By Robert Vislocky, Ph.D
After every tragic meteorological event there will inevitably be the alarmist cries that climate change was a substantial influence. The recent late-season tornado outbreak of Dec 10-11 is no exception. Usually most of these are appeals to emotion with no data to support the relationship between extreme weather and climate change. However, this tweet from Michael Mann pointed to a 2019 peer-reviewed study titled “Increasingly Powerful Tornadoes in the United States” by Elsner et al.1
This study, not so surprisingly based on the title, claims that tornado power has increased 5.5% per year on average in the period from 1994-2016 as displayed in Figure 1 below: https://myweb.fsu.edu/jelsner/PDF/Research/ElsnerFrickerSchroder2019.pdf
I found this result to be intriguing in the least, especially in light of the observation that intense tornadoes (EF3-EF5) appear to be declining in frequency. So I investigated further and did a bunch of number crunching with the actual tornadoes dataset available from the Storm Prediction Center (SPC), downloadable from the following web site:
According to the article, power for a tornado is computed using (i) an estimate of the wind speed (taken from the mid-point of each EF category’s wind speed range, or 97 m/s in the case of an EF5 which is unbounded), (ii) the area impacted by the tornado (computed by multiplying the path length by the width), and (iii) an estimate of the fractional path area for each EF category (for instance, an EF3 rated tornado will actually spend 53% of its life as an EF0, 27% as an EF1, 13% as an EF2, and only about 7% as an EF3 on average). These fractional estimates for each EF tornado rating can be found in Fricker & Elsner2: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0131090
It is item (ii) that will be the focus of this article, namely the historical path length & width data. At first glance going through the entire history, it was surprising to see so many tornadoes that happen to have a path length equal to 0.1 miles and/or a width equal to 10 yards. It’s almost as if these were some kind of default (or minimum) values that were assigned if the actual path length (PL) or width (PW) were unknown or not measured. The data below indicate the percentage of the time these values occurred within various time periods in the dataset. Keep in mind the authors’ results were for the period 1994-2016.
Anyone notice a big problem?! Of particular interest, it appears the discontinuity in PL occurs right when the F-scale was replaced by the EF-scale in January 2007. Digging deeper, shown below are the values for PL & PW at the 25th & 50th percentile levels for all tornadoes (EF0-EF5) across two time periods that span the authors’ study.
Results clearly show that the pre-2007 path lengths & widths are contaminated with an abundance of very small path lengths = 0.1 mi and path widths = 10 yd. It is meteorologically implausible that path lengths would increase by 2-3 times or that tornado widths would increase by 50% in just 13 years. Climate change is not that powerful! Unfortunately, this bias appears at higher EF rated tornadoes (EF2-EF5) as well.
The effect of the faulty path length & width data should be obvious for a study that spans across the two periods. Namely, the tornado energy dissipation data in the first part of the study (1994-2006) will be artificially low compared to the second part (after 2006) resulting in a false upward trend. As a result of this finding, the conclusion from the authors that tornadoes are becoming increasingly more powerful should be rejected outright.
Another minor flaw in the study is that the authors only plotted annual median, quartiles and 90th percentile tornado powers instead of plotting the sum total annual power. Using the former can result in misleading conclusions, especially if the number of tornadoes is increasing or decreasing through the time period. For example, suppose a given year had 100 tornadoes in each EF category and the next year had 200 tornadoes in each EF category. Plotting the median power or the 90th percentile value would show no increase from one year to the next, whereas plotting the sum total would provide a clearer picture.
Interestingly, assuming that the post-2007 data are reasonably representative of what truly occurred, below is a plot of total annual tornado power for all tornadoes (EF0-EF5) during the period 2007-2019 using the identical methods described in Elsner et al. Although the result shows a declining trend, keep in mind that this decline was not statistically significant. Unfortunately we will need to wait 10-20 years or more to find if a trend emerges. Hopefully, the NWS can take steps to ensure consistency in path length & width data (and EF ratings) going forward.
1Elsner, J.B., T. Fricker, and Z. Schroder, 2019: Increasingly powerful tornadoes in the United States. Geophys. Res. Lett., 46, 392–398.
2Fricker T, Elsner JB (2015) Kinetic Energy of Tornadoes in the United States. PLoS ONE 10(7): e0131090.
I don’t receive any money from big oil and I’ve never met the Koch Brothers.