Starting as pilot efforts at WFOs Sterling VA and Taunton MA during the winter of 2013-14, the NWS probabilistic snowfall experiment has grown to involve more than 60 NWS forecast offices across all parts of the CONUS during the winter of 2016-17, with additional expansion planned for 2017-18. The forecast methodology has evolved considerably during the last several winters. The approach utilizes NCEP's Weather Prediction Center's (WPC) probabilistic snowfall percentile accumulation values, generated from an ensemble of NWP snowfall forecasts, to create a cumulative probability distribution function with the official NWS WFO snowfall forecast as the mode of the distribution. Tenth and 90th percentile graphics, exceedance probabilities, and categorical range probabilities for a variety of thresholds are output and made available to end users for decision making. The 10th and 90th percentile values are communicated as the minimum (expect at least this much) and maximum (potential for this much) snowfall, respectively, while the official (e.g., NDFD) NWS WFO snowfall is communicated as the most likely snowfall amount.
Through forecaster and user feedback, including a formal social science study this past winter, much has been learned about how to effectively create, display, and communicate probabilistic snowfall information. This presentation will discuss how the methodology, graphics and web page presentation, and decision support messaging of the snowfall probability has evolved during the last several winter seasons. Emphasis will be on the impact of the changes made for the 2016-17 winter season and the resulting user feedback. Some 2016-17 season verification results will also be presented. Plans for the 2017-18 winter season, linkages to other experimental probabilistic forecasting efforts, and the potential for a national operational implementation for the probabilistic snowfall forecasts will also be discussed.