The local search methods such as Simplex method are quite efficient, but not very robust. That is, they normally converge in a small number of iterations, but often fall into local minima. On the other hand, modern heuristics such as Simulated Annealing are very robust, but unacceptably expensive. On the basis of analysis of the advantage and disadvantage of Simplex method and Simulated Annealing algorithm, we present a hybrid method named SMSA aimed at coupling the efficiency of Simplex method and the robustness of Simulated Annealing for solving the groundwater inverse problem. In this paper, we introduce the realization of the SMSA algorithm's operators and apply it to the inverse simulation of a large water source area. As shown by the calculation results, the SMSA algorithm has the strongpoint of high precision, fast speed. Furthermore, this SMSA algorithm is easy to parallelize.