A Multiple-scale, Pattern-based Approach to Sequential Simulation

Abstract

In the context of multiple-point geostatistics, a new algorithm (SIMPAT) is presented. This algorithm relies on several image processing concepts, such as image similarity, to borrow and reproduce patterns from training images constrained to hard and soft data. The method makes use of a new multiple-grid approach by which the scale relations between the training image patterns are better captured and reproduced.