Intercomparison of Meteorological Forcing Data from Empirical and Mesoscale Model Sources in the N.F. American River Basin in northern California

Abstract

The availability of forcing data required to drive distributed hydrological models is significantly limited within mountainous terrain and at higher elevations due to the spatial scarcity of observations. Previous studies have commonly used three methods of estimating and distributing forcing data within basins that have sparse in-situ observations: a) one or two low-elevation stations in combination with empirical models, b) gridded output from a mesoscale model, or c) a combination of the two. In this study, we evaluated each source of forcing data within the heavily instrumented North Fork American River Basin in California. For the mesoscale model source, we selected the Weather Research and Forecasting (WRF) model, which used lateral boundary conditions from the North American Regional Reanalysis. Finally, each case of forcing data was used to drive the Distributed Hydrology Soil and Vegetation Model (DHSVM), and we examined those variables whose sources resulted in significant differences in simulated snowpack and streamflow. Results indicated that the choice of the least biased forcing source was not uniform for every variable. Accumulated precipitation was dependent on the year examined, however, it is important that the WRF model performed as well as the single station combined with the climatological weighting from the Parameter Regression on Independent Slopes Model (PRISM). Simulated streamflow from DHSVM was more sensitive to the source of precipitation forcing than other variables, resulting in biased high/low flow during years when the precipitation biased was high/low. Simulations of snowpack melt rates were most sensitive to the source of radiation data used and the elevational range considered. While the empirical estimated long-wave irradiance at high-elevation sites resulted in melt rates lower than observations, at lower-elevations the same forcing caused mid-winter melt that was not observed. However, the sensitivity of simulated snowpacks at lower-elevations was significantly reduced under a forest canopy. Short-wave irradiance from the WRF model was consistently less biased than empirical estimates, especially during cloudy days. In general, these results support the use of output from the WRF model over empirical estimates, but stress the need for additional observations to allow a complete evaluation of the energy balance.