Evaluation of AMIP models from CMIP6 in simulating winter surface air temperature trends over Eurasia during 1998–2012 based on dynamical adjustment

The relationship between winter cooling in Eurasia and Arctic amplification during the period 1998–2012 under global warming has received increasing attention in recent years. This relationship is controversial and is often studied using model simulations. However, the process of evaluating these models is challenging as a result of the different internal variability among models and between the models and observations. We applied a dynamical adjustment method based on constructed circulation analogs to the model simulations and observations to remove the effects of the internal variability of the atmosphere and then evaluated the performance of the models in simulating the winter surface air temperature (SAT) trends over Eurasia from 1998 to 2012 based on 11 models of the Atmospheric Model Intercomparison Project (AMIP) from phase 6 of the Coupled Model Intercomparison Project. Our results show that the overall performance of all the model ensemble simulations was poor, but was much improved after applying dynamical adjustment, with the median values of the 11 AMIP ensemble simulations fairly close to the observed winter SAT trends averaged over Eurasia. When considering both the model-simulated SAT trends averaged over Eurasia and the skill scores of the trend pattern, the HadGEM3-GC31-LL simulation gave the best performance among the models with multiple runs. This method allows a more objective evaluation of the performance of models and provides an alternative way to evaluate the ability of models to simulate the “warm Arctic and cold Eurasia” trend pattern. The cold Eurasia, especially central Eurasia, in the observations is found to be mainly induced by the contribution from the internal variability of the atmosphere.