Sylta-2004
-
The aims of this study are to deduce improved methods for modeling the migration of oil and gas at geologic time scales, and develop methods that can be used to advance the incorporation of process based 3D basin scale hydrocarbon migration simulation techniques in exploration risk assessments. Critical analyses of existing methods and descriptions of new algorithms that can be used effectively to model the migration processes are included. The methods are based upon laboratory experiments, and validation of models use basin scale case studies and the matching of existing field data to support the conclusions.
The petroleum industry is increasingly relying on computerized methods to develop new oil and gas resources and exploit these resources. The use of seismic methods in combination with geological model building and fluid flow simulation techniques have allowed for significant increases in the amounts of oil and gas that can be produced from existing fields. During the last 10 years this increase in the producible oil and gas has contributed significantly to the ability to maintain a sufficient supply of petroleum for the continuously increasing demands of the world population. The increase in supply of petroleum has been one of the major requirements for the world economic growth, and a steady increase in the produced volumes is most likely needed to maintain the economic growth in the future.
The discovered hydrocarbon resources are a finite resource. Sooner or later each discovered field will be depleted, even when enhanced oil recovery (EOR) techniques are used. The EOR methods typically increase the ultimate recovery fraction of each oil field with in the order of 10-20%. Some fields, such as the Ekofisk Field in the North Sea, experience much higher EOR increases, and thus contribute to the reserves base for an extended period of time. But this can not change the overall trend of a limited potential for reserve growth due to EOR in existing fields.
During the last 10 years, the increased utilization of 3D seismic has helped in increasing the output from existing fields. The introduction of 3D seismic has, however, not been able to keep the exploration for oil and gas at a level where depleting reserves from the production of oil and gas can be replaced with resources from new discoveries. Therefore, most of the significant oil production increase in the 1990’s has been from EOR techniques in combination with faster depletion of existing fields. The latter can be seen in NPD statistics that show a larger number of producing wells than planned being completed in existing Norwegian fields.
The trend of insufficient exploration success and a lack of replacing existing resources through efficient oil and gas exploration programs need to be reversed in order to keep up with the long-term world consumption of petroleum. Many oil companies and nations also face a depleting resource base from which to provide future revenue. Norway is a typical example: If exploration success is not increased, Norway will deplete most of their oil reserves during the next 30-40 years (Figure 1), while gas can be produced over a period of at least 100 years. The oil production from existing fields is planned to be reduced to 20% of the present day peak production rates as early as 2025 (Figure 1). This is a dramatic decrease in production and revenues for Norway. NPD has, however, estimated that Norway’s reserve base can be increased significantly to delay the reduction in production (Stortingsmelding 38). Two development paths are outlined in Figure 2: “Forvitringsbanen” (“decline scenario”), which occurs if only the approved plans are followed, and a much more optimistic “langsiktig utviklingsbanen” (“long-term scenario”) which can be achieved if all the resources are exploited and produced.
Figure 1 Estimates of future Norwegian oil production (Berge, 2003).
The undiscovered resources of Norway are estimated to be 40% of the remaining resources. During the years from 1998 to 2003, NPD reports an average of 18 wells drilled per year. NPD therefore believes that it will be very difficult to find and produce these undiscovered resources with the expected discovery rates (Berge, 2003). A total of 30 wells will be required per year with the present day discovery rate, to find sufficient amounts of oil. It can become a challenge to achieve this level of exploration activity when the prospects become smaller.
The most striking feature of Figure 1 may be the growth of the uppermost green bar, representing the undiscovered resources, after 2020. In the optimistic scenario, 50% of the Norwegian petroleum production will be from fields that have not yet been discovered! And these fields are not likely to be found with the exploration activity that is now taking place! Clearly, a dramatic increase in the exploration success is required for Norway to achieve their goals in future petroleum production.
Dramatic increases in exploration success can be achieved by tax incentives, geological break-through, or technology development. Reducing taxes could obviously increase the number of wells to the 30 wells NPD considers necessary to find sufficient amounts of undiscovered resources. A geological break-through could be that an oil-play was discovered in the Vøring basin, or that the key to the hydrocarbons in the Barents Sea were unlocked. Many oil companies may want to rely on this approach, but it is risky, in that the probability of success of any of these two events occurring is perhaps less than 50%. In fact, our estimates of the probabilities are highly subjective, and not published outside the risking teams of oil companies.
Figure 2 Prognoses of Norwegian petroleum production: minimum and maximum scenario (Berge, 2003).
Technology development can provide an important contribution to achieve a sufficient success in exploration activities. Much of the costs involved in petroleum exploration are related to the process of drilling wells. Making each well less expensive, or reducing the number of exploration wells that it takes to find and delineate a new field could therefore significantly improve the economics of exploration.
Exploration wells are drilled to identify and describe new oil and gas fields from prospects that have been defined using well log or seismic data. The main purpose of exploration drilling is to confirm or reject a geologic model of the subsurface. Almost all exploration wells can be described as expenses for the oil companies. Only in rare cases where exploration wells are converted to production wells, will there be any income directly resulting from an exploration well. Therefore, there is no economic downside to reducing the number of exploration wells, or eliminating them altogether, if the confirmation of the oil and gas fields could be achieved using other methods.
The implementation of new methods and techniques that have the potential for increasing the success rate of exploration wells is a natural first step for increasing the exploration efficiency. The worldwide exploration success ratio (the probability of finding oil or gas in an exploration well) is around 30%, and the Norwegian success ratio also has hovered around this value during the 1990’s. An important aspect of this low probability value is that for each new exploration well, the most likely outcome will be that it is dry. This influences the planning of the strategy around the drilling of each exploration well. There is little use in making extensive plans for a positive outcome (a discovery) before the well has been drilled. Therefore each new exploration well tends to be followed by a period of analysis and further geological and/or geophysical modeling and assessment. This period can easily result in a year of no apparent activity within a licensed area. The result is that it takes many years to develop a license into a discovery and later a field for production.
If the methods that were applied in the exploration of oil and gas could increase the drilling success ratio to, say, 80%, then the entire setting of the exploration activity could be modified. The most likely outcome would then be a confirmation of the geological model, and the exploration teams could make more extensive and “firm” plans for the activities after each exploration well at an earlier stage in the ownership of a license. Still, contingency plans for the case of dry wells would have to be made, but in most cases one could speed up the entire exploration phase. The result would be a much faster and more economic exploration for oil and gas, in particular in offshore areas such as the North Sea, where drilling activities are expensive, and delays between the exploration (cost phase) and production (income phase) are costly.
The last 10 years have documented that the exploration success ratio can not be increased significantly by the use of 3D seismic alone. Seismic quality improvements during the last 10 years seem to have been instrumental in not decreasing the exploration success as prospects become harder to find, but not in increasing the success rate. This development should perhaps not come as a major surprise. Seismic reflections are influenced by the fluid and vapour properties of the media that the signal traverse, but the effects of the fluid and vapour are usually only minor compared to the effects of changes in lithology. Only in cases of gas-filled reservoirs can one expect to see clear effects with a fair degree of certainty. Even in these cases, a dry reservoir that resulted from an old gas accumulation that later leaked out may give almost the same seismic response as an existing gas accumulation.
Using seismic hydrocarbon indicators alone as decision criteria for drilling exploration wells can be a quite expensive methodology. If a model is built from e.g. seismic amplitude changes above crests of structures that may contain hydrocarbons, the first mistake (dry well) will not typically be sufficient to disprove the assumptions. Most likely the seismic indicators will have worked in another area, and therefore the a-priori mode of operation will be to “believe” in the seismic indicators. Only after the second dry well is there sufficient information to seriously doubt the seismic information. Still, one will often find reasons why it did not work for one or both of the wells, and argue for a third exploration well. An oil company can therefore easily end up drilling 3 dry wells based on seismic information alone before the concept is downgraded as not working well in the area. This would then constitute a cost of up to $100 million (700 MNOK) in areas such as the North Sea.
Processed based modeling of geological processes can be used as fairly independent decision criteria for exploration campaigns. Most companies use thermal modeling and source rock generation modeling to demonstrate that prospects or plays have the potential for receiving sufficient quantities of oil and/or gas before they drill. This is not sufficient to increase the success ratio from their present level. Very few, if any, oil companies have so far tried successfully to systematically describe the migration into and leakage out of all prospects that they consider drilling using quantitative process based methods in three dimensions. The proper 3D mass-balance modeling of all fluids that are part of geological systems that contain prospects must therefore still be considered a new and untested technique in petroleum exploration.
There are many reasons why well calibrated 3D fluid flow models of the geological system are not extensively used in the exploration process. Some of these are:
Lack of data of sufficient quality to construct a 3D geological model.
Insufficient time to make all the required interpretations.
The method has not been proven to add value to the exploration process.
Assumed costs of the work involved.
Lack of experienced personnel.
Existing tools are extremely demanding to use.
Parameters needed by existing models are not well constrained yet.
Lack of long-term planning of exploration activities.
Uncertainty as to how to use results of the analysis within the exploration workflow.
Existing commercial and research basin scale simulators aim to model subsidence and burial of the relevant sediments during geological time. The thermal evolution of source rocks, reservoirs and seals are computed using typically a heat flow history, surface temperatures and conductive and sometimes convective heat transport (Hermanrud, 1993). From the reconstructed temperature history, kinetic models are used to simulate the generation of oil and gas component groups through geologic time. Primary migration of hydrocarbons is modelled from the source to the nearest permeable carrier rock using Darcy flow. Secondary migration can be modelled using Darcy flow, percolation or ray-tracing methods, while attempts at modeling the capillary leakage out of traps use Darcy flow or percolation methods.
The most extensive simulators will take days or weeks to complete a full 3D basin scale simulation of all the above processes. The parameters of existing models are not sufficiently constrained to allow for the direct modeling of the hydrocarbon phases in existing fields. Therefore extensive calibration of the basin model is almost always required to find good matches to the observed data. Observed data include pressures and temperatures in drilled wells, maturity measurements from cores and cuttings, and oil and gas saturations/contacts in drilled wells. The long simulation times have so far prevented the probabilistic use of basin scale simulators. Most uses of these techniques have been in deterministic modeling studies. In deterministic modeling, results from one or very few simulations runs are used as best case scenarios for the geological and economic risk assessments. Alternative cases are then not fully explored, and it is easy to forget (miss) the very low and high side potentials of a prospect. In other words, the risk of not including the correct geological scenario in the analysis increases with a deterministic approach to modeling.
The full potential of using 3D basin scale hydrocarbon generation and migration simulators in combination with modern seismic data can only be achieved if the algorithms incorporated into the simulators give a fair to excellent description of the processes involved. The algorithms used have to be sufficiently fast to allow for multiple realizations over hours or days. This is critical to allow for accurate and rapid risk assessment of prospects during the short decision times allowed for exploration work.
The Darcy equation describes the relationship between fluid flow, permeability and pressures (potential differences) in porous media (discussed in e.g. Bear, 1972). Multi-phase flow can be described by including relative permeability and capillary pressures in the fluid flow equation(s). Basin modeling started off by adopting the methods that had been developed at the reservoir scale for basin scale simulations. The computing nodes were increased from e.g. 200*200*10m used in reservoir simulation to 1km*1km*100m for typical basin scale simulations (Johannesen et al. 2002). The size of the computing node was thus increased 250 times in order to keep simulations run times acceptable, with each simulation run taking a few days or weeks to complete.
The adoption of the reservoir simulation methodology, which often assumes constant hydrocarbon saturations within each computing node for each time step modelled, can be considered quite risky for simulating migration at basin scales. Selle et al. (1993 - Chapter 3), Sylta et al. (1998 - Chapter 4) and Sylta (2002 - Chapter 5) have shown that oil and gas migration in permeable carrier rocks occurs as focused, low saturation migration stringers. The vertical thickness of secondary migration stringers are typically less than 1m (Figure 3), and migration velocities can exceed 1000 km/My.
The hydrocarbon saturation varies vertically within even thin stringers (Figure 3), from zero at the base of the stringer, to typically 10-20% at the top of the stringer (Sylta et al., 1998 - Chapter 4). The average relative permeability within the migration stringer will be very low because of the low saturations, but the velocities of the oil and gas that migrate within the stringer are very high compared to the long geologic time periods modelled. One way of speeding up simulations can therefore be to assume that the hydrocarbons can migrate from the source rock to the traps within each time step used in the modeling of the process (Sylta, 1993 - Chapter 6). Time step lengths of basin scale fluid flow simulators will typically be in the range 0.1 Ma and with migration velocities approaching 1000 km /My. This means that oil and gas can migrate distances that are longer than 100 km within a single time step in a basin scale fluid flow simulator. The ray-tracing method (Sylta, 1987, 1993 - Chapter 6) and the percolation method (Carruthers, 1996) assume such high hydrocarbon migration rates, and these process descriptions therefore make use of the above assumption (and others) to achieve simulation times for large datasets that do not exceed a few minutes, at least for simple (fluid phase) conditions.
Figure 3 Cross-section outlining process of secondary migration with inserted colour-coded oil saturation distribution within carrier (modified from Sylta et al., 1998 and Selle et al., 1993).
In contrast, the classical Darcy basin scale fluid flow simulators will take days to run full 3D geologic models. Sylta (Chapter 10) discusses how this method can cause hydrocarbon saturations to be modelled to be too high, velocities too low, and therefore computed hydrocarbon losses to be too large. In an example taken from the literature, Sylta (Chapter 10) shows that oil volumes corresponding to a giant oil field are modelled to be lost in a spill route within the Tampen Spur area.
There are several possible approaches that can be used to modify the classical basin modelling (CBM) methods so that the hydrocarbon flow properties are treated more correct (Chapter 10). One way would be to use up-scaled permeability and capillary functions; another would be to use hybrid solutions for hydrocarbon migration within the carrier system. Hybrid systems (Hantschel et al., 2003) can make use of ray-tracing or percolation techniques (Carruthers, 1996) to describe the migration within the carrier system, while the normal Darcy approach would handle the rest of the geological system (Hantschel et al.,, 2003).
Faults are important features that need to be accounted for in many geological systems, but are frequently ignored or treated very simplistic in CBM. Faults can act as barriers to flow or be “open” to flow at geologic scales. Faults may act as barriers at production scales (years), while being completely open to flow at geologic time scales. Here we only discuss the basin scale processes. During the 1980s, the modeling capabilities with respect to faults as barriers to hydrocarbon migration at basin scales were quite limited. Sylta (1987) introduced sealing faults and geometric (juxtaposition) faults into a ray-tracing approach. Later, attempts were made to use entry pressures and fault transmissibilities to describe flow properties (Sylta, 1996). Flow transmissibility multipliers have been shown to be useful for simulating flow at the reservoir (time) scale (Manzoochi et al., 1999).
The experience with using fault transmissibility formulation in hydrocarbon migration simulations led to a conclusion that there is insufficient data to constrain these fault properties for basin scale simulations at the present time. There is also quite some uncertainty as to whether fault permeability is an important parameter for basin scale hydrocarbon migration. The supply rate of oil and gas from the source rocks is generally considered very low compared to the flow that a low-permeability fault can accommodate over long geologic time spans. As a first order approach, entry pressures are considered the more important process property to describe for oil and gas migration modelling at basin scales. The formulation of Childs et al. (2002 - Chapter 13) was therefore used to describe the influence of shale-smear gouge (SGR) in faults on hydrocarbon migration and leakage out of traps that have a component of fault seal.
Early case studies involving fault seal simulations focused on developing scenarios that could be used in the risking process. Each scenario involved setting a number of fault segments to various combinations of open or closed. The simulation results were used to assess into which prospects oil and/or gas migrated and when. From this information a geologic risk profile could be developed for each prospect (Schroeder and Sylta, 1993 - Chapter 16). One limitation of the open/barrier/juxtaposition approach to fault seal modeling is that the outcomes are very much either/or. The result is often that either a trap is filled with hydrocarbons or it is dry. Monte Carlo simulation approaches are not so easy to apply to this fault seal definition, because each scenario has to be defined manually by geologist(s).
In contrast, the clay-smear modeling (Childs et al., 2002 - Chapter 13) can be made to behave more smoothly, through the setting of properties in the entry pressures relationship, Pe(SGR). Faults can therefore be modelled to change from sealing, to partly sealing to completely sealing by changing fault properties. These properties may also be obtained from laboratory investigations of core samples. Sperrevik et al. (2002) showed how a database of entry pressure and permeability could be developed and a resulting Pe(SGR) formula was defined. It is still not certain that this kind of formulation will apply for entire faults (Childs, pers. comm., 2003). Until the Sperrevik et al. (2002) type of formulation is proven effective, many geologists will prefer a cut-off type formulation, where faults are open to flow when the computed SGR is lower than a limit value, and closed to flow when the SGR value is greater than this limit.
There is only one way to resolve these issues, and that is to apply the formulations to many case studies. Eventually one or the other (or both) formulation will fail, and a proper description can be deduced. Unfortunately, this process will take at least 5 to 10 years to complete, and we are therefore unable to conclude with respect to this problem in this work. At the very least, the incorporation of the various fault seal descriptions into a simulator (Childs et al., 2002, Sylta et al., 2003) has made it feasibly to resolve these issues in the future.
Vertical migration in mud-rocks has been modelled in both 2D (Ungerer et al. 1987) and 3D (Johannesen et al., 2002) by means of basin scale fluid flow simulators using the Darcy equation approach with relative permeability and capillary (entry) pressures. Sylta (2002 - Chapter 10) compared the Darcy approach to using the percolation method for mud-rocks and concluded that the percolation method could not provide a proper process description for oil and gas migration in low-permeability rocks. The reason for this conclusion is that the effective permeability for the hydrocarbon phases that leak out of traps is too low for migration to occur within the focused migration pathways that the percolation process requires. Vertical migration by capillary leakage from traps will tend to occur over larger, often several km2, areas. The hydrocarbon saturations within the cap-rock flow-paths will be low, as will migration velocities (Chapters 9, 10). Still, large quantities can leak out of traps at geologic time scales.
The capillary pressures at the interface of the cap rock and the reservoir has to exceed the entry pressures of the cap-rock for leakage to be initiated. Vassenden et al. (Chapter 8) showed in the laboratory that the column supported by the seal can become greater than the one supported by the entry pressures of the seal. When the supply of oil was stopped, their experiment showed that the column within the trap decreased substantially below the entry pressure of the seal, until a snap-off pressure was reached. The snap-off pressure in the laboratory experiment was observed to be only 35% of the entry pressure, suggesting that the capillary seals of cap-rocks can be much more dynamic than is normally considered in exploration work. The entry pressures sealing column capacity calculation often used in exploration should therefore be replaced by a more dynamic assessment of the cap-rock seal (Chapter 9). The time-dependent supply of oil and gas into a trap in combination with the capacity down to the spill-point of the trap are critical inputs to this analysis. It is not yet evident that this type of cap-rock seal analysis can at all be done successfully without a proper 3D migration modeling study.
The process of expulsion and primary migration from a source rock to the nearest carrier or reservoir rock is not dealt with here. A critical review of this process with respect to applying the correct process description in the fluid flow simulations has not yet been done. Most simulators rely on a Darcy type description for primary migration. The primary migration process is, however, surprisingly efficient, with efficient vertical migration reported to be occurring through thick mud-rock sequences.
Basin scale hydrocarbon migration modeling is suited for use in the geological and economic risk assessment of petroleum prospects. The application of this coupling is limited by our ability to provide sufficient input data, such as regional depth maps and 3D lithology property descriptions. The outputs of basin scale 3D migration modeling studies include volumes of oil and gas in mapped prospects. The quality of predictions made from these volumetric estimates can be tested by how well the model explains existing discoveries and dry wells in a study area.
Results from the first 3D migration modelling studies were immediately used in the risking process. Schroeder and Sylta (1993 - Chapter 16) and Skjervøy and Sylta (1993 - Chapter 17) showed how risking tables could be developed from “best case” scenarios and used to rank prospects. Both papers also show modelled amounts of oil and gas trapped in prospects versus geological time, thus allowing for a risking not only of the present day trapped oil and gas volumes, but also a risking of the palaeo-trapping of the hydrocarbons. The importance of this is that geologic events that are known to affect the trapping of oil and gas can be accounted for in a rigorous manner.
Sylta (1993 - Chapter 6) cross-plotted results from several simulation runs to elucidate dependencies between the amounts of gas trapped in one field (the Troll Field) and the amount of oil in a prospect in Block 35/9. The amounts of oil in 35/9 could thereby be estimated with some confidence to be between 50 and 95 106 Sm3. After the work was completed, well 35/9-1R discovered oil/gas in 1989. The details of the discovery could not be discussed in Sylta (Chapter 6) because of the 5 year confidentiality period for the drilling results.
Scenario modelling can be used as a term for the risking process described above. First, a “base case” that describes as many of the existing fields, discoveries and dry wells as possible is obtained. A match of the base case scenario to dry wells is not always required because mapping inaccuracies may have resulted in mapped traps that are not actually there. The best case simulation run defines a set of values for all input variables that has been obtained in the calibration process. Sensitivities are thereafter studied by varying important variables systematically, usually just one variable at a time, and each new set of values defines a separate case. Each case is simulated and the results are studied with respect to trapping of oil and gas in the prospects and calibration fields. Prospects that change little when variables are varied are considered to be robust, and therefore the risks are less. If most of the sensitivity cases result in dry prospects or small volume estimates, then the prospect should be risked accordingly.
The most important advantages of the scenario modelling approach to risking of prospects are that it is very flexible, only a small number of simulation runs are required, and the results are usually manageable for the geologist. A serious problem can arise when only small changes from the base case simulation run cause the calibration to become unacceptable (no hydrocarbons modelled into large fields etc). In this case, a recalibration of many of the sensitivity runs is needed, and this may take too much time in an exploration setting. Calibration is still, after all, very close to a manual task as long as inversion methods have not been adopted for hydrocarbon migration modelling.
The scenario modeling technique may be well suited to deal with fault seal scenarios. Experience from clay-smear modelling in faults and the resulting secondary migration pathways (Childs et al., 2002 - Chapter 13) suggest that faults can sometimes be treated either completely sealing or completely open to hydrocarbon migration. This leads to a definition of yes/no scenarios in the definition of hydrocarbon migration of some study areas. A typical example would be a scenario that defined a nominal fault A to be sealing or open to migration, a fault B to be sealing or open and a fault C to be sealing only when both fault A and B are sealing. This would lead to 4 simulation cases that could be simulated and the outcome used in the risking of one or more prospects.
Monte Carlo simulation techniques are very popular to use in the risking of prospects. A well established method for using Monte Carlo techniques is in the calculation of risked trapped hydrocarbons. A probabilistic quantification of the in-place oil and gas prospect volumes is usually required (Irwin et al. 1993). This can be achieved by estimating probability distributions of such parameters as gross rock volume, porosity, net/gross etc. These properties are then multiplied together using the Monte Carlo scheme (Figure 4). A few thousand calculations are typically used to compile the resulting probability distributions. The geologists also estimate the risk of success and multiply these together to find the aggregated probability of success, which in this case means the chance of finding economic quantities of oil and/or gas in the prospect.
Figure 4 Principle of Monte Carlo simulation approach applied to prospect risking.
a: Properties are multiplied together to give volumes of oil and/or gas.
b: Processes are simulated and results are compiled into volumes of oil and gas.
In the methods proposed by Krokstad and Sylta (1996 - Chapter 20), the same procedure is in principle used, except that the multiplication of values is substituted by a full pseudo 3D migration simulator that simulates the amounts of oil and gas in all traps throughout the geological history until the present day (Figure 4). The resulting present day volumes (Chapter 20) or oil/gas column heights (Sylta and Krokstad, 2003 - Chapter 21) can then be compiled into probability distributions. The overall uncertainty of the predictions can be extracted from the probability distributions and plotted as maps of e.g. most likely oil columns (Figure 5). This gives a good basis for understanding the spatial distribution of uncertainties, which is important for making decisions on where to drill the next exploration well in a study area.
Figure 5. Adoption of Monte Carlo approach to estimate map of most likely oil columns (m). See Chapters 21 and 22 for detailed discussions.
When significant computing resources are made available, the Monte Carlo simulation techniques described above can be used to also compile probability distributions of the variables that have been used to calibrate the basin simulator (Sylta, 2004 - Chapter 22). This means that we can describe not only the a-posteriori uncertainties of the input variables, and thus our geological knowledge, but also how much we have improved this knowledge by providing more information to the analysis. The new information may be the results from drilling one exploration well.
The probabilistic description of the input variables is a valuable quantity for comparing results from one basin to another, and from one exploration play to another. Computed standard deviations of input variables such as oil expulsion factors (Chapter 22) can be used to build world-wide experience databases of geological knowledge, and thereby make it simpler to assess the exploration status of different areas, and compare risks in different areas to each other.
Case studies play an extremely important role in the development of new exploration methodologies. The petroleum industry is conservative, and exploration risking techniques change slowly throughout the industry. This is one of the most important reasons that the included work has taken 10 years to complete.
Case studies are also the mechanism that has to be relied upon to show validity of the proposed process descriptions, methods and procedures. Geological processes can not be proven, as discussed by Oreskes et al. (1994). They argue that “verification and validation of numerical models of natural systems is impossible. This is because natural systems are never closed and because model results are always non-unique. Models can be confirmed by the demonstration of agreement between observation and prediction, but confirmation is inherently partial.” Therefore “models can only be evaluated in relative terms, and their predictive value is always open to question.” Oreskes el al. (1994) also discuss the practice of comparing numerical models with analytical solutions, and point out that the congruence between the analytical and the numerical solution does not mean that the numerical solution can be said to have been verified beyond the realm of the analytical solution.
The four included case studies (Chapters 16 to 19) demonstrate that the methods that have been introduced here based upon the process descriptions do demonstrate an agreement between observation and model results in these cases. This agreement is achieved by adjusting model variables until acceptable matches are found. The models are therefore not truly predictive. So far we have not documented any case where a modelling scenario could not be made to fit the data, except in cases where the input to the model was later considered wrong. Published documentation of such case studies is unfortunately difficult to achieve because of oil company confidentialities.
Each chapter in this work represents a self-contained work. Most chapters correspond to papers that have been published in international journals with review or conference proceedings with review. These papers are listed with the year they were published below (Table 1). Chapters 1, 2 and 23 (Conclusions) have not been submitted for publication. Many of the papers have been presented at international conferences. Table 1 provides the title of each chapter, while full paper information is presented in References to the end of Chapter 1.
Chapter 1 (this chapter) is a summary chapter that describes the background of the work, introduces the topic of the work, states the objectives, and gives an overview of the work. Chapters 2 to 5 discuss the migration process and how it should be described and quantified. Chapters 6 to 10 deal with simulation techniques for secondary migration and leakage modelling. Chapters 11 to 15 discuss the simulation techniques in more details, with focus on the effects of special geologic features: pressure compartments, hydraulic leakage, faults, burial reconstruction and palaeo-water depths. Chapters 16 to 19 present 4 different case studies, while Chapters 20 to 22 outline methods of risking using Monte Carlo methods. The risking methods make use of hydrocarbon migration simulators to build probability distributions of the results. Each chapter (paper) contains detailed discussion sections, while an overall discussion is included below. Chapter 23 presents the overall conclusions from the work.
Chapter 2 is a state-of-the-art discussion of processes and methods of evaluation relevant to hydrocarbon migration.
Chapter 3 presents experimental verification of the secondary migration process. This paper supports the use of focused migration stringers in the modelling of the secondary migration process.
Chapters 4 and 5 describe how the properties of migration stringers can be quantified using existing relative permeability and capillary pressures relationships. The results of this work includes graphs and equations that can be used to quickly assess migration losses, velocities etc, and equations that can be used to improve the calculations of secondary migration properties in 2D and 3D basin scale multiphase fluid flow simulators.
Chapters 6 and 7 discuss methods that can effectively be used to simulate the secondary migration of oil and gas from source to trap within carrier systems using computer technologies. The ray-tracing method is used for the simulation of the secondary migration. A multi-component hydrocarbon mixture can be simulated during migration and accumulation, although it is more time-consuming than using a gas-oil system.
Laboratory experiments described in Chapter 8 document the possible dynamic behaviour of traps during filling and leakage of hydrocarbons. The use of a simple entry pressure to describe capillary leakage out of traps may need to be revised and snap-off pressure concepts may have to be introduced.
Chapter 9 discusses the effects of dynamic hydrocarbon trapping on buried gas traps, with special emphasis on the delayed leakage that can result when trapped columns exceed the entry pressure capacities of the cap-rocks.
A review of the existing methods for simulating oil and gas secondary migration and leakage out of traps is performed in Chapter 10. There are limitations to the applicability of Darcy, ray-tracing and percolation type models. When the methods are applied outside the conditions where they are valid, erroneous conclusions and predictions most likely will result.
Chapters 11 to 13 discuss methods of handling faults as barriers to fluid flow, (water and hydrocarbons). Pressure modelling is discussed in Chapter 11 using a method that allows for the rapid modelling of fluid flow between pressure compartments. Lateral pressure differences are modelled to be controlled by averaged properties of the fault zones. Chapter 12 uses the pressure compartment methodology to assess vertical hydraulic leakage due to fracturing of the cap-rock seals. Chapter 13 introduces a method for using clay-smearing in faults to characterize the secondary migration properties of the faults and model migration with these properties.
Chapter 14 shows how advanced structural reconstruction techniques can replace the standard vertical subsidence and compaction modelling and thereby provide a more accurate geometry for use in hydrocarbon migration modelling at time steps before the present day. Chapter 15 also investigates effects of geometry on hydrocarbon migration, but this time the effects of reconstructing the bathymetry through time are studied. Comparisons are achieved by simulating migration for two cases, one with and one without the palaeo-water reconstruction effects.
Chapters 16 and 17 describe published case studies that use ray-tracing techniques to simulate hydrocarbon migration at basin scales. These are the first published case studies where hydrocarbon migration was simulated to yield volumetric estimates of the amounts of trapped oil and gas in un-drilled prospects for use in the risking process. Chapter 18 summarizes lessons learned during a number of North Sea migration case studies, including the one described in Chapter 17.
Chapter 19 contains a description of a basin modelling and secondary migration case study using ray-tracing techniques from a non-Norwegian basin, the East Irish Sea. The case study shows that a single simulation scenario can be created to explain the oil and gas distribution in a large number of fields. There is a complex hydrocarbon phase distribution in the traps within this basin.
Chapter 20 introduces the application of Monte Carlo simulation techniques in combination with hydrocarbon migration simulation techniques to provide uncertainty estimates and estimates of risked in-place oil and gas volumes for un-drilled traps. The probability distributions of the trapped volumes of oil and gas are here derived from probability distributions of the most important input variables, as defined by a user. Chapter 21 enhances the effort to provide improved risking tools by creating maps of estimates of probabilities, most likely oil and gas columns, and standard deviations of these columns. These maps can be used to choose the most optimum drilling location for the next well, and to plan drilling campaigns in an area.
Chapter 22 uses the established Monte Carlo simulation technique with more than 30.000 simulation runs to analyze the input variables. The results of each simulation run are weighted by how close the modelled column heights match observed hydrocarbon columns in drilled wells, and provide a-posteriori probability distributions of the geologic input variables. Geological uncertainties and estimates of measured improvements in geological knowledge are thus quantified
The conclusions from the work are listed in Chapter 23.
Table 1 Chapters & papers (*)
Sylta, Ø.: Introduction (with aims, background) and discussion.
Sylta, Ø.: Hydrocarbon migration, entrapment and preservation: Processes and Evaluation.
Selle, et al., 1993: Experimental verification of low-dip, low-rate two-phase (secondary) migration by means of gamma-ray absorption.
Sylta, Ø., Pedersen, J.I., Hamborg, M., 1998: On the vertical and lateral distribution of hydrocarbon migration velocities during secondary migration.
Sylta, Ø., 2002: Quantifying secondary migration efficiencies.
Sylta, Ø., 1993: New techniques and their application in the analysis of secondary migration.
Sylta, Ø., 1991: Modelling of secondary migration and entrapment of a multicomponent hydrocarbon mixture using equation-of-state modelling techniques.
Vassenden, F., Sylta, Ø., Zwach, C.: Secondary migration in a 2D visual laboratory model.
Sylta, Ø.: On the dynamics of capillary gas trapping: implications for the charging and leakage of gas reservoirs.
Sylta, Ø.: On the use of modelling techniques for hydrocarbon migration in carriers and seals.
Borge, H., Sylta, Ø., 1998: 3D Modelling of fault bounded pressure compartments in the North Viking Graben.
Lothe, A., Borge, H., Sylta, Ø.: Evaluation of late cap-rock failure and hydrocarbon trapping using a linked pressure and stress simulator.
Childs, C., et al., 2002: A method for including the capillary properties of faults in hydrocarbon migration models.
Huggins, P., et al.: Structural restoration techniques in 3D basin modelling: Implications for hydrocarbon migration and accumulation.
Kjennerud, T., Sylta, Ø., 2001: Application of quantitative palaeobathymetry in basin modelling, with reference to the northern North Sea.
Schroeder, F.W., Sylta, Ø., 1993: Modelling the hydrocarbon system of the North Viking Graben: a case study.
Skjervøy, A., Sylta, Ø., 1993: Modelling of expulsion and secondary migration along the southwestern margin of the Horda Platform.
Skjervøy, A., Sylta, Ø., Weissenburger, S., 2000: From basin modelling to basin management: reuse of basin-scale simulations.
Cowan, G., et al., 1999: Oil and gas migration in the Sherwood sandstone of the East Irish Sea Basin.
Krokstad, W., Sylta, Ø., 1996: Risk assessment using volumetrics from secondary migration modelling: assessing uncertainties in source rock yields and trapped hydrocarbons.
Sylta, Ø., Krokstad, 2003: Estimation of Oil and Gas Column Heights in Prospects Using Probabilistic Basin Modelling Methods.
Sylta, Ø., 2004: A probabilistic approach to improved geological knowledge and reduced exploration risks using hydrocarbon migration modeling.
Conclusions.
(*) Years are publication years for published papers. See also publication list at the end of Chapter 1.
There are a number of authors that have been involved in the work presented here. Much of the work presented involves simulation techniques that have been applied to data sets which have been worked up by others, in particular seismic interpretations, depth conversions, geological model constructions and geochemical and reservoir laboratory work. The thesis would not have been possible to complete without these contributions. Each chapter (paper) outlines the work done by the authors.
All the work described in the 9 single author chapters (1, 2, 5, 6, 7, 9, 10, 22 and 23) and which include 3 papers submitted for publication and 3 published papers, has been done by Sylta. Approximately 50% of the work of the 6 two-author papers has been done by Sylta. The first authors in these 6 papers (chapters 11, 15, 16, 17, 20 and 21) have done from 50% to 70% of the writing of the papers. The hydrocarbon migration simulations in these chapters were set up and run by Sylta.
There are eight multi-author papers included in the work (Chapters 3, 4, 8, 12, 13, 14, 18 and 19). My involvement in chapter 3 included proposing the concept of doing the experimental work, discussions during the planning stages, and participating in discussing the results and writing the published extended abstract. Most of the work in Chapter 4 was done by Sylta, including the method development, incorporating the numerical solution into the simulator and writing most of paper. The numerical solution of the loss equation was implemented by Pedersen.
Vassenden is the principal scientist behind the work in Chapter 8. My contribution has been in early discussions of the experimental set-up and in discussions of the experimental results and in contributing to the writing of the paper.
Lothe was the principal scientist in Chapter 12 and wrote most of the paper with continuous discussions with Sylta. The interpretation of the results was done jointly.
In Chapter 13 the method was first developed in joint discussions between Walsh, Childs and Sylta. The method was thereafter designed for Semi and implemented in the simulator by Sylta. My contribution to the paper also included participation in the writing of the paper.
Huggins was the coordinator and principal author of Chapter 14. My contribution included making the design of the 3DMove-Semi interface that allows for dynamic structural grids to be imported in Semi and used in the simulation of hydrocarbon generation, expulsion and migration. The migration modelling was done jointly by Tømmerås and Sylta. I also contributed to the writing of the paper itself and in the editing of the revised manuscript.
The inputs to Chapter 18 were taken from a number of case studies where I had been in charge of and/or performed the hydrocarbon migration modelling for Conoco. I also participated in the discussions and contributed to the writing, although the first author did most of the writing in this case.
Chapter 19 was based upon migration modelling work done by Hamborg and Sylta in a case study for BG, and where migration modelling was introduced to the BG staff. Most of the paper was therefore written by the BG staff with input from Hamborg and Sylta.
Petroleum migration can be modelled as a simple multi-phase fluid flow process in three dimensions. The complexity of the subsurface does, however, make this simple model description extremely challenging to solve. Subsurface sediments exhibit flow properties that are extremely dissimilar. Sandstones are permeable, and cause oil and gas to migrate almost instantly from source to reservoir at geological times. Mud-rocks have orders of magnitude lower permeabilities when at relevant burial depths and cause hydrocarbons to migrate at low velocities. It has therefore not been possible (nor attempted by the author) to arrive at a unified method that can be applied to hydrocarbon migration in all types of sediments. This does not, however, mean that such a method does not exist. One logical aim of further investigations may be to arrive at a method that can properly describe migration both in low and high permeability rocks. For such a method to be useful, it will also have to be fast, and account for special geologic features such as faults and fractures.
The influence of faults on pressures (mostly water flow) and hydrocarbon migration has been dealt with using somewhat different formulations in this thesis. This is partly because of the intention to focus on the compartmentalization properties of the pressure system in the earlier work, but also because of a lack of fault flow property descriptions during the late 1980’s and early 1990’s, when the pressure compartment modelling concept was developed. Presently, more accurate formulations of the fault flow properties have been suggested (Manzoochi et al., 1999), and these can be used. It is, however, important to note that the SGR formulations that are used for fault transmissibility calculations in reservoir simulators will not work well if they are used unmodified in basin scale simulators. The work of Borge and Sylta (1998) clearly shows that there has to be a depth (or temperature) dependent element in the fault transmissibility models. The fact that this has not been considered a requirement for reservoir simulation purposes may be because a producible reservoir will typically only span a few hundred meters in depth from top to base. Basin depths range over many kilometers. Permeability and entry pressure formulations such as the ones proposed by Sperrevik et al. (2002) and used in Sylta et al. (2003) may provide a better description of the process results. Still, the fact that pressures are controlled by fault permeabilities whereas entry pressures are the most important properties for the hydrocarbon system remains. Making predictions of hydrocarbon trapping by assessing pressure barriers (size and locations) are therefore inherently difficult and may sometimes be impossible.
The ultimate aim of the hydrocarbon migration modelling in this thesis has been to make use of improved process descriptions in an improved risking methodology for undrilled oil and gas prospects. The fact that many different geological features have to be accounted for in the simulations does make the implementation of an improved strategy to an extremely demanding task. One challenge is for instance that changes in the modelled sea bottom through the history of a basin may not be important to consider in many cases. But Chapter 15 shows that it can be very important in some cases: Migration spill routes from traps can be modified and modelled present day trapped hydrocarbon phases may change dramatically. One question to address in the future will therefore be: when do we need to consider this effect? If the answer is always, then the workflow has to be built accordingly. If we are able to create methods that can show when it is needed, then an overall less demanding, and therefore more efficient, risk assessment strategy may be devised.
The oil and gas migration effects of most of the geological processes that have been considered in this thesis show very non-linear behaviour. This leads to difficulties when trying to calibrate the migration models in different study areas. A large amount of freedom is given to the geologic modeller because there is generally a lack of data to calibrate the models to. This often makes it possible to come up with plausible migration scenarios to explain observations even when the input data are wrong. A typical case might be the tendency of some oil companies to use constant source rock properties within entire basins. The calibrated migration model may still provide an acceptable explanation of the discoveries that have been made even when later drilling shows that the source rock thicknesses change dramatically towards the basin depositional centres. When mismatches between the hydrocarbon migration model and the observed oil and gas pools are impossible, or very difficult, to remove by model calibration, then this may sometimes indicate that there is something wrong with the input data.
Frequently there will be insufficient data to make a reliable model of the carrier and source rocks at the time of running a migration case study. It may then be better to attempt to make alternative descriptions and use the migration simulator to study the sensitivity of the system. The Monte Carlo simulation approaches discussed in Chapters 20 to 22 can also provide important constraints on the geological system. There is still an additional need to develop more objective calibration methods. Many possible input parameter combinations can yield similar misfit results (see Chapter 22). Misfit is here defined as the error in calibration of the migration model (average difference between modelled and observed value). Using a single manually calibrated migration model may therefore lead to less accurate prospectivity assessments because the uncertainties in the modelling is not properly accounted for. Therefore, one future development of basin scale hydrocarbon migration simulators should be to develop inversion methods. These are methods that use the forward model iteratively to arrive at model parameters by minimizing a cost function. Inversion modelling can be combined with Monte Carlo simulation techniques to provide more accurate uncertainty predictions of amounts and locations of oil and gas in undrilled prospects.
With the many remaining challenges in the topic of hydrocarbon migration modelling and risk assessment, it may be tempting for oil companies to simply not consider using these methods in their exploration for oil and gas. It may indeed be possible to do this over the next few years. The main benefit would be that one can rely on old and tested methods, and simply hope to get slightly better at applying them than other oil companies. In the long run this strategy may be questioned: Modelling techniques will become better. Software will become easier to use, but more importantly: computers will become faster and faster, allowing for more and more accurate migration models to be used. Therefore companies that introduce these techniques into their risking workflow have the upside potential of significantly improving their predictions and becoming more and more efficient, possibly gradually becoming superior to companies that resist the use of these tools.
Whether this future scenario becomes a reality, remains to be seen. There are of course several future scenarios where this will not be the case, the most notable one being if determination of hydrocarbon phases can be made directly from seismic data. In the meantime, the conclusions listed in Chapter 23 can be utilized.
The paper in Chapter 22 was published in the Petroleum Geoscience in October 2004. The draft version, which was submitted to the committee, was therefore replaced with the printed version here. Only minor changes were made in the review of the paper, including a couple of paragraphs added in the discussions.
Berge, G., 2003. Norsk Petroleumsnæring – har den noen fremtid. Presented at Teknologiforum Rogaland 16 June, 2003. Presentation downloaded from NPD web-pages.
Bear, J., 1972. Dynamics of Fluids in Porous Media. Elsevier, New York.
Borge, H., Sylta, Ø., 1998. 3D Modelling of fault bounded pressure compartments in the North Viking Graben. Energy Exploration & Exploitation 16, 301-323.
Childs , C., Sylta, Ø., Moriya, S., Walsh, J.J., Manzocchi, T., 2002. A method for including the capillary properties of faults in hydrocarbon migration models. In: Koestler, A.G., Hunsdale, R. (eds): Hydrocarbon seal quantification. NPF special publication 11, 127-139.
Cowan, G., Burley, D., Hoey, N., Holloway, P., Bermingham, P., Beveridge, N., Hamborg, M., Sylta, Ø., 1999. Oil and gas migration in the Sherwood sandstone of the East Irish Sea Basin. In: Fleet, A.J., Boldy, S.A.R. (eds) Petroleum Geology of Northwest Europe: Proceedings of the 5th Conference, 1383-1398.
Hantschel, T., Kauerauf, A., Wygrala, B., 2003. Modelling capillary sealing and flow barriers in basin scale simulations. Presented at EAGE “Fault and Top Seals” conference in Montpellier, 8-11 September. Abstract p 21.
Irwin, H., Hermanrud, C., Carlsen, E.M., Vollset, J., Nordvall, I., 1993. Basin modelling of hydrocarbon charge in the Egersund basin, Norwegian North Sea, pre- and post-drilling assessments. In: Dore et al. (eds): Basin Modelling: Advances and Applications. NPF, Special Publications 3. Elsevier, Amsterdam, pp 539-548
Johannesen, J., Hay, J., Milne, J.K., Jebsen, C., Gunnesdal, S.C., Vayssaire, A., 2002. 3D oil migration modelling of the Jurassic petroleum system of the Statfjord area, Norwegian North Sea. Petroleum Geoscience, Vol 8, 37-50.
Kjennerud, T., Sylta, Ø., 2001. Application of quantitative palaeobathymetry in basin modelling, with reference to the northern North Sea. Petroleum Geoscience, 2001, vol. 7, no. 4, 331-341.
Krokstad, W., Sylta, Ø., 1996. Risk assessment using volumetrics from secondary migration modelling: assessing uncertainties in source rock yields and trapped hydrocarbons. In: Dore, A.G., Sinding-Larsen (eds.). Quantification and Prediction of Petroleum Resources, NPF Special Publication 6, 219-235.
Manzoochi, T., Walsh, J.J., Nell, P., Yielding, G., 1999. Fault transmissibility multipliers for flow simulation models. Petroleum Geoscience, 5, 53-63.
Oreskes, N., Shrader-Frechette, K., Belitz, K., 1994. Verification, validation, and confirmation of numerical models in the earth geosciences. Science, Vol. 263, 641-646.
Schroeder, F.W., Sylta, Ø., 1993. Modelling the hydrocarbon system of the North Viking Graben: a case study. In: Dore et al. (eds): Basin Modelling: Advances and Applications. NPF, Special Publications 3, 469-484.
Selle, O.M., Jensen, J.I., Sylta, Ø., Andersen, T., Nyland, B., Broks, T.M., 1993. Experimental verification of low-dip, low-rate two-phase (secondary) migration by means of gamma-ray absorption. In: Parnell, J. et al. (eds.). Geofluids'93. Contributions to an International Conference on Fluid Evolution, Migration and Interaction in Rocks, 72-75.
Skjervøy, A., Sylta, Ø., 1993. Modelling of expulsion and secondary migration along the southwestern margin of the Horda Platform; In: Dore A.G. et al. (eds.). Basin Modelling: Advances and Applications. Norwegian Petroleum Society Special Publication 3. Elsevier, Amsterdam, 499-537.
Skjervøy, A., Sylta, Ø., Weissenburger, S., 2000. From basin modelling to basin management: reuse of basin-scale simulations. In: Ofstad et al. (Eds). Improving the Exploration Process by Learning from the Past. NPF Special Publication 9, 141-157.
Sperrevik, S., Gillespie, P. A., Fisher, Q. J., Halvorsen, T., Knipe, R. J., 2002. Empirical estimation of fault rock properties. In: Koestler, A.G. and Hunsdale, R., Hydrocarbon Seal Quantification. NPF Special publication 11, 109-125.
Stortingsmelding nr. 38 (2001-2002) Om olje- og gassvirksomheten. Tilråding fra Olje- og energidepartemtentet av 28. juni 2002.
Sylta, Ø., 1987. SEMI - A program for the modelling of buoyancy driven, SEcondary MIgration of oil and gas by means of a ray-tracing technique. IKU Report 25.2403.00/01/87, 51p. Open.
Sylta, Ø., 1991. Modelling of secondary migration and entrapment of a multicomponent hydrocarbon mixture using equation-of-state modelling techniques. In: England, W.A., Fleet, A.J. (eds), Petroleum Migration. Geol. Soc. London, Spec Publication 59, 111-122.
Sylta, Ø., 1993. New techniques and their application in the analysis of secondary migration. In: Dore et al. (eds): Basin Modelling: Advances and Applications. NPF, Special Publications 3, 385-398.
Sylta, Ø., 2001. Quantifying secondary migration efficiencies. Geofluids (2002) 2, 285-298.
Sylta, Ø., 2004. A probabilistic approach to improved geological knowledge and reduced exploration risks using hydrocarbon migration modelling. Petroleum Geoscience, Vol. 10, no3, 187–198.
Sylta, Ø, Krokstad, W., 2003. Estimation of Oil and Gas Column Heights in Prospects Using Probabilistic Basin Modelling Methods. Petroleum Geoscience, Vol. 9, 243–254.
Sylta, Ø., Pedersen, J.I., Hamborg, M., 1998. On the vertical and lateral distribution of hydrocarbon migration velocities during secondary migration. In: Parnell, J. (ed.). Dating and Duration of Fluid Flow and Fluid-Rock Interaction. Geological Society London, Special Publication 144, 221-232.
Sylta, Ø., Childs, C., Sperrevik, S, Tømmerås, A. 2003. On the use of multi-carrier hydrocarbon migration modelling with clay-smearing in faults. Poster presented at the EAGE “Fault and Top Seals” conference in Montpellier, September 2003. Abstract only, p 83.
Ungerer, P., Doligez, B., Chenet, P.Y., Burrus, J., Bessis, F., Lafargue, F., Giroir, G., Heum, O., Eggen, S., 1987. A 2-D model of basin scale petroleum migration by two-phase fluid flow. Application to some case studies. In: B. Doligez (editor), Migration of hydrocarbons in sedimentary basins. Technip, Paris, 415-456.
Draft of papers not yet in print but included as chapters:
Huggins, P., Burley, D., Sylta, Ø., Tømmerås, A. Bland, S., Kape, S., Kusznir, N., Structural restoration techniques in 3D basin modelling: Implications for hydrocarbon migration and accumulation. Chapter 14.
Lothe, A., Borge, H., Sylta, Ø. Evaluation of late caprock failure and hydrocarbon entrapment using a linked pressure and stress simulator (in press). Chapter 12.
Vassenden, F. Sylta, Ø. Zwach, C., Secondary migration in a 2D visual laboratory model (in press). Chapter 8.
Sylta, Ø.: On the dynamics of capillary gas trapping: implications for the charging and leakage of gas reservoirs (in press) Chapter 9.
Sylta, Ø., On the use of modelling techniques for hydrocarbon migration in carriers and seals. Chapter 10.