A Geomorphology-centric ranking scheme for stochastic seismic inversion realizations

Abstract

Advances in the acquisition and processing of 3D seismic data have led to significant improvements in our ability to image subsurface reservoirs. The limitations of conventional 3D seismic measurements for reservoir characterization include its band-limited vertical resolution as well as the non-uniqueness of inverting seismic amplitudes for reservoir properties. These limitations have an impact on our ability to accurately model thin reservoirs for volumetric computations. Stochastic seismic inversion addresses these concerns by producing multiple, equally likely realizations, consistent with the available well and seismic data, at the fine-scale vertical resolution required for such reservoirs. The nature of the algorithm results in a large number of realizations (typically in excess of 200). We, therefore, require a methodology to rank the realizations in a way that is meaningful for the problem at hand and identify models corresponding to the P10th, P50th, and P90th percentiles. In the example presented here a feature recognized on a 3D deterministic seismic inversion result was interpreted as a mitten-shaped tidal bar using well-log data. The stochastic seismic inversion process generated realizations that showed a wide variation in the extent and geometry of the tidal bar. In this work we present an innovative ranking method used to classify the broad range of stochastic inversion results targeted at approximating this tidal bar geomorphological feature. From these results we were successful in identifying the various percentile models required for further analysis including input to reservoir simulation modeling.