Ninth International Geostatistics Congress, Oslo, Norway
June 11 – 15, 2012



Abstract No.:



Distance-based Kriging relying on proxy simulations for inverse conditioning


D. Ginsbourger, University of Berne, Departement of Mathematics and Statistics (CH)
B. Rosspopoff, Ecole des Mines de Saint-Etienne (FR)
G. Pirot, University of Neuchâtel (CH)
N. Durrande, Sheffield University (UK)
P. Renard, University of Neuchâtel (CH)


Given a reference tracer response curve and a set of candidate geological models, we consider the problem of identifying the set of models whose response computed by an precise forward flow and transport simulation matches the reference curve. In order to minimize the number of calls to the precise flow simulator, a distance-based approach relying on fast proxy simulations (with simplified physics and/or numerics) developed by Scheidt and Caers is revisited, and turned into an original non-stationary Kriging method. The covariance function is obtained by combining classical low-dimensional covariance kernels with the proxy function, hence generalizing the idea of random fields? deformation to high-dimensional Computer Experiments. Once the precise simulator has been applied for a set of geological models and an initial Kriging model has been inferred on those results, the predictive distributions of fits for the remaining geological models can be used as a guide to solve the inverse problem efficiently.




    Copyright 2012 International Geostatistics Congress

Produced by X-CD Technologies Inc.