Dust is one of the most important components of atmospheric aerosols. The main source of atmospheric dust is wind erosion in arid and semi-arid regions; it is estimated that the global atmospheric dust emission may be as high as 200–5000 Mt yr-1 (Goudie, 1983). Because of the large amount in the atmosphere, dust effects on the environment and the climate system have attracted much attention. The inhalation of dust aerosols can harm both human and animal health; it can also affect visibility and thus potentially increase the number of traffic accidents (Park and Kim, 2005). Dust aerosols also are important drivers of the global climate because of their direct radiative effects on the Earth–atmosphere radiation balance and temperature (Tegen and Lacis, 1996; Miller et al., 2004). They can alter the atmospheric hydrological cycle by acting as cloud condensation nuclei and thus can modulate both the regional and global precipitation (Rosenfeld et al., 2001). Satellite observations have shown that dust originating from the Taklimakan Desert can travel around the globe within 2 weeks and alter the interaction between the atmospheric CO2 and the global climate by providing nutrients to and interacting with the marine ecosystem (Uno et al., 2009).

East Asia is an important source region for dust (Zhang et al., 1996) and is home to more than half of the world's population. The lives of people in East Asia are deeply affected by the East Asian summer monsoon (EASM) and the relationship between dust aerosols and the EASM is of great interest to the scientific community. Simulations have shown that dust aerosols not only weaken the EASM (Sun et al., 2012; Guo and Yin, 2015), but can also reduce the atmospheric heat source over the Tibetan Plateau (TP) and delay the onset of the EASM (Sun and Liu, 2016). Aerosols, including dust aerosols, have been shown to affect the intensity of the EASM (Li et al., 2016) and variations in the EASM can modulate the spatiotemporal distribution of dust aerosols in East Asia. A recent modeling study by Lou et al. (2016) indicated that there was a negative correlation between the spring dust loading in eastern China and the East Asian monsoon.

Dust aerosols in East Asia are mainly derived from arid and semi-arid areas, including the Taklimakan and Gobi deserts. However, some studies have indicated that the TP itself may also be an important source region for dust (Zhang et al., 1996; Fang et al., 1999) and that the region is more conducive to the atmospheric transportation of dust due to its high altitude and can interact directly with the TP thermal pump (Wu et al., 2012). However, the source and spatiotemporal distribution of dust aerosols over the TP have not been established yet. At present, there are three viewpoints about the source of dust aerosols over the TP. First, an investigation by Fang et al. (1995, 1999) showed that there exists 2047.41×104 ha of desert within the TP, suggesting that the TP may be a potential source for dust. A numerical simulation by Chen et al. (2013) showed that dust aerosols were produced by local emissions over the TP in spring and winter. Second, satellite observations have shown that the aerosols over the TP are dominated by dust in spring and summer and that the dust aerosols were probably derived from the Taklimakan and Gurbantunggut deserts to the north of the TP (Huang et al., 2007; Jia et al., 2015). Third, some studies have indicated that the dust emitted from the south of the TP, such as from the Great Indian Desert, can also be transported over the Himalaya (Lau et al., 2006, 2010).

As a massive, elevated heat source, the TP can directly heat the upper troposphere. The heating anomaly over the TP has a great impact on the EASM (Yanai et al, 2006; Duan et al., 2012). Studies have shown that dust aerosols over the TP can alter the local atmospheric radiation balance, affecting both the heat source over the TP and the Asian monsoon (Lau et al., 2006, 2010; Chen et al., 2013; Sun et al., 2016). However, most previous simulation studies have focused on dust aerosols originating from the Taklimakan and Gobi deserts (Zhao et al., 2006; Wang et al., 2008; Huang et al., 2009; Sun et al., 2012), and there have been few investigations of the impact of dust aerosols emitted by the TP on the East Asian climate. The work reported here used the RegCM4.1 model to simulate climatic effects of distribution of dust aerosols surrounding the TP by performing numerical experiments with and without the emission of dust over the TP.

In this section we will first evaluate the CON simulation using the observed data described in the previous section. Then the results from CON and SEN experiments will be compared to determine the roles of dust aerosols generated from the TP play in the thermodynamic fields and circulations including the EASM.

Previous research has shown that dust emitted from Asian deserts can weaken the EASM (Sun et al., 2012; Guo et al., 2015; Li et al., 2016), although the details of weakening mechanisms are still unclear. It has been suggested by some authors that the weakening of the EASM is a result of the reduction in the thermal contrast between the land and the sea induced by dust aerosols (Guo and Yin, 2015; Li et al., 2016). However, the modeling result of Sun et al. (2012) showed that the EASM is reduced by the large-scale atmospheric circulation disturbances (cyclone–anticyclone–cyclone Rossby wave train) generated by the radiative cooling of dust aerosols. In the work reported here, we considered the effects of dust aerosols emitted only within the TP itself on regional climate and found that they can also reduce the EASM significantly by weakening the heat source (pump) over the TP and thus reduce the land–sea thermal contrast. The locally generated TP dust can cause surface cooling far downstream in Bohai Gulf and the China–North Korea border area through stationary Rossby wave propagation. Our sensitivity simulations showed there was a negative correlation between the EASM and dust aerosols emitted from the TP locally.

The spring dust aerosols from the TP have a close relationship with EASM. Although the cause–effect relationship is not immediately clear, the following processes are proposed as a possible mechanism based on the results in our simulation (Fig. 13). Firstly, increasing (decreasing) in dust aerosol over the TP in the heavy (light) dust years in spring can weaken (enhance) the TP heat source and thus reduce (increase) precipitation over the TP. Reduction (increase) in precipitation over the TP can also further enhance (diminish) dust emission over the TP (labeled 1 in Fig. 13). Secondly, the weakened (enhanced) TP heat source can persist from spring to summer and shrink (expand) the land–sea thermal contrast and thus weaken (enhance) the EASM. Therefore, the change of dust over the TP is anti-correlated with the variation of EASM circulation intensity (labeled 2). Thirdly, weakened (enhanced) monsoon circulation can reduce (increase) precipitation in East Asia (labeled 3). As a result, the precipitation variation of the TP presents a positive correlation with that of EASM.

It is worth noting that Sun and Liu (2016) demonstrated that dust emitted from Taklimakan and Gobi deserts weakens the Asian monsoon through large-scale atmospheric circulations by 2 m s-1 of wind at 700 hPa. This magnitude of reduction in the wind seems small compared to the values in the present study even though the emission source extent in the previous study is larger. We think both high-altitude source like the TP and low-altitude sources such as Taklimakan and Gobi deserts can weaken the EASM, but the mechanism could differ. The dust emitted from low-altitude source (mainly Taklimakan and Gobi deserts) reduces the EASM mainly by the large-scale atmospheric disturbances (Rossby wave train), while the dust emitted from high-altitude source weakens the EASM by the reduction in the TP heating and in thermal contrast in the middle troposphere between the land and sea. The column dust load induced by local emissions from the TP in heavy dust years accounted for 20 %, (CON - SEN) / CON, of the total loading over the TP; its impacts on Asian monsoon and climate seems more important than the low-altitude sources such as Taklimakan and Gobi deserts in East Asia. This disproportionately large impact from TP locally emitted dust is likely due to its higher elevation within TP itself so that the dust-induced cooling can more effectively weaken the TP from acting as a heat pump for the Asian monsoon. Further studies on this are certainly warranted.

One interesting finding of this study is the net negative radiative heating rate in the lower troposphere over the TP (Fig. 9). Dust direct radiative effects on the atmosphere have been reported to be predominantly positive (warming) over land areas in most previous research (Saeed et al., 2014; Osborne et al., 2011; Zhang et al., 2013; Banks et al., 2014; Chen et al., 2013, 2017). However, the research of Wang et al. (2011) reported strong cooling of dust aerosol in East Asian deserts, and their research demonstrated that dust storms with the same intensity over the East Asia deserts and nearby regions may have different or even opposite direct radiative effects on the Earth–atmosphere system, including its thermodynamic and dynamic structures, depending on season and time (of day) of dust storm occurrence. Here we offer following possible explanations for the lower tropospheric cooling: 1.

Magnitudes and even signs of dust aerosol direct radiative forcing in solar spectrum on the atmosphere are largely determined by the aerosol single-scattering albedo (SSA) and to a lesser degree by the albedo of the underlying surface. The SSA of dust aerosols is determined by size distribution, morphology and complex refractive index (Moosmüller et al., 2009). The chemical composition of dust also effects SSA values. For example, the SSA of fine mineral dust particles is determined by iron concentration (Moosmüller et al., 2012). Depending on their sources, the SSA of the dust aerosols can be quite different. Figure 14 shows the spatial distribution of SSA as calculated in the RegCM4.1. The SSA values are > 0.9 over the TP, considerably larger than those over Taklimakan, Gobi and Great Indian deserts, where SSA is about 0.7. Past studies have shown that the radiative roles of dust aerosol plays is largely dependent on SSA. A 5 % change in its value can significantly alter the magnitudes or even sign of SW radiative forcing (Hatzianastassiou et al., 2004; Solmon et al., 2008; Das et al., 2015b; Papadimas et al., 2012). Over the TP with SSA > 0.9, only less than 10 % of extinct solar radiation is absorbed by the dust aerosol, compared to 20–30% elsewhere, meaning that the TP dust aerosol is only one-third to one-half the efficiency of those over surrounding sources.

Why is SSA of the TP dust aerosol smaller than dust elsewhere? For a given wavelength, SSA depends on dust particle size, among other factors. The dust particle size spectrum varies among different sources. Those fine dust particles (0.01–1.0 µm) over the large part of TP contribute as much as 70 % of total dust particles mass compared to 50–60 % in Taklimakan Desert and some other areas (figure not shown). 2.

Unlike SW forcing that occurs only in the daylight hours, the LW cooling persists day and night. In addition, the TP dust tends to exist during night because of the nocturnal convergence driven by diurnal cycle of thermodynamics over the TP (Liu et al., 2009). This would further minimize the already weakened SW absorption (Fig. S1). The combination of reduced SW heating and maintaining of LW cooling resulted in a net negative radiative heating rate in the lowest 200–300 hPa of troposphere shown in Fig. 9.

Although Fig. 9 shows a negative heating rate over the TP within the lowest 200–300 hPa of the atmosphere locally, the absorbed SW by atmospheric column measured by the flux difference between the top of the atmosphere (TOA) and surface, as normally done in the literature, is still positive outside TP over major source regions (Fig. S2b minus Fig. S2a). Thus strictly speaking, by conventional definitions, the absorbed SW radiation is still mostly positive over the whole domain, even though the TP dust results in a cooling effect in the lower troposphere locally (Fig. S2e, f).

Finally it is noteworthy that, given the large variability of dust SSA among different sources in Asia, it is possible that the magnitudes or even signs of dust direct net radiative forcing on the atmosphere could vary among case studies and climate simulations over different continents. For example, the LW forcing of dust at Zhangye, China, was found to be about a factor of 2 larger than that over Saharan measured at Sal Island, Cabo Verde, owing to differences in the dust absorptive properties (Hansell et al., 2012).

It is very beneficial to study the impact of aerosols on climate using a regional climate model (RCM) instead of a coarse-resolution global climate model (GCM). However, limited-area RCMs naturally cannot fully account for external forcing remote from the domain of interest although the lateral boundary conditions allow large-scale features to propagate into the domain. Our domain size (9600 × 640 km) is reasonably large enough so that the weather and climate systems can have adequate spatial extent to develop within the domain, as attested by reasonable validation of wind pattern, temperature field and precipitation (Sect. 3.1). Caution should be exercised, however, as results from regional simulations could be somewhat domain-size dependent quantitatively although main results should not be affected. It is worth mentioning that the model's internal variability could influence the results; thus, we compared the standard deviation of summer surface temperature and precipitation in CON with the signal induced by the dust effects (CON minus SEN) during the heavy dust years. The signal induced by the dust is much greater than the standard deviations (figures not shown). Therefore, the dust effect reported in our simulation is significant in the heavy dust years.

Only direct radiative effects of dust were included in our model and future studies should include both direct and indirect effects. The semi-direct effects (Hansen et al., 1997) are included in this study as part of atmospheric feedback, but are not explicitly discussed here since they would be better discussed along with the indirect effects as both involve clouds. The simulated effects of dust aerosols on climate were highly sensitive to the physical characteristics of the dust aerosols, such as the SSA (Huang et al., 2014; Colarco et al., 2014; Das et al., 2015b). Therefore our results also need to be validated by sensitivity experiments using aerosols with different properties. A recent study by Tsikerdekis et al. (2017) demonstrated that simulated dust load and induced radiation change are sensitive to the dust particle size division in the model, so further sensitivity experiments using more dust size bins would be worthwhile. In addition, many other factors can also affect the EASM, including the El Niño–Southern Oscillation (Zhao et al., 2012; Liu et al., 2015), the North Atlantic Oscillation (Wu et al., 2009) and heat sources over the TP (Yanai et al., 2006; Duan et al., 2012). A recent numerical simulation by Wang et al. (2017) showed that aerosol emissions from outside East Asian play an important part in weakening the circulation of the EASM.

We conducted two numerical experiments to quantify the effects of dust aerosols emitted over the TP on the EASM in heavy/light dust years using a high-resolution RCM. Satellite and in situ observations were used to evaluate the simulated spatial distribution of dust aerosols and their seasonal and interannual variations. We analyzed the change in dust aerosols induced by emissions over the TP and their radiative effects on the EASM and summer precipitation in heavy/light dust years.

The spatiotemporal distribution of the dust AOD and their seasonal and interannual variation were captured well by the RegCM4.1 model compared with the MISR AOD and in situ observations from AERONET. Both the simulated and observed AOD were higher in spring/summer and lower in autumn/winter. The simulated dust AOD was higher in the Taklimakan, Gobi and Great Indian deserts, with peak values of > 0.6. The simulated dust AOD in the Qaidam Basin and the northern TP was also higher. The seasonal variation in the dust AOD at Nam Co was captured well by RegCM4.1 compared to the observed aerosol AOD.

Comparative analyses of the two simulations indicated that the dust aerosols generated over the TP had a profound influence on the EASM. The differences in the EASM index and column dust load between CON and SEN experiments are negatively correlated (r=-0.46). The index also weakened (strengthened) as the combined imported–local dust aerosol increased (decreased) over the TP. The net atmospheric heating rate was negative over the TP in heavy dust years as a result of the radiative cooling effects of the dust aerosols, leading to a 0.6 ∘C cooling in the surface and atmospheric temperatures. The land–sea thermal contrast and EASM were therefore both weakened, causing a 27 % reduction in precipitation in the southern monsoon region. The dust load over the TP in the light dust years was much less than in the heavy dust years, implying large interannual variability.