Bayesian Lithology-Fluid Inversion based on Well and Seismic Data

Information about petroleum reservoirs are collected in wells or seismic surveys. Well observations are precise but spatially sparse. Seismic data is indirect and fairly unprecise, but has good spatial coverage. Physical models for these data acquisition procedures are well established. Reservoir management should be supported by 3D representations of lithologies and brine/oil/gas saturations of the reservoir. The focus of this project is to provide methodology for inverting the available well and seismi c information into decision supporting 3D reservoir representations. Spatial coupling of reservoir variables, varying uncertainties with spatial dependence in the avaiable data, and complex relations between seismic data and reservoir variables constitut es major challenges. The problem will be phrased in a probabilistic setting - termed Bayesian inversion. The results of the project will improve exploration for new petroleum prospects and contribute to better reservoir management.