Can Varying Snow-Liquid Ratios Improve Probabilistic Snowfall Forecasts? (287)

Philip Schumacher, NOAA/National Weather Service, Sioux Falls, South Dakota

 

Abstract:

Snowfall prediction has two areas of uncertainty – the amount of precipitation that will fall at a location and the snow-to-liquid ratio (SLR) of the snowfall. Since the winter of 2015-16, the he National Weather Service have been producing probabilistic snowfall forecasts. The probabilities are determined from an ensemble of model solutions of forecast precipitation but only one SLR is applied to each model solution. Previous research has shown that the SLR can vary from as low as 5:1 to greater than 50:1 and is dependent upon various factors including temperature in the snow-producing layer and wind speed. Using one SLR per ensemble member may underestimate the probability of different snowfall amounts by not taking into account the uncertainty that exists in forecasting SLR.

This study examines whether addressing uncertainty in both SLR and precipitation improves the probabilistic forecasts for snowfall. Two tests are done. The first test uses past observed snowfall events to create an idealized ensemble of precipitation amounts. To create an ensemble from the observed precipitation, each precipitation forecast is randomly selected from a gamma distribution with a mean of the observation. Two ensembles of snowfall are created from this data set. The first ensemble uses one SLR for each forecast and the second uses several randomly selected SLRs for each forecast. The second test uses SREF forecast precipitation from the winters of 2016-17 and 2017-18. As with the idealized example, one SLR is randomly selected for each SREF member and multiple SLRs are randomly selected for each SREF member. For both tests, the reliability of the ensemble using a single SLR is compared to one using multiple SLRs.