Energy budget snow models (EBSMs) hold out the promise of more accurate simulation of snow processes compared to traditional conceptual snow models. A major part of the transition from conceptual models to EBSMs for operational river forecasting is to understand the sensitivity of these new models to data errors.

This abstract presents results of sensitivity tests of an EBSM conducted at several SNOTEL sites in the Carson River basin. Bias and random errors in surface temperature, surface downward shortwave radiation, and surface wind were considered in the experiments. Sensitivity tests were conducted for the entire water year of 1999 and for the snow melt season only. A temperature index-based model, SNOW-17 (currently in use at RFCs), was also run for tests on temperature as a reference. Statistical error factors were analyzed to show relative sensitivity levels for different variables.

Our results confirm that the EBSM is highly sensitive to errors in input meteorological forcing data. The results show that the EBSM is sometimes more sensitive to temperature perturbations than Snow-17. Among the tested input data with introduced bias, the EBSM is most sensitive to temperature, followed by solar radiation, and then wind, with the maximum error factors of 3.0, 1.2, and 0.43, respectively. For tests with random errors, the error factors were reduced to 1.5, 0.16, and 0.30, respectively. Sensitivity results also show that starting the simulations around the time of maximum accumulation may improve model performance in the subsequent melting period. This is especially true for temperature, where the error factor was reduced to 1.5 for bias and 0.45 for random error.