Communicating Forecast Uncertainties: Lessons Learned from the NWS Probabilistic Snowfall Experiment

Communicating forecast uncertainties associated with winter weather is a vital component of providing impact-based support to decision makers. During the past 5 winters, National Weather Service forecast offices have been producing and disseminating experimental probabilistic snowfall forecasts, with this effort expanding to 82 NWS offices during the 2017-2018 winter season. Probabilistic ice accumulations were also produced for the first time last winter for internal evaluation, and starting during the spring of 2018, experimental probabilistic QPF also began production at 7 test offices. Common output products and web displays have been designed for all three probabilistic precipitation suites to facilitate consistency in messaging as well as user interpretation and comprehension.

Much has been learned during the probabilistic snowfall forecast experiment about the communication and utility of snowfall uncertainty information. User surveys, feedback from core partners, and interviews and focus groups conducted by Eastern Research Group (ERG) have yielded considerable insights on user needs for winter weather information, and how uncertainty information is understood and potentially applied. Two important and related themes have emerged from these activities. First, consideration of the presentation media used is a critical aspect of effectively messaging uncertainty information. For example, adaptive approaches are needed for automated static web page graphics, vs. forecaster annotated graphics in published presentation slides and social media posts vs. live interactions such as webinar presentations and on-site IDSS forecaster presentations. Second, the value of placing uncertainty information in the appropriate context for the audience and the situation is vital. Such contextual enhancements can include additional explanations of the uncertainty including temporal and spatial differences across a forecast area, descriptions of forecaster confidence, and alternative scenarios, as well as combinations of each of these. Consideration of these factors appears to enhance user comprehension and application of uncertainty information. While considerable improvements have been made to the probabilistic snowfall products, presentation best practices are emerging and many challenges remain in communicating winter weather uncertainty information, including the subtle but important distinction between forecast uncertainty and forecaster confidence.