Modern Enviroinformatics: BME Mapping in Light of Uncertain Physical Knowledge Bases - The Equus Beds Case Study

CHRISTAKOS, George (george_christakos@unc.edu), SERRE, Mark (marc_serre@unc.edu), University of North Carolina, 111 Rosenau Hall, CB# 7400, DESE, Chapel Hill, NC 27599-7400

Key Words: spatiotemporal mapping, stochastic, uncertainty, enviroinformatics, terrain systems

This work is concerned with the Bayesian maximum entropy (BME) method, which can rigorously and efficiently handle space/time mapping applications of considerable practical importance. BME, which belongs to the field of Enviroinformatics, can integrate and process physical knowledge that belongs to two major bases: general knowledge (i.e., obtained from general principles and laws, summary statistics, and background information) and specificatory knowledge (i.e., obtained through experience with the specific situation). BME allows considerable flexibility regarding the choice of an appropriate spatiotemporal map, offers a complete assessment of the mapping uncertainty, and contributes to the scientific understanding of the underlying natural phenomenon. Valuable insight is gained by studying a spatiotemporal data set representing water-level elevations at the Equus Beds aquifer (Kansas). Numerical results show that, as was expected in theory, classical geostatistics results are obtained as special cases of the BME approach. In addition, a more accurate and informative analysis is possible by incorporating various sources of physical knowledge that cannot be processed by classical geostatistics methods.