Probability Field Simulation

Abstract

A new approach for conditional simulation is proposed. The key idea of this method is to dissociate the task of estimating the local probability distribution functions from the task of producing equi-probable images. Probability Field Simulation starts with the premise that the local conditional distributions are known. The conditional simulations are then obtained by drawing realizations from these cdfs. The probability values used to draw from the local cdfs constitute a probability field, and are viewed as outcomes of a random function characterized by a uniform distribution and a given covariance function. Probability Field Simulation, thus, consists of the following steps: inferring the univariate and bivariate characteristics of the probability field, generating a non-conditional simulation of the probability field and , finally, using the simulated probability values to draw from the local cdfs. Guidelines on how to infer the distribution characteristics of the probability field are discussed and an example of application is presented.