Robust changes in global subtropical circulation under greenhouse warming

The lower tropospheric subtropical circulation (SC) is characterized by monsoons and subtropical highs, playing an important role in global teleconnections and climate variability. The SC changes in a warmer climate are influenced by complex and region-specific mechanisms, resulting in uneven projections worldwide. Here, we present a method to quantify the overall intensity change in global SC, revealing a robust weakening across CMIP6 models. The weakening is primarily caused by global-mean surface warming, and partly counteracted by the direct CO2 effect. The direct CO2 effect is apparent in the transient response but is eventually dominated by the surface warming effect in a slow response. The distinct response timescales to global-mean warming and direct CO2 radiative forcing can well explain the time-varying SC changes in other CO2 emission scenarios. The declined SC implies a contracted monsoon range and drying at its boundary with arid regions under CO2-induced global warming.


ACCESS-CM2
X X ACCESS-ESM1-5    Because the SST warming pattern is not provided in the d4PDF, the change in surface air temperature is shown here as a reference for the SST forcing pattern.

Figure S1 .
Figure S1.Changes in the 850 hPa streamfunction (shading) in the SSP5-8.5 experiment relative to the historical experiment in each month.The contours represent the climatology of the 850 hPa streamfunction in the historical experiment (interval: 5´10 6 m 2 s −1 ).Hatching indicates that the change is robust (see Methods for the details of the criteria).

Figure S2 .Figure S3 .
Figure S2.As in Fig.2, but only for the Northern Hemisphere.

Figure S4 .
Figure S4.As in Fig.2, but for the divergent component of subtropical circulation represented by the velocity potential at 850 hPa in the SSP5-8.5 runs.

Figure S6 .
Figure S6.Monthly climatology of the 850 hPa streamfunction in the historical experiment.The red lines indicate the center of the subtropical circulation in each hemisphere (see Methods for details).

Figure S7 .
Figure S7.Changes in the 850 hPa streamfunction (shading) in the amip-p4K experiment relative to the amip experiment in each month.The contours represent the climatology of the 850 hPa streamfunction in the amip experiment (interval: 5´10 6 m 2 s −1 ).Hatching indicates that the change is robust (see Methods for the details of the criteria).

Figure S8 .
Figure S8.Changes in the sea surface temperature (shaded) and 850 hPa vector wind (vectors) in the amip-future4K experiment relative to the amip-p4K experiment in each month.

Figure S9 .
Figure S9.As in Fig.S7, but for the amip-future4K experiment relative to the amip-p4K experiment.

Figure S10 .
Figure S10.As in Fig. S5, but for the monthly and latitudinal intensity changes in 850 hPa streamfunction.There are 15 members for each future experiment.Only the results of the 15-member ensemble mean are shown at here.

Figure S12 .
Figure S12.Seasonal changes in 850 hPa zonal wind (120°E-120°W) projected onto the climatology under direct CO2 radiative forcing.The black curves indicate the maximum of zonal-mean 850 hPa climatological zonal wind in the amip experiment in each hemisphere.Hatching indicates that the change is robust (see Methods for the details of the criteria).

Figure S13 .
Figure S13.Changes in the surface temperature in the amip-4xCO2 experiment relative to the amip experiment in each month.

Figure S14 .
Figure S14.The prescribed diabatic heating and cooling for six regions used in the linear baroclinic model.

Figure S15 .
Figure S15.Vertical profile of the prescribed diabatic heating and cooling shown in Fig. S14.

Figure S16 .
Figure S16.The steady response in 850 hPa streamfunction to the prescribed diabatic heating and cooling shown in Fig. S14.

Figure S17 .
Figure S17.Taylor diagram of monthly climatology of the 850 hPa streamfunction in the historical experiment for 34 CMIP6 model covering 60°S-60°N with respect to data from ECMWF Reanalysis v5.The time span is from 1979 to 2008.

Table S1
CMIP6 models and experiments used in this study.