A model to forecast short‐term snowmelt runoff using synoptic observations of streamflow, temperature, and precipitation

Abstract

Snowmelt runoff is forecast with a statistical model that utilizes daily values of stream discharge, gaged precipitation, and maximum and minimum observations of air temperature. Synoptic observations of these variables are made at existing low‐ and medium‐altitude weather stations, thus eliminating the difficulties and expense of new, high‐altitude installations. Four model development steps are used to demonstrate the influence on prediction accuracy of basin storage, a preforecast test season, air temperature (to estimate ablation), and a prediction based on storage. Daily ablation is determined by a technique that employs both mean temperature and a radiative index. Radiation (both long‐ and short‐wave components) is approximated by using the range in daily temperature, which is shown to be closely related to mean cloud cover. A technique based on the relationship between prediction error and prediction season weather utilizes short‐term forecasts of precipitation and temperature to improve the final prediction. Verification of the model is accomplished by a split sampling technique for the 1960–1977 period. Short‐ term (5–15 days) predictions of runoff throughout the main snowmelt season are demonstrated for mountain drainages in western Washington, south‐central Arizona, western Montana, and central California. The coefficient of prediction (C_p) based on actual, short‐term predictions for 18 years is for Thunder Creek (Washington), 0.69; for South Fork Flathead River (Montana), 0.45; for the Black River (Arizona), 0.80; and for the Kings River (California), 0.80.