An optimization model is presented for selecting among and sizing potential reservoirs on a river basin. The provision of water to meet municipal and industrial, irrigation, and hydropower demands, as well as recreation and flood control at each candidate reservoir site are considered. A compact, nonlinear optimization formulation for the reservoir sizing problem is derived by decomposing the problem into simulation and optimization components. Reservoir storage capacities needed are determined using a modified sequent peak algorithm to simulate monthly reservoir operation. Simulation is also employed to determine optimal sizes for hydropower generators at each site. Similarly, average annual flood control and recreation benefits are determined through simulation. An optimization scheme that considers annual yields for each purpose with specified reliabilities and conservation and flood storage as decision variables is used to integrate these simulations into an optimal screening algorithm. Significant savings in computational and memory requirements over other formulations are offered at the expense of nonlinear functional forms. Applications of the model developed with data from sites on the Lower Bear River are presented.