Atmospheric moisture transport and moisture budgets during winter (December-April) simulated by the National Center for Environmental Prediction (NCEP) regional spectral model (RSM) are examined and used to simulate streamflow variations in the Sierra Nevada. The RSM was nested in 2°-latitude x 2°-longitude NCEP-reanalyzed atmospheric-data fields (as a surrogate for a perfect global-scale forecast). Precipitation, temperature, and solar insolation simulated by the RSM on a 50-km grid were used as inputs to precipitation-runoff models of the Merced, American, and Carson Rivers in an evaluation of the potential for global-to-local forecasting connections in the Sierra Nevada. Over the Western United States, the RSM captures both patterns and magnitudes of precipitation. The RSM has dry biases in the Southeast and over the Gulf of Mexico, but it captures interannual and intraseasonal variations well. During El Niņos, it simulates a precipitation-anomaly dipole over the Western States with wetter-than-normal conditions in California and drier-than-normal in the Pacific Northwest. The Northeast is relatively dry and the Southern States are wet, as in observations. The RSM did equally well during La Niņa winters.
The RSM simulations at grid points over the Sierra Nevada were fed directly to precipitation-runoff models of three river basins. The runoff models simulate daily water and energy balances within about 900 km2 of the American River basin, 700 km2 of Carson, and 500 km2 of Merced, with internal variations of elevation, slope, aspect, soils, and vegetation represented at resolutions of 100 m. The area-averaged altitudes are 1,250 m above sea level in the American River basin, 2,400 m in the Carson, and 2,800 m in the Merced. Because it comes from a lower, warmer basin, streamflow in the American River is about half wintertime rainfall runoff and half springtime snowmelt runoff; the higher and cooler Carson and Merced Rivers are dominated by springtime snowmelt runoff. These differences in altitude and temperature result in different requirements from, and sensitivities to, global-to-local downscaling: The snowmelt-dominated Merced and Carson River models accumulate cool-season precipitation in thick snowpacks that release most of their water in one major surge during the springtime. Downscaling for these basins is sensitive to the overall cool-season precipitation total and to the timing of the regional springtime transition from cool winter to warm spring temperatures. In contrast, downscaling to the mixture of rainfall and snowmelt runoff in the American River also depends on the timing, temperatures, and precipitation totals of individual winter storms, and thus is a more difficult task. The simulated streamflows are encouragingly like the historical flows in most years, but exceptions occur. Much of the success of this downscaling effort is attributable to the regional scale of the most hydrologically important weather fluctuations affecting Sierra Nevada rivers.