Stream depletion predictions using pumping test data from a heterogeneous stream–aquifer system (a case study from the Great Plains, USA)

This uniquely designed study investigates a fundamental issue—the feasibility of predicting stream depletion rates using linear uniform two-dimensional models. Required input for these models includes the hydraulic parameter estimates of the aquifer and the stream–aquifer interface, which may be obtainable through pumping test data analysis. This study utilizes pumping test data collected near the naturally meandering Prairie Creek, Platte River watershed, Nebraska, USA. Drawdown data were obtained in eight piezometer clusters, located on both sides of the stream, each containing three piezometers screened at different aquifer depths. Parameter estimates and, thus, stream depletion predictions varied over a wide range. Large parameter variance and a low degree of goodness of fit between the calculated and measured data encountered during the analysis suggest deficiencies of the uniform aquifer models in describing significant physical processes. This was also shown by additional field experiments that indicate lateral and vertical aquifer heterogeneity. Hydrogeological and sedimentological considerations of the meandering stream architecture (point bar versus cut bank) and the application of a linear piecewisehomogeneous model yielded a higher degree of goodness of fit and higher confidence in stream depletion predictions. Aquifer heterogeneity appears to be the major reason for uncertainty in stream depletion predictions, though other possible sources of uncertainty should be considered. These include the model linearity, the Dupuit assumption, the simplified representation of the stream–aquifer interface, the approximation of the stream as a straight line or a strip, and the impact of regional groundwater flow. q 2003 Elsevier B.V. All rights reserved.