The management of storage facilities (reservoirs, lakes, and ponds) is an important element of the utilization of a surface water resources system. Dynamic programming (DP) techniques have been used extensively in the analysis and resolution of reservoir systems operation. A major drawback of DP for a multireservoir system is the curse of dimensionality. To overcome this problem, this study developed a nonlinear programming (NLP) solution scheme which also integrated the objectives of several stakeholders, objectives which were noncommensurate. The NLP model employed a time series model to generate synthetic inflows into the storage facilities in order to simulate future operations of the system. Two sets of simulations were run, and a statistical test was used to terminate them. The expected values of the decision variables were tabulated and plotted, as well as their standard deviations and probability distributions. The model allows decision makers to adjust operation targets for different alternative operation scenarios. ¿ 2000 American Geophysical Union