Abstract: A method combining a mathematical programming model with an unsteady flow model is presented to determine the optimal sizes of main drainage projects in the reclaimed seabeach area, improving computional accuracy and reducing project costs. An objectiva is to minimize the cost, and constraits come from water blance, Samt-Venant equation, outflow formula of a sluice gate, the requirement of matching a sluice with a channel, and limitations to the size of a project. The optimization models are solved by dynamic programming method, in which each project is regarded as one stage, and the optimization problem is translated into a multi-stage decision process. In solving the unstead flow models, three finite difference methods, i. e., a method of characteristics, a diffusive scheme, and a two-step method are given, which automaticaly avoid nonconvergence. The results of case studies show that the above mentioned method is practicable and efficient.