A Bayesian/geostatistical approach to the design of adaptive sampling programs

Traditional approaches to the delineation of subsurface contamination extent are costly and time consuming. Recent advances in field screening technologies present the possibility for adaptive sampling programs--programs that adapt or change to reflect sample results generated in the field. A coupled Bayesian/geostatistical methodology can be used to guide adaptive sampling programs. A Bayesian approach quantitatively combines ``soft`` information regarding contaminant location with ``hard`` sampling results. Soft information can include historical information, non-intrusive geophysical survey data, preliminary transport modeling results, past experience with similar sites, etc. Soft information is used to build an initial conceptual image of where contamination is likely to be. As samples are collected and analyzed, indicator kriging is used to update the initial conceptual image. New sampling locations are selected to minimize the uncertainty associated with contaminant extent. An example is provided that illustrates the methodology