We used the Community Earth System Model version 1 (CESM1), a fully coupled ocean–atmosphere–land–sea-ice model, created by the National Center for Atmospheric Research of the US (Hurrell et al., 2013). The model is a version with a finite-volume approximation 1∘ horizontal resolution (latitude 0.9∘ × longitude 1.25∘ for the atmosphere and land, and 1∘ × 1∘ for the ocean) and 30-level vertical resolution, with a rigid lid at 4 hPa. CESM1 includes the primary anthropogenic forcing agents, such as GHGs, tropospheric and stratospheric ozone, sulfate, and black and primary organic carbon. The three-mode modal aerosol model that contains the Aitken, accumulation, and coarse modes has been implemented in the model (Liu et al., 2012). It can provide the number and mass concentrations of internally mixed aerosols for the three modes. The model also includes the physical representations of aerosol direct, semi-direct, and indirect effects for both liquid- and ice-phase clouds (Morrison and Gettelman, 2008; Gettelman et al., 2010; Ghan et al., 2012).

Anthropogenic and biomass burning emissions of aerosols and their precursors are based on Lamarque et al. (2010). However, the BC emission at the present day is adjusted due to the potential underestimation of BC heating in the atmosphere in CESM1 (Xu et al., 2013; Xu and Xie, 2015). BC emissions over East Asia and South Asia are increased by a factor of 2 and 4, respectively. The emissions are changed in all economic sectors (industrial, energy, etc.) and all seasons by the same ratio. Such an adjustment significantly improved the simulated radiative forcing compared to the direct observations.

This study used a series of simulations (Table 1):

Fully Coupled CESM1 simulations. The control case was a 394-year preindustrial simulation (referred to as PI). Two perturbed simulations, sulfur dioxide (SO2) (a precursor of SO4) and BC emissions, were increased instantaneously from preindustrial to present-day (PD) levels, but the GHG concentrations were unchanged (referred to as PDSO4 and PDBC). Starting from the end of the 319th year, the perturbed simulations were run for 75 years, with the last 60 years being analyzed. To increase the signal-to-noise ratio caused by BC forcing (a smaller forcing), we performed an ensemble of five perturbed simulations by altering the atmospheric initial conditions by an air temperature difference at roundoff level (order of 10-14 ∘C). The long averaging time (394 years for the PI case, 60 years in the SO4-perturbed simulations, and 60 × 5 years in the BC-perturbed simulations) can restrain the impact of decadal natural climate variability and obtain a clear effect due to aerosol forcings.

Figure 1 shows the changes in aerosol optical depths (AODs) at 550 nm from PI to PD induced by SO4 and BC. The AOD increases significantly over most of the globe except for some oceans due to the increase in anthropogenic aerosol loading. The change in AOD induced by SO4 is larger than that induced by BC. The prominent increase in AOD caused by SO4 appears over eastern China, the USA, India, and western Europe, while the AOD decreases over the tropical and subtropical oceans of the Southern Hemisphere (SH). The change in BC leads to a large increase in AOD over eastern China and South Asia but a slight reduction over the tropical Pacific and Indian Ocean of the Northern Hemisphere (NH), northern Atlantic, and high latitudes of the SH.

The fifth assessment report of the Intergovernmental Panel on Climate Change provided a new definition of radiative forcing named ERF, which is a better indicator of the climate responses (Myhre et al., 2013). The global distributions of simulated SO4 and BC ERFs at the TOA are shown in Fig. 2. The ERFs are calculated using the atmosphere-only model simulations with fixed SST by subtracting the net radiative flux at the TOA. There are fundamental differences between both aerosol ERFs. Reflecting SO4 gives rise to large negative ERFs, especially in East and Southeast Asia, Central Africa, western Europe, and the subtropical oceans. However, absorbing BC leads to marked positive ERFs over East and South Asia and Central Africa, where the BC emission is large. The simulated global annual mean SO4 and BC ERFs are -0.98 and +0.36 W m-2, respectively. The simulated SO4 forcing is close to those estimated by Zelinka et al. (2014) and Forster et al. (2016), while our results show a larger BC forcing. This is attributed to the correction of BC emission in our simulations (Xu et al., 2013). The difference between reflecting and absorbing aerosol forcings implies the substantially different climate responses.

Aerosol-induced SST change is an important part of the climatic effect of aerosols (Xu and Xie, 2015). Figure 3 shows the changes in SST caused by various aerosol species from the fully coupled simulations. Despite the essential difference between both types of forcings, the spatial pattern of SST change caused by SO4 is found to be similar to that caused by BC (opposite in sign). It is characterized by a large SST change over the mid- and high-latitude oceans of the NH but only a slight SST change in the SH. The SO4 forcing leads to a significant decrease in SST over the northern Pacific, northwestern Atlantic, and NH high-latitude oceans, with the largest cooling exceeding 1.5 K. However, the opposite occurs over those regions in response to BC, with the largest warming reaching 1 K. A unique characteristic of the SST response to BC is the obvious warming over the Indian Ocean–western Pacific warm pool. Similar patterns in SST changes were found by Chung and Seinfeld (2005), Friedman et al. (2013), and Ocko et al. (2014). However, Ocko et al. (2014) showed a weaker SST change in the NH high latitudes induced by SO4 or BC and a warming in the SH high latitudes caused by SO4, which was not seen in other studies. The simulated global annual mean SST changes caused by SO4 and BC are -0.44 K (NH: -0.7 K, SH: -0.24 K) and +0.12 K (NH: +0.17 K, SH: +0.07 K), respectively (Table 2). Such an interhemispheric asymmetric adjustment in SST has been used as a crucial index of climate change (Ocko et al., 2014).

The sign of the change in surface temperature is consistent with that of the forcing. Negative SO4 forcing leads to a marked surface cooling in summer over the East Asian monsoon region (EAMR), which increases with latitude (Fig. 4a). In particular, the cooling exceeds 1 K over most of the NH subtropical oceans. The anomalous northerly winds prevail over eastern China and the surrounding oceans between 20 and 40∘ N due to SO4 forcing (Fig. 4d), which signifies the weakening of the EASM circulation. As seen in Fig. 4, the slow responses of surface air temperature and winds at 850 hPa to SO4-induced SST change closely resemble their total responses to SO4.

The fast response of surface air temperature to SO4 forcing primarily features a cooling over continental East Asia, with the values being less than -0.3 K over most of continental East Asia (Fig. 4b), because the SST is fixed in these simulations and changes in SO2 emissions are concentrated over land. Such a change in surface temperature decreases the land–sea surface thermal contrast over the EAMR, thus weakening the EASM circulation (Fig. 4e). This is consistent with previous studies using other general circulation models with fixed SST (e.g., Jiang et al., 2013; Dong et al., 2016). However, note that the weakening of the EASM in the fast response to SO4 is too weak to explain the total response of the EASM to SO4, especially over eastern China (Fig. 4d and e). Therefore, we next elaborate the physical mechanism behind the slow response of the EASM to SO4.

Figure 5 shows the JJA mean responses of zonally averaged atmospheric temperature between 100 and 140∘ E to SO4 forcing over the EAMR. The SO4-induced slow climate response leads to a significant cooling in the whole troposphere (Fig. 5c), though SO4 does not largely affect the radiation in the atmosphere. It is responsible for a large fraction of the atmospheric cooling in the total response to SO4 (Fig. 5a and c). This is because the prominent decrease in the JJA mean SST caused by SO4 also occurs in the NH midlatitude oceans, with the values being less than -1 K over most of the northern Pacific and northwestern Atlantic (Fig. S1a in the Supplement). The interhemispheric asymmetric change in SST may distinctly affect the free troposphere by alerting the tropical circulations and midlatitude eddies (Ming et al., 2011; Hsieh et al., 2013; Ocko et al., 2014; Xu and Xie, 2015). The most remarkable feature of change in atmospheric temperature in the slow response to SO4 is a deep tropospheric cooling between 30 and 45∘ N (Fig. 5c). A similar temperature response to aerosol forcings was found by Rotstayn et al. (2014) based on the multi-model ensemble simulations, which indicates that this is a robust feature of climate response to aerosols. There is an anomalous cooling center in the upper troposphere (200–500 hPa), with the cooling exceeding 1 K (Fig. 5c), which leads to a prominent decrease in geopotential height at those altitudes (Fig. 5f). The geopotential height at about 200 hPa is reduced by more than 35 m.

The East Asian subtropical jet (EASJ) that is located around 40∘ N at 200 hPa is an important component of the East Asian monsoon. The change in geopotential height in the slow response to SO4 increases the poleward (equatorward) pressure gradient force to the south (north) of the cooling center region. Such a change in pressure gradient force leads to an increase (decrease) in westerlies to the south (north) of the EASJ center through the geostrophic balance between the Coriolis force and pressure gradient force (Yu et al., 2004). It is shown in Fig. 6a that the largest increase and decrease of more than ±1 m s-1 in westerlies occur at about 25 and 45∘ N, respectively. Consequently, the EASJ shifts southwards in response to SO4. The slow response dominates over the total response of the EASJ to SO4 (Fig. 6).

The north and south flanks of the jet axis correspond generally to the divergence and convergence areas in the lower atmosphere, respectively. The southward displacement of the EASJ center implies the southward spread of divergence areas, thereby resulting in an anomalous surface anticyclone over continental East Asia between 30 and 40∘ N. The anomalous subsidence motion in the lower atmosphere around 40∘ N (Fig. S2c) due to the large surface cooling also intensifies the anomalous surface anticyclone. To the east of the anticyclonic center, anomalous northerlies increase prominently (Fig. 4f). In addition, the interhemispheric SST gradient caused by SO4 (Fig. S1a) strengthens the ascending branch of the local Hadley cell between 20 and 35∘ N in the summer (Figs. S2c and S3c), thereby resulting in an anomalous cyclonic vortex over southeastern China and the western Pacific (Fig. 4f). To the west of the cyclonic center, anomalous northerly winds are further increased. Finally, the SO4-induced slow climate response leads to a more intense weakening of the EASM circulation than its fast response. Dai et al. (2013) also suggested that the thermal contrast in the mid-upper troposphere played a more important role than that in the mid-lower troposphere in impacting the strength and variations of the Asian summer monsoon circulations. Drop in tropopause height over the EAMR can suppress the convection and weaken the EASM. As seen in Fig. 7a, the tropopause north of 40∘ N in the summer declines significantly in the total response to SO4, which primarily contributed by its slow response. The sharp drop also coincides with the southward displacement of the NH subtropical jet, as the jet approximately divides the tropics (with higher tropopause) and extratropics (Ming et al., 2011). The above analyses indicate the importance of ocean response to SO4 forcing in driving the changes of the EASM circulation.

Note that the changes in atmospheric temperature and geopotential height due to the adjustments in clouds and atmospheric states in the fast response to SO4 (Fig. 5b and e) lead to the increase of more than 0.8 m s-1 in westerlies at the north of the jet center (Fig. 6b). The positive change of westerlies in the fast response is comparable to the negative change of westerlies in the slow response to SO4 due to the comparable changes in temperature and geopotential height. The change of the jet in the fast response to SO4 (Fig. 6b) is conducive to the enhancement of the EASM, which partially offsets the weakening of the EASM due to the decrease of the land–sea surface thermal contrast.

The weakening of the EASM circulation caused by SO4 forcing suppresses the transport of surface warm and moist air northwards and upwards, which results in a significant decrease in precipitation over eastern and southern China and the ambient oceans (Fig. 8a). In particular, the precipitation is decreased by more than 0.6 mm day-1 over most of southern China. The cooling in the lower troposphere and warming in the upper troposphere due to the fast response north of 20∘ N (Fig. 5b) can suppress the vertical ascending motion (Fig. S2b) and moisture transfer, thereby also contributing to the decrease in precipitation. However, the precipitation increases (yet not significantly) over some of the western Pacific due to the enhanced convection in the slow response to SO4 (Fig. S2c), with the maximum exceeding 1 mm day-1. Note that the SO4-induced slow climate response leads to a large increase in precipitation over western China, which is the opposite compared to its fast response (Fig. 8b and c). This is because the enhanced easterly anomalies in the lower troposphere between 25 and 35∘ N in the slow response (Fig. 4f and 6c) bring about more moisture into the inland regions of China, which is beneficial to the formation of clouds and precipitation. In a word, the decrease in precipitation over land in East Asia in the total response to SO4 is dominated by the fast response, while the change in precipitation over the adjacent ocean is dominated by the slow response.

Figure 9 shows JJA mean responses of surface air temperature and wind vectors at 850 hPa to BC forcing. Absorbing BC increases the surface air temperature over the EAMR, with the largest warming appearing at the NH midlatitudes, especially over the northwestern Pacific, with the maximum exceeding 0.8 K (Fig. 9a). This is mainly from the contribution of the slow climate response to BC (Fig. 9c). The small anomalous southerly winds at 850 hPa prevail and the EASM circulation is slightly enhanced, especially north of 30∘ N, in the total climate response to BC (Fig. 9d), which is mainly attributed to their fast responses to BC. The fast and slow responses of surface winds to BC over the EAMR are inverses of each other. The anomalous northerlies in the slow response that tend to weaken the EASM greatly offset the anomalous southerlies in the fast response to BC (Fig. 9e and f).

Now we explain why the enhancement of the monsoon in the fast response to BC forcing is strong. Firstly, the large surface warming in the fast response to BC occurs over continental East Asia, especially north of 30∘ N, with the warming exceeding 0.2 K (Fig. 9b). This increases the land–sea surface thermal contrast over the EAMR, thereby enhancing the EASM circulation (Fig. 9e). This mechanism is also at work in the fast response to SO4. Secondly and unique to the BC case, the direct absorption of solar radiation by BC leads to a deep tropospheric warming of 0.2 to 1 K at the NH midlatitudes (Fig. 10), which dominates over the tropospheric warming in the total response to BC. An anomalous warming center of more than 0.6 K appears and the geopotential height is increased by more than 10 m in the upper troposphere around 40∘ N (Fig. 10e). Consequently, the pressure increases in the uppermost troposphere, which strengthens the poleward (equatorward) pressure gradient force in the north (south) flank of the warming region. This results in an increase of 0.2 to 1 m s-1 (a decrease of 0.8 to 1.2 m s-1) in westerly winds in the north (south) flank of the EASJ center and the northward movement of the EASJ (Fig. 11b). The total response of the jet is consistent with the fast response of it to BC. With the change of the EASJ, an anomalous cyclonic vortex is formed over land in East Asia and anomalous southerly winds increase over eastern China. This second mechanism involving the EASJ change further magnifies the enhancement of the EASM caused by the increase in land–sea surface thermal contrast in the fast response to BC. The elevation of the tropopause between 40 and 50∘ N in the summer due to the fast response also implies the strengthening of the EASM (Fig. 7b). The fast response dominates over the total response of tropopause height to BC.

The BC-induced slow response is in the opposite direction of the fast response. Like the annual mean SST response, the significant increase in the JJA mean SST caused by BC occurs not only in the NH midlatitude oceans but also in the Indian Ocean–western Pacific warm pool, with the warming exceeding 0.2 K over most areas (Fig. S1b). This results in the deep tropospheric warming north of 40∘ N and a larger warming in the upper troposphere between 20 and 30∘ N, respectively (Fig. 10c). Keshavamurty (1982) found that the warming over tropical western Pacific could significantly enhance the convection motion in the western Pacific and that it was more efficient in producing atmospheric circulation anomalies. Therefore, the BC-induced slow climate response also strengthens the ascending branch of the local Hadley cell between 15 and 30∘ N in the summer (Figs. S4c and S5c). This leads to an anomalous cyclone in the lower atmosphere over these regions, thus increasing the anomalous northerly winds over eastern China (Fig. 9f). While the tropospheric temperature increases in the slow response to BC, the warming in the upper troposphere of around 40∘ N is less than that on both of its flanks (Fig. 10c). Such an adjustment in tropospheric temperature is conducive to a southward shifting of the EASJ (Fig. 11c). These EASJ changes cause the BC-induced slow response to weaken the EASM circulation, which even overcomes the strengthening of the EASM due to the increase in land–sea surface thermal contrast in the slow response (Fig. 9c). The opposite fast and slow responses of tropopause height to BC also indicate their adverse impact on the EASM circulation (Fig. 7b).

Lastly, the JJA mean response of precipitation to BC forcing over the EAMR is weaker than that found in the response to SO4 with less area in which the change is significant (Fig. 12), mainly because of a smaller radiative forcing. The total response of precipitation to BC manifests a spatial pattern of wetting–drying–wetting from north to south over the EAMR, with an increase of 0.1 to 0.6 mm day-1 over most of southeastern China and north of 40∘ N and a decrease of 0.1 to 0.5 mm day-1 over the Yangtze–Huai River valley (Fig. 12a). This is not consistent with that reported by Menon et al. (2002), which indicated that BC forcing primarily contributed to the wetter-south–dryer-north phenomenon in eastern China during the past decades. The change in precipitation caused by BC forcing is mainly in line with the change in monsoon circulation. The fast and slow responses of precipitation to BC are almost opposite over the EAMR (Fig. 12b and c) due to the opposite circulation changes. The deep tropospheric warming north of 40∘ N due to the fast response (Fig. 10b) can intensify the vertical ascending motion (Fig. S4b) and moisture transfer, which dominates over the increase in precipitation here. However, the warming in the troposphere and anomalous surface easterly winds between 20 and 30∘ N due to the slow response (Figs. 10c and 11c) are conducive to the development of ascending motion (Fig. S4c) and moisture transport from the oceans, which contributes to the increase in precipitation over these regions. In addition, the southward shifting of the EASJ in the slow response leads to an increase in surface divergence (Fig. 9f) and a decrease in precipitation over the Yangtze–Huai River valley. In general, the spatial distribution of the total precipitation response agrees well with that of the slow precipitation response to BC except north of 40∘ N, which also shows the significance of the SST change induced by BC forcing in impacting the EASM.