Niched Pareto genetic algorithm (NPGA) is a superior method to solve the multiobjective optimization problems because it is applicable to a large range of variable problems, and can search nonlinear and discontinuous space without need for continuity and secondorder partial differential operators. However, it is inefficient in finding the Pareto optimal solutions due to two shortcomings: premature convergence to local area and low convergence speed. In this paper, an improved NPGA (INPGA) is developed to promote the solving ability of algorithms. The main improvements of INPGA include three aspects: the Pareto solution set filter, the elite individual preservation strategy and the neighborhood space Mühlenbein mutation. Moreover, the message passing interface (MPI) for parallel computing and the operation library of individual fitness is introduced in the INPGA to improve calculation speed. Also, the INPGA is applied to a twodimensional hypothetical test problem to demonstrate the multiobjective optimal design of a groundwater pumpandtreat system. The comparison of results shows that the INPGA is superior to the NPGA in finding the tradeoff curve with a range of applicable Pareto optimal solutions.