An evaluation of a high-resolution hydrometeorological modeling system for prediction of a cool-season flood event in a coastal mountainous watershed

Abstract

This study used the atmospheric Fifth-Generation Pennsylvania State University-National Center for Atmospheric Research Mesoscale Model (MM5) and the University of Washington Distributed Hydrology-Soil-Vegetation Model (DHSVM) for the simulation of a complex rain-on-snow flood event that occurred from 28 December 1996 to 3 January 1997 on the 1560-km2 Snoqualmie River watershed in western Washington. Three control simulations were created with MM5 applied at 36-, 12-, and 4-km horizontal spacing and DHSVM at a horizontal spacing of 100 m. Results showed that the accuracy of the atmospheric fields increased with higher horizontal resolution, although underforecasting of precipitation was evident for all three resolutions. Simulated river flows captured 67% (36 km), 75% (12 km), and 72% (4 km) of the total flow and 52% (36 km), 58% (12 km), and 62% (4 km) of the event peak flow. Several sensitivity simulations of the modeling system (4-km spacing only) were conducted to improve on the control simulations. Adjusting the MM5 precipitation using observations led to a streamflow forecast that captured 90% of the total flow. Reduction of the model high-wind speed bias improved the simulated snowmelt, although the resulting effects on streamflow were relatively small. A sensitivity experiment that included the precipitation from an intense rainband that was not captured by MM5 revealed the importance of this high-intensity, short-lived feature; simulated streamflow from this experiment captured 93% of the total flow and over 82% of the peak flow, with a 4-h timing error. A final set of sensitivity simulations, using both a higher-and lower-elevation observation as the sole forcing of DHSVM (no MM5), revealed strong sensitivity to the observation location; using a slightly displaced (∼8 km) lower-elevation observation produced river flows that differed by over 18%. Both of the resulting simulated river flows forced by the two-station method were significantly lower than both the observed flows (35% and 53% of total observed flow) and the flows simulated with the MM5 input fields. A major cause of this low flow was that the temperatures at the observation locations were located in gap regions of the Cascade Mountains, were not representative of the basin-average temperature, and therefore caused too much precipitation to fall as snow.