Validating Prior Geological Scenario Uncertainty with Geophysical Data

Abstract:

Subsurface reservoir modelling, whether for groundwater, storage or oil/gas production relies on geophysical data for determining structure, rocks and fluid variations. The traditional approach depends on stochastic inversion of the geophysical image into subsurface models. However, in addition to geophysical data a wealth of geological information is available from analog or previous studies. Most of this information is ignored, and inversions resort to more mathematically-inspired priors often based on covariance models. In this presentation, using a real field application, we propose a method to validate a rich geological prior with geophysical data without the need for costly inversions. The result of this work is a wide, but geologically-realistic prior that can then be used in subsequent stochastic inversions. To achieve this, we propose to validate plausible geological models (from analog studies) with the observed geophysical data through a global, pattern-based measure of dissimilarity. This global dissimilarity measure is defined between the forward simulated geophysical response of a large variety of geologically plausible models and the observed field data. The proposed dissimilarity measure relies on a comparison of the wavelet decompositions between observed and forward simulated geophysical responses. The difference in frequency distribution of the wavelet coefficients is used via a JS-divergence measure to define the dissimilarity between all the subsurface models and the observed data.

The proposed approach is applied to a real field offshore reservoir in West Africa, where a 3D seismic cube is available. The uncertain geological parameters defined for this case are the rock physic model, the infill channels size, depth, sinuosity, the proportion of sand/shale and the stacking patterns.