flow in both matrix and discontinuities. We are currently developing a scheme to allow these two environments to provide feedback to one another, and so be able to better understand the way that seals can serve as pressure valves.

These simulation environments provide a means for understanding how mudrocks and pore fluids can interact in a subsiding basin to form a self-organizing system. The new simulation capabilities, in conjunction with others now available, will allow us to address topics such as:

  • The role of variability in primary deposition as it feeds through, in time, to create sealing/overpressure arrangements.

  • The question of: What IS a seal? Is it a well-compacted layer, or an under-compacted layer? Or a composite of both? Are there multiple physical types?

  • How under-compacted layers can be preserved to function as flow conduits.

  • How pressure transmission (from sites laterally or vertically removed) affects seal creation/failure processes.

  • How tectonism, or other structural events (including differential compaction), can be related to sealing capacity.

Gravity and Two Dimensional Basin Modeling, Offshore MyanmarmdashExplanation of Regional and Local Thermal Maturity and CO2 Variations

Martin D. Matthews1, Michael A. Kowalski2, Vaughn R. Robison2, David B. Ephraim2

1Argonne National Lab.; 2Texaco Exploration

Heat flow in the Northeastern Gulf of Martaban currently varies spatially by a factor of at least 2 and by nearly an order of magnitude over time. This variation is due to the recent rifting in the basin. Higher heat flows occur near the rift centers and decrease outward to a regional background in different domains, both for wells and oceanographic surface measurements. This variation is significant both to hydrocarbon generation/migration and CO2 origin. The magnitude of the regional background heat flow is unknown. It is at least as low as our lowest value, perhaps lower. The distribution of available data is too sparse for meaningful determination.

Basin modeling was performed using rock properties derived from logs and constraining thermal history by bottom hole temperatures and vitrinite. This study indicates that there are four thermal events, separated in time and space: formation of the sag, followed by a cooling-off period; renewed basin subsidence accompanied by faulting; progressive heating events proceeding from the Northern to the Southern platform.

The kerogen in the Miocene source rock, within the deep basin, is currently in the dry gas window. Kerogen essentially ceased to generate significant quantities of hydrocarbons about the time the source reached the depth of dry gas production. Thus, any recent production of gas from the deep basin should be from non-expelled heavier hydrocarbons. Maturation has continued to progress shoreward up the transition zone towards the platform.

The kerogen in Miocene source rock on the shallow platform recently entered the oil window. Kerogen transformation at the level of the Miocene source is 5% or less. In the areas of lower heat flow the source is immature. Generation in the transition between these two areas is gradual in the absence of a let-out fault, the upper zone currently undergoing active generation, the deeper zones having done so earlier. Where the let-out fault is present, the kerogen in the source section landward of the fault is usually less than 70% transformed to petroleum.

The migration of fluids is dominantly vertical in the deep basin and divided into 2 or 3 lateral flow units in the transition zone and the platform. The lower flow unit is constrained below the source rock.

The CO2 concentration is expected to be generated and largely confined below the overpressured Miocene section. The probability of CO2 generation is hypothesized to be greatest in, and adjacent to, regions of highest heat flow and in the first trap out of the basin.

The distribution of current heat flow has been divided into two classes, low and high. The spatial location of these classes appears to be related to Basement rifting and later intrusions. The existing satellite free-air gravity data is the key to understanding the spatial changes and origin of the different heat flow regimes.

One, Two, and Two and a Half Dimensional Combined Depositional and Fluid Flow/Maturation Modeling of the Southern Portion of Bohai Bay, China

Martin D. Matthews1, Kelly Dempster2, Jeff R. Johnson2, and Mona Bissada2

1Argonne National Lab; 2Texaco Exploration

Well data is used to constrain the rock properties in the Southern portion of Bohai Bay. Global cyclostratigraphic analysis is combined with well data and seismic stratigraphy to understand the spatial and time variation of stratigraphic patterns and tectonics in the region of study. This information is used to construct a 3 dimensional model of the stratigraphic evolution of the basin in order to construct 2-D fluid-flow/maturation cross sections and areal 2-D migration maps.

The early syn-rift, Kongdian and Shahejie 4, is characterised by locally derived rivers and sags in an alternating dry (saline lake) and wet (fresh lakes) climate, favoring the deposition of lith-arkosic sediments. This was followed by syn-rift conditions, Shahejie 3 to Dongying. River basins enlarged by headward erosion, providing a more constant water supply by transporting water and sediment form multiple climates into the basin of deposition, particularly when the Himalayas rose. Subsidence rates increased, resulting in the potential for deeper lakes. Chemical weathering increased with sand compositions averaging 1/3 quartz, 1/3 lithics, and 1/3 feldspar. Locally, sands rage upwards to 70% quartz. Post rift conditions, Guantao to recent, are typified by reduced subsidence and throughgoing river systems. These cause fine sediment to bypass the basin. Presently the Basin is covered by marine waters but is being rapidly reclaimed to sub-areal conditions by river deposition.

The themal history of the region is constrained by bottom hole temperatures and vitrinite data as a function of depth. Variation in the data is due to local uplifts and varying depth to the Moho. Pressure modeling, constained by RFT and DST measurements, indicates that both compaction and kerogen maturation are important pressure generation mechanisms in the region. Faults are believed to be important pathways for transmitting overpressure to the shallower sections.

A Conceptual Model for Thermal History of Sedimentary Basins Using Thermodynamics

Andy Carr

Advanced Geochemical Systems Ltd., 1 Towles Fields, Burton on the Wolds, Leics., LE12 5TD. UK

Hydrocarbon generation is a temperature dependent process and any uncertainty regarding the temperature history of sedimentary basins is a risk to successful hydrocarbon exploration. Two methods are available for predicting temperature history. Tectonic subsidence models (e.g. McKenzie, 1978) predict the temperature history as a function of extension (and subsidence), whereas calibration models use temperature dependent maturity parameters (e.g. vitrinite reflectance) to solve for temperature at a given time. In basins such as the northern North Sea the thermal history predictions obtained from the tectonic models differ significantly from those produced by calibration models. Both thermal history predictions cannot be correct, since for any point in the geological history of a basin, there can only be one correct value for heat flow at that time.

Burial or uplift of a sediment involves the movement of mass under the influence of energy supplied to that sediment. This movement of sediments is analogous to the movement of loads (masses) in many other technical disciplines, e.g. engineering. The method used to analyse the amount of energy required to move such loads involves the first law of thermodynamics:
Equation 1

where u = internal energy, q = energy (heat) supplied and w = work. Before using thermodynamics it is essential that the system and its boundaries are defined before any further analysis occurs. For this study the system is defined as a layer of rock/sediment, e.g. shale, which is buried during a given time interval, and whose boundaries are the layers of rock above and below, while two fault systems define the lateral boundaries. The extent (length, width and thickness) of the layer is used to calculate the mass.

The internal energy consists of the potential, kinetic and thermal energy of the sediment (system). The increase in temperature during burial is included in the thermal energy (m.c.dt) of the system. This term, as indeed all terms involving mass, must include the energy within both solids and liquids mass (m). Kinetic energy (0.5 mv2) is discounted due to the slow burial rate of all geological basins when expressed as metres/second. Potential energy is determined by the change in vertical height (m.g.h). Note that although the earth's surface is normally used as the reference surface, rocks would continue to fall if unsupported as a result of the potential energy still remaining within the rock. The calculation of the heat supplied needs to include heat conducted from below, heat transferred via convection and internal radiogenic heat, as well as accounting for heat lost via conduction and convection from the sediment.

The work (force (m.g) lowast distance moved) done by and on the system by external systems is the most difficult term to identify. The problem arises from the work done by layers above the sediment. Is the subsidence of a sedimentary layer caused by the load (mass of overlying layers)? If this is the case, then the force of the sediments above must exceed the opposite force of the sediments from below. Since the amount of mass below a sedimentary basin far exceeds any mass within the basin, the force (m.g) of the underlying rock mass will exceed the force of the overlying load. In this model the only work done by the overlying sediments is compaction, and subsidence is the result of other changes occurring within the system.

The heat flow values that occur with different amounts of subsidence are calculated. The results show that extensive basin subsidence occurs during periods of reducing heat flow, while regional inversion would occur during periods of increasing heat flow. Extensive periods (gt10 M.a.) of constant heat flow commonly used by basin modelers in extensional basins are inconsistent with subsidence that frequently occurs in the basins during the same time interval.

Multidimensional Burial and Thermal History Modeling of Mesozoic and Cenozoic Strata in the Mississippi Interior Salt Basin, Northeastern Gulf of Mexico

Ernest A. Mancini, William C. Parcell, and T. Markham Puckett

Department of Geology and Center for Sedimentary Basin Studies, Box 870338, University of Alabama, Tuscaloosa, Alabama 35487-0338

Multidimensional burial and thermal history modeling with an application to petroleum systems analysis and petroleum exploration has been performed for Mesozoic and Cenozoic strata of the Mississippi Interior Salt Basin (MISB), northeastern Gulf of Mexico. The MISB is the largest interior salt basin in the northern Gulf of Mexico. The tectonic and depobasin. Five regional cross sections consisting of 48 representative wells, regional structural maps on key stratigraphic horizons, and regional isopach maps of genetic stratigraphic intervals comprise the basis of the study. Information interpreted from well samples, outcrop data, cross sections, and subsurface maps include biostratigraphic, paleoenvironmental (water depths), lithologies and stratigraphic thickness of the units, compaction, sedimentation and subsidence rates, unconformities and faulting. Burial and thermal history modeling was performed using BasinModreg software.

Burial history modeling is consistent with the rift-related geohistory of'the basin. Tectonic subsidence rates were greatest during the Jurassic and Early Cretaceous and decrease progressively from the Jurassic and Early Cretaceous through the Late Cretaceous and Tertiary reflecting the post-rift history of the'basin. The highest mean sedimentation rates correlate to the highest tectonic subsidence rates. The variation in sedimentation rate is related to lithology, unit thickness, and duration of deposition. Deepest water depths in the basin occurred during the Oxfordian, Hauterivian-Albian, Cenomanian-Turonian and Priabonian. These events correspond to global rises in sea level indicating that eustasy contributed to the generation of accommodation space in the basin.

Thermal maturity modeling was accomplished utilizing vitrinite reflectance, thermal alteration index, Tmax, bottom hole temperature, type of kerogen, and total organic carbon data. Maturity modeling indicates that Oxfordian carbonate mudstones (rich in algal/microbial kerogen) were effective regional source rocks throughout the basin. Significant oil generation commenced from these carbonate mudstones in the Early to Late Cretaceous and continued into the Paleogene. Upper Cretaceous (Cenomanian to Turonian) black shales (rich in herbaceous kerogen) were effective source rocks locally in the area of the Perry sub-basin. Oil generation was initiated from these shales in the Paleogene. Lower Cretaceous sediments have experienced thermal conditions conducive to oil generation in the Perry sub-basin area, but the source rock potential of these sediments is dependent upon favorable organicfacies being present. The burial and thermal histories of Paleogene shales (high in organic matter) were not conducive for the generation of oil in the MISB.

By utilizing a multidimensional burial and thermal history modeling approach, petroleum companies searching to extend known oil and gas plays in the MISB should be able to generate new prospects and improve their ability to drill successful exploration wells (Figures 1 and 2).

Figure 1.

Figure 2.

Interactive Charge Modeling of the Qatar Arch Petroleum Systems

Zhiyong He1 and Thomas A. Berkman2

1ARCO Exploration Technology and Operations; 2ARCO Latin America

An easy-to-use map-based charge modeling package was developed at ARCO to model the history of petroleum generation, migration and accumulation. Building of a 3-D data cube is not required. The system can be used to simulate migration interactively, usually at much higher horizontal resolution than typical 3-D models can tolerate, and is amenable to limited input data sets (e.g., one map and few wells). Basin models can be updated so that the complete paleo-structure, maturation and migration history of the petroleum system can be reconstructed and visualized, as more data become available.

The Qatar Arch is a broad, north-trending anticlinorium, separating two deep salt basins. The Middle Jurassic Hanifa Formation is the primary source for the oil produced in the region. In this study, the maturation history of the Hanifa is modeled using kinetic models. The map view of the oil window through geological time is displayed as a Quick Time movie to show the spatial variations as well as timing of generation relative to structural development.

At 65 Ma, salt structures were already forming and were surrounded by an actively generating kitchen. A quantitative calculation of the charge volumes, based on the migration and fetch, shows that these structures trapped most of the oil generated in the oil window, therefore limiting charge to the crest of the arch. The crest of the arch thus produces limited amounts of mostly lower-maturity (lower-gravity) oils from undercharged reservoirs.

Interactive migration modeling based on the Hith structure surface, which mimics the top Arab reservoir, shows that migration patterns are affected by small-scale variations of the structure surface. Focusing by these minor features resulted in meandering migration channels as well as large areas of "migration shadows." These features are confirmed by active shows and production on or near the migration "channels," and by wells characterized by poor shows and lacking production in the migration shadows.

The predicted migration patterns are also consistent with geochemical analyses that demonstrate the existence of several oil families across the arch. Each of these oil families resides in a separate fetch area defined by the migration modeling.

Charge is a significant risk associated with the Cretaceous plays throughout much of Qatar because of the lack of identified Cretaceous source rocks and the presence of the 250–300-foot-thick Upper Jurassic Hith anhydrite seal between the Hanifa source and potential Cretaceous reservoirs. In addition, hydrodynamic effects may be present across the arch as evidenced by a northeast-tilted potentiometric surface and salinity increase. By accounting for the hydrodynamic processes and using salt piercements as possible entry points, migration pathways were simulated. The predicted locations and shapes of accumulations in the Cretaceous coincide with existing production from the Al Shaheen Field, located northeast of the structural crest.

We understand that most of the data required for petroleum system modeling carries a high degree of uncertainty. It is necessary to frequently modify assumptions to test alternative scenarios and to determine the sensitivity of modeling results to data input. In the Qatar and other studies using our modeling programs, the interactive nature of the programs allows rapid, multiple realizations and sensitivity analyses. For example, the hydrodynamic and hydrostatic cases can be compared by merely selecting a different button in the program.

Highly optimized algorithms in the migration model allowed us to take advantage of high-resolution map data. Running the application on a desk-top PC, we typically use map grids with 40,000–1,000,000 nodes, which far exceeds the limitations of most 3-D models. This enabled us to perform "what-if" modeling of detailed migration patterns and make accurate charge calculations on an interactive basis.

Some Thoughts on the Geometric/Mechanical Evolution of BasinsmdashNew Avenues for 2-D and 3-D Basin Models

G. D. Couples

Department of Petroleum Engineering, Heriot-Watt University, Edinburgh EH14 4AS, Scotland

As numerical technology makes it (nearly) affordable to apply 3-D modeling to our work, should we abandon the 2-D technology? Not yet. There is a lot of life left in the 2-D approach, especially when we consider how basin modeling can be used. By this I mean that 2-D basin modeling can be used to help us understand the range of processes and events that may have occurred in a basin, while retaining a simpler input and less overhead than in 3-D modeling. I favour an approach in which basin models are used to investigate how geologically plausible variants (stratigraphic age uncertainties, episodic deposition, paleo-hydrothermal events,etc.) may have impacted the system. If such variations have a major impact on the predictions of a basin model (such as whether hydrocarbons accumulate or by how much), then these risks need to be considered. I believe that it is important for basin modeling to develop good ways in which these "other" processes can be considered by geoscientists.

The following areas illustrate the (renewed) potential for 2-D modeling technology.

  • Basin models currently interpolate between control points. Those controls may be wells or pseudo-wells, maps (that have been smoothed by the mapping algorithms), or seismic picks. Most control points are in the sedimentary section, and rarely do they include actual basement. Yet we know a lot about basement behaviour. It is usually seen that basement rocks deform by semi-rigid motions of fault-bounded blocks. This appreciation could be incorporated into basin models, and indeed, the default could be arranged to include basement as an essential element of a model.

  • The advantages of making this change can be realized if basin models are also adapted to include realistic rock mechanics for the evolution of basin structures. There is a vast quantity of work available for the kinematics of structures, and this is now being subjected to full mechanical modeling. The use of realistic basement motions as a lower boundary condition will allow such modeling to contribute to the forward calculations of basin evolution. Staying with two dimensions is probably a sufficient challenge for this task.

  • Realistic tectonic motions are essentially missing from the current generation of basin models. Reverse faults (in 2-D) and strike-slip faults (definitely 3-D), believable folding, and even realistic normal faults are all very difficult to handle with the data structure that is presently used. One way forward may be to adopt a different formulation. Object Elementscopy is one potential solution; this method is derived from the object-oriented paradigm, and carries historical and state information within an evolving information structure that is linked with deformation.

  • Although it is an implicit element of the previous two points, a mechanistic treatment of compactional deformation is an essential improvement that is overdue to be included in basin models. 2-D models are a good place to start, since there are reasonable assumptions that allow simplifications of the full mechanical state.

  • On a broader scale, the change to a mechanical paradigm necessitates having material descriptions for the range of rocks comprising a basin. Such material descriptions can now be related to petrophysical alterations, and to changes in other material properties, so leading to a basin model in which the various transport processes are a function of the evolution of the basin. Even in 2-D, this area opens a'large range of opportunities for analysis and discovery, especially with regard to flow barriers and flow conduits.

  • Current basin models are rather poor at treating hydrogeological processes—especially convection. In part, this problem is related to the (normal) choice not to calculate water density and viscosity, but it is also related to the typical size of grid blocks. There is an avenue open for research to make it possible to include such hydrothermal effects (a form of process up-scaling) while retaining the efficiencies of larger grid sizes. The 2-D environment is perfectly suited for a development platform in this topic.

  • Another research/development avenue for basin modeling is in the area of geodynamics. Lithospheric processes are ultimately responsible for the motions of the basement blocks and for the resulting "tectonic" accommodation space. They are also the means by which heat is delivered into a basin. Present and future understanding in geodynamics can be better used in basin modeling if basin models are re-designed to be better multi-physics simulators, as suggested above.

  • Process models of deposition are now becoming available. These could be integrated into basin models to produce a "sensible" distribution of rock (material) types during the forward phase of the calculations. Such variations in material type will, of course, have effects on the distribution of deformation (including compaction), and thus there is considerable scope for feedbacks and self-organizing behaviour within the model.

  • We are all familiar with geoscientists' ability to employ a type of fuzzy logic when interpreting geological history. The good practitioners can "eyeball" how a line or plane can be fitted to a data set, and they are quite good at separating bad data from anomalous and, therefore, interesting, data. Why can't we develop some tools that assist this process within the complex environment of fully populated basin model? Examples might be: age uncertainties, possible turbidites within a deep-water section, slumps, etc.). Surely, a 2-D system is the appropriate starting point for such work.

  • Academic researchers are essentially end users of basin modeling technology. By this, I mean that we typically only have the use of commercial products (they are black boxes with some documentation about what they claim to be doing). Yet if academics are to develop their basin modeling ideas—beyond the point of making a list (like the one above)—there is a need to be able to implement those ideas. This requires a numerical environment in which "standard" process calculations are already implemented, but where new process calculations can be introduced. It is silly to think that an academic researcher could afford to re-invent a complete modeling tool. It would be much better, I think, if academics could gain access to existing environments. Is it, perhaps, time for the commercial vendors to enter into partnerships with selected academics? The current generation of 2-D models, if used in this way, would have even more life left than any of us may have imagined.

So, in summary, I believe that 2-D basin models have a rich (potential) future. However, I believe that this potential can only be realized if the modeling industry adapts to a new way of relating to the research community. Many of the opportunities noted above can only be actioned via a wholesale re-invention of basin modeling technology, and I suspect that this can occur only with vehicles that have already seen their payback. New ideas that are tested in 2-D can become available in next-generation 3-D models, so there is a commercial incentive (albeit a long-term one) for the proposal made here.

Sediments Behaviour Through Geological Times for Basin Modeling

Freacutedeacuteric Schneider

Institut Franccedilais du Peacutetrole, 1 amp 4 avenue de Bois-Preacuteau, 92852 Rueil-Malmaison Cedex-France

Basin modeling aims to reconstruct the accumulation of hydrocarbons at basin scale, and at geological time scale, taking into account the effects of kinematics displacements, sedimentation, erosion, compaction, temperatures history, overpressures and fluids flows (water and hydrocarbons). Furthermore, explorationists wish to address overpressure reconstruction in order to estimate the risks of drilling.

Kinematics at basin scale (faulting and/or folding) generally represents the displacements related to the tectonic stress or to the salt (or mud) tectonics. These phenomena are poorly understood, especially in term of rheology, and they are generally treated by using empirical geometrical rules (simple shear, flexural slip,hellip). The main driving forces during hydrocarbons migration are: the pressure gradient, the gravity, and the capillary pressure gradient. The displacement of the hydrocarbons is then the result of the coupling of the conservation laws (solid, water, and hydrocarbons) with the generalised Darcy's laws and with a compaction law.

Within the frame of basin modeling, this paper deals with recent advances related with compaction and hydrofracturing.

Two types of phenomena contribute to the compaction of sediments: (1) purely mechanical phenomena, mainly caused by the rearrangement of grains during burial, and (2) chemical compaction resulting from dissolution-precipitation mechanisms, generally induced by stress (pressure solution). Mechanical compaction is more efficient near the surface, whereas chemical compaction dominates at depth. The transition zone is probably at a depth of a few hundred meters for carbonate sediments and around 1.5 km for sandstones.

Mechanical compaction is described on a geological time scale by an elastoplastic model in which the elastic modulus and the strain hardening modulus increase when deformation increases. The plastic limit is the maximum vertical effective stress reached by the sediment.

The model also incorporates a viscoplastic term in the compaction equation. This component macroscopically considers viscous compaction phenomena such as pressure solution. The viscosity coefficient is considered to be a function of temperature.

The resulting mathematical formulation is then given by (Schneider et al., 1996):
Equation 2

where phis is the porosity, sigma is the vertical effective stress, T is the temperature, beta is the elasto-plastic coefficient and alpha is the visco-plastic coefficient.

One of the difficulties to quantify the sediment behaviour through geological time is that some of the known phenomena (e.g., pressure solution) are not completely reproducible at laboratory scale. In order to address this problem, the methodology developed for basin modeling consists of coupling a theoretical approach which gives the macroscopical laws with the calibrations of the parameters of the resulting laws with present-day observations. Using this methodology, it has been found that the activation energy is around 15 kJ/mole for chalk and sandstones.

The corresponding viscosities are 8 Gpa.Ma (2.51 023 Pa.s) for chalk and 90 GPa.Ma (2.8 1014 Pa.s) for sandstones (Figure 1).

Figure 1. Jurassic garn sandstones.

Another difficulty is that the parameters are not constant over the geological history of the sediments. Indeed, the diagenetic processes that include pressure solution modify "slowly" the sediment properties such as permeability or mechanical properties (chemical hardening). For example, from data given by Engstrom (1992), it could derive the following relationships between chalk cohesion and porosity : c = 20.9 minus 41.1phis, and between chalk friction angle and porosity: phiv = 56.1 minus 105.6phis.

Once the evolution of the mechanical properties of the sediment have been defined, at a given time (or at a given porosity), the elastoplastic yield surfaces of the sediment can be drawn in the (Q,Pprime) space, where Q is the deviatoric stress and Pprime the mean effective stress. In this formulation, it is supposed that the weigh of the overburden is one of the principal stresses (sigmav); the other principal stresses are supposed to be equals and horizontal (sigmah = sigmaH). The module of the horizontal stress is related to the vertical one through the relation (sigmah = Ko (z) (sigmav) where Ko, (z) is a function of the tectonical setting (Grauls, 1996).

The sediment compaction is also controlled by the permeability of the porous medium. Overpressures (pore pressure greater than the normal hydrostatic pressure) may be generated when permeability is low. In this case, the fluids that saturate the porous medium cannot escape from the sediment. This phenomenon, called compaction disequilibrium, can evolve toward hydraulic fracturing.

More generally, hydraulic fracturing is seen as an accommodation process during which the sediment permeability increases in response to an increase of the pore pressure. The permeability increase begins when one of the hydraulic fracturing criteria (tensile fracturing or shearing) is reached.

In particular, hydraulic fracturing is invoked to predict the behaviour of caprock above overpressured reservoirs, fracturing correlated with in situ generated overpressure in shale, and fracturing induced by the transformation of kerogen into hydrocarbons. For example, in the case that the total stress are constant, hydrofracturing can occur during an increase of the pore pressure when the "unloading" stress path cuts the vertical tensile fracturing criterion (Figure 2).

Figure 2. Example of normal compaction stress path.

Both the experimental and numerical results have shown that the permeability increase during hydraulic fracturing seems to be less important than thought. In the numerical tests performed, the maximum permeability needed to accommodate the increase of pore pressure is two or three times the matrix permeability.

REFERENCES CITED

Engstrom, F., 1992, Rock mechanical properties of Danish North Sea chalk: Fourth North Sea Chalk Symposium, Deauville.

Grauls, D., 1996, Minimum principal stress as a control of overpressures in sedimentary basins. Extended abstract, Compaction and overpressure, 8th conference on exploration and production. IFP, 9–10 December, I'FP report no. 43313.

Schneider, F., J. L. Potdcvin, S. Wolf, and I. Faille, 1996, Mechanical and chemical compaction model for sedimentary basin simulator: Tectonophysics, vol. 263, p. 307–317.

Influence of Changing Rock and Fluid Properties on Bulk Thermal Properties in Basin Simulation

Harald S. Poelchau

Institut fuumlr Chemie und Dynamik der Geosphaumlre, Forschungszentrum Juumllich GmbH, Germany

Basin simulation depends not only on knowledge of changing geometry of layers of the history of the basin but also on appropriate physical property values which are variable with time. Of particular interest and influence are thermal properties which determine the outcome of the reconstruction of the thermal field and its history, which in turn shape the maturation history of the source rocks and the generation of oil and gas. Thermal conductivity is one of the main parameters affecting the temperature simulation, and a chief concern is the method used to compute the bulk sediment parameters as well as the parameters entering into this determination.

For calculating the bulk thermal conductivity one needs to consider the thermal conductivities of the rock components and the fluids of the sediment, their changes with temperature and pressure and diagenesis.

Many different methods and models have been reported in the literature for calculating bulk thermal conductivity from component thermal conductivities. Most of these depend on additional data such as grain size or grain arrangement, which for practical purposes are not available in basin modeling situations. Therefore, the simple geometric mixing formula is normally used as a practical approximation. It works well enough for most situations and is generally implemented in most basin simulation packages. However, this equation gives poor results for partial gas (or air) saturation (Figure 1). This may be a case of two combined models.

Figure 1. Thermal conductivity as function of water saturation of a sandstone with 18% porosity. Square dots show experimental data (with shaded error envelope) from Reibelt (1991) in Clauser and Huenges (1995). For comparison, the curve calculated with the geometric mixing formula is shown.

The influence of temperature on the interstitial fluids is opposite to that on the solid grains or solid matrix. Thermal conductivity of solids tends to decrease with temperature except for those with very low thermal conductivity such as shales. On the other hand, fluid thermal conductivity increases with temperature. Pressure affects fluid thermal conductivity much more strongly at shallow burial, especially for gases (Figure 2). The balance of these effects depends to a large degree on porosity and its decrease with burial.

Figure 2. Thermal conductivity of methane as function of pressure and temperature. Data from Landolt-Boumlrnstein (Toumldheide et al., 1968).

The effects of some of these parameters will be shown for actual case histories with sensitivity analyses detailing the reaction of the temperature field in geological systems to changes of porosity, temperature and pressure and the composition of fluids and solids.

REFERENCES CITED

Clauser, C., and E. Huenges, 1995, Thermal conductivity of rocks and minerals, in T. J. Ahrens, ed., Rock physics and phase relations—A handbook of physical constants, Reference Shelf, v. 3: Washington, Am. Geoph. Union, p. 105–126.

Toumldheide, K., F. Hensel, and E. U. Franck, 1968, Waumlrmeleitfaumlhigkeit von Gasen, in K. Schaumlfer, ed., Transportphaumlnomene II—Kinetik—Homogene Gasgleichgewichte: Landolt-Boumlrnstein Zahlenwerte und Funktionen. II. Band: Eigenschaften der Materie in ihren Aggregatzustaumlnden, v. 5b: Berlin, Springer, p. 39–71.

High Resolution 2-D Diffusion Simulation of Hydrocarbons in Reservoir and Surrounding Sediments: Implications for Gas Losses and for the Interpretation of Reservoir Geochemical Data

Ganjavar Khavari-Khorasani1 and Kjell Oeygard2

1Petrotrak; 2BP Amoco

Calculation of diffusive loss from gas reservoirs has been widely attempted (e.g., Leythaeuser et al., 1982; Krooss et al., 1992a, b; Montel et al, 1993; Nelson and Simmons, 1995). Nelson and Simmons (1995), who calculated the diffusive gas loss though the cap rock of the McClave field, concluded that for a shale cap rock porosity of 5%, the entire volume of methane will be lost in less than 2.3 m.y., and 5.3 m.y. for ethane. Increasing the porosity to 10% they calculated that the total methane loss could occur within 500,000 years, and ethane in 1.2 m.y. The rapid losses estimated by Nelson and Simmons (1995) compared to the earlier calculations (Krooss et al., 1992a, b; Montel et al, 1993) are due to the fact that Nelson and Simmons (1995) fix the boundary condition with respect to gas concentration in water outside the reservoir at a level which imposes an extremely transient state at all times, but treat this as a steady state. It is impossible to maintain a zero methane concentration in water at some point, at the same time that large quantities of methane diffuses past the same point, as in the calculations performed by Nelson and Simmons (1995). That is, the calculations does not account for the time-dependent changes in gas concentration in the surrounding sediments, and hence result in unrealistic high losses, regardless of other considerations.

Most earlier authors have acknowledged the importance of the cap rock petrophysical properties, as well as the time-dependent geological factors on the diffusive flux of gas from reservoirs. However, the effect of reservoir geometry and scale on diffusive gas losses and on the pattern of the diffusive flux has been overlooked. In a recent study it was shown that the geometry and heterogeneity (of lithology) of the reservoir has a major impact on patterns of diffusive flux within a reservoir (Khavari-Khorasani et al., 1998). In this paper we discuss the effect of reservoir geometry, scale and petrophysical properties of the entire sediment package on diffusive losses, and on the pattern of the diffusive flux. While the diffusive flux originates from forces due to the pressure gradient, forces due to the temperature gradient, and forces due to the concentration gradient, in the context of this paper we are isolating the effect of the latter diffusion type (i.e., ordinary diffusion).

DILUTE MULTICOMPONENT, MULTIPHASE, 2-D NUMERICAL SIMULATOR

The simulator calculates the diffusive flux in 2-D within the reservoir and in the surrounding sediments, at very high resolution, while moving the reservoir through the burial geohistory. The geohistory calculations are in 1-D and are based on rock properties and other input parameters from a reference well which intersects the reservoir. The calculated results (e.g., pressure, temperature, porosity, etc.) from the geohistory are used as part of the input parameters to the 2-D diffusion calculations. The diffusive flux outside the computational 2-D domain is handled by a set of 1-D calculations up to the surface and down to the basement (or a given boundary condition). The 1-D solutions are semi-coupled to the 2-D solutions for every time-step. Problem formulation for diffusion calculations was discussed earlier (Khavari-Khorasani et al., 1998). The simulations are performed for C1–C2 hydrocarbons in oil, gas, water and kerogen.

MAIN FINDINGS

The effect of lithology of the sedimentary package, and of thickness and properties of seals

For otherwise identical conditions, the average flux per area is significantly higher for a reservoir located in a sedimentary package rich in sand- and siltstone lithologies, compared to one within a sedimentary package dominated by claystone. The example in Figure 1a shows the cumulative gas loss (m3/m2), over 100 My, for two gas reservoirs ("A" and "B") with identical width (1000 m), identical column heights (350 m), identical petrophysical properties, and identical top and bottom seals thickness and lithology. The only difference is the overall petrophysical properties of the total sedimentary package, where the former is being rich in sandstone and siltstone lithologies, and the latter in claystone and shale lithologies. The total diffusive loss (loss at 0 My b.p.) from the reservoir "A" is around 750 m3/m2, compared to 295 m3/m2 for the reservoir "B" (Figure 1a).

Figure 1. (a) Example of cumulative gas loss (m3/m2) with time for a gas reservoir located in a sedimentary package rich in sandstone and siltstone lithologies (A) and for a sedimentary package rich in claystone and shale (B). Both reservoirs have similar seal thickness and petrophysical properties. (b) Example of cumulative gas loss(m3/m2) with increasing the gas reservoir column height. The loss increases with increasing column height. (c) Example of cumulative gas loss (m3/m2) with increasing the reservoir width-to-column height ratio. The total loss is reduced with increasing the ratio. (d) Example of cumulative gas loss (m3/m2) as a function of reservoir gas column height for a reservoir slope of almost 0 (A) and of 0.06 (B). Note that losses are reduced with increasing reservoir slope.

In addition, the effect of variation in the thickness and petrophysical properties of the seals on the extent of diffusive flux is insignificant, compared with the petrophysical properties of the entire sediment package. That is, as far as the gas loss via diffusion is concerned, the overemphasis on sealing properties, without accounting for the critical importance of the petrophysical properties of the entire sediment package, can result in erroneous conclusions with respect to which factors have the largest influence on the total gas diffusive flux.

The effect of column height and of reservoir width-to-height ratio on the diffusive gas loss

For otherwise identical conditions, the total diffusive loss increases as a function of the reservoir column height. In the example in Figure 1b the only variable is the gas column height. The diffusive flux increases from around 400 m3/m2 to more than 800 m3/m2 with increasing gas column height from 50 m to 500 m. Furthermore, the larger the ratio of the reservoir width to its column height, the lower the total diffusive gas losses (see example in Figure 1c).

The effect of reservoir geometry on the diffusive gas loss

Reservoirs are commonly tilted rocks. In Figure 1d the only difference between the curves A and B is the slope of the reservoir (regional carrier slope = fill to spill). The first ("A") has a 0 slope, and the other ("B") has a slope of 0.06 m/m. The increasing slope has a significant decreasing effect on diffusive gas loss. This is because the higher the slope, the larger the area of the top seal which can act as a lateral diffusion barrier.

The effect of reservoir geometry on the pattern of diffusive flux within and outside the reservoir

The rates of change in concentration of diffusers in water in the sediments surrounding the reservoir display systematic time dependent patterns, which are essentially controlled by the geometry of the reservoir (e.g., Figures 2a). We observe these geometry-dependent patterns regardless of variations in other parameters.

Figure 2. The geometry-dependent pattern of diffusion in water (a) and in oil (b). These geometry-dependent patterns are formed, regardless of other factors, and are persistent over long geological time periods.

England et al. (1987) interpreted lateral compositional variation, in components which do not have a significant effect on fluid density, in many reservoirs, to be related to maturity gradients laterally, and hence that this gradient imaged the initial filling of the reservoirs. However, it was recently demonstrated (Khavari-Khorasani et al., 1998) that such lateral gradients can form effectively via ordinary diffusion, regardless of the initial filling history. Naturally, in a reservoir, other types of diffusion as well as bulk flow will constantly influence the patterns of diffusive flux; nevertheless, it is important to note that the lateral gradients formed via diffusion are very persistent over geological time scales. In the Figure 2b example, the methane concentration gradients is shown in a one-phase oil reservoir. This is formed in less than a few m.y., but persists over 100 My. The lateral concentration gradients of diffuser molecules in an oil reservoir can become large enough to result in an observable GOR variation. In such case, a comparison between the GOR of the fluids from a well which intersect the reservoir close to the crest and that of the fluids from a well which intersect the reservoir close to the spill point (Figure 2b) can be mistaken for gravitational segregation. Furthermore, the lateral compositional gradients with a similar variation to what is expected from maturity can simply represent the effect of ordinary diffusion. Hence, in the interpretation of reservoir molecular parameters and of reservoir fluid data, the effect of the reservoir geometry should be factored in.

Finally, it is important to emphasize that diffusion in reservoirs is at least a 2-D problem. Cross diffusion which can be accounted for in 2-D and 3-D simulations generally reduces the vertical flux, relative to predictions from 1-D evaluations.

REFERENCES CITED

Khavari-Khorasani, G., J. K. Michelsen, and J. Dolson, 1998, The factors controlling the abundance and migration of heavy vs. light oils, as constrained by data from the Gulf of Suez. Part II: Organic Geochemistry, v. 29, p. 283–300.

Krooss, B. M., D. Leythaeuser, and R. G. Schaefer, 1992a, The quantification of diffusive hydrocarbon losses through cap rocks of natural gas reservoirs—A reevaluation: AAPG Bulletin, v. 76, no. 3, p. 403–406.

Krooss, B. M., D. Leythaeuser, and R. G. Schaefer, 1992b, The quantification of diffusive hydrocarbon losses through cap rocks of natural gas reservoirs—A reevaluation: Reply: AAPG Bulletin, v. 76, no. 11, p. 1842–1846.

Montel, F., G. Caillet, A. Pucheu, and J. P. Caltagirone, 1993, Diffusion model for predicting reservoir gas losses: Marine and Petroleum Geology, v. 10, p. 51–57.

Nelson, J. S., and E. C. Simmons, 1995, Diffusion of methane and ethane through the reservoir cap rock: Implications for the timing and duration of catagenesis: AAPG Bulletin, v. 79, no. 7, p. 1064–1074.

Ductile Shear Deformation in Fault Zones: Implications for Cross-Formation Fluid Flow

Jeffrey A. Nunn

Louisiana State University, Baton Rouge, LA

Recent evidence suggests that sediments underlying growth faults, salt sheets and salt welds in the Gulf of Mexico are tectonically disturbed: overturned and repeated sections defined by paleontological assemblages, fluid pressures greater than 90% of lithostatic which diminish with depth to a more regional pore pressure gradient, unusual ductile sediment rheology, and highly pore-pressure-dependent permeabilities. Oil shows in the adjacent salt and fractures within the sediments suggest that these disturbed zones act as episodic conduits for fluid flow.

For example, the A-20ST Pathfinder well was drilled 2.44 km through a growth fault zone that forms the northern boundary of the Eugene Island 330 minibasin, offshore Louisiana, to determine in-situ conditions within and surrounding the fault zone. Measured pore fluid pressures reach approximately 93% of lithostatic pressure below 2 km depth. Below the growth fault, core plugs have a measured porosity of 30%, indicating undercompaction. Effective stress near the fault is low (10 MPa or less) with the lowest values in the footwall. A plot of effective stress vs. porosity indicates that compaction disequilibrium accounts for most of the overpressures. Sediments have an unusual rheology. Possion's ratio is above 0.4 and shear modulus is less than 1 GPa. Poisson's ratio for most rocks is 0.25 to 0.3. Poisson's ratio for a liquid is 0.5. The shear modulus for shale and sandstone is 15 plusmn 5 GPa. Core examination plus a downhole Formation MicroImage log showed numerous faults and fractures. Many of the faults and fractures contain oils of similar chemistry to oils being produced from reservoirs directly above the fault zone. Formation waters from these Pleistocene strata have been isotopically dated as Oligocene or older. Drill stem tests for a perforated and fracture-packed interval around the growth fault indicate that the permeability of geopressured sediments varied by more than 3 orders of magnitude from 100 md to less than 0.1 md as a function of pore fluid pressure.

Similar observations of highly overpressured (gt90% of lithostatic) sediments, pore-pressure-dependent permeabilities, unusual sediment rheology, and episodic fluid flow have been documented in shear zones in other tectonic environments including the sub-salt play in the Gulf of Mexico, deacutecollement zone sediments in the Barbados and Oregon/Cascadia accretionary wedges, and the San Andreas Fault.

I suggest that the development of highly overpressured sediments within the fault zone in Eugene Island is related to reduction in porosity and permeability associated with ductile shear deformation. During shearing as the fault zone moves in response to salt evacuation, intergranular pressure solution, grain rotation, and fracture sealing reduce the volume and increase the aspect ratio of pores. These processes tend to reduce the permeability and increase the pore fluid pressure within the shear zone. Thus, during periods of ductile deformation, the shear zone is an effective seal against cross-formational fluid flow not only because it has low permeability but also because it has higher pore fluid pressures than adjacent reservoir sands. As ductile deformation continues, pore fluid pressures may eventually exceed the fracture strength of the sediments and a short period of brittle deformation occurs. Hydrofracture within the shear zone enhances its porosity and permeability. Subsequent fluid flow along the fault causes pore pressures to drop and fluid from adjacent reservoir sands to be drawn upwards along the fault zone. Rapid decline in pore fluid pressures causes the fracture network to collapse and the shear zone to reseal. Cycles of ductile and brittle deformation may be repeated numerous times as shear movement continues. Once shear deformation stops, elevated pore fluid pressures within the shear zone will diffuse down to the regional level. However, low permeability will remain. Thus, shear zones such as growth faults or salt welds may be long-term barriers to fluid flow.

Using hydrostratigraphic information constructed from seismic and well data, I have simulated episodic expulsion of fluids from the geopressured zone along faults into individual sand layers in the overlying hydropressured zone of EI 330. The results of this study produced a number of important conclusions regarding fluid expulsion along growth faults. (1) Fluid expulsion must be episodic. The fluid pressure drawdown of sim10 MPa near the fault should cause fracture/pore collapse and consequent decrease in permeability in the undercompacted, compressible sediments in and near the fault zone. (2) Duration of expulsion events and the amount of fluid transported is also dependent on the permeability of sediments adjacent to the fault which determines their ability to transport fluids towards/away from the fault zone. (3) The amount of fluid moving from the fault zone into an adjacent sand depends on duration of expulsion, stratigraphic position, sand thickness and permeability, dip of sand layer, and fluid density and fluid pressure in the sand prior to the expulsion event. (4) Large quantities of fluid move from geopressured sediments into overlying sands during an expulsion event (sim100 million barrels per horizontal square meter).

Adsorptive Capacity of Oil in Source Rock—Implications from Experiments for Modeling Oil Expulsion from Source Rock

Akihiko Okui1 and Atsushi Moriya2

1Technology Research Center, Japan National Oil Corporation, 2-2, Hamada 1-Chome Mihama-ku, Chiba-shi, 261, Japan; 2Technology RampD Center, Toyo Engineering Corporation, 1818 Togo, Mobara-shi, Chiba 297-0017, Japan

Oil expulsion from source rock is one of the important processes to have oil and gas fields in basins. Although a lot of research and discussions has been done, the mechanism of oil expulsion cannot be fully understood. It was proposed that the adsorption of oil on and in kerogen is one of the major mechanisms to control oil expulsion. Even though geochemical analysis suggests 50–200 mgHC/gTOC required to expel oil from source rock, the nature of the adsorption phenomena cannot be fully revealed. Therefore, we experimentally studied the adsorptive capacity of hydrocarbons on and in coals under various physical conditions.

A series of liquid-phase adsorption experiments was conducted, utilizing several types of coals crushed in 200–400 mesh with adsorbent solution (Table 1). We measured the concentration of each solute at a constant interval by gas chromatography (detected at FID). When the concentration became almost constant, which is regarded as an equilibrium state, the amount of adsorbed solutes is estimated from the difference between the final concentration and the initial one. It was found that the amount of adsorbed solutes varies according to the type of hydrocarbon; phenol gt aromatic gt naphthene gt paraffin. However, it was also affected by the type of solvent. The influence of coal type was almost negligible in these experiments.

When we interpreted the experimental results, we applied the method without using fitting parameter to generalize the model. Therefore, we did not use usual adsorption equations such as Freundlich's equation or Langmuir's equation. At first, we applied Polanyi's adsorptive potential theory, but it could not explain the experimental results completely. Then we introduced Hildebrand's solution theory and combined it with Polanyi's theory. It was found that no fitting parameter is necessary for phenol except physical properties such as molecular volume and solubility parameter (equation 1, Figure 1). For other hydrocarbons, we were obliged to introduce experimental parameter alpha to correct the influence of solvents now (Figure 2).

Figure 1. Relationship between amount of adsorption and adsorptive potential for phenol.

Figure 2. Relationship between amount of adsorption and adsorption for phenanthrene. Parameter a is estimated 1500 [kcal/mol] for n-Hexane and Benzene, and minus1200 [kcal/mol] for Ethanol, respectively.

The amount of adsorption is expressed as a function of adsorptive potential epsilon (equation 2), and the maximum amount can be estimated by assuming epsilon = 0 in equation 2. 8 cm3/100g d.a.f. coal was derived for phenol, which is equivalent to 100 mgHC/gTOC assuming oil density as 1.0 g/cm3. This value is in the range suggested by the geochemical analysis on source rocks. The maximum amount of adsorption for each type of hydrocarbon will be derived for various kinds (different maturities) of coal, which can calculate the adsorptive capacity of coal as a function of maturity.
Equation 3
Equation 4

A: amount of absorption [cm3/100g d.a.f. coal], R: gas constant [kcal/mol/deg], T: temperature [K], V: molecular volume of solvent [cm 3/mol], X: concentration [mole fraction], alpha: experimental parameter [kcal/mol], delta: slubility parameter [(kcal/cm3)1/2]; epsilon: adsorption potential [kcal/mol].

Rapid Assessment of Mudstone Compaction and Poroperm Parameters from Wireline Logs

Andrew C. Aplin, Yunlai Yang, and Steve Larter

NRG, University of Newcastle, UK

Fluid flow and pressure development in sedimentary basins are largely controlled by low permeability mudstones, which comprise around 70% of basin fill. Accurate modeling of flow and pressure, both today and through geological time, therefore rests substantially on an accurate description of the way in which mudstones lose porosity and permeability with increasing burial or effective stress. Two central equations are those describing the relationships between (1) void ratio (or porosity) and effective stress and (2) porosity and permeability. Current practices for estimating these relationships include use of defaults (unacceptable) and fitting of pressure data from the occasional sands in the basin. The latter approach assumes that pressure variations are related specifically to poroperm variations and is limited in that it poorly constrains the evolution of fluid flow and pressure systems.

We have been trying to establish ways of determining the compaction and poroperm parameters from rock properties which can be either easily measured or estimated from wireline log data. We have been specifically interested in grain size since its influence on compaction coefficients is well established in soil mechanics, as is its influence on the permeability of sands or beadpacks. Having established the importance of grain size on these properties, we have used Artificial Neural Networks (ANNs) to rapidly and pragmatically obtain grain size data from wireline logs.

A convenient and reasonably robust measure of mudstone grain size is clay fraction, the percentage of lt2 micron diameter particles. Reworking of data published by Skempton (1944) and Burland (1990), plus our own new data, supports the idea that the compression coefficient (beta; the coefficient describing the porosity–effective stress relationship) of mechanically compacted, natural muds is a strong function of clay fraction (Figure 1).

Figure 1. Beta describes the relationship between effective stress and void ratio and is strongly related to clay fraction. The relationship has been estimated, with similar results, by direct experiment and via the relationship beween eL (void ratio at liquid limit) and clay fraction.

Establishing the most appropriate porosity-permeability functions is hindered by the almost complete lack of permeability data for well characterised muds. We have developed a model which determines permeability from pore size distribution, clay fraction and porosity data (Yang and Aplin, 1998). Calibration against measured permeabilities suggests that the model predicts permeabilities to plusmn3 of the true value. We have a significant database of modeled permeabilities of mudstones for which we also have clay fraction data. Some of the data are shown in Figure 2 and indicate the influence that clay fraction exerts on the permeability-porosity relationship.

Figure 2. Modeled permeability vs. porosity of mudstones, divided according to clay fraction. At a given porosity, finer grained mudstones are less permeable than siltier mudstones.

Since clay fraction can be used to assess the coefficients in poroperm and porosity–effective stress functions, can clay fraction be rapidly assessed from wireline data? Using data from gt10 North Sea and Gulf Coast wells, we have trained ANNs to predict clay fraction and grain density from a standard suite of wireline data. Since we are interested in the solid matrix of the muds, log data other than gamma and caliper (resistivity, density, neutron and sonic) were manipulated in order to minimise the influence of porespace on the wireline signature. Networks trained and tested well, predicting clay fractions to within plusmn6% of the measured value and grain densities to within plusmn0.05 g/cmminus3 of the true value (95% confidence; Figures 3 and 4). Although the ANNs require testing with a broader suite of geological, sedimentological and mineralogical data, these encouraging results raise the possibility of rapidly estimating mudstone:

Figure 3. Density log, sonic log, ANN-predicted and actual grain density, ANN-predicted and actual clay content: training well.

Figure 4. Density log, sonic log, ANN-predicted and actual grain density, ANN-predicted and actual clay content: test well.

  • Porosity (from density log and estimated grain density)

  • Permeability (from clay fraction and porosity data)

  • Compaction Coefficient (from clay fraction data).

The evaluation of clay fraction and porosity from wireline logs therefore enables us to:

  1. More accurately correlate mudstone sequences

  2. Decompact (backstrip) mudstone sequences with greater confidence

  3. Estimate overpressure from wireline logs, using porosity data and the appropriate porosity–effective stress relationship for the given lithology

  4. Obtain permeability and compaction coefficients of fine grained sediments for basin modeling.

REFERENCES CITED

Burland, J. B., 1990, On the compressibility and shear strength of natural clays: Geacuteotechnique v. 40, p. 329–378.

Skempton, A. W., 1944. Notes on the compressibility of clay: Quarterly Journal of the Geological Society of London, v. 100, p. 119–135.

Yang, Y., and A. C. Aplin, 1998, Influence of lithology and compaction on the pore size distribution and modeled permeability of some mudstones from the Norwegian Margin: Marine and Petroleum Geology, v. 15, p. 163–175.

Experimental Assessment of Fluid Transport Parameters of Pelitic Rocks and Their Application in Basin Modeling

B. M. Krooss, S. Schloumlmer, and M. Burkhardt

Institute of Petroleum and Organic Geochemistry (ICG-4), Forschungszentrum Juumllich GmbH, D-52425 Juumllich, Germany

Fine-grained (pelitic) sedimentary rocks play a prominent role in sedimentary basins due to their abundance and their impact on fluid transport processes on different length scales. Considering the importance and variety of this lithologic species the knowledge and information on petrophysical parameters and in particular fluid transport properties of pelitic rocks and the dependence of these properties on lithotype, compaction and diagenesis are still rather limited.

In petroleum geology/geochemistry three main topical fields can presently be identified where the assessment of mudrock properties is of particular technical and economic importance.

(i) The prediction and quantification of compaction-induced overpressure in thick shale sequences represents an important technical and safety issue and one of the major challenges for fluid flow modeling on the basin scale. (ii) Assessment and quantification of gas/oil sealing efficiency of pelitic top and fault seals is one of the cornerstones of prospect appraisal in petroleum exploration. (iii) Primary migration of petroleum (oil/gas) in and its expulsion from clastic source rocks are affected by mudrock lithology and the deterministic description of these two processes requires information on the mechanical and fluid transport properties and associated fractionation effects.

During recent years various studies have been performed by our group touching upon one or several of the topics mentioned above. The results of these investigations are presently being compiled into a mudrock database comprising lithological, mineralogical, geochemical and petrophysical information with emphasis on fluid transport properties.

The following passages summarize the main results and the present status of the ongoing research on mudrock properties.

PRESSURE-DRIVEN VOLUME FLOW (DARCY FLOW)

The efficiency of pressure-driven volume flow through the pore space of sedimentary rocks is expressed by the permeability coefficient of Darcy's law. Permeability coefficients (k) have been measured as a function of effective stress (sigma) for selected samples. The semi-logarithmic plot in Figure 1 shows two distinct regimes of the k vs. sigma relationships evident by different slopes of the regression lines. In the measurements performed perpendicular to the bedding the transition between these regimes occurs around 35 MPa. A similar effect was observed for permeability parallel to bedding where the transition occurs at lower effective stress values (lt 20 MPa). The results are basically in line with observations reported by Katsube and Coyner (1994). They indicate the importance of conducting permeability tests on mudrocks under controlled effective stress and reporting the stress conditions with the permeability data.

Figure 1. Permeability (perpendicular to bedding) vs. effective stress for selected mudrock samples.

CAPILLARY GAS SEALING EFFICIENCY

The capillary sealing efficiency of mudrocks is controlled by the pore size distribution and the geometry of the pore network. Omnidirectional mercury intrusion, the standard method for characterization of porous rocks, provides only incomplete and ambiguous information on the capillary sealing efficiency of shales. Therefore direct, unidirectional gas breakthrough measurements have been performed measuring at the same time displacement pressure and two-phase permeability as a function of pressure gradient. Gas/water capillary pressures measured with this method ranged mostly between 1 and 4 MPa while maximum and minimum values were 17 MPa and 0.3 MPa, respectively. Gas permeability values measured after gas breakthrough in the two-phase (water-gas) system varied between 10minus23 and 6middot10minus22 m2 (0.01 and 0.6 nDarcy) and showed a strong dependence on pressure gradients.

MOLECULAR DIFFUSION

Molecular diffusion of natural gas components (methane, ethane, N2) in mudrocks was investigated in various studies with respect to the transport efficiency of this process and fractionation processes associated with it. Experiments were carried out under controlled effective stress and a significant stress dependence of the effective diffusion coefficients and the steady-state diffusive flux was observed. TOC content was found to affect the molecular diffusion of methane and ethane to a larger extent than any other petrophysical parameter. Molecular nitrogen exhibits a higher molecular mobility (higher effective diffusion coefficients) in water-saturated sedimentary rocks than methane. However, due to the higher aqueous solubility of methane, the molar diffusive flux of this gas can, under otherwise equal conditions, be higher than that of nitrogen.

Isotopic analysis of diffused methane and ethane showed an enrichment in 12C components (Deltadelta13C = 6 minus 7 permil for methane, 3–4 permil for ethane; Deltadelta13C = delta13Csource gasminusdelta13Cdiffused gas) in the first fractions collected during non-steady state. At steady state the Deltadelta13C values decrease to 2–3permil for methane and to 1–2permil for ethane. The effect of TOC content on the isotopic fractionation during diffusion remains to be examined.

PETROPHYSICAL PROPERTIES AND FLUID TRANSPORT PARAMETERS

The petrophysical characterization of mudrock samples used in fluid flow experiments comprised mercury porosimetry, specific surface area and grain density measurements. Generally, no simple correlation exists between these petrophysical parameters and the permeability coefficients. The Kozeny-Carman equation appears to be of limited applicability for low-porosity mudrocks. More sophisticated fluid transport models, under development by several groups, rely on an extended experimental database for calibration and testing.

The response of resistivity and sonic logs has been reported to react on overpressure in shales (Hermanrud et al., 1998) which, in turn, is related to fluid transport parameters. In an attempt to establish a direct link to logging tools, sonic velocity measurements have been conducted on samples used in permeability tests. Figure 2 shows correlations between permeability and sonic velocity of oriented mudrock plugs based on results of a recent study.

Figure 2. Correlations of permeability and sonic velocity (water-saturated samples) for a mudrock sequence.

REFERENCES CITED

Hermanrud, C., L. Wensaas, G. M. G. Teige, N. Bolarings, S. Hansen, and E. Vik, 1998, Shale porosities from well logs on Haltenbanken (offshore mid-Norway) show no influence of overpressuring, in B. E. Law, G. F. Ulmishek, and V. I. Slavin, eds., Abnormal pressure in hydrocarbon environments: AAPG Memoir 70, p. 65–85.

Katsube, T. J., and K. Coyner, 1994, Determination of permeability(k)-compaction relationship from interpretation of k-stress data for shales from Eastern and Northern Canada: Geological Survey of Canada, Part E, Current Research Paper 94-1D, p. 169–177.

Thermal Conductivity As an Input Parameter to Basin Modeling

Kerry Gallagher

T.H. Huxley School of Environment, Earth Science and Engineering, Imperial College of Science, Technology and Medicine, London SW7 2AS, England

In order to model maturation and hydrocarbon generation in an evolving sedimentary basin, we need to specify a variety of physical parameters. Of these thermal conductivity and heat flow are two primary inputs. Heat flow is important as it is intimately linked to the tectonic processes driving basin subsidence and in principle allows us to explicitly build some geological reality into our basin modeling. Thermal conductivity is required to calculate the temperature as a function of depth, given the heat flow. In the simplest case of constant thermal conductivity, then the ratio of heat flow (Q) and thermal conductivity (k) defines the temperature gradient in the sedimentary column, through Fourier's law, i.e.,
Equation 5

There are some of the problems related to inferring thermal conductivities in basin modeling. For example, it is well known that the bulk thermal conductivity of rock varies with lithology (constituent minerals which determine the matrix thermal conductivity), porosity and the nature of the pore-filling fluid (determines the fluid thermal conductivity), effective stress and temperature, as well as often being anisotropic (e.g. Somerton, 1982). It is straightforward measure thermal conductivities from small (i.e. several cm3) rock samples using conventional divided bar or needle probe techniques. These experiments can be made for different pore fluids and under different temperature conditions. The problem arises how to extrapolate a limited suite of thermal conductivity data to the basin scale required for modeling. Furthermore, in order to predict thermal conductivity reliably, a suitable model is required linking porosity and matrix thermal conductivity to the bulk thermal conductivity. There is a wide variety of such models (e.g. Gallagher, 1987, Somerton, 1982) that represent a range from purely empirical to more physically based.

So an outstanding question in basin modeling is how to represent large scale thermal conductivity structure to solve the thermal history modeling problem. Typically we may have a few isolated thermal conductivity measurements and some idea of how porosity and lithology changes as a function of depth from logs and cuttings. We then estimate the variation in thermal conductivity downhole based on these limited data and a variety of assumptions.

Let us consider now what we actually have to constrain heat flow. Usually this comes down to some variable quality downhole temperature data combined with the inferred thermal conductivity. This lets us estimate the present day heat flow, again typically using some form of Fourier's law. But what about heat flow in the past? We can assume some tectonic model which allows us to predict the heat flow variation as a function of time. If this agrees with the observed subsidence and structural history in the basin, there is often no reason to reject such a model. However, this is not so common, and typically we need to incorporate some thermal calibration data, such as vitrinite reflectance or apatite fission track analysis. These data effectively represent some form of thermally activated chemical reaction or physical transformation and therefore provide a record of the thermal history in the basin. The heat flow/thermal model needs to predict this data adequately. The calibration data used for thermal modeling is sensitive to the temperature history. Temperature at a given depth can be determined from Fourier's law (which implicitly assumes 1-D steady state heat transfer with no advection). In this case the thermal conductivity is some average of the sediments between the surface (z = 0) and the depth of interest. However, if the thermal conductivity is wrong, then at first inspection we might expect that the temperature gradient and so our temperature estimate will be wrong also. However, the temperature gradient is the ratio of the heat flow to the thermal conductivity (as shown in equation 1), and in principle we can compensate for the error in the thermal conductivity by allowing the heat to flow sympathetically (i.e. if the heat flow is twice as high as it should be then by making the heat flow twice as high we keep the same temperature gradient). Therefore, if we can allow the heat flow to vary in order to fit the calibration data, it may not matter what we assume or infer for the thermal conductivity structure. As we should predict essentially the same temperatures then we will still fit the calibration data and so the model is acceptable.

This notion has been investigated by adopting a formal inversion scheme, based on Gallagher (1998). The philosophy of this approach is to find the simplest heat flow history that can adequately predict the calibration data. The corollary of this is that although more complex heat flow models may fit the data, the extra complexity is not warranted by the data constraints. Therefore, additional information is required to justify it. We have used a suite of vitrinite data from the well Inigok 1 in the Colville Basin, Alaska. This basin also has a series of thermal conductivity measurements on a range of different lithologies and formations. These were used to estimate the present day heat flow and provide internally consistent estimates for the matrix thermal conductivities of the major lithologies. The inversion scheme was run using the best estimates for all parameters and the reconstructed heat flow and data fit are shown in Figure 1.

Figure 1. Predicted and observed vitrinite reflectance data from Inigok 1, and the inversion generated heat flow history.

Subsequently, the inversion scheme was run 100 times, using Monte Carlo sampling of the distributions of the individual lithologies' thermal conductivities. Figure 2 shows the spread of the estimated heat flow histories, normalised to the present day heat flow value (calculated for each assumed thermal conductivity structure). As can be seen from this figure the general form is very similar to that obtained with the best estimates of the thermal conductivity and the spread in the solutions is generally small. This agreement occurs because although the heat flow is variable, more or less the same thermal history is required to fit the data. Therefore the heat flow compensates for the variation in thermal conductivity to achieve a similar thermal history and so a similar fit to the vitrinite data.

Figure 2. The average heat flow (plusmn1sigma) normalised by the present day value for the 100 simulations varying the thermal conductivity structure.

These results suggest that thermal history modeling is robust to inconsistencies in the thermal conductivity data. However, some case must be exercised when interpreting the results in terms of geodynamic models of basin formation as the absolute heat flow values may be misleading or even meaningless. However, as we are generally interested in the thermal history of the basin and the maturation of particular source rocks, it may be that the geodynamic relevance is secondary.

REFERENCES CITED

Gallagher, K., 1987, Thermal conductivity of sedimentary and basement rocks from the Eromanga and Cooper Basins, South Australia: Exploration Geophysics, v. 18, p. 381–391.

Gallagher, K., 1998, Inverse thermal history modeling as a hydrocarbon exploration tool: Inverse Problems, v. 14, p. 479–497.

Somerton, W. H., 1992, Thermal properties and temperature-related behaviour of fluid rock systems: Developments in Petroleum Science 37, Elsevier, p. 257.

Integration of Capillary Top Seal Leakage with Map Based Fluid Flow Modeling

G. Eric Michael and Peter D'Onfro

Conoco Inc., P.O. Box 2197, Houston, TX, 77079

In map based fluid flow modeling (pseudo 3-D, e.g., SEMI), vertical leakage occurs through hydraulic fractures or by capillary top seal failure related to capillary displacement pressure. Using empirical data sets, we have devised a method of estimation of capillary displacement pressure of top seals. This method is dependent on top seal lithology and porosity and estimations are given in a range between minimum and maximum and most likely (e.g. P50) scenarios. Recently this method has been incorporated into the SEMI (SINTEF) map based modeling program, providing the capability to model change in top seal capacity with depth (porosity). This enables integrating the timing of hydrocarbon charge and phase with top seal capacity evolution through time. In this presentation we show how the methodology is particularly useful in modeling fill and spill scenarios, but has potential to explain hydrocarbon phase in other charge scenarios.

Tertiary reservoir charge examples, using the SEMI program, of the top seal application are shown for a predominantly lateral migration charge scenario (South Viking Graben, Norway) and a vertical charge scenario (deep-water Gulf of Mexico). Fill and spill in the Tertiary sands of the South Viking Graben have been previously recognized (Barnard and Bastow, 1991) based on geochemistry and hydrocarbon phase. Previous work has concluded predominantly gas-prone charge for the South Viking Graben central basin areas and more oil prone for graben flanks. This can be explained on the basis of additional gas flushing of reservoirs near high maturity basin center gas kitchens present-day. However, analysis of top seal with respect to charge time reveals interesting insight that can explain phase of observed fields and more recent fields in basin flank areas where Jurassic source rocks are only marginally mature.

In relatively shallow reservoirs (e.g., two phase petroleum at reservoir conditions) such as the Tertiary sands of the South Viking Graben, early top seal development retains gas and spills oil, versus late top seal development which leaks gas and retains oil. Areas where the Tertiary section thins and/or pinches out would be areas of late top seal development and more likely to retain oil but leak gas. The retention of gas and spilling of an oil leg is even more enhanced by low relief structures (most less lt200 m) as observed in the South Viking Graben Tertiary fields (Figure 1).

Figure 1. Capillary top seal capacity converted to gas column height through time for three fields in the Norwegian S. Viking Graben. Based on this seal capacity model, the map based fluid flow modeling (SEMI) is adjusted appropriately to spill oil at Frigg and possibly Heimdal early and leak gas, retain oil until 5 mya at Balder Field.

An example of top seal development influence on hydrocarbon phase, South Viking Graben, is shown in Table 1.

Within the deep-water Gulf of Mexico, charge of Tertiary sands is often a combination of vertical and lateral fluid flow, but undoubtedly requires predominantly vertical charge from Jurassic and Lower Cretaceous sources. At least for Tertiary reservoirs not subjected to natural hydraulic fracturing due to high sedimentation rates, there appears to be a reasonable correlation between hydrocarbon phase type and quality (e.g., GOR, sulfur content) with integration of time of top seal competency and hydrocarbon charge time.

Due to uncertainties in depth conversion and/or depth of source rocks in the deep-water Gulf of Mexico, unequivocal evidence of the top seal control on phase is not possible. However, empirical evidence from study of Antioch (Garden Banks 216), Troika (Green Canyon 200) and Ursa (Mississippi Canyon 809) fields in different areas of the deep water suggest that charge and top seal timing relationships control hydrocarbon phase. Modeling of these fields suggests that primary products are a function of the last main hydrocarbon charge pulse after 80–90% of present-day top seal capacity is attained. Early oil window products, often lower gravity and higher sulfur, may be lost from the system if top seal formation has not developed adequate sealing capacity, particularly if structures are high relief (e.g., Troika). The remaining late oil window and gas stage products of lower sulfur and higher API gravity are retained as top seal capacity increases. The interaction between charge and top seal formation is particularly sensitive to the high sedimentation rates in the Gulf of Mexico (Figure 2).

Figure 2. Schematic representation of relation of charge and top seal capacity through time for fields in the Gulf of Mexico. The diagrams demonstrate two end member cases regardless of structure amplitude. It should be noted that structure amplitude will influence hydrocarbon phase as high relief structures, even with early top seal capacity formation, may leak early charge.

By similar analysis to the above examples, understanding top seal formation with respect to charge suggests it is important in areas of mixed oil and gas hydrocarbon charge. Experience in Southeast Asia where mixed terrestrial and lacustrine oil and gas prone shales and coals are present has shown that leaky top seals can be favorable for formation of oil accumulations. An example is the shallow (sim4700 ft), low relief, graben flexural margin, Belida field (gt400 MMBO) in the South China Sea. Top seal analysis suggests that significant gas column height cannot be sustained in this field, although the correlated synrift source rock section is highly mixed oil and gas prone. Geochemical evidence from downdip stained reservoir (Alu Alu E2) suggests that the extent of the Belida field has been larger in the past and a considerable amount of gas is contained in shallower sands above the main oil pay horizons.

Faults and Oil Migration in the Halten Vest High Pressure and Haltenbanken Normal Pressure Regimes, Offshore Norway

Dag A. Karlsen1, Kristian Backer-Owe1, Knut Bjoslashrlykke1, Rainer G. Schaefer2, Abid Bhullar1, Kristian Angard3,9, Inga Steinhoff4,10, Richard Olstad5,11, Kristin Dale6,12, Eirik Vik7, and Bente Nyland8

1Petroleum Geochemistry Program, Department of Geology P.O.Box 1047, Blindern, N-0316 Oslo 3, Norway; 2Institute of Petroleum and Organic Geochemistry, Forschungszentrum Juumllich D-52425 Juumllich, Germany; 3Petroleum Geochemistry Program, Department of Geology; 4Petroleum Geochemistry Program, Department of Geology. Norway; 5Petroleum Geochemistry Program, Department of Geology; 6Petroleum Geochemistry Program, Department of Geology; 7Statoil Forskningssenter Trondheim, N-7005 Trondheim, Norway; 8Norwegian Petroleum Directorate (NPD), P.O. Box 168, N-4001 Stavanger, Norway; 9Now at Norsk Hydro, P.O. Box 200 N-1321 Stabekk, Norway; 10Now at Statoil, Stavanger N-4035 Stavanger, Norway; 11Now at Exxon Exploration Company, 233 Benmar, Houston, TX 77210-4778, U.S.A.; 12Now at Smedvik P.O. Box 165 Skoslashyen N-0212 Oslo, Norway

The occurrence and geochemistry of fluid inclusions in authigenic mineral cements in addition to chemical characterization of bitumen and oil (Figure 1) were used to provide a model for hydrocarbon sourcing and migration in the Smoslashrbukk and Halten Vest region.

We have found evidences to suggest that the main fault zone west of Smoslashrbukk, which today forms a pressure seal, was open for petroleum migration from the west at least as early as 50 m.y.b.p. (Figure 2).

Figure 1.

Figure 2.

Three out of four structures in the so-called Halten Vest High Pressure Dry-Hole Region west of Smoslashrbukk were found to have contained petroleum of the same organic facies as found today in Smoslashrbukk, and these structures filled with oil from the Upper Jurassic aged Spekk Formation at least as early as did the Smoslashrbukk Field. Portions of this oil are today occurring in petroleum inclusions and in mineral cements (Figure 3) and in reservoir core extracts representing residual saturation. The burial history for Smoslashrbukk (Figure 4) is used together with the known temperatures for formation of inclusions in quartz to infer the time when the inclusions could have formed in Smoslashrbukk, and by analogy in the Halten Vest High Pressure Dry-Hole Region. The derived time is between 70 to 50 m.y.b.p.

Figure 3.

Figure 4.

The presently dry structures in the Halten Vest region experienced cap-rock leakage upon receiving petroleum from the Spekk Formation and the subsequent genetically associated development of too high petroleum columns was evidently the cause of leakage as the region experienced a buildup in pressure. The pressure buildup must have occurred following the time at which the fault zone west of Smoslashrbukk transformed from open to sealing. The same abundance of fluorescent petroleum inclusions in the Halten Vest dry wells as in Smoslashrbukk implies according to our model that the time during which the presently dry Halten Vest structures actually held oil is roughly similar to the time petroleum has been present in Smoslashrbukk.

This implies that cap rock leakage in Halten Vest must have occurred, in geological terms, very recently. Petroleum migration from the vest through the presently sealing Smoslashrbukk main fault zone should according to the homogenisation temperatures measured on the inclusions have occurred as early as 70–50 m.y.b.p. Subsequent fault movement and diagenesis on the fault plane led to a situation in which the Halten Vest region was effectively sealed off from pressure let-off in a proximal direction towards the east.

This led to seal failure in the Halten Vest region whilst in the normally pressured Halten region pressure bleed-off towards the east continued, preserving oil in the structures.

We have documented the close genetic relationship between the hydrocarbons in all the above mentioned wells and fields and ascribe the minor variations that exists, not to other source rocks than the Spekk, but rather to east to west lateral systematic variations in terms of proximal to distal source rock organo-facies development of the Spekk Formation source rock. Hence, our geochemical data from the Smoslashrbukk giant oil field, the Smoslashrbukk Soslashr and the Heidrun oil fields all point to the same marine source rock, but with local facies variation. It is this pattern of lateral migration distances corresponding roughly to the inter-field distances of about 20 km which allows us to rationalize the hydrocarbon migration patterns in the Haltenbanken region. More extensive migration is only likely for the Draugen and the Midgard Fields.

We exclude, in contrast to earlier published models by other workers, any quantitatively important contributions of light hydrocarbons (C5-C8) and heavier hydrocarbons (C15+) from the coal bearing Aringre Formation to the presently reservoired hydrocarbon charges in these fields. This suggest the oil generative capability of the Spekk Formation in deep basin depressions to be higher than often indicated from published models and we propose that most condensates in the Haltenbanken region, apart from the 6406/7-1 condensate, are evaporative condensates owing their existence to phase separation rather than to thigh maturity or an inferior kerogen quality.

Traps in the western Haltenbanken region like 6407/4-1, Trestakk and Smoslashrbukk contain mainly petroleum of high maturity, yet in the compartmentalized Smoslashrbukk field some medium maturity bitumen remains in particular reservoir sections. This is most likely due to strong compartmentalization caused primarily by the lithological heterogeneities and secondarily by diagenetic poroperm modifications.

Hydrocarbon contributions from the Aringre Formation in particular in the light hydrocarbon and gas range may have been present in Haltenbanken traps at an earlier stage. The general lack of medium to low maturity petroleum in drill stem tests, which represent the most mobile hydrocarbon phase in a reservoir, in the deeper western parts of Haltenbanken is interpreted to be a further testimony of the very dynamic nature of petroleum entrapment and petroleum displacement and possibly even leakage from structures in this area.

In conclusion we have documented: a) Oil to have been present in structures in the Halten Vest High Pressure-Pry Hole Region; b) Oil in petroleum inclusions and residual saturation in the dry wells in Halten Vest to be of the same organic facies and maturity as that found in inclusions and certain core extracts the Smoslashrbukk field; c) Migration of the earliest phase of petroleum into the Smoslashrbukk Field to have occurred from the west-southwest, through the presently sealing fault-zone, as the current drainage area of the Smoslashrbukk Field was not mature at the time of formation of the first petroleum inclusions in the authigenic cements in Smoslashrbukk; d) Cap rock leakage in Halten Vest is seen to have occurred pene-contemporaneously with the pressure buildup in Halten Vest and is causally related to the process of transformation of the fault zone west of Smoslashrbukk from open to sealing.

Petroleum Systems Analysis in the 21st CenturymdashWhat Should We Be Working On?

Andrew S. Pepper

BP Amoco, Houston, Texas, U.S.A

LATE 20TH CENTURY: THE ROLES OF PETROLEUM GEOCHEMISTRY AND BASIN MODELING

Over the last three decades, petroleum geochemistry advanced from a frontier science to an accepted component in every petroleum basin evaluation. In many respects, however, the advance has been constrained to the static elements in the petroleum system, such as identification of the effective source rock layer in a basin, the definition of oil families, or the correlation of an oil with its source rock.

The last decade has seen the increasing use of basin models, especially multi-dimensional fluid flow simulators, in exploration. Links with petroleum engineering have also grown, with an appreciation of the impact of phase behavior In the petroleum system; we also understand more about the behavior of seals in petroleum traps. Essentially, we have become more aware of the importance of the dynamics of the petroleum system. One unfortunate consequence of the expansion of these new models, however, is that the degree of specialization needed to run the basin model has spawned "basin modeling" as a separate discipline in some organizations.

FUTURE ROLE IN THE 21ST CENTURY: THE PETROLEUM SYSTEMS ANALYST

The increasing appreciation of the impact of high performing multi-disciplinary teams within which experts are co-located places pressure on the retention of geochemist and basin modeler as separate individuals. In an environment where costs dictate increasingly lean organizations, teams will benefit from incorporation of an individual capable of integrating these areas together with phase behavior and seal behavior: performing an analysis of the Petroleum System in its widest context. In order to create such individuals, we will need to refocus efforts as I suggest below.

WHERE CAN WE AFFORD TO STOP WORKING?

We should stop wasting our efforts in our "comfort zones." One of these is the area of kinetics of oil and gas generation via kerogen degradation and oil to gas cracking.

Given the current state of the art, two opposing philosophies exist: a simple two component scheme calibrated under both laboratory and field conditions (our Orgas2 scheme) vs. a range of "custom" bulk kerogen to multi-component kinetic schemes calibrated only at laboratory heating rates.

At the present time, there are actually no published data in support of the benefits of the multi-component schemes, or indeed many of the simpler "custom kinetic" schemes. A much simpler alternative, the five-part global organofacies classification proposed in Orgas2, may seem restrictive, yet in our experience it offers all the capability needed in practical petroleum prediction. For example, Figure 1 shows a test of the scheme in a non-marine lacustrine (Organofacies C) basin. This result—to our knowledge the only field test of a kinetic model for a lacustrine basin presently at maximum thermal stress—is typical of the kind of match that can be obtained. Consequently we have invested no new research efforts in this area since 1989 and our observations on the rest of the industry—almost ten years on—would be that far too much attention is still being directed towards fine-tuning of kinetic parameters using "custom kinetics" type models.

Figure 1. Prediction of the kerogen breakdown profile (HI) vs. temperature for the Sunda Basin, Indonesia, using the simple global model of Pepper and Corvi (1995) for an Organofacies C kerogen, compared with field data.

In another area—oil to gas cracking—again the tendency is to produce laboratory-derived kinetic models without performing the necessary field validations. Yet simple models seem to perform adequately (Figure 2).

Figure 2. Prediction of the amount of oil remaining in-reservoir vs. reservoir temperature using the simple global model of Pepper and Dodd (1995) compared with field and MSSV laboratory data (0.7 K/min). Data ex. 33/9/14 and 2/4-14 are from Horsfield et al (1992).

Concepts of petroleum sorption in kerogen have also offered insights into the behavior of diverse source rocks, including coals which have been a long-standing subject of debate (Figure 3). Such models show that retention of petroleum in kerogen induces much greater control on the expulsion profile of the source rock compared to small changes in generation kinetics.

Figure 3. Predicted expulsion profile for East Java Sea coals vs. thermal stress using the simple global model of Pepper and Corvi (1995) for an Organofacies D/E kerogen and an oil sorption capacity of 0.1 kg/kgC, compared with field data. Expulsion occurs when QI (= (S1+S2)/TOC) starts to decrease at about 130 degrees C.

WHERE SHOULD WE BE WORKING IN THE 21ST CENTURY?

The emphasis demanded by our increasing capability to numerically model basin flow processes dictates a refocusing of our efforts in the future. Many elements of the chemistry and physics of petroleum phase behavior and transport in sedimentary basins remain poorly understood. Future high-reward areas will include:

rarrExpulsion: further understanding of the high P-T sorptive capacity of kerogen, especially for oil (C6+) components, and their partition coefficients in the kerogen matrix. This will further improve our ability to predict the timing of, and the gas/oil mix of, petroleum molecules entering the pore system of the mineral matrix.

rarrPrimary migration: an understanding of the geometry, intrinsic permeability and petroleum/water relative permeability of the intergranular pore/capillary system of the mineral matrix; more accurate expelled compositional prediction (above) will allow us to better understand the phase state and physical properties (e.g. density and viscosity) of the petroleum fluid(s) present in the pore space. Overall, these advances will materially improve our ability to predict storage/saturation losses during primary migration as well as the rates and possible extent of vertical migration through thick packages of fine-grained rock. We note that many earth scientists still do not believe that petroleum can migrate through significant thicknesses of fine grained rocks; thus in order to validate such concepts there will need to be a focus on geochemical techniques for detection of migrant petroleum in such rocks.

rarrPhase behavior: linkages between phase behavior and the geochemical character of petroleum fluid systems. The promise here is that by better classifying petroleum fluid systems, we will be able to develop petroleum physical property corellations that alleviate the computational complexities of full equations of state in basin simulations (the latter being a major contributing factor to extended model run-times).

rarrSeal behavior: in addition to the compaction and mechanical behavior of fine grained rocks, an understanding of fluid properties (density/buoyancy/interfacial tension) is required. Interfacial tension between the water and petroleum phases is directly proportional to the petroleum column height that can be retained beneath a capillary seal, yet the last review of interfacial tension in petroleum migration was by Schowalter in 1979.

This list of topics is probably well outside the comfort zone of many individuals who currently regard themselves as geochemists or basin modelers. However, the challenge for the 21st Century is to develop a capability to integrate petroleum geochemistry, fluid phase behavior, and an understanding of fluid transport (especially in fine grained rocks) within the confines of a basin simulation software package: the age of the Petroleum Systems Analyst is here!

The Effect of Mixing of High and Low Maturity Expelled Fluids on the Molecular Maturity Parameters of Reservoir Fluids: Implications for Petroleum System Analysis

Johan K. Michelsen1, Ganjavar Khavari-Khorasani2, and Susan E. Palmer3

1Statoil; 2Petrotrak; 3Consultant.

Petroleum geoscientists use a variety of data from reservoired petroleum, both in exploration and in production operations. Geochemical data, which reflect bulk properties of the petroleum, such as isotope data, are relatively simple to interpret and different semi-quantitative interpretive schemes have been suggested. For geochemical data, which reflect trace quantities of the petroleum such as most biomarker molecules, however, there have been no attempts to evaluate semi-quantitatively how the petroleum system geometry-related mixing will affect the biomarker parameters. When ratios of such molecules are used for maturity estimation, the ratios observed in reservoir fluids are "calibrated" against the same ratios measured on source rock extracts from source rocks with "known" maturity. In a petroleum system analysis, it is clearly important to develop a consistent relationship between already discovered fluids and the conceptual basin model used in the operation. However, if the nature of the discovered fluid is misinterpreted, the basin model may be tuned in the wrong direction. This presentation focuses on the relationship between molecular maturity parameters measured on source rock extracts versus the same molecular parameters measured on reservoired fluids.

Most source rocks have some inclination, and hence during maturation will expel simultaneously fluids with different maturity. It is normally unavoidable that a fluid expelled from a given maturity will be "contaminated" with fluids of other maturities, expelled from the same source rock. Hence, the meaning of the "maturity" of a reservoired fluid is not clear. If we mean the "source rock transformation ratio range" which dominates the reservoired fluid (which is how we use the term here), then we must be certain that the parameter we use to assess this "maturity" will respond linearly and consistently to the dominant maturity mass fraction in the fluid. One of the problems with biomarker-based maturity indicators of reservoired fluids lies in that these components are expelled with large differences in concentration, as a function of source rock maturity. If a reservoir captures fluids with a wide maturity range from a kitchen, then depending on the mass fractions of the different maturity ranges and the biomarker concentration in each of these mass fractions, we can have a wide spectrum of biomarker maturity ratios for the same maturity (i.e., for the same dominant transformation ratio range). Below we perform mixing estimates, to evaluate the probable distortions of biomarker maturity estimates from reservoired fluids.

We start with a scenario where two molecules, A and B, are always expelled with the same concentration (A + B = e.g., 100 ppm), and the measured ratio (A/(A + B)) changes linearly from 0.3 to 0.7 with increasing source rock maturity (which is what is measured in the source rock for calibration). Figure 1a shows a map view of a cone shaped petroleum system, and Figure 1b shows a vertical section through this system. We assume that an equal amount of petroleum is expelled upwards and downwards, and all the petroleum expelled downwards is accumulated without carrier-bed losses, in the reservoir at the crest (Figure 1b). The petroleum system is buried through five timesteps (Figure 1c), and the maturity at the vertical slice labeled with a black dot has just started to expel petroleum at timestep 1. We assume that the petroleum yield is constant per volume unit rock in all the mature vertical slices of the geohistory in Figure 1c. However, since the area per vertical slice increases outwards, the amount generated at the deepest basin position is always larger than the shallower vertical slices (Figure 1a). The curve [a] in Figure 1d shows the molecular ratio at the deepest source rock slice for each timestep (Figure 1c). Figure 1d[b] shows the evolution of the same molecular ratio in the reservoir. In Figure 1a, we give the width of the drainage cell of concern as the angle alpha. Hence in this first example (Figure 1d[b]), this angle is 360deg. If this angle is small, the scenario is close to a 2-D section, and Figure 1d[c] shows how the same molecular maturity parameter would have evolved in the reservoir if alpha is zero, and there is no 3-D effect. It is hence clear that for the given example, the 3-D effect would by itself cause only a shift of the maturity parameter of around 0.1. Hence, the 3-D effect by itself is not very severe.

Figure 1. The relationship between molecular maturity parameters, at the source rock positions, compared to the evolution of the same molecular parameters in a petroleum reservoir, as a function of petroleum system geometry, and concentration variation at the source rock locations (Details are explained in the text.)

However, in the above examples, we assumed that the biomarker molecules were expelled with the same concentration for all source rock maturities (A + B is constant). Biomarker molecules are typically observed to occur in largely different concentrations as a function of maturity (e.g., Bishop and Abbot, 1993; Peters and Moldowan, 1993). The biomarker concentrations in source rocks normally decrease with increasing maturity, but observations from oils (Requejo and Halpern, 1990; Requejo, 1992) may indicate that both major increases and decreases are possible. In Figure 1d[d] is shown the evolution in the petroleum reservoir, if the concentration in the vertical slices with lowest maturity is ten times higher, than for the highest maturity. Furthermore, in Figure 1d[e] is shown the case where the concentration is 100 times higher at the lowest maturity, and 10 times higher at the intermediate maturity, compared to the highest maturity. It is clear that such concentration variations can have a severe influence on the relationship between the molecular maturity parameter and the mass fractions of the different "maturity" ranges in the fluids. (The bulk contribution from the different source rock maturities is identical in all the examples.) Differences in biomarker concentrations as a function of source rock maturity can make the biomarker molecular maturity parameters insensitive to major maturity variation, i.e., there is a strong non-linear relationship between the biomarker maturity ratios and the maturity of the dominant mass fraction in the fluid. The simple geological scenario given above is only one of a wide range of possible scenarios. However, it is hard to find a petroleum system where at least not some of the geometrical elements of importance are present.

CONCLUSIONS

The main outcomes of mixing exercises of the kind above are:

  1. When biomarker concentrations vary strongly with source rock maturity, it is not possible to tell from biomarker ratios of reservoired fluids the dominant maturity of the fluid.

  2. When biomarker concentrations decrease at the source rock level as a function of maturity, the biomarker maturity ratios of reservoired fluids underestimate the true "maturity" of the fluid. The degree of under-estimation can vary from small to extreme underestimates, depending on petroleum system geometry, the ratio between total charge volume to the reservoir volume, and the variation in expelled biomarker concentrations at source rock locations as a function of maturity.

  3. The continuous mixing of the maturity ranges, which by themselves can have orders of magnitude differences in biomarker concentrations, causes smooth continuous biomarker estimated maturity trends, or biomarker maturity invariance. In the latter case, major maturity variation is masked, and major differences in bulk fluid properties (API, GOR etc.) can be misinterpreted as due to secondary processes. (It is important to notice that many studies of biodegradation and phase-fractionation use biomarker maturity indicators as the framework to isolate the secondary effects.)

  4. The biomarker maturity parameters can be used for relative maturity assessment for fluids with relatively similar bulk composition. However, the ratios are not specific and provide little information on the magnitude of the maturity difference.

Phase fractionation can complicate the problem further, and the most extreme scenario represents a gas which reequilibrates with residual immobilized oil in the carrier system. Condensates precipitated from these gases can have biomarker signatures indicative of low maturity as described by Thompson and Kennicutt (1990). Continuous burial of residual oils can also lead to the formation of "gas-stripped" condensates, as the already gas-stripped residual oil is re-evaporated into the gas stream with increasing PT.

A low-maturity signature from molecular ratios measured on reservoir fluids is in general not significant by itself. High maturity light oils and gas condensates can have biomarker maturity ratios indicative of a low maturity as outlined with the mixing principles given in Figure 1. A low-maturity estimate from a biomarker molecular ratio must be backed up by bulk fluid properties (e.g., API and GOR). In contrast, a high-maturity estimate from molecular ratios, if based on hydrocarbons, does indicate a high maturity. However, this is not necessarily the case for non-hydrocarbon- based molecular ratios. Non-hydrocarbons are prone to fractionate between an asphaltene enriched liquid (Khavari-Khorasani et al., 1998a) and the main fluid, and also to be involved in partitioning between the fluid and inorganic or organic solids (Larter and Aplin, 1995). This fractionation effect will complicate the relationship between the reservoir fluid and source rock maturity. Spatial perturbation of the hydrocarbon compositions will also occur within the reservoir itself, due to both compositionally controlled bulk flow and molecular diffusion (Khavari-Khorasani et al., 1998b), which also needs to be addressed when interpreting reservoir fluid maturity.

REFERENCES CITED

Bishop, A. N., and G. D. Abbott, 1993, The interrelationship of biological marker maturity parameters and molecular yields during contact metamorphism: Geochimica et Cosmochimica Acta, v. 57, p. 3661–3668.

Khavari-Khorasani, G., J. Dolson, and J. K. Michelsen, 1998a, The factors controlling the abundance and migration of heavy vs. light oils, as constrained by data from the Gulf of Suez. Part II: Organic Geochemistry, v. 29, p. 255–282.

Khavari-Khorasani, G., J. K. Michelsen, and J. Dolson, 1998b, The factors controlling the abundance and migration of heavy vs. light oils, as constrained by data from the Gulf of Suez. Part II: Organic Geochemistry, v. 29, p. 283–300.

Larter, S. R., and A. C. Aplin, 1995, Reservoir geochemistry: Methods, applications and opportunities, in J. M. Cubbit and W. A. England, eds., The geochemistry of reservoirs, Geological Society Special Publication No. 86: The Geological Society of London, p. 5–32.

Peters, K. E., and J. M. Moldowan, 1993, The Biomarker Guide: Prentice Hall, Inc, p. 363.

Requejo, A. G., and H. I. Halpern, 1990, A geochemical study of oils from the South Pass 61 field, offshore Louisiana, in D. Schumacher and B. F. Perkins, eds., Geochemistry of Gulf Coast Oils and Gases: Their Characteristics, Origin, Distribution and Exploration and Production Significance: SEPM, p. 219–235.

Requejo, A. G., 1992, Quantitative analysis of triterpane and sterane biomarkers: Methodology and applications in molecular maturity studies, in J. M. Moldowan et al., eds., Biological Markers in Sediments and Petroleum: A Tribute to Wolfgang K. Seifert: Prentice-Hall, p. 222–240.

Thompson, K. F. M., and M. C. Kennicutt II, 1990, Nature and frequency of occurrence of non-thermal alteration processes in offshore Gulf of Mexico petroleums, in D. Schumacher and B. F. Perkins, eds., Geochemistry of Gulf Coast Oils and Gases: Their Characteristics, Origin, Distribution and Exploration and Production Significance: SEPM, p. 199–218.

Using 2-D Thermal Modeling to Predict Fluid Quality in Deepwater Exploration Areas

Gary A. Cole, Z. (Alan) Yu and Rick Requejo

BHP Petroleum, 1360 Post Oak Blvd., Houston, TX, 77056 U.S.A

The objective of this paper is to use 2-D modeling to try to understand the emplacement of the crude oils in deepwater exploration areas prior to drilling. Simply stated, what is the risk on drill potential sands containing biodegraded versus fresh oils?

Two areas of the world where exploration interest is high are the Kwanza and Lower Congo basins of Angola, West Africa, and the Gulf of Mexico deepwater regions. Both the deepwater Lower Congo and Gulf of Mexico basins have been successfully drilled, and results have indicated large accumulations of biodegraded and fresh oils. To minimize risk, 2-D modeling can be used to determine the timing of oil emplacement within mapped reservoir sections. The timing of oil charge can then be matched to a temperature versus depth profile, and these relationships can then be used to determine the risk on biodegradation.

These relationships have determined the following for both the Gulf of Mexico and Lower Congo/Kwanza basins of Angola:

  1. When charge occurs when reservoir intervals are lt50C, the likelihood of biodegradation is high.

  2. When charge begins at low temperatures (lt40–50C) but continues during burial (reservoir temperatures increase to gt60–65C), the oils are movable and consist of an initial biodegraded charge, but are retopped and usually overwhelmed by a fresher charge.

  3. When charge begins at temperatures gt50C, the oils are generally fresh and unaltered.

The Angolan and Gulf of Mexico models have modified how we think of biodegradation of oils. The worldwide published average for the biodegradation deadline is around 65–70C. These models now indicate a low biodegradation threshold where altered oils generally occur when oil charge occurs at lt50C, and when oil emplacement is above 50C, the oils are fresh and unaltered.

Modeling Overpressures: How Good Are We?

Gareth S. Yardley1, Richard E. Swarbrick2, and Andrew C. Aplin3

1GeoPOP, Department of Petroleum Engineering, Heriot-Watt University, Edinburgh, EH14 4AS, UK; 2GeoPOP, University of Durham, UK; 3GeoPOP, Newcastle University, UK

Much of the overpressure observed in sedimentary basins can be attributed to disequilibrium compaction. This process can be modeled using a variety of basin modeling software packages. The prospect of using such packages for quantitative pressure modeling and even pressure prediction is now being widely considered. Modeling sensitivity analyses and case studies are used in this presentation to examine the range of uncertainties that are associated with modeled pore pressures.

Accurate knowledge of sediment, especially shale, porosity and permeability evolution is crucial. This is particularly true for modeling deep overpressured targets where compaction behaviour derived from empirical data from shallow sediments and/or soil mechanics theory may not apply. Currently there a wide number of formulations describing porosity and permeability reduction in shales commonly used in basin modeling packages. Unconstrained use of these relationships can give large uncertainties of up to 20 MPa at 5 km in modeled pressures (e.g. Figure 1). New techniques recently developed at Newcastle University allow compaction curves to be more tightly constrained based on shale characterization using analysis of log data. This allows the uncertainty in the modeled pressures to be reduced, particularly where the majority of compaction is mechanically controlled.

Figure 1. There is a wide variety of porosity-permeability functions used to describe shale permeability evolution during burial. Unconstrained use of these can lead to large uncertainties in modeled pressure. The pressure-depth plot above shows modeled pore pressures for a 6000 m shale section buried at a constant rate of 100 m/Ma. The large difference in resultant pressures is due to the choice of the permeability loss function: MM-MM uses the Mann and Mackenzie (1990) function; MM-PM uses the PetroMod default function.

The burial history and the structural evolution of the system also play a key role in our ability to model pore pressures. At very high burial rates virtually all additional overburden load is carried by the pore fluids regardless of which particular relationships are used to described shale porosity and permeability evolution. With lower burial rates accurate knowledge of sediment porosity and permeability evolution becomes more important.

An accurate assesment of the fluid flow paths in 2-D and 3-D can also be important for pressure modeling. Lateral transfer of pressure along inclined aquifers that connect shallower structures to deeper, more overpressured parts of a basin can enhance the pore pressure at structural crests. The actual pressure contribution from 2-D and 3-D flow, even where extensive inclined aquifers exist, is determined by the structural evolution of the system. Therefore, it is not always necessary to build complex 2-D and 3-D models in order to perform representative pressure calculations. A case study of a Triassic field from the North Sea, with an extensive dipping aquifer, shows that the modeled pressures are the same in both 1-D and 2-D models. However, in other cases the lateral transfer of fluids along dipping aquifers can increase the pressure calculated in 1-D models by at least 10 MPa in reservoirs at any depth that are connected to aquifers with extensive reliefs (gt1 km). A modeled example is shown in Figure 2 where shales deposited over the last 30 Ma have caused an underlying aquifer to become inclined. Incorporation of the effects of the inclined aquifer into the flow modeling gives rise to a 7 MPa difference in the aquifer pressure predicted from the 1-D and 2-D models.

Figure 2. Contour plots of excess pressure (i.e. pressures in excess of normal pressure) variation in a subsiding shale basin with an inclined sandstone layer. Excess pressures are produced by disequilibrium compaction and are calculated in a basin modeling package. Results are shown for a 2-D section and for 1-D models generated at the crestal and down dip positions of the 2-D section. Modeled pressures are sensitive to the presence of the inclined high permeability sand layer and a good 2-D (and 3-D) representation of the geological structure is needed for accurate pressure modeling.

The possibility of performing quantitative pressure modeling of other overpressure generation mechanisms such as gas generation will also be addressed using modeling sensitivity analyses.

Can Sandstone Diagenesis Induce Fluid Overpressure?

R. H. Lander, H. M. Helset, J. C. Matthews, and L. M. Bonnell

Geologica, Stavanger, Norway

Geopressures in deeply buried Cenozoic basins can typically be explained by basin models that simulate the "compaction disequilibrium" that arises from the mechanical compaction of low permeability shales. Current basin models, however, have been less successful in predicting geopressures in pre-Cenozoic basins when they also honor measured shale and sandstone properties. This paper reviews a quantitative model of pore volume reduction in sandstones resulting from quartz diagenesis through geologic time and evaluates the potential role that quartz diagenesis could play in geopressure development in pre-Cenozoic basins.

The diagenetic model simulates mechanical compaction in sandstones as well as quartz cementation and associated chemical compaction. Intergranular volume serves as a proxy for the mechanical compaction state and is simulated to be an exponential function of maximum effective stress. Quartz cementation is modeled as a precipitation-rate limited process and silica is assumed to have been derived from within the modeled sandbodies at stylolites or clay-rich interfaces. Chemical compaction occurs at the dissolution zones in proportion to the extent of quartz cementation within the nearby sandbody and the concentration of detrital and authigenic quartz that occurs at dissolution zones.

Results suggest that at depths greater than 3000 m the rates of pore volume loss resulting from quartz diagenesis in sandstones can be more than an order of magnitude greater than loss rates from shale compaction and two orders of magnitude greater than pore volume expansion rates resulting from aquathermal expansion. Sensitivity analyses indicate factors that promote increased rates of porosity loss due to quartz diagenesis include high thermal gradients, surface areas for overgrowth nucleation, and rates of burial.

Evaluation of diagenetic modeling results for Jurassic sandstones from the North Sea indicate that peak rates of porosity loss due to quartz diagenesis in sandstones are comparable in magnitude to those associated with shale compaction. Importantly, a given rate of pore volume loss resulting from quartz diagenesis is more likely to generate geopressure within sandstone "pressure compartments" than a comparable rate from shale mechanical compaction. Most shale porosity loss occurred in the Cretaceous, thus permitting tens of millions of years for associated geopressures to dissipate. By contrast, diagenetic modeling suggests that porosity loss rates from quartz diagenesis are presently at or near peak values. Moreover, the peak rates in porosity loss from quartz diagenesis occur at depths gt2500 m where permeabilities are as much as five orders of magnitude lower than near the depositional surface where shale compaction is most rapid.

In Miocene sediments from the U.S. Gulf Coast, the relative significance of quartz diagenesis to shale compaction as an agent of geopressure generation is more limited. Quartz diagenesis, however, may be the most important factor for generation of geopressures over the last fifteen million years in sandstones at depths greater than about 4000 m. At such depths quartz diagenesis has the potential to increase geopressures that had formed earlier in response to shale compaction to the point where hydraulic fracturing occurs.

Buoyancy and Interfacial Force Effects on Two-phase Displacement Patterns and Implications for Secondary Oil Migration

Tomochika Tokunaga and Katsuro Mogi

Department of Geosystem Engineering, University of Tokyo, Tokyo, Japan

The mechanics of secondary migration of oil are considered to be well understood physical processes that can be dealt with quantitatively (Hubbert, 1953; Schowalter, 1979). We often apply two-phase fluid flow equations in porous media to express migration based on combining Darcy's law, mass balance equations for both water and oil phases, and the capillary pressure discontinuity relation. These equations are established for artificial production/injection processes in which fluid flow is mainly governed by large pressure differences. However, the importance of interfacial and buoyant forces during secondary migration are much different than for artificial production/injection process. Thus, it is questionable whether the two-phase equations are applicable to secondary migration.

To better understand the effects of interfacial and buoyant forces during oil-water displacement process and secondary migration and to evaluate the applicability of two-phase equations, one-dimensional vertical oil-water displacement experiments were conducted. Oil was injected at a constant rate from the lower inlet of a glass tube packed with sorted glass beads. The injection pressure and oil outflow rate were measured while observing the displacement pattern. Runs were carried out using different grain sizes and injection rates.

Two displacement patterns were recognized during the experiments: type A consisted of stable, piston-like displacement and type B consisted of capillary fingering. The difference coincided with the relative magnitudes of driving forces, which can be characterized by dimensionless numbers, the modified Bond number (Bo') (ratio between buoyant and interfacial forces) and the Capillary number (Ca) (ratio between viscous and interfacial forces). Type A pattern was produced for high Ca/Bo' ratios and type B for low Ca/Bo' ratios. A "phase diagram" showing the regions of the two displacement patterns was constructed in Ca-Bo' space, including the effects of gravity (Figure 1). The "phase diagram" obtained from our experiments and that from Lenormand et al.'s (1988) network simulator are not entirely consistent. This discrepancy might be due either to numerical simulation vs. physical experiments, different pore structure (pore/throat network model vs. packed glass beads), or inappropriate application of the Darcian velocity to calculate the Capillary number, or a combination of above reasons. Our results also showed that excess pressure for the non-wetting phase fluid to intrude into an porous medium was rate dependent (Figure 2).

Figure 1. Log-log plot of the modified Bond number and the Capillary number. Open circles indicate the type A pattern, solid circles the type B pattern, and open triangle 'type A?'. Crosses indicate the estimated conditions for secondary migration. Numbers attached to each dot indicate the water saturation at the final steady state condition, and those attached to each cross express the angle between the migration direction and horizontal direction.

Figure 2. The relationships between injection rate, permeability and pressure jump for the intrusion of oil into a porous medium.

Calculated Ca/Bo' ratios during secondary migration of oil suggest that horizontal oil migration falls within the type A region, whereas vertical migration falls within the type B region, due mostly to the different buoyancy effects (Figure 1). This result together with the rate dependency of the excess pressure suggests that ordinary two-phase fluid flow parameters (relative permeability and capillary pressure curves) obtained from conventional core analysis might not be applicable for modeling secondary migration.

REFERENCES CITED

Hubbert, M. K., 1953, Entrapment of petroleum under hydrodynamic conditions: AAPG Bulletin, v. 37, p. 1954–2026.

Lenormand, R., E. Touboul, and C. Zarcone, 1988, Numerical models and experiments on Immiscible displacements In porous media: Journal of Fluid Mechanisms, v. 189, p. 165–187.

Schowalter, T. T., 1979, Mechanics of secondary hydrocarbon migration and entrapment: AAPG Bulletin, v. 63, p. 723–760.

Evolution of Fluid Pressure and Compositional Histories Revealed by Petroleum Fluid Inclusions

Andrew Aplln1, Steve Larter1, Gordon Macleod1,4, Ashley Bigge3,5, Rebecca Lloyd, Richard Swarbrick2, and Gareth Yardley3

1NRG, University of Newcastle, UK; 2Petroleum Engineering, Heriot Watt University, UK; 3Geological Sciences, University of Durham, UK; 4Present address: Shell EampP Technology, Houston, U.S.A.; 5Present address: Fluid Inclusion Technologies Inc., Broken Arrow, Oklahoma, U.S.A

A central aim of basin models is to predict fluid flow, pressure and composition, both now and through the geological past. Constraining data are limited. Pressure and compositional data are available from drilled structures but historical data describing the evolution of the fluid system are essentially absent. This contrasts with reservoir models, which are continuously updated and constrained by matching predicted and real production data.

Here we show how petroleum-bearing fluid inclusions can be used to generate data which enable us to track the evolution of fluid pressure and petroleum composition through geological time. Confocal Laser Scanning Microscopy is used to generate three-dimensional images of single petroleum inclusions. The liquid petroleum fluoresceses and allows one to quantify the volume of liquid and vapour within the inclusion. Using PVT modeling software, the liquid:vapour ratio is used along with the homogenisation temperature to determine the saturation pressure, gas-oil ratio, viscosity, molar volume, density and surface tension of included petroleum to a precision of a few percent. These results give the saturation (minimum) pressure at the saturation (minimum) temperature. Coexisting aqueous fluid inclusions are needed to estimate true trapping temperatures and pressures.

Because the pressure, compositional and temperature data are generated on individual inclusions, analysis of inclusion suites allows us to determine the evolutionary pathways of fluid pressure and composition. If the accuracy of the pressure estimates is validated, these data will provide hitherto unavailable constraints on palaeopressure in a similar manner to the way in which vitrinite reflectance is routinely used to constrain thermal histories.

North Sea case studies illustrate the new technique's potential and the ways in which the data can be used to constrain both the inputs and outputs of basin models:

  1. In three cases the currently reservoired fluid is retrograde gas condensate. Fluid inclusions reveal that black oil was previously present in the reservoir and indicate the temperature and thus time at which the black oil was replaced by condensate as a result of injection of a volatile petroleum. Combined with thermal modeling of adjacent source kitchens, these data indicate migration pathways and reveal periods when reservoirs must have been connected to source areas. Combined with biomarker analyses of reservoired fluids, the data constrain the nature (maturity) of the late injected fluids and thus the maturity of source kitchens when the reservoir was at temperatures known from fluid inclusion data.

  2. By revealing pressure histories, the fluid inclusion data allow one to constrain permeability histories of key sedimentary sequences. In an example from the Central North Sea (Figure 1), the pressure history of Jurassic reservoirs can only be modeled by assigning the overlying chalk a similar permeability-porosity relationship as that for mudstones. Models suggest that currently high overpressures are largely the result of disequilibrium compaction resulting from rapid, late Tertiary burial. Recent gas generation provides minor additional overpressure. Some of the overpressure has, however, been present since the early Tertiary, retained by the low permeability chalk.

  3. A final example indicates how inclusion data were used to infer how and when oil migrated from Jurassic to Tertiary structures in the North Sea, through a thick Cretaceous mudstone sequence. The current Jurassic reservoir is highly pressured and contains condensate. However, inclusion data suggest that black oil migrated into the Jurassic structure when pressures were close to hydrostatic. Remigration to Tertiary reservoirs is unlikely to have resulted, as previously assumed, because of hydrofracturing resulting from very high fluid pressures. We suggest that the black oil in the overlying Tertiary structure migrated through the Cretaceous mudstones when the whole section was shallower, more permeable, hydrostatically pressured and exhibited lower capillary threshold pressures.

Figure 1. Evolution of hydrostatic, lithostatic and pore fluid pressure in three wells from a Central North Sea field.

Empirical Correlations to Describe Petroleum Phase Behaviour in Sedimentary Basins

W. A. England

BPAmoco

In order to describe the phase changes that occur during generation, expulsion and migration it is necessary to have a robust, unbiased estimator of the relative phase volumes of petroleum gas and liquids as a function of pressure and temperature. For resource valuation it is essential to be able to predict the gas-oil ratio in exploration prospects.

In this presentation, a simple two component model is described, together with its calibration on data from over 100 petroleum accumulations. Differences between major petroleum provinces were found, enabling the development of either a "global" predictive model or models for specific basins.

This model is too simple to show some of the complex phase behaviour associated with equation of state models, but in exploration applications, where inadequate tuning data is available, it offers a reliable method for basin/prospect evaluation.

Characterisation of Petroleum Fluids in the Froslashy Field and Rind Discovery, NOCS, Using Petroleum Geochemical, Fluid Inclusion and PVT Data

A. G. Bhullar1, D. A. Karlsen1, K. Holm2, and R. di Primio3,4

1University of Oslo, Department of Geology, P. O. Box 1047, Blindern, N-0316 Oslo, Norway; 2Elf Petroleum Norge As, P. O. Box 168, N-4001 Stavanger, Norway; 3University of Oslo, Department of Geology, Oslo, Norway; 4Now at Saga Petroleum, Norway

The characteristics of oils in the subsurface as well as the hydrocarbon filling history of the Froslashy Field and the Rind Discovery, Block 25/5 and 25/2 Norwegian North Sea, were studied using geological, organic geochemical, fluid inclusion and PVT data. Analysis of normal alkane and biomarker distributions in the three separate data sets provides us with data for understanding the filling history of these structures. Oils from the Rind Discovery showed significant differences in terms of maturity, gas to oil ratio (GOR) and source rock facies like %C28 sterane content and bisnorhopane/(bisnorhopane + norhopane), indicating this structure to have received oil from a basin different to that which sourced Froslashy Field. Detailed biomarker analysis of 600 core extracts indicated most oils encountered in the Froslashy Field to belong to one uniform population, a fact supported by the good correlation between formation volume factors and saturation pressures calculated using a PVT simulation package. The general homogeneity of core extracts from two main structural compartments indicated a good reservoir communication and does not seem to have been adversely affected by a major fault dividing the Froslashy Field. However, one compartment penetrated by well 25/2-A2 contains slightly biodegraded oil entirely different from the rest of the field, most likely sourced from Dunlin shales, indicating lack of communication with the main Froslashy structure.

Maturity differences between the sampled oils in Froslashy indicated that the most mature oil is encountered in the crest of the reservoir and oils of decreasing maturity occur towards the flanks. GOP, saturation pressure and formation volume factor trends supported this interpretation (Figure 1). The phase envelopes of oils from five wells situated at the crest, flank and intermediate positions of the reservoir reflected the typical evolution in terms of maturity and upward/vertical migration of petroleum. Oil from the crest of the reservoir showed the highest amount of the polar compounds (resins + asphaltenes). The higher concentration of polar compounds in this sample is interpreted to reflect asphaltene precipitation related to the mixing of lower and higher maturity fluids in the crest of the reservoir and production of these asphaltenes in suspension during well testing. Iatroscan TLC-FID analysis of this sample revealed high asphaltene contents, thus supporting this interpretation.

Using the burial history of Froslashy, filling of the subcompartment holding today biodegraded oil sourced most likely from Dunlin shales must have occurred about 40–50 mybp and before oil from the Kimmeridge age Draupne shales more recently filled homogeneously the main Froslashy structure. In Rind, migration of oil between the main southern compartment and the presently dry neighboring northern compartment can, based on the relative maturity differences of the oil in these two structures, be timed relatively. Evidently the high maturity oil filled the presently dry structure subsequent to filling the main structure, which in low and high properm units holds bitumen spanning a large maturity interval. Fault movement between these two compartments disrupted communication and cap rock was blown out due to pressure buildup in the presently dry northern structure, while the oil charge in the main structure was preserved from pressure buildup due to lack of communication with the deep generative basin. Using the difference of the homogenization temperatures of the water inclusions of these two systems and the burial history, a novel concept for the time for cap rock leakage is suggested. Information on time for reservoir filling and seal failure, when used to calibrate basin modeling, should provide a better model for the recognition of basin-scale migration patterns.

Figure 1. Saturation pressure vs. formation volume factor.

Gas Breakthrough Experiments on Mudrocks: Experimental Results and Implications for Two-Phase Flow

B. M. Krooss and S. Schloumlmer

Institute of Petroleum and Organic Geochemistry (ICG-4), Forschungszentrum Juumllich GmbH, D-52425 Juumllich, Germany

Gas breakthrough experiments have been performed to investigate the gas sealing efficiency of mudrocks. The laboratory measurements were conducted with initially water-saturated core plugs of different orientations over extended periods of time (weeks to months). The procedure consisted of maintaining a constant gas (methane) pressure in the upstream reservoir of a flow cell and recording, continuously, the pressure evolution in the closed downstream volume of the system.

The measurements produced characteristic breakthrough curves with an onset of gas transport across the rock sample occurring usually after a distinct lag time (Figure 1). The residual difference between the upstream pressure and the final downstream pressure is interpreted as the capillary entry pressure after re-imbibition of water into the largest interconnected pores.

Figure 1. Pressure curves recorded in a gas breakthrough experiment performed with methane on initially water-saturated red claystone.

Based on the pressure curves, temperature and gas viscosity data and the downstream volume of the setup the gas permeability of the samples was calculated as a function of time and pressure difference using Darcy's law for compressible media. A Klinkenberg correction for slip flow was not applied. The permeability coefficients determined by this method usually exhibited a rapid increase at the onset of gas breakthrough, followed by a plateau and a subsequent decrease.

A conceptual scheme based on a parallel capillary model was applied to compute the equivalent pore-size distribution of gas-conducting pores during the pressure-decline phase of the breakthrough experiments. The results of this evaluation imply that for the samples studied, gas transport under the experimental conditions occurs through pores with equivalent radii ranging from 100 to lt10 nm with pore-size distribution maxima showing distinct differences for different mudrocks (Figure 2).

Figure 2. Comparison of pore-size distributions of gas-conducting equivalent pores derived from the results of breakthrough experiments.

Towards Including Chemical Alteration in Basin Models

Peter Meulbroek1, Steven Losh2, Lawrence Cathles2, and Jean Whelan1

1Woods Hole Oceanographic Institute; 2Cornell University.

Hydrocarbon fluids can be altered during migration by undergoing phase separations. These separations can be caused by either migrating between regimes of differing pressures and temperatures, or by mixing of two partially miscible fluids. The alteration occurs when the chemically distinct phases physically separate, resulting in fluids that are fractionated with respect to the original mixture.

An important type of phase fractionation results when a migrating gas mixes with oil. If the amount of gas exceeds the carrying capacity of the oil, then the gas will fractionate the oil by removing its most soluble components. The amount of alteration is on par with such mechanisms as biodegradation, water washing, and geochromatography. Gas phase fractionation, known by terms such as "evaporative fractionation," "migration fractionation," and "gas washing," produces a very distinct signature of alteration in the affected fluids. This signature includes light-end depletion, enrichment in aromatic compounds, and a characteristic shift in the pattern of n-alkanes (see Figure 1).

Figure 1. Effects of gas phase fractionation.

Current basin models do not include the detailed hydrocarbon composition information necessary to show fractionation, nor do they include methodology for calculating the complex phase behavior of hydrocarbon mixtures. This is unfortunate, since compositional changes can form an important calibration for modeling. Furthermore, an understanding of the complex chemical changes due to phase fractionation can lead to better prediction of hydrocarbon quality, and, perhaps, hydrocarbon quantities, in deeper reservoirs of developed fields. We are in the process of incorporating detailed chemical analysis with pressure and temperature modeling to estimate changes in hydrocarbon compositions under realistic conditions.

The present work combines the modeling of phase fractionation with pressure and temperature modeling to better estimate the depth of fractionation, to constrain the pressure and temperature evolution, and to better understand the filling history of a field. Initial estimates of pressure evolution (perhaps estimated using downhole porosity measurements) can be used to estimate depths of washing. These depths can then be reconciled within a geological framework to refine pressure estimates. If the washing information is combined with sedimentation and subsidence rates, estimates of filling time can be determined (see Figure 2).

Figure 2. Basin model refinement methodology.

Through application of numeric modeling to a case study at Eugene Island Block 330, we have related the amount of alteration in hydrocarbon chemistry to the conditions under which the fractionation event took place. The numeric modeling has suggested several parameters that can quantify the changes in n-alkane composition due to gas washing. These parameters can be used to estimate the amount of gas that has "washed" (come into proximity and reach chemical equilibrium with) an oil, and the depth (pressure-temperature) conditions under which the oil has been washed.

The changes gas phase fractionation causes in migrating fluid can be more profound than previously thought. Results of this application have led to the conclusion that the oils at Eugene Island have been "gas-washed" by 12 moles of gas/mole of oil, at a depth of 2.4 km, or just below the current reservoirs. Applications of the insights derived from Eugene Island to several other locations in the Gulf of Mexico has led to the conclusion that gas phase fractionation is common at locations throughout the Gulf. At a study area in South Marsh Island, oils have lost much of their light ends to gas phase fractionation; in some cases, up to 93% by mole of the oil has been lost. This behavior is not isolated; at several fields in the Vermilion area, oils have lost over 90% by mole of their light ends to gas washing.

Geochemistry of Organic-Rich Rocks From Mud-Volcano Ejecta in Azerbaijan: A Novel Approach For Regional Assessment of Source Rock Quality

Gary H. Isaksen1 and Adil Aliyev2

1Exxon Exploration Co., Houston, Texas, U.S.A.; 2Geological Institute of Azerbaijan, Baku, Azerbaijan

The South Caspian Basin is a prolific oil and gas province. As a result of its very thick sedimentary package (up to 25 km), likely Jurassic oceanic crust, and rapid sedimentation rate (10 to 12 km of sediment fill deposited in the last 6 million years), the basin is relatively cool. Geothermal gradients range from 20degC/km in the Kura Depression to 15degC/km in the SCB proper. Consequently, source rocks for oil remain immature for oil generation down to approximately 6 km. The rapid burial and compressional regime has resulted in the formation of numerous mud volcanoes, which are common in the Gobustan area of eastern Azerbaijan and throughout vast regions of the SCB.

Rock ejecta brought to the surface by mud volcanoes offer a unique opportunity to characterize sedimentary units both within, and beyond, conventional drilling depths. Biostratigraphic analyses along with lithological comparison of the samples with age-equivalent outcrop samples have enabled age-dating of the ejecta. The organic-rich rocks of the Oligogene-Miocene Maikop Formation are the primary source rocks for oil and gas in the basin and are present as ejecta. These rocks have total organic carbon contents up to 7%wt. and hydrogen indices up to 500 mg HC/g org. C. They are dominated by marine, algal-amorphous organic matter which accumulated under dysoxic to anoxic conditions. The high content of algal material in the kerogen is also evidenced by a predominance of C27 regular sterane biomarkers. Based on biomarkers and stable carbon isotopic data we have identified depositional environments and integrated these data to reconstruct the paleo-geography of the source intervals.

Pre-Drill Pressure Prediction Provided by Multi-Dimensional Basin Modeling Techniques

S. J. Duumlppenbecker, M. J. Osborne, J. R. Bunney, and R. Duncan

BP Exploration

Basin modeling tools were traditionally used to study basin development, source rock maturation, petroleum generation and expulsion. The emphasis were very much on understanding processes at the regional to semi-regional scale, and references to fluid flow were taken to mean migration of petroleum fluids. More recently, the same tools have been used more at a prospect-specific scale.

Relatively little attention has been given to the reconstruction of the pressure regime in basin modeling studies. The main overpressure generating mechanisms are disequilibrium compaction and gas generation, both of which can be quantitatively modeled. Sedimentation rate and permeability structure of the section being modeled are primary controls on water flow in the subsurface, and ultimately on fluid pressure. Understanding the past and present fluid flow regime in the basin is an essential step to prediction of pressures at a prospect well location. In BPX, calibration of models to observed data such as pressures and temperatures measured in the well bore is an integral step in any basin modeling study. Where feasible, fluid inclusion data are also used to constrain paleopressures. When the modeling techniques accurately reproduce the fluid pressures at known points of control, the models are able to predict pressures ahead of drilling at new locations in the basin. Models are then continuously updated as new information becomes available as the well is drilled. Even in frontier areas (no wells), basin models can be used for risk analysis, or they can be used to constrain pressure predictions made using other methods, eg. seismic.

In BPX such prediction of subsurface pressure and temperature is used by drillers and engineers for well planning and highlights potential drilling hazards. The presentation will concentrate on application of 2-D and 3-D basin modeling for predicting subsurface conditions (Pressure and Temperature) ahead of the drill bit. The results show why numerical models gained the respect of the wider user community — exploration geologists, geophysicists and drilling engineers — and why usage is passing from the hands of the specialists to generalists in integrated multidisciplinary teams.

Mechanical and Chemical Compactions and their Applications in Sedimentary Basins

A. Revil1 and L. M. Cathles2

1CNRS-CEREGE, BP 80, 13545 Aix-en-Provence, Cedex 4, France; 2Cornell University, Dept. of Geological Sciences, 14853 Ithaca, NY

In recent years, there has been increasing interest in establishing realistic models of rheological rock behaviors and compaction to explain time dependent deformation in sedimentary basins. Such models are required in order to evaluate fluid flow and stresses in sedimentary basins, the rheology of faults, assessement of fluid overpressures, and land or seafloor surface subsidence resulting from man's activities and oil or gas withdrawal from reservoirs. Two major deformation processes in silico-clastic sediments are (1) mechanical compaction and (2) pervasive pressure solution tranfer (see Figures 1 and 2). Mechanical compaction results from slippage and rotations of grains, which change their position and orientation, but not their shape, in order to reach a more compact rearrangement and a higher mean contact coordination number between the grains. Pervasive pressure solution transfer is a chemical compaction phenomenon. In a stressed quartz porous aggregate, grain boundaries are more highly stressed than free faces in the pore space. Grains dissolve at their contacts. The solute, Si(OH)4, migrates from the grain boundaries into the surrounding pore space and, when the pore fluid is already at saturation with respect to the solute, the solute precipitates as euhedral quartz overgrowths on the free faces of the grains. Both compaction phenomena are supported by both petrographic evidences and field observations and experimental works. Mechanical compaction is described by a purely plastic model whereas pervasive pressure solution is described by a poro-visco-plastic (Voigt-type) model. Applications decribed in the poster concern (1) porosity versus depth in situations of equilibrium and disequilibrium compactions, (2) compartmentation modeling and fluid pressure assessement, (3) cement content versus depth, and (4) land subsidence due to fluid withdrawal.

Figure 1. Hydrostatic compaction of quartz sands in sedimentary basins. A. In the first hundred meters, quartz sands grains are loosely packed and the depositional porosity is in the range 0.50ndash0.60 depending of the deposition process. B. The first stage of compaction corresponds mainly to mechanical compaction. Sliding and rotation of the grains decreases the porosity until a random compact assemblage is reached with a porosity in the range 0.36ndash0.40 if the grain size distribution is unimodal inside a representative elementary volume. Pressure solution plays only a minor role at shallow depths (lt1.5 km). C. Pressure solution allows the porosity to decrease (in hydrostatic fluid pressure conditions) until a residual porosity of 0.01ndash0.05 is reached, which corresponds to a percolation level.

Figure 2. Pressure solution in clay-free sands. a. We consider a representative elementary volume (REV) of a clay-free sand saturated by an aqueous solution and submitted to a given effective stress history. b. Pressure solution starts by dissolution of the grain contacts, diffusion of the solute at the grain contacts, and precipitation on the free pore faces of the grains. c. The grain to grain contact area is formed by a gel layer of silicic acid chains. These chains form a brush of protruding "hairs", which can resist to high stress concentration due to short range steric repulsions. d. The deformation of the representative elementary volume follows a poro-visco-plastic (Voigt-type) rheological behavior. The spring describes a thermodynamic equilibrium state whereas the dashpots represent thermodynamic disequilibrium at the grain contacts (the dashpots "p" and "d" correspond to "precipitation kinetic " and "solute diffusion" limited processes, respectively).

REFERENCE CITED

Revil, A., 1999, Pervasive pressure-solution transfer: A poro-visco-plastic model: Geophysical Research Letters, v. 26, p. 255–258.

Application of 2-D Basin Modeling for Pre-Drill Pore Pressure Prediction in the Gulf of Mexico and Offshore West Africa

Z. (Alan) Yu, G. Cole

BHP Petroleum, Houston, Texas, U.S.A

The accurate prediction of formation pressures prior to drilling is critical for both cost reduction and drilling safety. With the advances of two-dimensional basin modeling during the past two decades, particularly in gaining a better understanding of the dynamic parameters controlling basin fluid flow/compaction, 2-D basin modeling has been used vigorously as a tool for pressure prediction within BHP and the petroleum industry.

Overpressured (pore pressure in excess of hydrostatic) sequences within a sedimentary basin develop when the fluid escape rate from sediments is less than the porosity/void reduction. Primary factors controlling flow rates are compaction behavior of dominant lithology (usually shales/muds), rock permeability, sedimentation rates, stratigraphic geometry, and the overall basin's burial history. In other words, the formation pressures and rock physical properties (permeability, porosity, and density) observed within in a basin are the results of the naturally dynamic fluid flow/compaction processes generated by a basin's geological history.

The offshore eastern deepwater Gulf of Mexico features a high sedimentation rate, recent deposition and significant salt sheet tectonics. High overpressure is observed throughout the basin resulting in significant and costly drilling problems, particularly when saltsheet presence is involved. Extremely high overpressure zones (or low effective stress) are reported in many subsalt wells while only mild or regional overpressure is observed in other subsalt wells. Therefore, predicting subsalt pressures is essential for drilling success in the Gulf of the Mexico.

Using Temispack, a commercially available 2-D Basin Modeling package, a generic composite geological model from the Gulf of Mexico deepwater is presented for pressure modeling, determining the mechanical seal integrity, and predicting the "gumbo" zone assuming various geological settings.

Other basin types may show different pressure relationships; therefore, a case study from the deepwater Lower Congo Basin will be presented.

Overall performance of pressure prediction from dozens of sub-basins with well data comparison in the Gulf of Mexico and offshore West Africa will be discussed. The highly encouraging results demonstrate the viability of using this technology prior to drilling. For best results, though, this technology should be applied industrywide for pressure prediction before, during and after drilling.

A Paleolatitude Approach to Assessing Surface Temperature History for Use in Burial Heating Models

Charles E. Barker

U.S. Geological Survey, Box 25046, MS 939, Denver, CO 80225 U.S.A.

Calculations using heat flow theory as well as case histories show that over geologic time scales, changes in mean-annual surface temperature (Ts) on the order of 10degC penetrate kilometers deep into the crust. Thus, burial-heating models of sedimentary basins, which typically span kilometers in depth and persist over geological time frames, should consider Ts history to increase their accuracy. In any case, Ts history becomes important when it changes enough to be detected by a thermal maturation index like vitrinite reflectance, a parameter widely used to constrain burial heating models. Assessment of the general temperature conditions leading to petroleum generation indicates that changes in Ts as small as 6degC can be detected by vitrinite reflectance measurements. This low temperature threshold indicates that oil and gas windows can be significantly influenced by Ts history.

A review of paleoclimatic factors suggests the significant and geologically resolvable factors affecting Ts history are paleolatitude, long-term changes between cool and warm geological periods (climate mode), the degree to which a basin is removed from the sea (geographic isolation), and elevation or depth relative to sea level. Case studies using geologically realistic data ranges or different methods of estimating Ts in a burial heating model indicate a significant impact of Ts when (1) continental drift, subduction, tectonism and erosion significantly change paleolatitude, paleoaltitude, or paleogeography; (2) strata are at, or near, maximum burial, and changes in Ts directly influence maximum burial temperature; and (3) an increase or decrease in Ts occurs near the oil or gas windows causing petroleum generation to or cease. Case studies show that during the burial heating and petroleum generation phase of basin development changes in climate mode alone can influence Ts by about 15degC. At present, Ts changes from the poles to the equator by about 50degC. Thus, in extreme cases, continental drift alone can seemingly produce Ts changes on the order of 50degC over a time frame of 107 years.

Structural Influence on the Evolution of Petroleum Systems in the Powder River Basin

C. N. Wold1, R. J. Coskey2, D. L. Rasmussen1, C. P. James II1, and J. E. Leonard1

1Platte River Associates, Inc.; 2790 Valmont Road; Boulder, CO 80304; 2Rose Exploration, Inc.; 518 17th Street, Suite 430; Denver, CO 80202

The Powder River basin in eastern Wyoming (U.S.A.) is an ideal laboratory for petroleum systems modeling. Over 30,000 wells have been drilled during three main phases of petroleum exploration: 1) discovery of oil seeps and drilling near those seeps on the eastern margin of the basin; 2) drilling on basin margin structural plays; 3) drilling of stratigraphic plays. The extensive drilling of the basin and regional structure-contour and isopach maps provide an excellent database for multi-dimensional petroleum systems modeling.

Multi 1-D models are used to reconstruct the timing and distribution of hydrocarbon generation and primary expulsion from various source rocks. The multi 1-D modeling also provides a backstripped reconstruction of carrier bed geometry through time. The shape of the carrier bed through time is the primary control on the location of traps, migration fairways, and migration shadows determined using a 2frac12-D model for secondary migration. The 2frac12-D model incorporates hydrocarbon bouyancy, capillary pressure gradients, and hydrodynamic forces to predict migration pathways through time. The predicted prospects can be compared to known oil fields to evaluate the success of the applied modeling techniques.

Use of Fault-Seal Analysis in Understanding Petroleum Migration in a Complexly Faulted Anticlinal Trap, Columbus Basin, Offshore Trinidad

Richard G. Gibson1 and Peter A. Bentham2

1BP-Amoco Trinidad Exploration; 1BP Amoco Upstream Technology Group, 501 Westlake Park Blvd., Houston, Texas 77079-2696

In this paper, we present an analysis of hydrocarbon migration pathways within a petroleum field located in the Columbus basin, offshore from the southeast corner of Trinidad, West Indies. The reservoir-bearing Pleistocene stratigraphic section in this area consists of shales interbedded with thick, poorly consolidated sands. The structure of interest is a broad 4-way-closed, NE-trending anticline cut by a series of NE-dipping normal faults. Stacked petroleum columns exist within the blocks bounded by these normal faults, and the faults can be clearly seen to bound the accumulations. Integrated fault-seal studies have been undertaken in the area to understand the controlling properties of these faults, and these studies have led to an better understanding of the active petroleum system within the fields of the Columbus basin.

Drilling over the past 30 years has defined the distribution of charged reservoirs within a series of fault-bounded compartments. Shallower reservoirs are petroleum-bearing in the eastern portion of the field. Farther westward on the structure, shallow reservoirs in valid traps are found to be wet and progressively deeper reservoirs become petroleum-bearing (Figure 1). It appears that the location of fault-sealed accumulations is closely controlled by the larger structural geometry, which changes significantly from the shallower reservoirs to the deeper ones.

Figure 1. Field cross-section showing the distribution of petroleum-bearing reservoirs.

Previous attempts to explain the spatial distribution of productive sands within the field relied on the assertion that the NE-dipping normal faults acted as the key routes for hydrocarbons to access the higher reservoirs within the structure. One such model involved initial migration along the large fault on the west side of the field and subsequent migration across both faults and stratigraphy from deep sands in the west to shallower sand in the east. Such models require that the faults acted as both conduits and barriers to fluid flow, either simultaneously or at different times in the filling history of the field. In addition, the tortuous migration pathways implied by such models have limited ability to predict the distribution of hydrocarbons in untested blocks within the field.

Post-appraisal of the most recent delineation drilling shows that most of the hydrocarbon columns in the structure are limited in height by either synclinal spill points within individual fault blocks or cross-fault spill points on the west side of fault blocks (Figure 2). The pattern of hydrocarbon-water contacts within the field suggests that petroleum filled and spilled its way from NE to SW across the structure within individual sands. Vertical migration of hydrocarbons along the bounding faults is not required to explain the distribution of productive sands, and this is consistent with both petrophysical data and the known sealing character of the faults. This reservoir filling model serves as a tool for predicting primary targets and column heights in untested fault blocks within the area.

Figure 2. Proposed migration model within the field.

Advances in and the Future of 3-D Structural Modeling

Alan D. Gibbs and Stephanie Kape

Midland Valley Exploration Limited, 14 Park Circus, Glasgow, G3 6AX, U.K

Structural Modeling has been revolutionized in the last few years with the development of 3-D modeling systems that have taken us far beyond what was possible with map and section representations of data. This revolution offers the potential for significant improvements in understanding geological systems and risk reduction in our industry.

3-D modeling technologies fall broadly into 3 classes of system: 1 Model building, characterization and description; 2 Process based kinematic analysis tools; 3 Process based volume analysis.

It is the second two systems that provide the most powerful tools, as they enable analysis in the fourth dimension: time. As a result, the completed 3-D model is valid not only in the sense of geometric and physical completeness, but using time as a constraint implies a geohistory and hence means that the model is also valid as an evolving geological structure.

Many advances have been made (and are being continually developed) with the aim of improving modeling geological processes with mathematical algorithms. These new developments are crucial in reducing risks inherent in the 3-D structure modeling process itself. The finite element analysis provides the latest risk solution; however, it is greatly constrained by economic viability as such models take a longer time to execute and often require a down-sampled 3-D grid. 3-D kinematic analysis does not contain the rock property information of a finite element analysis (and hence has a higher inherent risk) but is faster and more efficient to run, and may therefore be applied over a larger amount of data.

It has become clear, however, that the greatest use of 3-D structural modeling is to integrate the structural modeling with other geological disciplines. Frequently, the structure model is the base from which other modeling processes are carried out. It is therefore not only essential to have a valid structure model prior to performing other analyses, but it has become apparent that the time component of the structure model can be used to greatest effect when used in a multi-disciplinary approach. We will describe how 3-D structural modeling can be applied at various stages of the work flow in order to reduce technical risk.

INTEGRATION WITH SEDIMENTOLOGICAL DATA

3-D structure models have an implied geohistory, and as a result, it is possible to determine "palaeo-topographic" maps at each time stage of the model. When this information is combined with a ray or pathway analysis, a palaeo-flow path across a surface may be determined. This is useful in predicting the distribution of sediment fairways though time in areas where tectonics have had a controlling influence.

When the information on sediment fairways is combined with core and well log data, facies maps at each time stage may be derived. This information is particularly pertinent when used as a predictive tool in areas where well control on the data is sparse.

One of the new challenges is to further develop these technologies (using ray/pathway analysis or a full 3-D cellular system), so that full sequence stratigraphic analysis can be included in 3-D kinematic models. A number of groups are working on this approach.

INTEGRATION AT A BASIN MODELING STAGE

Ray and pathway analysis may again be used (analogous to the approach taken to determine sediment fairways) to integrate the 3-D structure model with the hydrocarbon system model. Basin analysis technologies are now being integrated with the 3-D structural model in many groups. There are clear benefits from having a knowledge of the changing pattern of sediment fairways when hydrocarbon migration pathways are considered. Facies maps across a basin may give a better indication of source distribution.

Although this approach is very much in its infancy, problems in resolving charge, recharge and sourcing may now be minimized by integrating 3-D structural modeling with sedimentary analysis and basin studies.

As well as full 3-D backstripping, the 3-D structural evolution of the basin can identify the development and switching of fluid fairways with time. In many cases this can demonstrate mechanisms to predict hydrocarbon type, timing and spill when integrated with a thermal geochemical model.

INTEGRATION AT A RESERVOIR SCALE

The 3-D structural model implies a geologic process, from which it is possible to derive strain data from a 3-D structural model. Both finite and dilatational strains may be derived from a 3-D process based kinematic model, and give information on the strains implied by the algorithm used to model rock deformation.

Spatial analysis of strains implied can be used in some cases to successfully predict fracture fairways, which have relevance to both fluid migration and trapping, as well as to reservoir performance. Integration of 3-D finite strain data with fracture network prediction analysis gives a better indication of fracture orientations and density between wells. For the first time this approach of integrating derived strain data with well and core data offers a significant advance in our ability to model and predict fractured reservoirs.

3-D strain derived from the volumetric analysis can be used to condition the reservoir model by providing a process based prediction of fracture distribution and intensity as well as stress orientation. Several studies have now completed this loop with significant success.

Fault seal analysis in 3-D incorporates a variety of ways in which the problem can be addressed. These include statistical techniques, fault population modeling using real data and physical models which can be used to generate visualizations of fault networks and 3-D cellular property arrays which can then be used in finite element simulation.

By using the constrained 3-D structural model a full 3-D stress model can also be derived, and a number of studies now exist which attempt to tie this back to well observations of fracture logs and breakout studies. At field and semi-regional scale a similar approach can be adopted to predict fluid leakage on fault systems due to stress reactivation.

The 3-D structural model, particularly when coupled to a kinematic or dynamic process model, offers exciting new opportunities which were not apparent even a few years ago. The authors believe that we have the potential for a significant revolution in the way in which we work and that we expect to see process based interpretation offering new breakthroughs in our understanding of hydrocarbon systems. For this to happen, however, we need to continue the development of the science and technologies but to improve integration between the disciplines using the 3-D structural model as the hub and to migrate these technologies out of the expert groups.

Modeling of Tectonic Mechanisms that could Explain the Post-Rift Basin Development in the Viking Graben, Offshore Norway

Willy Fjeldskaar

RF-Rogaland Research, Stavanger, Norway

To get an understanding of the evolution of the basin geometry and the temperature regime, it is necessary to study tectonic processes that take place in the basin. It is well known that a realistic reconstruction of the basin evolution is important for the temperature estimation and hydrocarbon migration. Particularly, the movement of mass connected to faulting can be critical. More large-scale processes could, however, also have significant importance, because they can affect the geometry and paleo heat flow.

Viking Graben is a playground for basin modeling, because of the extensive exploration work that has been done in the area. A detailed study of the basin evolution of the Viking Graben shows that the post-rift subsidence is not consistent with a simple McKenzie extension model. There is, e.g., a significant mid-Tertiary phase of uplift. This paper reports modeling of different tectonic processes that could explain the post-rift subsidence pattern in the Viking Graben.

The modeling is done on several large 2-D transects crossing the Viking Graben, and the various processes or modeling techniques are built into the BMT system. Therefore, the modeling is done with high resolution in time and space. Discriminating between the different candidates of tectonic processes can be done based on both magnitudes and wavelengths. The effect of the processes investigated here spans in wavelength from a few tens of kilometers to hundreds of kilometers.

The modeled process with the longest wavelength is phase boundary migration. The response of a phase transition in the lithosphere to pressure changes leads to changes in the upper lighter phase, and thus greater effects than expected from isostasy alone. The effect of intraplate stress could be significant, provided the elastic thickness of the lithosphere is low. The effect would then be of relatively short wavelengths. Lateral and temporal changes in the elastic thickness and the subsequent effect for the isostatic compensation have also been modeled. The effects could be significant, and of medium wavelengths. The effect of a finite necking depth will be uplift in parts of the area, but with relatively short wavelengths. Finally, the effect of the lithosphere's visco-elastic properties is tested, showing that significant basin flank uplift could result.

3-D Structural Geology: Restoration Algorithms for Basin Modeling

Graham Williams, Steven Kane, Andrew Richards, Andrew Dodds, and Stuart Clarke

Basin Dynamics Research Group, Keele University, England

Restoration of interpreted seismic volumes in 3-D is increasingly used to validate interpretations in structurally complex areas, thus reducing exploration risk. Retro-deformation in time steps provides sequential geological models of structural geometries back through time. Three-dimensional flexural isostatic modeling may be applied in the reverse modeling procedure. Variables used in 3-D flexural isostatic modeling include structural geometries, sediment versus water fill, thermal structure of the lithosphere and effective elastic thicknesses (Te). Using an exponential decay of porosity with depth the effects of decompaction of layers of varying lithologies and with laterally varying lithologies can be modeled at each time step. The calculated geometries of stratal surfaces stored at each stage in the retro-deformation may be used to produce a forward model to which is added hydrocarbon generation, migration and trapping, thus providing a full understanding of sedimentation and structuring through time and their controls on the petroleum system.

Structural restoration algorithms control deformation through time in the reverse model, and they must conserve volume, or allow volume to change in a controlled manner due to compactional/decompactional effects. Structural restoration based on simple shear constructions uses either vertical or inclined shear geometrical techniques. For inclined shear in 3-D restoration the main components are the horizontal displacement vector, the deformation plane that contains the principal strain axes and the vertical/inclined shear angle parallel to which the hanging wall deforms during translation. It is possible to vary the orientation of the displacement vector and shear angle in both azimuth and plunge. Restorations are very sensitive to both shear angle and to the movement direction. Using map view visualisation of faulted hanging-wall and footwall cutoffs, it is possible to obtain a good estimate of the original slip vector thus minimising errors in restoration.

The flexural flattening technique involves the restoration of complexly folded surfaces to a horizontal plane while conserving surface area and minimising finite strain. Multiple surfaces showing complex folding are restored using the flexural flattening method applied sequentially to progressively deep surfaces. Volumes between the uppermost flattened surface and underlying surfaces are preserved giving volumetric balance. The jigsaw fit of footwall and hanging-wall cutoffs of flattened surfaces in map view provides a unique restoration solution based on translation and/or rotation of the hanging-wall block. Sequential restoration of progressively deep surfaces may incorporate three-dimensional decompaction at each restoration step.

The newly developed flexural flow restoration algorithm involves the movement of particles in a fault hanging wall parallel to the underlying fault. The technique is is loosely based around the geometrical constructions of classical inviscid fluid mechanics, which means that hanging-wall fold geometry is determined primarily by the shape of the underlying fault, without the requirement for predetermination of hanging-wall fold geometry prior to restoration, or establishment of a direct relationship between fault shape and fold axial surface geometries prior to forward modeling. Stratal surface area and volume balance is achieved via the application of a heterogeneous fault parallel shear to the hanging wall during forward deformation and/or restoration. The technique uses a laminar, fault parallel flow field to define the movement of the sedimentary layers in the hanging wall of the fault and has been applied to complex, curviplanar fault geometries in 3-D. It is simpler than current fault bend fold techniques which rely on cumbersome geometrical analysis of hanging-wall fold shape prior to retro-deformation hanging-wall fold structures.

Computer models using ray tracing techniques are used to forward model hydrocarbon migration from source kitchens to traps. The forward models allow sequential stages of deformation, sediment accumulation and compaction, source rock maturation and fluid flow.

3-D Visualization and Analysis of Fault Seal: Offshore Myanmar

Linji An1, Russell K. Davies, Donald A. Medwedeff and Dennis Yarwood

ARCO Exploration, 2300 West Plano Parkway, Plano, Texas 75075; 1Current address: 5401 Independence Drive, Apt. 901, Plano, Texas 75023

BACKGROUND AND MOTIVATION

The Gulf of Martaban, offshore Myanmar, is underlain by a thick sand-shale sequence of Plio Pleistocene age. Regional extension of these sediments has set up potential fault seal traps against major normal faults. A recent exploration well in prospect Shwe Pyi Htay (SPH) tested seven stacked horizons (DHI-1 to 7) in the footwall of one these normal faults (fault S1). Six of the seven horizons were supported by anomalously bright seismic amplitudes. The well (SPH #1) encountered producible gas in only one prospective interval (DHI 6). The remaining intervals contained residual or low saturation gas. Our hypothesis is that the zones of residual gas are due to failed fault traps or poor top seal capacity.

The purpose of this study is to use the well results from SPH #1 to constrain the mechanisms for fault seal on fault S1 as a calibrated threshold for fault seal potential, and to use this calibration to mitigate fault seal risk on additional prospects in the same area. Interactive 3-D visualization of the seismic anomaly distribution, trap geometry and fault seal risk distribution projected onto the fault surface assist in the evaluation of the model calibration.

TECHNICAL APPROACH

Fault seal potential

We investigate the effects of two main processes that control fault seal in clastic sequences: juxtaposition and shale gouge. Juxtaposition seal quantifies the effects of permeability differences across a fault that juxtaposes a reservoir interval against a non-reservoir interval. Shale gouge reduces the fault permeability by mixing shale with reservoir and non-reservoir lithologies in the fault zone.

Juxtaposition

In this analysis, the sealing effect of stratigraphic juxtaposition is estimated using a relative sand quality or Kr (expressed in %) which is approximated from the Vshale fraction determined from the well log,
Equation 6

If Vshale is 1, Kr = 0% and the stratigraphic interval is like an impermeable shale. If Vshale is 0, Kr = 100% and the interval is like a permeable sand.

Shale Gouge

The sealing effect of shale incorporated into the fault zone is estimated using the shale gouge ratio or SGR. SGR is an estimate of the percent shale dragged into the fault zone with fault displacement. A high SGR is equivalent to a high potential for fault seal.

Fault Seal Potential

Juxtaposition and shale gouge ratios contribute simultaneously to the sealing potential of the fault. We represent this composite contribution as a combined parameter called the fault seal potential or FSP. At each point on the fault, the FSP is the greater of the SGR and the Vshale percent. A high potential for fault seal (FSP high) occurs when either SGR is high or Kr is low. FSP is low when SGR is low and Kr is high.

APPLICATION TO FAULT S1

We measured the horizon cut-off data along fault S1 for 11 seismically mapped horizons. The horizon cut-off data is expressed in seismic travel-time. The relative sand quality between these seismic horizons was divided into 56 unique strata based on Vshale values determined from well SPH#1. The stratigraphy was juxtaposed onto the fault surface assuming no lateral changes along the fault map length and the same stratigraphy in the hanging wall and footwall of the fault.

We calculated Kr, SGR, and FSP along the mapped fault surface from these data and displayed the results on a strike projection of the fault. A 3-D model of the interpreted horizon and fault surfaces helps to visualize the sensitivity of the fault seal parameters to the trap geometry and seismic anomaly distribution. Three seismic intervals (DHI-2, DHI-6, and DHI-7) were chosen for discussion and seal calibration because they have the best quality reservoir sands. In addition, DHI-2 and DHI-6 have thick top-seal shale intervals, reducing the ambiguity of the analysis.

RESULTS

The calculated minimum FSP for the prospective intervals of horizons DHI-2, DHI-6, and DHI-7 are 45%, 65%, and 30%, respectively. The primary parameters SGR and Kr rank in the same order. Thus, prospect DHI-6 has the highest potential for fault seal. DHI-6 is the only zone with producible gas, suggesting that FSP accurately predicts the degree of fault seal.

The gas-water contact calculated from the well data approximately corresponds to the lateral extent of the seismic amplitude anomaly. This contact allows independent corroboration of the FSP threshold by comparing the elevation of the gas-water contact to the distribution of FSP on horizon DHI-6. The FSP just below the gas-water contact on DHI-6 is 55%; the minimum FSP above this contact is 65%, as discussed above. If the gas-water contact in DHI-6 is controlled by the low FSP value this is a threshold value for fault seal. The minimum FSP of 65% for DHI 6 at the crest of the structure and 55% at the gas-water contact provides a threshold range for fault seal.

Determination of the threshold range of FSP for DHI-6 assumes that the limits of the gas charge correspond to the extent of the mapped seismic anomaly, which is contained entirely in the fault block. However, the velocity effects of shallow gas significantly distort time structure maps. Thus, within the precision of the velocity control, the gas-water contact determined from the well logs extends to the fault bounding the adjacent fault block. Integration of the FSP data analysis with this structural geometry of horizon DHI-6 indicates that the coincidence of the gas-water contact and the FSP minima may be fortuitous. Thus, the threshold FSP may only be constrained between 45% and 65%, based on comparison to prospects DHI-2 and DHI-7.

CONCLUSIONS

The fault seal risk of future exploration in adjacent areas of the Gulf of Martaban can be mitigated by pre-drill calculation of FSP and discounting of prospects with FSP values below the 45% to 65% threshold.

Hydrocarbon Migration Pathway Evolution Consequential of Basin Structuration

Paul Huggins1, Stuart D. Burley1,2, Oslashyvind Sylta3, Are Tommeras3, Stephanie Kape4, Stuart Bland4, and Nick Kusznir5

1BG Technology, Gas Research amp Technology Centre, Ashby Road, Loughborough, LE11 3GR, UK; 2Basin Dynamics Group, Department of Earth Sciences, Keele University, Keele, ST5 5BG, Staffs, UK; 3SINTEF Petroleum Research, Trondheim, Norway, N-7034; 4Midland Valley Exploration Ltd, 14, Park Circus, Glasgow, G3 6AX UK; 5Department of Earth Sciences, University of Liverpool, Liverpool, L69 3BXUK

Hydrocarbon migration in carrier systems and accumulation within traps are dynamic processes that evolve through time. Migration pathways within basins are controlled by a combination of high permeability conduits, such as faults and sand-rich carrier sequences, and the structural dip, or geometry, of such conduits. Both the distribution of conduits and their structural dip are intimately related to the structural evolution of the basin.

In basins where hydrocarbon migration is contemporaneous with the structuration of the basin, simulation of hydrocarbon secondary migration, accumulation and re-migration within structurally restored horizon surfaces are prerequisites to the accurate prediction of hydrocarbon occurrence. Changes in the dip of the migration pathway and the timing of trap formation in relation to hydrocarbon generation influence the hydrocarbon phase and GOR along individual fairways and in different structural traps.

Commercial 3-D secondary migration simulators do not incorporate structural restoration beyond vertical decompaction and backstripping, whilst structural restoration techniques do not simulate secondary hydrocarbon migration. One approach to incorporate restoration into migration simulators is to use sequential time step-output from structural restoration techniques as input to migration simulators. Results of integrating 3-D flexural isostatic backstripping, fault tectonics and unfolding into the burial history of horizon-based, ray-trace, secondary migration modeling demonstrate significant effects on the distribution, pattern and evolution of secondary migration fluid flow paths. These changes also influence the distribution and phase of resulting hydrocarbon accumulations, as well as the fill-spill history of the accumulations. Comparison of migration flowpath histories for structurally restored and "static" present-day horizon surfaces shows significant changes in migration directions, increased migration focus through spilling from temporary traps and sealing faults, and remigration concomitant with trap destruction.

Structural restoration induces geometrical artifacts into the modeled surfaces which are a function of both the validity and quality of the input interpretation, and of the algorithm used. These features are an inherent and useful part of the structural validation process, but, once carried into the migration model they create significant anomalies in fluid migration patterns. In the ray-tracing models, fluid flow responds to the smallest of structural features creating a dip change. Migration can thus be artificially focused into surface corrugations, produced during restoration, which may be oblique to the regional horizon dip. Additionally, inaccurate fault closure and fault cutoff drag/reverse drag artifacts channel flow along fault strike, the hydrocarbons never actually entering the fault zone, making open faults appear as sealing faults. Such geometrical features will require significant manual editing to condition structural models prior to migration modeling. The sensitivity of structurally restored model geometries to geological and model uncertainties needs to be carefully considered and appropriate methodologies applied to address specific geological questions at different scales.

3-D Modeling of Petroleum Migration—A Comparison of Various Approaches and Their Applicability

C. Zwach1, T. Hantschel2, G. E. Fladmark1, and N. Telnaeligs1

1Norsk Hydro AS, Research Center, P.O. Box 7190, N-5020 Bergen, Norway; 2IES, Integration Exploration Systems GmbH, D-52428 Juelich, Germany

The modeling of petroleum systems has become a powerful tool to understand complex inter-depending processes during the burial history of source and reservoir rocks in sedimentary basins. Application of such numerical tools offers therefore great help during the assessment of the exploration risk of hydrocarbon prospects.

The processes involved during secondary migration of hydrocarbons can be simulated in three dimensions with the help of newly developed numerical programs. Among others, two different approaches are presently widely applied: 1) map-based ray-tracing of migration paths in reservoirs and 2) advanced fully 3-D basin modeling with finite elements considering multiple components and fluid phases.

While the latter approach is honoring the physical formulation of the transport process with regards to all relevant driving forces (i.e. buoyancy, interfacial tension, pore and capillary pressure, relative permeability) it has in practice the disadvantage of depending on numerous input data and of rather long computer processing times (hours to days).

On the other hand, simpler methods such as map-based secondary migration modeling may oversimplify complicated situations where sub-vertical migration, hydrocarbon phase separation and mixing play an important role.

It is important in our opinion to critically choose between the available migration modeling approaches in order to reveal the best possible results with respect to project time frames, available input data and geological situations. The selection of one of the 3-D migration modeling tools varies ultimately with the scale of the study area from large regional studies in unexplored areas to the scale of petroleum fields with high 3-D geometry resolutions.

3-D Basin Modeling Developments with Applications from the Gulf of Mexico and Offshore Niger Delta

L. M. Cathles1, E. L. Colling2, and A. Erendi1

1Cornell University; 2Texaco Exploration and Production Technology Department

The development of 3-D basin models with complex structural features, fully coupled fluid flow computations, and network compositional kinetics is just becoming a reality. The new basin models are useful not only to reduce exploration risk on a basinwide scale, but also to prioritize blocks of interest on sub-basin or smaller scales. However, the models must be robust and easy to use to contribute in an industrial setting.

Stability issues largely involve the handling of faults in 3-D. In many cases faults are hydrocarbon migration conduits, especially at depths less than a few kilometers. This is clearly indicated by observations in extensional basins with high deposition rates, such as the Gulf of Mexico and the Niger Delta. Most gridding schemes either superimpose a regular grid on complex stratigraphies and structures, or capture some or all of the geologic complexity with disjointed grids, or grids that are highly deformed in fault zones. A practical alternative to these two extremes is to capture the geology accurately away from faults and re-grid inside faults to avoid grid distortion. This approach allows easy implementation of a broad set of algorithms for fault permeability, such as controlling permeability by the smear/gouge ratio, when the fault is active, or when fluid pressure exceeds some level. We illustrate this approach with the two complex fields shown in Figures 1 and 2.

Figure 1. A 9 times 11 times 9 km deep portion of the offshore Louisiana Gulf of Mexico centered on South Eugene Island Block 330.

Figure 2. Flow in a complex system of listric and antithetic faults in the offshore Niger Delta.

The new information available from basin modeling is largely in the organic chemistry of hydrocarbons. It is thus necessary that basin models accommodate both the kinetic and thermodynamic properties of hydrocarbon generation, expulsion, initial migration, phase fractionation, and mixing during migration. To complicate matters further, the combined physical properties of hydrocarbon gases interacting with coarse and fine-grained sediments can lead to unusual capillary sealing effects. Migrating hydrocarbons can change the permeability structure of a basin and this can significantly affect their later migration.

The area of the offshore central Niger delta shown in Figure 1 illustrates the factors important in developing a 3-D model. The major sequences, or delta cycles, in this area are controlled by progradation and retrogradation of the Niger delta, and comprise a complex mix of shoreface, estuarine, fluvial and delta plain deposits. Sediments deposited along the unstable progradational Niger delta complex were subject to syndepositional slumping and faulting. Present day reservoirs are located in Pleistocene, Pliocene, and Miocene non-marine and shallow marine sediments. 3-D modeling suggests that considerable vertical migration occurred in the complex structures associated with the known petroleum reservoirs. Maturity modeling indicates that the source rock was at its most productive during the Early to Middle Miocene. Therefore, deeper reservoirs could be the source of hydrocarbons that re-migrated to reservoirs younger than the Middle Miocene. Deep reservoirs are a possible exploration target.

The presentation will discuss the philosophy and essential elements of the GBRN/GeoGroup BasinView modeling system and illustrate these capabilities with models of the Niger delta and Gulf of Mexico systems shown in Figures 1 and 2. Academic and industry perspectives will both be given.

3-D Basin Modeling: Can We Avoid Coupled Dynamic Fluid Flow and Petroleum Migration Simulation? Lessons Learned from Case Studies

J. M. Gaulier1, J. Wendebourg1, F. Schneider1, O. Brevart2, and N. Schoellkopf3

1Institute Francais du Petrol; 2ELF Exploration Production; 3Chevron

3-D basin modeling yields a true volumetric petroleum charge and migration calculation. Indeed, fluid flow pattern analysis shows that in numerous cases, a simple 2-D line or a single carrier bed map cannot account for a realistic charge scenario in basins where regional carriers are redistributing petroleum fluids generated in a single kitchen area, where several reservoir layers are interacting in a "cascade" logic, or where paleo-accumulations are redistributed due to later geometric modifications (tilt, displacement of the sedimentation depot center, etc).

Building a 3-D instead of a 2-D model also helps checking the consistency of model input data. Assembling the isopach maps of the different horizons in the basin improves the level of confidence in the structural interpretation. Creating a suite of lithofacies maps implies careful use of well information together with a consistent view of the sedimentation framework throughout the basin's sedimentation and erosion history. Regional geodynamic processes, such as isostasy of the basin, subsidence history in relation to tectonic activity, etc., provide strong controls that can be integrated to validate and improve the geologic interpretation of the 3-D regional evolution.

The simulations themselves allow one to test the consistency of the basin's evolution as interpreted by the geologist with respect to the physical conservation laws that govern the thermal evolution, the generation potential of the source rock, and the dynamics of the fluids during the geologic history of a sedimentary basin. Different geological hypotheses can be tested by performing appropriate simulations. Their consequences can be compared to the actual geological data.

In this process of excluding unlikely scenarios, uncoupling of processes is difficult to justify a priori because the interaction effects of processes are a function of the local and dynamic geological situation which varies widely between case studies and therefore must be evaluated individually. This is particularly true for 3-D studies as compared with 2-D or 1-D studies where potential interactions are amplified by the additional spatial dimension(s). Example applications of IFP's TEMIS3-D basin modeling code will be shown from the West African Atlantic margin.

4-D Basin Modeling: The Dawn of Quantitative Prospect Appraisal

M. Giles, J. Meijerink, J. Toth, P. Dijkstra, P. Riviere, A. Loopik, M. Hordijk, E. Dufour, S. Bettemborg, A. van den Hoeven, P. Huysse, D. Hindle, P. van der Geest

Shell Research and Technical Services, Rijswijk, Neths

The volume and composition of hydrocarbons in an undrilled trap are the result of a series of complex series of temporally variant processes including source rock and reservoir deposition, thermal and organic matter maturation, HC expulsion, HC migration and overpressure and trap integrity history. The application of 1- and 2-D basin modeling has so far failed to realise the full potential of basin modeling in prospect evaluation because in reality 4-D processes dominate. Temporally variant geometrical effects mean that the use of 1- or 2-D models is inappropriate. For instance, the sealing history of a fault may change during its history, thus reconstruction of the fault movement and sealing properties is essential. Prospect appraisal frequently only pays lip service to the results of basin modeling, often trivializing the complexity of the process involved and giving a false sense of security by employing probabilistic approaches. However, the goal of basin modeling is to provide an estimate of the volume and composition of hydrocarbons in undrilled traps and as such is a primary tool for prospect appraisal. However, to be useable in prospect appraisal such models must predict the associated uncertainties. Before the advent of 4-D models it was not possible to adequately model the geometrical complexities, while the workstation computing capacities have only recently allowed the calculation of uncertainties. In this paper we will demonstrate the capability of 4-D basin models to estimate the volume and composition of hydrocarbons in undrilled traps and to estimate uncertainties, taking account of these effects on HC expulsion, HC migration, overpressures and trap integrity evolution.

3-D and 2-D Modeling of Salt Movements in the Gulf of Mexico and Their Control of Fluid Flow

Reinold R. Cornelius1, Alex Erendi2, and Larry M. Cathles III2

1GSCI Consulting; 2Cornell University

The Global Basins Research Network's (GBRN) basin reconstruction tool Ageohist and their finite element code Akcess.Basin are part of our present 2-D and 3-D modeling capability. Ageohist is a pre-processor to Akcess, which defines the evolving 3-D finite element macro grid that describes basin development over time. Akcess.Basin is a flexible finite element solver, which accepts differential equations in template form. The ".Basin" hooks input non-linear material properties as the calculations proceed.

Our focus in the studies reported here was to model salt evolution on two scales in the offshore Louisiana Gulf of Mexico, compare the GBRN models to published analyses, and evaluate the effect of salt movement on aqueous and hydrocarbon fluid flow.

The existing GBRN capabilities include:

  • Decompaction and back-stripping

  • Paleo-sea level reconstruction according to isostacy or, alternatively, paleo-shelf break progradation

  • Erosion

  • Overpressure and retarded compaction underneath a fixed or a migrating seal

  • Faulting

  • Salt diapirism and salt weld formation

  • Heat flow calculations and calibration

  • Maturation calculations with standard or custom kinetics

  • Separate or simultaneous solutions of temperature, pressure, stream function for convective flow, and salinity.

The GBRN capabilities do not include:

  • Extensional tectonics

  • Salt buoyancy

We calculated large-scale salt movements in a 100 times 200 kilometer area of the Gulf of Mexico extending from the Louisiana coastline to the base of the continental slope. The model reconstructs salt movements (both upward and laterally) from a base salt layer, which generally rises north to south, yet is highly undulating. The case includes a migrating seal, which fixes the top of overpressure at a prescribed depth and migrates through the stratigraphic section as sedimentation proceeds (Revil et al., 1998). The model does not consider sub-salt lithology or stratigraphy and does not address the mechanics of salt emplacement. However, salt movement is inferred from the sedimentation pattern and changing water column. We use the model to assess the effects of salt movements on the regional temperature, fluid pressure, and fluid flow. Regional flow from below the salt escapes through salt welds into mini basins flanked by salt.

Our smaller scale model investigates salt movement in more detail in an area about a quarter the size of the large-scale analysis. This area is located near the giant oil field of South Eugene Island Block 330. The model addresses the interaction among faulting, salt movement, and fluid flow as an allochthonous salt body is transformed by salt migration into a sediment-filled mini basin. The mini-basin analysis proceeds from a published seismic interpretation by Mark Rowan (1995). The basin is bounded by a growth fault on the proximal side and by a large salt body, flanked and topped by an anti-regional normal fault, on the distal side. In addition, it contains an antithetic fault splaying from the growth fault. We model the episodic flow of water and hydrocarbons from sub-salt pressure compartments through salt welds and faults into supra-salt strata below the top of overpressure.

Our structural model includes:

  • Passive diapirism. Salt rise is modeled by downbuilding. Salt fall is modeled by salt withdrawal due to higher than average sedimentation rates only. Salt burial is modeled once the salt is surrounded by salt welds.

  • Salt weld formation and the timing of salt weld formation. Weld ramps may dip in all directions.

  • Faulting. The heave of a fault is balanced by salt withdrawal at depth only, without basin extension.

  • Sub-weld, truncated sediments.

Salt migration is calculated automatically in our models from variations from the average sedimentation rate at each time step. We follow Nelson's (1991) concept of passive salt rise driven by differential sediment loading. The salt thickness is minimized at each well to coincide exactly with the present-day values. This method predicts the location and timing of salt weld formation. Since the algorithm is driven entirely by the sediment record, it is fully automatic and requires no user input.

Our basin evolution can be favorably compared to, and agrees almost exactly with, Rowan's independent structural reconstruction. The main difference from Rowan is that Ageohist does not consider extension and is therefore presently applicable only to tectonic regimes where sedimentation outweighs extension. It also presently does not consider multi-level salt sheets, multiple salt feeders, salt over-hangs or pinched-off salt bodies. Since it considers neither buoyancy nor extension, we do not model reactive diapirism, active diapirism, or true salt fall (Vendeville and Jackson, 1992a and b). The case studies from the Gulf of Mexico demonstrate, however, that simple models of passive downbuilding can provide an adequate simulation of the geologic evolution for the 3-D integration of salt kinematics, basin evolution, and fluid flow (Figure 1).

Figure 1. A partial view of the regional 3-D-model shows growth of salt domes and the formation of a salt weld at South Eugene Island Block 330. The view is from the southwest. The top of salt is shown as solid lines, the base of salt as dotted lines. Salt thickness, where present, is indicated by thick, vertical bars. Salt welds occur where dotted and solid lines merge. The base of salt surface is tinted gray, whereas the salt weld is left white. Figure a) shows a continuous salt sheet with an uneven base and incipient salt domes at 4.0 Ma. Figure b) shows the salt weld at South Eugene Island Block 330 at 0.45 Ma, surrounded by salt domes to the north, east, south, and west. The salt reconstruction was achieved entirely automatically, based on calculations of differential loading from the overlying sediment record. Temperature, pressure, and fluid flow are calculated as affected by salt growth and weld formation.

REFERENCES CITED

Nelson, T. H., 1991, Salt tectonics and listric-normal faulting, in A. Salvador, ed., The Gulf of Mexico basin: GSA, DNAG v. J, p. 73–89.

Revil, A., L. M. Cathles, J. D. Shosa, P. A. Pezard, and F. D. de Larouziere, 1998, Capillary sealing in sedimentary basins: GRL, v. 25, p. 389–392.

Rowan, M. G., 1995, Structural styles and evolution of allochthonous salt, central Louisiana outer shelf and upper slope, in M. P. A. Jackson, D. G. Roberts, and S. Snelson, eds., Salt Tectonics: A Global Perspective: AAPG Memoir 65, p. 199–228.

Vendeville, B. C., and M. P. A. Jackson, 1992a, The rise of diapirs during thin-skinned extension: Marine and Petroleum Geology, v. 9, p. 331–353.

Vendeville, B. C., and M. P. A. Jackson, 1992b, The fall of diapirs during thin-skinned extension: Marine and Petroleum Geology, v. 9, p. 354–371.

Comprehensive 3-D RTM Modeling: From Fractured Reservoirs to Salt Tectonics

P. Ortoleva, A. Park, D. Payne, K. Tuncay, X. Zhan

Laboratory for Computational Geodynamics, Dept. of Chemistry, Indiana University, Bloomington, IN 47405 U.S.A

Predicting the location and characteristics of reservoirs in basins requires a 3-D approach that accounts for a comprehensive set of coupled reaction, transport and mechanical (RTM) processes. The model must be 3-D to account for fracture orientation and the potential for the changing direction of basin compression/extension, heat anomalies, sedimentation patterns, and, where present, salt movement. The model must integrate a comprehensive collection of processes to capture diagenesis, rock rheology, oil and gas generation, and multiphase fluid flow, among other RTM processes. All of these processes affect one or more basin properties, and thereby each other.

Presented in this lecture is the most current RTM model of the Laboratory for Computational Geodynamics (LCG) at Indiana University. It is based on moving hexahedral finite element simulation of

  • diagenesis (grain growth/dissolution at free faces) and pressure solution at grain-grain contacts;

  • organic matter decomposition and resulting petroleum generation;

  • multiphase fluid flow and reaction;

  • poroelastic and irreversible continuous deformation;

  • fracture (3-D, oriented) nucleation, growth and healing dynamics; and

  • heat flow.

The equations solved for the descriptive variables (over 100 in every element) account for these processes in their 3-D implementation. The descriptive variables include grain size, shape and packing geometry parameters of each mineral; fracture length, aperture, and orientation; complete 3-D stress and strain rate tensors, rock deformation velocities as well as rock viscosities; saturations, wetting and composition of each fluid phase; fluid composition; and temperature. These characteristics are dynamic, i.e., they change during basin evolution.

The model incorporates geologic data on the present-day basin. With this basin data, it automatically constructs the basin's unique history of sedimentation and erosion, tectonics (compression/extension and uplift/subsidence) and basement heat flux. Specific data used include surface geology, well logs and seismic information. During the simulation, these histories are used to construct the boundary conditions needed to solve the aforementioned conservation of mass, energy and momentum equations.

The current version of the computer simulator Basin.RTM incorporates all of the above features. We believe it is the first of a new generation of advanced, fully integrated basin models. The model has been tested on a number of extremely different basin systems, from the young siliciclastic (Piceance Basin, Colorado) to older carbonate (Permian Basin, West Texas) systems at various scales and resolutions. In addition, the model applies to salt-tectonic regimes (as in the U.S. Gulf Coast). In the following, highlights from some of these applications are shown, and they and other examples will be the basis of our presentation.

FRACTURE NETWORK DYNAMICS

Production from tight reservoirs, such as those commonly encountered in Rocky Mountain Basins, is typically dependent on the presence of natural fractures. The attainment of accurate prediction of fracturing in such systems is extremely challenging as the industry requires an evaluation of a number of factors, including the present-day fracture network characteristics and spatial location; the natural state of stress; and the state of overpressuring and characteristics of fluid phases. Other aspects that are important for assessing petroleum reserves, however, require knowledge of events that occurred in the past; these include maturation and fracturing timing and paleo-flow paths.

Figure 1 shows results from a Piceance Basin simulation of an area 50 times 48 km encompassing the MWX site. Shown are fracture permeability and overpressure at 45 MA. The isosurface of fracture permeability of minus6.52 (base-10 log of darcy) indicates regions of appreciable permeability. The complex geometric details reflect the interaction among lithologic control, compaction, overpressuring, gas-generation, and fraturing histories. The strong control on these factors by sedimentation/erosion, subsidence/uplift is accounted for in this simulation.

Figure 1.

SALT TECTONICS

Salt tectonics depends on the mechanical properties of the nearby sediments, deeper-originating faulting, the geometry of the lithologies, thermal anomalies and the distribution of fluid pressure (by means of effective stress). These factors and the interplay of salt movement, sediment surface topography and the spatial distribution of sedimentation rate affect salt body and sediment (and, thus, reservoir) location and geometry. Our model integrates these factors to yield results as in Figure 2, where alternating layers of sand (shaded) and shale (unshaded) sediments are deposited above salt body (shaded). The system is 5 km wide and 4.4 km thick at 4 MY. Note that the salt diapirism affects the location and shape of the sand bodies, and also creates pockets of fracturing and preserved porosity/permeability that affect migration patterns of oil generated from shale source rocks.

Figure 2.

OTHER EXAMPLES

Other simulations using the simulator Basin.RTM show a richess of dynamic petroleum system phenomena, including episodic cycling of fracturing-fluid release healing, and formation of laterally periodically distributed vertical fracture zones in basins with complex sedimentation and tectonic histories. These examples demonstrate that 3-D integrated RTM models can be used to answer many of the questions that arise in basin exploration and production analysis.

Better Basin Modeling Through 3-D Visualization

Marek Kacewicz and Wenlong Xu

Unocal Corporation, 14141 Southwest Freeway, Sugar Land, TX 77479

Basin modeling requires data such as well logs, depth structure maps, paleobathymetry, sea level curves, temperature, vitrinite reflectance, biomarkers, seismic attributes, sequence stratigraphic interpretations of depositional systems, etc. In dealing with large data sets (thousands of wells, hundreds of oil samples or a 3-D survey), the amount of gathered information is overwhelming and quick data integration and analysis are becoming an issue. Usually, available data include a combination of old and new data sets. This may cause many quality-related problems, and may result in questionable conclusions, if work is not done carefully. 3-D visualization provides a natural integration of different data types, the ability to render spatial relationships between key controlling factors and the associated uncertainty, a very efficient way of eliminating poor quality data, and an excellent way of communicating and discussing results with our clients (business units) and peers. In combination with other tools such as spatial statistics, numerical algorithms, etc., 3-D visualization helps us to understand relationships between different data sets and to monitor how key factors evolve through geologic time. This presentation will demonstrate how 3-D visualization in combination with basin modeling helped us understand regional as well as prospect-scale geologic, geophysical, and engineering data, to build and calibrate an integrated basin model, and highlight prospective areas that otherwise could be missed (Figure 1).

Figure 1. Rendering of regional and prospect-scale data provides a powerful tool for quick data analysis, calibrating basin models, and testing multiple exploration scenarios.

3-D Temperature Modeling of the Permo-Carboniferous Saar Basin (Germany)

M. Hertle1, R. Littke1, T. Hantschel2, and D. H. Welte2

1Aachen University of Technology, Germany; 2Integrated Exploration Systems, Germany

The coal-bearing Saar-Nahe Basin is an intramontane sedimentary basin in the internal zone of the Middle European Variscan orogen. The SW-NE trending basin is mostly covered discordantly with Mesozoic strata. At least 8500 m of fluviolacustrine and alluvial sediments was accumulated from Late Carboniferous to Lower Permian time. The time of maximum burial was reached in Latest Carboniferous/Lower Permian.

The Saar Basin, as part of the bigger Saar-Nahe Basin, provides an excellent initial case study for our modeling approach, because a wealth of geological information exists due to long-time subsurface mining activity. From wells, coal mines and seismic data the structure of the study area is well known. Vitrinite reflectance values from coals, present-day temperature and heat flow data were used to calibrate the model. Apatite fission track analysis gave information about the Mesozoic reburial/reheating of the sediments. Thermal conductivity measurements for different lithologies were performed ranging between 0.2 for coal and 5.7 W/(m K) for conglomerate and were integrated into the temperature modeling. Average thermal conductivities of the strata were calculated taking into account the detailed interlayering pattern of coals, shales, siltstones, sandstones, conglomerates and carbonates.

With the newly developed PetroMod3-D software the calculation of the (three-dimensional) temperature field of sedimentary basins through time is possible. The model is based on a regular 3-D finite element grid with a size of about 200 times 200 grid points in the horizontal plane that contains input data and simulation results.

Recent thermal maturity of the sediments can only be explained by deep burial and moderate palaeo-heat flows during time of maximum burial, i.e. in the Permo-Carboniferous. In the Permian about 2000 to 3000 m of Permo-Carboniferous sediments was eroded. The 3-D-model shows the different sedimentation and erosion histories of, e.g., anticlines and synclines. Palaeo-heat flow values for the time of maximum burial range between 55 and 75 mW/m2 (Figure 1).

Figure 1. Location of the study area and geological map.

A Topologically Based Framework for Simulating Complex Geological Processes

Ulisses T. Mello1 and Paulo R. Cavalcanti2,3

1IBM T. J. Watson Research Center, Yorktown Heights, NY, U.S.A.; 2Federal University of Rio de Janeiro, Rio de Janeiro, RJ, Brazil; 3Present address: IBM T. J. Watson Research Center, Yorktown Heights, NY, U.S.A

Three-dimensional (3-D) Basin Modeling requires the best description of subsurface geological structures, or an earth model, as the data permit. The generation of numerical meshes derived from an earth model is a necessary critical step to provide specific data representations such as finite difference grids or finite element meshes for the solution of the partial differential equations governing heat and fluid transfer within heterogeneous porous media. These meshes must include the basin geometry and associated physical properties along the evolution of sedimentary basins. The generation and maintenance of numerical meshes can be quite complex for basins that have undergone extensive changes in geometry as a result of compaction, diapirism and fault motion.

The construction of basin, or earth, models has been a challenge not only due to the scarcity of subsurface data but also because of the geometrical complexity of geological structures. Since typical geological structures (eg., fractures, faults, salt domes) evolve with time due to various physical process, we use a topological framework to create and maintain earth models that allow the geometry to be modified during simulations and with new interpretations. Here, we briefly describe the framework used to create a reference model defining a spatial partition, which is used to represent multi-material objects. Multi-resolution and multi-structure meshes can be associated as attributes to each cell of the reference model, making it possible to have a flexible mesh management environment for numerical simulations. Note that this framework provides critical adjacency information for the spatial domain decomposition used in parallel computing.

The major sources of data on subsurface comes from wells and from interpretation seismic data. In general, the interpreted data is very patchy in nature and provides a limited amount of information for building the earth-model "puzzle." Because of the unique nature of geological data, commercial CAD systems generally designed for mechanical or architectural applications are not appropriate to build 3-D earth models. In addition, geological surfaces are frequently inconsistent with each other due to the uncertainties in their generation. Hence, it is frequently necessary to use pre- and post-processing tools to correct inconsistencies in order to produce a coherent 3-D earth model. Various commercially available geological interpretation systems do not enforce model consistency because they regard earth models as just a collection of 2-D surfaces embedded in 3-D space, which are used to generate 3-D raster earth models when a voxel representation is required.

Geological "CAD" systems such as gOcad and Pyramid (Wiggins et al., 1993) have been designed to build truly 3-D earth models. However, the design of these systems was also oriented toward specific tasks in the industry. These systems are appropriate for building static earth models, but this is a limitation if one wants to use these systems to model dynamic geological structures that evolve through time.

We have designed a framework to build dynamic 3-D earth models. The main characteristic of this framework is that the model topology and geometry are kept separated. This characteristic is very important because many rock deformations are topologically invariant in time and can be described by geometrical changes. Topological changes occur in few distinct events, and Mello and Henderson (1997) have described techniques to minimize significantly topological changes during simulations.

TOPOLOGICALLY BASED FRAMEWORK

The high-level architecture of our modeling framework is shown in Figure 1. This is a layered architectural software pattern in which each layer has a distinct role in the framework. In the base of the framework, we use a topological representation based on the Radial Edge Data Structure — REDS (Weiler, 1988), which is used to represent complex non-manifold topologies.

Figure 1. Architecture of our modeling framework.

We implemented the REDS and its topological operators using C++, and this implementation is very compact, having less than C++ 50 classes. The REDS is the component that stores the topological and geometrical representation of an earth model. Geological attributes and high-level operators have been designed to mimic geological processes and to process input surfaces defining geological objects. The MultiMesh Toolkit — MMT (Figure 2) is the layer that generates and manages numerical meshes associated with earth model sub-regions. It is important to note that meshes are treated as attributes of geological entities such as blocks, horizons, layers and faults. Hence, a mesh is not the model, but only one possible realization of a model or a sub-region of the model. This paradigm is very powerful for simulation of evolving processes where the geometry is changing with time and therefore some degree of re-meshing is frequently necessary. In addition, this paradigm can also be used to integrate legacy simulation codes that use distinct numerical techniques. Using MMT, the meshing operators can provide multiple mesh representations with multiple resolutions of a given earth model. These meshing operators are used to re-mesh specific regions of the model that have undergone excessive deformation and to transfer information among meshes. One particularly important application of these operators is the upscaling operation, in which it is necessary to downsize meshes to a resolution such that the flow simulation can be executed on available computers. Operations between coarse and fine resolution grids are greatly facilitated within this framework.

Figure 2. MultiMesh toolkit concepts. See text for details.

BUILDING AN EARTH MODEL

A typical object-based earth model is created from horizon and fault surfaces that define the geometry of a basin. These surfaces are usually supplied by a set of planar polygons or regular 2-D grids. The input polygonal surfaces are read, and each of its defining polygons is inserted incrementally as described in Cavalcanti et al. (1997). The process of incrementally adding polygons defines the spatial partition of the earth model and generates the set of polygonal faces bounding a geological object. To improve the execution time, we use a dynamic index structure for the spatial localization of the face set in the model. To keep track of the face-surface relationship, each face has an attribute index that points to its respective entry in the model attribute table, which contains all the surface attributes such as type, name, age and so forth.

MESHING OPERATORS

Meshing operators are used to create an alternative realization of the model. These operators are used for numerical solution of partial differential equations and for interpolation purposes. Meshing operators are hierarchical in nature, and most of the meshing operations start on cells of lower dimension (1-D), moving to higher dimensions (3-D) hierarchically. The higher dimension meshes are constrained by the lower dimensional ones previously generated. The most important meshing operators we implemented are: regular, rectilinear, triangular, tetrahedral, and curvelinear (partially).

APPLICATION: AN EVOLVING SALT DOME

In Figure 3, one application of our modeling framework is shown. In this application, we calculate the temperature and pressure field surrounding an evolving salt dome using finite element methods. Note the deflection of the fluid path around the low-permeability salt and thereby moving the fluid away from reaching the reservoir (on the top).

Figure 3. (a) and (b) Two stages of a salt dome evolution and (c) the pressure field and associated streaklines of fluid flow surrounding the dome.

REFERENCES CITED

Cavalcanti, P. R., P. C. P. Carvalho, and L. F. Martha, 1997, Non-manifold modeling: An approach based on spatial subdivision: Computer-Aided Design, v. 28, p. 209–220.

Mello, U. T., and M. E. Henderson, 1997, Techniques for including large deformation associated with salt and non-vertical fault motion in basin modeling: Marine and Petroleum Geology, v. 14 no. 5, p. 551–564.

Weiler, K. J., 1988, The radial edge structure: A topological representation for non-manifold geometric boundary modeling, in M. J. Wozny, J. L. McLaughlin, and J. L. Encarnaccedilatildeo, eds., Geometric modeling for CAD Applications: Holland, Elsevier Science Publishers, p. 3–36.

Wiggins, W., U. Albertin, and G. Stankovic, 1993, Building 3-D depth-migration velocity models with topological objects: Society of Exploration Geophysicists Annual Meeting Technical Program, v. 63, p. 170–173.

Control and Uncertainty of Multi-Dimensional Models

Reneacute O. Thomsen

Department of Geology amp Geophysics, Texas AampM University, College Station, TX 77843

Complex dynamic models have to be constrained by control data. Control data can be point data, line data or surface data. Through the process of calibration, specific model parameters are changed systematically to obtain minimal discrepancy between observed control data and model predictions. Often, model results are highly non-unique and it is important to obtain a measure of the non-uniqueness—or uncertainty—of the modeled parameters. A measure of the uncertainty can be obtained by evaluating the resolution of model parameters, such as constants that control compaction, in the set of control data available. As the dimensionality increases, this procedure becomes non-trivial and in order to maintain a certain level of confidence the requirements with respect to control data change. Examples of model variability within the limits of resolution of control data will be given for 1-D, 2-D and 3-D models with a discussion of model constraints, elements adding to the uncertainty, and procedures for evaluating requirements for constraining multi-dimensional models and assessing model uncertainty.

Constraining the Gamble — Stochastic Techniques and Quantifying Uncertainties in Basin Modeling

Christian Zwach and Arnd Wilhelms

Norsk Hydro ASA, E amp P, P.O.Box 200, N-1321 Stabekk, Norway

INTRODUCTION

Basin modeling is presently applied routinely in the petroleum industry and academic research to study geological processes on a quantitative basis. In hydrocarbon prospect evaluation this may lead to the identification of critical geological factors, thus giving important contributions to the pre-drilling risk assessment of identified traps. However, the uncertainty of basin modeling results is not very frequently addressed, as it is generally difficult to quantify. This paper introduces a new approach to quantify uncertainties of hydrocarbon expulsion from mature source rocks and discusses general application of stochastic techniques to basin modeling.

THE DETERMINISTIC DILEMMA

During the basin modeling work process geological factors (input parameter, e.g. heat flow) are typically studied by users more or less systematically in order to define the sensitivity on the results (output parameter, e.g. expelled hydrocarbons). However, many of these input parameters are difficult or impossible to predict beyond a certain spatial and temporal resolution due to the nature's complexity and the insufficiency of calibration data (especially for the geologic past, e.g. paleopressure). For example, present day vitrinite reflection and temperature data may be explained by a large number of different heat flow histories taking the uncertainty of the calibration data into account (Nielsen 1993).

Ideally, those geological factors have to be systematically studied over their whole probability range in order to reveal a quantified uncertainty of basin modeling results. Such study should include the simultaneous (random) variation of all important input parameters in each run. This procedure is thought to be essential to reveal a fair quantified uncertainty picture, and it is generally not sufficient to vary one input parameter exclusively while all others are fixed. Taking these considerations into account, this leads to a complicated and time-consuming work process applying deterministic programs. How do we then calculate the result of a certain probability ("worst case," "best case," P10, P50, and P90)? Are stochastic techniques in basin modeling helping us to quantify result uncertainties?

THE STOCHASTIC PARADISE?

Stochastic modeling techniques offer an attractive and still not widely applied alternative to honor uncertainties during basin modeling calculations. The most widespread established technique is Monte Carlo simulation to vary the input parameters for a number of calculations ("realizations") within the defined probability distributions. Recently, a number of publications addressed the usage of Monte Carlo simulation techniques (e.g. Nielsen 1993, Van Laer et al. 1998, Zwach and Carrathers 1998) in basin modeling.

However, stochastic modeling relies very much on the speed of the algorithm due to the number of realizations needed to represent sufficiently the probability range of each input parameter. This is obviously a challenge for future development of new basin modeling software.

A new fast stochastic approach ("QuickVol3-D") has been developed to calculate expulsion volumes from mature source rocks in drainage areas. QuickVol3-D is integrated with standard basin modeling software and relies on a minimum of input data. Thus, it is well suited to standard exploration situations in the petroleum industry. The hydrocarbon expulsion is simulated based on the source rock retardation concept (e.g. Pepper and Corvi 1995), and present-day depth grids of the source rock, maturity, initial TOC, kinetic model of hydrocarbon generation, and initial hydrogen index are required as input data. Obviously, geological factors such as initial source rock properties are poorly known in a prospective drainage area during exploration. A probability range can be assigned in the program to each of the input parameters to account for uncertainties.

During the simulation all defined input, parameters are varied between each realization according to their probability distribution and interrelations. The uncertainty of the hydrocarbon expulsion mount is finally quantified: the P10, P50, and P90 of the expelled mounts of hydrocarbons from defined source rocks in a drainage area are calculated after the total number of realizations. Single random expulsion grids (realizations) may later be used for stochastic migration simulations to assess the risk of traps being filled with the expelled hydrocarbons (Zwach and Carruthers 1998).

Stochastic techniques in basin modeling offer a great potential to quantify basin modeling uncertainties, a potential that has been largely unused in basin modeling software development at present day.

REFERENCES CITED

Nielsen, S. B., 1993, Fast Monte Carlo simulation of hydrocarbon generation, in A. G. Dor6 et al., eds., NPF Special Publication No. 3: p. 265–276.

Pepper, A. S., and P. J. Corvi, 1995, Simple kinetic models of petroleum formation. Part III: Modeling an open system: Marine and Petroleum Geology, v. 12, no. 4, p. 417–452.

Van Laer, P., J. Toth, and M. Giles, 1998, 4D basin modeling: Quantitative prospect appraisal: Abstract, IFE Symposium on Advances in Understanding and Modeling of Hydrocarbon Migration, Oslo, Norway, 7.–8.12.1998.

Zwach, C., and D. Carruthers, 1998, Honouring uncertainties in the modeling of migration volumes and trajectories: Abstract, IFE Symposium on Advances in Understanding and Modeling of Hydrocarbon Migration, Oslo, Norway, 7.– 8.12.1998.

Multiscale Petroleum Migration and Invasion Modeling: A Pilot Study

Philip Ringrose1, Eirik Vik1, and Dan Carruthers2

1Statoil's Research Centre, Postuttak, Trondheim 7005, Norway; 2Permedia Research, 53 Fourth Avenue, Ottawa, Ontario, Canada

Petroleum distribution within a mid-Norway offshore gas-condensate reservoir is variable. Biomarker analyses indicate multiple petroleum sources, some of which are difficult to reconcile with the distributions of fluid in place. Reservoir data show quite variable hydrocarbon/water contacts and wetting properties, which are related in some way to the reservoir filling history.

In this pilot study, we have used a recently developed invasion percolation (IP) oil migration tool to test several alternative field-filling scenarios. Field-filling scenarios are defined on a 2-D regional cross-section, with stratigraphic and petrophysical properties defined from seismic and available well data. Alternative geological scenarios (different fault, sand and shale properties) and petroleum charge models (rates, locations) are tested against resulting reservoir fill patterns. Results show how the locations of reservoir petroleum fluids and their invasion time sequence depend on the assumed sources, their timing, and the geological model. A subset of plausible models can be identified which helps us to select the more likely field-filling models. Charging from a lower source rock leads to a bottom-up reservoir filling sequence (Figure 1a) whereas charging from an upper source rock leads to a top-down reservoir filling pattern (Figure 1b). Reservoir fluid distributions appear to require contributions from both sources.

Figure 1. Examples of field-filling models for the example prospect (section is 55 km long by 2 km deep; grey tones indicate stratigraphy, dark tones show oil migration pathways and accumulations): A. Feeding only from lower source leading to a bottom-up filling sequence. B. Additional later contribution from an upper source-rock horizon leading to a mixture of bottom-up and top-down reservoir filling sequences.

The main recovery mechanism for the field is by gas injection. We have used the same modeling approach (IP) to look at simple gas-injection patterns (non-wetting gas invading an in situ wetting phase) in reservoir models at the lithofacies and formation scale. The IP model allows us to look at fluid invasion patterns in detailed models of reservoir heterogeneity and to quantify the likely distribution of flow units invaded by injected gas. These gas invasion models (simplified fluid physics/high resolution) are compared with more conventional reservoir models (Eclipse Black-Oil model) of the gas injection process (better fluid physics/lower resolution) to test and compare the approaches.

The study illustrates the advantages of explicitly modeling fluid invasion patterns using the fast IP method to rapidly test field-filling and production scenarios, integrating exploration and production data in the field development phase.

Construction and Application of a Stratigraphic Inverse Model

Timothy A. Cross, Margaret A. Lessenger

Department of Geology and Geological Engineering, Colorado School of Mines, Golden, CO 80401

Stratigraphic inversion is a quantitative technique that extracts values of process parameters—such as tectonic movement, lithosphere strength, sea-level change, sediment supply and basin topography—from stratigraphic data. A stratigraphic inverse model contains: (1) a forward model which simulates stratigraphy through the operation of a set of input process parameters and algorithms which describe the behavior of the stratigraphic process-response system; (2) a set of observed data which are comparable in type and form to forward model predictions; and (3) a set of equations and algorithms that compares the values of forward model predictions with observations, and simultaneously adjusts values of all forward model parameters to create a better match between predictions and observations. The inverse model iteratively reduces differences between forward model predictions and observations until a best match is achieved. The model calculates the degree of accuracy and uncertainty of values of stratigraphic predictions. Constructing an inverse model requires the following steps: (1) selecting a stratigraphic forward model; (2) designing simple mathematical functions that most accurately describe the real stratigraphic processes that operated in a basin and that make inversion computationally possible; (3) measuring data that correspond in type to the output of that forward model and transcribing those data into the form of a numerical vector; (4) selecting an appropriate parameter optimization algorithm; (5) building a stratigraphic inverse model that connects components of steps 1 through 4. One purpose of stratigraphic inverse modeling is prediction of stratigraphic attributes (e.g., facies, geometry, distribution, volume) with calculated estimates of accuracy and uncertainty. Once the range of parameter values is calculated by the inverse model, a population of forward models may be run which should contain the true stratigraphy. The population of forward models is used to predict the geographic and stratigraphic positions and extent of potential reservoir, source and seal rocks. We show an example of accurate stratigraphic predictions using inverse modeling of the Mesa Verde Group, San Juan Basin, Colorado and New Mexico, U.S.A.

Uncertainties and Sensitivities in Petroleum System Analysis

Jianchang Liu and Phillip A. Levine

EPTD, Texaco, 3901 Briarpark, Houston, TX 77042

One of the major tasks of basin analysis is to search for potential hydrocarbon accumulation locations by modeling the history of a basin. Certain simplifications and assumptions in model construction and data sampling are necessary in order to simulate a basin's geological history, but inherent uncertainties and errors within modeling results are therefore unavoidable.

In addition to the numerical instabilities within the applied models and various inaccuracies of thermal and lithological data and parameters, paleo-topographical and geometric uncertainties are the other major sources of the modeling uncertainties. Any unrecognized paleo-bathymetry, undetected differential erosion and compaction rates, unbalanced faulting, or migration modeling in only one and two dimensions may well cause different migration and accumulation histories for a basin. Some of these factors may not be resolved in a frontier basin, while others remain undetermined due to the limitations of the applied models.

Based on petroleum system modeling using various scenarios, the uncertainty and sensitivity of simulation results on the above factors are investigated. It is suggested that migration modeling in an exploration stage tends to predict general migration trends and potential accumulation locations/directions rather than a detailed migration mechanism simply because too many uncertainties are involved.

The economic and functionality analysis of 1-D, 2-D and 3-D modeling in routine petroleum system studies also indicates that 3-D modeling is desirable because it can reduce some of the basic uncertainties introduced from either 2-D areal or 2-D cross sectional migration models. It is also justified to use a 2-D cross section model for investigating time relationships of different petroleum system components, a 2-D areal model for accumulation potential in a reservoir formation, and a 1-D model for maturity analysis.

Prospect Evaluation by Probabilistic Basin Modeling

A. Corradi, B. Maragna, D. Ponti, P. Ruffo and G. Spadini

ENI - AGIP Division, Via Emilia 1, 20097 San Donato Milanese MI, Italy

Basin modeling has been employed in the past with a strictly deterministic approach, one of the reasons being the high computational times required. However, many of the parameters involved in the numerical simulation of the processes that lead to hydrocarbon entrapment are often highly uncertain, and defining a range of possible values for these parameters is often more realistic than specifying exactly a single value. Although basin modeling programs are generally deterministic—i.e., they provide a unique set of output parameters, given a set of input parameters—they can be used with a probabilistic approach in a simple way, combining basin modeling with statistical techniques, like for instance Monte Carlo. A large number of (deterministic) simulations is performed, sampling the assigned probability distributions of the input data, and obtaining a probability distribution for the output data. A probabilistic approach to basin modeling can thus provide ranges of uncertainty for output data, as a result of the uncertainties in the input parameters. Obviously this can be done only if simulation times for a single run are not too high.

Unlike basin modeling, prospect evaluation has commonly used Monte Carlo techniques to predict the volume of hydrocarbons in place, assuming probability distributions for the volumetric input parameters. However, the hydrocarbon column, which is undoubtedly a critical factor in the final result, is often guessed by the explorationists without taking into account the geological and physical processes that can lead to the accumulation of hydrocarbons in a trap. Basin modeling can be of great value in giving such information, via quantitative estimates of generation, expulsion and migration of hydrocarbons, therefore enabling the explorationists to make better hypotheses about the volumes of oil and gas that can have migrated into a prospect.

The aim of our work is to develop a methodology, integrating basin modeling results in prospect evaluation, that can be applied to real cases by basin modelers, and possibly by explorationists. Therefore a compromise is to be found between the complexity of the processes to be simulated and the need to keep computational times within reasonable limits.

The first part of our workflow includes the calculation of temperature and its evolution through time, which is a fundamental factor in the process of generation of hydrocarbons, and the calculation of the amounts of hydrocarbons generated and expelled from the source rock. Expulsion maps for oil and gas in the area of interest during the evolution of the basin are obtained with two different approaches: a multi-1-D simulation and a full 3-D simulation.

A multi-1-D simulation is performed with ENI proprietary 1-D basin modeling tool (GET) to calculate pressure and temperature history, generation and expulsion in selected points of the basin. Geostatistical techniques are then applied to obtain expulsion maps all over the area from these values of expelled oil and gas. Alternatively, ENI proprietary 3-D basin modeling package (SEBE-3), a full 3-D program based on finite elements technique, is used to reconstruct the burial history of the basin and the evolution of pressure and temperature, finally providing expulsion maps for oil and gas at several timesteps.

For the simulation of secondary migration of hydrocarbons we have chosen a program based on the ray-tracing method (SEMI, developed by IKU-Sintef). Being computationally very efficient, the ray-tracing technique allows us to apply a probabilistic approach to secondary migration, which would be impossible with simulators based on other techniques. Having assigned probability distributions for critical input parameters, a Monte Carlo method is applied to generate input data for secondary migration simulations. The final output results are the probability distributions for oil and gas volumes in the traps that can be used directly in the prospect evaluation.

In basin modeling, sensitivity analysis is performed in order to have a better understanding of the influence of input parameters on the results and to identify the critical factors in the simulated processes. A common practice is that of performing sensitivity analysis varying one of the parameters within its range of uncertainty and keeping the others to a fixed value. A method of systematic search in the uncertainty hyperspace of the critical parameters can be applied to run a number of simulations whose input data uniformly cover the whole range of uncertainty for all the selected parameters.

A calibration of the model can be carried out with a combination of sensitivity analysis and Monte Carlo technique, comparing simulated oil and gas volumes in traps to the observed volumes. By so doing, the ranges of uncertainty of the input parameters are narrowed, thus providing constraints for the prediction of probability distributions of hydrocarbon volumes in undrilled traps.

From a geological point of view this probabilistic methodology seems very suitable to model petroleum systems that are usually difficult to define with simple, deterministic parameters. The use of input ranges of parameters instead of single values, often difficult to assess, and the large number of parameters that can be set provide a better link among conceptual geological models, physical reality and mathematical modeling, thus helping to perform more realistic prospect assessments.

The work done so far shows that the integration with statistical techniques is a very promising direction for further developments in basin modeling. The continuous improvement of hardware capacities will allow more and more processes to be simulated with a probabilistic approach and will enable basin modeling to be employed in prospect appraisal and risk assessment.

Determining Uncertainty and Sensitivity in Basin Modeling by Experimental Design and Response Surface Modeling Techniques

J. Wendebourg

Institut Franccedilais du Peacutetrole, Rueil-Malmaison, France

Basin models are deterministic and therefore modeling results are unique. However, exploration scenarios are uncertain and model results should therefore be expressed in a probabilistic manner. Such probability distributions may be used as a surrogate of the actual uncertainty because a basin model includes the physical principles and interdependencies of a petroleum system that can be calibrated to local geological data. Uncertainties arise when defining values of input parameters and depend on the modeling objective. A systematic sensitivity test of all input parameters or Monte Carlo sampling of input parameter distributions is not feasible because multi-dimensional basin models require generally many input parameters and, given the average run time of typical applications, may become prohibitively costly.

Experimental design techniques have been used originally in chemistry to quantify the outcome of complex reactions among several reactants with a minimum of experiments. Similarly, a basin model can be considered a complex reaction with many input parameters and a model run can be considered an experiment with a unique result. Multivariate statistical analysis allows us to quantify the relative importance of input parameters. A response surface can be fitted to the experiment outcomes that represents the complex model in a simplified form from which probability distributions of model results can be drawn. In order to avoid large sets of experiments, a number of input parameters assumed to be influential needs to be chosen a priori. The response surface may also be constrained by calibration data. Example applications using IFP's GENEX and TEMIS basin modeling codes will be shown to illustrate this approach.

Methodology to Handle Geologic Models in Reservoir Engineering

D. Gueacuterillot

Institut Francais du Petrole

Problems of great economical interest such as improving recovery rate of hydrocarbon reservoirs, optimizing gas storage fields or controlling underground pollution require a very detailed knowledge of porous media, and, in particular, of the spatial variations of their hydraulic properties. To describe the hydrocarbon reservoirs, geological models using statistical concepts are more and more used. They bring a new horizon for reservoir engineering studies. Because most of the current geological models are obtained from the knowledge of the geological environment and the interpretation of static data, this presentation summarize methods for constraining these models by dynamic data such as well test data and production data. Then, some problems raised by the introduction of these geological models are discussed: sorting images, averaging of petrophysical data, meshing of the reservoir, etc. When production data are available, the author proposes the "Scenario Matching" to estimate uncertainties in production forecasts.

The presentation will be composed of the following topics:

  • Methodology to Handle Geostatistical Models in Reservoir Engineering

  • Integration of Well Test Data in Geological Models

  • Optimal Meshing of Heterogeneous Geological Models

  • Upscaling of Petrophysical Parameters

  • Sorting Images

  • History Matching and Scenario Matching.

Scaling Reservoir Data: A Fact of Life

Philip Ringrose

Statoil's Research Centre, Postuttak, Trondheim 7005, Norway

Scaling reservoir data is a challenge which is often given only a passing treatment or even avoided completely. There remain both theoretical and practical problems with the derivation of "true" macroscopic properties for multiphase flow in heterogeneous systems. Common practice usually comprises making a set of assumptions (e.g., use the average and variance of the observed well data) and applying them directly to the reservoir model without checking the scale dependence of the property. Conclusions about the large-scale flow behaviour are consequently often made on a fairly weak foundation. In homogeneous rock systems this approach may be acceptable, but in most cases it is not.

This problem is illustrated with reference to detailed analyses of well data (mini-permeameter, core-plug and well-log) from recent Statoil studies of two offshore mid-Norway reservoirs (Figure 1a). In particular, the tidal sedimentary systems of the lower Jurassic Haltenbanken reservoirs present a difficult problem as they are very heterogeneous at the 25-cm scale, and have a significant and pervasive mud content. Between 30% and 50% of the permeability variance is found to occur at the sub-25 cm scale (Figure 1b). Using detailed bed-scale and facies-scale reservoir models, we have estimated the effective larger-scale flow properties for single-phase, and multiphase flow (water and gas injection) using steady-state methods (Figure 1c).

Figure 1. (A) Multi-modal permeability distribution of core plug and well log data for one reservoir unit. (B) Analysis of variance as a function of scale: log-normal distributions of core plug permeability data compared with 1 m and 5m running averages (which respectively retain 70% and 40% of the core plug sample variance). (C) Reservoir flooding pattern for a fully upscaled thin-bedded reservoir unit (well spacing is 2 km, reservoir unit is 25 m thick).

The findings are not simple: some perceived aspects of small-scale heterogeneity turn out not to be very significant, while other aspects have a large impact on the reservoir-scale behaviour. For example, in heterolithic facies vertical permeability drops rapidly at mud fractions of around 30% and higher mud fractions have very little effect on flow, while horizontal permeability may remain high even when mud fractions are as high as 80–90%. Horizontal barriers, mud-layering at the sub-metre scale and sand-facies connectivity at the 10m scale are found to be especially important factors. The general problem is the correct representation of rock connectivity within a discrete-grid flow simulator. The choice of parameters to represent pore-scale processes (especially the relative permeability end points) is consistently found to be a major uncertainty, ranking alongside the uncertainty associated with upscaling.

A pragmatic solution to this problem is the use of relatively simple flow models at a range of scales in order to identify the critical factors and to estimate the effective properties. This has the advantage of being explicit about "the facts of life" (significant parameter variability as a function of scale) and leads to important insights into reservoir flow behaviour. However, some of the basic problems with upscaling theory remain unsolved (e.g., how do you upscale when there is no separation of length scales?) and the flow property estimates using the pragmatic multi-step approach can be difficult to validate.

Pay-Per-View Purchase Options

The article is available through a document delivery service. Explain these Purchase Options.

Watermarked PDF Document: $14
Open PDF Document: $24