Some Practical Considerations for Visualization and Operational Interpretation of Probabilistic Guidance from the Warn-on-Forecast System

A key practical challenge of ensemble forecasting is visualizing information from an ensemble that may contain dozens of individual members to provide a forecast for a specific weather event as well as a measure of uncertainty related to that forecast. Ensembles run at convection-allowing resolution for short-term prediction of thunderstorms and their associated hazards face a specialized set of challenges in producing useful visualizations of probabilistic guidance. For example, small amounts of ensemble spread in the location of simulated radar reflectivity associated with an individual thunderstorm will result in low predicted probabilities of thunderstorm occurrence at the grid scale and excessive smoothing in the ensemble mean, which produces an unphysical representation of the intensity of the storm. Therefore, neighborhood post-processing techniques are typically applied to convection-allowing ensemble forecasts to spread guidance to larger spatial scales. Probabilistic guidance is then visualized to provide a measure of event likelihood (e.g. probability of exceeding a specific threshold), potential severity (e.g. the ensemble maximum), or feature-based aggregate of individual member forecasts (e.g. spaghetti or paintball plots).

The experimental Warn-on-Forecast System is a convection-allowing ensemble designed to provide frequently-updating (every 30 minutes), short-term (0–6 hr) guidance of hazards within individual convective storms. The short duration and rapid issuance of WoFS forecasts presents additional challenges for producing useful probabilistic guidance products. First, the short duration of forecasts requires rapid generation and dissemination of probabilistic guidance products to be useful to operational meteorologists. Additionally, meteorologists making short-term predictions of hazards within convective storms are under time pressure to incorporate information from many different data sources to produce the best possible forecast. This time pressure requires information in WoFS guidance to be visually communicated to operational meteorologists in an efficient manner that minimizes the additional workload placed on the forecaster. This presentation will provide an overview of post-processing and visualization strategies to overcome these challenges associated with producing useful real-time, probabilistic WoFS guidance. Additionally, situations where specific products may provide value, or produce misleading interpretations, of the likelihood and potential severity of convective storm hazards are provided using examples from severe thunderstorm and flash flood events in central Oklahoma on 2 May 2018.