Optimization of Monitoring Well Networks Using Spatial Winnowing and Temporal Thinning Applied to a Contaminant Plume at the Massachusetts Military Reservation

摘要

The emerging techniques of spatial winnowing and temporal thinning were instrumental in optimizing the characterization network to provide adequate but economical long-term monitoring for the FS-12 plume at the Massachusetts Military Reservation. The optimized network for long-term monitoring reduces sampling costs by 84%. The final network consists of 108 wells, which will be sampled semiannually or annually. Other factors affected selection of wells for the long-term monitoring network, including the need for sentry wells at key points outside the plume boundary and downgradient of the remedial system, and the need to address specific regulatory and stakeholder concerns. Spatial winnowing reduced the number of wells by 36% without significantly affecting the plume geometry, internal mass distribution, and total plume mass. Winnowing utilized three-dimensional (3-D) concentration kriging to examine spatial redundancy in the characterization network. A modified version of GSLIB provided two key measures of the importance of each well to the kriging process: the global kriging weight (GKW), and the number of cells influenced in the kriging grid (nCell). A plot of GKW versus nCell revealed a low-GKW population with a linear relation to nCell, and a high-GKW population uncorrelated with nCell. For the FS-12 plume, all of the low-GKW population was removed from the network. Temporal thinning indicated that annual instead of quarterly sampling would be sufficient to detect long-term trends. The thinning analysis utilized Sen's Method, a non-parametric statistical procedure for detecting trends, to evaluate temporal redundancy in the characterization network. Points were removed from quarterly time-series data until calculated trends were no longer within the 95%) confidence interval of the trend using the full data set for a given well.