1 Introduction

Atmospheric aerosols, which are fine particles suspended in the air, comprise a mixture of mainly sulfates, nitrates, carbonaceous (organic and black carbon) particles, sea salt, and mineral dust. Atmospheric aerosols have a strong impact on the Earth’s radiation budget and climate (Stocker 2014). Aerosols interact with climate system through scattering and absorption of radiation (direct effect) and through modification of the microphysical properties of clouds (indirect effect). Absorbing aerosols, such as black carbon (BC), organic carbon (OC) and dust, can both absorb and reflect sunlight, thus heating the lower-troposphere and cooling the surface. In contrast, non-absorbing aerosols, such as sulfate, generate mainly surface cooling but weak atmospheric heating. In recent decades, the BC emissions are particularly large in China and India due to energy combustion and biomass burning and have brought the world’s attention (Ramanathan and Carmichael 2008; Yang et al. 2022; Wei et al. 2022).

Extensive studies concerning the roles of aerosols in the Earth’s climate did not begin until the 1990s. Aerosols can affect monsoon rainfall, and regional climate change through radiative forcing and microphysical effects (Rosenfeld 2000; Li 2004; Nakajima et al. 2007; Li et al. 2007, 2011a, 2011b; Huang et al. 2014; Guo et al. 2016). Many general circulation model (GCM) studies have investigated the impacts of aerosols on global and regional changes in precipitation (Menon et al. 2002; Lau et al. 2008; Wang et al. 2009a, b; Bollasina et al. 2011, 2013; Cowan and Cai 2011; Ganguly et al. 2012). The weakening of monsoon circulation and rainfall reduction have been attributed to aerosol effects (Ramanathan et al. 2005; Chung and Ramanathan 2006; Ramanathan and Carmichael 2008; Liu et al. 2009; Cowan and Cai 2011; Bollasina et al. 2011; Salzmann et al. 2014; Krishnan et al. 2016) and equatorial Indian Ocean warming due to increased GHG (Ramanathan et al. 2005; Chung and Ramanathan 2006; Annamalai et al. 2013; Lee and Wang 2014; Sabeerali and Ajayamohan 2017).

As an integral component of the Earth’s hydrological cycle, the Indian summer monsoon (ISM) is the largest monsoon system critical for the well-being of over two-thirds of the world’s population. The observational evidence of ISM circulation experienced a significant declining trend from the 1950s together with a weakening local meridional circulation and notable precipitation decreases over north-central India and the west coast that are associated with a reduced meridional temperature gradient (Ramanathan et al. 2001, 2005; Krishnan et al. 2013; Goyal 2014; Roxy et al. 2015; Praveen et al. 2020). Although it could potentially be altered by multidecadal variations (Shi et al. 2018) arising from internal modes of climate variability such as the Atlantic multidecadal variation (AMV) and the interdecadal Pacific oscillation (IPO) (Krishnan and Sugi 2003; Ding et al. 2008, 2009; Zhou et al. 2008; Cheng and Zhou 2014; Salzmann and Cherian 2015; Jiang and Zhou 2019), the high aerosol emissions in South Asia has made the role of aerosol effect a critical issue. The aerosol effects on the energy-water cycle and monsoon dynamics are strongly dependent on aerosol distribution and characteristics as well as its spatial and temporal variations. Elevated deep layers of radiation-absorbing aerosols can potentially affect the water cycle by significantly altering the energy balance (Ramanathan et al. 2005; Lau and Kim 2006; Lau et al. 2006).

The increase of aerosols and associated impact on the ISM has been documented in many modeling studies (e.g., Ramanathan et al. 2005; Lau et al. 2006; Meehl et al. 2008; Ganguly et al. 2012; Bollasina et al. 2013). Ramanathan et al. (2005) used a coupled ocean–atmosphere model to show that absorbing aerosols over India could decrease monsoon precipitation by reducing surface shortwave radiation, which limits the amount of evaporation, as well as increasing atmosphere heating, which stabilizes the troposphere over South Asia. The high aerosol regions of the northern portion of the Indian Peninsula and the Arabian Sea are cooled relative to the oceans to the south, leading to a reduction of the meridional thermal gradient and a slowing down of the local meridional circulation. The slower circulation reduces surface evaporation and provides a positive feedback, further weakening the monsoon. In a somewhat different modeling approach, Lau et al. (2006) used an atmospheric model together with an observational analysis of Lau and Kim (2006) to emphasize the importance of temporal and regional distributions of both natural and anthropogenic aerosols not only as a forcing agent but also as an integral part of a dynamical feedback mechanism involving clouds, rainfall, and winds that can alter the evolution of ISM system. They proposed the elevated heat pump (EHP) theory which posits that atmospheric heating by deep layers of BC and dust accumulated over the Indo-Gangetic Plain (IGP) and the Tibetan Plato slope (TP) can induce a moisture convergence feedback, leading to increased precipitation in northern India during March to June. Specifically, the accumulation of BC and dust over the northern and southern slopes of TP absorbs shortwave over the longitudes of the Indian Peninsula during pre-monsoon periods and heats the lower and middle troposphere around the TP. The heated air rises via dry convection, creating a positive temperature anomaly in the mid-level to upper troposphere over the southern slope of TP relative to the region to the south. The rising hot air forced by the increased heating in the upper troposphere from the North Indian Ocean draws in more warm and moist air over Indian Peninsula, setting the stage for the onset of the ISM.

In our recent study (Chen et al. 2018), the anthropogenic aerosols lead to surface cooling through the decrease in surface short wave flux from October to December. The surface cooling region extends downwind of major emission source areas to western and northern India, as the aerosols are transported by the prevailing winds. The precipitation strongly reduces during October to December over western and northern India, and the reduction is mainly contributed by the vertical convergence term that is associated with changed vertical motion by anthropogenic aerosols. Furthermore, Wang et al. (2013) further demonstrated that absorbing aerosols were particularly important in influencing the circulation and precipitation over the northern Indian regions in winter.

Similar to the radiation effect of BC and OC, the natural dust also absorbs solar radiation that causes atmospheric heating and surface cooling. Long-range transport of dust aerosols originated in the Middle East is known to be a significant source of aerosol over the Arabian Sea and Indian region during the pre-monsoon season. As a result, enhanced aerosol loading in Indian Peninsula during the pre-monsoon season is likely contributed by both anthropogenic and natural aerosols (Satheesh and Srinivasanan 2002; Badarinath et al. 2010). A recent study by Wei et al. (2022) reveals a decreasing dust transport by altered monsoon flow caused by reduced BC burden over the northern (70°–88°E, 25°–35°N) but increased BC burden over southern (70°–88°E, 15°–25°N) India during the lockdown of COVID-19. Their results indicate that the solar heating in April and May is decreased by BC reduction in northern India and the surface albedo of the TP southern slope is increased due to the reduced BC snow-darkening effect. The northern Indian atmosphere responds to this cooling with a descending motion and enhanced atmospheric stability. Furthermore, the surface and near-surface cooling over the southern slope of TP by increased albedo lead to southward cold air advection. The cooling over northern India and warming over southern India from increased crop residue burning BC emissions induce an anomalous southward pressure gradient force. Therefore, an anomalous northward Coriolis force as the emergence of easterly wind anomalies appears to balance the anomalous pressure gradient force. Owing to anomalous easterly wind, the eastward transport of dust from the Middle East and Sahara as well as local dust emissions in the Thar Desert are suppressed.

Despite a large number of studies, there still exists a gap in our knowledge and understanding of anthropogenic aerosols and their climate effect. Aerosol processes are still poorly observed and treated in numerical models. Here, we plan to revisit the aerosol-induced monsoon variability along with aerosol’s temporal transition within the monsoon evolution. In difference from the prescribed simulates aerosol radiative forcing using simulated transport and simulated aerosol spatial distributions. We use the historical emission inventory which quantified anthropogenic and biomass emissions of climate-relevant species for the period 1850–2000 by Lamarque et al. (2010). We focus especially, on the regional anthropogenic aerosols forcing over India to identify the ISM evolution responses. The paper is structured as follows. A brief description of the model and experimental design is given in Sect. 2. The analyses of ISM characteristics are shown in Sect. 3, including Indian monsoon characteristics in observation data and model simulation. The changes in ISM onset/withdrawal date and rate of changes are also discussed in this section. Each period during Indian monsoon evolution from pre-monsoon to monsoon withdrawal are discussed in Sects. 4, 5, and 6, including the effect of anthropogenic aerosols on the circulation and precipitation over the Indian Peninsula. The natural dust distribution response to BC climate impacts over monsoon evolution is discussed in Sect. 7. At the end of the text, a summary and discussions are given in Sect. 8.

2 Model and experiment designs

For the current climate simulation study, we use the National Center for Atmospheric Research Community Earth System Model (CESM) v1.0.3. The atmospheric component of CESM is the Community Atmosphere Model (CAM) version 5.1 (Neale et al. 2012). CAM5.1 is the atmosphere model of the NCAR CESM1 supported by the US National Science Foundation and DOE. It has a horizontal resolution of 1.9° × 2.5° and a vertical resolution of 30 levels from the surface to 3.6 hPa with a finite volume dynamical core. The properties and processes of major aerosol species included black carbon, primary organic matter, secondary organic aerosol, mineral dust, sulfate, and sea salt are treated in the three-mode modal aerosol scheme (MAM3, Liu et al. 2012). The aerosol size distribution is represented by three log-normal modes: Aitken, accumulation, and coarse modes. The associated mass mixing ratios of different aerosol components and the number concentration in each mode are predicted. The ZM scheme (Zhang and McFarlane 1995) with dilute convective available potential energy modification is used for deep convection parameterization. The shallow convection parameterization uses the Park and Bretherton scheme (Park and Bretherton 2009). The Bretherton and Park moist turbulence scheme (Bretherton and Park 2009) is used to parameterize the stratus-radiation-turbulence interactions in boundary layer. The Morrison and Gettelman scheme (Morrison and Gettelman 2008) is used for stratiform-cloud microphysics and the mechanism of Park (2010) is used for macrophysics. The Rapid Radiation Method for GCMS (RRTMG, Iacono et al. 2008) is used for the radiative transfer calculations. The land process simulation uses the Community Land Model (CLM) version 4.0 with explicit representation of land hydrological processes and land–atmosphere interactions (Lawrence et al. 2011; Oleson et al. 2010).

In this study, only the aerosol direct effect is considered by assuming number/mass fraction of aerosols activated to zero in cloud microphysics schemes. We carry out two experiments subject to the same prescribed climatological annual cycle of monthly mean sea surface temperature (SST) and the greenhouse gas concentration at the level of the Year 1850, and the annual mean surface aerosol emissions by the Year 1850 data without seasonal cycle (Lamarque et al. 2010). But the two experiments are given by different anthropogenic aerosol emissions in the India Peninsula within (70°–95°E, 5°–35°N) where the emission in the control simulation (CTL) is kept the same as the Year 1850 but changed to the Year 2000 in the aerosol forced experiment (AERO) that contains additional anthropogenic emissions over India Peninsula also without seasonal cycle (Lamarque et al. 2010). The emission of anthropogenic aerosols containing black carbon (BC), organic carbon (OC), and sulfate over the Indian Peninsula for the Year 2000 is shown in Fig. 1, and the value of each species are shown in Table 1. The area of highest emission resides over the IGP extending longitudinally against the Himalayan range from Bangladesh to northwestern India and Pakistan. The distribution of emission in Fig. 1 is consistent with the aerosol distribution by satellite-measured aerosol index (e.g. Bollasina et al. 2008). The emission of BC, OC and sulfate have 70%, 21% and 8% in total emission of anthropogenic aerosols as the Year 2000 data (Table 1). The areal mean aerosol emission over the Indian Peninsula for the Year 2000 and its difference from the Year 1850 are shown in Fig. 2 (dashed line). The natural aerosol i.e., dust and sea salt emissions in both simulations (CTL and AERO) are given by the Year 1850 data (Lamarque et al. 2010). And the aerosol optical depth (AOD) of natural aerosol also shows in Fig. 2 grey line.

Fig. 1
figure 1

Emission of anthropogenic aerosols (BC, OC, and sulfate) of Year 2000 (shaded, \({10}^{-1}\bullet {\mathrm{Mg}\bullet \mathrm{m}}^{-2}\bullet {\mathrm{month}}^{-1}\)) from the IPCC AR5 emission data set (Lamarque et al. 2010). The sources of emission include industrial, domestic and agriculture activities. Surface topography is superimposed in contour at 200 m interval and selected contours highlighted as following: 200 m (blue dash), 600 m (red dash), 1000 m, 2000 m, 3000 m and 4000 m (thick solid). The Tibetan Plateau is denoted by areas with elevation above 4000 m and shaded grey

Full size image
Table 1 Emission of anthropogenic aerosols (BC, OC, Sulfate) averaged within Indian region (70°–90°E, 5°–35°N) with Year 2000 data and percentage of emission differences
Full size table
Fig. 2
figure 2

Time series of mean anthropogenic aerosols (BC, OC, sulfate) emission (\({\mathrm{Mg}\bullet \mathrm{m}}^{-2}\bullet {\mathrm{month}}^{-1}\)) of Year 2000 (blue dash), and the simulated annual cycle of anthropogenic aerosol optical depth (AOD, blue solid) in the Indian sub-continent region, and the differences of aerosol emission (red dash) and AOD (red solid) between Year 2000 and Year 1850, and dust AOD (grey line) of Year 1850

Full size image

For the two experiments, the CESM is integrated for 30 years. The outputs of each year are considered an ensemble member, and we use the last 25 years to form ensemble averages as the annual cycle. The annual cycle of areal mean AOD corresponding to the surface emission in the India Peninsula region for the Year 2000 and its difference from the Year 1850 are shown in Fig. 2. The figure shows that the AOD of the Year 2000 increases sharply from February to April–May and then rapidly decreases in June–August following the onset of monsoon rains, and rise again mildly at the end of the rainy season in October. This is similar to the climatological annual cycle of the aerosol index over IGP based on satellite data in Bollasina et al. (2008) except the minimum AOD in AERO appears in September, one month later than that of the satellite climatology.

3 Simulated and observed climate of ISM

3.1 Climatological evolution over India Peninsula

We first examine the model simulated climatological evolution of the ISM against the observed ISM evolution based on the 25 years of AERO outputs and the ERA-interim and GPCP data for 1979–2020. Figure 3 shows the time-latitude distributions of observed and simulated climatological fields over the Indian longitude sector (70°–90°E) of 500-hPa temperature, surface air temperature, 850 hPa zonal wind and precipitation. Following the movement of the Sun and resultant land-sea thermal contrast, surface temperature over India starts to rise first in the central region in March (> 24 °C) and then over the whole sub-continent in April through mid-June when the IGP reaches above 28 °C. The temperature remains above 24 °C from July to August and cools in September and October. The 500-hPa temperature rise in spring and summer lags that of the surface temperature rise by about one month, and falls earlier in August and September than the surface temperature does (Fig. 3a, b). The rapid establishment of warm temperature at 500 hPa over northern India and the Tibetan Plateau from late April and early May marks an abrupt shift of zonal wind in the lower troposphere (u850) from easterly to westerly (Fig. 3c), a northward shift of the Intertropical Convergence Zone (ITCZ) and the start of monsoon rainfall over India (Fig. 3d). Likewise, the fall of 500 hPa temperature in late September and early October is associated with a reversed u850 change and termination of monsoon rainfall.

Fig. 3
figure 3

Time-latitude distribution of climatological fields over India longitude sector (70°–90°E) a 500 hPa temperature, b surface air temperature, c 850 hPa zonal wind, and d precipitation, derived from ERA-interim and GPCP for 1979–2020. The model simulated climatological fields for the corresponding fields are shown in e–h derived from 25 years of AERO outputs

Full size image

The model simulated annual cycle in AERO corresponding to the observed fields is shown in Figs. 3e–h. The figure shows that CESM simulates the overall evolution of ISM but overestimates rainfall in northern India to the south of the Tibetan Plateau (Fig. 3h) and the associated regional temperature (Fig. 3e) and wind (Fig. 3g) in the lower troposphere near the Tibetan Plateau. We will further evaluate the simulated circulation and rainfall in three stages of the ISM evolution below.

3.2 Evolution of ISM

Previous studies proposed different indices for describing the ISM evolution. In this study, we follow Wang et al. (2001) who defined the ISM evolution based on the difference of the 850 hPa zonal winds between a southern region of 40°–80°E, 5°–15°N and a northern region of 70°–90°E, 20°–30°N. This index is defined because low-tropospheric circulation responds to convective heating better than the upper-level circulation or vertical shear does, and the low-tropospheric vorticity is highly indicative of the strength of boundary layer moisture convergence and precipitation in regions away from the equator. In addition, such an index reflects both the intensity of the tropical westerly monsoon and the lower-tropospheric vorticity anomalies associated with the ISM trough. The ISM onset date is defined as the transition date when the circulation index changes from negative to positive values in spring. Following the same approach, we define the ISM withdrawal date as the transition date when the index reverses from positive to negative values in fall.

The evolution of 25-year ensemble-mean zonal wind index calculated from CTL and AERO are compared against the observed climatological index (Fig. 4). Both CTL and AERO simulate the same ensemble-mean monsoon onset date on May 12 (Fig. 4a, b) in agreement with the observed onset date (Fig. 4c). On the other hand, the mean withdrawal date in CTL is similar to the observed withdrawal date around October 20 but the withdrawal date in AERO is 10 days earlier on October 10. Besides the ensemble mean dates, we also examine the frequency distributions of ISM onset and withdrawal dates of the 25 years of simulations in Fig. 5a and c. Considering both ensemble mean and frequency distribution, the onset dates in AERO are not significantly different from that in CTL (Fig. 5a) but the withdrawal dates in AERO are well separated from that in CTL indicating that withdrawal occurs significantly earlier in AERO than in CTL (Fig. 5c). Note also that the onset and withdrawal dates in AERO have a larger spread than that in CTL. In addition, we define the rate of change in onset (or “onset rate” in short) and withdrawal (or “withdrawal rate”) by calculating the differences of zonal wind index within the transition periods within 2 pentads before and after the mean monsoon onset and withdrawal date (Fig. 4, orange shadings). The frequency distributions in Fig. 5b and d show that the onset rate in AERO is significantly larger than in CTL, but the withdrawal rate in AERO is significantly weaker than in CTL. Possible reasons for the aerosol influences on the monsoon onset and withdrawal are discussed in Sect. 4.

Fig. 4
figure 4

Climatological evolution of the zonal wind index as defined in Wang et al. (2001) (blue line), all-India precipitation (mm, solid line), and Kerala precipitation (mm, dash line, used by India Meteorological Department to define monsoon onset) for (a) CTL, (b) AERO and (c) ERA-interim/GPCP. The standard deviation of the simulated zonal wind index of the 25 years simulations from CTL and AERO are shown in a, b by blue shadings. The 25-year mean zonal wind index changes sign from negative to positive on May 12 in both CTL and AERO (monsoon onset), and from positive to negative on Oct. 20 in CTL and Oct. 10 in AERO (monsoon withdrawal). The onset and withdrawal dates are marked by vertical dashed lines, and the transition periods are shaded orange within 2 pentads before and after the mean monsoon onset and withdrawal dates. Total monsoon precipitation is 571 mm in CTL and 539 mm in AERO

Full size image
Fig. 5
figure 5

Frequency distributions of India monsoon onset and withdrawal dates (a, c) and their corresponding rate of changes (b, d) of the 25 ensemble years in CTL and AERO. The statistics is shown in box plots drawn from the first quartile (bottom) to the third quartile (top) and a horizontal line through the box at the median with the mean value marked by a cross and the date of 90% and 10% ranking marked by the sign “ − ” linked to the box by vertical lines

Full size image

The characteristics of ISM evolution are also shown in precipitation over Indian Peninsula. India Meteorological Department defines the ISM onset by the precipitation of Kerala when it exceeds 9 mm/day. In our simulations, the ISM onset and withdrawal defined by precipitation over Kerala is consistent with the phase transition defined by the zonal wind index (Fig. 4). The climatological evolution of ISM in terms of Kerala precipitation is similar to the evolution of all-India precipitation as also shown in Fig. 4. During the ensemble mean ISM period of AERO (151 days) and CTL (161 days), total precipitation of India Peninsula is 539 mm and 571 mm, corresponding to 3.56 mm/day and 3.54 mm/day, respectively.

We further evaluate the simulated fields of circulations and precipitation against the observed climatology in three periods of the ISM evolution: pre-monsoon, developed monsoon, and withdrawal (Fig. 6). The pre-monsoon period in each of the 25 years from AERO and CTL is selected the same, i.e. March 1 to April 10 in early spring before the significant increase in monsoon rainfall (Fig. 4). Such a selection ensures that the climatological signal of seasonal transition is excluded from the difference fields of circulation between AERO and CTL assuming the climatological seasonal cycle of the two experiments is the same. For the monsoon season, we choose two and half months from May 26 to August 3 to define the fully developed monsoon state. The pre-monsoon and developed monsoon periods for both CTL and AERO are the same as that of the observed climatology. For the monsoon withdrawal period, we choose the four-week period centered on the withdrawal date as the transition period. So the transition period of the CTL is October 3 to November 6, the same as the observed climatology, while that of AERO is September 23 to October 27.

Fig. 6
figure 6

Observed climatological fields of precipitation (shaded), horizontal winds at 850 hPa (\({\mathrm{m}\bullet \mathrm{s}}^{-1}\), vector) and geopotential height at 500 hPa (m, red dash) in a pre-monsoon, b post-monsoon, and c withdrawal, derived from GPCP and ERA-interim. The simulated ensemble mean fields of 25 years of AERO corresponding to a–c are shown in d–f. The two rectangles enclosing the area (40°–80°E, 5°–15°N) and (70°–90°E, 20°–30°N) are used to define the zonal wind index [U850(1) − U850(2)] for the evolution of ISM

Full size image

The observed climatological circulation and precipitation in the pre-monsoon period (Fig. 6a) indicates that the circulation in the pre-monsoon period is characterized by subtropical high over the northern Indian Ocean and surrounding land region as shown by the 500-hPa high-pressure ridge along 10°N and anticyclonic flow at 850 hPa centered around the Arabian Sea, Bay of Bengal, and the Indo-China flanked by equatorial easterlies and subtropical westerlies. Precipitation is confined to the eastern equatorial Indian Ocean and Sumatra. Following the sun in boreal summer, the climatological flow in the fully developed monsoon period shows pronounced cross-equatorial flow and the prevailing westerlies over the Indian Ocean and the Bay of Bengal. The strong westerly monsoon encounters the Indian subcontinent and Indo-China Peninsula to produce heavy monsoon rain over the western coastal regions and neighboring oceans. The associated subtropical high at 500 hPa develops a strong meridional structure with a strong anticyclonic center over the Arabian Peninsula and a cyclonic center over India and the Bay of Bengal. In the monsoon withdrawal period, the overall circulation and equatorial precipitation largely resemble that of the pre-monsoon period but precipitation resides over southern India and the Bay of Bengal. This is known as the asymmetric seasonal transitions between boreal spring [March–May (MAM)] and fall [September–November (SON)] monsoon regimes, i.e. the maximum rainfall remains mostly south of the equator during boreal spring without a northwestward progression that would retrace the path of the boreal fall progression (e.g., Lau and Chan 1983; LinHo and Wang 2002; Hung et al. 2004; Chang et al. 2005). The above climatological features are reasonably simulated in AERO except for an obvious positive rainfall bias along the southern Tibetan peripheral primarily over the northern Ganges–Brahmaputra basin. The bias is most extensive in the developed monsoon and withdrawal periods. The model also overestimates rainfall in the eastern Arabian Sea and western India and underestimates rain over the Bay of Bengal in the developed monsoon period. As a result, the model-simulated southwestlies over the tropical western Indian Ocean is stronger than the observed monsoon flow.

The natural dust in Indian Peninsula is also associated with monsoon circulation. In our simulations, the AOD of Indian dust shows a seasonal cycle and has a high value from April to July (Fig. 2). We evaluate the simulated fields of dust against the observed climatology in the three periods of ISM evolution (Fig. 7). In the pre-monsoon period, high dust AOD appears over the middle Arabia Peninsula and the Tarim Basin in the observed climatology. The monsoon circulation transports dust from the Middle East, leading to relatively high dust AOD. In the developed monsoon period, the high dust AOD appears in Arabia Peninsula and extends to northwestern India where the dust is carried by monsoon circulation into Ganges Basin and accumulated. The dust AOD decreases during monsoon withdrawal over the Arabia Peninsula to Pakistan. In addition, the Tarim Basin also shows around 0.4–0.5 AOD of dust in those three periods in the observed climatology. The climatological features of dust are also reasonably simulated in AERO except for an overestimate of dust AOD in the Tarim Basin. That is because the dry deposition and transportation in the MAM3 aerosol scheme are underestimated, leading to the overestimation of dust AOD in the source region over the Middle East and South West East (Liu et al. 2012; Wu et al. 2020). The overestimate is most notable in the developed monsoon.

Fig. 7
figure 7

Observed climatological fields of dust AOD (shaded), horizontal winds at 850 hPa (\({\mathrm{m}\bullet \mathrm{s}}^{-1}\), vector) in a pre-monsoon, b post-monsoon, and c withdrawal, derived from MERRA2 and ERA-interim. The simulated ensemble mean fields of 25 years of AERO corresponding to a–c are shown in d–f

Full size image

4 Effect of anthropogenic aerosols on pre-monsoon climate

The regional circulation in the pre-monsoon period of AERO (March 1 to April 10) is shown by the ensemble-mean fields of horizontal winds and vertical velocity (− ω) at 850 hPa and 500 hPa are shown in Fig. 8a and b, and the vertical-meridional cross-section of mean − ω within 70°–90°E in Fig. 8c. The low-tropospheric circulation (850-hPa winds in Fig. 8a, the same as Fig. 6d) is characterized by anticyclonic flow over the Arabian Sea and Bengal-Indo China surrounded by westerlies along the southern Tibetan peripheral in 20°–30°N and easterlies around 10°N. Associated with the circulation, deep ascending motion appears in the tropical eastern Indian Ocean as shown by positive values of − ω at 850 hPa and 500 hPa (Fig. 8a and b) and shallow ascending motion over India Peninsula at 850 hPa capped by descending motion at 500 hPa (Fig. 8a–c), indicating the emergence of a thermal low over the India Peninsula. At 500 hPa, broad subtropical anticyclonic flow and descending motion reside over India and the neighboring Arabian Sea and Bay of Bengal that is accompanied with a strong ascending motion to the north over the southern slope of the Tibetan Plateau (~ 30°N) (Fig. 8b, c).

Fig. 8
figure 8

Ensemble mean fields in pre-monsoon period of AERO (March 1 to April 10) a vertical velocity, − ω (shaded) and horizontal wind (\({\mathrm{m}\bullet \mathrm{s}}^{-1}\), vector) at 850 hPa, b − ω and horizontal wind at 500 hPa, c 70–90°E averaged − ω, d–f AERO-CTL differences corresponding to the fields shown in a–c. Differences of − ω are shown by contours (− ω < 0 with dashed line, > 0 with thick solid line) with interval 0.04. A confidence level greater than 90% (shaded) use the Student’s t-test

Full size image

To examine the effect of anthropogenic aerosols on the pre-monsoon climate, we compare the fields of AERO against the difference fields of (AERO–CTL). First, we show the mean AOD in AERO and the differences (AERO–CTL) in Fig. 9. The figure shows that the overall suppressed condition in the lower-tropospheric anticyclonic flow (shown in Fig. 8b) in the pre-monsoon condition in AERO results in aerosols accumulation over the India Peninsula, especially downstream over eastern India and along eastern IGP without being washed out by rainfall as shown by the mean AOD (Fig. 9, shaded). The AOD differences (AERO–CTL) in Fig. 9 show the anthropogenic aerosols add further accumulations in the Ganges basin over 20°–30°N along the southern Tibetan peripheral and over eastern India and neighboring Bay of Bengal (10°–20°N). The distribution of AOD differences (AERO–CTL) are obviously determined by lower tropospheric winds and topography.

Fig. 9
figure 9

Ensemble mean aerosol optical depth (AOD, shaded) and horizontal wind at 850 hPa (\({\mathrm{m}\bullet \mathrm{s}}^{-1}\), vector) in pre-monsoon period of AERO (March 1 to April 10). The AOD differences (AERO–CTL) > 0 are shown by red contours with contour interval 0.02 begin from 0

Full size image

By absorbing and scattering solar radiation, anthropogenic aerosols generally lead to atmospheric heating and surface cooling as revealed in temperature differences (AERO–CTL) near the surface (Fig. 10a) and areal mean vertical profile of temperature (Fig. 10b) over the Indian Peninsula. The negative (positive) temperature difference near surface (in the lower troposphere) Fig. 10b shows an enhanced stability by aerosols in the pre-monsoon period. The surface energy budget averaged over the region 70°–90°E, 5°–35°N for AERO and AERO–CRL is shown in Table 2. The budget differences (AERO–CTL) are dominated by the net surface fluxes of shortwave (positive downward) and longwave (positive upward) by − 7.26 W m−2 and − 2.01 W m−2, respectively. The reduced surface radiation in both downward shortwave and upward longwave is consistent with reduced surface temperature. The horizontal distribution of AOD in AERO and differences (AERO–CTL) in Fig. 9 and the vertical-meridional distribution of aerosol concentration differences (AERO–CTL) within (70°–90°E) in Fig. 10c indicate that anthropogenic aerosols are transported by westerly and southwesterly winds (Fig. 8d and e) and accumulated over the terrain of Himalayas, eastern India and coastal ocean. The accumulated aerosol concentrations coincide well with the solar heating rate (Fig. 10c, shaded). Note that the heating source is all over the mountainous terrain from the foothill to the top of the Himalayas. Over 10°–20°N, the anomalous aerosols also heat the atmosphere, but the altitude of the heating source is lower without terrain effect.

Fig. 10
figure 10

Ensemble mean differences (AERO–CTL) in pre-monsoon period (March 1 to April 10) a Surface temperature (°C, shaded), b The vertical profile of areal mean temperature anomalies (°C) shows in 75°–90°E, 15°–25°N, and c Solar heating rate (°C \({\bullet\;\mathrm{day}}^{-1}\), shaded) and anthropogenic aerosols concentration (\({\mathrm{mg}\bullet \mathrm{m}}^{-3}\), contours at interval 0.2). A confidence level greater than 90% (shaded) use the Student’s t-test

Full size image
Table 2 Ensemble-mean AOD, surface temperature, and surface energy fluxes averaged within Indian region (70°–90°E, 5°–35°N) and pre-monsoon season for AERO and AERO-CTL
Full size table

From the (AERO–CTL) differences in aerosols and solar heating (Figs. 9 and 10c), we further examine the differences in regional circulation in Fig. 8. During the pre-monsoon period, anthropogenic aerosols accumulate mostly to the east of India (Fig. 9), and the ascending motion over India Peninsula tends to shift eastwards (Fig. 8d). On the other hand, the aerosols are transported by the southwestlies and accumulate over the foothills of the Himalayas, heating the atmosphere by absorbing solar radiation (Fig. 10c). Elevated ascending motions over the southern slopes of the Tibetan Plateau are thermally driven by the heated atmosphere (Fig. 8e, f). The differences of horizontal and vertical flow (AERO–CTL) show a strengthening and southward expansion of upward motion from the Ganges basin to the Indian Peninsula within 70°–90°E (Fig. 8e, f), consistent with the elevated heat pump mechanism suggested in Lau et al. (2006). But the aerosols also result in zonal changes in circulation, i.e. a zonal shift of − ω in lower-tropospheric wind within 10°–20°N over the India Peninsula (Fig. 8d) and strengthened subsidence (or weakened ascending motion) west and east of the positive − ∆ω at 500 hPa over India and Tibet within 70°–90°E. Can the aerosol-induced heating and circulation changes in the pre-monsoon period explain the variability of monsoon onsets as shown in the frequency distributions of ISM onset dates and change rate of zonal wind direction in onset in the 25 years of AERO and CTL (Fig. 5)? Here we examine the co-variability of the onset date and change rate of zonal wind direction with the 500 hPa ascending motion above TP (25°–35oN, 75°–100oE) in pre-monsoon season in each of the 25 years in AERO and CTL (Fig. 11). The TP ascending motion is significantly correlated with the onset date of ISM (correlation coefficient − 0.71 in AERO and − 0.8 in CTL) with stronger TP ascending motion correspond to earlier onset. The TP ascending motion varies in a wider range in AERO (0.02 to 0.23 P s−1) than in CTL (0.02 to 0.14 P s−1) that is quasi-linearly related to the onset date that varies from + 13 to − 9 day in AERO and + 3 days to − 4 days in CTL and the change rate of onset that varies from 5.5 to 1 in AERO and 3 to 0.2 (m s−1) in CTL. So the more scattered and abrupt onset can be explained by the elevated aerosol heating over TP that altered the pre-monsoon circulation that further affect the onset. This is consistent with the conclusion that the elevated heating can change the march of ISM (Lau et al. 2006).

Fig. 11
figure 11

Correlation of 500 hPa TP ascending motion with (a) onset date and (b) ISM change rate over the pre-monsoon season (March 1 to April 10) in AERO (red dots) and CTL (blue dots). The correlation coefficient of TP ascending motion with onset date is − 0.71 (− 0.8) in AERO (CTL) and with change rate is 0.69 (0.78) in AERO (CTL)

Full size image

5 Effect of anthropogenic aerosols on developed monsoon climate

The regional circulation in the developed monsoon period of AERO (May 26 to August 3) as discussed in Sect. 3.2 (Fig. 6e) transports aerosols from Arabian Peninsula across the Arabian Sea into Indian Peninsula. As a result, the horizontal distribution of AOD in AERO (Fig. 12) shows a higher amount of aerosol over the west coast, northern India and IGP, a lower amount of aerosol over southern India and the northern Bay of Bengal due to wash out by monsoon rainfall. The ensemble-mean fields of horizontal winds and − ω at 850 hPa and 500 hPa and the vertical-meridional cross-section of mean − ω within 70°–90°E in the developed monsoon period of AERO (May 26 to August 3) (Fig. 13) show an overall ascending motion over India subcontinent and strong cyclonic westerly monsoon flow from Arabian Sea to Bay of Bengal.

Fig. 12
figure 12

Ensemble mean aerosol optical depth (AOD, shaded) and horizontal wind at 850 hPa (\({\mathrm{m}\bullet \mathrm{s}}^{-1}\), vector) in developed monsoon period of AERO (May 26 to August. 3). The AOD differences (AERO–CTL) > 0 are shown by red contours with contour interval 0.02 begin from 0

Full size image
Fig. 13
figure 13

Ensemble mean fields in developed monsoon period of AERO (May 26 to August 3) a − ω (shaded) and horizontal wind (\({\mathrm{m}\bullet \mathrm{s}}^{-1}\), vector) at 850 hPa, b − ω and horizontal wind at 500 hPa, c 70°–90°E averaged − ω

Full size image

To investigate the anthropogenic aerosol-induced changes, we first show the difference (AERO–CTL) fields of regional circulation and precipitation in Fig. 14. The cyclonic westerly monsoon flow is weakened as revealed by the anticyclonic flow (Fig. 14a, b) and reduced rainfall (Fig. 14d) over India and Bay of Bengal. The weakened monsoon is associated with NW–SE oriented bands of vertical motion: descending motion within 10°–25°N, ascending motion within 0°–10°N, and localized ascending and descending motion along the southern slope of Tibet centered within 20°–35°N (Fig. 14a–c).

Fig. 14
figure 14

Ensemble mean differences fields (AERO-CTL) during developed monsoon period (May 26 to August 3) of − ω are shown by contours (− ω < 0 with dashed line, > 0 with thick solid line) with interval 0.04 and horizontal wind (m s−1, vector) at (a) 850 hPa, and (b) 500 hPa, (c) 70°–90°E averaged − ω. A confidence level greater than 90% (shaded) use the Student’s t-test. d The differences of precipitation (mm) are shown by shaded with 850-hPa horizontal winds (m s−1, vector). e The vertical profile of areal mean temperature anomalies (°C) shows in 75°–90°E, 15°–25°N, and f the solar heating rate anomalies (°C day-1, shaded) are shown with anthropogenic aerosols concentration anomalies (\({\mathrm{mg}\bullet \mathrm{m}}^{-3}\), contour); each contour is 0.2 \({\mathrm{mg}\bullet \mathrm{m}}^{-3}\)

Full size image

Based on the difference field of 850 winds (Fig. 14a, d), higher AOD over the Tibetan peripheral as revealed in the difference distribution of AOD (AERO–CTL) in Fig. 11 corresponds to northward and eastward flow at the northern portion of the anticyclonic flow over India and Bay of Bengal. The high aerosol concentration in the vertical-meridional cross-section of the corresponding difference fields within 70°–90°E coincides with solar heating above the Himalaya slope and top within 25°–35°N (Fig. 14f). Compared with the pre-monsoon period, the areas with increased anthropogenic aerosol accumulation are more concentrated on the southern slope of the plateau. Even if the washout effect occurs during the monsoon season, the daily anthropogenic aerosols that are continuously emitted into the atmosphere will still accumulate through the northern part of the anticyclonic flow and continue to provide a heat source toward TP for atmospheric heating over the southern slope, thereby strengthening the upward motion. In addition, a corresponding increased descending motion above the Indian Peninsula over 10°–20°N (Fig. 14b, c) suppresses monsoon precipitation over the region (Fig. 14d), leading to a slight increase in solar heating rate (Fig. 14f).

Combining the horizontal and vertical-meridional distribution of difference fields of aerosols and circulation, the anthropogenic aerosol-induced changes in the developed monsoon appear to consist of two parts. One is an anthropogenic aerosol-induced weakened monsoon over India and the surrounding tropical oceans due to the reduced land–ocean contrast by aerosol forcing (Fig. 14e). The other part is the elevated heating and ascending motions over the southern slopes of the Tibetan Plateau due to accumulated aerosol heating (Fig. 14c, f).

6 Effect of anthropogenic aerosols on monsoon withdrawal and post-monsoon climate

For the monsoon withdrawal period, we choose the four-week period centered at the withdrawal date as the transition period that is further divided into two: the two weeks before the withdrawal date as Phase − 1 and the two weeks after as Phase + 1. The Phase − 1, withdrawal date, and Phase + 1 correspond to (Oct. 3 to Oct. 18), October 20, and (Oct. 22 to Nov. 6) for CTL, and (Sep. 23 to Oct. 8), October 10, and (Oct. 12 to Oct. 27) for AERO, respectively.

The ensemble-mean AOD, 850 hPa winds, and precipitation in the Phase − 1 and Phase + 1 of AERO and CTL are shown in Figs. 15 and 16. The figures show a clear weakening and southward shift of the cyclonic Indian monsoon flow and rainfall, and a strengthening of the subtropical high over Indo-China from Phase − 1 to Phase + 1. The above circulation transition leads to a change of low-tropospheric winds in India and Arabian Sea from westerly to easterly and the consequent transports of aerosols. These changes in transition are better revealed by the difference fields of precipitation and winds (Phase + 1 minus Phase − 1) in Fig. 17a. The difference fields show anticyclonic flow and suppressed rainfall within 10°–30°N, cyclonic flow and enhanced rainfall within 0°–10°N. Persistent easterlies reside within 10°–20°N between the two zonal bands of anticyclonic and cyclonic flow.

Fig. 15
figure 15

Aerosol optical depth (AOD, shaded) and 850 hPa horizontal circulation (\({\mathrm{m}\bullet \mathrm{s}}^{-1}\), vector) in withdrawal period (September 23 to October 27) of AERO. The Phase − 1 and the Phase + 1 represent the 2 weeks before and after the withdrawal date of ensemble-mean fields. Red contour means AOD anomalies (AERO–CTL) > 0, each contour is 0.02

Full size image
Fig. 16
figure 16

Precipitation (mm, shaded) and 850-hPa horizontal winds (\({\mathrm{m}\bullet \mathrm{s}}^{-1}\), vector) averaged in a CTL Phase − 1 (Oct. 3 to Oct. 18), b CTL Phase + 1 (Oct. 22 to Nov. 6), c AERO Phase − 1 (Sep. 23 to Oct. 8), and d AERO Phase + 1 (Oct. 12 to Oct. 27). Black dash line equals to 2 mm, and red dash line equals to 9 mm

Full size image
Fig. 17
figure 17

Difference fields of precipitation (mm, shaded) and 850-hPa horizontal winds for a Phase + 1 minus Phase − 1 in AERO, and b withdrawal (Phase − 1 to Phase + 1) differences (AERO minus CTL)

Full size image

The change in aerosol transport and the weakening rainfall washout result in more anthropogenic aerosol accumulation over the IGP and westward expansion of AOD over the Indian Peninsula from Phase − 1 to Phase + 1 (Fig. 15). The change of AOD distribution in the transition phases also shows in the AOD differences (AERO–CTL), i.e. the contours and shadings in Fig. 15 resemble each other, indicating the changes are primarily anthropogenic origin.

The influences of anthropogenic aerosols on the monsoon transition are shown by the difference fields of precipitation and winds (AERO minus CTL) for Phase − 1 and Phase + 1 (Fig. 17b). The figure shows an anticyclonic flow from Indo-China to eastern India within 10°–25°N and cyclonic flow over the eastern Indian Ocean within 0°–10°N. The aerosol-induced changes in Fig. 17b largely resemble the transition changes east of 80°E in Fig. 17a. This is attributed to the earlier withdrawal in AERO relative to CTL. Such an anthropogenic aerosol influence on monsoon evolution is consistent with Chen et al. (2018) who attributed post-monsoon precipitation reduction over South Asia to aerosols.

7 Circulation changed influence on natural dust distribution in monsoon evolution

In our experiments, we apply the same dust emission data (The year 1850) in both CTL and AERO so that the desertification effect is excluded to simplify the experiment. Even the anthropogenic aerosols are the only variation in our study, dust emission can be changed by surface wind speed in the model, leading to differences of dust distribution by circulation change induced by anthropogenic aerosols. We evaluate the influences of dust distribution under anthropogenic-aerosol-induced circulation change in simulations. The dust AOD differences with a confidence level greater than 90% are shown in Fig. 18 and with 850 hPa wind field differences. In the pre-monsoon period, the dust AOD increases in the northwest side of the anthropogenic-aerosol-induced anomalous anticyclone above the Arabia Sea, inducing the transportation of dust from middle Arabia to Pakistan. The dust AOD increases about 20% in middle Arabia and southeastern Pakistan (Fig. 18a). More natural dust is transported along the topography into the Ganges Valley from the Middle East under anthropogenic aerosol-induced circulation change during the pre-monsoon period. For differences of aerosols in the Indian Peninsula (70°–90°E, 5°–35°N), the concentration of BC increases 60%, OC increases 17%, sulfate increases 7% and dust also increases 16% in total aerosols (Table 3). Natural dust can heat the atmosphere and cool the surface by radiation absorption which is similar to the effect of BC and OC. Calculating the differences AOD of dust and anthropogenic aerosols, the dust contributes 26% in increased total AOD over Indian Peninsula and 32% in northern India (Table 4). As a result, the local climate change in the Indian pre-monsoon period may have about 1/3 of the contribution by dust. The anthropogenic-aerosol-induced anticyclonic flow over the Indian Peninsula results in suppressed conditions and weakened monsoon that causes a decreased transportation of dust from Arabia in the developed monsoon period (Fig. 18b). In the Indian developed monsoon period, the part of changes in the dust is negligibly small, the dust concentration differences decrease 0.34% (Table 3), and the corresponding AOD differences decrease 3% (Table 4). Indeed, the anthropogenic aerosols dominate the local climate change by radiation budget changes as discussed in Sect. 5. In the first half of the withdrawal period (Phase − 1), the dust AOD increases 21% over the western Arabian Sea and the southern coast of the Arabian Peninsula. The increased dust AOD is limited to the west of the sea by the anomalous easterly wind over the Bay of Bengal and southern India in 10°–20°N (Fig. 18c). In the second half of the withdrawal period (phase + 1), the differences of dust AOD are only located over the Middle East and the Arabian Peninsula because of the anticyclonic anomalous above India and strengthened easterlies over 10°N (Fig. 18d). The Indian aerosol concentration differences of dust in the atmosphere increases 0.37% in the Phase − 1 and decreases 0.34% in the Phase + 1 (Table 3). For the total AOD differences over India, the dust AOD differences contribute 6% in the Phase − 1 and − 9% in the Phase + 1 (Table 4). According to Wei et al. (2022), the decreased dust burden by the easterly wind anomalies is a result of the meridional heating contrast due to reduced (increased) BC burden over northern (southern) India in April and May. In our study, the meridional asymmetrical heating of the atmosphere in India is opposite to Wei et al. (2022). The atmospheric heating by anthropogenic aerosols in the north of 20°N is greater and higher than the south (Fig. 10b) by topography effect that cause an opposite pressure gradient force in our study. The westerly wind anomalies which is opposite to Wei et al. (2022) bring more dust from the Middle East in pre-monsoon period. In developed monsoon period, the atmospheric heating concentrates in the southern slope of TP and the overall India is suppressed (Fig. 14a–c) that the opposite anomalous meridional heating of atmosphere disappeared. The suppressed phase extends to withdrawal period that cause the weakened transportation of dust from the Middle East to India during developed period to withdrawal. As a result, the anthropogenic aerosols may have a major effect on local climate over India from developed period to withdrawal in this study. However, the natural dust coming from the Arabian Peninsula is also important to the local climate change over India in the pre-monsoon period, even if the anthropogenic-aerosol-induced circulation change influences more on the distribution of dust over upstream of dust transportation.

Fig. 18
figure 18

Differences of dust AOD (contour, ∆ < 0 with dashed line, ∆ > 0 with thick solid line) and 850-hPa horizontal winds (\({\mathrm{m}\bullet \mathrm{s}}^{-1}\), vector) averaged in a pre-monsoon (March 1 to April 10), b developed monsoon (May 26 to August 3), c withdrawal Phase − 1 (Sep. 23 to Oct. 8), and d AERO Phase + 1 (Oct. 12 to Oct. 27). The Contour plot with interval 0.02, and significant ΔAOD with a confidence level greater than 90% using the Student’s t-test

Full size image
Table 3 Concentration differences of each aerosol species; BC, OC, Sulfate and dust in atmosphere averaged within Indian region (70°–90°E, 5°–35°N) with percentage of differences
Full size table
Table 4 AOD differences of dust and anthropogenic (Anthro) aerosols averaged within Indian region (70°–90°E, 5°–35°N) and northern India (70°–90°E, 20°–35°N) with percentage of differences
Full size table

8 Summary and discussions

In this study, NCAR CAM 5.1 is used to study the influences of aerosol direct effect on regional monsoon evolution by the control (CTL) and aerosol (AERO) experiments that differ by annual mean distribution of anthropogenic emissions over India Peninsula within (70°–95°E, 5°–35°N) with an areal average of 1.77 Mg m−2 month−1. The two experiments are integrated for 30 years subject to the same perpetual SST annual cycle. The model reasonably simulates the overall monsoon evolution except the positive bias in rainfall over the northern Ganges–Brahmaputra basin, stronger westerly monsoon over tropical western Indian Ocean and the downstream rainfall near western India.

Our analysis of the differences of the two experiments shows that the transition of the ISM from spring to fall in AERO exhibits a similar ensemble-mean onset date with larger spread and more abrupt change, and an earlier but more gradual withdrawal relative the ISM evolution in AERO. To further summarize the effect of aerosols on the overall ISM evolution in study, we note that previous studies have provided a background knowledge that anthropogenic emissions over India alter the radiative flux distribution causing atmospheric heating and surface cooling that leads to a general tendency of weaker ISM. But the ISM responses to aerosol radiative forcing are further changed by advection, topography, and rainfall washout.

The AOD differences between the two experiments (AERO–CTL) show a common feature of aerosol accumulations in Ganges–Brahmaputra basin over 20° to 30°N from spring to fall. The difference fields show additional regions of high AOD in the three stages of monsoon evolution: eastern India and neighboring Bay of Bengal (10°–20°N) in the pre-monsoon period; northeast India in the developed monsoon; northeast India and neighboring Arabia Sea in the period after monsoon withdrawal (post-monsoon).

In the pre-monsoon environment, aerosols lead to stronger upward motion from Tibet through the Ganges basin to the Indian Peninsula within 70°–90°E along with strengthened anticyclonic circulation over the Arabia Sea and eastern Bay of Bengal and a zonal shift of lower-tropospheric wind within 10°–20°N. The strengthened anticyclonic caused an increased natural dust transport from the Middle East. Furthermore, those changes in circulation are well correlated with the dates and change rates of ISM onset, i.e. the stronger TP ascending motion in the pre-monsoon period corresponds to earlier and more abrupt onset. This causal correlation supports the explanation that the aerosol-induced elevated heating over Tibet causes enhanced meridional overturning circulation that further causes more spread in ISM onset dates and more abrupt transitions in ISM onset. Other factors like intraseasonal variability can also influence the ISM evolution to cause more scattered onsets. These possible causes await further studies.

In the developed monsoon period, aerosols result in weakened cyclonic westerly monsoon flow and rainfall over India and Bay of Bengal but a stronger NW–SE oriented band of ascending motion within 0°–10°N (ITCZ). Aerosols also lead to localized ascending along the southern slope of Tibet around 75°–85°E and 30°N that corresponds to high concentrations of aerosols and solar heating. But the localized ascending motion is sandwiched by descending motions along NW–SE direction. These features suggest the two aerosols-ISM interaction mechanisms. One is a weakened land-sea contrast by aerosols over India causing a northward shift of monsoon flow such that westerly within 10°–20°N is weaker but the tropical ITCZ (0°–10°N). The other is the aerosol heating from the lower troposphere over central India to the atmosphere above the southern slope of the Tibetan Plateau causing a local meridional circulation that adds to the suppressed monsoon and rainfall over southern India.

The influences of aerosols on monsoon withdrawal feature a tendency to form anticyclonic flow from Indo-China to eastern India within 10°–25°N and cyclonic flow over eastern Indian Ocean within 0°–10°N. The aerosol-induced changes resemble the transitional changes in monsoon withdrawal over eastern India and Bay of Bengal. This helps to explain the earlier monsoon withdrawal in AERO relative to CTL, consistent with the finding in Chen et al. (2018).

Overall, the increased anthropogenic aerosol emissions lead to a weakened land-sea contrast and a more stable lower troposphere by aerosol-radiative forcing in India. Furthermore, the aerosol-induced elevated heating over Tibet causes local circulation changes in both pre-monsoon and developed monsoon periods. The differences in regional circulation in the two monsoon periods is assumed to the responses to the elevated heating in the different mean flows. The elevated heating also affects the timing and change rate of ISM onset.

We only show experiments with aerosol direct effect in this report although we did carry out experiments including aerosol indirect effect on Indian monsoon that is not reported here because of large uncertainties in the treatment of aerosol-cloud interactions in the model. With only aerosol direct effect in the model, our results summarized above are in general agreement with those studies using coupled ocean–atmosphere climate models with more careful treatment of aerosol forcing (e.g. Ramanathan et al. 2005).

Similar to Lau et al. (2006), the SST annual cycle is prescribed in the model and our experiments show aerosols lead to stronger upward motion over the Ganges basin and the southern slop of Tibetan Plateau by accumulation and induce the meridional overturning that is similar to the results in Lau et al. (2006) and Lau and Kim (2006) who suggested the EHP theory as the cause. Note however that the spatial and temporal distribution of the meridional overturning circulation over Indian subcontinent in pre-monsoon period in this study bears notable differences from the results in Lau et al. (2006) as discussed in Sect. 4. A likely cause is the different treatment of aerosol forcing that is motion dependent in our study but fixed in Lau et al. (2006).

9 Open research

ERA-interim and GPCP data for 1979–2020; monthly sea surface temperature (SST) climatology.

The observation and reanalysis data sets were downloaded from the following sources: Global Precipitation Climatology Project (GPCP) merged analysis of pentad precipitation (1979–2016; https://ftp.cpc.ncep.noaa.gov/precip/GPCP_PEN/; accessed 2 January 2019); NOAA CDR OLR (1979–2016; https://www.ncei.noaa.gov/data/outgoing-longwave-radiation-daily/access/; accessed 24 March 2019); NCEP/DOE atmospheric reanalysis (1979–2016; ftp://ftp.cdc.noaa.gov/Datasets/ncep.reanalysis2.dailyavgs/pressure/; accessed 2 January 2019).

The NCAR Community Earth System Model was downloaded from the following site: https://www.cesm.ucar.edu/models/cesm1.0/cam/

The greenhouse gas concentration at the level of Year 1850, and the annual mean surface aerosol emissions by Year 1850 data (Lamarque et al. 2010).