The Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2) was developed by NASA’s Global Models and Assimilation Office (GMAO) aiming to establish an integrated Earth system analysis datasets coupled with four segments: atmosphere, ocean, land and chemistry (Randles et al., 2017). MERRA-2 assimilates the meteorological observations and the aerosol data retrieved by satellite through a global data assimilation system. The atmospheric module of this dataset contains common meteorological factors: wind, temperature, pressure, humidity, etc. The aerosol module contains the dust emissions, dust mass mixing ratios and DOD (Gelaro et al., 2017). Although Xian et al. (2020) claimed that there exists a large uncertainty in the reanalysis datasets, MERRA-2 have been proved as a reliable product including both aerosol characteristics and meteorological factors (Jordan et al., 2020; Sun et al., 2020). Here, the MERRA-2 datasets are used to verify the simulated results in this paper. The monthly mean dust emission, DOD, 850 hPa wind and geopotential height from 1980 to 2015 with a horizontal resolution of 0.625° × 0.5° are used to evaluate the simulated results. All the data can be downloaded online at: https://disc.gsfc.nasa.gov/.

The Global Precipitation Climatology Project (GPCP) is an international research project with a monthly average data set of global precipitation (Huffman, 1997). The datasets combine the microwave data from the low-orbit satellites, the infrared data from the geosynchronous orbit satellites, and the rain gauge observations from the ground stations. The high precision microwave low-earth orbit datasets are used to calibrate and adjust the orbit satellite infrared remote sensing data, and the gridded datasets based on the rain gauge ground observation data are used to further adjust the rainfall distributions (Adler et al., 2003). The monthly averaged precipitation datasets with a horizontal resolution of 2.5° × 2.5° from 1980 to 2015 are used to validate the simulated model results for models in our study (https://climatedataguide.ucar.edu/climate-data/gpcp-monthly-global-precipitation-climatology-project).

The Community Atmospheric Model version 4 with a Bulk Aerosol Model (CAM4-BAM) consists of the following components: dust emission, dust transport, wet and dry depositions and radiation scheme (Neale et al., 2010). The released version has been described by Neale et al. (2010). The Dust Entrainment and Deposition (DEAD) model (Zender et al., 2003) has been used as the dust emission scheme in CAM4-BAM. The size distribution is fixed in four bins based on a log-normal distribution (Zender et al., 2003). This model has been successfully applied within the Community Earth System Model (CESM) in simulating dust emission over the Middle East and North Africa (MENA) region (Parajuli et al., 2016). The soil texture data set is from the International Geosphere-Biosphere Programme (IGBP, Bonan et al., 2002). The vertical dust mass flux has been calculated according to the source modes and wind speed (Schulz et al., 1998). The dust transport processes come from the CAM4 tracer advection scheme (Neale et al., 2010) and the dry deposition processes include the gravitational and turbulent parts (Zender et al., 2003). The wet deposition processes consist of the in- and below-cloud precipitation based on a prescribed dust solubility of 0.15 (Neale et al., 2010). For the optical parameters, CAM4 uses the OPAC scheme (optical properties of aerosols and clouds) with the short-wave radiation (Hess et al., 1998). The radiation scheme is from the CAM radiative transfer (CAMRT) scheme (Collins et al., 2006). However, there are some issues in the released version of CAM4-BAM in simulating the dust cycle (Albani et al., 2014), hence, we used the updated version with the following improvements in this study from Albani et al. (2014): 1) The soil erodibility maps have been optimized based on the observation. 2) The emitted dust size distribution is “scale invariant” in four bins (the ratio: 0.011, 0.087, 0.277 and 0.625), which is based on the brittle fragmentation theory (Kok, 2011). 3) The shortwave and longwave optical parameters are from Yoshioka et al. (2007) which has been found closer to the observations. 4) The dust solubility in the wet deposition process is increased from 0.15 to 0.3 based on the observation. The simulating data is at 1.25° × 0.9° horizontal resolution and the model simulations are conducted at 26 vertical levels.

The dust parameters (such as the dust size distribution, dust surface mass concentration and the deposition have been extensively evaluated by Albani et al. (2014). Regarding the radiative effect, a lot of comparison work has been done by Neale et al. (2013) based on the Clouds and the Earth’s Radiant Energy System (CERES) datasets, in situ datasets and reanalysis datasets. CAM4-BAM has been widely used to reveal the aerosol-climate interactions (Albani et al., 2014; Xie et al., 2018; Shi et al., 2019).

In order to investigate the roles of the fast and slow responses of ISM to the DRE of dust aerosols respectively, two sets of experiments were carried out: Coupled model with dust radiative feedback (CMD) and coupled model without dust radiative feedback (CMN). Similarly, two additional sets of simulations, uncoupled model with dust radiative feedback (UMD) and uncoupled model without dust radiative feedback (UMN), were also conducted. Therefore, the difference CMD-CMN can represent the total climate response, the difference UMD-UMN represents the fast response, and finally the difference (CMD-CMN)-(UMD-UMN) reflects the slow response induced by the dust DRE. For the CM experiments, both the CMD and CMN experiments were simulated for 60 years with the SST calculated and interacted synchronously with the atmosphere, and the simulated results for the last 30 years are taken for the analysis, allowing the first 30 years for spin up. For the UM experiments, we take the results of the last 15 years of the 21-years simulations with the prescribed-fixed SST getting from the monthly mean corresponding to 2000 AD.

According to the MERRA-2 reanalysis results, the highest dust emission flux (Figure 1A) and the DOD (Figure 1D) values in the WAD are mainly concentrated in the Nafud desert, the Dahna desert and the Rubal Khali desert in the Arabian Peninsula. The spatial distribution of the dust emission flux simulated by the CMD (Figure 1B) and the UMD (Figure 1C) corresponds well with the MERRA-2 reanalysis results in the most parts of the WAD, however, they underestimate in Oman, the United Arab Emirates, and overestimate in Iran. The DOD shows larger value over the Arabian Peninsula, the northern Arabian Sea and the Iranian Plateau (Figure 1D), which are also reproduced by the CMD and UMD simulated results. Both the coupled and uncoupled models can mainly capture the spatial distributions of the annual averaged dust emission flux and dust optical depth (DOD) in the WAD as compared to the MERRA-2 reanalysis. As for the annual averaged dust emission flux values and the DOD values, it shows that the simulations from the CMD are closer to MERRA-2 reanalysis comparing with those from the UMD simulations (Table 1).

Our model not only captures the spatial distributions of dust emission flux and the DOD, but also their annual cycle characteristics (Figure 2). The annual cycle characteristics of the dust emission flux and DOD regional averaged values in the WAD simulated by the CMD and UMD are basically consistent with the reanalysis results. For the dust emission flux (Figure 2A), the MERRA-2 reanalysis shows the higher values are in summer and then in spring (blue line). Both the CMD (red line) and UMD (black line) can capture these features, but the values in summer are higher than those in the reanalysis results. Meanwhile, the highest DOD values occurred from May to August (Figure 2B) and the simulated variations are in agreement with the reanalysis results. We conclude that both the CMD and UMD simulated results can well represent the spatio-temporal features of the WAD dust.

Dust aerosols in the WAD can be transported to the remote Arabian Sea and then influence the precipitation in India through the radiative effects (Jin et al., 2014; Vinoj et al., 2014). Such long-range transport is possible because high dust concentration is observed at 5–7 km height in the Saudi Arabian region in MPL-LIDAR data (Parajuli et al., 2020). In this study, we mention the CEI region (Figure 3A) as the main study domain, which is highly correlated with the precipitation trend at the central grid (23°N, 83°E) of ISM region following Jin et al. (2014), Vinoj et al. (2014). The annual cycle of the regional averaged precipitation showed that the precipitation values begin to increase in May and June compared with those in winter and then peak in July and August (Figure 3B), which we called the early monsoon season and the late monsoon season, respectively (Lau and Kim, 2007).

To evaluate the simulated results of ISM, the spatial distributions of 850 hPa circulation and precipitation from MERRA-2 reanalysis (850 hPa circulation), GPCP (precipitation), CMD simulations, UMD simulations and their differences in the early monsoon and late monsoon seasons are compared in Figures 4, 5. It is shown that the southwesterly flows over the Arabian Sea have emerged in the early monsoon season (Figure 4A), carrying large amounts of water vapor which favors the precipitation increase (Figure 5A). In the late monsoon season, both the Somali jet (Figure 4F) and precipitation in CEI (Figure 5F) are stronger than these in the early monsoon season, which are at their most powerful time in a year. Both the coupled and uncoupled models can capture the temporal and spatial distribution characteristics of the circulation except the CMD results in early monsoon which have underestimated the 850 hPa winds and overestimated the geopotential height (Figure 4D). Regarding precipitation, both the CMD and UMD simulations could performance well in the CEI region but overestimated the values in the south of Tibetan Plateau (Figure 5), this might be due to weaknesses in cumulus convective parameterization in the module because of the complex topography (Sun et al., 2020). For a quantitative evaluation of our simulations, we calculate the regional mean differences, the spatial correlation coefficient (R) and the root mean square error (RMSE) of the observed data [MERRA-2 for 850 hPa winds and geopotential height (GPH), GPCP for precipitation] and the simulated data for 850 hPa U, V, GPH and precipitation over the region in Figures 4, 5 (0–40°N, 30–90°E) during the early and late monsoon seasons (Table 2). The quantitative results also show that, CMD and UMD can both capture the meteorological fields in these two seasons.

Previous studies show that the DRE of the WAD dust aerosols can influence the ISM (Jin et al., 2014; Vinoj et al., 2014). From the (CMD–CMN) monthly precipitation anomalies (the total response), the averaged values in CEI induced by dust DRE show an increase during the early monsoon season (Figure 6D, mean value: 0.15 mm/day) but a decrease during the late monsoon season (Figure 7D, mean value: −0.25 mm/day). Since the opposite precipitation anomalies in these two periods, the relevant analyses are carried out respectively for the early monsoon (May–June) and late monsoon (July–August) seasons in the following.

Figure 6 shows the 850 hPa wind and precipitation anomalies from the total response, the fast response and the slow response to the dust DRE during the early monsoon season. As a typical representative branch of ISM, the southwesterly flows over the Arabian Sea occur in the early monsoon season (Figure 4A) and strengthen in the lower troposphere over the Arabian Sea (Figure 6A), which bring more water vapor to the Indian subcontinent from the Arabian Sea and lead to the increase of the precipitation in CEI (Figure 6D) from the total response to the dust DRE. The fast response of 850 hPa wind to the dust DRE (Figure 6B) strengthens over the Arabian Sea, and the precipitation increases in the CEI (Figure 6E), which are similar to the total response. However, the slow response of the ISM shows opposite changes, the northeasterly flows strengthen (Figure 6C) to inhibit the water transport to the Indian Subcontinent from Arabian Sea and decrease the precipitation in CEI (Figure 6F). In summary, although the fast and slow response work in opposite direction by increasing and decreasing the rainfall, respectively, the total response results in increased precipitation over the region of interest in the early monsoon season.

Figure 7 shows results similar to those in Figure 6 but for the late monsoon season. The total response of the 850 hPa wind induced by the dust DRE shows the strong easterly flows over Arabian Sea (Figure 7A), inhibits the water transport to the land and decreases the precipitation in CEI (Figure 7D). The fast response of the precipitation slightly increases (Figure 7E) in CEI (Mean value of precipitation anomalies in the CEI: 0.08 mm/day) because of the strengthened westerly flows (Figure 7B) over the Arabian Sea. However, the slow response of 850 hPa wind (Figure 7C) to the dust DRE shows the strengthened easterly flows over the Arabian Sea and the decreases on precipitation in CEI (Figure 7F), which are similar to these of the total response. These results show that the dust DRE weakens the ISM in the late monsoon season, which is dominated by the slow response.

The fast response of ISM to the dust DRE dominates the enhancement of southwesterly flows over the Arabian Sea and precipitation in the CEI during the early monsoon season, and the involved physical mechanisms are revealed in this section.

The radiative forcing differences induced by the dust DRE can contribute to the temperature changes at the surface and in the atmosphere, thereby affecting the local atmospheric circulation (Miller and Tegen, 1998; Xie et al., 2018). Figure 8 shows the differences of short-wave, long-wave and net radiative forcing at the surface, the top of the atmosphere (TOA) and in the atmosphere induced by the dust DRE in the early monsoon season. The net radiative forcing is the sum of long-wave and short-wave radiative forcing, and the radiative forcing at the TOA is the competition results between those at the surface and in the atmosphere. The dust DRE short-wave forcing decreases at the surface (Figure 8A) but increases in the atmosphere (Figure 8C), which results in a positive anomaly at the TOA (Figure 8B). The variation of net radiative forcing induced by the dust DRE at surface (Figure 8G), at TOA (Figure 8H) and in the atmosphere (Figure 8I) are consistent with those of shortwave radiative forcing over the WAD. As for the long-wave radiative forcing, a positive change at the TOA (Figure 8E) induced by the dust DRE over the WAD contributes to the positive surface changes (Figure 8D) and the negative atmospheric changes (Figure 8F).

The surface temperature anomalies induced by the dust DRE in regions at the low-latitudes such as the WAD and the Arabian Sea correspond well with the changes in the net radiative forcing at the TOA (Figure 8H). The significant warming occurs over the WAD in the total responses of temperature at surface (Figure 9A) and at 850 hPa (Figure 9D) in response to the dust DRE, which corresponding well with the low-pressure anomalies at 850 hPa over the Arabian Peninsula (Figure 9G). As stated in the above Section 4.1, the enhancement of the ISM induced by the dust DRE is mainly caused by the fast response during the early monsoon season. Because of this, the fast response (Figures 9B,E,H) of the temperature at surface, at 850 hPa and the geopotential height anomalies are almost identical to the total response to the dust DRE over the WAD. The low-pressure anomalies over the WAD and its surroundings at 850 hPa (Figure 9H) strengthen the southwesterly flows over the Arabian Sea in Figure 7B. On the contrary, the slow response of the 850 hPa geopotential height (Figure 9I) to the dust DRE shows the high-pressure anomalies because of the widespread cooling of the Arabian Sea at 850 hPa (Figure 9F) and at the sea surface (Figure 9C). Hence, the reduced SST over the Arabian Sea weakens the ISM during the early monsoon season.

As for the water vapor transport, both the total (Figure 10A) and fast (Figure 10B) responses of 850 hPa moisture flux to the dust DRE enhances over the Arabian Sea, which contributes to the transport of water vapor from the ocean to the CEI. Meanwhile, the total and fast responses of specific humidity at 850 hPa to the dust DRE increase in the CEI (Figures 10D,E), which are in agreement with the precipitation changes induced by the dust DRE during the early monsoon season. But the slow responses of the 850 hPa moisture flux (Figure 10C) and specific humidity (Figure 10F) are both weakened.

During the early monsoon season, the dust aerosols heat the atmosphere in the lower troposphere over the Arabian Peninsula through the dust DRE, resulting in the formation of the thermal low-pressure, which strengthen the southwesterly winds over the Arabian Sea, and thus strengthen the water vapor transport from the ocean to the land, leading to the increasing precipitation in the CEI. These results are mainly caused by the fast response processes, while the slow response processes weaken the ISM and inhibit the precipitation in CEI.

The dust DRE weakens the ISM through cooling the Arabian Sea (slow response processes) in the late monsoon season as shown in Section 4.1 (Figure 7). In the early monsoon season, the strengthened southwesterly winds over the Arabian Sea reinforce the ISM, and further transport more dust aerosols from the WAD to the ocean (Jin et al., 2014). According to the annual cycle of dust emissions in the WAD, it showed that the emissions in the late monsoon season are higher than those in the early monsoon season (Figure 2A). Meanwhile, the regional averaged DOD in the Arabian Sea (0.32) during the late monsoon season is also higher than that (0.24) during the early monsoon season due to the enhancement of the dust transportation.

The short-wave, long-wave and net radiative forcing anomalies at the surface, TOA and in the atmosphere caused by the dust DRE over the Arabian Sea (Figure 11) are basically consistent with the spatial distribution characteristics of those in the early monsoon season (Figure 8). The dust DRE leads to reduction in radiative forcing anomalies at the surface (Figure 11G) and TOA (Figure 11H) over the Arabian Sea, but increases in the atmosphere due to the presence of absorbing dust aerosols (Figure 11I). It is worth noting that the increases in the net radiative forcing at the TOA over the Arabian Peninsula caused by the dust DRE are not significant in the late monsoon season (Figure 11H).

The total and slow responses of temperature at the surface (Figures 12A,C) and 850 hPa (Figures 12D,F) to the dust DRE show negative anomalies over the Arabian Sea, and the geopotential height at 850 hPa show positive anomalies over this area (Figures 12G,I). The pressure anomalies produce the easterly flows over the Arabian Sea (Figures 12G,I), inhibit the water vapor transport (Figures 13A,C), thus decrease the specific humidity (Figure 13F) in CEI and weaken the ISM (Figures 7C,F). On the contrary, the fast response of the ISM to dust DRE shows the positive effects (Figure 13E) but insignificant on westerly flows over the Arabian Sea (Figure 13B).

To sum up, the dust DRE strengthens the ISM by heating the lower troposphere over the WAD during the early monsoon season, but weakens it by cooling the sea surface and lower troposphere over the Arabian Sea during the late monsoon season. Further, the enhancement of the ISM during the early monsoon season is dominated by the fast response while the weakening during the late monsoon season is dominated by the slow response.