Direct Sequential Co-Simulation: A New Stochastic Modelling for Environmental Applications

Abstract

The direct sequential simulation, without the transformation of theprimary variable, is a commonly used algorithm for the spatial characterization.of categorical variables, e.g. lithotypes. Journel (1994) showedthat the direct simulation of a continuous variable succeeded inreproducing the covariance model provided that the simulated valuesare drawn from local distributions centered at the simple krigingestimates with a variance corresponding to the simple kriging estimationvariance. However, the direct simulation does not reproduce the histogramof the original variables, one of the main requirements of any simulationalgorithm. Soares (2000) proposes a simple algorithm of direct sequentialsimulation that reproduces the histogram and the variogram of thevariable. However, the main advantage of the proposed algorithm isthat it allows the co-simulation procedure which does not call forany transformation of original variables A set of examples coveringdifferent situations illustrates the direct sequential co-simulationtechnique.