A Comparison between Geostatistical Inversion and Conventional Geostatistical-simulation Practices for Reservoir Delineation

Abstract

Abstract Geostatistical inversion provides a quantitative way to integrate the high vertical resolution of well logs with the dense aerial coverage of poststack threedimensional seismic amplitude data. A systematic field study is presented in this chapter to understand the relative merits of geostatistical inversion over standard geostatistical-simulation procedures that do not make explicit use of threedimensional seismic amplitude data. It is shown that, by making quantitative use of the poststack seismic amplitude data, geostatistical inversion considerably reduces the space of stochastic realizations that honor both the well-log data and the spatial semivariograms. Sensitivity analysis also shows that geostatistical inversion remains less affected by a perturbation of semivariogram parameters than standard geostatistical simulation. Tests of cross-validation against well-log data show that geostatistical inversion yields additional information over the average trends otherwise obtained with stochastic simulation. In the vicinity of existing wells, geostatistical inversion can potentially infer vertical variations of resolution similar to that of well logs and, at worst, of vertical resolution equal to that of the seismic amplitude data at locations distant from wells. A drawback of geostatistical inversion is the need to convert well-log data from depth to seismic traveltime. In addition, geostatistical inversion may be rendered computationally prohibitive when applied to large seismic and well-log data sets.