Description and Evaluation of a Hydrometeorological Forecast System for Mountainous Watersheds

Abstract

This paper describes and evaluates an automated riverflow forecasting system for the prediction of peak flows during the cool season of 1998?99 over six watersheds in western Washington. The forecast system is based on the Pennsylvania State University?National Center for Atmospheric Research (Penn State?NCAR) fifth-generation Mesoscale Model (MM5) and the University of Washington Distributed-Hydrology-Soil-Vegetation Model (DHSVM). The control simulation used the forecasts produced by the University of Washington's real-time MM5 forecasts system as input to the hydrologic model. A second set of simulations applied a correction scheme that reduced the long-term precipitation bias identified in the MM5 precipitation field. A third set of simulations used only those observations that are available in real time for forcing the hydrologic model. The various MM5?DHSVM forecasts are also compared with those issued by the National Weather Service Northwest River Forecast Center. Results showed that the observations-based simulation produced the most accurate peak flow forecasts, although it was susceptible to inadequate input data and the overdependence on the few available observations. The control simulation performed remarkably well, although several poor synoptic (MM5) forecasts, in addition to a model wet bias, produced a significant overprediction of peak flows over one watershed. The bias correction scheme did not prove worthwhile for peak flow forecasting, but may be useful for longer-term modeling studies where the emphasis is on long-term discharge rather than peak flow forecasting. A real-time updating procedure that incorporated meteorologic observations in the creation of initial hydrologic states showed considerable promise for forecast peak streamflow error reduction.