Historically, water supply forecasts have been based on deterministic regression procedures (or single streamflow forecasts) that reflect only historical procedural uncertainty. The National Weather Service (NWS) has developed an Ensemble Streamflow Prediction (ESP) forecasting system that also incorporates meteorological uncertainty and can provide risk-based decision-makers with forecast likelihood information. By using present basin conditions and historical meteorological time-series as model inputs, numerous streamflow scenarios (traces) are created. A statistical analysis of an ensemble of traces is performed, from which a forecaster can make probabilistic statements about the anticipated seasonal discharge. The University of Arizona has implemented an ESP evaluation study with the help of the NWS Colorado River Basin Forecast Center (CBRFC) and the NWS Hydrology Lab (HL). Simulated historical streamflow predictions for fourteen watersheds in the Colorado River Basin have been created using the NWS ESP Verification System (ESPVS). The complete evaluation considers scalar, probabilistic, and conditional verification methods, and highlights the importance of a comprehensive diagnostic approach. Preliminary statistical results indicate that ESP provides an average 10-30% improvement in Ranked Probability Score over climatology forecasts, with the highest improvement occurring later in the forecast season. Evaluations identify, for different watersheds, the relative importance of climate forecasts, estimates of basin conditions, and hydrologic modeling.