Abstract

We describe an approach to seasonal streamflow forecasting for the Snake River Basin in the Pacific Northwest that uses the variable infiltration capacity (VIC) macroscale hydrology model in conjunction with adjustments of the model's initial snow state using moderate resolution imaging spectroradiometer (MODIS) imagery for Winters 2000–2004. We evaluated seasonal streamflow forecasts made on March 1, April 1, and May 1 through the end of July, as well as short lead forecasts for two-week durations beginning on April 15 and May 15 retrospectively for 2000–2004 (2004 forecasts are archived forecasts made in real time).

In general, reductions in mean absolute error were more prevalent and larger for the two-week forecasts than for the longer seasonal forecasts, and the forecast skill improvements resulting from updating were greatest for forecasts made earliest in the season. Inclusion of MODIS data resulted in forecast error reduction (or no change in forecasts) in 59% of the seasonal forecasts (54% of the two-week forecasts), however when only the largest adjustments were considered, the fraction of improvement was much higher. We also evaluated the effect of MODIS adjustment on reservoir storage volume forecasts using a monthly time step reservoir simulation model. For two reservoirs (Dworshak and American Falls) where the reservoir model performed well in retrospective simulations, storage forecast errors were reduced for most forecasts at Dworshak, but were mostly unchanged or degraded at American Falls. The differences are attributed to reservoir operating characteristics, which are more constrained at American Falls than at Dworshak.