Facies Modeling Using Multiple-Point Statistics: An Example from a Carbonate Reservoir Section Located in a Small Part of a Large Shelf Margin of Arabian Gulf, UAE

Abstract

Abstract A new breakthrough in reservoir modeling is using Training Images (TI) with Multiple Point Statistics (MPS). A TI is a structural image of a reservoir layer or stratigraphic time interval, which represents all the relevant events (e.g. facies) at a time independently from a real location. Therefore, the TI's capsulate much broad geological sense with conceptual patterns or multiple joint relations beyond the traditional geostatistics where it may merely offer two-point correlations (e.g. variograms) with sparse well data. Each time sequence presents a transgression within Arabian plate along hundreds of km's distance length one would not see the whole sequence of a platform (e.g. lagoon, ramp and shoal deposits) through a relatively small reservoir area (30km x 30 km). This brings a common challenge in facies modeling of carbonate reservoirs regarding lateral changes in deposition settings that are rarely crossed along the small scale reservoir sequences although a diversity of facies types can be observed vertically. The Arabian Gulf plate reservoir of Lower Thamama around Abu Dhabi was studied to understand how a relatively limited reservoir area was deposited through the help of TI's. A rock-type study based on core and log data was compiled, and 23 facies types varying vertically were identified. TI's were drawn for each time interval between high resolution sequence stratigraphy (HRSS) picks. It was not eligible to drive properties through seismic attributes since the seismic attributes were weak for tuning the facies heterogeneity. The analogs, representing the depositional strata sketch were mapped with outcrop data, core descriptions and solid geological knowledge of the reservoir. Given these conceptual images conditionally, all possible scenarios of the deposition were simulated with MPS algorithms. Uncertainty analysis was also performed to infer the variability of the different TI's.