Unsaturated and Saturated Flow in Response to Pumping of an Unconfined Aquifer: Numerical Investigation of Delayed Drainage

Abstract

AbstractIn this study, numerical methods are used to examine the capillary fringe hypothesis proposed by Nwankwor et al. (1992) to explain variations in the storage property of an unconfined aquifer during pumping. Based on field observations, the hypothesis suggested that the response of an unconfmed aquifer to the stress imposed by pumping is largely controlled by the magnitude of vertical hydraulic gradients developed above the moving water table. After validation with a 1‐D hysteretic laboratory experiment and a 2‐D axisymmetric variably saturated semianalytical model, the numerical solution was used to simulate the pumping test of Nwankwor et al. (1992). Numerical parameters used for the simulation were chosen from experiments independent of the pumping test. There was good agreement between numerical results and field data at different depths and different radial distances from the pumping well. Numerically obtained variations in the magnitude of vertical gradients above the water table and the resulting variations in values of apparent specific yield also showed similar trends with the field data. The good agreement between the numerical and field results provides convincing evidence that the field observations and the resulting capillary fringe hypothesis concerning time variations of specific yield, are indeed consistent with the physical principles of variably saturated flow. The high storativity values frequently obtained by fitting early‐time data to type curves were shown to correspond to the period of increasing hydraulic gradients, and may not be related to intrinsic properties of the aquifer, such as compressibility. In agreement with the “volume‐balance method” of Nwankwor et al. (1984), it was also shown that the results confirm that the ultimate specific yield of an unconfmed aquifer is more related to the drainable porosity than the low values often obtained from type‐curve analysis.