Research

Forecast Methods/SWE Assimilation

Because winter snow water equivalent (SWE) is a predictor without equal of summer streamflow in many western U.S. river basins, use of SWE observations (from the 600+ station USDA/NRCS SNOTEL network and several ASP stations in BC, Canada, run by Environment Canada) to estimate the initial snow state for the forecasts should benefit the forecasts by improving upon the state estimated solely by hydrologic simulation. A one-time assimilation routine blends the station anomalies at the time of the forecast with the hydrologic model SWE estimates to create the initial state used in the forecasts.

To achieve a high station density, anomalies were taken relative to station means from a shorter period (1990-1999) than is available for many stations, and stations were screened to have at least 9 (of 10) years' data present in that period. The station anomalies were applied to each model grid cell mean from the same period, for each snowband elevation resolved by the 1/8 degree model (up to about 6), and the result was merged with the original simulated SWE values according to a combination of weightings for a) distance from the grid cell to the station, b) elevation difference between the station and the elevation band, and c) the possible contributions of other stations in the area. Radii of station influence ranged from 50 km (with a linear dropoff to no effect) to 200 km (for the sparse stations in Canada); and weighting for elevation also dropped off linearly, over +/-3000 meters from the station elevation. For example, a station located at a grid cell center, at the same elevation as an elevation band, outside the range of all other stations, would have a weighting of 1.0, while the simulated value would be weighted 0. These are initial, somewhat arbitrary choices that may be modified in the future as more analysis is completed. Stations used in the assimilation are shown below, and the effects are shown on the Current Conditions pages.