Conclusions

This thesis has investigated hydrocarbon migration modelling for use in quantitative prospect risk assessment. Migration process descriptions have been analyzed and applicable modelling techniques have been tested. The effects of overpressures, hydraulic leakage, fault clay-smear sealing, palaeo water-depths and structural restorations have been investigated. The overall conclusion is that the use of migration modelling to perform quantitative prospect risk assessments is feasible. Volumetric estimates of the amounts of oil and gas in undrilled traps can be obtained by using the migration modelling techniques. This conclusion is arrived at through a series of investigations of the different aspects of migration modelling. Models can be confirmed by demonstration of agreement with observation and prediction, but confirmation is inherently partial. Thus, alternative explanations to geologic observations can be put forward and confirmed by other authors. Several case studies have therefore been used to corroborate the experimental and analytical findings.

The overall conclusion is supported by the following:

• Experimental verification of low-dip, low-rate two-phase secondary migration has shown that the process is vertically focused.

• Analytical and numerical analysis suggests that secondary migration of oil and gas will be focused within a thin zone below vertical barriers. The hydrocarbon saturated zone will typically be less than 1m thick.

• Secondary migration saturations and relative permeabilities of most hydrocarbon flow-paths within carrier rocks will be low, while migration velocities will be very high, typically approaching 1000 km/My.

• There are inherent problems in modelling the secondary hydrocarbon migration process using full Darcy flow numerical schemes with constant saturations within each computing node, mostly due to a lack of resolution.

• Vertical capillary leakage out of a trap has been observed by a visual 2D laboratory model. The leakage process is dynamic, with maximum column heights that exceed the column height defined by the cap-rock entry pressures (h_Pe), reducing slowly to a snap-off pressure when supply is stopped.

• The measured experimental snap-off pressure was 35% of the cap-rock entry pressure, and cap-rock snap-off pressures above real traps may therefore also be lower than their cap-rock entry pressures (h_Pe).

• The efficiency of capillary gas leakage through cap-rocks is low, and dynamic columns may result when gas input to traps exceed gas leakage. When gas supply stops, or is reduced, the reduction in the dynamic column may take millions of years, thus maintaining a rate of gas leakage long after source rocks have stopped expelling hydrocarbons.

• Vertical capillary leakage through cap-rocks can, in general, not be modelled with percolation methods, because the flow-paths within the cap-rocks are not focused and migration occurs at low saturations.

• Fault property descriptions can be combined with a pressure compartment formulation into a method that effectively can simulate single phase fluid flow between compartments using averaged fault transmissibilities. Pressure histories of compartments can then be estimated.

• Estimated pressures in compartments can be combined with effective stress considerations to study the sensitivity of hydraulic leakage with respect to timing of leakage events and the influence upon pressures in neighbouring compartments. The timing of hydraulic leakage will have major impacts on hydrocarbon trapping.

• Secondary migration can be modelled efficiently by a ray-tracing technique where migration losses are calculated as properties of the flow-paths.

• Multicomponent hydrocarbon mixtures can be simulated during secondary migration and trapping. The difference in mixture patterns between traps may be used to estimate migration properties.

• The effects of structural restoration can be accounted for in hydrocarbon migration modelling. Hydrocarbon accumulations may be modelled to be created, moved and/or destroyed through geologic time. The present day pool contents may change as a result of applying these techniques.

• A method of incorporating clay-smear in estimating the hydrocarbon sealing properties of faults has been incorporated into a hydrocarbon secondary migration simulator. The sensitivity of migration and trapping to clay-smear fault sealing can be studied with this approach.

• A method of determining palaeo water-depths through time has been used in the construction of carrier bed burial histories. It is shown that often, the modelled hydrocarbon migration is not influenced to any significant degree by the proper inclusion of palaeo water-depths. For some traps, however, the modelled volumes of trapped oil and gas are severely affected, and the modelled hydrocarbon phases and volumes can be very different. Only migration modelling can elucidate whether this is an important effect at the present time.

• Hydrocarbon migration simulators can be used to estimate the present day volumes of oil and gas in prospects and fields, but only after calibration of the model to observed well data. Calibration may use estimated oil and gas volumes in drilled traps or observed oil and gas columns in drilled wells.

• Case studies have documented that basin scale hydrocarbon migration modelling can be used to understand the petroleum system and to make quantitative predictions about the amounts of oil and gas in undrilled prospects. Simulation results can be used to assess uncertainties of the predictions.

• Monte Carlo simulations can make use of basin scale migration simulations and thereby provide probabilistic estimates of the amounts of oil and gas in prospects.

• A misfit calculation can weigh the results from each simulation run within the Monte Carlo simulation loop by the difference between modelled and measured oil and gas column heights in wells or trapped oil and gas volumes in fields.

• The Monte Carlo simulation results can be used to study not only probabilities of trapped oil and gas as independent probability distributions, but also the combined probability of the two phases, P(oil-volume, gas-volume).

• The oil and gas probability distributions can be constructed from a total of approximately 1000 simulation runs, while a-posteriori probability distributions of the geological (input) variables require 1 to 2 orders of more simulation runs in order to cover the non-linear parameter space.

• The Monte Carlo approach combined with 3D basin scale hydrocarbon migration simulations may later be developed into a continuously improving description of the geological model: as more and more simulation runs are completed, the probability distributions of the geological variables become better defined because high misfit simulations will have less influence when more low-misfit simulation runs are sampled.

• A screening filter may be constructed using statistical data fitting to e.g. the first 10.000 simulation runs. A (careful) application of such a filter may increase the efficiency of the Monte Carlo simulation technique by discarding input variable combinations that will produce high misfit results.

Hydrocarbon migration and prospect risking is a complex and challenging area of research. We have investigated a number of effects here. There are, however, many more effects that need to be investigated by further research. Each oil and gas migration topic that can be properly described and understood will contribute to an improved understanding of the risks involved in drilling for new prospects. Some, but not all, of the results from further research will reduce the risk of drilling dry wells. Most importantly however, is that all results may contribute to improving the description of the uncertainties, and therefore allow for more realistic decision processes by oil companies. This, in turn, will lead to a higher utilization of existing resources in the exploration for oil and gas, it is hoped.