Is human activity really making hurricanes stronger?

Yes and no. But mostly yes.

The media tends to treat it as obvious: I mean, all you have to do is look out the window to know hurricanes have gone crazy, right?

Not quite.

Detection and attribution is a field onto itself, and it’s trickier than you might suspect.

Roger Pielke Jr., the famous contrarian climate policy analyst who’s also a bit of a hero of mine, has spent a lot of energy explaining why the media consensus on this is just plain wrong.

He notes the latest IPCC assessment report fails to pick out the anthropogenic signal from the natural variability noise with any degree of confidence. You can just eye-ball this chart and see why:

High quality data only go back a few decades, and there are only a few hurricanes per year. If you squint, you can sort of see that more of the hurricanes are now categories 3-5 than before. But it’s not a super distinct trend, is it?

Turns out this is true not just for hurricanes and typhoons, but for a wide variety of extreme weather events: floods, droughts, landslides and more.

In fact, the IPCC says it doesn’t expect to be able to detect an anthropogenic change in the pattern of tropical cyclones due to human activites before the end of this century, (page 1856) even in an unrealistic scenario with very high emissions. If you stick closely to the evidence we have, you just can’t say whether human influence has in fact strengthened tropical cyclones or not.

So Pielke is right, but only in a sense that’s so narrow and technical it ends up being misleading.

Pielke grants that the attribution bar the IPCC has set itself is hard to meet. Extreme weather events are, by definition, rare. High quality data is sparse. Natural variability, including from one decade to another, is quite high. Natural variability is a real thing. The decade with the most major hurricane landfalls in U.S. history is still the 1940s. That wasn’t anthropogenic, it was just a fluke. Flukes do happen. When your N is small, flukes can send you for a loop.

The bottome line is that there’s a lot of noise in these data, and not that much signal. That makes formal detection and attribution very demanding. But it doesn’t mean the signal isn’t there. It just means there’s a hell of a lot of noise. 

Pielke illustrates the point well with a little fable. Imagine you go to a casino to play blackjack, and you start to suspect that the croupier is messing with the deck, adding aces for some reasons. You don’t know if he’s slipped one ace in, or two, or three, or four. Maybe you’re wrong and he didn’t slip in any.

How many hands would you have to play to detect, with high statistical confidence, how many aces the croupier slipped into the deck?

Drawing two aces on your next hand doesn’t prove anything: that outcome is consistent with any of these hypotheses. To establish tampering, you’re going to have to prove that aces are turning up with a frequency that would be highly unlikely if the deck hadn’t been tampered with. For that, you’re going to need data. A lot of data.

How many hands would you have to play? Pielke calculates it would take almost 11,000 hands to detect the tampering using the standard for “high confidence” the IPCC adopted. But with hurricanes, you only get to “play” 5-15 hands per year. At that rate, it could take centuries for an unmistakable trend to arise in the data.

But why wait that long? In our situation, there isn’t any actual uncertainty about whether our croupier is fiddling with the deck. We know he is! We emit a gigaton of carbon dioxide equivalent a week. The croupier is stacking the deck in plain view.

Is that a blackjack table you want to play at?

Now that we’re withdrawing cloud cover from the oceans, it’s even worse. To extend the analogy, we’re at a blackjack table where, for the last 60 years, the croupier has been sliding four extra aces into the deck at the start of the night, but his manager has been tut-tutting him and taking three of the extra aces back out again before the start of play. Suddenly, in 2020, the manager stopped showing up to work. Maybe he got Covid. The croupier is still slipping those four extra aces into the deck each night, but nobody’s taking any out before the start of play. And a suspicious number of two-ace hands seem to be getting dealt. 

Are you really going to sit at that table and play 11,000 hands just to gather the statistical evidence to prove that the game’s rigged?

You already know it’s rigged.

In the end, the narrow technical question of whether we have detected a change in hurricane trends and attributed it to human activities is not really that relevant. What people want to know is whether they’re at increased risk of finding themselves in the path of a very destructive storm. That’s subtly different, and here the answer isn’t really in doubt.

We have a sturdy scientific basis to expect that if humans do what we’ve been doing, oceans will get hotter. And we have a strong basis for expecting that if oceans get hotter, hurricanes will strengthen. And oceans are getting hotter. And we have a bunch of very strong storms.

I don’t think anybody sane needs is going to stay at this blackjack table until 2100 before they draw their conclusions.