Analysis of and Ways to Improve Tornado Warning False
Alarms at NWS Louisville (107)
Ryan Sharp, National
Weather Service, Louisville, KY
Zack Taylor, National
Weather Service
Theodore Funk, National
Weather Service
Abstract:
In 2017, National Weather
Service (NWS) Louisville (LMK) took a detailed look at our tornado warning
statistics over the past ten years, centered on false alarm rate (FAR). Our FAR
has been high for many years, in part due to the more tenuous quasi-linear
convective system (QLCS) tornadoes that frequent the Ohio Valley. We assessed
county-by-county statistics as well as larger geographical zones to determine
where FAR needs the most improvement across the forecast area. We also studied
the effect of human factors on the warning decision process as determined from
a local forecaster survey. Potential ways to improve statistics and overall
warning decision support were discussed, including recommendations on warning
polygon vs. county shape.
The study also stressed the
importance of a mesoscale analyst to assess hourly near storm environmental
data and to consider whether certain parts of our forecast area were more prone
to tornado development. Previous research at LMK correlated Storm Prediction
Center mesoscale data sets (e.g., Effective Storm Relative Helicity, etc.) with
tornado warnings in the Ohio Valley. A tool was developed at LMK to integrate
this research into real-time events. Given the multitude of existing and new
data sets (e.g., GOES-16 and MRMS), we recommended employing 2-person warning
teams to monitor mesoanalysis data and issue warnings.
This presentation will
highlight the procedures outlined above, including lessons learned from our
office's severe weather roundtable discussion in October 2017 and case studies
since then.