Thread by @hausfath

CMIP6 is the next generation of climate models being run in the lead-up to the 2021 IPCC 6th Assessment Report (AR6). Over at Carbon Brief we've put together an in-depth explainer of what scenarios and experiments are being run and of results so far carbonbrief.org/cmip6-the-next… 1/18

CMIP6 stands for "Coupled Model Intercomparison Project 6". It consists of around 100 distinct climate models being produced across 49 different modeling group. It is currently a year or so behind schedule, and not all 100 model runs will be done in time for the AR6. 2/18

In the lead up to the IPCC AR6, the energy modeling community has developed a new set of emissions scenarios driven by different socioeconomic assumptions. A number of these SSP scenarios have been selected to drive climate models for CMIP6: carbonbrief.org/explainer-how-… 3/18

While the four "RCPs" used in CMIP5 live on in new forms (e.g. SSP1-2.6, SSP2-4.5, SSP4-6.0, and SSP5-8.5), a number of new scenarios have been added. These include a scenario designed to limit warming to 1.5C (SSP1-1.9) and one exploring a ~2.5C warming outcome (SSP4-3.4) 4/18

Additionally, there is a better exploration of no-policy baselines, with SSP3-7.0 providing a more plausible counterpoint to the worst-case SSP5-8.5 scenario: carbonbrief.org/explainer-the-… 5/18

The new scenarios analogous to the old RCPs are not identical, even though they lead to nearly the same 2100 forcing levels. Here are the CO2 emissions in the new SSPs compared to their old RCP analogues: 6/18

Many of the new CMIP6 models available so far have notably higher equilibrium climate sensitivity (ECS) than the older CMIP5 models. While CMIP5 ECS ranged from 2.1C to 4.7C, the 19 CMIP6 models that provide runs needed to calculate ECS range from 2.3C to 5.6C. 7/18

About half of the individual CMIP6 models available so far have an ECS that exceeds the 1.5C to 4.5C "likely" range highlighted in the last IPCC report. While 80 or so more models will eventually be included, its clear that the high end of the sensitivity range has gone up: 8/18

15 CMIP6 models currently have "historical" runs available, where estimates of past CO2 concentrations and other climate forcings are used to simulate climate of the past 150 years. This figure shows CMIP6 historical runs compared to both CMIP5 and @NASAGISS observations: 9/18

Long term 1880-2019 warming rates are quite similar in both CMIP5, CMIP6, and observations. Both CMIP5 and CMIP6 show slightly higher average warming rates than observations since 1970 (10% and 20% respectively), but observations are within the range of different models: 10/18

There is not really evidence that CMIP6 runs available so far do a worse job at simulating past temperatures than CMIP5, despite the notably higher sensitivity of many of the models. 11/18

Now for the exciting part: future warming projections in CMIP6.

Here higher ECS (and transient climate response) of models available to-date really shows up. Figure below shows the range of warming for each CMIP6 scenario, w/ number of models currently available in () 12/18

While the multimodel mean will change as more models are added, the results so far are stark. The 1.5C SSP1-1.9 scenario has a mean warming of 1.6C relative to preindustrial (1880-1900 here). The well-below 2C SSP1-2.6 scenario has a mean warming of 2.1C. 13/18

The new SSP3-7.0 baseline scenario has around the same warming (4.5C) as the old RCP8.5 scenario did. The new SSP5-8.5 scenario shows a staggering 5.5C mean warming. 14/18

These higher warming projections are clear when we compare analogous scenarios in CMIP5 and CMIP6: 15/18

So what are we to make of these new results? Clearly these are preliminary and will change as more models are added, but they are certainly worrying. 16/18

At the same time, models are not our only way to estimate the sensitivity of the climate to future emissions, and paleoclimate, emergent constraints, and the observational record all suggest that ECS > 4.5C is somewhat unlikely. 17/18

Going forward we should try and use multiple lines of evidence in assessing future climate changes, rather than just relying on the CMIP6 multi-model and spread in isolation. 18/18