May 22 - July 28
How Forecasters Anticipate Nocturnal, Cool-Season Southern Tornado Events
David Nowicki, Ariel Cohen, Alan Gerard, Kim Klockow, and John Hart
What is already known:
- Steps towards the probabilistic forecast and warnings of Warn-on-Forecast have been taken using a variety of tools, including improved radar detection and novel ensemble numerical weather prediction models.
- Nocturnal, cool-season tornado events in the southeast United States are difficult to forecast/warn for, for a plethora of reasons.
- The public in this area of the country are particularly vulnerable to these events and would benefit by having a longer warning lead time.
What this study adds:
- This study offers insight into what tools/measures forecasters are using currently when in the kinds of situations described above.
- Forecasters’ beliefs regarding the potential utility of the Statistical Severe Convective Risk Assessment Model as well as forecasters’ propensity to use such a model, were it available, are shown.
- The existence of a gap between forecasters’ general knowledge and their ability to use SSCRAM to improve their forecasting accuracy in these situations shows a need to further examine current and potential use of conditional probabilities in forecasting.
Abstract:
In this study, forecasters across the Southeast were surveyed to find out current practices/tools used when issuing forecasts and warnings during nocturnal (0300 UTC-1200 UTC), cool-season (November through May) tornado events. Additionally, forecasters were asked to rate personal beliefs regarding the possible forecasting utility added by novel statistical models as well as beliefs about potential personal use of such models both before and after viewing output from an existent statistical model. Readings from the well-calibrated, climatologically based Statistical Severe Convective Risk Assessment Model (SSCRAM) were shown to forecasters halfway through the survey to serve as the treatment for the sample. SSCRAM output was analyzed by Bunker (2017) to discover conditional probabilities of tornado occurrence – given certain environmental parameters in varying ranges – for this region during this specific kind of event. SSCRAM averages of conditional probabilities of tornado occurrence were found for six parameters in specific ranges; then, these averages were compared to forecasters’ subjective estimates of conditional probabilities, given the same parameters in the same ranges. Results of the study show a gap between forecasters’ knowledge and their calibration with the environment as well as a shift in personal beliefs regarding SSCRAM’s potential utility and use after being shown an example of that model’s output.