Evaluation of Ensemble Snowfall Forecasts Using Operationally Used Snow-to-Liquid Ratios in Five Winter Storms

Abstract

Abstract The prediction of snow accumulation remains a forecasting challenge. While the adoption of ensemble numerical weather prediction has enabled the development of probabilistic guidance, the challenges associated with snow accumulation, particularly snow-to-liquid ratio (SLR), still remain when building snow-accumulation tools. In operations, SLR is generally assumed to either fit a simple mathematical relationship or conform to a historic average. In this paper, the impacts of the choice of SLR on ensemble snow forecasts are tested. Ensemble forecasts from the nine-member High-Resolution Rapid Refresh Ensemble (HRRRE) were used to create 24-h snowfall forecasts for five snowfall events associated with winter cyclones. These snowfall forecasts were derived from model liquid precipitation forecasts using five SLR relationships. These forecasts were evaluated against daily new snowfall observations from the Community Collaborative Rain Hail and Snow network. The results of this analysis show that the forecast error associated with individual members is similar to the error associated with choice of SLR. The SLR with the lowest forecast error showed regional agreement across nearby observations. This suggests that, while there is no one SLR that works best everywhere, it may be possible to improve ensemble snow forecasts if regions where SLRs perform best can be determined ahead of time. The implications of these findings for future ensemble snowfall tools will be discussed. Significance Statement Snowfall prediction remains a challenge. Computer models are used to address the inherent uncertainty in forecasts. This uncertainty includes aspects like the location and rate of snowfall. Meteorologists run multiple similar computer models to understand the range of possible weather outcomes. One aspect of uncertainty is the snow-to-liquid ratio, or the ratio of snow depth to the amount of liquid water it melts into. This study tests how common predictions of snow-to-liquid ratio impact snowfall forecasts. The results show that snow-to-liquid ratio choices are as impactful as the models’ differing snow rate or snow location forecasts, and that no particular snow-to-liquid ratio is most accurate. These results underscore the importance of better snow-to-liquid ratio prediction to improve snowfall forecasts.