The Antisocial Darwinist

A lot of people are apparently convinced that climate change is responsible for Hurricanes Harvey and Irma. These destructive storms are conspicuous and terrifying, and therefore they provide an easily visible and emotionally powerful centerpiece for political arguments about the need to curtail fossil fuel emissions. Only problem is that the data connecting storms to climate change is sketchy at best.

I’m not going to go into the output of real climate scientists here. My understanding is that the prediction that the future will hold more and stronger tropical storms is based on models, much like predictions about the future of Earth’s temperature. These models are based on a combination of historical observation and theory about the underlying causes of weather patterns. Models are very useful for these sorts of things — but I’m often miffed at how science reporters fail to convey to the general public the limitations of models, or give an honest representation of the “error bars” on future predictions. Regardless, doesn’t matter, in this blog I’m not going to talk about the models at all.

Also, I want to preface this by saying that I don’t consider myself to be a climate change skeptic. Number one, I’m a scientist that works on questions involving anthropogenic global change. Moreover, I believe humans are changing the composition of Earth’s atmosphere and that those changes appear to be causing the planet to warm at some rate. I agree with most other scientists that for that reason, and a host of others, humans need to figure out another way than fossil fuels to provide for their energy requirements.

But also as a scientist, I get angry when people misrepresent facts and overstate conclusions. The argument that storms will get stronger in a warming world makes theoretical sense, but the problem is that it isn’t clear i) that it’s actually happening or ii) if it is, how much of an effect it actually has. For instance, Hurricane Irma has record-breaking wind speeds of 185 mph. How much of that, if any, is due to forming over warmer oceans? If Irma had formed in 1990, would she have had 181 mph winds? 150 mph winds? Would she not have formed at all? These possible effects of climate change run from the insignificant to the apocalyptic, and all seem plausible to me at first glance. Moreover, hurricanes are relatively infrequent, so my gut-level intuition as a scientist who works with real datasets is that we can’t come to any conclusions about the effects of climate change on storms unless those effects are quite large.

Fortunately, the internet has lots of data, so it’s pretty easy for a concerned citizen to go download some pretty good numbers to do some back-of-the-envelope calculations for a first pass at assessing these claims about climate change and tropical storm intensity, which is just what I did. I whole-heartedly encourage all of you to do the same about this and any other controversial political claims — don’t just rely on the prognostications of Internet Science Guys, pull the numbers yourself. The analysis I’m going to talk about here took me less than half an hour to do.

First, let’s consider the hypothesis that storms are becoming more frequent with time. I pulled data for Atlantic tropical storm frequency from 1851 to 2015 from the awesome website WeatherUnderground.com. Per year, it lists number of total tropical cyclones, the number that eventually became hurricanes, and the number of deaths reported.

 Figure 1

By doing a simple linear regression (figure 1), we see that the number of storms has significantly increased (p << 0.001) over this time period, by about 0.04 storms per year each year.

 Figure 2

The number of hurricanes has also increased (figure 2, p = 0.004) but by a lesser amount of 0.01 hurricanes per year, per year. This seemed odd to me — if rising temperatures are creating storms, wouldn’t we expect a similar increase in both types of storms?

 Figure 3

So I looked at the change in the percentage of tropical storms that became hurricanes eventually (figure 3), and discovered that this value actually significantly DECREASED (p << 0.001) by 0.13% per year over the time frame considered.

These observations may support the idea that climate change is increasing storm frequency, but it contradicts the idea that it’s making storms worse. Another, more likely possibility is that storms aren’t actually becoming more frequent or milder, but that our technology for detecting them is just improving, meaning that storm frequency for the earlier years in the dataset was underreported, and was biased toward large storms (hence in some earlier years 100% of reported storms became hurricanes). To test this, I constrained my analysis to just the “satellite era” when we gained the ability to see storms from space even if no one was around to experience them. This technically started in the 1960’s, but I used a relatively arbitrary cut-off of 1976 to give the technology some time to come into routine use. With this limited dataset, we still see storms in general increasing significantly (p < 0.001) at the even faster rate of 0.22 storms per year per year, but hurricanes no longer increase significantly, and the decrease in the percentage of storms becoming hurricanes becomes even more pronounced (-0.005% per year, p = 0.01). I feel this tentatively supports the hypothesis that the appearance of increasing storm frequency is an artifact of increasingly sophisticated and careful reporting; but regardless, it fails to support the hypothesis that climate change is producing stronger storms, even if it may be favoring a greater number of less powerful ones.

I wasn’t able to find easily mineable data for top wind speed or lowest barometric pressure for historical storms, but another metric we can use to estimate storm intensity is the loss of human life. It’s relatively easy to lose track of simple tropical storms, harder to ignore hurricanes, but it’s really hard to ignore dead bodies. So I ran a regression analysis of number of deaths from tropical storms per annum from 1851 to 2016, and discovered there was no significant increase in the lethality of storms during this time period, nor was there a difference when the analysis was run just using the 1976-2015 data. But this understates the case. When I normalized deaths against the US census for the population size of the state of Florida across the same time period (hopefully a reasonable proxy for general population increase over the same time frame), we see that the risk of death by tropical storm DECREASES significantly (figure 4, p = 0.001) by about 0.013% per year over this time period.

Figure 4

This isn’t surprising, since our ability to predict landfall and prepare for its effects have surely increased dramatically since 1851. For instance, in 1870, you had a ~1.5% chance of being killed by a hurricane, the equivalent of Irma killing 250,000 Floridians today! However, when we just consider 1976-2015, we still don’t see an increase in risk of death by hurricane, despite the fact that we know that large storms like Katrina are still capable of killing many people despite our best preparations. Thus, the hypothesis that storms are becoming more lethal is soundly refuted by this simple analysis.

I’m sure there are much more sophisticated data analysts out there, and accept that they might be able to come up with superior analyses to the very simple one I’ve done here. However, I think it’s important to look at this, because the reality is that the assertion that climate change is affecting hurricanes, and that the deaths in Houston and the Caribbean this year can be put on President Trump’s bill, really doesn’t pass the first smell test. There is ample reason to be skeptical of these claims, and they aren’t based on ignorance or climate myopia.