A Vision for the Development and Implementation of the Warn-on-Forecast Concept (Invited Presentation)

Deterministic modeling approaches to forecasting severe weather presume a single solution for the evolution of storms and the environments in which they form. Similarly, deterministic severe weather warnings assume severe weather impacts are equally likely within the warning’s spatial and temporal extent. These approaches fail to represent forecast uncertainty that we know exists owing to the chaotic nature of the atmosphere, imperfect observations, imperfect prediction models, and other causes. Advancements in data assimilation, ensemble modeling, high-performance-computing, and probabilistic forecast and verification methods provide the means and opportunity to better represent these uncertainties within severe weather prediction systems. These advancements support the emerging paradigm shift from a deterministic to probabilistic approach to forecasting, called FACETs (Forecasting a Continuum of Environmental Threats).

The Warn-on-Forecast project at the National Oceanic and Atmospheric Administration (NOAA) National Severe Storms Laboratory in Norman, OK, USA, aims to improve short-term severe weather forecasts, warnings, and decision support for high-impact events (e.g., tornadoes, hail, wind, and flooding) by leading and collaborating with others on convection-scale research and development activities that enable convection-resolving, ensemble model forecasts become a key resource for NOAA National Weather Service (NWS) operations, especially between the watch-to-warning temporal and spatial scales. The result of these efforts is the development of a prototype Warn-on-Forecast system called the NSSL Ensemble Warn-on-Forecast System for ensembles (NEWS-e).

NEWS-e is a frequently updated, regional-scale, on-demand convection-permitting ensemble analysis and prediction system, nested with an hourly convection-allowing ensemble forecast system. The 2017 version of this system assimilates radar, satellite, and surface data every 15 minutes, and generates new probabilistic 3-hour and 1.5-hour forecasts at the top and bottom of each hour, respectively, at grid spacing O(~3 km). This data assimilation system uses the advanced research version of WRF, version 3.8+ (ARW) to produce storm-scale ensemble analyses and forecasts. Details of the system configuration will be provided within the presentation.

NEWS-e seeks to improve 0–3-h predictions of individual convective storms and mesoscale aspects of convection that provide enhanced probabilistic forecast guidance. Key to the usability of this guidance are post-processing, visualization, and verification methods developed and tested through researcher-stakeholder interactions. Forecast swath products, such as exceedance probabilities and ensemble percentiles of key predictive parameters (e.g., reflectivity, accumulated rainfall, updraft helicity) and real-time object-based verification are expected to revolutionize forecasters’ ability to anticipate not only storm location, mode, intensity, but also high-impact threats and their impacts on society. Toward this end, this prototype system is being tested in real time during peak severe weather season in the U.S. Following each storm season, findings from rigorous quantitative and qualitative case study evaluations are used to direct system enhancements. Example cases and the associated forecast verification will be shown. Additionally, next steps, challenges, and paths to operational implementation will be discussed.