Basin Dynamics Research Group
Explore this Section
Stratigraphical Uncertainty in Fault Seal Analysis
S.M. Clarke, M. Littler, S. Googan & D. Hughes
Faults are important controlling elements of hydrocarbons migration systems in sedimentary basins. When faults undergo displacement, their fluid transmissibility properties change as a result of juxtaposing different lithologies across the fault, by smearing semi-impermeable rocks within fault zones and by pumping or valving aqueous fluids along or across the fault.
Consequently, many researchers have explored numerical techniques of modelling the effects of faults on hydrocarbon flow in one, two and three dimensions. Modern numerical modelling techniques in three-dimensions go some way to reducing inaccuracies inherent in their one- and two-dimensional counterparts and hence improve the prediction of fault seal properties. However, such deterministic approaches can produce results that extremely sensitive to input parameters such as structural architecture and stratigraphical variation. These parameters can be amongst the most poorly defined of modelling inputs.
In this study, three-dimensional structural models, with associated stratigraphical and petrophysical data are used as a best-case interpretation of the geology. Fault juxtaposition relationships and the development of argillaceous smears are predicted in three dimensions using full, three-dimensional algorithms, and migration pathways are determined using fast, statistical flow pathway modelling methods. The result represents a deterministic solution for the best-case model. Multiple realisations of the same model, with varying input parameters are used to determine the sensitivity of the outcome to specific stratigraphical uncertainties.
Based on these results, uncertainty in both stratigraphical geometry and sequence are introduced as an inherent part of the modelling process. A large number of possible scenarios with variations in these parameters are automatically generated constrained by set limits determined from the muliple detemenistic vartiations.
Three-dimensional fault juxtaposition relationships and argillaceous gouges are determined for each individual scenario and combined with fast flow pathway modelling techniques to produce an analysis of potential migration pathways, leak points and fault-trapped hydrocarbons. The results of each scenario are combined to produce a stochastic model of the likelihood of hydrocarbon migration or entrapment at any point within the model given the structural architecture and uncertainty in stratigraphy.
The developed techniques are applied to a numerical model of a prospect from the Southern Gas Basin, UK, in which uncertainties in stratigraphy are a particular issue and have a significant impact on the likely locations of hydrocarbon accumulations.
The Stratigraphical 'best case' Model
In order to assess the effects of stratigraphical uncertainty on fault seal and hydrocarbon migration, the three-dimensional numerical model and its associated geological data is considered the best-case interpretation rather than an exact answer. The model is constructed from a number surfaces that represent faults and the major lithological contacts of the faulted blocks, and their variation in geometry in three-dimensions. The detail of the stratigraphy (lithology and lithological properties) is derived from detailed well data and mathematically interpolated to the model using standard, geospatial interpretation routines.
At desired points within in the model, best-case lithological logs are generated from the best-case model. Each lithological unit within these best-case logs is assigned a lithological uncertainty (see below). Best-case lithological logs can be generated for different points and assigned independent uncertainties such that the uncertainty of a given lithological unit can vary spatially. The actual wells themselves, from which data were interpolated to the best-case model, can also be treated as best-case lithological logs and assigned lithological uncertainty.
In the example below, a single listric normal fault cuts a lithological succession of three major sequences. The sequences are represented by three surfaces within the footwall and hanging-wall. The detailed lithology between these surfaces is derived from a number of wells that penetrate the model (not shown) and mathematically interpolated to the remainder of the model to generate a best-case interpretation. Four points are indicated from which best-case lithological logs can be extracted: one such log is shown. These best-case logs will be assigned lithological uncertainty.
Careful positioning of the best-case lithological logs can be used to reflect the variation in uncertainty in lithology over the extent of the model or explore specific stratigraphical issues at specific points.
Uncertainty in the detailed lithological succession of the subsurface arises from two major sources (below):
1) Uncertainty in the thickness of any interpreted lithological unit or, more correctly, uncertainty in the position of the contacts between lithological units.
2) Uncertainty in the composition of any lithological unit. For example, units interpreted as clean sand may contain shale partings.
Uncertainties in both these aspects of each best-case well log are used to generate random variations in the lithological succession.
To introduce unit thickness uncertainty, each best-case log is assigned an uncertainty value that represents the degree to which positions of lithological contacts within the log (and therefore unit thickness) may vary. The tops and bases of each major lithological succession within the log, represented by the model surfaces, are considered fixed points with zero uncertainty and hence each major succession within each log can be considered separately. This constraint is necessary to limit the generated random variations to geologically plausible scenarios that fit the structure of the model and produce a coherent geological interpretation in three dimensions. In the example shown above, the positions of the tops of major successions within the best-case log are indicated.
To introduce compositional uncertainty, each best-case lithological log is first translated into a log of 'reservoir potential' units (RIGHT). Each lithological unit is replaced by a 'reservoir potential' unit of the same thickness with a 'percentage' attribute that represents the percentage of that unit that can be considered potential reservoir (HC permeable). Clean sandstones translate into 100% reservoir units (coloured green), shale units translate into 0% reservoir units (red) and mixed sand and shale units, or carbonate units, are translated into reservoir units with a percentage that corresponds to the likely reservoir potential of that unit. In the image below, the best-case lithological log (from the model above) is shown translated into corresponding reservoir potential units.
Units that are not zero or one-hundred percent 'potential reservoir' are subdivided into a succession of separate units of zero and one-hundred percent reservoir potential in a random combination that, overall, are in the relative proportions of the original reservoir potential for that unit.
Random variations in unit thickness and composition occur concurrently. In the image shown below, possible random variations of the lithological units that comprise the lowermost major succession of best-case lithological log are shown. All units show variations in thickness and those units with a reservoir potential between 0% and 100% are subdivided accordingly. For example, Unit 11 is a mixed sand and shale with 80% sand. This unit is represented by a reservoir potential unit of 80% and subsequently divided into units of 0% reservoir potential (shale) and 100% reservoir potential (sand) of random thicknesses and distribution but such that 80% of the resultant unit thickness is reservoir (sand).
In order to populate the numerical model with geological properties that reflect randomly generated variations in lithology, synthetic petrophysical logs are derived from petrophyscial data corresponding to the best-case lithological log. For reservoir potential units of 0% or 100%, original well logs are stretched or condensed to fit the random variations in thickness of these units. For units with a reservoir potential between these extremes, representative 'reservoir' and 'non-reservoir' values for each petrophysical property are derived from the original log and used to general a bimodal synthetic log curve that alternates between these values in accordance with the randomly generated sequence of reservoir and non-reservoir fractions of these units.
|In the image shown left, a petrophyscial well log derived for best-case lithological log is used to derive logs for the randomly generated variations in lithology.|
Finally, the new petrophysical logs for each best-case lithological log are used to re-interpolate the properties of the model, calculate fault seal and determine hydrocarbon flow pathways and accumulations.
By generating 500+ random variations in lithology using this method, and using each to re-interpolate data, calculate fault seal and model hydrocarbon flow within the same numerical model, it is possible to combine results and generate stochastic solutions that highlight the most likely leak and seal points on faults and the most likely fault controlled hydrocarbons given the properties of the fault and faulted sequence and the uncertainty in the lithology of the faulted sequence. The example below shows fault controlled hydrocarbon accumulations within the Southern Gas Basin model, coloured according to the likelihood of their presence, given the sealing capacity of the faults and the uncertainty in the stratigraphical succession that comprises the model.
Further examples of the use of uncertainty modelling to assess the effects of lithological uncertainty on fault seal and hydrocarbon migration can be seen in the Southern Gas Basin example.
Combining multiple deterministic realisations and stochastic modelling in one workflow exploits the strength of both approaches. A single case, purely deterministic model produces results that can be heavily sensitive to input parameters and therefore places considerable (and generally unjustifiable) faith in the original interpretation of the geology. Multiple realisations allow the effects of specific uncertainties to be examined and sensitive uncertainties defined, but only a very limited number of possible scenarios can be generated. Purely stochastic modelling will, at best, reduce efficiency by modelling uncertainties to which the result is insensitive, and, at worst, has the potential to generate random unconstrained variations within the model that are not geologically plausible and may bias the result. By combining multiple realisations and stochastic modelling it is possible to use the result of end-member deterministic scenarios to condition the uncertainty modelling within the stochastic simulation and hence tailor the model to those critical uncertainties and constrain stochastic variations to geologically plausible combinations.
The three-dimensional modelling presented here was performed using in-house modelling software developed by the Basin Dynamics Research Group at Keele. This software has now been developed into a commercial three-dimensional fault seal analysis and flow pathway modelling software package called Qfault. The software combines all of the three-dimensional modelling techniques presented here with a powerful advanced visualisation and model-building tool. Images produced specifically using Qfault have the Qfault logo on them.
For further information on Qfault or any of the techniques demonstrated here please contact the corresponding author: