May 23 - July 29
Intra-Storm Tornado Warning False Alarms During Tornado Outbreaks
Ev Bohlmann (Valparaiso University), Dr. Matt Flournoy (OU/CIWRO & NOAA/NSSL), and Dr. Kenzie Krocak (OU/IPPRA & NOAA/SPC)
What is already known:
- Warning verification is one of several ways to assess tornado warning performance, alongside the Probability of Detection (POD) and lead time.
- A variety of factors can influence tornado warning performance on a daily scale. In particular, tornado warnings issued during outbreaks tend to perform better than those issued during isolated events, and warning performance on the first tornado of the day is generally worse than on subsequent tornadoes.
- Relationships between these factors and performance in terms of lead time and POD generally also exist on the storm scale in tornado outbreaks (i.e. first tornadoes from individual storms still have a lower POD and shorter lead times, etc).
- A high POD, longer lead time, and low False Alarm Ratio (FAR) are ideal, but these metrics are intertwined and there are tradeoffs.
What this study adds:
- The numbers of tornadic vs. non tornadic storms were nearly identical, as were the number of false alarm warnings issued on each, but the average number of warnings (verified or not) on a tornadic storm was more than double the average number of warnings on a non tornadic storm.
- A majority of the warnings issued on tornadic storms were verified.
- The first warning issued on a particular storm during a tornado outbreak is much less likely to be a false alarm, consistent with the established link between a low FAR/low POD
- Approximately 80% of false alarms on tornadic storms were issued after the first warned tornado, potentially indicating uncertainty regarding whether a previously tornadic storm will continue to produce tornadoes.
Abstract:
Tornado outbreaks carry the potential for catastrophic, widespread damage to property and loss of life, making timely, accurate warnings a necessity. Effective forecasting in these outbreak situations has resulted in a relatively high probability of detection and adequate lead times. A consequence of this, however, is the common issuance of “false alarm” warnings (meaning, a tornado does not occur within the warning area during the valid time). The approach to assessing warning performance has typically been on a larger scale. None so far have assessed the occurrence of false alarms on the scale of an individual storm within a tornado outbreak. We were particularly interested in where “false alarm” warnings tend to occur within a storm’s lifetime (e.g., whether they are more common before the first tornado that the storm produces, after the last, or between two separate events). By matching both verified and unverified warnings issued during tornado outbreaks to individual storms, we found that the majority of false alarm warnings were issued after the first tornado that a storm produces (both between tornadoes and after the last), while the first warnings on each storm had a very low false alarm rate. However, storms that remained nontornadic had fewer warnings issued on average than tornadic storms. This shows a lack of understanding of the phases before a storm produces its first tornado and after it has produced its final tornado. Future research could focus on these stages in order to improve forecasting skill and therefore warning performance.