Originally, I thought we would have a brief look at the subject of attribution before we went back to the IPCC 6th Assessment Report (AR6). However, it’s a big subject.
In #8, and the few articles preceding, we saw various attempts to characterize “natural variability” from the few records we have. It’s a challenge. I recommend reading the conclusion of #8.
In this article we’ll look at a paper by G J van Oldenborgh and colleagues from 2013. They introduce the concept of assessing natural variability using climate models, but that’s not the principle idea of the paper. However, it’s interesting to see what they say.
Their basic idea – we can compare weather models against reality because we make repeated weather forecasts and then can see whether we were overconfident or underconfident.
For example, one time we said there was a 10% chance of a severe storm. The storm didn’t happen. That doesn’t mean we were wrong. It was a probability. But if we have 100 examples of this 10% chance we can see – did we get approximately 10 instances of severe storms? If we got 0-3 maybe we were wildly overconfident. If we got 30 maybe we were very underconfident.
Now we can’t compare climate models outputs of the future vs observations because the future hasn’t happened yet – there’s only one planet and climate forecasts are over decades to a century, not one week.
We can, however, compare the spatial variation of models with reality.
To see the whole article, visit the new Science of Doom on Substack page and please consider suscribing, for notifications on new articles.
The Science of Doom 