Probabilistic petrophysical-properties estimation integratingstatistical rock physics with seismic inversion

摘要

A joint estimation of petrophysical properties is proposed thatcombines statistical rock physics and Bayesian seismic inver-sion. Because elastic attributes are correlated with petrophysicalvariables (effective porosity, clay content, and water saturation)and this physical link is associated with uncertainties, the petro-physical-properties estimation from seismic data can be seen as aBayesian inversion problem. The purpose of this work was to de-velop a strategy for estimating the probability distributions ofpetrophysical parameters and litho-fluid classes from seismics.Estimation of reservoir properties and the associated uncertaintywas performed in three steps: linearized seismic inversion to esti-mate the probabilities of elastic parameters, probabilistic upscal-ing to include the scale-changes effect, and petrophysical inver-sion to estimate the probabilities of petrophysical variables and litho-fluid classes. Rock-physics equations provide the linkbe-tween reservoir properties and velocities, and linearized seismicmodeling connects velocities and density to seismic amplitude. Afull Bayesian approach was adopted to propagate uncertaintyfrom seismics to petrophysics in an integrated framework thattakes into account different sources of uncertainty: heterogeneityof the real data, approxiniation of physical models, measurementerrors, and scale changes. The method has been tested, as a feasi-bility step, on real well data and synthetic seismic data to showreliable propagation of the uncertainty through the three differentsteps and to compare two statistical approaches: parametric andnonparametric. Application to a real reservoir study (includingdata from two wells and partially stacked seismic volumes) hasprovided as a main result the probability densities of petrophysi-cal properties and litho-fluid classes. It demonstrated the applica-bility of the proposed inversion method.