In this paper, a modified real-coded genetic algorithm is introduced to handle these groundwater simulation-optimization problems by combining a minimal generation gap model and the Pareto strength index procedure. The routine weighing the combination pattern of an objective function and the degree violating the constraints are altered to vector weighing the combination pattern. Thus the original constrained optimization problem is transformed to a two-objective optimization. The method is verified as feasible through application of a benchmark groundwater optimization problem and comparison with other optimization methods.