BEWARE OF TORNADO NUMBERS
Before I get into the recent fascinating and troubling story on tornado warnings, let me tell you a story -- a true story -- from my days with the National Weather Service.
I interned at the National Hurricane Center in Miami in the 1970s (not a typo). Part of the office housed the local National Weather Service office that covered all of South Florida. My boss (the Meteorologist-In-Charge) was in charge of some things for the entire state, which included keeping official statistics on Florida tornadoes.
This guy (who shall remain nameless, partly because I can’t remember his name) had an unusual philosophy about tornadoes. He didn’t believe in simply inspecting storm damage and determining whether it was a tornado. We only did that with the “bigger” ones, which in Florida means practically no tornadoes. His orders were: “It’s a tornado unless we have evidence that it wasn’t." So, if there was damage, and we heard that a witness saw a funnel, it was a tornado. If we read a newspaper article the next day that quoted someone saying, “It sounded like a freight train," it was a tornado. Or if any local official told us so, it was a tornado.
Surprise, surprise! Florida ranked second in the country in the number of official tornadoes for much of the 1970s. But did it really? I assume other NWS offices had different philosophies, so how can we accurately compare tornado numbers? It’s not like comparing temperatures, rain, snow, wind, etc.
NOW TO THE TROUBLING TORNADO STATS…
Again, keep in mind that tornado statistics are not precise. If one knocks over a few trees in the woods, it probably won’t be counted. And determining whether it was a tornado is not an exact science. I should know-in my later years with the NWS in Georgia, I was the one responsible for inspecting damage and making determinations. The weaker the tornado, the harder it was.
An article from Jason Samenow of the “Capital Weather Gang” at the Washington Post showed the numbers, and the study that had the bad news.
So, only 58% of tornadoes in the U.S. were detected before touching down, compared to 75% only a few years ago.
Are they being more selective in issuing warnings, in order to minimize the problem of “false alarms?" There’s not much evidence of that:
The false alarm rate (FAR) hasn’t dropped much. Note that 7 out of 10 tornado warnings are “false alarms”- that’s a problem.
Is the NWS becoming bolder in issuing warnings in order to detect more of them? If so, the warning time should be going up. But it isn’t-it’s going down.
The “lead time” is down to only 8 minutes, compared to a record 15 minutes in 2011.
WHY IS THIS HAPPENING?
These numbers are an unpleasant surprise. They should be improving, with the new generation of “Dual-Pol” Doppler Radars. Warnings for severe weather are the highest priority of NWS. Here is a theory from Harold Brooks, an expert on this subject:
“Forecasters have faced increased pressure to reduce tornado ‘false
alarms’……’motivating some forecasters to issue fewer warnings’
There is a little evidence of that, with FAR dropping a few percent. But there may be other issues:
1. Tornado warning areas may be shrinking in recent years, as forecasters try to get more precise with warnings. In the old days, entire counties were included, which would result in fewer “false alarms." If they narrow the warning area and still reduce the FAR, that is much better news.
2. Have a lot of older, experienced forecasters retired in recent years, leaving less experienced ones issuing the warnings? As an older, experienced forecaster myself, I would understand that theory.
3. There have been more small and “weak” tornadoes in recent years. They are harder to detect, and hard to give more than a few minutes warning. But Brooks says “lead times have dropped by 3-4 minutes for all intensity scales.”
4. The sample size is too small to definitively prove the changes. We’re talking about tiny storms over a huge country.
5. All tornado statistics are questionable, due to changes in verification philosophy across the country and from time to time. Maybe we’re comparing “apples and oranges." I hope so.
HELP-JUST IN TIME!
Less than ten days ago, the President signed a bill into law to improve forecasts and warnings in the future. “The Weather Research and Forecasting Innovation Act” is concentrated on getting better severe weather warnings for tornadoes and hurricanes, among others. It aims to improve data, computer models, and other research. This should lead to longer lead times and lower false alarms in the future. And maybe it will even help the U.S. computer models catch up with the world-leading European model.
There was strong bipartisan support for this bill, which shows how important better forecasts and warnings are to EVERYONE in ALL PARTS OF THE COUNTRY. I don’t think we’ll ever see one-hour lead times for tornado warnings. Most tornadoes are too small and short-lived for that. But getting back to that 15 minute lead time would be a great start.