Locating Nested Monitoring Wells to Reduce Model Uncertainty for Management of a Multilayer Coastal Aquifer

Abstract

Although the aquifers below the Eastern Shore of Virginia serve as the region’s only sources of fresh water, increased development and intensifying groundwater demands have put this resource at risk. Simulation modeling is used as a tool to evaluate impacts of groundwater withdrawals on the aquifers. Using available monitoring data, a regional three-dimensional variable-density groundwater model (SEAWAT) was previously calibrated to the Eastern Shore’s aquifer system. In general, distributed parameter values were determined by kriging values around specified or calibrated parameter values at 21 pilot point locations spread across the Eastern Shore of Virginia. After calibration, an uncertainty analysis of the model’s performance was completed, and it was concluded that in order to most cost-effectively improve the model accuracy to facilitate management, it is important that any additional data collection be focused on the key modeling objectives, which are predicting hydraulic head and chloride concentrations under pumping conditions. The work described in this paper develops and demonstrates a methodology to identify the most useful locations, with respect to improving model accuracy, for installation of new multilayer nested groundwater monitoring wells. First, critical subregions where additional data may be useful were identified. Then a set of hypothetical new nested observation wells were evenly distributed throughout each critical subregion within the model. For each potential observation well site, the cumulative impact of data collection at that location on the prediction of hydraulic head values was estimated using the cumulative sensitivity of the relevant pilot point parameters to data at the new observation well. This cumulative sensitivity was calculated by summing the dimensionless scaled sensitivities of the new observation over all the relevant parameter values. The impact of collection of new hydraulic head information on the prediction of chloride concentrations was also estimated by determining the cumulative sensitivity of the pilot point parameter most sensitive to chloride concentrations based on the prediction scaled sensitivities. Once cumulative sensitivities were calculated for each potential observation well site, maps of the influence of new observations were created and used to identify the most effective location within each critical subregion to install a new nested observation well.