TESTING AND IMPROVING TEMPERATURE THRESHOLDS FOR SNOW AND RAIN PREDICTION IN THE WESTERN UNITED STATES

The phase of precipitation at the land surface is critical to determine the timing and amount of water available for hydrological and ecological systems. However, there are few techniques to directly observe the precipitation phase and many prediction tools apply a single temperature threshold (e.g., 0 degrees C) to determine phase. In this paper, we asked two questions: (1) what is the accuracy of default and station optimized daily temperature thresholds for predicting precipitation phase and (2) what are the regions and conditions in which typical temperature-based precipitation phase predictions are most suited. We developed a ground truth dataset of rain vs. snow using an expert decision-making system based on precipitation, snow depth, and snow water equivalent observations. This dataset was used to evaluate the accuracy of three temperature-threshold-based techniques of phase classification. Optimizing the temperature threshold improved the prediction of precipitation phase by 34% compared to using 0 degrees C threshold. Developing a temperature threshold based on station elevation improved the error by 12% compared with using the 0 degrees C temperature threshold. We also found the probability of snow as a function of temperature differed among ecoregions, which suggests a varied response to future climate change. These results highlight a current weakness in our ability to predict the effects of regional warming that could have uneven impacts on water and ecological resources.