Modified dynamically dimensioned search algorithm for identification of regional groundwater net-recharge rate

The Modified Dynamically Dimensioned Search Algorithm (MDDS) is proposed to efficiently and accurately identify the net recharge rate for a basin-scale groundwater model. MDDS is an enhancement to the Dynamically Dimensioned Search (DDS) heuristic. The efficiency and accuracy is examined by comparing the calibration results between MDDS and DDS. One issue that DDS has is that it cannot efficiently improve the objective value because it randomly selects the parameters to calibrate leading to unnecessary work being done. In this study, DDS is improved by the addition of three new search mechanisms. First, a new random parameter selection method is developed based on the calibration residuals. Second, the adjustment direction of each selected parameter is determined by comparing the observed with simulated heads. Finally, a two-phase search, which mimics the concept of global-local search using different step sizes, is also applied in the proposed method to improve the convergence efficiency. The proposed MDDS is applied to a highly dimensional parameter identification problem of the groundwater basin of Chou-Shui River Alluvial Fan. In this application, the simulation variable is hydraulic heads. To evaluate the convergence efficiency, 20 scenarios are examined with each scenario require 20 model simulation runs. The mean and standard deviation of the optimal objective values show that the proposed algorithm and two-phase search obtain superior results compared to DDS. The relation coefficient between the simulated and observed heads in the case of Chou-Shui River Alluvial Fan is 0.999. This study demonstrates the correctness and practicability of MDDS as an alternative to DDSs for calibrating a large-scale model with high parameter dimension.