Influence of aquifer heterogeneity and return flow on pumping test data interpretation

Analytical solutions of drawdown in unconfined aquifers are widely applied for determining the specific yield, Sy, and the horizontal and the vertical hydraulic conductivity Kr and Kz, respectively. In many previous studies, estimates of Sy and Kz were observed to be highly variable and physically unrealistic. This has been attributed to the conceptualization of flow above the declining water table and aquifer heterogeneity in the applied models. We present the analysis of time-drawdown data from a pumping test instrumented with depth-differentiated observation piezometers arranged in clusters. Applying homogeneous anisotropic aquifer models in combination with nonlinear least squares parameter identification techniques, the data were analyzed in different groups: analysis of data from individual piezometer clusters and simultaneous analysis of the entire data set from all piezometer clusters (global analysis). From the cluster analyses, estimates of Sy and Kz exhibit large variances and depart from a priori estimates inferred from the hydrostratigraphy. Parameter estimates from the global analysis do not fall within the parameter bounds (minimum and maximum values) defined by the cluster analyses. While heterogeneity appears to be the important reason for large parameter variances, we discuss the influence of rarely considered aquifer return flow on drawdown and the inconsistent results from the cluster and global analyses. We corroborate our findings with data on hydraulic gradients, slug test data, and results from the application of a more realistic numerical flow model.