Reduced spread of simulated global warming patterns among CMIP6 models with accelerated pace of warming

Uneven economic impacts of climate change have been caused by differentiated warming rates across different geographical regions, threatening the well-being of millions to billions of people. Region-dependent historical and future warming rates are often obtained from global climate models, which, however, exhibit wide spreads in both global mean temperature change and regional deviates. While the multi-model spread in global mean warming rate has been widely reported in past literature, the multi-model spread in terms of global warming pattern and its temporal evolution remain unclear. Here we show that the multi-model spread in the simulated global warming pattern is dependent on the level of warming. We find that the simulated global warming pattern deviates substantially among CMIP6 models before 1985. The multi-model consistency rises afterwards, as the greenhouse gases level and global mean warming rate increase. Furthermore, the consistency of model-predicted future warming pattern varies by emission scenario. Models predict highly consistent warming patterns under the high emission scenario during the entire 21st century; whereas under low and intermediate emission scenarios, future warming patterns diverge among these models around middle of the 21st century. While our study detects the anthropogenic signal in the temporal evolution of multi-model consistency in the global warming pattern, the physical mechanisms underlying such varying multi-model consistency in the warming pattern merits further investigation.