Recently NWS offices are using a temperature-index snow model
in river flood prediction. While overall performance of this type of
model is reasonable, it can fail in unusual weather condition. One of
the possibility for the improvement may be transition to more
physically sound energy budget models. Question is if available
operational input data to drive the model have a reasonable level of
the accuracy to support river flood prediction. This presentation
focuses on this question. Comprehensive tests were performed with NCEP
energy model and NWS index-type model to analyze their sensitivity to
noisy data, and to quantify model errors to data uncertainty.