A Geostatistical Data Assimilation Approach for Estimating Groundwater Plume Distributions from Multiple Monitoring Events

Knowledge of the distribution of groundwater contaminant plumes is needed to avoid pumping contaminated water, assess past exposure to contamination, and design remediation schemes to contain or treat the contaminated area. In most field cases, however, contamination is discovered by a small number of fortuitously located wells, and the full distribution of the plume is never known. This paper presents a stochastic geostatistical data assimilation approach capable of estimating the plume distribution at any time before, during or after the monitoring history of a site. The approach uses concentration data from all available monitoring events and results in plume estimates that are also consistent with groundwater flow and transport at the affected site. One of the unique features of the approach is that measurements taken at times subsequent to the time for which the plume is to be estimated can be used as an additional constraint on the plume distribution. The method is demonstrated using two hypothetical examples. In the first example, the distribution of a plume is estimated based on multiple sampling events from a sparse monitoring network. In the second example, the plume distribution is recovered using temporal breakthrough curves from downgradient monitoring wells.