The mixing state of aerosols over the Indo-Gangetic Plain and its impact on radiative forcing

Authors


Abstract

Seasonal variations in mixing states of aerosols over an urban (Kanpur) and a rural location (Gandhi College) in the Indo-Gangetic Plain (IGP) are determined using the measured and modelled optical properties, and the impact of aerosol mixing state on radiative forcing is examined. Different fractions of black carbon (BC) and water-soluble aerosols in core-shell mixing emerged as the probable mixing state during winter, monsoon and post-monsoon over Kanpur. The degree of mixing, i.e. the percentage mass fraction of aerosols involved in core-shell mixing, is found to exhibit seasonal variations. Owing to the abundance of mineral dust (MD) during the pre-monsoon, MD coated by BC emerges as the most probable mixing state. Top-of-atmosphere (TOA) forcing changes its sign from positive for external mixing to negative for different probable mixing states during the pre-monsoon over both locations, as single scattering albedo is lower for external mixing. However, for other seasons, the TOA forcing is negative for external and different probable core-shell mixing states of aerosols. Surface aerosol forcing for probable mixing state during the post-monsoon is higher (−44 W m−2) over Kanpur, and is lower (−24 W m−2) over Gandhi College. A regression between instantaneous model-derived aerosol forcing and AERONET-measured forcing yielded r2 > 0.9, which confirms the robustness of the methodology adopted to retrieve aerosol optical properties and estimate forcing. Heating rates over Kanpur and Gandhi College during the pre-monsoon and monsoon are ∼0.75 K d−1 and ∼0.5 K d−1 respectively. Differences exist between measured and model-derived asymmetry parameter, g, owing to the non-sphericity of aerosols. However, aerosol radiative forcing is found to be weakly sensitive to the variation in g due to high (> 0.2) surface albedo. The modelling study provides new insights into the state of aerosol mixing, and indicates that aerosol mixing can vary depending on the type and abundance of aerosol species. Copyright © 2012 Royal Meteorological Society

1. Introduction

Aerosols modulate the Earth–atmosphere radiation balance by scattering and absorbing the incoming solar and outgoing terrestrial radiation, and indirectly by influencing the lifetime and albedo of cloud (IPCC 2007). In the atmosphere, different types of aerosol (viz. soil/mineral dust and insoluble organics, water-soluble nitrates, sulphates and organics, black carbon (BC), and sea salt) are present. The presence of different types of aerosol over a location is due to local sources and long-range transport which can result in different mixing states because of aging and interaction among different aerosols. In external mixing, there exists no physical and chemical interaction among the different aerosol species. In core-shell mixing, one type of aerosol (e.g. BC) can get coated by another type of aerosol (e.g. sulphate). In homogeneous internal mixing, all types of aerosol can be mixed together, resulting in an aerosol entity with the same chemical composition throughout the particle. The hypothesis of homogeneous internal mixing is termed unphysical, and reality is said to lie between the externally mixed and core-shell mixed scenarios (Jacobson, 2000). The changes in the size distribution of aerosols, their life cycle and radiative effects due to mixing of aerosols are yet to be explicitly incorporated in the majority of climate models (e.g. IPCC, 2007). Knowledge of the mixing state of aerosols is important for an accurate assessment of aerosols in climate forcing, as assumptions regarding the mixing state of aerosol and its effect on optical properties can give rise to uncertainties in modelling their direct and indirect effects (IPCC 2001; IPCC 2007). In most models, external mixing among the aerosols is assumed, which is the simplest approach from a computational perspective as there are difficulties in realistically determining the degree of the mixing state (e.g. Lohmann et al., 1999). Different types of aerosol mixing states are observed in several field studies and they strongly vary with location and season (e.g. Ramanathan et al., 2001; Hasegawa and Ohta, 2002; Zhang et al., 2003; Mallet et al., 2004; Arimoto et al., 2006). These observations show that in the atmosphere aerosols can exist in external, core-shell and/or internally mixed states.

The climatology and seasonal variability of aerosol optical properties over the Indo-Gangetic Plain (IGP), including the influence of dust storms on the aerosol optical properties, and aerosol radiative effects have been studied earlier (Dey et al., 2004; Prasad and Singh, 2007; Gautam et al., 2009, 2010, 2011; Eck et al., 2010; Giles et al., 2011). In this study we derive the seasonal variation in probable mixing state of aerosols over the IGP in India (Figure 1), which is influenced by anthropogenic and natural aerosols, and show distinct seasonal characteristics (Tare et al., 2006; Dey and Tripathi, 2008). The IGP is surrounded by the Himalayan mountain range to the north, the Thar Desert and the Arabian Sea to the west, the Bay of Bengal to the east and the Vindhyan Satputra ranges to the south (Figure 1). The IGP is one of the most populated and polluted river basins in the world, and rich in fertile lands and agricultural production (Dey and Tripathi, 2008). A map of aerosol optical depth (AOD) at 0.55 µm derived from the MODerate Resolution Imaging Spectroradiometer (MODIS) is shown in Figure 1. MODIS level 3 collection V005 monthly mean Terra and Aqua AOD from January 2007 to December 2009 at 1°×1° latitude–longitude resolution are utilized (Remer et al., 2008), and seasonal means are calculated. AOD is higher over the IGP than the other Indian regions during all the seasons (Figure 1). Industrial growth and increasing use of fossil fuel as well as biomass burning have been suggested as the reasons for higher aerosol loading in the region (Singh et al., 2004; Dey and Tripathi, 2008). Both the anthropogenic pollution and natural aerosols (mostly dust) contribute to the regional aerosol loading in the pre-monsoon season (Figure 1).

Figure 1.

Map of aerosol optical depth (AOD) at 0.55 µm and surface winds over the Indo-Gangetic Plain during (a) winter, (b) pre-monsoon, (c) monsoon and (d) post-monsoon. The shaded contours denote AOD, with surface winds (ms−1) represented by arrows. The study locations Kanpur and Gandhi College are marked. The seasonal mean seven-day wind back-trajectories over (e) Kanpur and (f) Gandhi College for different seasons are also shown. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

The chosen study locations, Kanpur and Gandhi College, represent different environments within the IGP and are governed by different aerosol types (Figure 1). Kanpur is an urban location and is influenced by local anthropogenic pollution and long-range transport, while Gandhi College is a rural location (∼ 250 km from Kanpur) and situated downwind of major urban areas. Due to different environments and availability of various aerosols, the most probable mixing states over these two locations are determined to highlight the seasonal variability and the differences in mixing state of aerosols over this region. The modelled optical properties in conjunction with measured optical properties–aerosol optical depth (AOD), single scattering albedo (SSA) and asymmetry parameter (g)–are used to estimate the probable mixing states over the IGP. In addition, information is utilized on aerosol mass concentration reported only at Kanpur during the winter, pre-monsoon and post-monsoon seasons. In the only previous study, Dey et al. (2008) derived the most probable mixing state over Kanpur during particular months of 2005/06, making the assumption that the total mass concentration (100%) of different aerosol species were involved in core-shell mixing. In the present study, the mass fractions of different aerosol species involved in core-shell mixing are allowed to vary, and the seasonal variations in aerosol mixing state are determined. This is important since observations (e.g. Hasegawa and Ohta, 2002; Arimoto et al., 2006; Vester et al., 2007) show that the total mass of two species need not be completely involved in core-shell mixing, but can have a variable mass fraction of even about 10%. In this study, for the first time, we derive the probable mixing state of aerosols and the seasonal variation over two environmentally different locations, and also discuss the radiative implications. Measurements of the mixing state of aerosols are not available over the IGP at present, but these results are expected to offer new insights into the state of aerosol mixing in this area, where both natural and man-made aerosols coexist and also exhibit significant seasonal variations.

2. Study locations and meteorology

Kanpur (26.51°N, 80.23°E, 123 m AMSL) and Gandhi College (25.87°N, 84.13°E, 60 m AMSL) are situated in the IGP (Figure 1). Kanpur is an urban and industrial city in Uttar Pradesh state and is densely populated (> 4 million), while Gandhi College is a rural village located in the Ballia district southeast of Kanpur. As Gandhi College is situated downwind of major urban areas (viz. Delhi, Kanpur and Lucknow; Figure 1), the atmosphere over this location is influenced by a mixture of rural (local) and urban (downwind and long-range transport) aerosol emissions. The IGP experiences four distinct seasons, winter (December–February), pre-monsoon (March–May), monsoon (June–September) and post-monsoon (October–November). There is no significant seasonal variation in winds over Kanpur and Gandhi College due to their mixed origin over the entire year. The wind speed is < 5 m s−1 except during the monsoon (Figure 1). Winds are mainly northeast or northwest during pre- and post-monsoon seasons, while during the winter and monsoon the winds come from different directions (Figure 1). Seasonal mean seven-day wind back-trajectories arriving at the study locations at an altitude of 500 m show the transport of air parcels from the Arabian Sea and the desert areas during winter and pre-monsoon; during the monsoon the winds are of marine origin (Figure 1). Over Gandhi College the winds are a mixture of southwesterly and easterly during the post-monsoon (Figure 1). However, the winds during this season are mostly local, and speeds are lower. The northwesterly winds during the pre-monsoon transport dust from the Thar desert (e.g. Chinnam et al., 2006). Relative humidity (RH) is highest during the monsoon and lowest during the pre-monsoon season over both locations (Table 1).

Table 1. Seasonal mean and standard deviation of columnar ozone, water vapour and relative humidity over Kanpur and Gandhi College during 2007–2009.
Atmospheric parameterWinterPre-monsoonMonsoonPost-monsoon
Kanpur    
Columnar ozone (DU)258±15280±11274±8261±9
Columnar water vapour (cm)1.4±0.12.2±0.54.9±0.92.2±0.4
Relative humidity (%)41±632±478±2258±10
Gandhi College    
Columnar ozone (DU)258±15282±12277±9260±9
Columnar water vapour (cm)1.7±0.22.7±0.85.5±0.62.9±0.5
Relative humidity (%)46±738±691±570±12

3. Data

Seasonal mean spectral AODs, SSAs and g derived from the monthly level 2.0 quality-assured and cloud-screened Aerosol Robotic Network (AERONET) (Holben et al., 2001) data over Kanpur and Gandhi College are utilized. In the present work, the required seasonal mean aerosol optical properties could not be calculated for a single year over the study locations, because data were not available in the seasons over a whole year. Therefore, monthly AERONET data from January 2007 to December 2009 are utilized, which are then used to calculate seasonal means. The total number of data points over Kanpur and Gandhi College during 2007–2009 were 146 and 181, respectively. The interseasonal (season to season) variation in AOD is found to be higher (≥30%) than the interannual variation ( ∼20%) over Kanpur and Gandhi College in the 0.34–1.02 µm wavelength range. These features are consistent with earlier results over Kanpur using AERONET data; Singh et al. (2004) reported a strong seasonal variability in AODs. The interannual variations in SSA and g are ∼2% and 6% of the corresponding average values respectively over both locations in the wavelength range 0.44–1.02 µm. The interseasonal variation in SSA is similar to interannual variation in SSA, while interseasonal variation in g is higher than interannual variation and ranges between 1 and 17% of their respective seasonal mean values. Since the interseasonal variations in the measured spectral aerosol optical properties over Kanpur and Gandhi College are higher than the interannual variations, the probable mixing states and their radiative effects are determined and reported for the different seasons.

3.1. Aerosol optical depths

The Aerosol Robotic Network (AERONET) sun/sky scanning radiometer measured AODs at seven different wavelengths (0.34, 0.38, 0.44, 0.50, 0.675, 0.87 and 1.02 µm) (Dubovik and King, 2000) over Kanpur and Gandhi College are used. The sun/sky radiometer measures direct and diffuse solar radiances in the wavelength range 0.34–1.02 µm with a field of view of 1.2°. The absolute uncertainty in AOD under clear-sky conditions is < ±0.01 when λ > 0.44µm, and < ±0.02 for shorter wavelengths when AOD at 0.44 µm is less than 0.2 (Smirnov et al., 2000). The relative standard error in estimated AODs based on Smirnov et al. (2000) is found to be < 10% for all the wavelengths for the range of AODs obtained over Kanpur and Gandhi College during the study period.

3.2. AERONET inversion data: SSA and g

SSA and g obtained from AERONET sun/sky measurements at 0.44, 0.675, 0.87 and 1.02 µm over Kanpur and Gandhi College are utilized. SSA and g are retrieved using an inversion algorithm which searches for the best fit of all data to a theoretical model considering the different magnitudes of the accuracy in the fitted data (Dubovik and King, 2000). The uncertainty in the SSA retrieved by AERONET is within 0.03 for high aerosol loading for an optical depth (at 0.44 µm) > 0.5, while error in SSA increases to 0.05–0.07 for lower AOD (Dubovik and King, 2000). The angular distribution of scattered light (described by g) is an integral characteristic of aerosols, and the uncertainty in the retrieval of g from AERONET lies in the range 3–5% (Andrews et al., 2006).

3.3. Vertical profiles of aerosol extinction coefficient

Insufficient information on the vertical distribution of aerosols can introduce uncertainty in the estimation of aerosol radiative forcing and heating rate as a function of altitude (IPCC 2007). As ground-based measurements of simultaneous collocated lidar data are not available over the study locations during the study period, in the present study the aerosol extinction profiles obtained from the CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations) lidar measurements are used. Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) on board CALIPSO has been providing information on the vertical distribution of aerosols and clouds and also on their optical properties on a global scale since June 2006 (Winker et al., 2007). In the present work, level 2 aerosol extinction profile data for 0.532 µm available at a horizontal resolution of 40 km are obtained over the study locations. The measurement uncertainty in CALIPSO-derived aerosol extinction products is reported to be about 40 (http://wdc.dlr.de/sensors/calipso/). Since an exact CALIOP overpass over a particular location is possible only once in about two months, seasonal mean extinction profiles are constructed by averaging all the available profiles from CALIOP tracks inside a 1° × 1° box around the study locations.

4. Additional inputs: atmosphere, ozone, water vapour and surface albedo

The additional inputs (viz. atmospheric profiles of temperature, pressure, columnar ozone, water vapour and spectral surface reflectance) are necessary to perform aerosol radiative forcing calculations. Standard tropical atmospheric profiles of temperature and pressure are used in the present study (McClatchey et al., 1972). Monthly mean columnar ozone and water vapour are obtained from the Ozone Monitoring Instrument (OMI) and National Center for Environmental Prediction (NCEP) reanalyses, respectively. Though surface reflectance is not an aerosol property, it plays an important role in determining the magnitude and sign of aerosol radiative forcing. In the present study, surface reflectance measured by MODIS on board Terra and Aqua satellites (8-day, level 3 global 500 m ISIN grid product, MOD09A1 (Terra) and MYD09A1 (Aqua)) at seven wavelength bands centred at 0.645, 0.859, 0.469, 0.555, 1.24, 1.64 and 2.13 µm are utilized to calculate the Terra-Aqua mean surface reflectance over the study locations. The 8-day surface reflectances obtained from MODIS are used to estimate the seasonal mean surface reflectance over Kanpur and Gandhi College at seven different wavelength bands during 2007–2009 (Table 2). The surface reflectance values are higher during the pre-monsoon and monsoon. The aerosol radiative forcing calculation requires the surface reflectance values as a function of wavelength for the entire short-wave range. Therefore, the spectral surface reflectance is constructed by combining sand, vegetation and water in appropriate proportions such that the resultant spectrum of surface reflectance matches the observed MODIS-derived surface reflectance (Ramachandran and Kedia, 2010), and is used in aerosol radiative forcing calculations.

Table 2. MODIS-derived seasonal mean surface reflectance over Kanpur and Gandhi College during 2007–2009.
Wavelength (µm)WinterPre-monsoonMonsoonPost-monsoon
Kanpur    
0.4690.059 ± 0.0190.081 ± 0.0170.074 ± 0.0280.059 ± 0.009
0.5550.102 ± 0.0140.136 ± 0.0250.129 ± 0.0350.107 ± 0.010
0.6450.108 ± 0.0140.158 ± 0.0340.134 ± 0.0490.115 ± 0.016
0.8590.273 ± 0.0220.284 ± 0.0220.311 ± 0.0340.263 ± 0.019
1.240.277 ± 0.0150.318 ± 0.0320.311 ± 0.0430.290 ± 0.019
1.640.239 ± 0.0160.308 ± 0.0440.277 ± 0.0660.258 ± 0.030
2.130.173 ± 0.0160.243 ± 0.0450.194 ± 0.0750.187 ± 0.035
Gandhi College    
0.4690.034 ± 0.0040.072 ± 0.0230.067 ± 0.0290.046 ± 0.012
0.5550.076 ± 0.0050.118 ± 0.0310.107 ± 0.0340.085 ± 0.011
0.6450.067 ± 0.0080.139 ± 0.0420.099 ± 0.0480.087 ± 0.017
0.8590.303 ± 0.0230.266 ± 0.0280.321 ± 0.0340.265 ± 0.029
1.240.275 ± 0.0190.308 ± 0.0380.310 ± 0.0350.286 ± 0.018
1.640.197 ± 0.0120.296 ± 0.0630.236 ± 0.0680.231 ± 0.027
2.130.115 ± 0.0090.226 ± 0.0650.145 ± 0.0850.149 ± 0.034

5. Methodology

Among the different aerosol components, the most suitable aerosol components based on aerosol source regions, long-range transport and earlier reported data over the IGP are found to be water insoluble (IS), water soluble (WS), black carbon (BC), mineral dust (MD) and sea salt (SS) (e.g. Chinnam et al., 2006; Tare et al., 2006; Dey and Tripathi, 2007). In the determination of probable mixing states, the optical properties (AOD, SSA and g) of aerosols are estimated assuming external and core-shell mixing scenarios. The coated-sphere Mie calculation requires the refractive index of core and shell species, and the radius of core and shell particles. Core to shell radius ratio (CSR) is calculated from the geometry of core-shell particles, which depends upon the mass, M, and density, ρ, of the core and shell. All the aerosol species are assumed to follow a log-normal size distribution. The size distribution parameters and refractive indices of different aerosols are taken from the OPAC database (Hess et al., 1998). Refractive indices of different hydrophilic and hygroscopic aerosol species at 0.55 µm are given in Tables 3 and 4. Extinction, absorption and scattering efficiencies of single core-shell mixed particles are obtained from the Mie calculation, then these efficiencies are integrated for the log-normal size distribution to get the extinction, absorption and scattering coefficients. It should be noted that the measured behaviour of the coated particles was found to be consistent with Mie calculations of core-shell particles in laboratory experiments (Riziq et al., 2008).

Table 3. Real and imaginary parts of refractive index of hydrophilic aerosol species at 0.55 µm, as given in Hess et al. (1998) and used in the present study.
Aerosol speciesRefractive index
Insoluble1.53 – i 8.00 × 10−3
Black carbon1.75 – i 4.40 × 10−1
Mineral dust1.53 – i 5.50 × 10−3
Table 4. Real and imaginary parts of refractive indices of hygroscopic aerosol species at 0.55 µm, as given in OPAC (Hess et al., 1998) for different relative humidities, and used in the present study.
RH Water solubleSea saltSea salt
(%) (accum. mode)(coarse mode)
01.53 – i 6.00×10−31.50 – i 10.0×10−91.50 – i 10.0×10−9
501.44 – i 3.18×10−31.37 – i 3.89×10−91.37 – i 3.88×10−9
701.41 – i 2.47×10−31.36 – i 3.32×10−91.36 – i 3.31×10−9
801.40 – i 1.99×10−31.35 – i 2.98×10−91.35 – i 2.97×10−9
901.38 – i 1.36×10−31.34 – i 2.56×10−91.34 – i 2.55×10−9
951.36 – i 0.90×10−31.34 – i 2.29×10−91.34 – i 2.28×10−9

The size distribution parameters of core-shell mixed particles (viz. mode radius and width of log-normal curve) are assumed to be the same as those of the shell species distribution. This assumption is supported by scanning electron microscopy (SEM) analysis of Asian dust coated by BC which showed that the mode radii of BC aggregated with dust are closer to the shell BC radii (e.g. Arimoto et al., 2006). The dry (0% RH) mode radii of BC, WS and IS aerosol species are 0.0118, 0.0212 and 0.471 µm respectively (Hess et al., 1998). The mode radii of mineral dust in nucleation, accumulation and coarse modes are 0.07, 0.39 and 1.90 µm respectively. The mode radii of SS aerosols in accumulation and coarse modes are 0.209 and 1.75 µm respectively (Hess et al., 1998). However, mode radii of hygroscopic aerosols will increase with increase in RH (Hess et al., 1998). RH plays an important role when hygroscopic aerosols are present in core or shell (Lesins et al., 2002). Since RH shows a large seasonal variation over the IGP (Table 1), optical properties of core-shell mixed and externally mixed aerosols are estimated for the seasonal mean RH (Table 1) by taking into account the mode radii of different aerosol species given in the 0–95% RH range in OPAC. The mode radius of BC aerosols is taken as 0.0118 µm in the study. Jacobson et al. (2005) reported that the mode radius of BC spherules can be in the range 0.013–0.017 µm. However, since it has been shown that any change in mode radius of BC results only in a subtle change in aerosol optical properties (Lesins et al., 2002), this assumption is not expected to modify the derived aerosol optical properties.

Observations during Aerosol Characterization Experiment (ACE–Asia) revealed that typically 15 to 30% of dust particles' surface was covered by BC (Arimoto et al., 2006). Similar results were obtained over an urban location in Germany where between 20 and 40 of the complex secondary aerosol particles contained soot inclusions (Vester et al., 2007). Following this, different mass fractions ranging from 20 to 100% of aerosol species are considered to take part in core-shell mixing in the present study; the remaining mass fraction of aerosol species which are core-shell mixed, and other aerosol species not involved in core-shell mixing, are assumed to be externally mixed with the core-shell composite particles. In the case of external mixing, total extinction, scattering and absorption coefficients are calculated by summing the respective coefficients of the individual aerosol species; this is acceptable because of the absence of physical and/or chemical interaction among the aerosol species. Spectral aerosol optical properties (AOD, SSA and g) estimated for different mixing scenarios are then compared and contrasted with measured optical properties, and mixing states are determined.

5.1. Determination of probable mixing states

Chemical characteristics of aerosols over Kanpur during winter, pre-monsoon and post monsoon have been reported by e.g. Chinnam et al. (2006), Dey et al. (2008) and Tare et al. (2006). Higher BC and dust were reported during winter and pre-monsoon respectively, and BC mass fraction at the surface was lower during the monsoon over Kanpur (Chinnam et al., 2006; Tripathi et al., 2007). Chemical characterization studies suggested that BC, WS (NH4SO2 and NOequation image), salt (K+, Na+, Cl) and dust were the dominant aerosol species over the IGP (Dey and Tripathi, 2007). The mean mass concentration of dust (which consists of soil dust or fly ash from coal-based thermal power plants (Prasad et al., 2006), industrial emissions, and long-distance dust) was 1 µg m−3 in fine mode, and 5.5 µg m−3 in coarse mode, and contributed ∼ 6% to total mass concentration during winter (Dey and Tripathi, 2007). Based on the above observation, the dust mass has been divided into (1) soil dust or fly ash as IS aerosol, and (ii) transported mineral dust, following the specific aerosol types given in Hess et al. (1998). The mixing state of aerosols for years 2007–2009 is determined using the reported mass concentration over Kanpur for 2004, since mass concentration measurements for 2007–2009 are not available. This is not expected to influence the results significantly as the interannual variability in aerosol properties is less than the seasonal variability (e.g. Singh et al., 2004). Over Gandhi College, as the mass concentration of aerosol species are not available, we consider the aerosol mass concentration over Kanpur as a basis, and modify the mass concentrations of different aerosol species taking into account the meteorology and long-range transport over Gandhi College. This is based on inference that types of aerosol species present over the IGP are considered to be similar while the percentage contribution varies during the year as a function of season and aerosol type (e.g. Singh et al., 2004; Prasad et al., 2006). In order to derive the aerosol mass concentrations over Kanpur and Gandhi College, the number concentration of each aerosol component mentioned above is altered suitably until the following conditions are met:

(1) The root mean square (rms) difference between the measured AOD and modelled AOD spectra is < 0.03, thus constraining the rms difference to within 15% of AOD.

(2) The Ångström wavelength exponent α obtained from the measured AODs in the 0.38–0.87 µm wavelength region should be consistent with model-derived α values.

(3) The OPAC-estimated total mass concentrations should lie within ±1σ of the earlier reported mass concentrations. This condition is applicable only for Kanpur.

The mass concentrations of different aerosol species obtained using the above criteria are used to estimate the AOD, SSA and g spectra for different mixing scenarios. Note that only AOD spectra are used as a constraint in the determination of mass concentration of aerosol species, while all the three spectral optical properties are used for identifying the probable mixing state of aerosols. It should be noted that, in AERONET retrieval of SSA, aerosol particles are considered as homogeneous (uniform refractive index) spheres and modelled sun/sky spectral radiances are fitted with the measured spectral radiances (Dubovik and King, 2000), as stated earlier. This suggests that there are no assumptions on the refractive indices of different aerosol species in AERONET retrieval which distinguishes absorbing aerosols from scattering aerosols. In the present study, SSA is estimated for different aerosol mixing scenarios using the mass concentrations of different aerosol species.

In contrast to the earlier study (Dey et al., 2008), where only spectral SSA was used as a constraint, we use all the three measured aerosol parameters. The cases for which rms differences in AOD, SSA g spectra are < 0.03 are designated as the probable mixing states. The rms difference of 0.03 corresponds to the lowest uncertainty in the AERONET retrievals of AOD, SSA and g. However, there are cases where rms differences in the spectra of AOD and SSA are < 0.03, but for asymmetry parameter g this condition is not satisfied (Tables 5, 6), which could be due to the asymmetry in the shape of aerosol particles. In the present study, the optical properties are estimated assuming aerosols as spherical. However, in the atmosphere aerosols are present in various shapes (e.g. Guieu et al., 1994). The AERONET algorithm was modified to take into account non-spherical shapes of aerosol particles such as mineral dust (Dubovik et al., 2006). Therefore, the spectral g values derived in our calculation can differ from the measured g values, since in the model calculation all aerosols are treated as spherical. Variations in g values have been found to influence aerosol radiative forcing only negligibly (Mishchenko et al., 1997), an aspect which is re-examined later in this study.

Table 5. Aerosol species (viz. insoluble, water soluble, black carbon and mineral dust present in core and shell) and their percentage mass fractions in mixing along with rms differences between the measured and modelled optical properties in the wavelength range 0.34–1.02 µm over Kanpur. See text for details.
 Mixing state Rms differences
 Core Shell    
SeasonSpecies% Species% AODSSAg
  1. AOD = aerosol optical depth; SSA = single scattering albedo; g is the asymmetry parameter.

WinterBlack carbon100 Water soluble20 0.0190.0200.029
Pre-monsoonInsoluble20 Black carbon20 0.0170.0210.076
 Black carbon20 Insoluble20 0.0170.0210.074
 Mineral dust20 Black carbon20 0.0180.0280.087
MonsoonBlack carbon20 Water soluble50 0.0300.0100.054
 Mineral dust100 Black carbon100 0.0300.0260.090
 Mineral dust20 Water soluble20 0.0200.0130.064
Post-monsoonInsoluble20 Water soluble20 0.0270.0300.052
 Black carbon100 Water soluble20 0.0290.0210.034
Table 6. As Table 5, but for Gandhi College.
 Mixing state Rms differences
 Core Shell    
SeasonSpecies% Species% AODSSAg
WinterInsoluble20 Water soluble20 0.0240.0090.035
 Black carbon50 Water soluble50 0.0230.0240.031
 Mineral dust50 Water soluble20 0.0220.0150.028
 Sea salt100 Water soluble20 0.0230.0170.028
 Insoluble50 Sea salt20 0.0200.0270.096
Pre-monsoonMineral dust20 Black carbon50 0.0240.0250.086
 Black carbon20 Insoluble50 0.0250.0240.072
MonsoonInsoluble50 Water soluble50 0.0240.0140.036
 Mineral dust50 Black carbon20 0.0280.0160.102
 Mineral dust50 Water soluble20 0.0230.0050.040
 Sea salt20 Black carbon100 0.0270.0130.107
 Sea salt20 Water soluble20 0.0240.0170.060
 Mineral dust50 Sea salt20 0.0290.0180.098
Post-monsoonInsoluble20 Water soluble50 0.0270.0100.027
 Black carbon20 Water soluble20 0.0280.0220.023
 Sea salt50 Black carbon100 0.0280.0270.081
 Mineral dust20 Black carbon100 0.0280.0270.083

5.2. Calculation of aerosol radiative forcing

The SBDART (Santa Barbara DISORT Atmospheric Radiative Transfer) model developed by Ricchiazzi et al. (1998) is used for radiative transfer calculations. SBDART solves plane-parallel radiative transfer in clear- and cloudy-sky conditions within the Earth's atmosphere. The fluxes calculated by SBDART were found to lie within 2% of direct and diffuse irradiance measurements (Michalsky et al., 2006). Aerosol radiative forcing calculations are performed using eight radiation streams at 1 h intervals for a range of solar zenith angles in the wavelength range 0.2–4.0 µm, and 24 h averages are obtained, which are then used in calculating the monthly and seasonal means. The radiative fluxes are computed for clear-sky conditions in the presence of aerosols and for an aerosol-free atmosphere. The aerosol optical parameters (AOD, SSA and g) obtained for probable mixing states of aerosols and for external mixing are input to the SBDART model to estimate aerosol radiative forcing. The CALIOP seasonal mean aerosol extinction profiles are scaled by the respective seasonal mean AODs (e.g. Ramachandran and Kedia, 2010), and vertical profiles of aerosol radiative forcing and heating rate are calculated. It should be noted that only the information on the vertical structure or distribution of aerosols (including the presence of elevated layers) are taken from the CALIOP-derived aerosol extinction profiles, and not the absolute values of aerosol extinction. Therefore, the uncertainty in CALIOP-derived aerosol extinction is not expected to alter significantly the results on vertical profiles of aerosol radiative forcing and heating rates.

The amount of absorbed energy is termed the atmospheric heating (cooling) rate (K day−1) and can be calculated as

equation image(1)

where ∂T/∂t is the heating rate (K day−1), g is the acceleration due to gravity, Cp is the specific heat capacity of air at constant pressure and P is the atmospheric pressure (Liou, 1980). For the estimation of vertical profile of heating rate, the forcing and pressure difference corresponding to two consecutive altitudes are taken. The absolute total uncertainty in the modelled aerosol radiative forcing was found to vary in the range 0.2–3.1 W m−2 by considering different aerosol optical properties, and SSA was found to be the largest contributor to the above uncertainty (McComiskey et al., 2008). The relative standard error in aerosol radiative forcing and heating rate reported here is found to be < 15% by taking into account the uncertainties in aerosol input parameters, additional inputs and flux estimates.

6. Results and discussion

6.1. Aerosol mass fractions over Kanpur and Gandhi College

The mass fractions of different aerosol species obtained using the criteria described in section 5.1 over Kanpur and Gandhi College are shown in Figure 2. BC mass over Kanpur in winter is found to be 12 µg m−3. This is consistent with measurements made during winter, which showed higher BC mass with mean of 12 µg m−3 and contributed ∼ 10% to total mass concentration of aerosols over Kanpur (Tare et al., 2006). During winter, biomass burning is most prominent over the IGP (Di Girolamo et al., 2004; Habib et al., 2006), and organic aerosols are the single largest species of biomass burning aerosols in addition to BC (IPCC, 2001, and references therein). Organic aerosol species are accounted for in OPAC in both IS and WS components (Hess et al., 1998). Thus, IS and/or WS aerosols contribute the highest to the total mass over Kanpur and Gandhi College during winter (Figure 2). BC aerosol reduces to 1% of total mass during the monsoon due to wet removal. The decrease in BC mass fraction from winter to monsoon was also observed and reported over Kanpur (e.g. Tripathi et al., 2007) and over other locations in India (e.g. Ramachandran and Kedia, 2010, and references therein). The mass of dust is higher and contributes ≥ 40% to total mass during pre-monsoon and monsoon seasons over Kanpur and Gandhi College (Figure 2) as northwesterly winds transport dust aerosols from the Thar desert to the study locations (e.g. Chinnam et al., 2006). Mass fraction of WS aerosol is found to be higher (> 50%) during winter, while it is lower during the pre-monsoon (equation image) due to lower RH (Table 1). WS (containing organic carbon) and BC aerosols are higher during winter and less during summer (Figure 2). The seasonal variation in mass concentrations of various aerosol species are consistent with measurements reported in (Chinnam et al., 2006; Prasad et al., 2006; Tare et al., 2006). Aerosol mass concentrations at Gandhi College are estimated to be lower than at Kanpur (Figure 2). The fine mode fraction of aerosols was found to be higher over Gandhi College than Kanpur throughout the year (Ramachandran and Kedia, 2012). Fine mode fraction was higher during winter and post-monsoon than pre-monsoon and monsoon seasons over both the locations (Ramachandran and Kedia, 2012). Higher fine mode fraction over Gandhi College gives rise to a lower aerosol mass, whereas a higher fraction of coarse aerosols in the size distribution results in higher mass concentrations over Kanpur.

Figure 2.

Percentage mass contribution of different aerosol species to total mass concentration (µg m−3) over Kanpur and Gandhi College during (a, e) winter, (b, f) pre-monsoon, (c, g) monsoon and (h, i) post-monsoon respectively. The total mass concentration in each season is shown in parentheses. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

6.2. Probable mixing states in the IGP

6.2.1. Kanpur

AERONET-derived spectral aerosol optical properties and the modelled properties corresponding to probable and external mixing states of aerosols during winter, pre-monsoon, monsoon and post-monsoon are shown in Figure 3. The ±1σ standard deviation (shown as vertical bars) in the AERONET data correspond to both intraseasonal and interannual variations in aerosol optical parameters. The most probable mixing states obtained and the rms differences in the measured and modelled optical properties during different seasons are given in Table 5. During winter, coating of 20% WS aerosol over all the BC (100%) is found to be the probable mixing state of aerosols. During the pre-monsoon, the presence of dust aerosols which are transported by northwesterly winds give rise to a core-shell mixing of 20% BC over 20% dust mass (Table 5). 20% of insoluble aerosols (mainly resuspended dust and IS organics) coated by 20% mass of BC is also found to be a probable mixing state.

Figure 3.

Spectral aerosol parameters: (a–d) aerosol optical depth, (e–h) single scattering albedo and (i–l) asymmetry parameter from AERONET, for external mixing and probable mixing states of different aerosol species, viz. insoluble (IS), water soluble (WS), black carbon (BC), mineral dust (MD) and sea salt (SS) during winter, pre-monsoon, monsoon and post-monsoon seasons over Kanpur. Table 5 gives further details. Vertical bars in AERONET data represent ±1σ deviation from the mean. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

The occurrence of this probable mixing state is supported by the observations taken during ACE–Asia where BC coating over the dust was observed (Arimoto et al., 2006; Clarke et al., 2004). 20% BC and dust is found to be coated by 20% WS aerosols. However, dust coated by 20% BC also results in a probable mixing state during the monsoon. WS aerosol (e.g. sulphate and nitrate) coating over dust is found to be most common due to the secondary transformation of sulphate and nitrate on dust particles during the long-range transport of dust (Bauer and Koch, 2005; Wang et al., 2005). IS and WS in small fractions (20%), and large fraction of BC (≥ 50%) with 20% WS are probable core-shell mixing states during the post-monsoon. External mixing is also probable during the monsoon and post-monsoon seasons, while SSAs in external mixing are lower than AERONET measurements during the winter and pre-monsoon. The previous finding of probable mixing states over Kanpur by Dey et al. (2008) is consistent for pre-monsoon and post-monsoon. However, there are differences in the mixing states of aerosols between the present study and that of Dey et al. which could be due to:

(1) In the present study, fractions of mass of aerosols ranging from 20% to 100% are considered to be involved in core-shell mixing, while Dey et al. assumed the entire mass of aerosol species which acts as core to be coated by the species acting as shell.

(2) All the three optical properties (AOD, SSA and g) and reported mass concentrations are used as constraints to determine the probable mixing state of aerosols in the present study, while Dey et al. used mass concentration of species and SSA values only.

(3) The mixing states are determined when the rms difference between measured and modelled AOD and SSA spectra is < 0.03, while in Dey et al.'s case the rms difference in SSA spectra was much higher than the stringent condition adopted in the present study.

(4) The present study is done on a seasonal basis using data obtained from 2007–2009, while Dey et al.'s results come from three specific months (October 2005, December 2005 and March 2006).

6.2.2. Gandhi College

Spectral AOD, SSA and g measured by AERONET and modelled for different probable mixing scenarios and external mixing are shown in Figure 4. The vertical bars (±1σ) in AERONET data include both the intraseasonal and interannual variations. The various core-shell mixing states and rms differences between measured and modelled properties are given in Table 6. A small mass fraction of IS (soil dust and organics) and SS/WS is found to be most probable in core-shell mixing during winter. This type of mixing was reported in a previous study (Andreae et al., 1986) in which the fraction of silicate from dust was found to be internally mixed with SS. Various mechanisms such as collision of SS and silicate, Brownian coagulation of aerosols, and electrostatic attraction between silicate and SS, were proposed for such types of mixing (Andreae et al., 1986). During the pre-monsoon, dust originating from the Thar desert can reach Gandhi College by passing through many urban locations (Figure 1). Therefore dust, which comes through the polluted regions, could be mixed with the local pollutants (BC), and get coated during the pre-monsoon. The BC and IS core-shell mixing is also probable during the pre-monsoon. Analogous to the observations, the core-shell mixing of WS aerosols with other aerosols (IS, BC, MD, and SS) are found to be the probable mixing states during the monsoon season which exhibits the highest (91%) RH (Figure 4, Tables 1 and 6). The probable mixing states during the post-monsoon are IS coated by WS and BC coated by WS. The uptake of WS sulphate and nitrate on other aerosols (e.g. dust) has been found to be more prominent in various observations (e.g. Jordan et al., 2003; Bauer and Koch, 2005). During the post-monsoon, SS and MD coated by BC are also found to be probable mixing states over Gandhi College. Thus, in the IGP, different core-shell mixing states are observed and the state of mixing varies with location depending upon the availability of different types of aerosol and environment.

Figure 4.

As Figure 3, but for Gandhi College. Table 6 gives further details. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

6.3. Comparison of aerosol radiative forcing: AERONET and SBDART model

AERONET version 2.0 inversion data also include the solar spectral and broadband fluxes and aerosol radiative forcing at the surface (SFC) and at the top of the atmosphere (TOA) (García et al., 2008). The retrieved spectral aerosol properties from AERONET are used to estimate broadband fluxes in the wavelength range 0.2–4.0 µm. AERONET-retrieved spectral real and imaginary refractive indices are interpolated/extrapolated for the above wavelength range. In version 2.0 AERONET data, the dynamic (spectral and spatial) satellite and model estimation of the surface albedo, including the bidirectional reflectance distribution function (BRDF), is accounted (García et al., 2008). The integration of atmospheric gaseous absorption and molecular scattering effects are accounted using the radiative transfer Global Atmospheric ModEl (GAME) (Dubuisson et al., 1996). AERONET computes broadband radiation and aerosol radiative forcing using the spectral dependence of the aerosol optical properties and surface albedo as inputs to the radiative transfer model. A comparison between the instantaneous aerosol radiative forcing retrieved from AERONET and estimated by SBDART shows good agreement (r2 > 0.9, Figure 5). Despite the differences in radiative transfer algorithms and methodology, the good agreement between AERONET and SBDART radiative forcing suggests that the approach adopted in the study to estimate aerosol optical properties and the resultant radiative forcing is quite robust.

Figure 5.

Modelled versus AERONET aerosol radiative forcing (a) at the surface and (b) at the top of the atmosphere for all seasons during 2007--2009 over Kanpur. The linear fit between the modelled and AERONET forcing is also shown. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

6.4. Aerosol radiative forcing, forcing efficiency and heating rate over the IGP

6.4.1. Kanpur

Aerosol radiative forcing at SFC, in the atmosphere (ATM) and at TOA over Kanpur during different seasons estimated for the external and all the probable mixing scenarios (listed in Table 5) are shown in Figure 6. The vertical profiles of radiative heating rate are also shown in Figure 6. Aerosol forcing for external mixing shows higher deviations from those for probable mixing cases during winter and pre-monsoon since the SSA for external mixing is lower than the measured and modelled SSA for different core-shell mixing scenarios during these seasons (Figure 3). Aerosol forcing for all the mixing states are higher during the post-monsoon due to higher AODs. Forcing for external mixing and all the core-shell mixing scenarios agree within the uncertainty limits during the monsoon and post-monsoon seasons. TOA forcing changes its sign from positive for external mixing to negative for probable mixing cases during the pre-monsoon (Figure 6) since SSA is lower (< 0.8) for external mixing than for measured and probable mixing cases (SSA > 0.8) (Figure 3). The aerosol radiative forcing at the surface for probable mixing states are higher during the post-monsoon (−44 W m−2), and lower during the monsoon (−30 W m−2). Similar seasonal variations in aerosol radiative forcing were also obtained during 2001–2005 over Kanpur by Dey and Tripathi (2008).

Figure 6.

Aerosol radiative forcing at the surface (SFC), in the atmosphere (ATM) and at the top of the atmosphere (TOA) for external and different probable mixing states of different aerosol species, viz. insoluble (IS), water soluble (WS), black carbon (BC), mineral dust (MD) and sea salt (SS) over Kanpur during (a) winter, (b) pre-monsoon, (c) monsoon and (d) post-monsoon. The vertical profiles of heating rate (K day−1) for different mixing scenarios are also plotted. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

The inclusion of vertical distribution of aerosols is found to modify only negligibly (∼5%) the net aerosol radiative forcing for SFC, ATM and TOA. Vertical profile of short-wave heating rate shows larger values for external mixing (Figure 6) than for probable core-shell mixing during all the seasons. The heating rate profile is found to peak at 1.25 km with the mean value of ∼1.5 K day−1 for external mixing and ∼0.75 K day−1 for probable mixing states during the pre-monsoon over Kanpur. Vertical profiles of aerosol heating rates follow the aerosol extinction profiles and exhibit structures consistent with aerosol extinction at different altitudes. During winter, the primary heating rate peak of ∼ 1 K day−1 occurs at 0.5 km and a secondary peak of ∼ 0.6 K day−1 at 1.75 km altitude. Vertical profile measurements over Kanpur by Tripathi et al. (2007) suggested that the enhancement of radiative heating rate can occur due to the presence of high BC fraction at the higher altitudes. Heating rate for probable mixing states during the winter and post-monsoon are similar (∼1 K day−1).

6.4.2. Gandhi College

Aerosol radiative forcing at different levels over Gandhi College during different seasons estimated for various probable mixing scenarios (listed in Table 6) and external mixing scenarios are shown in Figure 7. Vertical profiles of heating rate for different seasons are also shown in Figure 7. Aerosol radiative forcing at SFC and ATM is higher during the pre-monsoon and significant difference in forcing is observed for probable and external mixing states of aerosols. Aerosol radiative forcing estimated for different probable mixing states is similar for a given season. During the post-monsoon, aerosol forcing for external mixing is higher than those obtained for probable mixing states since SSA is lower for the external mixing (Figure 4). Aerosol forcing at all the levels is higher during winter and lower during the monsoon (Figure 7). The higher forcing during winter is attributed to higher AOD and lower SSA, while lower forcing occurs during the monsoon due to higher SSA and lower AOD. TOA forcing is positive when aerosols are externally mixed, while it is negative for probable mixing state during the pre-monsoon, while TOA forcing is always negative for other seasons for all the mixing states. The aerosol radiative forcing at the surface for probable mixing states is higher during winter (−35 W m−2), and lower during the monsoon and post-monsoon (−24 W m−2), which is in contrast to Kanpur where aerosol surface forcing is higher during the post-monsoon (−44 W m−2). Aerosol forcing efficiency at the surface for probable mixing state over Gandhi College is lower during the monsoon, since SSA is higher during the monsoon (Table 7). Aerosol forcing efficiency at SFC and ATM over Gandhi College is found to be lower than at Kanpur during the winter and pre-monsoon (Table 7).

Figure 7.

As Figure 6, but for Gandhi College. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

Table 7. Aerosol forcing efficiency at the surface, in the atmosphere, and at the top of the atmosphere (TOA), for probable mixing state over Kanpur and Gandhi College during 2007–2009.
 Forcing efficiency (W m−2/AOD)
 Kanpur Gandhi College
SeasonSurfaceAtmosphereTOA SurfaceAtmosphereTOA
Winter−5538−17 −4730−18
Pre-monsoon−5850−8 −5446−9
Monsoon−4328−15 −4431−13
Post-monsoon−4630−16 −5030−20

In winter, the heating rate at SFC is ∼ 1.75 K day−1 over Gandhi College (Figure 7). The vertical profile of heating rate shows a smaller secondary peak at 1.75 km, while during the pre-monsoon and monsoon the peak in heating rate appears at 1 km. During the post-monsoon, the heating rate is lower (∼ 0.5 K day−1) at Gandhi College, but the heating rate is higher (∼ 1 K day−1) at Kanpur and peaks at 0.75 km.

The aerosol forcing efficiency (aerosol radiative forcing per unit AOD) at SFC and ATM for probable mixing states is found to be highest during the pre-monsoon, while forcing efficiency is lowest at the TOA (Table 7). Aerosol forcing efficiency is independent of AOD and is governed by the type of aerosols (SSA) and surface albedo. Thus, lower SSA during the pre-monsoon leads to higher surface forcing efficiency and lower TOA forcing efficiency than in the other seasons; spectral surface albedo is marginally higher during the pre-monsoon and monsoon than in the winter and post-monsoon (Table 2).

6.5. Sensitivity of the asymmetry parameter to aerosol radiative forcing

The asymmetry parameter (g) value can change depending on the sphericity of aerosols. Dust-like aerosols are found to be non-spherical in shape (e.g. Guieu et al., 1994). In this study, the aerosols are assumed to be spherical, therefore the modelled g can be different from measured g when dust-like aerosols are mixed with other aerosols. This is corroborated by the fact that the rms differences between the measured and modelled g spectra are larger than those obtained for AOD and SSA (Tables 5, 6). The sensitivity of g to aerosol radiative forcing is estimated to examine the effect of shape of dust-like aerosols on the radiative forcing. Aerosol radiative forcing over Kanpur during the pre-monsoon, monsoon and post-monsoon for g values obtained for different mixing scenarios, and for AERONET - measured g values are shown in Figure 8. The sensitivity calculation has not been done for winter since g for different mixing cases during the winter matches well (rms difference ≤ 0.03) with the measured g (Table 5).

Figure 8.

Sensitivity of the asymmetry parameter (g) to aerosol radiative forcing at the surface, in the atmosphere and at the top the atmosphere over Kanpur during (a) pre-monsoon, (b) monsoon and (c) post-monsoon. Filled bars correspond to g obtained for different mixing states of different aerosol species, viz. insoluble (IS), water soluble (WS), black carbon (BC), mineral dust (MD), and unfilled bars represent the aerosol forcing for measured g. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

Aerosol radiative forcing at SFC and TOA for modelled and measured g deviate a little, but agree within the uncertainty of forcing estimation during all the seasons. Higher g values correspond to higher forward scattering and hence result in a higher reduction of radiation at SFC and also lower TOA forcing. BC coating of dust during the monsoon has the largest rms deviation (0.09) in g spectra (Table 5) which results in a 3 W m−2 lower value in both SFC and TOA forcing (Figure 8). The deviations in aerosol radiative forcing at SFC and at TOA cancel each other and result in the same value of atmospheric radiative forcing. The aerosol radiative forcing is weakly sensitive to g when surface albedo is ≥ 0.1 (McComiskey et al., 2008). The broadband surface albedo is higher (≥ 0.2) over the IGP due to which aerosol radiative forcing is seen to be less sensitive to variations in g. Mishchenko et al. (1997) suggested that the influence of particle shape on the aerosol radiative forcing at TOA is negligibly small and can be accurately estimated using the Mie theory of spherical particles. Thus, this calculation reiterates that aerosol radiative forcing at TOA and SFC is weakly sensitive to g, and any change in aerosol forcing due to differences in g is smaller than the uncertainty limits of the forcing computation.

7. Summary and conclusions

The most probable mixing states of aerosols have been determined using measured and modelled optical properties–aerosol optical depth (AOD), single scattering albedo (SSA) and asymmetry parameter (g)–over an urban (Kanpur) and a rural (Gandhi College) location in the Indo-Gangetic Plain. The seasonal variations in aerosol mixing state and their impact on aerosol radiative forcing has also been estimated and discussed. The sensitivity of aerosol radiative forcing to non-sphericity of aerosols has also been assessed.

The major findings are:

  • 1.The probable mixing states of aerosols over Kanpur and Gandhi College are found to exhibit seasonal variations. Different fractions of BC–WS aerosols in core-shell mixing are found to be probable during the winter, monsoon and post-monsoon, while MD and BC core-shell mixing is found to be probable during pre-monsoon and monsoon. In addition, WS aerosols coated over MD can also be a probable mixing state during the monsoon. External mixing is probable during the monsoon and post-monsoon seasons.
  • 2.Over Gandhi College, the probable mixing states are found to be an IS–SS core-shell mixture during the winter, MD/IS–BC core-shell mixture in the pre-monsoon and monsoon, while in the monsoon MD–WS, BC–WS and WS–SS core-shell mixtures are also probable in addition to MD–BC core-shell mixing. During the post-monsoon, IS–BC, SS–BC, and MD–BC core-shell mixing are the most probable. External mixing is found to be most probable during the winter and monsoon.
  • 3.Aerosol radiative forcing over Kanpur and Gandhi College at the surface, in the atmosphere and at the top of the atmosphere for various probable mixing states agree well for a season. TOA forcing changes its sign from positive for external mixing to negative for probable mixing states during the pre-monsoon over both locations. Surface aerosol forcing for probable mixing state during the post-monsoon is higher (−44 W m−2) over Kanpur, and lower (−24 W m−2) over Gandhi College. Instantaneous aerosol radiative forcings computed from a model and obtained from AERONET show good correlations with r2 > 0.9, which shows the robustness of the methodology used.
  • 4.The heating rate over Kanpur and Gandhi College is found to be similar during the pre-monsoon (∼0.75 K day−1) and monsoon (∼0.5 K day−1), while differences occur during the other seasons. During the post-monsoon, the heating rate is higher (∼1 K day−1) over Kanpur than at Gandhi College (∼0.5 K day−1).
  • 5.Differences in modelled and measured asymmetry parameter (g) arise due the non-spherical shape of aerosols. Sensitivity results reconfirm that aerosol radiative forcing at TOA and SFC is less sensitive to g, and the difference in g is found to produce only a small change in aerosol forcing due to higher surface albedo.

The knowledge of the probable mixing states of aerosols over the IGP and their impact on aerosol radiative forcing will be useful in assessing the role of aerosols on regional and global climate.

Acknowledgements

We thank B. N. Holben, R. P. Singh and S. N. Tripathi for their efforts in establishing and maintaining the AERONET at Kanpur and Gandhi College, the monthly mean data of which are used in this study Wind speeds and relative humidity were obtained from National Climatic Data Center, USA via http://www.cdc.noaa.gov. The air back trajectories are calculated using HYSPLIT model from http://www.arl.noaa.gov/ready/hysplit4.html. MODIS aerosol optical depths and OMI columnar ozone values were downloaded from GES-DISC, NASA. CALIPSO data were obtained from the NASA Langley Research Center Atmospheric Science Data Center. The PRL high-performance computing cluster was used for the coated sphere Mie calculation and radiative forcing computation.

Ancillary