Five snowmelt models: a comparison of prediction accuracy

Snow melt forecasts may be made using models that range from simple temperature indices to complex heat and mass flux budgets. The process-based models offer improved predictive capability but require extensive and expensive data collection equipment. Depending on the use of the results, the improved prediction accuracy may not warrant the extra expense and the greater chance of equipment failure.The extensive hydrometeorologic data base at the USDA Forest Service's Central Sierra Snow Laboratiory was used to estimate the energy fluxes that produce melt. Estimates of hourly and daily melt were made using five melt models of varying complexity. Predicted melt during April and May was compared to outflow data from meltpans sited both in a forest and in a forest opening. Predicted melt was also compared to hourly snow pillow and daily isotopic profiling snow gauge snow water equivalents. The most complex long- and shortwave model overpredicted melt by 30% compared to the observed melt. The net radiation model produced better results than the simple temperature index model.