1    

In this study, we used 26 CMIP5 models with a total of 39 members. Detailed information of the CMIP5 models and members are described in Table S1.

Table S1. The CMIP5 models used in this study

Model Institution NO. of members

Spatial Resolution (Lon×Lat, degree)

1. ACCESS1.0 Commonwealth Scientific and Industrial Research Organization (CSIRO), Australia and Bureau of Meteorology (BOM), Australia

1 1.875×1.25

2. ACCESS1.3 1 1.875×1.25

3. BCC-CSM1.1 Beijing Climate Center, China Meteorological Administration 1 2.81×2.77

4. CanESM2 Canadian Centre for Climate Modelling and Analysis, Canada 1 2.81×2.79

5. CCSM4 National Center for Atmospheric Research, USA 3 1.25×0.9 6. CESM1-BGC National Science Foundation, Department of Energy,

NCAR, USA 1 1.25×0.9

7. CESM1-CAM5 1 1.25×0.9 8. CMCC-CM Euro-Mediterraneo sui Cambiamenti Climatici, Italy 1 0.75×75 9. CMCC-CMS 1 1.875×1.86

10. CNRM-CM5 Centre National de Recherches Meteorologiques, Meteo-France, France 1 1.41×1.40

11. CSIRO-Mk3.6.0 Commonwealth Scientific and Industrial Research Organization (CSIRO), Australia 3 1.875×1.86

12. GFDL-ESM2M NOAA Geophysical Fluid Dynamics Laboratory, USA 1 2.5×2.0 13. GFDL-ESM2G 1 2.5×2.0 14. HadGEM2_CC Met Office Hadley Centre, UK 1 1.875×1.25 15. HadGEM2_ES 4 1.875×1.25 16. INM-CM4 Institute for Numerical Mathematics, Russia 1 2.0×1.5 17. IPSL-CM5A-LR

Institut Pierre-Simon Laplace, France 3 3.75×1.875

18. IPSL-CM5A-MR 1 2.5×1.25 19. IPSL-CM5B-LR 1 3.75×1.875 20. MIROC-ESM Atmosphere and Ocean Research Institute (The University

of Tokyo), National Institute for Environmental Studies and Japan Agency for Marine-Earth Science and Technology

1 2.81×1.77

21. MIROC-ESM-CHEM 1 2.81×1.77 22. MIROC5 3 1.41×1.39 23. MPI-ESM-LR Max Planck Institute for Meteorology, Germany 3 1.875×1.85 24. MPI-ESM-MR 1 1.875×1.85 25. MRI-CGCM3 Meteorological Research Institute, Japan 1 1.125×1.125 26. NorESM1-M Norwegian Climate Centre 1 2.5×1.875

2    

Figure S1. Number of CAO events per winter during 1981-2000 among the five reanalysis products including NCEP/NCAR Reanalysis 1 (NCEP1); NCEP-DOE Reanalysis 2 (NCEP2); ERA-40; ERA-Interim and NARR, CMIP5 ensemble mean (26 models with 39 members), CCSM4 and WRF.

Figure S2. Percentage change of the mean winter CAO duration simulated by the CMIP5 models (top panel), and the CAO changes due to mean warming (second panel), change of standard deviation (third

3    

panel), and change of higher moments (bottom panel) in the 2030s (2026-2045) under RCP 8.5 (left) and the 2090s (2081-2100) under RCP 4.5 (right) compared to the present day (1981-2000) using the present threshold. In Figure S2a-d, all areas satisfy the criterion that 75% of models agree with the sign of change from the CMIP5 ensemble mean. The black diamonds in Figure 2e-h indicate areas where 75% of models agree with the sign of change from the CMIP5 ensemble mean. In areas where models show agreement, the white squares indicate areas where 75% of the models show statistically significant change of winter CAO duration at 95% level.

Figure S3. Changes of temperature (left) and standard deviation (right) under RCP 8.5 2090s (2081-2100) compare to present (1981-2000) using CMIP5 mean. Only areas satisfying the criterion that 75% CMIP5 models agree with the sign of CMIP5 mean change are shown.

4    

Figure S4. Distributions of monthly mean snow water equivalent in 1981-2000: Left: the National Aeronautics and Space Administration (NASA) Earth Observations (NEO: http://neo.sci.gsfc.nasa.gov); right: WRF simulations.

Figure S5. Duration of the longest CAO event at present and future from the CMIP5 mean (26 models and 39 members): black crosses indicate where the CMIP5 mean events last longer than 5 days in future scenarios

5    

Figure S6. The cold air outbreak durations at present (1981-2000) using wind chill temperature (WCT_MIN,WCT_MAX and WCT_AVG represent minimum, maximum and average wind chill warning temperature, respectively; 13 models with 17 members) and daily average temperature (26 models with 39 members)

Figure S7. Change of winter mean wind speed. Black stipples indicate model agreement (75% of models agree with the sign of the CMIP5 mean); red stipples indicate areas with model agreement and statistical significant change (75% of the models showing agreement shows statistical significance)

6    

Figure S8. The same as Figure S7 but for winter wind speed during CAO events