Abstract

The integration of well log and seismic data has traditionally been done using deterministic model-based approaches such as acoustic impedance inversion. More recently, statistical techniques have been applied to the problem. These approaches include geostatistical methods such as kriging with external drift (KED) and cokriging, multilinear transforms involving seismic attributes, and neural networks. In this paper, we propose a new technique that incorporates both multiattribute transforms (either linear or nonlinear) and geostatistics. We illustrate this technique with a 3D channel sand example from Alberta