The interaction between the unsaturated zone, aquifer, and stream during a period of groundwater withdrawal

By combining laboratory and field experiments we study the release of water from above a falling water table and the implications this has on pumping test analysis and prediction of stream depletion. Undisturbed core samples from the field site were used in laboratory experiments to measure drainage responses to water-table drawdown. The responses can be sufficiently modeled by estimating the specific yield and five exponential time constants of a Moench/Boulton type model of delayed drainage. The average specific yield is thus estimated to 0.24 which is in agreement with previous small scale and large scale estimates. In addition two pumping tests were conducted near the stream. The drawdown analysis was made using a 3D numerical groundwater flow model facilitating use of the Moench/Boulton type boundary condition at the water table. The drawdown observations were fitted by calibrating two alternative models: one simulating the release of water at the water table as instantaneous; the other simulating the release of water as delayed. Comparing the calibration results for the two alternative models showed that they fit the drawdown observations nearly equally well, and most estimated parameter values are realistic and nearly identical. However, estimation of the specific yield is found to be sensitive to whether drainage is modeled as instantaneous or as delayed. In the first case the estimated specific yield is about half of the estimate from the core experiments and of the previous estimates; in the second case the estimate (0.17) is in better agreement with core and previous estimates (0.24). The analysis indicates that relatively fast drainage, and the existence of two drawdown dependent sources of groundwater recharge (the storage and the stream), complicates pumping test design to obtain unique parameter estimation. The analysis supports that an essential factor in parameter estimation by pumping test analysis for (at least some) unconfined aquifers is the use of a model that accounts for time-varying drainage from the vadose zone. Finally, when predicting stream depletion beyond 1 day of pumping it is acceptable in the present case to use a model simulating drainage as instantaneous if unbiased parameter estimates are used.