All the ground-based measurements reported in this paper were made with Cimel Electronique CE-318 Sun–sky radiometer, which is a part of the AERONET global network. This instrument works on the principle of measuring the solar radiation intensity at some specified wavelengths and converts them into optical depth by knowing the corresponding intensities at the top of the atmosphere (Holben et al., 1998). The Cimel Sun photometer is a solar-powered, hardy, robotically pointed Sun and sky spectral radiometer. A microprocessor computes the position of the Sun based on time, latitude and longitude, which directs the sensor head to within approximately 1∘ of the Sun, after which, a four-quadrant detector tracks the Sun precisely to a programmed measurement sequence. After the routine measurement is completed, the instrument returns to the “park” position, awaiting the next measurement sequence. A “wet sensor” exposed to precipitation will cancel any measurement sequence during rain events. Data in the memory of the Sun photometer are transferred to the PC and are sent to NASA for processing and archival of products. The columnar precipitable water content is determined based on the measurements at 940 nm (H2O absorption peak) and at 1020 nm (no absorption by water content). The details of the retrieval of water content and the types of errors involved in it can be found in the works of Schmid et al. (2001), Devara et al. (2001), Smirnov et al. (2004) and Alexandrov et al. (2009). The “Almucantar” scenario denotes a series of measurements taken at an elevation angle of the Sun for specific azimuth angles relative to the position of the Sun. These measurements are taken at 440, 675, 870 and 1020 nm channels. During Almucantar measurements, observations from a single channel are made in a sweep at a constant elevation angle across the solar disk and continue ±180∘ from the Sun in about 40 s. This is repeated for each channel to complete an almucantar sequence (Holben et al., 1998).

The calibration of the instrument was performed regularly at the Goddard Space Flight Center (GSFC) by a transfer of calibration from reference instruments that were calibrated by the Langley method at Mauna Loa Observatory (MLO), Hawaii. The combined effects of uncertainties in calibration, atmospheric pressure and total ozone amount (climatology is used) result in a total uncertainty in derived aerosol optical depth of ∼0.010–0.021, with the largest error in the UV (Eck et al., 1999). The AERONET algorithms compute the aerosol optical depth retrievals in near-real time (Holben et al., 1998). Data are quality checked and cloud screened following the methodology of Smirnov et al. (2000), which relies on the greater temporal variance of cloud optical depth vs. aerosol optical depth. The AERONET archive is divided into three quality levels: level 1.0 for raw data, level 1.5 for cloud-screened data and level 2.0 for quality-assured data (Holben et al., 1998; Smirnov et al., 2000) and can be downloaded from the AERONET website (https://aeronet.gsfc.nasa.gov/, last access: 13 June 2019). The AERONET data (from January 2005 to December 2015) used in this study belong to version 2.0 at level 2.0, which are cloud screened and have been manually screened, ensuring the data quality (Giles et al., 2019). A number of studies have already described the instrumentation, data acquisition, retrieval algorithms and calibration procedures, which conform to the standards of the AERONET global network, as well as the uncertainty in final products and the applied cloud-screening procedures (Holben et al., 1998, 2001; Eck et al., 1999; Smirnov et al., 2000, 2002a, 2002b). More details about AERONET instrument used in the present study in Pune, India, have been reported by Kumar et al. (2011).

In the present study, the aerosol radiative forcing (ARF) at bottom of the atmosphere (BOA), in the atmosphere (ATM) and at the top of the atmosphere (TOA) has been estimated by following the method reported by Xie et al. (2018). This method involves forcing calculations from broadband fluxes between 0.2 and 4.0 µm from aerosol size distribution (ASD), spectral AOD, single scattering albedo and phase function by using the radiative transfer module integrated in the AERONET inversions (García et al., 2008, 2012). In this method, the AERONET-defined RFTOA and RFBOA values have been directly used to calculate RFATM=RFTOA-RFBOA.

The Moderate-resolution Imaging Spectroradiometer (MODIS) is a scientific instrument (radiometer) aboard the NASA Terra and Aqua satellite platforms. The Terra and Aqua satellite platforms were launched in 1999 and 2002, respectively, to study global dynamics of the Earth's atmosphere, land, ice and oceans. Thus, the instrument collects a variety of global datasets. The Terra and Aqua satellites, with MODIS instruments attached, fly on Sun-synchronous orbits at 705 km altitude and pass over the same spot of the Earth at about the same local time every day, i.e., 10:30 LT in the case of Terra and 13:30 LT for Aqua. Due to the large swath of data collected by MODIS (over 2300 km wide), it is possible to observe almost the entire Earth's surface every day. MODIS measures reflected solar and emitted thermal radiation in a total of 36 bands ranging in wavelength from 0.4 to 14.4 µm and at varying spatial resolutions (2 bands at 250 m, 5 bands at 500 m and 29 bands at 1 km). Detailed descriptions of the MODIS aerosol retrieval and its evolution since the start of MODIS operation are given in the work of Remer et al. (2005). In this study, the Deep Blue product of AOD at 550 nm and water vapor observations was used.

The Ozone Monitoring Instrument (OMI) measures the solar radiation backscattered by the Earth's atmosphere and surface over the entire wavelength range from 270 to 500 nm with a spectral resolution of about 0.5 nm. OMI measurements are highly synergistic with the other instruments on the Aura platform. The 114∘ viewing angle of the telescope corresponds to a 2600 km wide swath on the surface, which enables measurements with a daily global coverage. OMI continues the Total Ozone Mapping Spectrometer (TOMS) record for total ozone and other atmospheric parameters related to ozone chemistry and climate. Also, total columns of gases like NO2, BrO and SO2 will be derived. The US Environmental Protection Agency (EPA) has designated these atmospheric constituents as posing serious threats to human health and agricultural productivity. These measurements are made at near-urban-scale resolution and track industrial pollution and biomass burning. Furthermore, aerosol and cloud parameters will be determined from the OMI measurements. The OMI instrument is a contribution of the Netherlands's Agency for Aerospace Programs (NIVR) in collaboration with the Finnish Meteorological Institute (FMI) to the Aura mission. In this study, we used the OMAEROev003 product of AOD at 442 nm measurements for comparison with ground-based observations of AERONET data during the study period.

The amount of solar radiation trapped in the atmosphere by aerosols, as quantified by the atmospheric heating rate (HR; K d-1), has been analyzed. The HR due to aerosol absorption is calculated according to the first law of thermodynamics and assuming hydrostatic equilibrium, as suggested by Liou (2002): ∂T∂t=gCpARFATM∇P, where ∂T∂t is the HR in K d-1, g is the acceleration due to gravity, Cp is the specific heat capacity of air at constant pressure (i.e., ∼1006 J kg-1 K-1), and ∇P is the atmospheric pressure difference between the surface and 3 km altitude, where most aerosols are present.

The discrimination of aerosol types increases accuracy of the assessment of the aerosol radiative impact and is therefore important to climate modeling (Diner et al., 1999). Previous studies showed that different aerosol types have different effects on climate because their diverse morphology, size distribution, hygroscopic properties and chemical components will lead to different aerosol optical properties (Giles et al., 2011, 2012; He et al., 2018; Kumar et al., 2018). For example, dust aerosols are often large particles and have a scattering tendency (Vijayakumar et al., 2014), whereas black carbon aerosols are usually smaller particles and have an absorbing nature (Tan et al., 2016). Various optical and microphysical parameters have been used for aerosol classification. The spectral dependence of AOD, expressed by the Ångström exponent (AE), is a good indicator of particle size. These two parameters are commonly used in aerosol remote sensing to infer dominant aerosol types given knowledge of the source region or typical aerosol transport mechanisms (Kaskaoutis et al., 2009; Pathak et al., 2012; Vijayakumar and Devara, 2013; Verma et al., 2015). Gobbi et al. (2007) used this parameter and its spectral curvature to propose a graphical method for evaluating the contribution of fine-mode particles to AOD and to track mixture of pollution containing dust. Schuster et al. (2006) determined that variation in AE wavelength pairs is sensitive to aerosol composition. In addition, the spectral AOD and AE data were used in deriving the curvatures (a1 and a2) correlated with AOD and can be effectively used for discriminating different aerosol types (Vijayakumar et al., 2012; Kumar et al., 2013) information content from these relationships varies from generic identification of major aerosol particle types (e.g., dust mixed and urban industrial pollution) to specific degrees of absorbing aerosols. In the Indian subcontinent, Kaskaoutis et al. (2009) have made the first attempt to distinguish different aerosol types originating from variety of sources over Hyderabad. Furthermore, Vijayakumar et al. (2012) over Pune and Vijayakumar and Devara (2013) over Sinhagad have discriminated different aerosol types over different regions in the Indian land mass.

For the classification of aerosols into specific types, some “appropriate” threshold values are required. Our approach to infer major aerosol types is based on the combination of particle size and absorption–scattering information, following Kaskaoutis et al. (2009), Vijayakumar et al. (2014) and Lee et al. (2010). The AOD–AE patterns have been utilized to describe different aerosol types (e.g., clean maritime, continental clean, desert dust, urban industrial/biomass burning and mixed-type aerosol). In this study, we used AOD at 440 nm and an Ångström exponent of 440–870 nm. The threshold values used in this study represented a slight adjustment from those previously used by authors, as cited above. Cases of AOD440nm<0.2 with AE440-870nm values of small (<0.9) or large (>1.0) represented clean maritime-influenced aerosols (CM) and continental clean type represented as background aerosols (CC), respectively. AOD440nm>0.3 and AE440-870nm>1.0 can be used to characterize long-range-transported biomass burning/urban industrial aerosols (BB/UI) and AOD440nm values >0.6 associated with AE440-870nm values <0.7 are indicative of desert dust (DD) particles transported over oceanic areas. Finally, the aerosols which do not fall under any of the above categories (types) are considered as mixed aerosol type (MA), considering different effects of aerosol-mixing processes in the atmosphere (e.g., coagulation, condensation, humidification, gas-to-particle conversion). The classification method (AOD440nm vs. AE440-870nm) is illustrated in Fig. 3a.

The Cimel Sun–sky radiometer derived fine-mode fraction (FMF) at 500 nm has been used to represent the dominant aerosol size mode, because FMF provides quantitative information for both fine- and coarse-mode aerosols. SSA has been used to quantify the aerosol absorption–scattering at 440 nm, which is the shortest wavelength of AERONET channels and generally used to distinguish absorbing from non-absorbing aerosols (Lee et al., 2010). The categories of aerosol types are distinguished between FMF500nm and SSA440nm is polluted dust (PD, dominantly dust with anthropogenic aerosols), polluted continental (PC, dominantly anthropogenic mixed with dust), absorbing and non-absorbing aerosols. The threshold values of groups are as follows: (i) FMF500nm<0.4 for any range of SSA440nm indicates predominantly coarse-mode aerosols and hence is assigned to PD, (ii) introducing a safety margin of 0.4 between fine- and coarse-mode aerosol, 0.4≤FMF500nm≤0.6 for any range of SSA440nm represented PC. All FMF500nm>0.6 were considered as fine-mode aerosols which were further discriminated as absorbing (mainly black carbon, BC) or non-absorbing depending on values of SSA440. (iii) FMF500nm>0.6 with SSA440nm>0.95 as non-absorbing (NA, which includes sulfates, nitrates and aged water-soluble organic carbons), (iv) FMF500nm>0.6, 0.90<SSA440nm≤0.95 as slightly absorbing (SA), (v) FMF500nm>0.6, 0.85<SSA440nm≤0.90 as moderately absorbing (MA), whereas (vi) FMF500nm>0.6 with SSA440nm≤0.85 represented highly absorbing (HA) fine-mode aerosols. It should be noted that this classification must be considered only qualitatively not quantitatively, since the percentages may be strongly modified with a change in the AOD440nm, AE440-870nm, FMF500nm and SSA440nm threshold values. The method of classification (SSA440nm vs. FMF500nm) is illustrated in Fig. 3b.

In order to obtain a holistic view of the regional aerosol characteristics during the period of study, the seasonal variations in aerosol quantities are grouped and shown in Fig. 7. The different panels in the figure display variations in different aerosol products. Different panels in this figure indicate variations in aerosol optical depth (Fig. 7a); water vapor (Fig. 7b); Ångström exponent (Fig. 7c); total, fine-mode and coarse-mode fractions (Fig. 7d); aerosol size distribution (Fig. 7e); single scattering albedo (Fig. 7f); asymmetry factor (Fig. 7g); aerosol refractive index, real (Fig. 7h); refractive index, imaginary (Fig. 7i); effective radius (Fig. 7j); and volume-weighted mean radius (Fig. 7k).

Figure 7a displays the seasonal variation of AOD at different wavelengths. At all wavelengths, AODs show lower values in the winter season, which could be due to strong inversions, and whatever aerosols due to various human activities (domestic cooking, vehicular and industrial emissions, etc.) are let out into the surface layer get trapped in the lower atmosphere due to less ventilation.

Also, due to calm wind conditions, aerosols of soil-dust type are less common during winter. In the pre-monsoon, the greater AOD values be attributed to the increased concentration of continental aerosol loading due to higher temperatures in the study region. Moreover, processes such as long-range transport are decreasing from winter to pre-monsoon which may be associated with air masses consisting of dust from west of the Arabian Sea (Devara et al., 2002; Kumar et al., 2011). In addition, Vijayakumar and Devara (2012) reported that the anthropogenic aerosols are abundant in the pre-monsoon season could also be due to festive celebrations. In India, a festival called “Holi” is celebrated every year during the pre-monsoon season. More popularly, this festival is also known as the “Festival of Colors”. During this festival, people play with colored powder and paste. The associated Holika Dahan (fire burning) activities will enhance the particle concentration and alter chemistry of the local environment. Higher aerosol loading during the monsoon period at all wavelengths was noticed in the present study due to paucity of useful data points and prevalence of turbid atmospheric conditions.

Figure 7b shows the seasonal variation in water vapor. The monthly mean values of the water vapor varied from 1.39 to 3.98 cm, representing a minimum value in January (1.39±0.27) and maximum in July (3.98±0.23). It is observed that water vapor is lower during the months of January and February. It starts increasing from March with the onset of the hot summer season and reaches a maximum during the months of the southwest monsoon season (June–September). However, the number of days of observations during June–September is lower because of a smaller number of clear-sky/partly clear-sky days. Therefore, the relative magnitude of water vapor during these 4 months could be partly due to sampling bias. However, the average value of water vapor during the monsoon season is still significantly higher when compared to other seasons. Water vapor starts decreasing once the monsoon season ends in September. Lower values are observed in winter, slightly increasing up to monsoon season, and thereafter water vapor decreases due to monsoon rains and decreased aerosol input due to colder ground surface. The lower possibility of hygroscopic growth of aerosols due to low water vapor content may also contribute to this. On a seasonal scale, the increase water vapor from the winter (December–February) to the pre-monsoon (March–May) season is about 39.72 % and from the pre-monsoon to monsoon (June–September) seasons is as high as 80.95 %.

Figure 7c illustrates the mean seasonal variation in Ångström exponent. The Ångström exponent (α440-870) is determined from the spectral dependence of the measured optical depth and is a measure of the relative dominance of fine (submicrometer) aerosols over the coarse (micrometer-size) aerosols (Kumar et al., 2011) as shown in Fig. 7d. Higher values of α indicate the dominance of fine particles, whereas lower values indicate the dominance of coarse particles and relatively less concentration of fine particles. It is expected that when the aerosol particles are very small, on the order of air molecules, α440-870 should approach 4, and it should approach 0 for very large particles (Holben et al., 2001). The decrease of α from winter to pre-monsoon and monsoon months is indicative of increased coarse-mode particle contribution, consistent with dust particles (>10 µm) in the aerosol loading. This strong decrease in the α440-870 value could be due to mixing of air originating from oceanic and desert regions. The notable lower values of the Ångström exponent during June–September are ascribed to cloud contamination of data retrievals caused by thin invisible cirrus (Chew et al., 2011; Huang et al., 2012).

Figure 7e portrays mean seasonal variation in aerosol size distribution. It is clear from the figure that during the monsoon season, coarse-mode volume concentration is higher than the fine mode, which suggests that there is dominance of coarse-mode aerosol particles over the site due to the monsoon activities that start during this season and also due to local meteorological conditions, land-surface and long-range transport processes. Tripathi et al. (2005) found that there was an increase of 50 % in volume concentration in coarse mode over the Indian region during the monsoon season. On the other hand, there is a very small variation in mean radius of fine-mode aerosol particles in comparison to coarse mode during post-monsoon and winter seasons, which suggest that there may be different types of sources of coarse-mode particles. The errors associated with the particle retrieval in the size range (0.1≤r≥7 µm) do not exceed 10 % in the maxima of the size distribution and may increase up to about 35 % for the points corresponding to the minimum values of dV(r)/dln⁡r in this size range (Dubovik et al., 2002).

Figure 7f depicts the seasonal dependence of single scattering albedo (SSA) during different seasons. SSA is calculated from the scattering optical thickness which is obtained from the normalized aerosol phase function using diffuse radiance measured at different angles. The detailed method of determining the SSA was given by Dubovik et al. (1998). SSA has nearly a unit value for purely scattering aerosols (e.g., sulfate aerosols) and has low value for strongly absorbing aerosols (e.g., black carbon and/or mineral dust). The mean seasonal variation of SSA at different wavelengths during the study period shows increase in SSA from winter to monsoon due to the dominance of anthropogenic aerosols in the atmosphere. It can also be seen that during post-monsoon and winter seasons, SSA decreases with wavelength due to dominance of absorbing aerosols over the experimental site, which is attributed to presence of a mixture of aerosols from multiple sources like vehicular and industrial pollution, and biomass burning in the field. But in the pre-monsoon season, SSA values are slightly higher at 1020 nm, compared to 440 nm. This suggests that the dominance of dust and marine events, thus enhancement in the scattering contribution of coarse particle. The atmosphere contains more water-soluble particles, in conditions like those found by Singh et al. (2004) in Kanpur, India. In addition, for locations closer to the ocean, the air is more humid during summer, leading to enhanced water uptake of the water-soluble particles fraction. But from monsoon to post-monsoon, the SSA values are slightly low, which suggests a possible combination of urban industrial particles and sea salt aerosols over the region.

Figure 7g depicts the seasonal dependence of the asymmetry parameter (AP) during different seasons. Theoretically, the range of AP lies between -1 (for the backward scattered radiation) and +1 (for the forward scattered radiation). However, the zero value represents symmetric scattering. There is relatively little variation in AP among these four seasons at shorter wavelengths (440 nm), while AP was larger at 440 nm in the monsoon season compared to other wavelengths (675, 870 and 1020 nm). The greater relative contribution of coarse-mode particles to the aerosol size distribution might have resulted in a phase function shift toward greater forward scattering at the longer (infrared) wavelengths and very little change in the shorter (visible) wavelength. The AP is also wavelength dependent and varies from 0.71±0.01 to 0.65±0.04 during the whole period.

Figure 7h–i depict the wavelength dependence of real and imaginary parts of refractive index (RI), respectively, during different seasons. The refractive index is a complex quantity, expressed in terms of real n(λ) and imaginary k(λ) parts, which provide an indication of highly scattering or highly absorbing types of aerosols, with higher n(λ) values corresponding to the scattering type and higher k(λ) values corresponding to the absorbing type (Sinyuk et al., 2003). Real and imaginary parts, n(λ) and k(λ), are not independent of SSA and the retrieved size distribution of the aerosols in the region, but some differences in trends may be observed because of the presence of different types of aerosols (Dubovik et al., 2002). The useful information about the RI comes from aureole radiance, which is strongly affected by errors in the angle-pointing bias. The errors are estimated to be 30 %–50 % for the imaginary part and ±0.04 for the real part of the RI (Dubovik et al., 2002). These estimated errors are for high aerosol loading (AOD440nm≥0.5) at a solar zenith angle >50∘.

The seasonal variation of real part of RI shows increase in the real part of values with an increase in wavelength, while n(λ) is highest (>1.52) at all wavelengths during the pre-monsoon season in the years 2005–2015, showing the higher scattering optical state of the atmosphere during this period. This is also supported by the higher SSA in the years 2005–2015. However, lower n(λ) values during the post-monsoon season could probably be associated with higher relative humidity and resultant hygroscopic growth, similar to the conditions found over Goddard Space Flight Center (Dubovik et al., 2002). During the pre-monsoon season, n(λ) values at higher wavelengths are close to the n(λ) values of dust (1.53) found from several models (WMO, 1983; Koepke et al., 1997), clearly indicating the contribution of dust to the optical properties. However, n(λ) during the monsoon and winter seasons has intermediate values. The lowest value in post-monsoon is due to the anthropogenic activities. The real part of refractive index of dust aerosol is usually greater than that of the anthropogenic aerosols (Alam et al., 2011).

The imaginary value is found to be higher in winter and lower in the monsoon season, with higher values relating to absorbing anthropogenic aerosols and the lower values to dust aerosols. The imaginary part is highest during December, which shows that anthropogenic aerosols are dominant during this period. The higher imaginary part values at the two shortest wavelengths (440 and 670 nm) are attributed to the absorption of organic carbon/black carbon (Arola et al., 2011). On the contrary, the imaginary part values are lower in the monsoon season. This suggests that dust aerosols are dominant during the monsoon season, similar to the conditions found in the Kanpur region, India (Singh et al., 2004).

Figure 7j–k indicate seasonal variation in effective radius (Reff) and volume-weighted mean radius (Rv), respectively. Effective radius (Reff) is quite representative of the optical properties of coarse-mode particles, whereas for fine particles, volume-weighted mean radius (Rv) is the more appropriate parameter (Tanré et al., 2001). Reff is found to be higher during the pre-monsoon and monsoon seasons. The high values in Reff during the monsoon season are attributed to the abundant transport of aerosols of natural origin and also the surface-level anthropogenic aerosols, which increases the loading of coarse-mode particles (Vijayakumar et al., 2012). The increase and decrease in Reff during post-monsoon and winter seasons, respectively, are interesting. Although no significant coarse-mode particle loading takes place during the post-monsoon season and the hygroscopic growth of these particles is unlikely, the effective radius of coarse-mode particles may increase if the fine-mode particles get attached to the surface of the coarse-mode particles. However, Reff has not been found to increase significantly, because it is more influenced by the number concentration compared to the volume concentration. The increase in Rv is attributed to the hygroscopic growth of the fine-mode particles. The seasonal variability in coarse-mode particles (Reff) is found to be greater than that of fine-mode particles (Rv) with a maximum during the monsoon season due to changes in circulation, land-surface and long-range transport processes.

Figure 8 shows the pie diagram of different aerosol types over Pune using above threshold values. The analysis reveals that the mixed-type (MT) aerosols clearly dominates in the pre-monsoon (47.07 %) and monsoon (56.82 %) seasons and in some fractions (40.55 %, 39.97 %) during winter and post-monsoon seasons, while the BB/UI type exhibits its highest presence during the post-monsoon (31.55 %) and more rarely (5.18 %) in the monsoon season. It is interesting to note that CM aerosols are absent during the post-monsoon period. Polluted continental aerosols are dominant in the pre-monsoon (16.04 %), compared to the monsoon (8.46 %) and post-monsoon (6.64 %) seasons due to rainout/washout effects.

The absorbing aerosols have much lower contribution in the monsoon season. Overall, the MT (44.85 %) aerosols are dominant compared to other types of aerosols over study region. Furthermore, the total contribution of various types of aerosols is as follows: CM (1.02 %), CC (0.47 %), BB/UI (22.57 %), DD (2.27 %), PD (1.94 %), PC (10.36 %), SA (2.7 %), MA (5.56 %), HA (7.93 %) and NA (0.33 %). However, it should be noted that the present fractions may be considered rather qualitatively and not quantitatively, since they correspond to specific threshold values, while an alteration in the thresholds would cause changes in the fractions but not in the general pattern of the seasonal distribution of aerosol types. A comparison of the present results of Pune with those obtained from similar analyses performed over other locations on the globe is not an easy task due to different land-use and environmental characteristics as well as differences in the time periods of the observations and in the seasonal pattern of the air masses and long-range transport, the specific influence of anthropogenic pollution, etc. The aerosol types inferred from the present study are compared with those obtained from other Indian sites in Table 1.

Aerosols modify incoming solar and outgoing infrared radiation. The ARF at TOA or BOA is defined as the difference in the net solar fluxes (down minus up) (solar plus longwave; in W m-2) with and without aerosol. The difference between these two quantities gives the ARF in the whole atmosphere. Generally, negative values of TOA, implying the presence of aerosols, result in an increase in the radiation loss to the space (by enhanced backscattering), leading to a cooling in the Earth–atmosphere system, while positive values imply an atmospheric warming. At BOA, the forcing will always be negative because aerosols reduce the surface reaching solar radiation, and these values are more sensitive to aerosol loading only. The difference between the radiative forcing at TOA and BOA is defined as ATM. It represents the amount of energy trapped within the atmosphere due to the presence of aerosols. If ATM is positive, the aerosols produce a net gain of radiative flux to the atmosphere leading to a heating (warming), while negative ATM indicates loss and thereby cooling. Generally, the intensity of ARF depends on the aerosol loading, and therefore it is difficult to consistently intercompare the radiative forcing by the typical aerosol types. Seasonal variations of radiative forcing at TOA, BOA and ATM observed in the present study are compared with such studies available at other Indian sites in Table 3.

The monthly average ARF variations at the TOA, BOA and within the ATM during the study period are plotted in Fig. 14. The decreasing trend of radiative forcing in the ATM and at TOA reveals dominance of hygroscopic (scattering) particles, whereas the increasing trend of radiative forcing at BOA indicates the dominance of hydrophobic (absorbing) aerosol particles. The BOA forcing is found to be negative, while ATM is positive for all months. These interactions between aerosols and solar radiation can be attributed to combination of aerosol properties (i.e., types), surface properties (i.e., albedo) and geographical parameters (latitude, season) (Yu et al., 2006). The large difference between TOA and BOA forcing demonstrates that solar radiation is being absorbed within the atmosphere, and as a result the atmosphere gets warmer, but the Earth's surface gets cooler (Alam et al., 2011; Kumar and Devara, 2012b). This can substantially alter the atmospheric stability and influence the dynamic system of the atmosphere (Li et al., 2010). For the ARF during winter, the BOA is more strongly negative associated with the corresponding TOA, giving rise to the highest ATM. In contrast to this, during pre-monsoon and post-monsoon seasons, ATM appears to be reduced with reduced heating rate. ATM forcing will increase if aerosol forcing at the TOA is more toward the positive side, while there is a large negative forcing at surface level. But TOA shows almost minor variation between positive and negative.

The ARF for the whole observation period at the TOA is in the range of +11 to -46 W m-2 (average -17±10 W m-2) and at the BOA from -32 to -152 W m-2 (average -82±19 W m-2), increasing the ATM forcing from +15 to +149 W m-2 (average +64±25 W m-2). But for the radiative forcing during winter, the BOA (-73.3±18.2 W m-2) is more strongly negative, associated with corresponding TOA (-14.8±9.5 W m-2), giving rise to the highest ATM value of 58.4±15.5 W m-2 with a resulting heating rate of 1.95±0.5 K d-1 during this season. In contrast to this, during pre-monsoon and post-monsoon seasons, ATM appears to be reduced with the reduced heating rates of 1.7±0.5 and 1.8±0.5 K d-1.

Figure 15 depicts the comparison between long-term climatology of AOD and water vapor datasets from AERONET observations with MODIS and OMI observations. Here, the MODIS satellite gives AOD at 550 nm, while AERONET data had no corresponding wavelength to match it exactly. So, the AERONET AOD at 550 nm was obtained by interpolation between 440 and 675 nm following the Ångström equation. All figures clearly show an increasing trend except the water vapor parameter. Figure 15b clearly shows that AERONET AOD observations were relatively higher overestimations compared to OMI AOD. The possible reasons for relatively lower correlation observed between the AODs recorded by AERONET and OMI have been explained. Figure 15c shows that the AERONET H2O trend is higher compared to the satellite H2O trend, but magnitude-wise both are following similar trends. Figure 16 shows the correlation between AERONET observations and satellite observations. They show a strong correlation between satellite observations and AERONET data (Fig. 16a, c). A weak correlation (Fig. 16b) was observed between OMI AOD442nm and AERONET AOD440nm. Apart from a small sensing wavelength difference of 2 nm, higher AOD values by AERONET and relatively lower values by OMI have also been reported by Humera et al. (2015) due to anthropogenic activity and biomass burning. Now the correlation has been improved by reducing the scatter (eliminating the significant out-layered data points) between the observations.