Bayesian Approaches for Adaptive Spatial Sampling: An Example Application

摘要

BAASS (Bayesian Approaches for Adaptive Spatial Sampling) is a set of computational routines developed to support the design and deployment of spatial sampling programs for delineating contamination footprints, such as those that might result from the accidental or intentional environmental release of radionuclides. BAASS presumes the existence of real-time measurement technologies that provide information quickly enough to affect the progress of data collection. This technical memorandum describes the application of BAASS to a simple example, compares the performance of a BAASS-based program with that of a traditional gridded program, and explores the significance of several of the underlying assumptions required by BAASS. These assumptions include the range of spatial autocorrelation present, the value of prior information, the confidence level required for decision making, and inside-out versus outsidein sampling strategies. In the context of the example, adaptive sampling combined with prior information significantly reduced the number of samples required to delineate the contamination footprint.