What about those tornadoes?

Are there more tornadoes because of global warming? Are they stronger? Do they occur more frequently outside of the usual tornado season, or are they more common in areas that formerly had few tornadoes?

There are problems with all of these questions, and the main problem is the fact that the tornado data isn’t very informative.

Here’s the raw data from the NOAA tornado database, showing the number of tornadoes per year of all intensities greater than one mile long on the ground:

Screen Shot 2013-12-05 at 8.49.06 AM
(Click on the graph to see the whole thing in case it isn’t showing for you.)

This looks like more tornadoes are happening. We could leave it at that but we’d be doing bad science if we did so. The problem is that over time, the way tornadoes are observed and measured has changed, and owing, probably, to changes in population distribution in the region that gets most of the tornadoes in the US, there may be a number of tornadoes in the earlier years that were not observed. But, we don’t know that. In fact, we have no idea whatsoever how to make these data useful. These data could represent a reasonably accurate picture of tornado frequency over time in the US, or they might not, but we have no quantification of how biasing effects might work over time.

A while back I tried to see if I could make the data speak more clearly by measuring the total length and width of tornadoes in the database and adding them up for the period 1990 to 2012. This eliminated the problem of missing tornadoes because I was only looking at recent times when the data would be better. The resulting graph looks like this:

Screen Shot 2013-05-30 at 11.11.53 AM
(Click on the graph to see the whole thing in case it isn’t showing for you.)

This seems to show a dramatic increase in overall effects of tornadoes. But, it turns out that the way “width” of tornadoes was measured was changed during this period, so these data are still not that useful.

Why am I showing you bad data? Here’s why. Consider the possible interpretations of graphs like these. For example:

1) There are more tornadoes over time, possibly because of global warming.

2) There are fewer tornadoes over time, despite global warming.

3) We can’t say anything about changes in frequency, severity, or land coverage of tornadoes over time because the data suck.

4) The evidence shows that there is no effect of global warming on tornadoes over time.

If this was a multiple choice question, the correct answer would not be 1, because the data are not useful. The correct answer would not be 2, because the data are not useful. The correct answer would be 3, because the data are not useful.

But often, we hear people who want to minimize the effects of global warming and deny the importance of climate change claim that number 4 is true. But we can’t say this because … wait for it … the data suck! The data as presented here are not sufficient to say that there are more tornadoes, but the default fallback null model is NOT that there is no relationship between climate change and tornado frequency, severity, or landscape coverage.

Peter Sinclair has a post linking together the tornado question and a spate of climate science denialism in this post: Tracking the Truth About Tornadoes. Go have a look.

In that post Sinclair quotes Michael Mann:

Actual atmospheric scientists know that the historical observations are too sketchy and unreliable to decide one way or another as to whether tornadoes are increasing or not…

So one is essentially left with the physical reasoning…

That physical reasoning is, from a livescience’s piece by Michael Mann:

…warm, moist air is favorable for tornadoes, and global warming will provide more of it. But important, too, is the amount of “shear” (that is, twisting) in the wind. And whether there will, in a warmer world, be more or less of that in tornado-prone regions, during the tornado season, depends on the precise shifts that will take place in the jet stream — something that is extremely difficult to predict even with state-of-the-art theoretical climate models. That factor is a “wild card” in the equation.

So we’ve got one factor that is a toss-up, and another one that appears favorable for tornado activity. The combination of them is therefore slightly on the “favorable” side, and if you’re a betting person, that’s probably what you would go with.

I’d like to add to this: Storminess in “Tornado Alley” is likely to increase. It may be more straight line winds and severe thunderstorm, more ALH (amazingly large hail), or more tornadoes. Personally, I suspect it will be all of this but with the actual rate of tornado formation varying a great deal from year to year, depending on effects such as wind shear, but that’s just a gut feeling. Vertical wind shear may help attenuate the development of tornadoes when it is happening, but it won’t make the energy go away, and that energy may be manifest in other ways through storms that don’t happen to form tornadoes. We’ll see, I guess.

Comments

  1. #1 Beth
    December 5, 2013

    You need to reverse the inequality sign in your graph. It should be > 1 not < 1.

  2. #2 Greg Laden
    December 5, 2013

    Ugh. Thanks Beth. I’ll try to do that when I am at an appropriate computer in a about two days from now.

    For now, let us arbitrarily assign for the purposes of this post only the meant “Greater than” to the symbol “<”

  3. #3 Lawrence
    December 5, 2013

    Along the lines of the last paragraph –

    Is there a scientifically accepted definition of ‘derecho’ and a database of them? It would be interesting to track frequency and size of these big straight line storms along with tornados, making it easier to see how shear plays out.

    The historical data is probably even more sparse than what we have for tornados, but it would be interesting to see going forward.

  4. #4 Greg Laden
    December 5, 2013

    Yes, there’ s a definition: damage across more than 240 miles, wind gusts of 58 mph+ along most of the length of the storm, distinct 75 mph+ gusts frequently, straight line (not tornadic) winds.

    The storm prediction center has a database of storms that can be mined for derecho events:

    http://www.spc.noaa.gov/climo/online/sp3/plot.php

    and there is a derecho database here:

    http://www.nssl.noaa.gov/users/mcon/public_html/derlist.htm

    I’ve not worked with any of these data myself.

  5. #5 Harold Brooks
    December 6, 2013

    The final graph only goes to 2011. There’s really no reason not to include data for area further back in time. F0 tornadoes contribute a few % to the F1+ tornado areas. For example, in 2011, the F0s only cover ~75 sq. miles out of the >2800 total. Looking at F1+ is relatively consistent for many purposes, but the F0 area is small enough, it doesn’t matter a lot. It’s not clear what impact the change in width definition from mean to max in ’95 had on the data, but more attention to damage surveys for major tornadoes can add to both the width and length estimates.
    I was having a conversation with a forecaster a couple of days ago about the derecho definition problem. We use an outcome-based definition, rather than a physical process definition, so that it’s more susceptible to report changes.

  6. #6 Greg Laden
    December 6, 2013

    This was put together roughly. It is years since 1990 so it goes to 2012. This was originally made to display differences oner a few decades to test a specific question in a different post about or peoples recollection of changes in tornado frequency and severity.

    It would be nice to distill this diwn to energy levels.

  7. #7 Harold Brooks
    December 6, 2013

    Starts in 1990, 22nd year is 2011. 2011 is a huge area year (the last two weeks of April). The only comparable year in the database is 1974. 2008 (Super Tuesday and a few days in May) is the 2nd one on this graph. ’98 (x-axis=9) was the biggest area of the 1998. Large area years tend to be dominated by the big outbreaks. 2012 is much lower than 2008 or d2008, below the 2010 number. I couldn’t quite draw this chart freehand without looking at the data, but I could get most of the details.

  8. #8 Greg Laden
    December 6, 2013

    Actually ,what is going on here is that for some reason the right hand side of the graphic is cut off. Click on it and see the last year.

Current ye@r *