For our study, we use the regional climate model version 4 (RegCM4) coupled with an atmospheric chemistry and aerosol module. RegCM4 is a limited-area model with a hydrostatic dynamical core and sigma-pressure vertical coordinates (Giorgi et al., 2012). The Coordinated Regional Downscaling Experiment (CORDEX) South Asia domain is used in this study (Fig. 1) with 50 km horizontal grid spacing, 18 vertical levels and a model top at 50 hPa (Giorgi et al., 2012; Nair et al., 2012; Usha et al., 2020). The meteorological initial and lateral boundary conditions for our simulations are provided by the ERA-Interim reanalysis (Dee et al., 2011) for the three winter seasons (December to February) from November 2014 to March 2017, where the first month of simulation (November) of each year is discarded from the analysis as model spin-up. Optimum-interpolation weekly sea surface temperature data from the National Oceanic and Atmospheric Administration (NOAA) are used as lower boundary conditions over the ocean. The parameterization schemes used in the simulations are (1) for boundary layer, the University of Washington (UW) planetary boundary layer scheme; (2) for convection over land and ocean, the Tiedtke scheme; (3) for radiative transfer, the Community Climate Model (CCM3) scheme; (4) for cloud microphysics, a subgrid explicit moisture scheme; and (5) for land surface, the Biosphere–Atmosphere Transfer Scheme. More details on the model configurations and physics have already been discussed in earlier papers (Ajay et al., 2019; Giorgi et al., 2012; Usha et al., 2020). Note that in the validation and analysis of the model output we always consider the average over the three winter seasons unless otherwise specified.

RegCM4 has an online chemistry module, which is extensively used for understanding aerosol–climate interactions (Nair et al., 2012; Shalaby et al., 2012; Solmon et al., 2015). The aerosol and trace gas emission fluxes are adopted from the IIASA dataset, and chemical boundary conditions are from the global model MOZART (Model for Ozone and Related chemical Tracers). Gas-phase chemistry is based on the CBMZ (Carbon-Bond Mechanism version Z) scheme and ISORROPIA II is used for inorganic aerosols (Shalaby et al., 2012). The aerosol scheme includes sulfate, nitrate, ammonia, sea salt (two size bins), mineral dust (four size bins), black carbon (hydrophilic and hydrophobic) and organic carbon (hydrophilic and hydrophobic). The sources, sinks, atmospheric processes and transport of the aerosol species are detailed in Solmon et al. (2006) and Zakey et al. (2008). The mass concentration of each aerosol species is converted into optical properties using a mass extinction cross section, which depends on the ambient relative humidity and species-specific hygroscopic-growth functions (Solmon et al., 2006). We use the growth function described by Kiehl et al. (2000) for sulfate, nitrate and ammonia aerosols, in which the growth function increases exponentially with humidity. For example, at 80 % relative humidity, the particle mass extinction cross section increases by a factor of 3 compared to its dry value. The model emits carbonaceous aerosols as hydrophobic, and the optical properties of these carbonaceous aerosols are invariant with relative humidity. These particles change from hydrophobic to hydrophilic due to ageing at a fixed timescale of 1.15 d (∼ 27.6 h; Solmon et al., 2006). The hydrophilic part of the carbonaceous aerosols (black carbon and organic carbon) has a weak affinity to water compared to sulfate aerosols (Kasten, 1969). At 80 % relative humidity, the growth factor is 1.37 and 1.49 for hydrophilic black carbon and organic carbon, respectively. Mineral dust is considered fully hydrophobic. The sea salt size increases by 2 times at 80 % relative humidity compared to in dry conditions, but this is of little importance over the IGP during winter. In general, most of the aerosols grow with relative humidity, and this enhances the optical depth.

RegCM4 has been widely used to investigate aerosol–hydroclimate interactions and the impacts of various forcings on monsoon characteristics (e.g. Solmon et al., 2015, and Usha et al., 2020). The model can be run with and without radiative feedbacks of aerosols. In the former case, aerosols affect the radiation balance and change the surface temperature and meteorology through direct and indirect pathways. In the latter, the aerosol radiative forcing is calculated but it does not have effects on the meteorology, surface temperature and atmospheric thermodynamics. In the present study, model simulations with (ambient) and without (dry) hygroscopic growth of aerosols are named as experiments 1 and 2, respectively. Experimental details are summarized in Table 1. In the case of the dry-aerosol simulations, we forced the aerosol hygroscopic-growth functions to be equal to 1 in the model code. The model runs without meteorological feedbacks (Exp1 and Exp2) are considered control runs for the ambient- and dry-aerosol cases. Since the effect of aerosol–radiation interactions on land and meteorological parameters are switched off (no meteorological feedback), the difference between these two experiments provides the contribution of hygroscopic growth to the aerosol properties (ΔAODRH): 1ΔAODRH=AODAmbientnofeedback-AODDrynofeedback. Experiments 3 and 4 are similar to Exp1 and Exp2 but with meteorological feedbacks (Table 1). The difference between Exp3 and Exp4 provides the net effect of the hygroscopic growth of aerosols on the regional climate, which includes (i) changes due to the radiative forcing of the aerosol hygroscopic growth and (ii) its meteorological feedback. For example, the change in temperature (T) due to ΔAODRH and its feedback is estimated as 2ΔTRH=TAmbientfeedback-TDryfeedback. The effect of the total aerosol system on meteorological variables (e.g. temperature) is estimated as 3ΔTaerosol=TAmbientfeedback-TAmbientnofeedback. The ΔTRH is a subset of ΔTaerosol.

The change in AOD solely due to the meteorological feedback of the total aerosol system (dry + hygroscopic growth) is estimated as 4ΔAODfeedback=AODAmbientfeedback-AODAmbientnofeedback. Hence, the AOD(Ambientfeedback) is the sum of dry AOD, its hygroscopic growth and the meteorological feedback: 5AODAmbientfeedback=AODDrynofeedback+ΔAODRH+ΔAODfeedback.

Several earlier studies have used RegCM4 to simulate the meteorology of South Asia (Ajay et al., 2019; Giorgi et al., 2012; Nair et al., 2012; Usha et al., 2020). In general, the model captures the basic features of seasonal precipitation (monsoon), though it is sensitive to the convection schemes used and the extent of the ocean–atmosphere coupling (Ajay et al., 2019; Di Sante et al., 2019). Studies on the validation of chemical constituents simulated by the model are rather limited over South Asia (Nair et al., 2012; Solmon et al., 2015; Usha et al., 2020). A qualitative inter-comparison of the chemical composition of aerosols simulated by RegCM4 with in situ measurements at distinct locations over the Indian region is shown in Fig. 2. The measurements of columnar aerosol optical depth are taken from AERONET (Aerosol Robotic Network) radiometer observations at Jaipur, Kanpur, Gandhi College and Delhi. The black carbon measurements were carried out using Aethalometers installed under the ARFI (Aerosol Radiative Forcing over India) project. Since the measurements of aerosol composition are sparse over the Indian region, we have taken values reported from earlier studies for the winter season, and therefore the organic-carbon and sulfate measurements are for different years than those of the model simulations (Ali et al., 2019; Aswini et al., 2019; George et al., 2008; Ram et al., 2010, 2012; Rengarajan et al., 2011; Safai et al., 2008; Satsangi et al., 2012). Despite these shortcomings, compared to earlier studies, the present model configuration simulates black carbon mass loading and AOD which are closer to the observed values (Nair et al., 2012). Though RegCM4 has a simple scheme for organic aerosols, the organic carbon mass concentration broadly matches the observed seasonal mean values. In fact, most climate and chemical transport models fail to capture the high aerosol loading over this region, a problem which has been mostly attributed to the parameterization of stable boundary layer conditions (Nair et al., 2012), unaccounted for emissions (e.g. Nair et al., 2012, references therein), and biases in simulating relative humidity and precipitation (Chatani and Sharma, 2018; Feng et al., 2016). In the present study, AOD and the near-surface mass concentrations of black carbon, organic carbon, sulfate and dust simulated by the model are in the range of values reported by several studies over the region. One of the main challenges in simulating the AOD over the region is the accurate simulation of relative humidity, which affects hygroscopic growth (Chatani and Sharma, 2018). The mean relative humidity estimated from several reanalysis datasets and measurements in different locations of the IGP varies from 60 % to 80 % (Chatani and Sharma, 2018; Gautam et al., 2007; Ghude et al., 2017; Goswami and Sarkar, 2015). A good agreement between measured and simulated (with aerosol feedback) relative humidity over the IGP (Fig. 2e) is important for the estimation of aerosol optical properties from simulated speciated aerosol mass concentrations. The RegCM4-simulated relative humidity values are approximately 10 % higher than the measured RH, with a mean absolute error of 6.6 % and root mean square error (RMSE) of 7.3 %. During winter, aerosols are confined within the boundary layer, and changes in the boundary layer height play a major role in the dilution (ventilation) of aerosols over the IGP (Bharali et al., 2019; Nair et al., 2012). Shallow and stable boundary layers prevail during the simulated winter seasons. A comparison with IGRA (Integrated Global Radiosonde Archive) radiosonde data indicates that the UW scheme is able to simulate the boundary layer height over the region reasonably well (Fig. 2f) with a mean absolute error of 200 m (26 % of mean boundary layer height) and RMSE of 240 m. Further intercomparison of modelled and measured PM2.5, AOD and black carbon (BC) mass concentration at various locations over the Indian region is shown in Figs. S1, S2 and S3 in the Supplement.

The particulate mass loadings (PM2.5) measured over most cities in the IGP during winter are well above the air quality standards, as shown in Fig. 1. The total column AOD at ambient humidity simulated by RegCM4 with meteorological feedbacks shows very high values over the IGP, mostly consisting of boundary layer aerosols. The comparison of measured and simulated AOD for dry and ambient humidity conditions (Fig. 3) over the distinct environments of the IGP indicates an almost 2-fold increase in AOD due to the hygroscopic growth of aerosols and its meteorological feedback. The mean observed AODs (simulated ambient and dry AODs) for Jaipur, Kanpur, Pune and Gandhi College are 0.53 (0.51, 0.26), 0.83 (0.81, 0.34), 0.48 (0.32, 0.15) and 1.15 (0.92, 0.36), respectively. In general, both the magnitude and large day-to-day variability in AOD (Fig. 3) are captured by the model when the hygroscopic growth (ΔAODRH) and meteorological feedback (ΔAODfeedback) are included in the simulation. Except for dust and freshly emitted black carbon, most of the aerosol species have an affinity towards water vapour, with the hygroscopic growth of inorganic aerosols such as sulfate, ammonia, nitrate and sea salt being more pronounced compared to organic carbon (Solmon et al., 2006). Due to ageing, a fraction of hydrophobic black carbon also becomes slightly hydrophilic (Solmon et al., 2006) and contributes, though not to a large extent, to the AOD at ambient humidity. Observations close to the Thar Desert (Fig. 1a, Jaipur) show lower hygroscopic growth (43 %) compared to other IGP stations because of the dominance of dust aerosols and relatively low humidity conditions prevailing there.

Our study shows that the large day-to-day and spatial variability in AOD over the IGP is mostly contributed by relative humidity (Fig. 3), with lower contributions from changes in source characteristics and synoptic-scale circulation. The simulated AOD at ambient relative humidity matches well with the measured AOD (mean absolute error of 0.12 and RMSE of 0.15), whereas the dry AOD is significantly lower and less variable in time. The standard deviation of measured and simulated AOD at ambient humidity conditions is nearly half of the mean AODs (coefficient of variation of 50 %; Fig. 3), whereas, for dry aerosols, the standard deviation of the AOD is low (0.05 to 0.1), with a coefficient of variance ranging between 20 % and 30 %. All the meteorological conditions, associated processes (winds, chemistry, deposition and transport) and anthropogenic emissions remained the same for dry and ambient AOD simulations, except the hygroscopic growth of AOD with relative humidity for the ambient AOD. Compared to the dry AOD, the ambient AOD shows large variability associated with the variability in humidity. The AOD due to dry aerosols has 10 % to 15 % less variability compared to that of the total AOD.

The AOD due to hygroscopic growth (ΔAODRH) estimated using Eq. (1) is shown in Fig. 4a. The ΔAODRH values are high over the IGP, especially the eastern IGP and Bangladesh (ΔAODRH>0.3). The ratio of AOD(Drynofeedback) to AOD(Ambientnofeedback) shown as contours in Fig. 4a indicates that dry aerosols contribute 50 % to 60 % of the AOD over most of the Indian subcontinent (Fig. 4a), except over western India, where the contribution of hydrophobic dust to the total AOD is high even during winter. The contribution of the aerosol hygroscopic growth is more than 50 % of the total AOD over the northern Indian Ocean, which is attributed to the dominance of hydrophilic aerosols (sea salt) and high relative humidity. It is interesting to note that the increase in AOD due to hygroscopic growth (ΔAODRH= AOD(ambient) - AOD(dry)) is ∼ 0.23 over the IGP (Fig. 4a), which is higher than the global mean AOD (0.155±0.018) reported by Watson-Parris et al. (2020) using various satellite observations. Mandariya et al. (2020) have reported that wintertime aerosols over the central IGP show an increase in particle diameter of 1.34±0.07 times that of the dry particles at 85 % relative humidity. This increase in particle diameter corresponds to a doubling of the scattering coefficient for anthropogenic aerosols (McInnes et al., 1998). In the present study, the mean dry AOD increases almost 1.7 times due to relative humidity over the IGP, which is slightly lower than the estimated AOD for the hygroscopic factor reported by Mandariya et al. (2020). Since organic aerosols dominate the aerosol system over the IGP (Ram et al., 2012), there could be a substantial decrease in the hygroscopic growth of particles as discussed by several studies. Mandariya et al. (2020) have reported that 150 nm particles are more hygroscopic than 100 nm particles due to the higher organic fraction present in the 100 nm particle mass concentration. In addition, although most observational studies consider 40 % as the reference humidity level (dry) for estimating the hygroscopic growth, the model estimates the dry AOD for zero relative humidity. Titos et al. (2016) reported that the assumption of no growth at 40 % humidity contributes to higher uncertainty in the estimates of measurement-based hygroscopic growth.

In general, the mean AOD over the IGP stations (Kanpur and Gandhi College) is ∼ 0.7, indicating significant surface dimming of solar radiation (Fig. 4b). The increase in AOD due to the hygroscopic growth alone reduces the surface solar flux by 10 to 20 W m-2 over the IGP. Nair et al. (2016a) reported a mean surface forcing of -28.2±12 W m-2 due to the total aerosol system over the entire Indian region during winter. The surface dimming due to aerosol–radiation interactions results in the reduction in surface temperature, heat fluxes (sensible and latent) and boundary layer height (Bharali et al., 2019; Ding et al., 2016; Li et al., 2017). As a result, aerosol forcing leads to the accumulation of pollutants in the boundary layer (as discussed later). Since water vapour contributes significantly to the total AOD, most of the effects of aerosols on regional climate, visibility and air quality are strongly associated with meteorological conditions.

The effects of the increase in AOD solely due to the hygroscopic growth of particles (ΔAODRH) on surface temperature, relative humidity and cloud fraction (Eq. 2) are illustrated in Fig. 5a–c, which shows a surface cooling of about 0.5 ∘C and a 3 % increase in relative humidity over the IGP during winter. This is mostly due to the direct solar dimming at the surface and to a cloud cover change (up to 3 %) induced by the ΔAODRH (Fig. 5c). The total effect of aerosols on temperature, relative humidity and cloud fraction (Eq. 3) is shown in Fig. 5d–f. Nearly 40 % to 50 % of the aerosol-induced surface cooling is contributed by water vapour through the hygroscopic growth of particles. The cloud cover increases by 3 % to 5 % over the IGP due to ambient aerosol forcings (Fig. 5f). These effects are more prominent over the IGP and central India, whereas high ΔAODRH and surface dimming is seen over the eastern IGP (Fig. 4). Although AOD and ΔAODRH are high, the change in relative humidity due to the hygroscopic growth of aerosols (Fig. 5b) is negligible over Bangladesh and the eastern IGP. The presence of relatively high cloud fraction over the eastern IGP mask the aerosol-induced surface cooling which is attributed to the contrast in the spatial pattern of surface temperature and AOD. The change in shortwave flux at the surface shows an almost similar pattern to that of change in temperature (Fig. S4). Bharali et al. (2019) have reported that the strength of land–atmosphere interactions through the exchange of heat and moisture fluxes plays a major role in determining the aerosol-induced dimming effects on meteorology (more details are available in Li et al., 2017). At the surface, an observed decrease in evapotranspiration (and associated rainfall) due to solar dimming further confirms a slowing down of the hydrological cycle induced by the aerosol hygroscopic growth (Liepert et al., 2004; Ramanathan, 2001). In other words, water vapour itself reduces the evaporation through dimming associated with the aerosol hygroscopic growth.

The mean vertical profiles of change in temperature and relative humidity over the IGP due to total aerosol and hygroscopic effects are shown in Fig. 6. Significant cooling and associated moistening are noticeable in the boundary layer followed by a weak drying in the lower free troposphere (1 to 3 km) and a negligible influence above (> 3 km). Based on extensive measurements from Delhi during wintertime fog campaigns, Ghude et al. (2017) have reported high values of relative humidity (70 %) within the boundary layer and dry conditions (< 20 %) above 3 km. A similar pattern is observed for specific humidity, with an increase of 0.6 g kg-1 at the surface, which confirms that the hygroscopic growth is the single largest contributor to the total aerosol effects over the IGP during winter (Fig. S5).

Aerosol-induced surface cooling and weak entrainment of dry air mass from the free troposphere lead to an increase in relative and specific humidity at the surface (Fig. 5a). Based on extensive measurements of aerosol and humidity profiles, Feng et al. (2016) reported up to 30 % underestimation of relative humidity within the boundary layer and 20 % overestimation in the free troposphere by the WRF-Chem model over a north Indian station (Nainital) during the winter season. They cited this dry bias in boundary layer humidity as one of the major reasons for the underestimation of AOD by climate models over the region. Most reanalysis datasets (National Centers for Environmental Prediction – NCEP – and European Centre for Medium-Range Weather Forecasts – ECMWF) have a negative bias (underestimate) in simulating relative humidity over this region, especially during winter (Chatani and Sharma, 2018; Feng et al., 2016), which significantly affects the simulation of aerosol optical depth, aerosol radiative forcing and fog prediction. The present study shows that the negative bias (underestimation) of relative humidity could be reduced, especially at high relative humidities (RH > 70 %), by including the meteorological feedback of ambient aerosols (Exp4). In contrast to boundary layer moistening, drying in the lower free troposphere (1.5 to 3.5 km) by 1 % to 2 % was attributed to the aerosol-induced warming, which decreases humidity and suppresses mixing of air mass between the free troposphere and boundary layer. Compared to dry conditions, the warming of the top of the boundary layer occurs at a lower altitude for ambient aerosol forcing.

We analysed reanalysis data and in situ observations for the long-term trend in wintertime relative humidity within the boundary layer and free troposphere over the IGP during the last 4 decades (Fig. 7). In contrast to the global scenario, relative humidity increased over the Indian region at the rate of 1 % per decade, together with a concurrent increase in aerosol loadings (Babu et al., 2013) and the number of foggy days (Ghude et al., 2017; Syed et al., 2012). Based on the above discussions (Fig. 6), it can be argued that the aerosol forcing contributes significantly to the observed increasing trend in wintertime moistening of the boundary layer and enhanced occurrence of fog events over the IGP. The aerosol radiative forcing increases the stability of the atmosphere and decreases the transport of moisture from the boundary layer to the free troposphere, which favours the drying trend in the lower free troposphere shown in Fig. 7b. Earlier studies have also attributed the humidification of the boundary layer and drying of the free troposphere during winter to the aerosol forcing (Li et al., 2017; Tie et al., 2017). Our analysis shows that aerosol feedback processes significantly increase the near-surface relative humidity due to weak turbulent diffusivity in the stratified boundary layer (Bharali et al., 2019; Li et al., 2017). The increase in relative humidity, when surface temperature decreases, is due to the increase in water vapour content in the atmosphere as seen in the specific humidity trend reported by Mukhopadhyay et al. (2017). In contrast to the negative trend in relative humidity in the lower free troposphere during winter (Fig. 7b), Mukhopadhyay et al. (2017) show a positive trend for annual mean humidity, which is largely dominated by the positive trend during summer. Global analyses have shown that most continental regions exhibit a long-term warming trend and decrease in relative humidity (Dai, 2006), which has strong implications for the land–ocean warming contrast (Hodnebrog et al., 2019). The long-term trend in the moistening of the boundary layer and drying of the free troposphere over the Indian region during winter is further amplified by the aerosol radiative forcing, primarily through the hygroscopic growth. However, the masking effects of surface warming by aerosols and the contribution of aerosols to the increasing trend in humidity are yet to be quantified.

A regional mean AOD increase of ∼ 0.23 is contributed by the hygroscopic growth of aerosols (Fig. 3), which further increases the AOD through meteorological feedbacks (ΔAODfeedback; Fig. 8a), whose contribution can be estimated using Eq. (4). Overall, the hygroscopic growth (ΔAODRH∼ 0.23) and associated meteorological feedback (ΔAODfeedback∼ 0.17) contribute significantly to the total AOD (0.72) over the IGP. This positive feedback is stronger over the central IGP than the eastern IGP, though the eastern IGP has higher AOD and relative humidity. The moisture content in the atmosphere increases the AOD directly by its hygroscopic effect (Fig. 3) while high AOD increases the surface relative humidity through its radiative and meteorological feedbacks (Fig. 8a). Even though the spatial patterns of change in AOD and relative humidity due to aerosol feedbacks are slightly different, the anomalies in AOD and relative humidity show a significant correlation over most of the Indian subcontinent, especially over the IGP (Fig. 8b). Since the aircraft and lidar observations over the IGP clearly showed that aerosols are mostly confined within the boundary layer during the winter season with lower contributions from the free tropospheric aerosols (Nair et al., 2016b), we have used AOD and surface-level parameters like PM2.5, and visibility to investigate the implications of aerosol-induced forcing on air quality and visibility.

The frequency of occurrence of AOD and PM2.5 for dry and ambient humidity conditions with meteorological feedback over the IGP is shown in Fig. 9. The inclusion of meteorological feedback and hygroscopic aerosol growth (ΔAODRH) produces an increase in AOD along the IGP (Fig. 5). The narrow frequency distribution for dry AOD broadens substantially when including the effect of ambient humidity, which implies an increase in the number of days with hazy skies, high AOD and significant dimming at the surface. The frequency distribution of the ratio of AOD for ambient and dry humidity with and without meteorological feedback is shown in Fig. 9b. The no-feedback simulation shows a mean ratio of 1.72, which increases to 2.3 at ambient humidity with meteorological feedbacks. Due to the positive feedback as mentioned above, AOD at ambient humidity increases to more than 3 times its dry AOD, depending on the strength of the land–atmosphere and aerosol–boundary layer interaction. The magnitude of the meteorological feedback due to the total aerosol system (ΔAODfeedback) is comparable to ΔAODRH over the IGP. This clearly indicates that not only the hygroscopic growth of aerosols but also the meteorological feedback is important over the IGP. Therefore, the aerosol feedbacks on regional meteorology need to be investigated further to better understand and predict the fog events over the IGP (Bharali et al., 2019). Although the dominance of organics in the aerosol mass loading over the IGP is expected to decrease the hygroscopic growth of pure inorganic aerosols (Mandariya et al., 2020), the meteorological feedback processes (Eq. 3) increase the AOD by moistening the boundary layer (ΔRHaerosol). The frequency of occurrence of high values of PM2.5 mass concentration for ambient humidity shows deterioration of air quality due to aerosol–radiation interaction, and the aerosol forcing itself increases the aerosol loading through aerosol–boundary layer interaction.

In fact, generally calm winds (< 2 m s-1), low temperatures, a shallow boundary layer and a slow descending motion prevail over the IGP during winter, which favours the accumulation of aerosols within the boundary layer (Ghude et al., 2017; Kumar et al., 2015a; Nair et al., 2007). In addition, the aerosol effects further weaken the horizontal and vertical circulations and decrease the boundary layer height (Fig. S6), which results in more accumulation of aerosols and thus poorer air quality (Bharali et al., 2019; Ding et al., 2016). Although wintertime aerosols are mostly confined within the boundary layer, near-surface PM2.5 and columnar aerosol loading show poor correlation over the IGP as reported by Kumar et al. (2015a). Our study also shows a poor correlation between ambient AOD and PM2.5; however, a high correlation is found between the dry AOD and PM2.5. This result has implications for the estimation of PM2.5 from satellite observations or for finding an empirical relationship between AOD and PM2.5 for air quality applications since the high relative humidity causes the association between AOD and PM2.5 to deteriorate. By and large, the offline filter-based PM2.5 measurements have issues associated with sampling aerosols at ambient humidity conditions. This also points to the need for dedicated field experiments focusing on the size-segregated hygroscopic-growth functions of physical and optical properties of aerosols over the IGP (Mandariya et al., 2020).

As shown in Fig. 10, there also exists an enhancement in near-surface aerosol mass loading (carbonaceous, inorganics and PM2.5) due to the suppression of the boundary layer height induced by the meteorological feedbacks from the aerosol–radiation interaction. This results in further degradation of the air quality of this region. The aerosol concentration increases along the IGP due to aerosol-induced weak ventilation. Figure 11 shows the estimated visibility using the aerosol extinction coefficient. Due to high aerosol loading and high humidity conditions, visibility is low over the region. Compared to the dry aerosols, there is a 60 % reduction in visibility because of the aerosol hygroscopic growth (Fig. S7). In addition, the enhancement in relative humidity due to aerosol forcing again decreases the visibility. This analysis further confirms that the air quality problems in the IGP are associated not only with the aerosol source strength but also with the ambient humidity conditions and their (aerosol) radiative forcing and meteorological feedback.

As discussed above, the AOD increase via hygroscopic growth of aerosols under high-relative-humidity conditions further decreases the incoming solar radiation at the surface, which results in enhanced surface cooling (compared to dry aerosols) and a decrease in the exchanges between the boundary layer and free troposphere. Aerosol–radiation interactions lead to a significant increase in cloud cover over the IGP region (2 % to 8 %), which is further increased by the hygroscopic effect of aerosols. The increase in cloud cover leads to more sulfate formation and thus again influence the radiative balance. Hence, the hygroscopic effects of aerosols significantly strengthen the observed aerosol–boundary layer interactions over the region (Bharali et al., 2019). Recent studies have shown that the high aerosol concentrations at the surface over urban centres are strongly related to positive-feedback processes associated with the boundary layer and water vapour (Ding et al., 2016; Huang et al., 2014; Li et al., 2017; Tie et al., 2017). The surface solar dimming and atmospheric warming due to black carbon in the upper boundary layer decrease the height of the mixed layer (increasing stratification) and increase the accumulation of aerosols within the boundary layer (Bharali et al., 2019; Ding et al., 2016; Nair et al., 2007). The aerosol-induced increase in relative humidity (1 %–10 %) over most land regions (Fig. 5), especially the central IGP, further increases the AOD. Tie et al. (2017) have reported that a decrease in the dispersion of water vapour leads to a self-amplifying feedback mechanism through which an increase in relative humidity further increases the AOD due to hygroscopic growth (ΔAODRH). All these pathways further increase moisture in the boundary layer and, consequently, the AOD (Tie et al., 2017). These feedback processes significantly cause air quality conditions during winter to deteriorate (Li et al., 2017).