Probabilistic Model for Identifying Groundwater Contamination Sources

Abstract

If contamination is observed in an aquifer, the source of contamination is often unknown. We present a probabilistic approach for identifying source (or prior) locations using knowledge of the present-day spatial distribution of a contaminant plume. The results are probability distributions describing either the prior position of the observed contamination (backward location probability) or the travel time from a particular upgradient position to the observation location (backward travel time probability). We build from earlier work on backward probabilistic models in which transport is modeled backward in space and time, from the receptor (defined by the sample location and sampling time) to all possible sources, using information about the observation location and time, but ignoring the observed concentration. The relative concentrations of two or more observations provide additional information that can be used to identify possible source locations and travel times. In this paper, we present a method for using these concentration data to reduce the variance of the probability distributions, thereby providing better information about the groundwater contamination source location and release time. We present the theoretical development of travel time probability conditioned on measured concentrations, and we illustrate the method using a hypothetical, two-dimensional aquifer.