From the American Geophysical Union | WASHINGTON, D.C. – Peak tornado activity in the central and southern Great Plains of the United States is occurring up to two weeks earlier than it did half a century ago, according to a new study whose findings could help states in “Tornado Alley” better prepare for these violent storms.
Tornado records from Nebraska, Kansas, Oklahoma, and northern Texas – an area of high tornado activity dubbed “Tornado Alley” — show that peak tornado activity is starting and ending earlier than it did 60 years ago.
Peak tornado activity, which occurs in the region from early May to early July, has moved an average of seven days earlier in the year over the past six decades. The study’s authors observed the shift in tornado activity for all categories of tornadoes that occurred in the region from 1954 to 2009.
The research team published its findings last week in Geophysical Research Letters, a journal of the American Geophysical Union.
Additional, more-selective analyses by the authors show that for some states in the region and for stronger tornadoes the season advances an average of 14 days compared to 1954.
“If we take Nebraska out [of the data], it is nearly a two-week shift earlier,” noted John Long, a research scientist in the Department of Land Resources and Environmental Sciences at Montana State University in Bozeman, Montana, and lead author of the new paper. For tornadoes rated above F0, the lowest rung on the original Fujita scale of tornado strength, the shift is also close to 14 days, according to a preliminary analysis by Long and his colleagues that’s not included in the new paper.
F1 tornadoes have winds between 117 and 180 kilometers per hour (73 and 112 miles per hour), while the strongest tornadoes, F5, have winds between 420 and 511 kilometers per hour (261 and 318 miles per hour), according to the original Fujita scale. Although the Fujita scale was updated in 2007, Long and his colleagues stayed with the original Fujita scale because most data in this new study originates from prior years.
The new research does not attribute the shift in tornado activity in the region to any single cause. However, the earlier tornado activity seen in the study is in-line with what could be expected in a warmer climate, the study’s authors said.
The new research could help residents in the region be better prepared for severe weather, said Long. About 1,300 tornadoes hit the U.S. every year, killing an average of 60 people, according to the National Weather Service’s Storm Prediction Center. This year, the majority of the 309 tornadoes that have hit the U.S. occurred in May and the deadliest storms were in April, according to the Storm Prediction Center.
“From a public safety perspective, if this trend (of an earlier tornado season) is indeed occurring, then people need to begin preparing for severe weather earlier in the year,” said Greg Carbin, the warning coordination meteorologist at the Storm Prediction Center in Norman, Oklahoma, who was not involved in the new study.
The new research analyzed National Weather Service tornado data for Tornado Alley from 1954 to 2009. The authors broke the data into ten-year time frames and analyzed how the dates of peak tornado activity changed over time.
The analysis showed the date of peak tornado activity in the region moved earlier at a rate of 1.55 days per decade over the time period studied. In the heart of Tornado Alley, an area with the highest density of tornadoes, peak activity shifted by seven days: from May 26 in the 1950s to May 19 in the early 2000s.
Although there is a consistent movement in the region toward earlier tornado activity, it is difficult to pinpoint a cause, said Paul Stoy, assistant professor in the Department of Land Resources and Environmental Sciences at MSU and co-author of the new study. Records of tornado activity in the U.S. only date back to the 1950s, making it difficult to study changing trends in tornado activity. Furthermore, tornadoes can be influenced by many regional factors, including topography of the land and areas where cooler air meets warm, subtropical air, making it difficult to attribute the shift in the tornado season to any one factor, he said.
Carbin, of the Storm Prediction Center, said a warmer climate might play a role. “If winters are not as cold, or if spring times are warmer, the location of the jet stream is most likely displaced north of where it has been in the past,” he said. This would cause tornado activity to shift earlier in the year, like what is seen in the new study, Carbin said.
The study has revealed a connection between one global climate pattern and tornado activity, specifically in the state of Oklahoma. When El Niño conditions occur between January and April, peak tornado activity in Oklahoma occurs earlier in the spring, the researchers report. El Niño, an oscillation of the ocean-atmosphere system that is associated with warm ocean waters in the Pacific Ocean, changes the air surface pressure and atmospheric circulation.
“The relationship we do see in Oklahoma is a light but significant connection to El Niño,” Stoy said. “This makes one suspect that if global climate change is changing these larger circulations, then there is a connection between a global [variability] and tornado activity.”
###
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
So they computed a downward shift in the mean start and end time statistic. That’s fine. But I would guess that this data (like most climate data) is somewhat noisy. So they need to look at the variance of these data points over this period to judge the statistical significance of the shift.
So my question is: is this shift statistically significant at some high, say, 95% or 99% p levels?
show that peak tornado activity is starting and ending earlier than it did 60 years ago.
==========
how is “ending earlier” consistent with global warming? they have ignored the implications of both the start AND end being earlier.
today: national news: weather radar shows a tornado forming at this instant outside Buttfork, Nebraska.
60 years ago: national news: a farmer outside Buttfork, Nebraska reports evidence that a twister must have come through his fields sometime in the past couple of weeks.
is this shift statistically significant at some high, say, 95% or 99% p levels?
=========
and under what PDF assumption? Does the significance change if you assume the data is a Fractal distribution as compared to a Normal distribution.
Is it even appropriate to assume a Normal distribution when studying a dynamic system like Climate and Weather? Doesn’t this underestimate the probability of outliers? Isn’t most of our statistical theory misplaced when applied to dynamic systems?
“.. under what pdf assumption? …”
The first task in assessing any statistical signficance is to determine how the statistics are distributed, so if it’s not a normal distribution (or approximately normal) then their are other tests for signficance. There are even tests that don’t depend on distributions (e.g. non-parametric).
The main idea of signficance remains the same: is this earlier-onset shift so large that it can’t be attributed to natural variance in the data?
Signficance levels are subjective in the sense that you can pick the “level” at which you want to report it. So, formally, p-values represent the probability that the null hypothesis could be rejected by just guessing. Informally, confidence in the prediction is (1-p). (But statistical testing cannot prove a null hypothesis. Can only be used to reject the hypothesis).
A p=0.01 level represents a case where guessing or random chance would only have 1 chance out of 100 to succeed. At 10% signficance then a random guess might succeed 10% of the time, inspiring less confidence in the outcome.
So how much confidence should we have in the statistics generated by this report?
Even if the claim were true, I would trade earlier tornados if the over all number or severity were down.
http://www.spc.noaa.gov/wcm/torngraph-big.png
Sick bastards ALWAYS looking at a small change in the nefarious, ridiculous metric known as global temperature for their cause. Idiots all!!
maybe I missed it (please tell me so) but what was the result if nebraska left in and why did they remove it to begin with?
this confused me.
Given that we now have technology to identify tornadoes and incipient tornadoes (thunderstorms with internally circulating winds) almost 100% of the time, technology that was only a dream 60 years ago, one must wonder how much cognitive bias is in play.
Innumeracy enters into it too. With the present means of dissemination, folks learn about a dog finding a fire hydrant in real time these days complete with a tweet and a picture. People over-react to things.
I’m sure cold phases of the PDO & AMO in the 50s, 60s, & 70s have nothing whatsoever to do with it… Cycles…vroom, vroom.
There are a number of problems with this.
First of all, this appears to be another case of cherry picking the starting point of the study – putting it right in the middle of the 1940-1975 cool-down, so that they get a little bit of that period and then mostly the period is the 1975-1998 warm-up. This makes the trend look steep.
Point #1: If they can look back to 1954 – when it was pre-doppler – why not take it back to 1880 or 1930? Anthony says that the records only go back to the 1950s. That seems weird. I mean, the HAD newspapers out in the Great Plains, didn’t they? And weren’t tornadoes a big deal for the people out there before the 1950s? But he knows his stuff, so if he says so, I will take it at face value. It just seems weird to me.
But I WILL go on record with one of my first reactions to the whole global warming thing: The history is so short, how can they even imagine that they can determine anything about the overall history, based on such a short record?
Point #2: This seems to be an overall shift of the pattern, even as it is – considering that then END also shifts earlier.
Point #3: Also, it doesn’t say if the trend flattens out in the last 17+ years.
Point #4: HOW SO? The paper isn’t linked to, so it is not possible to know what evidence this “expectation” is based on.
Q1: What measure is used to determine “peak activity”? Total energy? Or a simple count of tornadoes?
Q2: Moving the peak earlier does not necessarily mean that tornadoes are genuinely starting earlier. Without the link to the paper (which is probably behind a paywall, anyway), it isn’t possible to determine.
Q3: If the peak is only moving 7 days, what is with the “up to two weeks” b.s.?
Q4: Doesn’t this sound like Nebraska actually is doing the OPPOSITE? That with Nebraska included the trend would be less HAS to mean that Nebraska is pulling down their “average”.
Q5: At the same time, if OK’s peak is “only” 7 days earlier, the non-BE and non-OK states must really be showing something even more so – but for some reason that wasn’t rolled into the article. Why not?
Anthony:
Point #5: This makes me want to tie it to the high sunspot activity in the latter part of the 1900s, since the El Niño history shows a correlation between El Niños and sunspots. In which case, if more heat is entering the system (however the sunspots are tied to that, which is still a mystery), then spring and summer would come sooner. But also last longer.
Study finds less tornadoes in late May and June. This is consistent with what we would expect from global warming.
Now I wonder why we don’t see this headline?
The tornado data is not nearly good enough to make any credible claim about trends in counts, occurrence date, etc. The data has several sources of inhomogeneity that make any analysis of trends meaningless. Before 1997, the reports relied totally on human spotters. After 1997, radar is used. Comparing data before and after is not advisable. Any study that claims global warming is going to cause some bad outcome will always be published, regardless of the poor quality of the data. It is very frustrating to see so many studies like this one published in journals with big reputations.
As there are less tornadoes now than in the 1950’s and 70’s, what we are probably seeing is a similar number of early tornadoes but less later ones
The data I’ve seen shows the US temperature peaks around 1940. The cooler climate since then is also “in-line” with the earlier in peak tornadic activity over the last 60 years.
Naturally, other factors such as better identification and reporting of tornados over time “may” play a role.
I’m sure they didn’t consider the massive increase in buildings and paved roads since the 1950s. When I fly r/c sailplanes the first places I look for fat thermals is over hot blacktop or a simmering rooftop. The increase in blacktop and concrete in those states over the past 60 years has got to be staggering.
Why do artificial boundaries – state lines – (Nebraska?!) seem useful in a study about atmospheric activities? Did Earth Mother mention her acceptance of these lines, and I missed it.
I can see rationally how global warming could decrease the frequency of tornados by reducing the Arctic air-gulf air temperature difference, but I cannot think of a reason why there should be changes in SEASONALITY of tornados .
As others have pointed out, throwing out Nebraska is a blatant example of cherry picking. There’s nothing special about Nebraska other than that it makes up about 1/7 of the data points or so. When you’re playing around with large fractions of the data, you can get funny results . What would the results have been if Texas, or Oklahoma, or Kansas had been thrown out- would tornado season peak a week later rather than two weeks earlier?
I suspect Doppler radar plays a part as more tornados are detected now than in the 1950’s.
Doppler radar could result in a spurrious INCREASE in tornado stats, but it shouldn’t affect the RELATIVE frequency of tornados- and it’s RELATIVE frequency that was being discused in the paper.
To me this is clearly just a publication that fulfills a publication quota and it’s quite silly for them to make this claim. What would be the average tornado peak date be if all data were known for the past 10,000 years? That’s obviously unknown, but the law of averages suggests that the 1954-2013 average date would be closer to the long term average than the 1954-64 average alone. As more and more data are collected the average peak date is going to move. They are attributing this to a trend caused by unknown forces over the past 60 years but it is much more likely that the average is simply moving closer to the long term mean.
And practically speaking, no one here in tornado alley is preparing for the peak of tornado season earlier. Not only are we not expecting tornados earlier than we used to, there is no point in worrying about it. You don’t need to prepare with days of supplies like you do for hurricanes and boarding up your windows is not going to help. We simply keep an eye on the sky on a day to day basis.
I’m skeptical of any tornado analysis that purports to show reliable trends, especially based on data from the very earliest period from which we have any remotely accurate data or which includes weaker tornadoes. Tornadoes are much more likely to be observed now due to spotter activity, radar, and higher population. In Oklahoma, it is my observation, that the strongest tornadoes occur later in the season and on the average, the earlier ones are weaker. When you add the fact that measurements of tornado strength is based on damage alone( strong twisters in open territory may well be rated weaker than they really are due to lack of available damage parameters), you include even more uncertainty, especially in such a relatively short time period as 60 years. Were I in a classroom with those who published this as instructors, I would have some ready made questions. They might indeed have the answers, but I’d like to see them.
Several commenters above have already noted that a big part of tornado territory in the US was left out of the study. What really are the most tornado prone states in the US?
I extracted data from a NOAA map, keyed to Average annual tornadoes per 10K square miles per state (1991-2010)
12 KS
10 FL
9 OK, IA, IL, MS
8 AL, MD, SC
7 AR, LA, NE,
6 TX, MO, TN, NC, IN,
5 CO, MN, GA, KY, ND, SD
The six most tornado-prone states are: Kansas, Florida, Oklahoma, Iowa, Illinois and Mississippi. Tornado Alley is in fact a tornado ellipse.
http://www.spc.noaa.gov/wcm/
http://wattsupwiththat.files.wordpress.com/2012/05/tornadotracks_56_years.jpg
http://wattsupwiththat.com/2012/05/31/stunning-map-of-noaa-data-showing-56-year-of-tornado-tracks-shed-light-on-the-folly-of-linking-global-warming-to-severe-weather/
While I detest the left coast, I am kinda glad I live in Oregon!
Oregon is indeed beautiful area, but the tornado threat, while real, is also one you can anticipate with a little
information, education, and common sense. – often a week ahead of time in the generalities, and much closer in the particulars. From 2010-2013, there were 6 EF4-EF5 tornadoes within 10 miles of my location and I escaped any damage or injury. An EF1 came within 100yards of where I was sheltered, damaging trees, roofs and some apartments, but I had anticipated it 15 min. prior to any warning, just using my experience and cell phone radar..
You can’t do the same with earthquakes. They just happen and over a much wider area, without warning.
I’ll take Oklahoma. We might go years without a siren being blown in a down cycle as well, until the next
up cycle. In 2014 we have had almost no tornadoes in the state.
That’s seems like a lot of tornadoes, but it’s only 56 years worth of data superimposed. It would be interesting to see an animation of these tracks by year, and even better to have a lot more data. Otherwise, it’s just weather.
Even a quick glance at the map reveals that, excluding the Appalachians, tornadoes can strike just about anywhere in the United States east of the Rocky Mountains. If you’re going to study tornadoes, you need to study them all, but the authors are playing another game:
In other words, if we can jigger the data, our results will be even more impressive.
So let’s see. Starting earlier and ending earlier sounds like a shift, not an expansion. If they had said the tornado season was starting earlier and ending at the same time or ending later, then we might have something to worry about. Simply shifting the season earlier is interesting (perhaps even useful) information, but not a substantive concern, presumably.
I’m curious about the physical mechanism that would cause a two week shift in tornadoes as a result of a 50-100 ppm increase in CO2. I’m not saying there isn’t such a mechanism. Just curious to know what it would be.
http://research.aerology.com/wp-content/uploads/2010/03/Tornados-by-date.jpg
http://research.aerology.com/wp-content/uploads/2010/03/Tornados-by-date-LD.jpg
graphs from
http://research.aerology.com/lunar-declinational-affects-on-tornado-production/
Past patterns as a forecast for the years 2008 through 2010;
http://research.aerology.com/wp-content/uploads/2010/03/6558day2008-2010.jpg
The yellow spheres are the actuals from the 2008 and 2009 seasons. The other colors are from the past three cycles.
The pattern for 2013 in above the size scale in the middle of this graph, 2014 next on the right. This is the pattern I expect to repeat out to 2021 on the far right side, shows the effect on the lunar tidal effects on the atmosphere as the declinational varies on the 18.6 year declination angle at culmination.
http://research.aerology.com/wp-content/uploads/2010/03/6558days-cycle.jpg