Constraining internal surface temperature variability and its implications for detection and attribution

The latest generation of climate models (CMIP6) have very different large-scale surface temperature variability and inconsistencies with observed climate have been found in the variability in several regions. Given that detection and attribution, in common with many climate analyses, relies on model internal variability for uncertainty ranges, it is crucial to better constrain this variability. Here, we compare the latest climate models to observed variability to determine where and on what timescales discrepancies occur, with the models found to be, in general, too variable on annual timescales over land and with not as much variability as the observations particularly over the Southern oceans at multi-decadal timescale. We further use paleo-proxy reconstructions, supported by observational datasets finding that the majority of models have variability consistent with large-scale mean temperature on multi-annual and multi-decadal timescales. Finally, the presentation will explore the implications of these findings on key detection and attribution analyses, in particular the attribution of warming since pre-industrial times to anthropogenic forcings.