This paper addresses the problem of uncertainty in optimizing water networks in process industries. Due to the fact that wastewater flow rates as well as the levels of contaminants may vary widely as a result of changes in operational conditions and/or feedstock and product specifications, optimal wastewater network designs should be resilient and able to accommodate such changes. Uncertainties considered in this study are derived from actual operational practice of major water-using units in a typical oil refinery of 400,000 barrels/day throughput. Rather than directly varying the concentrations and mass loads, only seasonal effects have been considered in this research to illustrate applications of the models. Sensitivity analyses reveal that introducing uncertainty in operating conditions results in considerable changes in the connectivity of the units involved in wastewater reuse. The proposed stochastic optimization model produces a flexible wastewater network which is capable of accommodating uncertainties in operating temperature. In the presence of uncertainties, the optimal network minimizes the impact on the reuse connectivity (topology) by providing 32.2 t/h of freshwater in addition to the condensing steam. The stochastic approach adopted in this research has been found to be effective in handling uncertainties and has resulted in flexible and resilient wastewater networks with low expected value of perfect information (EVPI). (c) 2004 Elsevier Ltd. All rights reserved.