Model reduction for combined surface water/groundwater management formulations

A methodology for clustering stresses on a groundwater system based on similarity of hydrologic response produced by the stress is presented. The method is relevant to the computationally efficient calculation of response matrixes for use in groundwater management and other applications. The procedure is presented for the case in which the impact of pumping withdrawals on streamflow is of interest. The method uses cluster analysis on multiple, transient responses and a simplified response matrix for the modeled system. It is demonstrated on a field scale hypothetical management problem applied to a portion of the Republican River basin in the High Plains Aquifer of the United States with a multi-decadal planning horizon that maximizes well withdrawal while requiring that minimum streamflows be maintained. For this case, the clustering approach both reduces the computational requirement and helps to produce response coefficients with meaningful precision. The effectiveness of the model reduction is tested by comparing the optimal solutions produced by a full scale formulation and two reduced size formulations.