Pengfei LIN, Bowen ZHAO, Jilin WEI, Hailong LIU, Wenxia ZHANG, Xiaolong CHEN,Jie JIANG, Mengrong DING, Wenmin MAN, Jinrong JIANG, Xu ZHANG,Yuewen DING, Wenrong BAI, Chenyang JIN, Zipeng YU, Yiwen LI,Weipeng ZHENG, and Tianjun ZHOU

1State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG),Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China

2College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing 100049, China

3Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190, China

4Group of Alpine Paleoecology and Human Adaptation (ALPHA), State Key Laboratory of Tibetan Plateau Earth System,Resources and Environment (TPESRE), Institute of Tibetan Plateau Research,Chinese Academy of Sciences, Beijing 100864, China

5Key Laboratory of Western China’s Environmental Systems (Ministry of Education), College of Earth and Environmental Science, Lanzhou University, Lanzhou 730000, China

Table S1. The names of 29 CMIP6 models whose historical and SSP5-8.5 simulations are used in the study.

Fig. S1. The RMSE of annual global SAT (GSAT) compared with the 110-member ensemble mean of FGOALS-g3. The x-axis label is the sample size of subsets from the 110 members, and the y-axis label is the RMSE from subset-estimated GSAT compared with the 110-member estimated “true” GSAT forced response. The black line is the mean RMSE for GSAT for subset members 2 to 110. The RMSE is computed 10000 times, so the gray shaded area shows the range of 10000 RMSEs for each subset.

Fig. S2. The observed SAT in (a) BEST and (b) HadCRUT5 for 1951-80. The simulated SAT bias in FGOALS-g3 relative to (c) BEST and (d) HadCRUT5 for 1951-80. The simulated SAT bias in (e) CSIRO large ensembles, (f)CESM large ensembles, and (g) SMHI large ensembles relative to HadCRUT5 for 1951-80. The global mean SAT bias (ensemble mean minus observstion) and PCC for climatology between observations and the large ensemble mean of model are shown.

Fig. S3. The zonal averaged warming amplification factor, defined as the change in zonal average temperature relative to the global mean temperature change (Holland and Landrum 2015), for FGOALS-g3 and 29 CMIP6 models. The change is computed as the average for 1995-2014 minus the average for 1961-90. (a) FGOALS-g3 super-large ensembles, (b) 29 CMIP6 models (Models’ name see Table S1). The black line in (a) is the ensemble mean in FGOALS-g3 large ensembles, and (b) is the multi-model ensemble mean (MME) in CMIP6, respectively.The red line is the HadCRUT5.

Fig. S4. The simulated (a) SAT and (b) land precipitation change (1995-2014 minus 1961-1990) over the globe in FGOALS-g3 large ensembles and observations. The observed SAT and land precipitation are from HadCRUT5 and CRU, respectively.