Definition of Dry Thunderstorms for Use in Verifying SPC Fire Weather Products

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Sunday, 4 January 2015
Paul X. Flanagan, Oklahoma Climatological Survey, Norman, OK; and C. J. Melick, J. W. Rogers, I. L. Jirak, A. R. Dean, and S. J. Weiss
Manuscript (1.6 MB)

Handout (1.3 MB)

Forecasts for fire weather are performed on a routine basis by the National Weather Service (NWS) Storm Prediction Center (SPC). The SPC provides outlooks, out to eight days, predicting the likelihood of dry thunderstorms within the prescribed area. This study evaluates several classification methods for identifying cases of dry thunderstorms from observational data in order to develop a more specific definition that can be used to verify fire weather outlooks. Observations from 148 days in 2013 were analyzed using ten various combinations of thresholds from QPE, lightning flash data, and other environmental parameters. For the evaluation, a 2x2 contingency table was created by comparing the events classified by each method and the initial Day 1 SPC dry thunderstorm outlooks. Traditional forecast verification statistics (e.g., critical success index [CSI]) were then calculated for individual days as well as in an accumulated sense for the entire period considered in the investigation.

Results showed that by including environmental parameters into the classification process, an increase in the restrictiveness of the method was found. This meant that dry thunderstorm events were not defined in the observations in areas where dry thunderstorms were not forecast to occur. However, this improvement was often offset by increases in false alarms given the reduced number of observed events found within the outlook areas. Despite the aforementioned drawback, statistical analysis still showed that these more realistic approaches tended to outperform methods that only used lightning flash data and QPE. Finally, a neighborhood diagnosis was also pursued to ensure that isolated observed events were not considered in the verification, which more closely follows the coverage criteria used in SPC fire weather outlooks.