38 Probabilistic Tornado Warning Plumes

Monday, 3 November 2014
Capitol Ballroom AB (Madison Concourse Hotel)
Mark J. Mitchell, NOAA/NWS, Pleasant Hill, MO
Manuscript (322.3 kB)

Handout (1.1 MB)

In 2012 the National Weather Service (NWS) office in Pleasant Hill, Missouri participated in the Impact-Based Warning project. A review of the project was conducted where NWS forecasters were asked to set aside the current NWS text warning system and brainstorm how severe weather threats could best be conveyed to customers. One of the suggestions was to create a grid of probabilities for severe weather threat in the path of the storm. This grid set could portray forecaster confidence and diagnosis for a range of different threats and impacts. Although the tools to do this directly are not currently available to NWS meteorologists, there is enough information in NWS tornado warning text to make some general assumptions regarding threat levels. Specifically, using the forecaster defined initial storm location and motion, threat levels downstream can be determined. A climatology was developed using NWS tornado warnings and tornado reports from late 2007 through 2013 for all NWS forecast offices. Warnings are separated into bins based upon initial storm speed. A comparison of the distribution of tornado reports relative to the initial storm location is used to calculate a high-resolution grid of strike probabilities in proximity to the storm. Applying regression techniques to the grid smoothes the probability grid, which in turn can be used to produce a warning plume highlighting the highest threat area. Calculation of the climatological tornado threat plumes is presented along with basic verification statistics for comparison against NWS warning polygons. Although skill scores for the NWS warning polygons were better when compared to the warning plumes, there was significant improvement noted in the warning plumes in reduction of “false alarm” area.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner