Stream-aquifer interactions under pumping conditions in an unconfined aquifer considering three-dimensional flow, aquifer heterogeneity, and anisotropy

Stream-aquifer interactions, as major processes of aquifer recharge and discharge are of great significance for natural resources sustainability. Consideration of stream depletion (SD) under pumping conditions is paramount for the understanding and quantification of water budgets on the watershed scale. Direct techniques for determining SD rates, such as stream discharge measurements, were often used unsuccessfully in cases of natural streams. Indirect techniques, such as the application of inverse models to drawdown data obtained via pumping tests, regained attention with the advent of improved theoretical models of stream-aquifer interactions. At the naturally meandering Prairie Creek (Platte River watershed, east-central Nebraska), a multi-component study has been carried out to test the applicability of theoretical models of SD to real stream-aquifer systems. Prairie Creek is hydraulically connected to the underlying alluvial aquifer, which is associated with the braided-river depositional environment of the Platte River. A test site near a high capacity pumping well close to Prairie Creek (57 m) has been instrumented with a three-dimensional (3D) piezometer network, stream gauging station, and an observation well. Pumping tests under various hydrologic conditions and slug tests were used for hydraulic aquifer characterization on different spatial scales. The pumping tests data were analyzed with two-dimensional (2D) and 3D (semi-) analytical and numerical models of stream-aquifer interactions. In the framework of uniform aquifer models, anomalous time-drawdown responses were used to delineate preferential flow paths that are characteristic of braided-river deposits. For these models uncertainties in parameter and SD rate estimates were quantified. Inverse 3D numerical models consistently explained the measured drawdown responses under different hydrologic conditions. Although, the approximation of the stream in the 2D analytical models is of little physical merit, such models arrive at a good proxy of SD rates as follows from the comparison with results from the 3D inverse models. Techniques applied in this study proved to be feasible for SD rate estimation from pumping test data and aquifer heterogeneity delineation