Truncated Gaussian simulation of discrete-valued, ordinal coregionalized variables

This paper deals with the modeling and cosimulation of ordinal coregionalized variables, such as indicators, counts or continuous-valued variables discretized into a limited number of classes. The proposed model relies on truncations of a set of cross-correlated stationary Gaussian random fields. We provide guidelines and algorithms for inferring and validating the structural model (direct and cross variograms of the underlying Gaussian random fields) and constructing realizations conditioned to data. The algorithms are implemented in a set of computer programs and are illustrated with applications to datasets in pest management and mineral resources evaluation.