Right this moment, three years after his dying, and earlier than the 10-year anniversary of World Climate Attribution, the final paper Geert Jan and I labored on collectively is revealed.
The paper presents a quantitative statistical synthesis technique now we have developed during the last eight years of conducting fast probabilistic occasion attribution examine. It’s a statistics targeted paper, one thing he favored loads and one thing I can do if want be, which is why it was on my desk for thus lengthy. It’s most likely not an thrilling learn for many. Nevertheless, having the ability to mix totally different traces of proof into one quantity, the ultimate end result describing the overarching affect of local weather change on the depth and probability of an excessive climate occasion, has been a key milestone in World Climate Attribution’s methodological improvement and the science of occasion attribution at giant. We name this step the hazard synthesis.
Many attribution research solely use local weather fashions or climate observations, quite than each, or deal with only one side of an excessive occasion, such because the low stress system that led to very heavy rainfall, however not the function of local weather change in ensuing rainfall. Our technique, utilizing observations and fashions, and crucially, combining them within the synthesis, extra realistically captures the general affect of local weather change on an excessive climate occasion.
Whereas the concepts introduced within the paper had been developed with Geert Jan over a few years, a few of the limitations of the method have solely develop into obvious in the previous few years. For instance, it isn’t potential to estimate how more likely an excessive climate occasion has develop into if the occasion wouldn’t have been potential in a 1.3°C cooler world with out local weather change. We’ve seen this in heatwaves within the Mediterranean and the Sahel this 12 months and in Madagascar, Southern Europe and North America, Thailand and Laos final 12 months. When the change in chances are infinite, a numerical illustration simply turns into an illustration that goals to point out simply how a lot of a recreation changer human induced local weather change is.
An issue we steadily encounter comes when local weather mannequin outcomes don’t agree with the fundamental physics that governs the climate. We all know from the basic Clausius-Clapeyron relationship {that a} hotter environment can maintain extra water vapour, which ends up in heavier downpours – about 7% with every 1°C of world warming. Nevertheless, after we studied excessive floods within the Philippines, Dubai, or Afghanistan, Pakistan and Iran this 12 months, climate observations confirmed a rise in heavy rainfall as anticipated, however the local weather fashions indicated both reducing rainfall or no change in any respect.
This disagreement means that the fashions are unable to duplicate all of the bodily processes at work in the actual world. Sadly, poor mannequin efficiency is widespread for international locations within the World South that usually have restricted means to fund local weather science programmes. For brief period occasions, primarily based on Clausius-Clapeyron, we will level to local weather change to clarify the elevated rainfall. For longer period occasions over weeks or months, although, we will’t hyperlink the rise to local weather change as altering climate patterns could have a task.
When climate observations and local weather fashions do align, we will perform the synthesis described within the first a part of the paper (statistic geeks, you’re welcome!) and confidently report modifications to the depth and probability of an occasion. For instance, in 2022, we discovered that local weather change made the lethal heatwave in Argentina and Paraguay was made 60 occasions extra seemingly, whereas earlier this month, we discovered that local weather change elevated Hurricane Helene’s rainfall by about 10%.
Whereas the methodology within the paper may be very statistics-heavy, it additionally highlights necessary questions that everybody ought to ask when evaluating the outcomes of an attribution examine:
- Does the statistical mannequin match the noticed knowledge properly? Or is the document too brief?
- Are the observations now we have of top of the range? Or are there giant discrepancies between totally different datasets?
- Are the mannequin outcomes constant between totally different local weather fashions? Are there any identified deficiencies within the local weather fashions?
- Do fashions and observations agree?
- Do we discover totally different outcomes when taking a look at greater warming ranges?
- Are fashions and/or observations in step with what we all know from bodily science, e.g. the Clausius–Clapeyron relationship?
- How do our outcomes evaluate to different revealed analysis or analysis syntheses reminiscent of included in IPCC stories and authorities assessments?
Typically, the solutions to those questions should not trivial. They affect how we interpret and talk the ultimate end result. So, in the event you ever questioned why we don’t simply automate the evaluation of the hazard, or let AI do it, that is why.
As Geert Jan used to say: “you want time and expertise to know when your numbers lie”.
By Friederike Otto, Co-founder of World Climate Attribution