Suarez-Gutierrez, LauraMaher, Nicola2026-01-312026-01-312041-1723PubMed:41390522ORCID:/0000-0003-3922-9833/work/203984359https://hdl.handle.net/1885/733805138Changes in temperature variability affect the frequency and intensity of extreme events, as well as the regional range of temperatures that ecosystems and society need to adapt to. While accurate projections of temperature variability are vital for understanding climate change and its impacts, they remain highly uncertain. We use rank-frequency analysis to evaluate the performance of eleven single model initial-condition large ensembles (SMILEs) against observations in the historical period, and use those that best represent historical regional variability to constrain projections of future temperature variability. Constrained projections from the best-performing SMILEs still show large uncertainties in the intensity and the sign of the variability change for large areas of the globe. Our results highlight poorly modelled regions where observed variability is not well represented such as large parts of Australia, South America, and Africa, particularly in their local summer season, underscoring the need for further modelling improvements over crucial regions. In these regions, the constrained projected change is typically larger than in the unconstrained ensemble, suggesting that in these regions, multi-model mean projections may underestimate future variability change.N.M. was supported by the Australian Research Council Discovery Early Career Researcher Award DE230100315. L.S.G. received funding from the European Union’s Horizon Europe Framework Programme under the Marie Skłodowska-Curie grant agreement No. 101064940. We thank the Deutsches Klimarechenzentrum (DKRZ) for providing the necessary computational resources to carry out this work. We thank Sebastian Milinski for computing the warming levels used in this project and both Sebastian Milinski and Jochem Marotzke for their intellectual contributions to the development of this project in its early stages. We additionally thank Thomas Frölicher and Dirk Olonscheck for providing data and access for GFDL-ESM2M and early access to MPI-GE-CMIP6, respectively. We acknowledge the US CLIVAR Working Group on Large Ensembles for their provision of the Multi-Model Large Ensemble data and we the World Climate Research Programme, which, through its Working Group on Coupled Modelling, coordinated and promoted CMIP6. We thank the climate modelling groups for producing and making available their model output, the Earth System Grid Federation (ESGF) for archiving the data and providing access, and the multiple funding agencies who support CMIP6 and ESGF. Open access funding provided by Swiss Federal Institute of Technology Zurich.16en© 2025 The Authors202610.1038/s41467-025-67005-y105027174977