In part 1 of our investigative series, we examined 67 years of US tornado data to address the notion that tornado seasons have lately been disappointing for storm chasers. In addition to fewer tornadoes occurring between 2012 and 2016 than the prior 5 years, our research suggests that during the latter period tornadoes were generally more difficult for storm chasers to spot.
This “difficulty to spot” is based on our anecdotal view of when and where storm chasers tend to chase. Our view of the when was fairly straightforward as described in part 1.
“We’ll consider tornado season in America as March through June, outside of which tornadoes are already too infrequent to promote disappointment. As for the time of day, noon through 9 PM is the ideal storm chasing window, as anything later is probably too dark to observe and anything earlier is likely occurring within a high-precip convective system.”
Although these time and date windows are very crude to say the least, it offered a convenient way to filter away tornadoes that the hoards of chasers aren’t likely pursuing. As for the where, we developed the concept of Chasecationland.
The ideal chasecastion region hinges on where tornadoes most frequently occur during peak season (i.e. Tornado Alley) as well as where the terrain is most compatible for tornado viewing.
In an even more crude fashion, we semi-seriously defined Chascationland as the footprint of the old Big XII conference. Although the results in part 1 showed a decline in chaseable (i.e. the when) tornadoes occurring in Chasecationland, it is unlikely that this decline was uniform across the region. In addition, other areas of America may have experienced increases in chaseable tornadoes. The goal of part 2 is to map the trends in chaseable tornadoes in order to identify the spatial sources of storm chaser disappointment as well as offer potential tornado prominent alternatives to Chasecationland.
Starting with the full SPC tornado dataset (1950-2016), we examine the expected annual number of chaseable tornadoes (Noon-9pm LT, Mar-Jun) occurring within a 50 mile radius of each county’s centroid. 50 miles seems like an appropriate distance threshold. It’s broad enough to smooth out the luck factor at the county scale while anything further would be difficult for a storm chaser to reach from their initial target location.
It’s worth noting that this long-term average includes the early years where observed tornadoes were less frequent, not because they occurred less frequently, but because there were less people in the area to spot them. Although it is unclear when exactly the likelihood of observing a tornado caught up to its current probability, we can be certain that sampling the most recent 30 years will more closely represent the true tornado occurrence rates. Furthermore, this 30-year period, which somewhat arbitrarily became the golden standard for climate normals, filters out tornado seasons that are too old to shape the opinions of today’s storm chasers. I highly doubt any of us were chasing storms over 30 years ago.
Although the data becomes more splotchy thanks to the reduced sample, we see a significant increase in tornado density east of the Rockies, particularly in rural areas. Compared to the surrounding regions, the Ozarks still exhibit a noticeable lack of tornadoes indicating that either the storm spotters or the tornadoes themselves want nothing to do with the hilly woodland topography.
Overall, the data suggests that our idea of Chasecationland should be shifted slightly northward, but it’s fair to assume that the historical bulk of chaseable tornadoes occur within the broadly perceived Tornado Alley. The darker orange/red shades are where storm chasers commonly flock to and ultimately judge each tornado season based on how many tornadoes they see.
If we assume that the 30-year average density aligns with storm chaser expectations, we can measure their disappointment levels using the simple meme formula:
(Eq. 1) Disappointment = Expectation – Reality.
The 30-year average of chaseable tornadoes for each county is the expectation while the 5-year averages (2007-20011, 2012-2016) are the realities for the periods in which we wish to measure disappointment.
One unfortunate artifact of our formula is that “DOMINATING!” isn’t the best adjective choice to describe the April 2011 super outbreak over the Southeast. Instead, let’s focus on Chasecationland where aside from a few pockets, storm chasers were generally rewarded during their 2007-2011 expeditions.
The qualitative visual difference between the 5-year periods is fairly remarkable, with the latter period showing much larger areas of dissapointment. Although a nice swath of above-average tornadic activity exists in the highly-chaseable plains of N. TX/W. OK/W. KS, nearly all remaining areas of chasecationland struggled to meet their annual hype further reinforcing the rationale behind chaser disappointment.
Referring back to the “what have you done for me lately” mentality of storm chasers that we discussed in part 1, it’s likely that expectations during the 2011-2016 period were influenced by not only the long-term averages but also by the previous 5-year period’s results. With this in mind, we can tweak our disappointment formula to incorporate both contributors to expectation. Our expectation now weighs the average annual chaseable tornado density (AACTD) for both the long-term and 2007-2011 periods, which can be expressed as:
(Eq. 2) Expectation = (AACTD30-year + AACTD2007-2011)/2.
Along with the updated expectation, applying the 2012-2016 AACTD as the reality in Eq. 1 provides a recency-biased view of chaser disappoint.
Aside from Southwest Oklahoma and a few smaller pockets, that’s a plethora of storm chaser disappointment across Chasecationland. The fact that there’s still green on the map however, indicates that expectations can be exceeded even during periods of overall chaseable tornado decline. Although the task becomes more difficult, and perhaps the RadarScope dots more concentrated, the more dominant storm chasers will continue to dominate.
Special Thanks to Rob Hammond for Python coding assistance with generating the underlying mapping data.