Origin of the seasonally-dependent response of the subtropical highs

Origin of the seasonally-dependent response of the subtropical highs and tropical precipitation in a warming climate Fengfei Song, L. Ruby Leung, Jian Lu, and Lu Dong Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, Richland, Washington, USA Submitted to Nature Climate Change January-02-2018 Corresponding authors: Fengfei Song (Fengfei.Song@pnnl.gov) and L. Ruby Leung (Ruby.Leung@pnnl.gov) 1

The subtropical highs are semi-permanent atmospheric features that strengthen during April-September and exert large influences on regional precipitation 1-5. Previous studies of their future changes mainly focused on their peak season (June-August) 6-9. Here we find a robust seasonally-dependent response of the subtropical highs to warming in a suite of multimodel simulations. Both the North Pacific and North Atlantic subtropical highs strengthen more in April-June than July-September, with opposite changes in the Southern Hemispheric counterpart. These responses are closely related to a southward shift of tropical precipitation in April-June relative to July-September, manifesting in a seasonal delay of tropical precipitation and monsoon rainfall onset in the Northern Hemisphere 10-11. From theory and analysis of atmospheric energetics, it is found that in a warmer world, the Northern Hemisphere needs more latent energy to warm up during April-June as constrained by the Clausius-Clapeyron relation, and the opposite occurs in the Southern Hemisphere. The interhemispheric energy contrast drives a southward shift of tropical precipitation that strengthens the Hadley cell and the subtropical highs in the Northern Hemisphere in April-June. These changes scale linearly with warming, so they have increasing implications for projecting regional climate changes in the tropics and subtropics as the warming continues. The subtropical highs are associated with the descending branch of the Hadley cell and shaped by the land-sea distribution. In the Northern Hemisphere, they are most evident over the subtropical oceans, known as the North Pacific subtropical high (NPSH) and North Atlantic subtropical high (NASH). The Southern Hemisphere counterpart is more zonally distributed due to the weaker land-sea contrast. During April-September, the NPSH and NASH strengthen and exert substantial influences on the precipitation over East Asia and North America, homes to more 2

than 2 billion people, by modulating the moisture transport 1-2 and tropical cyclone tracks 3-5 directed to the regions. Meanwhile, the Southern Hemisphere subtropical high also strengthens and the associated trade winds transport substantial moisture to the Northern Hemispheric monsoon regions, including the Sahel, South Asia, East Asia and North America 2,12-13. Previous studies suggest that under global warming, the NASH will intensify and move westward during June-August 6-7, but changes in the NPSH are more uncertain 6-9. Few studies have paid attention to seasons other than June-August, even though the subtropical highs are also well defined and substantively strong during April and May (Supplementary Fig. 1). Even less well studied is how the Southern Hemisphere subtropical high may change under global warming 14. Here we find that changes of the subtropical highs under global warming have marked seasonal dependence, which is related to the tropical precipitation delay noted in previous studies 10-11. These seasonally-dependent changes originate from the climatological seasonal cycle of temperature under the constraint of the Clausius-Clapeyron relation. These results have significant implications for projecting changes in the global hydrological cycle and the regional climate of the tropics and subtropics. Characteristics of the seasonally-dependent responses under global warming Here the historical (HIST) and Representative Concentration Pathway 8.5 (RCP8.5) simulations from 27 CMIP5 (Coupled Model Intercomparison Project phase 5; ref. 15; see Methods) models are used to represent the current and future climates, respectively. Figure 1 shows the present day subtropical highs and future changes in April-June and July-September. Under global warming, both the NPSH and NASH intensify more evidently during April-June than July-September (Fig. 1c). For the NPSH, its changes in both April-June and July-September show large inter-model spreads (Supplementary Fig. 2), but the differences between the two seasons are 3

more consistent: 24 out of 27 models exhibit more strengthening during April-June. There is a greater consensus on the change of the NASH during April-June, with all models unanimously showing strengthening. As a regional manifestation of the NASH, the Great Plains low-level jet also strengthens more during April-June than July-September, consistent with a previous study 16. Land-sea thermal contrast has been broadly invoked to explain the summertime changes of the subtropical highs and low-level jet 8-9,16. For both the East Asia-Pacific and North America-Atlantic regions, land-sea thermal contrast is enhanced under warming, but with slightly more enhancement during July-September, so it cannot explain why the NPSH and NASH are enhanced more during April-June. This seasonal dependence also occurs in the Southern Hemisphere counterpart, with stronger enhancement during July-September than April-June instead (26 out of 27 models). Consistently, the mid-latitude westerly at the poleward flank of the subtropical high also shows a greater increase during July-September in all models. Besides the subtropical highs, tropical precipitation also exhibits a clear seasonal dependence under global warming (Fig. 2). Broadly, the precipitation response is characterized by the wet-get-wetter 17-18 and warmer-get-wetter 19-20 patterns, with the former representing an enhancement of the climatological seasonal cycle pattern due to the increase of moisture with warming, and the latter manifesting as larger precipitation increase at the equatorward flank of the climatological rainband following the accentuated equatorial sea-surface temperature (SST) warming 21 (Figs. 2a-d). Closer inspection reveals evident deviations of the tropical precipitation change from the climatological seasonal cycle (Figs. 2b and d) and a conspicuous dipole change straddling the climatological peak during June-August at about 10 N (Figs. 2a and c), signaling a seasonal delay in the arrival of monsoonal rainfall 10-11. Removal of the wet-get-wetter and warmer-get-wetter components (see Methods) amplifies the suppression of precipitation in the 4

early warm season in both hemispheres during their respective transition from spring to summer (Figs. 2e-f). Figure 3 shows the seasonal differences in the precipitation and circulation changes between April-June and July-September. Besides the stronger NPSH and NASH changes during April-June, high-pressure anomalies extending from North Atlantic to the Eurasia continent form a quasi-zonal high-pressure band over the northern subtropics (Fig. 3a). An opposite but more zonally-uniform differential response is found in the Southern Hemispheric subtropical high and mid-latitude westerly at the poleward flank. Notably, the seasonal differences of tropical precipitation change also exhibit a zonally-uniform dipole pattern, with negative and positive changes north and south of the equator, respectively. Changes in the position of tropical precipitation and the associated peak heating have a profound asymmetric influence on the strength of the Hadley cell in both hemispheres. The inter-seasonal dipolar precipitation anomaly (Fig. 3a) is expected to enhance the Northern Hemispheric Hadley cell and weaken the Southern Hemispheric counterpart, according to the Hadley cell theory under interhemispherically asymmetric forcing 22. This is confirmed by the alternating bands of vertical velocity response in Fig. 3b, showing an enhanced/weakened descent near 35 N/35 S (also see Supplementary Fig. 3 for the streamfunction response) in April-June relative to July-September under warming. Coinciding with the descending branch of the Hadley cell, the subtropical highs in the two hemispheres change correspondingly (Fig. 3a). The center of the Hadley cell and the tropical precipitation shift southward in almost all models during April-June relative to July-September (Supplementary Fig. 4), agreeing with the robust differences in the changes of subtropical highs among the models. 5

To understand the physical processes of the seasonally-dependent changes, the climate response can be decomposed into components associated with direct radiative CO 2 forcing and indirect SST warming. We use AMIP (Atmospheric Model Intercomparison Project) experiments in which SST is prescribed (see Methods) to estimate the two components. The component associated with direct radiative forcing can be quantified by quadrupling the CO 2 concentration without SST changes (AMIP4xCO2), while the component associated with the indirect SST warming can be estimated by adding an SST warming pattern derived from the coupled atmosphere-ocean model response to increased CO 2 (AMIPFuture) or increasing SST uniformly by 4 K (AMIP4K). The combined responses of atmospheric circulation and tropical precipitation to direct radiative forcing and indirect SST warming resemble the response under the RCP8.5 scenario (Supplementary Fig. 5), lending credence to our approach. Under the direct radiative forcing (Figs. 3c-d), seasonal differences in the changes of the subtropical highs in both hemispheres are generally weak. In contrast, both experiments with the SST warming pattern (Figs. 3e-f) and uniform SST warming (Figs. 3g-h) reproduce the seasonallydependent responses of RCP8.5 well (Figs. 3a-b). The seasonal delay of tropical precipitation is also reproduced well under the SST warming pattern (Supplementary Fig. 6). Consistent with the anomalies of tropical precipitation, the two SST warming experiments also reproduce the strengthening and weakening of the Hadley cell in the Northern Hemisphere and Southern Hemisphere, respectively, in April-June compared to July-September (Supplementary Fig. 3). Hence, the seasonally-dependent responses, including the seasonal differences in the weakening/strengthening of the subtropical highs and the tropical precipitation seasonal delay, can be attributed to SST warming. Origin of the seasonally-dependent response to global warming 6

Recent advances 23-25 in understanding the shift of tropical precipitation from an energetics perspective provides a framework to understand why SST warming can cause a seasonallydependent atmospheric response globally. In the energetics framework 23-25, a southward displacement of tropical precipitation corresponds to a northward cross-equatorial atmospheric energy transport anomaly, and vice versa. This relation is dictated by the atmospheric structure in the tropics: the direction of atmospheric energy transport is determined by the upper branch of the Hadley cell, while the interhemispheric moisture transport is in an opposite direction, dominated by the lower branch of the Hadley cell. The atmospheric energy transport is determined by the difference between the net input energy F #$% and the tendency of the moist static energy &'() (Eq. (3) in Methods). With uniform SST warming or SST warming pattern, the seasonal difference in the change of F #$% between April-June and July-September has strong meridional variations but small interhemispheric contrast (Fig. 4a). However, the seasonal difference in the change of the negative moist static energy tendency &'() shows a strong interhemispheric contrast, with the Southern Hemisphere 1.74 and 2.16 W m -2 higher than the Northern Hemisphere in the uniform SST warming and SST warming pattern, respectively (Fig. 4b). Further decomposing h suggests that the seasonal difference in the change of the negative latent energy tendency &', -.) (Fig. 4c) accounts for most of the interhemispheric contrast shown in Fig. 4b. The interhemispheric energy contrast requires a northward cross-equatorial energy transport for energy balance (Fig. 4d), which is 0.42 and 0.24 PW in the SST warming pattern and uniform warming, respectively. The cross-equatorial energy transport is closely related to the southward shift of tropical precipitation in both experiments (Supplementary Fig. 7), with the moist static energy tendency contributing most to 7

the energy transport. Among the moist static energy, the contribution from latent energy is dominant, accounting for 70% and 85% under the SST warming pattern and uniform warming, respectively. Based on the Clausius-Clapeyron relation, the change in the latent energy tendency &', -.) under global warming is proportional to the climatological temperature tendency &0 1 (see the derivation in Methods): &', -.) a T 5 &0 1 (1) where a is a parameter determined by the present climate state, T 5 is the temperature change. Since both a and T 5 have weak seasonal variations, the seasonal cycle of &', -.) is mainly determined by the seasonal cycle of &0 1, with correlation coefficients of 0.94 and 0.90 for the uniform SST warming and SST warming pattern, respectively (Supplementary Fig. 8). From Eq. (1), it is easy to see why the latent energy tendency change &', -.) has a striking interhemispheric contrast between April-June and July-September: during April-June, the Northern Hemisphere begins to warm up ( &0 1 > 0) and the Southern Hemisphere begins to cool down (&0 1 < 0), while July-September is the warmest and coldest season in the Northern and Southern Hemispheres, respectively, with weak temperature tendency ( &0 1 0; Supplementary Fig. 9). In a warmer world (i.e., positive T 5 ), more latent energy is needed to warm up the Northern Hemisphere during April-June, and vice versa for the Southern Hemisphere. To achieve an energy balance, the atmosphere must transport more energy from the Southern Hemisphere to the Northern Hemisphere. Thus, the Hadley cell and tropical precipitation shift southward during April-June relative to July-September. This anomalous energy transport originated from the climatological 8

seasonal cycle of temperature is also evident, but in the opposite direction, during October- December (Supplementary Fig. 10), when the Northern Hemisphere begins to cool down and the Southern Hemisphere begins to warm up. In summary, the climatological seasonal cycle of temperature dictates that under global warming, the transitional season (April-June and October-December) engenders larger interhemispheric latent energy transport to the hemisphere that is warming up seasonally, as required by the Clausius-Clapeyron relation. Under this energy constraint, tropical precipitation shifts to the hemisphere that is getting cold seasonally, hence resulting in a seasonal delay in the migration of tropical precipitation from the cold hemisphere to the warm hemisphere. Relative to July-September, the southward shift of the tropical precipitation during April-June strengthens the Hadley cell in the Northern Hemisphere and weakens it in the Southern Hemisphere. As an integral part of the descending branch of the Hadley cell, the NPSH and NASH are also intensified, while the Southern Hemispheric subtropical high is weakened in April-June relative to July-September (Supplementary Fig. 11). Since the change in the latent energy tendency scales linearly with global warming from Eq. (1), it is expected that the seasonally-dependent responses of the subtropical highs and tropical precipitation delay will become more evident as global warming continues. This has increasing implications for projecting future changes in regional precipitation. As both the NASH and NPSH tend to strengthen more during April-June consistently across the models, their influences on East Asia and North America may be enhanced in this season. For example, CMIP5 models project precipitation increases over the northern US and decreases over the southern US during April-June relative to July-September, consistent with the enhanced NPSH and Great Plains low-level jet (Supplementary Fig. 12). As moisture transport by the Great Plains low-level jet is important for 9

the mesoscale convective systems in the central and midwestern US 26-27, which have yet to be resolved in global climate models, the robust enhancement of the low-level jet during April-June may provide an important constraint for projecting the future precipitation change. 10

Methods 1. Model simulations In this study, monthly mean output from a suite of Coupled Model Intercomparison Project Phase 5 (CMIP5) model simulations are used (see Supplementary Table 1). The latter include two sets of atmosphere-ocean coupled experiments for the historical (HIST) and RCP8.5 simulations from 27 CMIP5 models, and four sets of fixed-sst atmosphere-only experiments (AMIP, AMIP4xCO2, AMIP4K and AMIPFuture simulations) from 11 of the 27 CMIP5 models that include outputs of all the selected AMIP-type simulations. The present and future climates are defined as RCP8.5 during 2056-2099 and HIST during 1962-2005, and the response to global warming is defined as their difference. The standard AMIP simulation is run with observed SSTs and sea-ice and prescribed anthropogenic forcing. In addition to the standard AMIP forcing, three additional AMIP experiments include quadrupling of CO 2 forcing (AMIP4xCO2), uniform 4 K SST warming (AMIP4K) and SST warming pattern from the 1% CO 2 coupled CMIP3 model experiments at the time of CO 2 quadrupling (AMIPFuture). AMIP4xCO2 AMIP estimates the climate response to direct radiative forcing, AMIP4K AMIP estimates the climate response to uniform SST warming, and AMIPFuture AMIP estimates the climate response to SST warming pattern. The CMIP5 datasets used in this study can be found at http://www.ipccdata.org/sim/gcm_monthly/ar5/reference-archive.html. 2. Definitions of atmospheric circulation systems and land-sea thermal contrast The NPSH is defined by the sea-level pressure averaged over (25 N-45 N, 180 E-130 W) and the NASH is defined by the sea-level pressure averaged over (25 N-45 N, 70 W-20 W). The Great Plains low-level jet is defined by the 925 hpa meridional wind averaged over (26 N-38 N, 11

92 W-102 W), the Southern Hemispheric subtropical high is defined by the sea-level pressure averaged over (20 S-40 S, 180 E-180 W), and the Southern Hemispheric mid-latitude westerly is defined by the 925 hpa zonal wind averaged over (40 S-55 S, 180 E-180 W). The East Asia- Pacific land-sea thermal contrast is defined as the surface temperature difference between the land points and ocean points over (20 N-50 N, 100 E-180 E); the North America-Atlantic land-sea thermal contrast is defined as the surface temperature difference between the land points over (30 N-50 N, 60 W-120 W) and ocean points over (30 N-50 N, 10 W-100 W). The center of the Hadley cells is defined as the latitude of zero crossing of the stream function averaged between 400-600 hpa between the Northern and Southern hemispheric cells. The tropical precipitation asymmetry is defined as the zonal-mean precipitation difference between the northern tropics (0-25 N) and the southern tropics (0-25 S). 3. Removing the warmer-get-wetter pattern and wet-get-wetter pattern To remove the warmer-get-wetter pattern, tropical precipitation change is regressed upon the tropical SST change between RCP8.5 and HIST, and the regressed change is removed from the simulated changes. To remove the wet-get-wetter pattern, tropical precipitation change is regressed upon the climatological tropical precipitation, and the regressed change is considered as the wet-get-wetter pattern and removed from the simulated changes. 4. Atmospheric energy analysis According to the atmospheric energy equation, the divergence of atmospheric energy transport AHT is equal to the difference between the net energy input to the atmosphere F #$% and the moist static energy tendency &'() : AHT = F #$% &'() (2) 12

Where, <> represents the vertical integration between the surface and the top of the model, F #$% = F?@A + F 0CD, F?@A is the net heat flux entering the atmosphere from the surface, F 0CD is the net radiation at the top of the atmosphere, h = c F T + gz + L J q is the moist static energy, c F is specific heat at constant pressure, T is air temperature, g is acceleration due to gravity, z is the geopotential height, L J is latent heat of vaporization, q is specific humidity. The energy transport AHT φ at latitude φ can be calculated by integrating Eq. (2) as follows: AHT φ = R N S TQ N NQ R (F #$% &'() ) cosφdλdφ (3) Where, R is the radius of the Earth, φ is latitude and λ is longitude. Positive (negative) AHT φ means northward (southward) energy transport. Specifically, the cross-equatorial energy transport AHT 0 is expressed as: AHT 0 = R N R TQ N NQ R (F #$% &'() ) cosφdλdφ (4) 5. Relating the latent energy tendency change to the climatological temperature tendency The latent energy L J q is closely related to specific humidity q, which is the product of relative humidity r and saturation specific humidity q 5 : L J q = L J rq 5 (5) The saturation specific humidity q 5 is a thermodynamic function of pressure and temperature. For simplicity, it can be written as q 5 = Z$ 1(0) F (6) 13

where e 5 is the saturation vapor pressure and ε is the ratio of the gas constants for dry air and water vapor. The dependence of saturation vapor pressure e 5 on temperature is controlled by the Clausius-Clapeyron equation: ] $ 1 ^$ 1 ^0 =, - _ - 0` (7) Where R J is the gas constant for water vapor. Eq. (7) can be integrated from e 5R and T R to obtain an explicit relationship between e 5 and T: a- e 5 = e 5R eb- ( c de Tc d ) (8) Combining Eqs. (6) and (8), we can obtain the dependence of the saturation specific humidity q 5 on the temperature T: q 5 = Z$ a- 1e e b- ( c de Tc d ) (9) F At a given pressure level p, the saturation specific humidity tendency can be obtained by the time derivative of Eq. (9): &. 1 = Z, a- -$ 1e e b- ( c de Tc d ) &0 F_ - 0` (10) The change of saturation specific humidity tendency between the present and future climatological states (denoted with subscripts c and f, respectively) can be written as: &. 1 = Z, -$ 1e a- b- e ( c de T c ) d &0 g g F_ - 0 g` a- eb- ( c de T c dh ) &0 h F_ - 0 h` Z, -$ 1e (11) T i = T j + T, where T is the temperature change. Since T is an order of magnitude smaller than T j, Taylor expansion can be used to simplify Eq. (11) as: 14

&. 1 Z, a- -$ 1e( b- TN0 h) F_ - 0 k h a- eb- ( c de T c dh ) T &0 h (12) Considering the relative humidity r doesn t vary much at the seasonal cycle and in the future change at the hemispheric scale, we can integrate Eq. (12) vertically, relate q 5 to L J q using Eq. (5) and obtain the vertically-integrated latent energy tendency change &', -.) as: &', -.) < Z, a- -`l$ 1e b- TN0 h F_ - 0 k h a- eb- c de T c dh T &0 h > (13) Since the vertically-integrated moisture change is mainly determined by the lower troposphere, T and &0 h can be estimated by the lower tropospheric or near-surface air temperature T 5. Therefore, Eq. (13) can be further simplified as: &', -.) a T 5 &0 1 (14) Here, a =< Z, -`l$ 1e a- b- TN0 h F_ - 0 k h a- eb- c de T c dh >, T 5 is the near-surface air temperature. Hence, the latent energy tendency change &', -.) is determined by three factors: the first is a associated with the climatological temperature, the second is the temperature change T 5, and the third is the climatological temperature tendency &0 1. Based on the simulated values, the seasonal variations of the first two terms are only about 7% of the annual-mean in both the AMIP4K and AMIPFutue experiments at the hemispheric scale, so the seasonal variation of latent energy tendency change is mainly determined by the seasonal variation of the climatological temperature tendency. This is also confirmed in Supplementary Fig. 8, as the correlation coefficient between these two variables is 0.90 and 0.94 in the AMIPFuture and AMIP4K experiments, respectively. 15

Acknowledgement This research is supported by the U.S. Department of Energy Office of Science Biological and Environmental Research as part of the Regional and Global Climate Modeling program. This work has benefited from helpful discussion with Dr. Zhe Feng and Dr. Bob Houze of PNNL. PNNL is operated for the Department of Energy by Battelle Memorial Institute under contract DE-AC05-76RL01830. We acknowledge the World Climate Research Program s Working Group on Coupled Modeling, which is responsible for CMIP, and thank the climate modeling groups (listed in Table S1) for producing and making available their model output. For CMIP, the US DOE s Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals. 16

References 1. Li, W., Li, L., Fu, R., Deng, Y. & Wang, H. Changes to the North Atlantic subtropical high and its role in the intensification of summer rainfall variability in the Southeastern United States. J. Clim. 24, 1499 1506 (2011). 2. Zhou, T. & Yu, R. Atmospheric water vapor transport associated with typical anomalous summer rainfall patterns in China. J. Geophys. Res. 110, D08104 (2005). 3. Colbert, A. & Soden, B. Climatological variations in North Atlantic Tropical Cyclone Tracks. J. Clim. 25, 657-673 (2012). 4. Stowasser, M., Wang, Y. & Hamilton, K. Tropical cyclone changes in the Western North Pacific in a global warming scenario. J. Clim. 20, 2378 2396 (2007). 5. Wu, L., Wang, B. & Geng, S. Growing typhoon influence on east Asia. Geophys. Res. Lett. 32, L18703 (2005). 6. Li, W., Li, L., Ting, M. & Liu, Y. Intensification of Northern Hemisphere subtropical highs in a warming climate. Nat. Geosci. 5, 830 834 (2012). 7. Shaw, T. A. & Voigt, A. Tug of war on the summertime circulation between radiative forcing and sea surface warming. Nat. Geosci. 8, 560 566 (2015). 8. Seo, K.-H., Ok, J., Son, J.-H. & Cha, D.-H. Assessing Future Changes in the East Asian Summer Monsoon Using CMIP5 Coupled Models. J. Clim. 26, 7662 7675 (2013). 9. He, C. & Zhou, T. Responses of the Western North Pacific Subtropical High to Global Warming under RCP4.5 and RCP8.5 Scenarios Projected by 33 CMIP5 Models: The Dominance of Tropical Indian Ocean Tropical Western Pacific SST Gradient. J. Clim. 28, 365 380 (2015). 17

10. Biasutti, M. & Sobel, A. H. Delayed Sahel rainfall and global seasonal cycle in a warmer climate. Geophys. Res. Lett. 36, L23707 (2009). 11. Dwyer, J. G., Biasutti, M. & Sobel, A. H. The effects of Greenhouse Gas-induced Changes in SST on the Annual Cycle of Zonal Mean Tropical Precipitation. J. Clim. 27, 4545-4565. 12. Wang, B. et al. Northern Hemisphere summer monsoon intensified by mega-el Niño/southern oscillation and Atlantic multidecadal oscillation. Proc. Natl Acad. Sci. USA 110, 5347 5352 (2013). 13. Pathak, A., Ghosh, S., Martinez, J. A., Dominguez, F. & Kumar, P. Role of Oceanic and Land Moisture Sources and Transport in the Seasonal and Inter annual variability of Summer Monsoon in India, J. Clim. 30, 1839-1859 (2017). 14. Li, W. et al. Intensification of the Southern Hemisphere summertime subtropical anticyclones in a warming climate. Geophys. Res. Lett. 40, 5959 5964 (2013). 15. Taylor, K. E., Stouffer, R. J. & Meehl, G. A. An overview of CMIP5 and the experiment design. Bull. Am. Meteorol. Soc. 93, 485 498 (2012). 16. Cook, K. H., Vizy, E. K., Launer, Z. S. & Patricola, C. M. Springtime intensification of the great plains low-level jet and midwest precipitation in GCM simulations of the twenty-first century. J. Clim. 21, 6321 6340 (2008). 17. Held, I. M. & Soden, B. J. Robust responses of the hydrological cycle to global warming. J. Clim. 19, 5686 5699 (2006). 18. Chou, C., Neelin, J., Chen, C. & Tu, J. Evaluating the rich-get-richer mechanism in tropical precipitation change under global warming. J. Clim. 22, 1982 2005 (2009). 18

19. Xie, S-P. et al. Global warming pattern formation: Sea surface temperature and rainfall. J. Clim. 23, 966 986 (2010). 20. Johnson, N. C. & Xie, S-P. Changes in the sea surface temperature threshold for tropical convection. Nat. Geosci. 3, 842 845 (2010). 21. Huang, P., Xie, S. P., Hu, K. M., Huang, G. & Huang, R. H. Patterns of the seasonal response of tropical rainfall to global warming. Nat. Geosci. 6, 357 361 (2013). 22. Lindzen, R. S. & Hou, A. Y. Hadley circulations for zonally averaged heating centered off the equator. J. Atmos. Sci. 45: 2417-2427 (1988). 23. Kang, S. M., Held, I. M., Frierson, D. & Zhao, M. The response of the ITCZ to extratropical thermal forcing: Idealized slab-ocean experiments with a GCM. J. Clim. 21(14):3521 3532 (2008). 24. Kang, S. M., Frierson, D. & Held, I. M. The tropical response to extratropical thermal forcing in an idealized GCM: The importance of radiative feedbacks and convective parameterization. J. Atmos. Sci. 66(9):2812 2827 (2009). 25. Schneider, T., Bischoff, T. & Haug, G. H. Migrations and dynamics of the intertropical convergence zone. Nature 513, 45 53 (2014). 26. Feng, Z. et al. More frequent intense and long-lived storms dominate the springtime trend in central US rainfall. Nat. Commun. 7, 13429, doi:10.1038/ncomms13429 (2016). 27. Wang, S.-Y. & Chen, T.-C. The late-spring maximum of rainfall over the U.S. central plains and the role of the low-level jet. J. Clim. 22, 4696 4709 (2009). 19

Figure 1 Seasonal dependence of the changes in the subtropical highs under global warming. Climatology of the 925 hpa wind (vectors; unit: m s -1 ) and sea-level pressure (shading; unit: hpa) in the historical (HIST) experiment during (a) April-June (AMJ) and (b) July-September (JAS). (c) Future changes between RCP8.5 and HIST during AMJ (orange) and JAS (blue). From left to right: North Pacific subtropical high (NPSH; unit: hpa); North Atlantic subtropical high (NASH; unit: hpa); the Great Plains low-level jet (GPLLJ; unit: m s -1 ); land-sea thermal contrast (LSTC) over the East Asia-Pacific region (LSTC-Pac; unit: K); LSTC over the North America-Atlantic region (LSTC-Atl; unit: K); Southern Hemispheric subtropical high (SHSH; unit: hpa) and Southern Hemispheric mid-latitude westerly (SHMW; unit: m s -1 ). The error bar shows the standard deviation among models. 20

Figure 2 The seasonal delay of tropical precipitation under global warming. (Left panel) (a) Precipitation (unit: mm day -1 ) change between the RCP8.5 and HIST runs; (c) climatology of tropical precipitation in the HIST run and (e) is the same as (a) but with the wet-get-wetter and warmer-get-wetter patterns removed. (Right panel) The climatology (black line) and change (red line) in the precipitation averaged over (b) northern tropics (0-25 N) and (d) southern tropics (0-25 S). (f) The precipitation changes in the northern and southern tropics after removing the wetget-wetter and warmer-get-wetter patterns. 21

Figure 3 The difference of precipitation and atmospheric circulation changes between AprilJune and July-September under global warming. (Left panel) sea-level pressure (shading; unit: hpa), 925 hpa wind (vector; unit: m s-1) and precipitation (contour; unit: mm day-1) changes under (a) RCP8.5, (c) direct radiative forcing, (e) SST warming pattern and (g) uniform SST warming. 22

The green and purple contours on the left panels indicate positive and negative precipitation anomalies, respectively. The contour interval is 1 mm day -1 with the first positive and negative contours corresponding to 0.5 and -0.5 mm/day, respectively. (Right panel) Same as the left panel, but for the vertical-meridional distribution of zonally averaged vertical velocity (shading; unit: - 5*10-2 Pa/s) change. 23

Figure 4 The difference in atmospheric energy change between April-June and July- September under the indirect SST warming. (a) net atmospheric energy flux F #$% (unit: W m -2 ); (b) vertically-integrated negative atmospheric energy tendency &'() (unit: W m -2 ); (c) vertically-integrated negative latent energy tendency &', -.) (unit: W m -2 ) changes between AMIP4K (blue line)/amipfuture (red line) and AMIP. The pink and blue shadings represent the standard deviation among models for AMIP4K and AMIPFuture changes, respectively. (d) The cross-equatorial energy transport change (red bar) and its contributions from F #$% (blue bar), &'() (orange bar) and &', -.) (pink bar) between AMIP4K/AMIPFuture and AMIP experiments. The error bars represent the standard deviation among models. 24