A realtime coupled atmospheric-hydrologic prediction system, which combines the Penn State-NCAR MM5 model with the University of Washington Distributed Hydrology Soil Vegetation Model (DHSVM), has been developed for short-term (12 to 72 hour) operational flood forecasting in the mountainous terrain of the Pacific Northwest. This completely automated system provides flood forecast guidance on par with that issued by the NWS River Forecast Center.

This paper presents results from the use of this system for weather and streamflow forecasting for hydropower operations in the Pacific Northwest. The MM5-DHSVM system has been applied since the winter of 2000 to Seattle City Light's hydropower operations in the Skagit Watershed, which drains an area of approximately 5000 square kilometers of the Cascade Mountains in Northern Washington State. Hydropower operation in the Skagit is complicated by the demand for flood control and the requirement to satisfy downstream flow targets to preserve salmonid habitat.

Results from the operational system are presented for two points on the river network: inflow to the main reservoir and a downstream fish flow target. These results are compared to guidance issued by the RFC and SCL's current forecast system. The relative error of streamflow forecasts is compared to the relative error of the atmospheric forecasts showing that atmospheric uncertainty dominates hydrologic model uncertainty. Preliminary results from the use of ensemble-based NWP to obtain streamflow ensembles are discussed.