Computational Investigations of Bayesian Maximum Entropy Spatiotemporal Mapping

The incorporation of various sources of physical knowledge is an important aspect of spatiotemporal analysis and mapping. There exist two major knowledge bases: general knowledge and specificatory knowledge. The latter includes hard data (exact measurements) and soft data knowledge bases (such as measurement intervals, probability assessments, expert views, etc.). Due to its sound epistemological background, mathematical rigor and considerable flexibility in incorporating various sources of physical knowledge, Bayesian Maximum Entropy (BME) is a powerful technique of spatiotemporal analysis and mapping. In this work we investigate the numerical performance of the BME method. The results obtained demonstrate the superior performance of BME over Simple and Indicator Kriging at a computational cost that is small for modern day computers.