Living in Missouri as a kid, John Long grew up with tornadoes.
He went through the same tornado drills that all school children from tornado-prone parts of the country know well: Filing into school hallways and crouching against walls with a textbook or hands covering the head. Tornadoes were a part of life.
But growing up, Long said, he and his schoolmates knew that they weren’t likely to see a tornado while classes were still in session. June, after schools had let out for the summer, was when tornadoes came to his area of western Missouri.
Long, who hasn’t lived in the area in nearly three decades, hadn’t thought much about tornadoes or school drills or finding cover. That is until an EF2 twister hit the city of Branson, Mo., during the Leap Day outbreak of Feb. 28-29, 2012, much earlier than Long recalled from his childhood.
“I just remember reading that and thinking, ‘Wow, that’s really early’,” he told Climate Central.
Long talked to older relatives in Missouri to canvas their memories. Along with a healthy dose of “when I was your age” and “uphill both ways” comments, they said they felt tornadoes were happening earlier today than when they were young.
That set Long, whose work at Montana State University is normally in remote sensing, off on a side project to see what the tornado data said about the timing of the peak of tornado season in Tornado Alley, which was defined as Kansas, Oklahoma, most of Nebraska and northern Texas. (Unfortunately for Long’s relatives, Missouri had to be left out because different parts of the state fall into different tornado regions, with, for example, its north part of the so-called Hoosier Alley and its south part of Dixie Alley.)
“I just started poking around,” he said, “And pretty soon, the story just popped out.”
What he and co-author Paul Stoy found was that the peak of the tornado season had shifted seven days earlier over the previous six decades. The shift was even larger — up to two weeks — when the weakest tornadoes were excluded, and for particular states.
Of course, the key question — what’s causing the shift, including the possibility of effects from global warming — remains unanswered, but ripe for further study.
Of course, looking for trends in tornado data is fraught with potential pitfalls owing to changes in how tornadoes have been reported over time. Tornadoes, especially weaker ones, are much more likely to be reported now, because of larger populations in affected areas and the ability to spot tornado signatures on radar. The more rigorous methods used to rate tornadoes now are only a few decades old, making it hard to compare tornadoes now to those in the middle of the 20th century and earlier.
But, Long and Stoy reasoned, these factors shouldn’t affect the reporting of when tornadoes happen, as there’s no reason why people would be more likely to see tornadoes in April than May.
With that assumption, a reasonably sound one according to tornado experts not involved in the study, Long and Stoy set about using tornado observation data gathered by the National Weather Service’s Storm Prediction Center to look for shifts in the peak of activity.
The rhythms of the tornado season are familiar to those growing up in the broad swath of land from the Midwest down through Texas and through the Southeast. While “tornadoes can happen any day of the year,” Stoy said, the season tends to ramp up starting in April, first in the Southeast, then moving westward and northward as spring and summer progress.
The probability of a tornado occuring on a given day somewhere in the U.S.
The national average peak is June 12, but the peak in particular regions can be anywhere from early May to early July, when warm, moist air from over the Gulf of Mexico can venture northward and clash with other air masses, creating an unstable atmospheric environment.
Focusing on the Tornado Alley area, Long used a statistical tool not used often, but one that he says is perfect for looking at trends that vary with the seasons, called circular statistics.
Data like the number of tornadoes per day would typically be pictured as a line, from the number on day 0 (Jan. 1) to the number on day 365 (Dec. 31), with no relationship assumed between tornado occurrence on the start and end dates. But, of course, the tornado environments on Jan. 1 and Dec. 31 are very much related, and circular statistics allows researchers to capture how such relationships wrap back on themselves, as in a circle.
What they found was that “it’s a one-week shift when we look at the whole area and we include all the tornadoes,” Long said. The results were detailed in the journal Geophysical Research Letters in early September.
When EF0 tornadoes, the weakest rating on the Enhanced Fujita scale, were removed in a preliminary analysis not published in the paper, the shift broadened to nearly 14 days. Removing Nebraska from the data also produced a nearly two-week shift in the remaining states.
Of the shift, Stoy said, “I think it’s actually quite profound . . . this is over the course of people’s lifetimes.”
Greg Carbin, a warning coordination meteorologist with the SPC, said the findings were “very interesting,” especially coming from non-climatologists, and had sparked discussion in the tornado research community. (Carbin did a basic analysis using data that is independent of the tornado reports used in the study and found a signal “that would suggest that their findings are valid,” he told Climate Central.)
The shift in the timing of the peak of tornado season in Tornado Alley (defined here as Kansas, Oklahoma, most of Nebraska and northern Texas). The 'J' at the top of the clock feature is January.
Credit: Long & Stoy, 2014
What remains to be determined is exactly what is causing the shift in peak tornado activity. The study looked at some possibilities, but other than an apparent link between El Niño and tornado activity in Oklahoma, couldn’t find any links to major climate cycles, such as the Pacific Decadal Oscillation or the Atlantic Oscillation.
Stoy and Long hope that scientists more steeped in tornado and climatological knowledge will take their results and run with them to find what might be behind the shift, including the possibility that global warming is playing a role.
Carbin and his colleague Harold Brooks, a senior scientist with the National Severe Storms Laboratory and a reviewer of the paper, have both been working on looking for trends in tornado data and the possible explanations behind them, including warming.
Of the trend Long and Stoy found, both said that warming was a possibility, as it could be moving forward the timing of the return of warm, moist air from the Gulf of Mexico that is a key to tornado formation, Carbin said.
“I can see why this would be occurring with a warming climate,” Carbin said. But “I don’t want to completely commit to the fact that it’s all to do with warming.” As Brooks noted, the overlapping record of tornadoes and global warming is a short one, which makes finding a relationship harder.
Regardless of when the peak of activity occurs, devastating tornadoes can happen far before or after that peak, Carbin said. So in terms of individual outbreaks, “I think it’s dangerous to take too much away from a study like this,” he said.
And as with all tornado trend studies, the biggest problem is being able to say something about the activity of a given year ahead of time, Carbin said, “because ultimately what we want to do is try to be able to predict some of this stuff.”
This article is reproduced with permission from Climate Central. The article was first published on September 24, 2014.