Accurate estimation of water supply reliability is often hindered by a paucity of historic streamflow data and the expense of adequate system modelling. A procedure using synthetic streamflow generation and screening models in determining water supply reliability in a cost-effective manner is presented. A series of screening models are developed to evaluate numerous synthetic streamflow data to select from them desired flows for incorporation into an optimization model. Results from the optimization model are used to generate cumulative distribution functions of system reliability. The water supply system serving Seattle, Washington is used as a case study.