Impact of aerosols on global climate
This section focuses on the impact of aerosols on the global and regional climates, beginning with the global climate. The performance of the HIST simulation was investigated previously in Watanabe et al. (2011). The following subsection focuses on the reproducibility of the Asian monsoon precipitation, which is important for our purposes. Figure 1 presents the time series of global SAT, precipitation, aerosol optical depth (AOD), precipitable water, and cloud effective radius for the 100-year period from 1906 to 2005. To investigate long-term changes, seasonal components and those with periodicity of less than 10 years were suppressed by using a running mean with a window of 121 months.
It is well known that the global mean SAT has increased due to increased concentrations of GHGs, such as carbon dioxide. All three members of the HIST and piAERO experiments showed similar long-term increases (Fig. 1b), respectively. The simultaneous drops in SAT in the 1960s, around 1980, and around 1991 were considered to be caused by large volcanic eruptions, i.e., the effects of natural aerosols (Fig. 1a), although the drops were unclear due to the effect of the digital filter. These drops in SAT were also simulated in piAERO, as long-term changes in the natural aerosol burden were similar to those in HIST, although the simulated long-term trends in SAT in the piAERO were different from those in the HIST. Compared to the HIST, the increase in global mean SAT was larger in piAERO after the 1950s. The timing (from 1955 to 2005) when the differences in global mean SAT between the HIST and piAERO has gradually become larger corresponded to that in AOD (Fig. 1d). The increase in AOD was due to human activity.
In our experiments, during 1905–2005, the global mean SAT increased by approximately 0.6–0.8 K in HIST and by approximately 1.0–1.2 K in piAERO. These simulations indicate that approximately one third to one half of the total increase in global mean SAT was likely suppressed by the increase in anthropogenic aerosol burden. This estimated sensitivity to global warming was greater than that seen in the multi-model results of the IPCC-AR5 (IPCC 2013).
We also examined changes in the atmospheric hydrological cycle. Concurrent with the long-term increase in SAT, over the 100-year period, the global mean precipitation and precipitable water increased in the piAERO experiments (Fig. 1c, e). In general, an increase in global mean SAT results in an increase in precipitation, due to an increase in the global mean precipitable water. However, the global mean precipitation in the HIST decreased after 1950, although the global mean SAT gradually increased. The HIST and piAERO simulations of long-term changes in the global mean cloud effective radius showed clear differences between the HIST and piAERO (Fig. 1f). Smaller cloud effective radii occurred under higher aerosol loading conditions. These results can be explained by the cloud lifetime effects of aerosols (e.g., Twomey et al. 1984; Albrecht 1989).
Reproducibility of the Asian monsoon precipitation and circulation
Before discussing the regional characteristics and seasonality of the aerosol impacts, we examined the reproducibility of precipitation over the Asian monsoon region. Figure 2 presents the climatology based on observations from GPCP and JRA55 and simulations from MIROC-ESM.
The simulation revealed spatial peaks in precipitation during boreal summer (June, July, August; JJA) over the Asian monsoon region (Fig. 2b), similar to the observations (Fig. 2a). However, a major peak in precipitation was simulated over the western coast of the Indian subcontinent while this was observed over the western coast of the Indochina Peninsula. In addition, compared to the observations, weaker monsoon westerlies and a weaker monsoon trough around the Philippines were found in the HIST experiments. Precipitation over the southern Maritime Continent was also overestimated in HIST.
During boreal winter, although the simulation reproduced the spatial distribution of precipitation and circulation, the precipitation over the equatorial Indian Ocean was greater in HIST than that in the observations. As in simulations of boreal summer, precipitation over the southern Maritime Continent was overestimated, and precipitation over the southern Pacific Ocean was underestimated in HIST.
Although there were some differences between simulated and observed precipitation, the general features and spatial patterns of precipitation and low-level circulation and their seasonal changes were reproduced by HIST.
Impact of aerosols during boreal summer (JJA)
Geographical pattern of AOD changes
To understand how aerosols affect the regional climate of the tropical Asian monsoon, we produced a global map of differences between the HIST and piAERO in AOD, SAT, clear-sky downward DSRS, the difference between all-sky and clear-sky DSRSs, cloud cover, and precipitation (Fig. 3). Because changes outside of the tropical Asian monsoon region may affect the regional climate of the tropical Asian monsoon region, we analyzed the changes globally, focusing on the tropical Asian monsoon region. The difference between all-sky and clear-sky DSRSs will be explained later. To detect statistically significant signals, we used all three ensembles in both experiments.
An increase in AOD during JJA was observed over and around the regions of East Asia, South Asia, Maritime Continent, Western Europe, Western Africa, and South America (Fig. 3a). The signal of AOD was weaker over North America because the timing of improvement air quality problems was different from the other regions. During the period considered, considerable changes have occurred in land use, such as deforestation, conversion from natural vegetation to cropland, and urbanization.
The major source regions of aerosols are located in the Northern Hemisphere. Over this study region of the Asian monsoon, China and India are the major source regions. The aerosols were partly transported downstream of the monsoon flows (see Fig. 2). Notably, over the South Pacific Ocean, there is no major source of aerosols, which induced the meridional contrast of AOD globally, particularly over the Pacific Ocean. The increase in AOD ranged from 0.05 to 0.20.
Surface air temperature response
Over the tropics, statistically significant cooling signals were simulated over the entire Indian Ocean, North Pacific Ocean, and the North Atlantic Ocean (Fig. 3b). Note that the tropical South Pacific Ocean was not cooled by the increase in aerosols. Thus, the east-west inter-oceanic contrast in SAT between the Indian Ocean and Western Pacific Ocean was dominant (Fig. 3b), which may have affected the atmospheric hydrological cycle of the tropical Asian monsoon. In addition, no clear SAT signals were simulated over the Asian monsoon land regions, i.e., the Indian subcontinent and the Indochina Peninsula, indicating that the cooling over land was not systematically simulated. Cooling signals were also simulated over the middle and high latitudes in the Northern Hemisphere (Fig. 3b), whereas less cooling was simulated over the Southern Hemisphere. Notable cooling was simulated over a zonal band from 30° N to 70° N and not only near the source regions of aerosols.
Due to the increases in AOD, negative signals in clear-sky DSRS were observed over the major source regions (Fig. 3c). The spatial distributions of the differences in AOD and clear-sky DSRS were similar, indicating that direct radiation effects resulted in cooling over the major sources of aerosols. However, the changes in clear-sky DSRS due to increased AOD (Fig. 3c) cannot explain all of the changes in SAT (Fig. 3b).
Changes in cloud cover and precipitation
To understand the changes in SAT, we also examined the indirect effects of aerosols by considering changes in cloud cover, which are the parts of the atmospheric hydrological cycle. Of course, we also considered the responses of precipitation and precipitable water. Here, we defined the indirect effects of the aerosols as the difference between the all-sky and clear-sky DSRS (following Mukai et al. 2008; Mukai and Nakajima 2009). For this estimation, the indirect effects were considered to be the changes in cloud cover due to different aerosol conditions. However, this is not simply a sum of the first and second indirect effects; it involves many effects related to changes in the aerosol conditions, so we cannot estimate only the indirect effects of aerosols.
An impact of cloud cover changes on the decrease in SAT due to aerosol-cloud interactions was clear over the tropical regions, such as equatorial Indian Ocean, equatorial Western Pacific, and the Indian subcontinent (Fig. 3d), although significant cooling over the mid-latitudes was also simulated over downstream regions of the aerosol sources. We confirmed that changes due to the indirect effects could be explained by changes in cloud cover because their spatial patterns were similar (Figs. 3d, e). It is noteworthy that the effects of clouds due to the changes in aerosol condition contributed significantly to the creation of inter-oceanic and inter-hemispheric contrasts in SAT over the tropical Asian monsoon region and the Pacific Ocean.
Changes in precipitation were remarkable primarily over the tropical regions (Fig. 3f). An increase in precipitation was simulated over part of the Maritime Continent-southern equatorial Pacific and Amazon regions, whereas a decrease was observed over the Asian monsoon land, the southern equatorial Indian Ocean, and Central Africa. The changes in precipitation can be explained by changes in water vapor (Fig. 3g). The changes in water vapor mostly corresponded to those in SAT. The decreases in water vapor were simulated over the entire Asian monsoon region and the Indian Ocean. Only the equatorial South Pacific Ocean showed a positive signal in water vapor.
Changes in atmospheric circulation and coupling with clouds and precipitation
To understand the changes in meridional and zonal circulation on the Asian monsoon regional scale, we produced meridional-vertical cross-sections of winds over the Asian monsoon region and the Pacific Ocean, and a zonal-vertical cross-section of winds along the equatorial region (Fig. 4). To associate atmospheric circulation changes with the precipitation changes, we analyzed zonal and meridional averaged changes in precipitation (Fig. 5).
Over the Asian monsoon region, anomalous descending motions were simulated south of the equator and along the zonal band of 25–35° N (Fig. 4a), consistent with the decreased cloud cover and precipitation (Figs. 3e, f and 5a). These results revealed no changes in local Hadley circulation over the Indian Subcontinent and the Indochina Peninsula (along the zonal band of 10–25° N), which is different from the finding of Bollasina et al. (2011). A weak enhancement of local Hadley circulation over the Asian monsoon region was simulated within 10° S–10° N, which did not cover the Asian landmass (Fig. 4a). Over the Pacific Ocean, only the anomalous ascending motions were simulated over the equatorial regions (Fig. 4b), which is consistent with the changes in precipitation (Fig. 5b). The anomalies in the equatorial zonal circulation showed an eastward shift of the convective center of the Walker circulation or enhancement of the Walker circulation over the Indian-Pacific trans-ocean sector (Fig. 4c), which is consistent with the changes in precipitation (Fig. 5c). This suggested that the changes in equatorial zonal circulation can more coherently explain the large-scale anomalies in cloud cover and precipitation than can changes in local Hadley circulation. The distinct changes in the Walker circulation can be confirmed by the spatial map of surface winds (Fig. 6). These anomalous descending (ascending) motions were consistent with the decreases (increases) in cloud cover and precipitation, making it difficult to determine the cause-effect relationship between the cloud-precipitation and circulation changes.
On the more detailed spatial pattern of the precipitation response, the Asian summer monsoon precipitation decreased along two zonal bands, 10° S to the equator and 20°–30° N (Figs. 3f and 5a). Concurrent with the decreases in precipitation over the Asian monsoon region, an anomalous divergence in surface winds over the equatorial Indian Ocean and an anomalous anti-cyclonic circulation over South Asia and the Tibetan Plateau were simulated (Fig. 6). In addition, an increase in precipitation over the equatorial Pacific was consistent with an anomalous convergence in surface winds. Strong westerly anomalies over the eastern Indian Ocean and Maritime Continent, which modulated the Walker circulation in this region, were simulated. The precipitation signals also corresponded to the anomalous vertical motions (Fig. 4a). This spatial precipitation pattern matched the observed pattern of interannual precipitation variability, i.e., a zonally extended striped pattern (monsoon trough), as indicated in previous observational studies (e.g., Goswami and Mohan 2001; Fujinami et al. 2011; Takahashi et al. 2015). The spatial pattern in precipitation is associated with the tropical cyclone activity (Takahashi et al. 2015; Takahashi and Yasunari 2006), which suggests that the signal is associated with a modulation of tropical cyclone activity. The decrease in precipitable water (Fig. 3g) can weaken low-level water vapor convergence, even if the strength of wind convergence by a tropical disturbance does not change, and may reduce the development of tropical disturbances along the monsoon trough.
Impact of aerosols during boreal winter (DJF)
Geographical pattern of changes in AOD
An increase in AOD during December, January, and February (DJF) was simulated over and around East Asia, Southeast Asia, Western Europe, Western Africa, and North America (Fig. 7a). Over the tropical Asian monsoon region, there were large differences in AOD between JJA and DJF over the eastern region of Eurasia. Although the major sources of aerosols were similar to those during boreal summer, AOD was greater over Southeast Asia and was lower over the mid-latitude North Pacific, compared with those during boreal summer. The differences in spatial distribution of AOD can be explained by the increase in local biomass burning (e.g., van der Werf et al. 2008) over Southeast Asia and by equatorward transportation of the East Asian aerosols due to the climatological winter monsoon winds in the lower troposphere, which contribute to the stronger zonal contrast in AOD between the Indian Ocean-Maritime Continent sector and the equatorial Pacific Ocean in boreal winter.
Surface air temperature response
Over the tropics, cooling was simulated over the tropical Indian Ocean, North Pacific Ocean, and tropical Atlantic Ocean. No clear cooling was simulated over the South Pacific Ocean. This SAT response results in an enhancement of the zonal gradient in SAT due to the increased aerosols (Fig. 7b), consistent with the spatial pattern of AOD (Fig. 7a). These features were similar to those in boreal summer. In addition, cooling in the mid-latitudes was weaker than that in boreal summer, which was consistent with a seasonal change in AOD.
Next, we investigated the aerosol-radiation interaction. Figure 7c shows the differences in clear-sky DSRS during DJF. There was noticeable cooling along the southeastern coast of Eurasia, which corresponded to AOD changes (Fig. 7a). Again, although the geographical pattern of the changes in SAT was not entirely explained by the direct effect of aerosols due to the increased AOD, the sum of direct and indirect effects (Fig. 7c, d) can explain the changes in SAT. It was found that the spatial pattern of cloud cover was similar to that of indirect effects of aerosols. Moreover, the spatial distribution of the indirect effects (Fig. 7d) was different from that seen during JJA (Fig. 7d). The simulated cooling in the mid-latitudes due to the indirect effects of aerosols over the North Pacific Ocean at around 30° N was weaker than that in boreal summer.
Changes in cloud cover and precipitation
During DJF, east-west contrasts in cloud cover and precipitation were simulated over the tropics (Fig. 7e, f), corresponding to the stronger zonal gradient in SAT and precipitable water over the Indian Ocean and western Pacific Ocean (Fig. 7b, g). Decreases in precipitable water were simulated over the entire tropical region except for the tropical South Pacific Ocean, which can almost entirely explain the decrease in precipitation over the tropical Asian monsoon region. Although the spatial patterns of cloud and precipitation during DJF were somewhat different from those during JJA, the dominant east-west contrasts in cloud cover and precipitation were similar between boreal summer and winter.
Moreover, there were different responses of cloud-precipitation between the tropics and mid-latitudes. In the mid-latitudes, the cyclonic anomalies and enhancement of the winter monsoon outflows over and around Japan were associated with an increase in cloud cover and a decrease in precipitation (Figs. 7 and 8). This response can be explained as the well-known lifetime effect of aerosols. Interestingly, the signs of changes in precipitation corresponded to those of changes in cloud cover over the tropics (Figs. 3 and 7), which were different from the responses in the mid-latitudes. The different responses are discussed in the “Uncertainties of effects of aerosols on climate” section.
Changes in atmospheric circulation and coupling with clouds and precipitation
This subsection discusses changes in cloud cover and precipitation during boreal winter, which are associated with atmospheric circulation. Takahashi and Watanabe (2016) reported that the decreased volcanic sulfate aerosols caused partial enhancement of trade winds over the equatorial western Pacific Ocean from 1991 to 2010, which can be explained by the zonal gradient in SAT. Their study showed that an aerosol forcing can change the large-scale circulation over the tropics, although the forcing of aerosols in our experiment is quite different from that in Takahashi and Watanabe (2016). The differences can be clearly seen from the spatial pattern of SAT. Furthermore, a few observational studies showed co-variability of the aerosol-cloud-precipitation over and around the Maritime Continent on interannual timescales (e.g., Yamaji and Takahashi 2014).
As with the boreal summer, we investigated changes in meridional and zonal circulation (Fig. 9) and changes in precipitation for winter (Fig. 10). The changes in the local Hadley circulation over the Asian monsoon region were not clear (Fig. 9a). Over the Asian monsoon region, anomalous descending motions (Fig. 9a) were consistent with the decreased precipitation (Fig. 10a). Over the Pacific Ocean, anomalous ascending motions were consistent with the increase in precipitation (Figs. 9b and 10b), which can be understood as the southward shift of the inter-tropical convergence zone (ITCZ). In addition, this southward shift of the ITCZ (Figs. 7 and 10) was partly consistent with the annual mean response simulated by Takemura et al. (2005). However, we found that the responses over the Asian monsoon region and the Indian Ocean were quite different.
Changes in the Walker circulation were dominant (Fig. 9c). The anomalous convective center of the Walker circulation was simulated over the western equatorial Pacific Ocean, and this distinct signal was statistically significant (Fig. 9c and 10c). Thus, changes in part of the Walker circulation, rather than in local Hadley circulation, were strongly coupled with the changes in precipitation over the tropical Asian monsoon region. Note that our results regarding changes of the Walker circulation coupling with cloud and precipitation in association with the changes in zonal gradient of SAT can be explained in a similar manner to Takahashi and Watanabe (2016). Nevertheless, the changed spatial patterns of precipitation, circulation, and SAT are different because of differences in experimental design, such as the treatment of forcing.
The simulated decreases in cloud cover and precipitation coupled with increased aerosols over the Indian Ocean and Asian monsoon regions were symmetric with respect to the equator (Figs. 7e, f and 10a). The spatial pattern of this symmetrical response along 10° S and 10° N over the Indian Ocean corresponded to that of the activity of cyclone pairs that are symmetrical with respect to the equator (e.g., Kiladis and Wheeler 1995; Compo et al. 1999; Ito and Satomura 2009, Takahashi et al. 2011). The activity of cyclone pairs is dominant in this season. This correspondence suggests that modulation of the activity of the tropical disturbances may explain the decrease in cloud cover and precipitation. The decrease in precipitable water (Fig. 7g) due to cooling of the SAT (Fig. 7b) and anomalous descending motions (Fig. 9a, c) can weaken the low-level convergence of water vapor and/or the activity of the tropical disturbances over the equatorial Indian Ocean. Notably, the signals were clear along the major route of the tropical disturbances. These characteristics were common to both boreal summer and winter, although the frequent passage routes of tropical disturbances are different between boreal summer and winter.