Mar 27, 2023

Climate Modeling: CMIP3 and CMIP5

CMIP3 and CMIP5 are two versions of climate modeling used in IPCC AR4 and AR5 respectively. Learn how they differ and how well their projections have been doing since 2009.

CLIMATE_CHANGE SCIENCE ENVIRONMENT

Gavin Schmidt

Climate scientist, occasional juggler, even more occasional author, curious about how the world works.

Member of Climate Science

In which Judith Curry gets confused about CMIP versions (easily done!) and shows a lack of familiarity with the details of climate modeling (also easily done!). https://t.co/m5cFpD8a3v
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
For reference, CMIP3 was used in IPCC AR4 and the model runs happened in 2003-4, using historical forcings to 2000, and the SRES scenarios subsequently. I've been plotting how well those projections have been doing since at least 2009. https://t.co/MrjBQIa3HT
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
CMIP5 was put together prior to IPCC AR5 around 2011, using historical forcings t0 2005 and RCP scenarios subsequently. I have been plotting how well those projections have been doing since 2014. https://t.co/9u7rfLO8dl
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
The graph @SenWhitehouse showed at the hearing was a 2019 vintage version of the graph I've been making since 2009, specifically taken from an earlier version of this feature (it's since been updated): https://t.co/QiU7g2cDP6
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
Specifically, this graph, which was a nicer presentation of the CMIP3 results I'd been showing previously, and which was discussed in Hausfather et al, 2019: https://t.co/83b04tEAfN pic.twitter.com/XRevwJxUQ5
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
The latest update to this figure was included in a 'blog post' in January this year (including data from 2022). https://t.co/nXaMoo9Gad

There is now has more than two decades over which the CMIP3 projections can be assessed - and they are doing pretty well! pic.twitter.com/bL3aKlJQh5
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
Let's compare how these figures and observations have changed though from 2009 to 2023: pic.twitter.com/KQNXa6WpcB
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
The figure is now stylistically better, clearer text, less jargon, a couple of useful labels etc. The model ensemble has not changed at all. There was some difference in the observations earlier (HadCRUT3 & GISTEMP), which now (HadCRUT5/GISTEMP4) is much less.
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
I don't expect everyone to keep up with 'blog posts' on @Realclimate or my papers (life is too short!), but many people do. I do find it curious that Judith Curry has never seen any of this in over a decade of me posting it, but I too see a lot without really taking it in.
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
Judith confuses these results with those from CMIP5 (which, you will recall, was the next iteration of the model ensemble - don't ask about what happened to CMIP4!). CMIP5 was the basis for @ed_hawkins figure in IPCC AR5, and I've been plotting that too since 2014. pic.twitter.com/ZPQYcGZLqC
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
Two differences stand out. Firstly, the model spread is narrower than from CMIP3. This is due to a greater conformity of the forcings and model completeness (in earlier rounds, some models didn't have solar forcing, or ozone depletion, or used very different inputs). Secondly...
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
... the projections are not as good a fit to the observations post 2005. There are (it turns out) three reasons why. First, some of the forcings the models all used were slightly biased (enough to make ~0.1ºC difference) https://t.co/nFcxSji0sV
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
Second, the observations themselves were slightly biased (due to insufficient coverage in the polar regions, continuing issues with ocean measurements, etc). And thirdly, the models and the observations aren't quite measuring the same thing (~0.05ºC). pic.twitter.com/EHFeFxK5Uw
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
The upshot is that the concern that models were significantly diverging from the observations that so animated the community in 2013/14 in retrospect, was not that big a deal. Again, I don't expect folks to pay attention to the details here unless they are really interested.
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
I share Judith Curry's concern that Congressional hearings can be more about theater than substance, but then I don't invited just to be a contrarian voice. 🤷 I am happy to say "I don't know" when asked about graphs I don't recognise though.
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023
It's really ok not to know everything about this topic. It's big & complex. But even in the heat of the moment, declaring something you don't recognise to be a 'flawed analysis' and then doubling down on your confusion in a subsequent 'blog post' does not seem responsible. YMMV.
— Gavin Schmidt (@ClimateOfGavin) March 27, 2023