Use of input uncertainty and model sensitivity to guide site exploration

Three Quantitatively Directed Exploration (QDE) methods to identify optimum field sampling locations based on model input covariance and model sensitivity are presented. The first method bases site exploration only on the spatial variation in the uncertainty of input properties. The second method uses only the spatial variation in model sensitivities. The third method uses a first-order second-moment (FOSM) method to estimate the spatial variation in the output covariance. The FOSM method estimates output uncertainty using the product of the input covariance and model sensitivity. The three methods are illustrated by means of a synthetic groundwater site simulated with MODFLOW-2000. The groundwater-flow model computes piezometric head and the sensitivity of head to changes in input values. The QDE methods are evaluated by comparing model results to the "true" head. For the synthetic site used in this study, the most effective QDE method was the FOSM method.