2022 tornado statistics in context

[adlyons]

According to SPC (Tornado FAQ #Tornado Climatology and Data):
Tornado reports have increased, especially around the installation of the NEXRAD Doppler radar system in the mid 1990s. . . . The increase in tornado numbers is almost entirely in weak (EF0-EF1) events that are being reported far more often today due to a combination of better detection, greater media coverage, aggressive warning verification efforts, storm spotting, storm chasing, more developmental sprawl (damage targets), more people, and better documentation with cameras (including cell phones) than ever. . . . To compare tornado counts before Doppler radars, we have to either adjust historical trends statistically to account for the unreported weak tornadoes of before, or look only at strong to violent (EF2-EF5) tornadoes, whose records are much better documented and more stable. When we do that, very little overall change has occurred since the 1950s. . . .

Great Summary Pierre. Yes, there has been an uptrend in tornado reports with the advent of doppler radar, population growth, and different reporting practices. The largest change is in weak tornadoes which were previously underreported and interestingly enough that increase isn't uniform over the whole US. The NCEI report database is problematic the farther back in time we go we can chat about that If youd like as it is a very well-known problem here. The data on the graphs I posted are not adjusted for significant trends. I did try to keep any trends minimized by selecting the time period and geographic areas over the most stable portions of the archive. This isnt meant to be rigorously scientific by any means. This was just an interesting side project to look at multiple years and gauge activity levels quickly. I have a formal journal article in the works detailing some better methods and data to estimate tornado occurrence. Here's a pre-print using some of the data. https://www.spc.noaa.gov/publications/lyons/tormodes.pdf.

 

[adlyons]

Jason,

I'm not 100% sure, but I don't think I'm being unreasonable in inferring that the numbers that Adlyons is going by in the chart are the official SPC counts. To my knowledge, there is no available breakdown in *method* of reporting (ham, public safety, Twitter, etc.). As Jeff mentioned, technically almost all reports are crowdsourced, as few if any reports come from degreed mets in the field during the course of their duties. On top of that, many tornadoes get reported by multiple methods (example: Tornado A2 is called in on ham, but also reported by LE/Fire, and yet another report comes in from Twitter). While some LSRs do note the method by which it was reported if it's a single report, it does not break down if several reports came in on the same storm. And there are many days where the overall number comes down because they're able to ascertain during the surveys that x number of reports actually involved the same tornado.

As far as mobile vs stationary, I suspect it would be significantly more difficult to break that down even over trying to get a rough estimate on method of reporting. Unlike the official SPC report, those breakdowns would be at best, educated guesses, and possibly nothing more than spitballing (or even pertinent in the grand scheme of things).

Drew you were correct. This data is the official counts from SPC but also certified in NOAA NCEI Storm Data.

 

[Jason N]

thanks for the additional feedback and information Adlyons!

 

[Pierre Ducray]

The largest change is in weak tornadoes which were previously underreported and interestingly enough that increase isn't uniform over the whole US. The NCEI report database is problematic the farther back in time we go we can chat about that If youd like as it is a very well-known problem here.

1. Could there be a correlation between that non-uniform increase and population density?... For example the Plains v the Southeastern region...

2. Anyway it seems that there's another bias as well: Supercell tornadoes are much stronger and wider than damage-based ratings indicate.

And in this 2021 PNAS paper, this (not so) amazing sentence: 'It is no exaggeration to state that, until now, statistics concerning even the most basic characteristics of tornadoes, including intensity and size, could not be quantified with confidence.'

3. About your pre-print (I like it!) 'detailing some better methods and data to estimate tornado occurence', it looks like a sort of new contribution to 'RIP Tornado Alley'?!