Analysis of Aerosol Optical Properties due to a Haze Episode in the Himalayan Foothills: Implications for Climate Forcing

A super haze episode occurred over the Himalayan region in October 2010. This haze reduced the air quality in the region and spread across India and Pakistan. The purpose of this study is to investigate the optical and radiative properties of aerosols during this episode using data from the MODerate resolution Imaging Spectroradiometer (MODIS) and the AErosol RObotic NETwork (AERONET). Maximum Aerosol Optical Depth (AOD) values on 19 October exceeded two at various locations in Pakistan (Sialkot = 2.56, Faisalabad = 2.67) and India (Ambala = 2.03, Amritsar = 3.4, Ludhiana = 4.29). Maximum AOD values recorded on 20 October were slightly lower in parts of Pakistan (Lahore = 2.5) and India (Gurdaspur = 1.89, New Delhi = 1.90, Batala = 2.89, Bathinda = 1.89, Kanpur = 1.6). Data for aerosol properties such as Volume Size Distributions (VSD), Single Scattering Albedo (SSA), Refractive Index (RI), and ASYmmetry parameter (ASY) suggest that fine mode aerosols were predominant relative to coarse mode aerosols during the haze episode. The dominant aerosol types were classified by analyzing AOD vs. Ångström Exponent (AE) and Extinction Ångström Exponent (EAE) vs. Absorption Ångström Exponent (AAE). The results revealed that during the haze episode, the prevailing aerosol types were biomass burning and urban/industrial aerosol. The Aerosol Radiative Forcing (ARF) values were computed during the hazy and non-hazy days using the Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART) model. The shortwave ARF values at the Top Of the Atmosphere (TOA), SuRFace (SRF), and within the atmosphere were found to be in the range of –17.6 to –81.6, –64 to –193, and +47 to +119 W m, respectively, over Lahore. Likewise, over Kanpur, the ARF values were found to be in the range of +15.32 to –91.6, –38 to –134 and +33 to +75.91 W m, respectively.


INTRODUCTION
South Asia has experienced rapid economic growth over the past few decades, characterized by expanding urbanization with increasing population, industrialization, and anthropogenic activity.These changes have resulted in high aerosol loadings (Li et al., 2013a), reduction in visibility (Ramanathan et al., 2001), and alterations in aerosol radiative forcing (Kim and Ramanathan, 2008;Alam et al., 2014b) and air quality (Oanh and Leelasakultum, 2011;Li et al., 2017) in the Indo-Gangetic plain (Ram et al., 2012).Areas especially affected are the eastern areas of Pakistan such as the megacity Lahore (Alam et al., 2011(Alam et al., , 2012;;Gupta et al., 2013;Tariq et al., 2016), the coastal part of Karachi, and parts of southern territory of Pakistan such as around the Indus River plain.
Satellite data such as from the MODerate resolution Imaging Spectroradiometer (MODIS) are widely used for analyzing the aerosol distributions from regional to global scales.Satellite imagery is used to monitor aerosol optical thickness, which is an effective way to assess air pollution levels.Satellite observations also indicate that haze pollution not only exists in the urban regions, but also exists in rural areas (Tao et al., 2012).Ground-based remote sensing data from networks such as the AErosol RObotic NETwork (AERONET) are also suitable for the assessment of aerosol optical/radiative properties and aerosol loading at a particular site (Sorooshian et al., 2011;Alam et al., 2014;Dumka et al., 2014;Boiyo et al., 2017).
Atmospheric haze episodes in South Asia have recently received great attention due to increased frequency (Ramanathan et al., 2001;Chung et al., 2002;Gautam et al., 2007;Jose et al., 2015).Haze is increasingly seen as a risk to health, economic prosperity, and human security, particularly as it is linked with climate dynamics.To assess regional and global climate change caused by haze episodes in South Asia, detailed information is required about the atmospheric concentrations and properties of fine and coarse particles in the region.In the past few decades, there have been a number of studies on the physical, optical, and radiative properties of haze particles in different regions of South Asia (Ramanathan et al., 2001;Chung et al., 2002), the Southern Hebei province of China (Yang et al., 2013), Eastern China ( (Li et al., 2013a;Tao et al., 2013, Yang et al., 2013, Wang et al., 2014), Thailand (Oanh and Leelasakultum, 2011), Korea (Kim et al., 2008a), and the Himalayan foothills and Indo-Gangetic Plain (Ram et al., 2012).Despite the long history of aerosol research, knowledge of the optical properties of haze particles is still far from being sufficient (Kim et al., 2008b).Therefore, the present study has a potential to fill a scientific gap in our present knowledge about the haze characteristics and their optical and radiative properties.Such investigations are important for climate modeling studies, particularly over the monsoon region of India and Pakistan.
In the present study, we report aerosol properties measured from the AErosol RObotic NETwork (AERONET) and MODerate resolution Imaging Spectroradiometer (MODIS) over India and Pakistan during the heavy haze episode of October 2010.We examine daily mean spatio-temporal variations in Aerosol Optical Depth (AOD) over 12 sites in the studied region extending towards the Himalayan Mountains.Furthermore, we characterize daily mean variations in more detailed aerosol parameters (i.e., Ångström Exponent (AE), aerosol Volume Size Distributions (VSDs), spectral variation in Single Scattering Albedo (SSA), ASYmmetry parameter (ASY), and Refractive Index (RI)) over Lahore and Kanpur.A classification scheme is discussed that is used to identify dominant aerosol types, which is based on AOD, AE, Extinction Ångström Exponent (EAE), and Absorption Ångström Exponent (AAE).Additionally, Aerosol Radiative Forcing (ARF) at the Top Of the Atmosphere (TOA), SuRFace (SRF), and within the atmosphere are reported over studied sites.Finally, backtrajectories calculated from the National Oceanic and Atmospheric Administration Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model are reported as a way to investigate the source region and pathways of air masses arriving at the studied sites during the haze episodes.

Study Area
The Himalayas are the highest mountain ranges in the world and give rise to major rivers in Asia.The Himalayan range runs west to east, from the Indus River valley in northern Pakistan to the Brahmaputra River valley in northern India and Tibet.The kingdoms of Nepal and Bhutan are located along the Himalayan ranges, and the Tibetan plateau lies to their north.Though India, Nepal, and Bhutan cover most of the Himalayan regions, Pakistan and China also occupy parts of them such as Kashmir, Kaghan, Kohistan, Deosai, and Chilas.
Lahore (31°15'00'' and 31°45'00'' North and 74°01'00'' and 74°39'00'' East) is one of the most densely colonized cities in Asia, especially in Pakistan.Lahore has an area of ~404 km 2 .It is the second largest industrial and commercial center in Pakistan.Vehicular emissions are considered to be a major contributor of atmospheric aerosol in this region (Alam et al., 2014a).Lahore features a hot semi-arid and monsoon climate, with low average precipitation levels.
Kanpur (26°28'00'' North and 80°24' 00'' East), located in the state of Uttar Pradesh, is the second largest industrial region in north India.This city is situated on the banks of river Ganga, covering an area of over 1642 km 2 .The climate of Kanpur is generally defined to be tropical monsoon, with long and hot summers and short and mild winters.Important sources of pollution include power plants, biomass burning, leather industries, and fossil-fuel combustion.

MODerate Resolution Imaging Spectroradiometer (MODIS)
Data are used from MODIS instruments on-board NASA's Terra and Aqua satellites that monitor columnar air pollution based on measurements in 36 different wavelength bands (0.41 µm-14.4µm).We use MODIS for quantification of columnar aerosol extinction, represented as AOD (Remer et al., 2005).The expected error in MODIS AOD is generally within ± (0.03 + 0.2 AOD MODIS ).This error represents the fusion of absolute (0.05) and relative (20%) uncertainties that arise from combined errors in assumed boundary conditions (e.g., surface reflectance, instrument calibration) and errors in assumed aerosol model types (such as in single scattering albedo) (Levy et al., 2010).
In this study, we use MODIS Terra Level-2 daily AOD data with 10 × 10 km spatial resolution.More detailed information on algorithms for the retrieval of aerosol and different cloud parameters is available at http://modisatmos.gsfc.nasa.gov.

AErosol RObotic NETwork (AERONET)
AERONET is a ground-based remote-sensing aerosol network established by NASA (Holben et al., 1998).This network is based on almost 400 permanent and temporary sites globally.AERONET provides data in three different levels, Level-1 (un-screened data), Level-1.5 (cloud-screened data) and level-2 (quality assured data), which can be retrieved from the AERONET website (http://aeronet.gsfc.nasa.gov/).Furthermore, it uses direct sun and diffuse sky radiances within the spectral ranges of 340-1020 nm and 440-1020 nm to provide important characteristics of atmospheric aerosol particles.The uncertainty AOD retrievals under cloud free conditions is < ± 0.01 for longer wavelengths (> 440 nm) and is < ± 0.02 for shorter wavelengths, which is less than ± 5 for the retrieval of sky radiance measurements (Dubovik and King, 2000).
In this study, the AERONET Level 2.0 daily AOD data were used along with AE, VSD, SSA, ASY, and RI for the super haze episode that occurred over the Himalayan region in October 2010.The data can be downloaded from the AERONET website (http://aeronet.gsfc.nasa.gov/).

Aerosol Optical Properties
The volume size distribution (VSD) is retrieved from the spectral sun and sky radiance data with 22 radius size bins in the range of 0.05-15 µm.All properties of aerosols depend on their size and thus this parameter is critical to characterize in order to study the behavior of atmospheric aerosols.The VSD has a two-mode structure that can be characterized by the sum of two lognormal distributions (Dubovik and King, 2000) as follows: where C v,i denotes the particle volume concentration, r v,i is the median radius, and σ i is the standard deviation (Dubovik and King, 2000).Depending on the aerosol type, for each particle radius bin within the 0.1-7 µm range, retrieval errors in dV(r)/dlnr typically do not exceed 15-35%.The errors for very large particles (7-15 µm) and very small particles (0.05-0.1 µm) may be as large as 35-100% for a given particle radius bin.SSA is the ratio of scattering coefficient to total attenuation coefficient, mathematically represented as, sca sca where ω  is the SSA, Q sca is scattering coefficient and Q ext is the total attenuation coefficient.SSA values range from zero (purely absorbing medium) to one (purely scattering medium).SSA is an important parameter in ARF calculations (Singh et al., 2010;Alam et al., 2014b).As far as the uncertainty between SSA and wavelength is concerned, SSA values can be influenced by particle morphology (shape and size) and its physical and chemical composition (Pandithurai et al., 2004).SSA is expected to have an accuracy of < 0.01 and uncertainty of 0.03-0.05depending on aerosol type and loading (Dubovik and King, 2000).The Ångström Exponent parameter (AE) is important for quantification of ARF (Ångström, 1961) and represents the wavelength (λ) dependence of AOD: The AE parameter varies with wavelength and is thus a qualitative indicator of particle size (Dubovik et al., 2002).High AE values represent smaller particles and vice versa.
Refractive index is an important optical property of aerosol particles and key indicator of aerosol chemical composition.The real part n(λ) of RI reflects aerosol scattering.The imaginary k(λ) part of RI is the representative of absorbing characteristics of aerosol particles.Values of RI used in models and calculations are often obtained from look-up tables such as in the Global Aerosol Data Set, where n values typically range from 1.5-1.53 at 532 nm with some exceptions (sulfuric acid, soot) (Koepke et al., 1997).Airborne measurements at 532 nm have shown that RI tends to be fairly constant (1.52-1.54)for numerous air mass types (Shingler et al., 2016).

Formation of Haze and Meteorological Situation
Haze occurred in the southern slopes of the Himalayan region in late October 2010 as evident from NASA Terra MODIS imagery on 19 October 2010 (Fig. 1).The haze stretched over hundreds of kilometers and reduced air quality in many places in India and Pakistan.The red dots in the figure are the hotspots areas, where MODIS detected unusually high surface temperatures.The hotspots were most likely agriculture fires, which contributed to the formation of haze in the region.
The variation in meteorological parameters such as temperature, relative humidity, visibility, wind speed, and wind direction during October 2010 are shown in Fig. 2. The maximum temperature ranged from 23 to 36°C and 29 to 34°C in Lahore and Kanpur, respectively.The minimum temperature in Lahore and Kanpur ranged from 17-25°C and 14-26°C, respectively.Likewise, the minimum/maximum relative humidity (%) was 18-57/73-100 and 29-66/82-100 over Lahore and Kanpur, respectively.The visibility ranged from 2-4 km during haze episodes over Lahore and Kanpur.The average wind speed varied from 2-16 m s -1 and 0-13 m s -1 over Lahore and Kanpur, respectively.The winds were predominantly southerly/southeasterly, and westerly over both Kanpur and Lahore.

Aerosol Optical Depth (AOD) and Ångström Exponent (AE) Variations during the Haze Episode
Variations in AOD during the October 2010 haze episode were studied using MODIS and AERONET data at a wavelength of 550 nm for 12 different Himalayan regions such as Ambala, Amritsar, Chandigarh, Kanpur, Ludina, Patiala, and Sheikhpura in India, in addition to Faisalabad, Gujranwala, Kasur, Lahore, and Sialkot in Pakistan.The maximum values of MODIS AOD in Ludina, Amritsar, and Lahore on 19 October were 4.2, 3.17, and 3.1, respectively (see Figs. 3(b), 3(h), and 3(i)).The maximum MODIS AODs (≥ 2.6) in Sheikhpura and Faisalabad were recorded on the 19 October (Figs. 3(d) and 3(k))).Over the study sites, the AERONET AOD on 22 October ranged from 1.08 to 2.25 with an average of 1.68 ± 0.33, while on 23 October, AOD decreased to be within the range of 0.7 to 1.01 with an average of 0.87 ± 0.1.These AOD values demonstrate the influence of atmospheric circulation on transporting pollution across the study region, including over heavily populated urban areas (Cao et al., 2015;Tao et al., 2016).
An increase in AOD values was also observed during pre and post monsoon seasons in India (Sinha et al., 2013), South Asia (Chung et al., 2002), and over Lahore (Alam et al., 2012;Ali et al., 2014).The average AOD 440 was higher over Pretoria, South Africa during the spring and summer seasons due to biomass burning and seasonal effects (Kumar et al., 2017b).Furthermore, the highest AOD values reported for New Delhi, India by Pandithurai et al. (2008) were between 0.55 and 1.18 during March-June 2006.It is worth emphasizing that, regions where the highest density of agricultural and industrial activities are located and where anthropogenic sulfate aerosol and soil particle emissions are consistently high throughout the whole year (e.g., southeast Asia, the central-eastern, southern and eastern coastal regions of China) experience annual mean AODs ranging from 0.50-1.80(Wang et al., 2008).
The global mean AOD values in different regions of the world have been found to be larger than 1.0 during different haze episodes (Xu et al., 2011;Yu et al., 2012;Bi et al., 2014).Based on the remote sensing observations during haze days (2011)(2012) in Beijing, Li et al. (2013b) found a large AOD value of 3.0 at 440 nm.The AOD values during the haze episode in Penang Island, Malaysia were found to be lower than AOD values for the Lahore region.
Our AOD values over Lahore were slightly higher than the values reported for Johor Bahru, Malaysia (Kanniah and Yaso, 2010).The AOD values reported by Yang et al. (2013) from Beijing, China are consistent with results of this study.AOD values for eastern China (Wang et al., 2013) and Beijing (1.78;Wang et al., 2009) are lower than the range obtained from our own results.Some of AOD values observed during different haze episodes conducted in past decades over different regions are listed in Table 1.
High values of AOD and AE on 18 October 2010 reveal the presence of fine particles over Lahore and Kanpur regions (Fig. 4).High AE values of 2.4 and 1.7 were observed over Lahore and Kanpur, respectively.However, on high AOD days over Lahore and Kanpur, AE values were 1.8 and 1.3 respectively, suggesting the presence of fine particles (Li et al., 2017;Salinas et al., 2013).Our results are consistent with those of Yu et al. (2011) over Beijing, as they observed high AOD (1.26) and high AE (1.11) during haze-fog days between 2001 and 2007.In this study, maximum AE values were found to be 1.41 and 1.33 on 18 and 19 of October, respectively, with an AOD of 1.7 and 2.4.The AE values of haze particles over eastern Asia and Korea have been found to be around 0.80 (Takamura et al., 2007;Kitakoga et al., 2014).Moreover, high AE values (> 1.45) were recorded by Jung and Kim (2011) over the Asian continent and downwind over South Korea during the severe haze episodes of October 2005.They found that the AE value of urban aerosols increased ~11% during the haze period.During the case study over the tropical Indian Ocean, Franke et al. (2003) reported that the AE (for short wavelengths 355/400/532 nm) ranged mostly from 1 to 1.6 for Southeast Asia aerosol particles.Over Lahore, Ali et al. (2014)

Volume Size Distribution (VSD)
Daily variation in VSD was investigated by using mean radii values between October 15-30 2010.Fig. 5 shows a bimodal distribution with peaks between 0.05 and 0.8 µm.Moreover, particles can be characterized in two groups: (i) particles with radii < 0.6 µm belonging to the fine mode, and (ii) particles with radii > 0.6 falling in the coarse mode category (Dubovik et al., 2002).The total, fine, and coarse mode particle volume variation was found to be 0.07-0.21µm 3 , 0.15-0.36µm 3 , and 0.05-0.32µm 3 , respectively, throughout the haze event.The fine mode effective radius in the event was about 0.05-0.09µm larger than during   non-hazy periods, while the coarse mode radius during the haze episode was about 0.1-0.41µm less than that during non-hazy periods over Lahore and Kanpur.The fine mode radii varied in the ranges 0.04-0.25,0.06-0.42,and 0.1-0.35µm across the entire event of October 15-30 (Fig. 5), while values of the coarse mode radii were about 3.11-6.46,2.5-6.39,and 2.09-5.39µm.The total, fine, and coarse mode particle volume variations were found to be 0.02-0.36,0.03-0.36,and 0.01-0.16µm 3 , respectively.In addition, the fine mode aerosol VSD exhibited a two to four-fold increase during the haze event as compared to non-hazy days.Li et al. (2013b) found that the contribution of black carbon and brown carbon to the total aerosol loading can reach between 25 and 38%, respectively, during haze/pollution events.The fine mode aerosol particles (0.05-0.65 µm) exhibited a significant increase in volume concentration during the hazy days (October 19-22), reflecting the presence of anthropogenic aerosol particles (fine particle) in the haze event as reported by Yu et al. (2016).Furthermore, these results are in strong agreement with the recent study reported by Wang et al. (2013) for haze/fog event that occurred in Nanjing, China in the winter of 2010, where they showed comparatively lower values for fine particles during the haze episode event.It should be noted that there were also some variations in the relationship, which is likely due to changes in the physical composition and size distribution of fine particles during the study period.Ye et al. (2011) found that the hygroscopic growth factor is also affected by aerosol size distribution or vice versa in Shanghai, China.

Single Scattering Albedo (SSA)
Fig. 6 shows the spectral variation in SSA before, during, and after the haze event for Lahore and Kanpur regions.SSA exhibits a decreasing shift during the heavy haze event (October 19-22, 2010) as compared to non-hazy days.This decreasing shift ranged from 0.93 to 0.90 for Lahore and from 0.96 to 0.91 for Kanpur.Nevertheless, the significant increasing shift in SSA (0.8-0.94) was observed for both areas after the event.In addition, average SSA values at 675 nm for the whole study period over Lahore and Kanpur were 0.95 ± 0.01 and 0.96 ± 0.01, respectively, which is suggestive of the influence of fine particles.Kitakoga et al. (2014) reported the mean SSA to be 0.96 ± 0.01 at the Fukue-jima Islands (2003)(2004), and the variation in SSA over Amami-Oshima Islands during yellow sand (Kosa) events to be 0.99-1 during April 2004, which is consistent with our results at Kanpur during the haze episode.Furthermore, our results are comparable with SSA values (0.89-91) reported from Beijing, China (Wang et al., 2009).Moreover, our SSA values over Lahore were found to exhibit close correspondence with SSA (0.89) derived from Hisar in India during haze conditions (Ramachandran et al., 2006).
Probably, the origin of the fine particles is local combustion and industrial activity (Che et al., 2009).SSA increases with the inverse of wavelength confirming that SSA is wavelength dependent.This decreasing tendency in scattering is primarily due to the absorption of solar radiation by specific aerosol constituents (Chameides et al., 1999).These results are consistent with previous observational and modeling studies by Tripathi et al. (2005)  The SSA reduction with increasing wavelength beyond 620 nm during the haze event most likely can be attributed to the dominance of absorbing fine aerosols particles (Singh et al., 2004).A similar trend was found by Sinha et al. (2013) over Hyderabad, India, by Li et al. (2013b) over Beijing, and by Xu et al. (2011) over Jinan, China.Furthermore, in many monitoring studies the calculated SSA (e.g., 0.87 ± 0.08 in Granada, Spain (Lyamani et al., 2010) and 0.72 ± 0.12 in Nagoya, Japan (Khatri et al., 2009)) reflect the presence of fine particles.As compared to other Asian regions, relatively low SSA values were obtained at the Goson supersite (0.88 ± 0.02) (Kim et al., 2005) and the downwind sites (0.88 ± 0.03) (Anderson et al., 2003) of Korea during biomass burning periods.

ASYmmetry Parameter (ASY)
This section deals with the spectral analysis of ASY (Figs. 7(a)-7(f)).The ASY parameter is the cosine averaged scattering angle, which is dependent on the diameter and chemical composition of aerosol particles.ASY generally fluctuates between -1 and 1 for purely backward to forward scattering cases.Furthermore, if the scattering is uniform, ASY is taken to be zero (Zege et al., 1991).The mean ASY between 440 and 1020 nm was ~0.69 during the haze period over Lahore and Kanpur.Two different features can be found when examining ASY as a function of wavelength (Figs.7(a)-7(f)).One feature is during the heavy haze period (Figs.7(b) and 7(e)), during which there is a decreasing trend in ASY with increasing wavelength.This decreasing trend varies from 0.78 at 440 nm to 0.62 at 1020 nm for Lahore, and from 0.74 at 440 nm to 0.63 at 1020 nm for Kanpur, which is indicative of the presence of fine particles and consistent with severe haze events over Beijing (Yu et al., 2016).An overall decrease in the ASY value was also observed over Kanpur by Bibi et al. (2016a), suggesting a relative abundance of fine aerosol particles.The other (Figs.7(c) and 7(f)) key feature is the leveling off of ASY with increasing wavelength above ~675 nm at values between 0.6-0.7.This alternating shift reflects the presence of both fine and coarse particles after the haze episode.ASY values for fine particles were shown to be smaller than that of coarse particles at fixed AOD (Bi et al., 2014) during heavy haze episodes of January 2013 over Beijing.The results from Figs. 7(b) and 7(e) also suggest that the ASY variation on 19-22 October was higher as compared to non-hazy days.However, it is slightly greater than the values (0.65) recorded in hazy conditions in Beijing, China ( Wang et al., 2009) and (0.66) during the summer haze event (2011) at Tanggu, China (Zhang et al., 2013).Herber et al. (2002) conducted a climatological study of Arctic haze events between 1991-1999 in Spitsbergen (Norway).They found the mean ASY to be 0.62 during summer, and 0.81 for Arctic spring haze events, which is the highest recorded value among the investigated haze episodes in the past literature, while the former (0.62) is consistent with this study's results.
The difference in daily variations in ASY (see Fig. 7) on hazy days at a shorter wavelength (440 nm) was found to be in the range of 0.003-0.019.Additionally, the differences in daily variations (in ASY at 1020 nm on hazy days) were in the range of 0.001-0.02.The ASY 440 in the hilly area of Pune (India) was recorded to be 0.70 during dry weather conditions (2001)(2002) (Pandithurai et al., 2004), 0.64 at Chennai during 2001 (Ramachandran, 2005), and 0.65 in Beijing during haze episodes between 2001-2007 (Yu et al., 2011).All the previous results, including our measurements, indicate the presence of absorbing aerosols, presumably mainly due to black carbon, in these Asian regions.

Refractive Index (RI)
Table 2 presents the daily mean RI during haze events in October (15-30, 2010) over Lahore.The imaginary k(λ) parts for all cases exhibit relatively little change, with the daily mean values at four different spectral variations concentrated between 0.001 and 0.01 over Lahore.Results for the Kanpur region also showed the same statistical behavior (see Table 3).Compared to k(λ), n(λ) parts exhibited spectral variation, with the 675 nm value significantly higher than that of other bands.The retrieval results from a total of four haze days  showed that n(λ) values were higher for longer wavelengths, reflecting the predominance of fine (soot) aerosols, which are mostly considered to be absorbing particles.In addition, the mean values of the real part at 675 nm varied between 1.41-1.51for Lahore and between 1.49-1.52 for Kanpur city during the haze event (19-22 October).These results are in strong agreement with the findings of Li et al. (2013b) having n(λ) values between 1.44 and 1.51 for two winter heavy haze events in 2011 and 2012 in Beijing, China.However, at 675, 870, 1020 nm, n(λ) was found to be in the range of 1.41-1.42and 1.42-1.46at Lahore and Kanpur, respectively, while for non-hazy days (15-18 and 23-30 October 2010), n(λ) exhibited variation between 1.54-1.56for both regions.This shows that coarse particle (dust) concentration was also present during the haze event (Yu et al., 2016).Furthermore, k(λ) decreased after the hazy days (23)(24)(25)(26)(27)(28)(29)(30).This decrease in k(λ) after hazy days was also documented by Wang et al. (2013) at Nanjing, China, during the haze episode in winter 2010.The enhanced absorption (larger k(λ)) at 440 nm reveals the presence of spectrally dependent absorbing particle types (e.g., black carbon, brown carbon) during the haze period (Yu et al., 2016).In addition, the mean value of the n(λ) was found to be 1.45 ± 0.02 for haze event during (19-21 October), which is consistent with the values reported by Noh et al. (2009) for anthropogenic aerosol particles.They observed mean n(λ) values for anthropogenic aerosols as 1.44 ± 0.01 during haze episodes at Gwangju, Korea in October 2005.Müller et al. (2006) reported n(λ) for urban haze particles at two different sites of China.They found the mean value of 1.57 ± 0.11 and 1.62 ± 0.11 in the Pearl River Delta (South China) and in the urban city Beijing (North China), respectively.Forestfire smoke particles transported from North America to Leipzig (Germany) also showed a higher value (1.52 ± 0.01) (Müller et al., 2005).Furthermore, the effective mean n(λ) value obtained in this study was slightly lower, but still comparable with those from the forest fire smoke particles in the North American region (1.47 ± 0.07) (Muller et al., 2005).The n(λ) obtained in this study was even larger than those reported by Herber et al. (2002) during 1991 to 1999.They found a mean n(λ) value of 1.42 during Arctic haze (spring) episodes in the Spitsbergen (Norway).
In contrast to the results for n(λ), k(λ) was quite different Table 2. Daily mean RI (real and imaginary) with corresponding standard deviations for Lahore.

Classification of Aerosol Types
Different aerosol types originate from a various of sources having diverse physical and optical properties, leading to different atmospheric impacts (Dubovik et al., 2002).Classification of aerosol data into aerosol types is important as models often conduct calculations of aerosols based on their stratification into specific air mass types.Prevalent aerosol types can be characterized based on relationships between aerosol absorption and size parameters (Giles et al., 2011).Threshold values for the aforementioned aerosol parameters need to be selected to determine aerosol types (Giles et al., 2011;Mishra and Shibata, 2012;Kumar et al., 2015;Adesina et al., 2017;Bibi et al., 2017;Kumar et al., 2017a).The selection of threshold values depends on location, aerosol characteristics, and values of AOD (Tariq et al., 2016).The correlation of AOD with AE is a commonly used clustering technique for classification of different types (Sharma et al., 2014;Kumar et al., 2015;Verma et al., 2015;Tariq et al., 2016;Bibi et al., 2016b).Likewise, the classification of aerosols can be achieved by other clustering techniques through the correlation of different optical properties such as AAE and EAE.AAE is a function of aerosol composition (Mishra and Shibata, 2012) and EAE is an indicator of particle size (Russell et al., 2010).
For the identification of aerosol types during high pollution episodes, the selected threshold values of AOD and AE were as follows: biomass burning and urban/industrial = 0.7 < AOD < 4 and 0.9 < AE < 1.6; background conditions = 0 < AOD < 0.5 and 0.4 < AE < 1.6.The leftover scattered points were classified as mixed type aerosols.Fig. 8(a) and 8(b) shows the relationship between AOD and AE, in which the specific cluster indicates the different types of aerosol in October over Lahore and Kanpur.Biomass burning and urban/industrial aerosol types were prominent during the haze epsidoe over Lahore and Kanpur.Mixed type aerosols were observed during, before, and after the haze episode and on some of the haze days owing to a mixture of natural and anthropogenic aerosols (Yu et al., 2016).Recently, Tariq et al. (2016) reported that the pronounced aerosol type during haze episode was biomass burning and urban/industrial over Lahore.
Figs. 9(a) and 9(b) illustrates the relationship between AAE and EAE over Lahore and Kanpur in October 2010.Three significant types of aerosol which were classified as follows: biomass burning and urban/industrial = 1.0 < EAE < 1.4 and 1.0 < AAE < 2.0; background conditions = 0.5 < EAE < 0.8 and 1.2 < AAE < 2.0.The mixed type aerosols were again observed before, during, and after the haze episode resulting from the mixture of different size particles.It is observed that during the haze episode, the biomass burning and urban industrial aerosols were prominent, which highlights the significant anthropogenic activities.Similarly, Che et al. (2015) carried out AAE and EAE clustering technique to sort the aerosol types into mixed, urban/industrial and biomass burning during the heavy haze period in Beijing.Previously, Alam et al. (2012) grouped the aerosol into two categories (mineral dust and urban/industrial) using the same technique over Karachi and Lahore.

Aerosol Radiative Forcing (ARF)
The ARF at the TOA, SRF, and within the atmosphere is defined as the difference in the net solar irradiance with and without aerosol in the atmosphere.Here the net irradiance means the overall (down minus up) radiative fluxes including short and longwave radiation.The differences of the TOA and SRF give rise to the net ARF in the whole atmosphere.The Santa Barbra DISORT Atmospheric Radiative Transfer (SBDART) code, developed by Ricchiazzi et al. (1998), is used for radiative transfer calculations.This model is used to conduct a wide variety of radiative transfer calculations in the short wavelength (0.3 to 4.0 µm) and long wavelength (4-100 µm) regions.The main input parameters essential in SBDART model are AOD, SSA, ASY, columnar water vapour, ozone, and surface albedo.Another input parameter is the solar zenith angle, which is computed using the SBDART model by identifying a particular latitude, longitude, date, and time.The uncertainties in calculating ARF may arise from differences in the mid-latitude atmospheric dynamics and the real atmosphere, uncertainties in the surface reflectance, modeled AOD, and SSA measurements.
In the present analysis, the total flux at the TOA, SRF, and in the atmosphere, was computed separately in the wavelength range of 0.3-4.0µm using the SBDART model.The atmospheric heating rate due to aerosol can be calculated as suggested by Liou (2002): where ∂T/∂t is the atmospheric heating rate in K day -1 , g is the acceleration due to gravity (9.8 m s -2 ), C p is the specific heat capacity of air at constant pressure (i.e., 1006 Jkg -1 K -1 ) and ∆P (300 hPa) is the atmospheric pressure difference between top and bottom layers of the atmosphere.Figs.10(a) and 10(b) summarizes the ARF and atmospheric rate values in October 2010 over Lahore and Kanpur.The ARF at TOA, SRF, and in the atmosphere were found to be in the range of -17.6 to -81.6, -64 to -193 and +47 to  +119 W m -2 , respectively, for Lahore and 15.32 to -91.6,  -38 to -134 and +33 to +75.91 W m -2 for Kanpur.The mean daily variation in ARF during the haze event (19-22 October) at the SRF was on the order of -162 ± 21 W m -2 and, at the TOA, it was about -62 ± 5 W m -2 giving rise to an atmospheric ARF of 99 ± 15 W m -2 .The ARF in the atmosphere after the haze episode (23-30 October 2010) was found to be a maximum (i.e., 114 ± 31 W m -2 ), while the minimum value (91 ± 25 W m -2 ) occurred before the episode started (15-18 October 2010).Generally, the values of atmospheric heating rate during the haze event (19-22 October) were higher, reaching maximum values of 2.1 K day -1 over Kanpur on 23 October and 3.3 K day -1 on 21 October over Lahore.These high values of atmospheric heating rate represent the strongest atmospheric absorption during the haze event.After the haze event, relatively lower values of atmospheric heating rate were observing indicating less absorption.
Similarly, Bi et al. (2014) reported that the overall mean ARF at SRF and TOA were found to be in the range of -57.0-9.9 and -30.6-4.1 W m -2 , respectively, leading to atmospheric ARF values of 4.5-26.4W m -2 during January 2013 over Beijing.Singh et al. (2010) found the ARF at the TOA to be between -1.4 and 21 W m -2 and at the SRF it was found to be between -46 and -110 W m -2 for the Delhi region, whereas in the atmosphere it was in the range of 46 to 115 W m -2 throughout the year 2007.Overall, the mean ARF values at both TOA and SRF have been found to be negative by the previous assessment at global and regional basis (Yu et al., 2006;Kim et al., 2008a;Ge et al., 2010).Generally, in the case of negative forcing the variation in mean ARF is more significant at the shorter wavelength, while in case of positive forcing the rate of change in the mean ARF is most visible at the longer wavelengths.In the most recent analysis in Asian regions, it has been reported that positive forcing are generally influenced mostly by coarse mode aerosols, whereas negative forcing are more due to fine mode particles (Alam et al., 2014b).Yu et al. (2016) noticed that the mean ARF values were -112.29 and -58.61W m -2 at SRF during haze and non-haze days, respectively, whereas values were -45 and -18 W m -2 at TOA leading to atmospheric ARF of 66.51 and 39.97 W m -2 during January 2013 over Beijing.They also found the corresponding atmospheric heating rate of 1.61 ± 0.48 K day -1 and 1.12 ± 0.31 K day -2 , during haze and non-haze events, respectively.Jose et al. (2015) pointed out that hazy conditions were characterized by higher negative ARF values at the SRF (-87.82W m -2 ) as compared with normal days (-53.90W m -2 ), giving rise to large positive atmospheric ARF (30.56 W m -2 ) during haze day.For Karachi, Alam et al. (2012) found seasonally positive atmospheric ARF (40.6 ± 2.3 W m -2 for winter and 43.3 ± 8.4 W m -2 for summer seasons).Nevertheless, the sign of ARF changing to positive over the same region is mainly due to lower SSA, which consequently cause atmospheric absorption and reflect higher surface albedo.The ARF values reported from the northern region of China by Ge et al. (2010) at the SRF were -8 and -35 W m -2 , which are much lower than our calculated results.

Identification of Aerosol Transport and Source Regions
In order to analyze the probable transport routes and potential source regions of air masses, air-mass backtrajectories were computed using the HYSPLIT model (Draxler and Rolph, 2003).Fig. 11 shows the 5-day back trajectories of air masses at 500, 1000 and 1500 m for selected days (before, during, and after the haze period) over the studied sites.The results suggest that before the haze episode, the majority of the trajectory pathways originated from local sources.During hazy days, most of the air masses were arriving from local sources (Pakistan and India), but some of the trajectories also originated from Afghanistan.Gautam et al. (2007) and Sharma et al. (2014) reported that high AOD during haze periods were mostly because of local and/or regional pollution originating from anthropogenic activities and biomass burning over India.Tariq et al. (2016) also noted that the transport of aerosols from near southwest burning areas to Lahore were mostly due to winds at 1000 m above the ground during haze episode.Likewise, Yu et al. (2016) documented that local and regional aerosol sources (e.g., industrial emissions, biomass burning, and vehicular emissions) are mainly responsible for high aerosol loading during haze episodes over Beijing.

CONCLUSIONS
Aerosol characteristics, including their optical and radiative properties, have been analyzed during hazy and non-hazy days of October 2010 over the Himalayan region using MODIS and AERONET data.The variations in AOD during the studied period were examined for Ambala, Amritsar, Chandigarh, Kanpur, Ludina, Patiala, and Sheikhpura in India and Faisalabad, Gujranwala, Kasur, Lahore, and Sialkot in Pakistan.Additionally, aerosol optical and radiative properties were also analyzed over Lahore and Kanpur.The major conclusions of the present analysis are: • AOD values reached a maximum on 19 October at observational sites in India and Pakistan.These values exceeded 2.03 over Ambala, Sialkot, Faisalabad, Amritsar, and Ludhiana.Similarly, the highest AOD over Gurdaspur, New Delhi, Batala, Bathinda, Kanpur, and Lahore on 20 October exceeded 1.61.• The fine mode radius in the haze event was ~0.05-0.09µm larger than non-hazy periods, while the coarse mode radius during the haze episode was about 0.1-0.41µm less than that during non-hazy periods for both Lahore and Kanpur.
• SSA shows an increasing trend with respect to wavelength (440-1020 nm) after the haze event (23-30 October 2010) as compared to hazy days (19-22 October 2010) over Lahore and Kanpur.• Two different relationships were observed between ASY and wavelength.Firstly, there was a reduction in ASY as a function of increasing wavelength, which was associated with the heavy haze period.This trend varied from 0.78 at 440 nm to 0.62 at 1020 nm for Lahore and from 0.74 at 440 nm to 0.63 at 1020 nm for Kanpur, suggestive of the presence of fine particles.Secondly, ASY initially decreases with increasing wavelength and then levels off between 0.6-0.7 with the increasing wavelength from 675-1020 nm.• The relationship between AOD and AE reveals that biomass burning and urban/industrial aerosol types are prominent during haze episodes.• ARF values were computed during hazy and non-hazy days.The shortwave ARF values at the TOA, SRF, and within the atmosphere were found to be in the range of -17.6 to -81.6, -64 to -193, and +47 to +119 W m -2 , respectively over Lahore.In contrast over Kanpur, the ARF values were in the range of 15.32 to -91.6, -38 to -134, and +33 to +75.91 W m -2 , respectively.• Maximum atmospheric heating rates were observed during the hazy days in both Lahore and Kanpur.
• HYSPLIT airmass back-trajectory analysis showed that most of the air masses were from loacal sources.

Fig. 1 .
Fig. 1.MODIS Terra imagery data over the study area on 19 October 2010 showing the severe haze episode.The AERONET stations are located at Lahore and Kanpur.
found the highest value of AE in winter (1.20) during December 2009 to October 2011.They found AE values ranged from 0.52 ± 0.23 in June to 1.27 ± 0.07 in December with an annual mean difference of 1.04, which exceeds values recorded in this study.

Fig. 4 .
Fig. 4. Daily variations in AERONET AOD at 500 nm and AE (440-870) over (a) Lahore and (b) Kanpur during the haze episode in October 2010.The haze days are marked with circle in each figure.

Fig. 6 .
Fig. 6.Spectral variations of AERONET SSA for (a-c) Lahore and (d-f) Kanpur before, during, and after the severe haze episode.

Fig. 7 .
Fig. 7. Daily variations in AERONET ASY as a function of wavelength before (a, d), during (b, e) and after (c, f) the haze episode.

Fig. 8 .
Fig. 8. Scattered plot for AOD and AE shows the clusters of aerosol types over (a) Lahore and (b) Karachi during haze episode.

Fig. 9 .
Fig. 9. Scattered plot for EAE and EAE shows the clusters of aerosol types over (a) Lahore and (b) Karachi during haze episode.

Fig. 10 .
Fig. 10.Daily averaged variation of ARF at TOA and SRF along with associated atmospheric heating rate over (a) Lahore and (b) Kanpur during the month of October 2010.

Fig. 11 .
Fig. 11.5-day HYSPLIT back trajectories representing the origins and pathways of air masses at 500, 1000, and 1500 m above ground level (AGL) before (a) during (b) and after (c) haze episode over Lahore and Kanpur.

Table 1 .
AOD values observed in earlier studies during haze episodes over various locations.