Sabine
patreon
Extreme Weather Becomes Worse Faster Than Predicted
🕑 Added 2024-01-23 16:00:09 +0000 UTC[This is a transcript with links to references.]
Floods, droughts, heat waves, hurricanes, storm surges. Extreme weather events receive a lot of media coverage. In recent years, these events have frequently been attributed to climate change. This “extreme event attribution” how it’s called is a way to quantify how climate change supposedly increased the likelihood of a specific weather event by so and so much. But it’s becoming increasingly clear that these numbers are underestimates. Yep, that’s right, reality is worse than they told us it would be.
The idea of “event attribution” was first put forward by Myles Allen in a Nature commentary in 2003. Allen, a professor at the University of Oxford, was trying to figure out whether he might one day be able to sue the fossil fuel industry because his street was flooded. But well Oxford was under water again last week, and so far no one’s figured out how to sue Shell for it.
Allen’s idea didn’t take off until his student Frederike Otto, now a professor herself, got on the topic. That’s the same Allen by the way who wasn’t happy about a quote from Otto in the Guardian which I talked about a few weeks ago but, hmm, better we stick with the science.
Otto together with Geert Jan van Oldenborgh, who has since passed away, understood that the numbers for event attribution might not be of much use in lawsuits, but they’re gold for newspapers.
If you can say that climate change made this flood or that heat dome twice as likely, then that’s much more tangible than talking about the consequences of a 0 point 1 degree increase of global average temperatures and I don't want to shout but while we were talking, I saw you nodding out.
The idea of event attribution is simple enough. You specify what type of extreme event you’re interested in. Say a flood above a certain amount of rainfall in some region of the planet. Then you take a climate model and make two predictions for the frequency of that extreme event. One with the greenhouse gases at the current level. One with pre-industrial levels. Then you compare the two cases.
Even leaving aside the issue of people nodding out, this idea has still has multiple problems which I talked about extensively in an earlier episode. One is that the frequency by which the event occurs in the model depends strongly on the exact definition of the event. The more details you add the less likely the event is going to appear in any of your simulations, so you’ll end up comparing zero to zero which tells you zero. Since the definition is up to you, you can fumble with it until you like the result which, just so we’re on the same page, isn’t how science should work.
But the much bigger problem is that climate models were not coded to simulate extreme weather events and they’re in fact not any good at it. Most climate models underestimate both the frequency and severity of extreme weather events.
You don’t have to take my word for this, if you look at papers on the topic of event attribution you can find this information there in clear words. For example, in this review from 2021 you can read “climate models have not been designed to represent extremes well” and “Trends can differ widely between different climate models” and “The results of the attribution study can depend strongly on [the quantitative definition of the event]”. It’s just one of those things that they don’t tell you in the news articles.
What this all means is that all those numbers which you come across in the headlines are likely to be underestimates. The new paper which just appeared confirmed this problem for extreme rainfall events. They looked at a set of 21 of the best climate models and compared the models’ predictions with the actually observed changes. They find that, quote “virtually all climate models significantly underestimate the rates at which increases in precipitation extremes have scaled with global temperatures historically.”
I find the entire idea of extreme event attribution extremely problematic because it suggests a false sense of accuracy. If you tell policy makers that this recent once-in a century flood has become twice as likely, they’ll think they have 50 years until the next hits them. But this is not what the numbers mean. They mean the event has become at least twice as likely. Maybe it now happens every year. Or twice a year.
Basically I think this entire event attribution business should be regarded with great scepticism. But while it might be difficult to attribute extreme events to climate change, it’s much easier to attribute the popularity of the attribution to climate change. Which is why I herewith want to propose the new research area of attribution attribution. You heard it here first.
Comments
Sources, that should perhaps be taken more in account are the reports of the insurance companies. They show the increase of catastrophical happenings caused by weather events since decades, and their statistical calculations reckon with a further increase for the future. It´s hard for conspiracy theorists to integrate them into their ideas of evil.
One obstacle to having more precise and higher space and time resolution climate models is, as it has been from the beginning, a need for faster computers with larger storage capacity, as well as the algorithms for processing very large amounts of meteorological data both fast and precisely enough.
Climate models underestimate the likelihood of extreme events, but why? Is a problem of the model, of the implementation, can it be solved?
