Probabilistic seismic inversion based on rock-physics models

Abstract

A site-specific rock-physics transform from porosity, mineralogy, and pore fluid to elastic-wave velocities is used to invert seismic amplitude data for clay content, total porosity, and saturation. The implementation is Bayesian and produces probabilistic values of the reservoir properties from seismic measurements and well data. This method focuses on an exploration setting where minimal data exist. Two key assumptions reduce the problem and keep the prior information as noncommittal as possible. First, a prior interpretation of the seismic data is required that provides a geobody on which to perform the inversion. Second, the reservoir thickness is assumed to be constant, as are the rock properties within the reservoir. The prior distributions of the reservoir properties are assumed to be uncorrelated and independent, although this is not an essential assumption. Central to theinversion is the generation of a complete set of earth models derived from the prior distribution. A site-specific rock-physics model translates these properties (clay content, porosity, and saturation) into the elastic domain. A complete set of forward seismic models accompanies the earth models, and these seismic models are compared to the real data on a trace-by-trace basis. The reservoir properties corresponding to the seismic models that match the real data within predefined errors are used to construct the posterior. This method was tested on well and seismic data from offshore western South Africa. Initial results at calibration and test wells indicate an overprediction of porosity and uncertain predictions of clay content and saturation. This is a result of the constant-thickness assumption. However, a highly negative correlation between porosity and thickness is predicted, which manifests the success of this method.