Thread by @luki_brunner

Is the strong future warming found in several next-generation #CMIP6 climate models realistic? We apply weights based on the models’ historical performance and independence to find out.
Our study is now published in @EGU_ESD 👇
doi.org/10.5194/esd-11…

1/🧵

We calculate distributions of change in different variables, like the Transient Climate Response (a comparable measure for how much a model warms due to CO2 🏭). Weighting can considerably shift these distributions and increase their skill.

2/🧵

We find the multi-model distributions of Transient Climate Response as well as future warming to be shifted downward by applying the combined performance-independence weights. Decisive climate action is still needed but the most extreme warming might be less realistic. 🌡️

3/🧵

Our results hold for both, the low emission pathway SSP1-2.6 as well as the high emission pathway SSP5-8.5 and during the entire 21st century. 📈

4/🧵

A large effect comes from several high-warming models (orange labels) receiving systematically lower weights.

5/🧵

What about independence of information (when calculating statistics)? Some climate models are "related" due to, e.g., shared components. We identify them based on their output, cluster them in a #CMIP6 family tree 🌳, and account for the dependence.

6/🧵

Evaluation of the method: since there are no observations from the future we use all the last-generation #CMIP5 models as pseudo observations. Median skill increases for all cases by 12-22%! 💪

7/🧵

We also provide additional information such as weights and warming per model in the supplement to our study. 📊🤓
doi.org/10.5194/esd-11…

8/🧵

Bonus: The Climate model Weighting by Independence and Performance (ClimWIP) method is freely available on @github:
github.com/lukasbrunner/C…

And now also within the @ESMValTool:
docs.esmvaltool.org/en/latest/reci…

9/🧵

Thanks a lot to all the collaborators in this project! 🙏 @apuffycloud @ClimateFlavors @AnnaMerrifield1 Ruth Lorenz and @Knutti_ETH

Thanks also to @EU_H2020 for funding @EUCP_H2020 & @CRESCENDO_H2020 and to all the data and software providers!

/end