Facies as the key to using seismic inversion for modelling reservoir properties
Although the general distribution of facies and corresponding reservoir properties should be predicted from geological understanding, the particular characteristics and heterogeneity of a field may not be readily anticipated, especially if well coverage is limited. Seismic data offer the possibility to provide additional constraints on the spatial distribution of facies. Seismic data on their own do not provide the complete solution, so a thorough integration of seismic, well and geological input, and any other available information, is required to reduce uncertainty and produce meaningful and predictive models. This has three serious implications for facies model building. Firstly, the seismic, well and geological constraints must be applied simultaneously to ensure a consistent and unbiased integration of all data. Secondly, the seismic and reservoir properties must be related through a predictive rock physics model that should be established through analysis of well log data and include facies-dependent depth trends. Thirdly, the facies definition must be meaningful in all domains: elastic, reservoir and geological. Naturally, the facies should also be identifiable from petrophysical well log data to ensure that wells can be used to constrain the models. Bayesian stochastic inversion provides a framework that is well suited to achieve all these goals.