ANDERSEN, Peter F., pandersen@hsigeotrans.com, WEEBER, P.A., pweeber@hsigeotrans.com, HSI GeoTrans, Roswell, GA 30076

Groundwater models are often applied to assist in design of engineered facilities for waste sites, mining sites, water supply systems, or other activities requiring prediction of groundwater movement, flow rates, or changes in levels.  Although detailed results from the models are used in the design phase, the accuracy of the model predictions are rarely checked following construction and operation of the facilities.  The comparison exercise is seldom undertaken due to two primary factors: 1) the design is often modified during construction or operation such that it is not completely consistent with that which was modeled, and 2) the cost required to “verify” the model predictions may not be considered a useful expense.  Verification of site-specific model applications is important because the models may be useful for optimization of designs or operations following construction.  In a more general sense, model verification is important because, if successful, it provides confidence in the predictive capabilities of groundwater models.  The predictions from a groundwater model used to design an extraction/treatment/injection system at a military ammunition facility were re-evaluated using site-specific water-level data collected approximately one year after system startup.  Comparison of observed water levels, water-level changes (drawdown and pushup), and capture zones to predicted results was generally favorable.  The predicted capture zone was borne out although predicted water level changes were greater than observed, particularly in the deeper, un-pumped zones of the aquifer.  In general, the comparison indicated that performance specifications for the design had been achieved.  The model was later adjusted to provide an improved match to observed water-level changes.  Calibration to water-level changes due to extraction and injection was found to be a more effective technique than calibrating to absolute water levels.  Primary model adjustments were to increase the horizontal hydraulic conductivity and decrease the vertical hydraulic conductivity.  Following these adjustments the model was used to predict the hydraulic response to various alternative operational scenarios, including reductions in extraction, non-continuous operation, and re-allocation of extraction rates of individual wells.  The study shows the importance of a post-audit of model predictions for optimizing system performance and provides confidence in the predictive capability of appropriately applied groundwater models.