Ensemble contaminant transport modelling and Bayesian decision-making of groundwater monitoring

Abstract

An ensemble-based Monte Carlo framework in conjunction with a particle tracking method was used to investigate contaminant movement into the subsurface environment from an instantaneous leak emanating from a random location near the ground surface. Our findings include the following. A large number of wells, exceeding in all cases 12 monitoring wells were required in order to detect contaminants with some degree of confidence. The optimum distance that returned the maximum probability of detection Pd varied based on the heterogeneity and dispersion of the geologic medium. A low dispersive environment required larger distances in order for plumes to be able to be captured by the monitoring network, whereas high dispersive environments allowed detection close to the contamination source. Low dispersive geologic media made selection of the location of the monitoring system relatively insensitive to the distance from the source, whereas in high dispersive environments there appeared a narrow region, where the system needed to be placed in order to achieve high probabilities to detect. In highly dispersive media sampling influenced the Pd significantly. Finally, optimization of a monitoring network needs to consider concurrently the maximization of the probability of detection and the minimization of the contaminated volume.