Inverse Solution of Groundwater Models - Indirect Approach

The main purpose of this study was developing a framework for the calibration of groundwater models. The automatic calibration with indirect approach has been considered and inverse problem in groundwater flow modeling was defined as an optimization problem. For solving the nonlinear optimization problem, genetic algorithm has been used. Minimizing the root of mean square deviation between observed and the corresponding computed heads in MODFLOW was considered as a calibration and evaluation criterion. Also, the hydraulic conductivity and specific yield (with known zonation) have been considered as model parameters. Changing some part of MODFLOW-2005 source codes and embedding the genetic algorithm, an optimization program (MF2005GA_P) has been developed in FORTRAN 90. Internal exchange of main variables (e.g. RMSE) has decreased the execution time noticeably compared to the approach that linked optimization and simulation codes. The program has been developed and primarily evaluated based on a hypothetical model. Next, Abhar aquifer was selected as a case study and the program performance in a real scale has been investigated. The results have shown about 40 percent decrease in RMSE compared to the trial and error calibration results.