The Promises and Pitfalls of Direct Simulation

Abstract

The idea of direct simulation is to simulate in the space of the original data units, with minimal assumptions or transformations about the data distribution. A common approach to direct simulation is to proceed in a sequential fashion: direct sequential simulation (DSS). While the idea is not new, full development of the framework remains to be seen. The benefits of multiscale data integration, avoidance of the “Gaussian disease”, and flexible distribution considerations are offset by problems with histogram reproduction, the pervasive influence of Gaussianity, and proportional effect reproduction. This paper examines the promises and pitfalls of direct simulation with some illustrative examples, and also discusses the future of DSS as a practical alternative for natural resource characterization. The future of DSS requires a procedure to account for the dependency between the local variance and mean.