| Techniques for addressing uncertainty and incommensurate planning goals within the context of screening (optimization) models are evaluated and applied to water resources problems. Among the topics discussed are the formulation of the objective function in order to incorporate risk aversion, the characterization of uncertainty through a moderate number of scenarios, and the solution of large-scale models via decomposition.;The techniques are demonstrated through a number of example problems and a detailed case study. The example problems involve urban water supply planning with uncertainty in future water availability, groundwater contaminant plume containment with imperfectly known aquifer parameters, and reservoir operations with uncertain future inflows. The value of considering more than just the first moment (expected value) of the objective function is demonstrated, as well as the importance of testing solutions which are found to be "optimal" based on a small or moderate number of scenarios.;The case study addresses water resources planning for the Edwards Aquifer area in Central Texas. Planning objectives for the region include the provision of reliable, low cost water supplies for municipal, industrial, and agricultural use; the preservation of endangered species dependent on springflows emanating from the aquifer; and the fulfillment of instream flow and estuary inflow requirements. Each of these goals is quantified explicitly in economic terms except the preservation of species, which is given a weight to be selected by a decision maker. A spectrum of solutions is generated to illustrate the trade offs and risks involved, and use of the model for determining system operating policies is demonstrated. Results indicate that either large infrastructural investments or substantial reductions in water demands will be required to protect species with a high degree of reliability.;Throughout, some of the barriers to actual use of optimization models in water resources decision making are addressed. It is argued that, if used in the proper context, models such as those developed herein can promote understanding of complex systems, help in the evaluation of trade offs involving risk, and support the generation of alternatives which promise to perform well under a number of contingencies. |
