A climatology of aerosol optical and microphysical properties over the Indian subcontinent from 9 years (2000–2008) of Multiangle Imaging Spectroradiometer (MISR) data

Authors


Abstract

[1] We present the first detailed analysis of a 9 year (2000–2008) seasonal climatology of size- and shape-segregated aerosol optical depth (AOD) and Ångström exponent (AE) over the Indian subcontinent derived from the Multiangle Imaging Spectroradiometer (MISR). Our analysis is evaluated against in situ observations to better understand the error characteristics of and to corroborate much of the space-time variability found within the MISR aerosol properties. The space-time variability is discussed in terms of aerosol sources, meteorology, and topography. We introduce indices based on aerosol size- and shape-segregated optical depth and their effect on AE that describe the relative seasonal change in anthropogenic and natural aerosols from the preceding season. Examples of major new findings include the following: (1) winter to premonsoon changes in aerosol properties are not just dominated by an increase in dust, as previously thought, but also by an increase in anthropogenic components, particularly in regions where biomass combustion is prevalent; (2) ∼15% of the AOD over the high wintertime pollution in the eastern Indo-Gangetic basin is due to large dust particles, resulting in the lowest AE (<0.8) over India in this season and likely caused by rural activities (e.g., agriculture, etc.) from the densely populated rural area; (3) while AOD decreases from the Indo-Gangetic basin up to the Tibetan Plateau, a large peak in AE and the fraction of AOD due to particle radii <0.7 μm exists in the foothills of the Himalayas, particularly in the premonsoon season; and (4) the AOD due to nonspherical particles exhibits a strong ocean-to-land gradient over all seasons because of topographical and meteorological controls.

1. Introduction

[2] Understanding and quantifying the climatic effects of aerosols have made significant progress since the International Panel on Climate Change (IPCC) first assessment report in 1990. The estimates have improved significantly in the last two decades because of integration of in situ and satellite observations of aerosol optical and microphysical properties with model simulations, but still the level of uncertainty is far less than that of greenhouse gases [Forster et al., 2007]. Part of the difficulty in further reducing the level of uncertainty lies in the extreme heterogeneity in aerosol optical and microphysical properties over a wide range of spatial and temporal scales, of which we have incomplete knowledge. Thus it is necessary that we continue to improve our characterization of aerosols over all regions of the globe with high spatial and temporal frequency, particularly in regions with high population where the impact of aerosols on human health is also a concern.

[3] The Indian subcontinent is one region requiring improvements in characterization of aerosol properties because of an insufficient number of in situ observations and high aerosol loading. More than one billion people (one sixth of the world's population) living on the Indian subcontinent are exposed to enormous pollution. The region's economy depends heavily on agriculture, which in turn depends on rainfall that is mostly confined to the monsoon season (June–September). High aerosol loading over the region can affect precipitation [Ramanathan et al., 2001], the degree to which remains an active area of research that also requires improved aerosol characterization.

[4] Investigation of aerosols in India started as early as the 1960s, when Mani et al. [1969] studied Ångström turbidity from solar radiance measurements. In 1985, a multiwavelength radiometer was developed by the Indian Space Research Organization (ISRO) and deployed successfully at Trivandrum [Moorthy et al., 1999] to measure spectral aerosol optical depth (AOD) and, in the same year, aerosol vertical distribution measurements by ground-based lidar was initiated at Pune [Devara et al., 2002]. These events marked the beginning of systematic ground-based measurements of aerosol properties in India. The Indian Ocean Experiment [Ramanathan et al., 2001] first addressed the issue of climatic effects of aerosols in this region in a comprehensive manner, which led to enhanced, coordinated efforts for in situ multi-instrument and multiplatform measurements by ISRO under the Geosphere Biosphere Program (GBP) [Moorthy et al., 2005; Tripathi et al., 2006; Nair et al., 2008]. ISRO has conducted several focused campaigns in recent years over India and its surrounding oceans to characterize the chemical, microphysical and optical properties of aerosols. Many ship-based campaigns have been conducted in the Arabian Sea [Babu et al., 2008; Satheesh et al., 2006a and references therein] and the Bay of Bengal [Nair et al., 2008; Satheesh et al., 2006b and references therein] over multiple seasons. Two ground-based observatories were set up to collect data continuously: Port Blair in the Bay of Bengal [Moorthy and Babu, 2006] and Minicoy in the Arabian Sea [Vinoj et al., 2008] (Figure 1). Over the mainland, continuous in situ measurements were initiated as part of the ISRO-GBP network (Figure 1), but they are mostly restricted to measurements of spectral aerosol optical depth (AOD), that too at very few places. Good information about aerosol optical and microphysical properties from long-term measurements is available in the literature only at ten sites within the interior of the Indian subcontinent (Figure 1), namely Trivandrum [Moorthy et al., 2007a and references therein], Pune [Devara et al., 2008 and references therein], Kanpur [Dey and Tripathi, 2007, 2008], Nainital [Dumka et al., 2008 and references therein], Ahmedabad [Ganguly et al., 2006a], Hyderabad, Anantpur [Badarinath et al., 2009 and references therein], Bangalore [Satheesh et al., 2006c], Visakhapatnam [Madhavan et al., 2008 and references therein] and the Atmospheric Brown Cloud (ABC) observatory at Godavari near Katmandu in Nepal [Ramanathan and Ramana, 2005; Adhikary et al., 2007, 2008]. Long-term in situ observations at ISRO-GBP sites along with AErosol RObotic NETwork (AERONET) measurements at Kanpur and Hanimaadhoo, and the ABC observatory in Nepal (Figure 1) have provided deep insight into the seasonal variability of aerosol optical properties [Devara et al., 2002; Singh et al., 2004; Sagar et al., 2004; Ganguly et al., 2006a; Adhikary et al., 2007; Moorthy et al., 2007a; Dey and Tripathi, 2008]. For example, long-range transport of desert dust during the premonsoon season (March–May) has been found to affect regional aerosol optical properties [Dey et al., 2004; Prasad and Singh, 2007], and mixing of dust with anthropogenic particles [Chandra et al., 2004; Dey et al., 2008] has made the characterization of the aerosols more difficult. The presence of absorbing dust at elevated heights over the Indo-Gangetic basin (IGB) and Tibetan Plateau has raised several climatic issues [Lau et al., 2006; Moorthy et al., 2007b; Satheesh et al., 2008], while detailed radiative studies [Kahnert et al., 2007; Mishra et al., 2008] have revealed the importance in considering the nonsphericity of these dust particles in quantifying their radiative effects. Unfortunately, observations of the nonsphericity of particles are missing in the studies discussed above.

Figure 1.

Ground-based long-term measurement sites shown on a topographic map of the Indian subcontinent. The major geographic regions (names in red) considered for climatological mean (± standard deviation) values shown in Table 2 are delineated by black solid lines.

[5] Besides these continuous measurements, aerosol microphysical and optical properties were measured in India during four major field campaigns in the last ten years. The first land campaign was carried out in south India during February–March 2004 [Moorthy et al., 2005] followed by a second campaign focused in the IGB during the winter season (December–February) of 2004–2005 [Tripathi et al., 2006; Nair et al., 2007]. The major scientific objective of these two campaigns was to quantify aerosol direct radiative effects at several locations by using simultaneous in situ measurements of aerosol microphysical, chemical and optical properties. The Integrated Campaign on Aerosol and Radiation Budget (ICARB) was carried out in the premonsoon season of the year 2006 to measure the physicochemical and radiative properties of aerosols and trace gases and their vertical distributions [Satheesh et al., 2009]. In the premonsoon and monsoon seasons of the years 2008 and 2009, aircraft and ground-based measurements were carried out over the IGB and central India as part of Continental Tropical Convergence Zone campaign to quantify the aerosol indirect effect [Department of Science and Technology, 2008].

[6] All of these studies have greatly improved our understanding of aerosol microphysical and optical properties and their variability at several locations across the Indian subcontinent and its surrounding waters; however, a complete spatial analysis remains hindered because of the spatially limited nature of these data sets. This is where satellite data become very useful and can complement the in situ measurements. Satellite retrievals of aerosol properties over land have only been available in recent years and a few studies have been done using these data over the Indian subcontinent. Di Girolamo et al. [2004] were the first to study the spatial distribution of AOD over India using Multiangle Imaging Spectroradiometer (MISR) in the winter season during 2001–2004, where they were able to explain the enormous pollution observed over the IGB based on meteorology, topography and potential aerosol emission sources. All subsequent studies using Moderate Resolution Imaging Spectroradiometer (MODIS) data have confirmed this observation [Jethva et al., 2005; Prasad et al., 2006; Ramachandran and Cherian, 2008] with additional information on the seasonal variability of AOD and fine mode fraction, to some extent.

[7] Here, in a continued effort to develop an integrated picture of the spatial and temporal characteristics of aerosol properties over the Indian subcontinent, we analyze 9 years (March 2000 to November 2008) of MISR data. In addition to AOD, MISR is able to provide information on particle optical and microphysical properties based on its multiangle and multispectral capabilities, under good but not necessarily ideal viewing conditions [Kahn et al., 2001, 2007]. Particle properties include Ångström exponent (AE), optical depths segregated by particle size and shape (i.e., AOD fractions of “fine,” “medium” and “large” particles and “spherical” and “nonspherical” particles) and single scattering albedo (SSA) in two-to-four groupings. The retrieval of many of these aerosol properties from a space-based passive sensor is unique to MISR, thus offering a first opportunity to examine their spatial and temporal characteristics. In particular, we seek to address the following questions: (1) How do the aerosol optical and microphysical properties vary spatially and seasonally over the Indian subcontinent? (2) Is the observed space-time variability segregated by aerosol property (e.g., large and nonspherical versus small and spherical particles)? (3) Can the observed variability be understood in terms of the variability in meteorology and aerosol emission? Section 2 evaluates the quality of the MISR aerosol product and the analysis performed to derive the seasonal climatology. Section 3 provides an overview of the associated factors such as topography, meteorology, emission inventories and demography, which influence the seasonal climatology and interannual variations of aerosol properties. The seasonal climatology of aerosol properties is discussed in section 4 followed by implications of the results and a summary of major conclusions in section 5.

2. MISR Aerosol Properties

[8] MISR, onboard the NASA Earth Observing System's Terra spacecraft, is in a Sun-synchronous orbit that crosses the equator at ∼1030/2230 local time. It is a push broom imaging instrument operating at four spectral bands centered at 446, 558, 672 and 867 nm, in each of nine separate cameras oriented along the orbital track with surface viewing zenith angles ranging from ±70.5° [Diner et al., 2008]. Its ∼400 km swath provides global equatorial coverage every 9 d and 2 d coverage at polar latitudes. Aerosol retrievals are performed on 16 × 16 patches of 1.1 km subregions, yielding an aerosol product at 17.6 × 17.6 km spatial resolution, referred to as a “Level 2” product [Martonchik et al., 2002].

[9] Kahn et al. [2009] have provided an excellent overview and details on the MISR Level 2 aerosol product. In brief, MISR performs retrievals of AOD, SSA, the fraction of AOD due to “fine” (particle radii <0.35 μm), “medium” (particle radii between 0.35 and 0.7 μm) and “large” (particle radii >0.7 μm) particles as well as the fraction of AOD due to “spherical” and “nonspherical” particles at the four MISR spectral bands, and AE based on the retrieved spectral AOD. Throughout this manuscript, the discussion of the relative size categories strictly corresponds to the ranges given above. This may differ from other literature; for example, the term “fine” mode often refers to all submicron particles [O'Neill et al., 2001]. In the MISR Version 22 retrieval algorithm, the measured radiances are matched with simulated radiances from 74 types of modeled aerosol mixtures. In each individual mixture type, the aerosol is assumed to be an external mixture of various proportions of pure components, each pure component having fixed microphysical properties and uniform composition [Diner et al., 2008]. The particle properties of the mixture with the lowest chi-square value out of all successful mixtures passing the threshold test during matching with measured radiances are reported as “best estimates” within the product file. The terminologies used in the Level 2 aerosol product to represent various aerosol parameters are listed in Table 1. The additional information in the Level 2 product about the algorithm type flag, surface type flag, retrieval success flag and best estimate quality flag (Table 1) help in understanding the quality of the aerosol properties at a particular region; hence, they are important to keep track of in scientific analysis [Kahn et al., 2009]. Here, we have analyzed the latest version (Version 22) of the Level 2 aerosol product of MISR from March 2000 to November 2008, which includes AOD at 558 nm, AE, optical depth fractions of fine (ff), medium (fm), large (fl), spherical (fsp) and nonspherical (fnsp) particles, and spectral SSA. Optical depth due to a particular category has been derived by multiplying the AOD with the corresponding optical depth fraction reported in the product file (compare Table 1).

Table 1. MISR Level 2 Version 22 Aerosol Products at 17.6 × 17.6 km Resolution Used in This Studya
ProductDescription
RegBestEstimateSpectralOptDepthAOD at four wavelengths
RegBestEstimateÅngströmExponentAE (derived from retrieved AOD at four wavelengths)
RegBestEstimateSpectralSSASSA at four wavelengths
RegBestEstimateSpectralOptDepthFractionFraction of optical depth due to “small” (r < 0.35 μm), “medium” (0.35 < r < 0.7 μm), “large” (r > 0.7 μm), “spherical” and “nonspherical” particles
RegBestEstimateQAInformation about the best estimates reported in previous rows (0 = estimates from one successful mixture, 1 = more than one successful mixture, 2 = one or more successful mixtures in nearby region, 3 = insufficient information for estimate)
RegParticlePropertyQAQuality of the estimated particle property (0 = good quality, 1 = poor quality)
RegClassIndRegion type during retrieval (0 = clear region, 1 = solar oblique region, 2 = topographically complex region, 3 = cloudy region, 4 = no valid data in region)
RegSurfTypeFlagSurface type during retrieval (0 = dark nonpolar water, 1 = shallow/coastal nonpolar water, 2 = nonpolar land, 3 = polar dark water, 4 = polar shallow/coastal water, 5 = polar land)
AerRetrSuccFlag1 = no success matches with aerosol models, 2 = no potential matches, 3 = retrieval algorithm failure, 4 = retrieval not attempted, 5 = insufficient data to perform retrieval, 6 = inadequate scene contrast to perform retrieval, 7 = successful aerosol retrieval, 8 = unsuccessful aerosol retrieval, default value reported
NumSuccAerMixtureNumbers of mixtures passing chi-square test

2.1. Quality Assessment

[10] MISR-derived AOD has been validated globally against coincident AERONET observations [Kahn et al., 2005], showing that 63% of MISR-retrieved midvisible AOD values fall within 0.05 or 20%, and about 40% fall within 0.03 or 10% of AERONET AOD. Validating particle property is an ongoing MISR science team effort [Kahn et al., 2009]. An up-to-date quality statement is available at http://eosweb.larc.nasa.gov/PRODOCS/misr/Quality_Summaries/L2_AS_Products.html. According to the quality statement, the ongoing validation efforts of the MISR aerosol product using a combination of AEROENT sunphotometer and field campaign data suggest that spherical versus nonspherical particle distinction is the most robust property retrieval in the Version 22 product, while particle size and AE retrievals show mixed results. Here, we contribute to this validation effort by examining the quality of the MISR-retrieved aerosol properties with coincident data from the Kanpur AERONET station (Figure 1) in India, which has been operating for more than 8 years. Although not all of the MISR-retrieved aerosol properties can be directly validated by AERONET retrievals, the comparison does provide enough information to understand the MISR data quality that is needed to guide our analysis. Further evaluation of the quality of the particle properties is done by examining the logical spatial and temporal behavior of the particle properties with meteorology, topography and emission sources [cf. Di Girolamo et al., 2004], as well as with published knowledge gained through in situ measurements at ground-based sites (Figure 1). MISR-retrieved aerosol properties at 17.6 × 17.6 km spatial scale (i.e., one MISR pixel at Level 2) surrounding the Kanpur AERONET station are compared with AERONET-retrieved Level 2 cloud-screened and quality assured aerosol properties [Holben et al., 1998] within a 2 h time window centering the MISR overpass time [after Kahn et al., 2005] for data collected between February 2001 and November 2008. Altogether we have 339 d of MISR overpass over Kanpur during this period, but only 69 d for the comparison because either MISR or AERONET did not have successful retrievals on the remaining days, most likely because of cloud or very thick haze.

[11] We find very similar results to other AOD comparison studies (Figure 2a), where MISR AODs are found to be biased low relative to AERONET AODs [Di Girolamo et al., 2004; Kahn et al., 2005, 2007]. This bias increases with increasing AOD [Kahn et al., 2005], and we can expect the AOD to be underestimated in the high AOD regions at midvisible wavelength as discussed by Di Girolamo et al. [2004]. The bias in MISR-retrieved AOD compared to AERONET retrieval (ΔAODMISR-AERONET) at Kanpur is quantified by the relation

equation image

Higher SSA assumed in MISR aerosol retrieval algorithm compared to the real particle SSA contributes to the systematic bias in this region, because higher AOD is required to match the observed top-of-atmosphere reflectance as particle SSA decreases [Kahn et al., 2005, 2009; Chen et al., 2008].

Figure 2.

Scatterplot between (a) MISR-retrieved AOD558 and AERONET-retrieved AOD500 (the dashed lines are 1:1 line and envelope of ±0.05 or 20% × AOD, whichever is larger), (b) MISR-retrieved and AERONET-retrieved AE, (c) MISR-retrieved nonspherical fraction of midvisible AOD (MISR-fnsp) and AERONET-retrieved AE (AEAERONET), (d) MISR-retrieved small particle fraction of midvisible AOD (MISR-fs) and AERONET-retrieved fine mode fraction (AERONET-fmf), and (e) MISR-retrieved SSA at 446 nm and AERONET-retrieved SSA at 440 nm over Kanpur in the Indo-Gangetic basin for data during February 2001 to November 2008.

[12] AE, the spectral dependence of AOD, provides a firsthand qualitative approximation about the dominant size of the particles, with higher AE implying dominantly small particles and vice versa [Ångström, 1929]. MISR-retrieved AE is found to show good (and significant) correlation (R = 0.64) with AERONET retrievals as shown in Figure 2b, but the comparison reveals a high bias in MISR-retrieved AE, particularly when the AERONET-retrieved AE is very low (<0.6). The absence of wavelengths larger than 867 nm in MISR contributes to this overestimate, along with the limited selection of medium spherical particles in the MISR Version 22 components and mixtures (see the MISR quality statement). In the presence of coarse dust particles, the increase in AOD at wavelengths greater than 867 nm is higher than the increase at shorter wavelengths [Dey et al., 2004]. Hence, the absence of larger wavelengths in MISR suppresses the diminished spectral dependence of AOD, causing an overestimation of AE in heavy dusty cases. This can be seen from Figure 2b, where all the points lie above the 1:1 line for AERONET-retrieved AE less than 0.6, an indicator of heavy dust conditions in Kanpur [Singh et al., 2004].

[13] The fnsp from MISR cannot be validated directly with AERONET, because AERONET does not retrieve nonspherical particle fraction. Instead, we have compared MISR-retrieved fnsp with AERONET-retrieved AE, because AE can be used as a proxy to qualitatively determine the influence of dust on spectral aerosol optical properties in this region, as dust particles are dominantly of large size and nonspherical shape. Previous studies [Singh et al., 2004; Dey et al., 2004] have shown that an AE of 1 can be used as a rough divide between dusty and nondusty days over Kanpur. The MISR retrieval groups fnsp at 0.2 intervals because of the sensitivity of the MISR algorithm to nonspherical particles. MISR's medium-mode dust model was validated for Saharan dust [Kalashnikova and Kahn, 2006]. However, as pointed out by Kalashnikova and Kahn [2006], there is no satisfactory coarse-mode dust optical model, and in addition, there are hardly any field data available for dust types other than Saharan dust to directly evaluate MISR's retrieval. As shown in Figure 2c, AERONET-retrieved AE is found to increase with decreasing fnsp and the relation (R = −0.81) is significant with a 99% confidence interval. Nonzero values of fnsp for AE < 1 supports the successful retrieval of MISR to detect nonspherical dust particles in this region. MISR science team is working to improve the representation of coarse mode dust optical model in the aerosol algorithm; however, even with the current limitation, we are able to logically explain the spatial distribution of optical depth due to nonspherical particles (AODnsp) by topography and meteorology in the premonsoon and monsoon seasons (sections 4.2 and 4.3).

[14] MISR-retrieved size-segregated optical depth fractions also cannot be directly compared with AERONET-retrieved optical depth fraction, because AERONET retrieves “fine mode fraction” from optical measurements of spectral variation of AOD [O'Neill et al., 2003] that is defined differently than MISR. AERONET retrieval does not consider any microphysical cutoff particle size to define “fine” and “coarse” mode, and assumes neutral spectral dependence of coarse mode optical depth. AERONET fine mode optical depth (AODfine) and coarse mode optical depth retrievals are indirect retrievals from the retrieved AOD spectrum. They are also not validated. In eight cases out of 69 coincident sampling days, AERONET-retrieved AODfine was found to be higher than AERONET-retrieved total optical depth; hence they are excluded from the comparison (Figure 2d). In any multimodal particle size distribution, the fine (radii < 0.35 μm) and medium size (radii in the range of 0.35 to 0.7 μm) particles (hereafter grouped together and called small particle, i.e., fraction of small particles to the AOD, fs = ff + fm) retrieved by MISR are most sensitive to AERONET-defined fine mode, and large particles (>0.7 μm) retrieved by MISR are most sensitive to coarse mode as defined by AERONET. As shown in Figure 2d, the weak but statistically significant correlation (R = 0.36) between MISR-retrieved fs and AERONET-retrieved fine mode fraction (fmf = AODfine/AOD) indicates agreement in a broad sense between MISR and AERONET in retrieving the contribution of small particles to AOD. The correlation between MISR-retrieved AODnsp and AODl (not shown here) reveals that 65% of the variation in AODnsp can be explained by the variation in the AODl alone, and if the “medium” size particles are also considered nonspherical, it can explain 78% of the variation in AODnsp. This implies that AODnsp reported here may be even higher particularly during the dusty days.

[15] Unlike the particle properties discussed so far, MISR-retrieved SSA at 446 nm shows no correlation with AERONET-retrieved SSA at 440 nm as shown in Figure 2e. While MISR-retrieved SSA values are very high (mostly higher than 0.97) over Kanpur, AERONET retrievals indicate the presence of absorbing aerosols in this region [Singh et al., 2004; Dey and Tripathi, 2007, 2008]. The high aerosol absorption has been attributed to high concentrations of black carbon as measured from ground and aircraft [Tripathi et al., 2005a, 2005b; Ganguly et al., 2006b] and the mixing of black carbon with other aerosol components [Chandra et al., 2004; Dey et al., 2008; Satheesh et al., 2008]. A review of the mixtures of aerosols assumed by the MISR Version 22 aerosol retrieval algorithm reveals that it lacks an aerosol mixture containing both carbonaceous aerosols and dust that are characteristic of the region. Kahn et al. [2009] discusses this missing mixture as the most likely explanation for the poor SSA retrieval, which also contributes to the low bias in AOD in this region. Future versions of the MISR aerosol algorithm will address this issue. Until then, SSA from MISR is not examined further in this study. All other MISR aerosol properties are used to generate seasonal climatology of aerosol properties, and where possible, tied to past ground-based observations on the seasonality of aerosol properties.

2.2. Analysis Procedure

[16] The Indian subcontinent region, defined here to be bound within 40° north latitude and the equator and 65° and 99° east longitudes, is subdivided into 5440 grids of size 0.5° × 0.5°. This grid resolution increases the number of samples for a grid cell beyond a grid at the original resolution of the aerosol product (17.6 × 17.6 km), while remaining small enough to capture the detailed spatial variation in terms of emission inventories, topography and meteorology. All 17.6 × 17.6 km pixels whose central coordinates are within a single 0.5° × 0.5° grid cell during any particular season are considered for seasonal climatology for that particular grid. The values of a given aerosol property of all the successfully retrieval pixels of each day are averaged to derive the daily mean climatological value for each 0.5° × 0.5° grid cell. For example, the mean midvisible AOD for day k in year y (equation image) in a given season for a given grid cell is derived as:

equation image

where, AOD is the AOD at one successfully retrieved 17.6 × 17.6 km pixel, m is the total number of successfully retrieved pixels within that grid in that particular day, and the summation is carried over all m pixels. The mean seasonal AOD (AODseason) for that particular grid was derived as:

equation image

where, n is the total number of days of MISR data during a given season over all 9 years of data. The seasonal climatology of all other aerosol properties has been derived similarly. Since MISR data are available from March 2000, we have eight winter (December–February) seasons and nine premonsoon (March–May), monsoon (June–September) and postmonsoon (October–November) seasons during the time period considered in this study. The total number of successfully retrieved 17.6 × 17.6 km pixels for a given season within each 0.5° × 0.5° grid cell that go into the statistics (i.e., sample numbers) is shown in Figure 3. Lower sample density (<100 per grid cell) occurs over regions and times of more frequent cloud cover (e.g., during the monsoon season), more frequent snow cover (e.g., winter at higher elevations), and over topographically complex regions (e.g., the Himalayan mountain range), as one would expect based on algorithm design [Diner et al., 2008].

Figure 3.

Spatial distribution of total number of samples (in log scale) used to generate the seasonal climatology of aerosol characteristics over the Indian subcontinent during the period March 2000 to November 2008.

3. Overview of Factors Influencing Aerosol Variability

[17] The Indian subcontinent is very diverse in topography, population distribution, meteorology and emission sources, which influence the observed spatial and seasonal variability of aerosol properties. Ramachandran and Cherian [2008] have provided a detailed description of the geographic, climatic and source diversity of various regions of India. Here, the climatology of AOD, AE, fs, fl, fsp and fnsp are discussed in view of eight major geographic regions over land and ocean (Figure 1). The oceanic region comprises of the Arabian Sea, Bay of Bengal and north Indian Ocean. Large anthropogenic fraction in the measured aerosol properties at far distances over the oceanic region has been observed during October to May (often called the dry season) because of transportation of aerosols dominated by anthropogenic particles from the Indian subcontinent landmass, while the aerosol properties are dominated by natural particles during June to September (wet season) [Ramana and Ramanathan, 2006]. The IGB, the world's most populated river basin at >700 million population, stretches from Pakistan in the west to Bangladesh in the east, encompassing most of northern India (Figure 1). Central India is mostly hard rock terrain with undulating topography. Large areas of the northwestern part of India and the eastern part of Pakistan are covered by the Great Indian Desert, while the west coast of India surrounding the mega city Mumbai is highly industrialized and heavily populated. Southern India, the peninsular part of India, is the only region to receive significant precipitation in the summer monsoon as well as the winter monsoon (October–December). Precipitation in all other regions is mostly confined to the summer monsoon (hereafter “monsoon”) season. The northeastern part of India is sparsely populated and is the most underdeveloped region in terms of industrialization. The Tibetan Plateau is bordered by the Himalayan mountain range to the south and the Taklamakan and Gobi Deserts to the north. The Tibetan Plateau and the Himalayas have a profound dynamical and thermal influence on atmospheric circulation, and recent studies have pointed out the potential impact of pollution over this region on the south Asian monsoon rainfall through an “elevated heat pump” effect [Lau et al., 2006].

[18] The meteorological fields are obtained from National Centers for Environmental Prediction (NCEP) reanalysis data (http://www.cdc.noaa.gov) to understand the influence of synoptic-scale features on the spatial and temporal distribution of aerosol properties. Figure 4 shows the seasonal climatology of vertical winds at 850 mb, wind speed and wind direction at the surface and at 850 mb averaged over the same period of MISR data. In the winter, northerly to northwesterly wind transports a large anthropogenic component to the aerosol field over the oceans [Ramanathan et al., 2001]. The vertical winds show subsidence over the eastern part of IGB, Tibetan Plateau and oceanic regions close to the east and west coasts of India. In the premonsoon season, surface wind speed decreases over the ocean as compared to the winter season. The westerly to northwesterly winds transport desert dust from the Great Indian Desert and Arabian Peninsula affecting the aerosol optical properties over the Indian subcontinent [Dey et al., 2004; Prasad and Singh, 2007; Satheesh et al., 2008]. Increase in near surface wind speed in the monsoon season as compared to the premonsoon season allows enhanced production of maritime aerosols [Satheesh et al., 2006d] and transportation of them by southwesterly winds to the coastal regions. Precipitation data taken from Global Precipitation Climatology Project during the same MISR data period during the monsoon season show strong spatial (Figure 5a) and interannual (Figure 5b) heterogeneity. Precipitation in the other seasons (not shown here) is mostly lower than 200 mm over the land region. The synoptic meteorology in the postmonsoon season is generally similar to that of the winter season, but the surface wind speed is lower over the oceanic region compared to the winter season.

Figure 4.

Seasonal climatology of vertical wind at 850 mb (top panel), wind speed (in m/s) and wind direction (shown by arrows) at the surface (middle panel), and at 850 mb (bottom panel) over the Indian subcontinent during the same period of MISR data.

Figure 5.

Spatial distribution of (a) Global Precipitation Climatology Project–derived mean accumulated rainfall (in mm) in the monsoon season during 2000 to 2008 and (b) standard deviation of accumulated rainfall during the same period of MISR data.

[19] As the particle microphysical and optical properties depend on their sources, it is important to account for the spatial distribution of various emission sources beforehand to understand the climatology of aerosol properties. In the Indian subcontinent, most of the industrial sectors are associated with large cities (i.e., cities with >500,000 population). The major sources of aerosols in the urban/industrial areas are fossil fuel combustion in thermal power plants, industries (e.g., refineries and petrochemical, fertilizers, food, heavy engineering, cement, steel, textile, paper, nonferrous metals, brick making, etc.), domestic purposes and transportation, which generates mostly spherical and smaller (as compared to natural sources) particles [Reddy and Venkataraman, 2002a]. These particles can be moderate to highly absorbing depending on the mass fraction of carbonaceous aerosols. Biomass combustion (biofuel combustion and open biomass burning) and soil dust from rural roads, agricultural and bare lands and mining activities are the major sources of aerosols in the rural areas [Reddy and Venkataraman, 2002b; Ramachandran and Cherian, 2008]. From a MISR perspective, in a broad sense, small and spherical particles can be assumed to have a large anthropogenic component. The desert dust and maritime aerosols are not only larger in size than the anthropogenic aerosols, the proportion of nonspherical particles in desert dust [Kahnert et al., 2007] and even in maritime aerosols under suitable conditions [Chamaillard et al., 2006] is also higher. It should be noted that it is not possible to distinguish transported desert dust from local soil dust based on the nonspherical fraction retrieved by MISR, nor are there any in situ databases available that make such distinctions. Hence, we do not attempt to categorize desert dust from local soil dust, and instead use the term “dust” to interpret the climatology of nonspherical particles.

[20] Emissions from fossil fuel and biofuel combustion do not show much seasonal variability [Reddy and Venkataraman, 2002a]; rather they are responsible for the observed spatial variability in aerosol characteristics. However, emission from open biomass burning (i.e., forest burning and crop waste burning) is seasonal in India [Venkataraman et al., 2006]. Crop wastes are burnt from one month after the start of the harvest season until the end of the season, which is region specific in India. The western part of IGB has the highest crop waste availability followed by central and south India, whereas central India shows highest forest burning followed by the eastern part of IGB and the northeastern part of India [Venkataraman et al., 2006]. Forest burning peaks in the premonsoon season. Rabi (e.g., wheat, burley, mustard, etc.) and kharif (e.g., rice, millets, pulses, groundnut, cotton, sugarcane, etc.) crops are sown during late winter and monsoon seasons, respectively, and the crop wastes are burnt after the harvest, leading to seasonal peaks in open biomass burning in May (for rabi crop harvest) as well as in October (for kharif crop harvest) in the IGB. In central India, crop waste burning occurs during January–April and October–December, while in south India it takes place during February–March consistent with the January–February harvest season [Venkataraman et al., 2006]. Regional aerosol optical properties are strongly influenced by these emissions. However, the emission inventories in India have large uncertainties [Streets et al., 2003; Bond et al., 2004; Venkataraman et al., 2005, 2006] and the spatial and temporal distributions of emission data for all emission sources are not available. We have used population data as a proxy to link anthropogenic activities to the observed spatial distribution of particle properties. Population data of the Indian subcontinent has been collected from the Socioeconomic Data and Application Center Web site (http://sedac.ciesin.columbia.edu/gpw/global.jsp#), and the population density for the year 2005 is shown in Figure 6 at a 0.5° × 0.5° spatial resolution. Regionally, 41 large cities with population >500,000 are situated all along the IGB from west to east, while 16 large cities are located in western India, of which 11 are concentrated near the west coast of India, and 7 and 14 large cities are located in central and south India, respectively. Further categorization of the population data of India, Pakistan and Bangladesh into rural and urban population (not shown here) is done based on district data collected from the Census of India under Ministry of Home Affairs of Government of India (www.censusindia.net), the Statistics Division of Ministry of Economic Affairs and the Statistics of Government of Pakistan (www.statpak.gov.pk) and Bangladesh Bureau of Statistics (www.bbs.gov.bd), respectively. Rural population contributes 80% of the total population in the high population density (>600 km−2) region in the eastern part of IGB, while urban population is high in south India and is mostly centered on the large cities with high population density elsewhere.

Figure 6.

Spatial distribution of population density (km−2) over the Indian subcontinent.

4. Climatology of Aerosol Properties

[21] Aerosol optical and microphysical properties averaged over the Indian subcontinent show strong seasonal variability as revealed by the normalized frequency distributions plotted as function of seasons from the 2000 premonsoon season to the 2008 postmonsoon season in Figure 7. Kalashnikova and Kahn [2008] and Chen et al. [2008] have demonstrated that the MISR sensitivity to particle properties is best for AOD > 0.15; over the Indian subcontinent this occurs ∼84% of the time when there is a successful retrieval. Low AOD (<0.15) is mostly found over the Tibetan Plateau region (Figure 8). The seasonal distribution of AOD over the Indian subcontinent (Figure 7a) shows a median value in the range 0.194 to 0.325 with a strong seasonal cycle. The median value is highest in the premonsoon season followed by the monsoon season, when the overall frequency distributions are wider than those in the postmonsoon and winter seasons as revealed by a larger difference in first- and third-quartile values. The wide frequency distribution of AOD found in the monsoon season (Figure 7a) is attributed to highly heterogeneous spatial distribution of AOD (Figure 8). While median values of AOD tend to peak in the premonsoon season, the peak of the normalized frequency distribution occurs in the winter season. In winter, the peak of the distribution occurs in the 0.25–0.35 AOD bin, and the width of the distribution is narrowest compared to all other seasons.

Figure 7.

Normalized frequencies of mean seasonal (a) AOD at 0.1 bin width, (b) AE at 0.2 bin width, (c) fs at 0.1 bin width, and (d) AODnsp at 0.03 bin width, along with the seasonal median (black solid line), first-quartile (white solid line), and third-quartile (olive-green solid line) values for MISR data during the 2000 premonsoon season to the 2008 postmonsoon season. Monsoon season in the x axis index is abbreviated as “M”.

Figure 8.

Spatial distribution of climatological mean midvisible AOD (first panel), standard deviation of AOD, AODσ (second panel), mean AE (third panel), mean fs (fourth panel) and mean AODnsp (fifth panel) during the winter, premonsoon, monsoon, and postmonsoon seasons over the Indian subcontinent retrieved by MISR during the period March 2000 to November 2008.

[22] The seasonal median, first- and third-quartile values of AE vary in the range 0.72–1.06, 0.58–0.92 and 0.92–1.24, respectively, as shown in Figure 7b. The median value is lowest (AE < 0.8) in the monsoon season followed by the premonsoon season. AE frequency distribution is wider in the monsoon and postmonsoon seasons than the winter and premonsoon seasons as revealed by a larger difference in the first- and third-quartile values (Figure 7b) owing to larger spatial heterogeneity (Figure 8). The frequency distribution of fs shown in Figure 7c is similar to the frequency distribution of AE; lower median values in the monsoon and premonsoon seasons than the postmonsoon and winter seasons. The seasonal median, first- and third-quartile values of fs vary in the range 0.6–0.72, 0.55–0.63 and 0.67–0.81, respectively. The reduction of AE and fs in the premonsoon season along with a rise in AOD are primarily due to an increase in dust emission as observed in earlier studies [e.g., Dey et al., 2004; Jethva et al., 2005; Prasad and Singh, 2007], while higher production of maritime aerosols and dust contribute further to the reduction of AE and fs during the monsoon season [Satheesh et al., 2006d]. Unlike the AE frequency distribution, the fs distribution is wider in the postmonsoon and winter seasons than the other two seasons. The relative influence of anthropogenic particles is high over the Indian landmass and the surrounding oceans during the postmonsoon and winter seasons, while coarse particles still dominate over the Tibetan Plateau, thus contributing to wider frequency distribution in fs.

[23] The frequency distribution of AODnsp shown in Figure 7d is different from the other three aerosol parameters. The median value of AODnsp is highest in the monsoon season followed by the premonsoon season, and it is lower by a factor of two in the postmonsoon and winter seasons. The highest (>0.35) normalized frequencies are observed at AODnsp bin of 0–0.03 during the winter and postmonsoon seasons, along with its narrowest distribution (smallest difference in the first- and third-quartile values), indicating that nonspherical particles are at low levels over most of the region. The normalized frequency at the largest bin (AODnsp > 0.18) increases substantially in the monsoon season. This is attributed to the increase in the nonspherical particles over the oceans (Figure 8), which also leads to a wider frequency distribution in this season. The seasonal cycle in nonspherical particle loading is so strong that the first-quartile values in the premonsoon and monsoon seasons are similar or higher than the median values in the postmonsoon and winter seasons (Figure 7d). However, thin cirrus frequently observed in the Indian subcontinent in the monsoon season [Wang et al., 1996] may influence the retrieval of nonspherical particles (section 4.3). In summary, MISR captures many known seasonal variability of the particle properties, further substantiating the quality of the data, but the spatial distribution reveals finer details of the aerosol characteristics influenced by natural and anthropogenic emissions, topography and meteorology. The spatial heterogeneity of the aerosol parameters, as shown in Figure 8, is discussed separately below for each season.

4.1. Winter Season

[24] The spatial distribution of AOD (Figure 8) in the winter season reveals high AOD over the IGB and its outflow to the northern Bay of Bengal because of high anthropogenic emission sources, as previously observed from satellite [Di Girolamo et al., 2004; Jethva et al., 2005; Prasad et al., 2006; Ramachandran and Cherian, 2008] and ground-based [Singh et al., 2004; Tripathi et al., 2006; Nair et al., 2007] measurements. Figure 8 shows that AOD averaged over eight winter seasons is highest (>0.4) over the eastern part of IGB, which was referred to as the “Bihar pollution pool” by Di Girolamo et al. [2004]. Aerosols transported across the IGB from west to east by northwesterly winds (Figure 4) encounter a narrowing valley floor (Figure 1) and are trapped efficiently within the atmospheric column in the eastern part of the IGB by subsiding air (Figure 4). AOD is also higher than 0.3 over the industrialized sector in the west coast of India surrounding the Mumbai metropolitan area. Low (<0.2) AOD in south India, despite high anthropogenic activities revealed by high population density (Figure 6), is due to blow out of the aerosols to adjacent oceans by northeasterly to easterly winds [Di Girolamo et al., 2004], and as the outflow encounters subsidence over the ocean (Figure 4), it rises above 0.3 (Figure 8). AOD is lower than 0.2 over the topographically high Western Ghats, Eastern Ghats and Vindhyan Mountain ranges (Figure 1) and Myanmar (aka Burma) and even lower than 0.1 over the Tibetan Plateau in this season. Regionally, mean (± standard deviation (SD)) AOD over the IGB (0.36 ± 0.09) is >50% higher than any other region (Table 2), while it is comparable over western India (0.24 ± 0.07), central India (0.24 ± 0.04) and south India (0.22 ± 0.05) with small interannual variability as suggested by low AODσ values. However, the interannual variability is high over the eastern IGB, outflow region in the northern Bay of Bengal and the Taklamakan Desert (AODσ > 0.06). The strong latitudinal gradient of AOD over the oceans is due to transport of anthropogenic aerosols from the landmass, consistent with previous findings [Ramanathan et al., 2001; Satheesh et al., 2006a, 2006b]. Mean (±SD) AOD is slightly higher over the Bay of Bengal (0.27 ± 0.05) than over the Arabian Sea (0.25 ± 0.04), in part, because of the larger outflow of aerosols from the IGB.

Table 2. Mean Seasonal Climatological Values Averaged Over Major Geographic Regions of the Indian Subcontinent for Midvisible AOD, AE, fs, fl, fsp, and fnspa
ParametersWinterPremonsoonMonsoonPostmonsoon
  • a

    Mean is ±standard deviation; geographic regions are delineated in Figure 1; midvisible measured at 558 nm.

Indo-Gangetic Basin
AOD0.36 ± 0.090.45 ± 0.060.45 ± 0.080.35 ± 0.08
AE0.94 ± 0.110.96 ± 0.151.03 ± 0.141.04 ± 0.11
fs0.65 ± 0.040.67 ± 0.100.70 ± 0.060.70 ± 0.04
fl0.35 ± 0.060.33 ± 0.070.30 ± 0.090.30 ± 0.07
fsp0.89 ± 0.050.80 ± 0.080.84 ± 0.080.86 ± 0.07
fnsp0.11 ± 0.010.20 ± 0.040.16 ± 0.070.14 ± 0.06
 
Central India
AOD0.24 ± 0.040.36 ± 0.050.36 ± 0.050.25 ± 0.04
AE1.13 ± 0.091.00 ± 0.131.08 ± 0.111.21 ± 0.09
fs0.66 ± 0.030.67 ± 0.050.71 ± 0.040.71 ± 0.04
fl0.34 ± 0.050.33 ± 0.040.29 ± 0.070.29 ± 0.05
fsp0.92 ± 0.070.81 ± 0.040.83 ± 0.050.92 ± 0.04
fnsp0.08 ± 0.010.19 ± 0.060.17 ± 0.060.08 ± 0.01
 
Western India
AOD0.24 ± 0.070.39 ± 0.060.53 ± 0.100.27 ± 0.07
AE0.99 ± 0.180.76 ± 0.140.83 ± 0.141.00 ± 0.18
fs0.62 ± 0.070.56 ± 0.060.60 ± 0.060.64 ± 0.08
fl0.38 ± 0.020.44 ± 0.060.40 ± 0.090.36 ± 0.07
fsp0.92 ± 0.120.79 ± 0.030.85 ± 0.090.92 ± 0.09
fnsp0.08 ± 0.020.21 ± 0.120.15 ± 0.070.08 ± 0.03
 
South India
AOD0.22 ± 0.050.33 ± 0.060.31 ± 0.080.23 ± 0.06
AE1.06 ± 0.091.11 ± 0.171.01 ± 0.221.12 ± 0.12
fs0.68 ± 0.050.70 ± 0.040.69 ± 0.070.71 ± 0.05
fl0.32 ± 0.030.30 ± 0.080.31 ± 0.060.29 ± 0.04
fsp0.86 ± 0.080.82 ± 0.070.77 ± 0.070.77 ± 0.05
fnsp0.14 ± 0.020.18 ± 0.040.23 ± 0.060.23 ± 0.03
 
Tibetan Plateau
AOD0.09 ± 0.040.26 ± 0.080.19 ± 0.070.09 ± 0.04
AE0.88 ± 0.240.83 ± 0.160.83 ± 0.160.88 ± 0.22
fs0.51 ± 0.090.57 ± 0.060.61 ± 0.070.49 ± 0.08
fl0.49 ± 0.030.43 ± 0.100.39 ± 0.080.51 ± 0.05
fsp0.78 ± 0.060.85 ± 0.090.79 ± 0.050.78 ± 0.03
fnsp0.22 ± 0.020.15 ± 0.030.21 ± 0.020.22 ± 0.02
 
Arabian Sea
AOD0.25 ± 0.040.33 ± 0.050.47 ± 0.100.29 ± 0.04
AE1.08 ± 0.070.80 ± 0.130.53 ± 0.101.04 ± 0.07
fs0.77 ± 0.040.67 ± 0.040.59 ± 0.030.77 ± 0.03
fl0.23 ± 0.030.33 ± 0.060.41 ± 0.080.23 ± 0.03
fsp0.76 ± 0.070.61 ± 0.030.49 ± 0.090.69 ± 0.03
fnsp0.24 ± 0.010.39 ± 0.100.51 ± 0.070.31 ± 0.07
 
Bay of Bengal
AOD0.27 ± 0.050.33 ± 0.050.35 ± 0.060.25 ± 0.04
AE1.07 ± 0.101.02 ± 0.120.73 ± 0.170.98 ± 0.11
fs0.77 ± 0.040.71 ± 0.040.62 ± 0.050.76 ± 0.04
fl0.23 ± 0.050.29 ± 0.050.38 ± 0.060.24 ± 0.05
fsp0.78 ± 0.090.73 ± 0.050.63 ± 0.050.72 ± 0.04
fnsp0.22 ± 0.010.27 ± 0.040.37 ± 0.110.28 ± 0.08
 
North Indian Ocean
AOD0.21 ± 0.040.21 ± 0.050.22 ± 0.050.18 ± 0.06
AE0.98 ± 0.100.92 ± 0.110.66 ± 0.140.98 ± 0.14
fs0.74 ± 0.040.69 ± 0.040.61 ± 0.040.74 ± 0.06
fl0.26 ± 0.030.31 ± 0.050.39 ± 0.060.26 ± 0.05
fsp0.76 ± 0.060.67 ± 0.030.50 ± 0.050.72 ± 0.03
fnsp0.24 ± 0.020.33 ± 0.050.50 ± 0.070.28 ± 0.10

[25] AE is mostly greater than 1.0 over the Indian subcontinent, except the high AOD zone in the eastern IGB and the Tibetan Plateau, implying a high relative influence of small (high anthropogenic component) particles on spectral aerosol optical properties. In addition to low AE (<0.8), slightly higher AODnsp (>0.06) in the eastern IGB than the other parts, suggests high concentration of coarse dust particles emitted possibly by rural activities from the densely populated rural population. Mean AE is highest over central and south India (Table 2) in this season. Although the AE is lower in the eastern IGB, fs is in the range 0.6–0.7 across the IGB. ff is higher in the central and western parts of IGB than the eastern part (not shown here), while fm is higher in the eastern IGB, thus resulting in almost a similar fs (as fs = ff + fm) across the IGB, but a lower AE in the eastern IGB because of higher fm. Mean (±SD) fs (0.65 ± 0.04) and fsp (0.89 ± 0.05) over the IGB are at lower and higher end of the anthropogenic fraction derived from in situ measurements during ISRO-GBP second land campaign [Dey and Tripathi, 2007] as well as model-based estimates [Ramanathan and Ramana, 2005]. The close similarities in the MISR retrieved fs and fsp and in situ measurements reveal the quality of MISR aerosol retrievals in assessing relative contributions of anthropogenic and natural sources. For example, the high AOD over the industrialized greater Mumbai area is clearly attributed to dominant anthropogenic emissions as suggested by very high AE (>1.2) and fs (>0.8).

[26] Small particles contribute 60–70% to AOD over most of the landmass, but contribute 55–65% over the Great Indian Desert and arid to semiarid lands in western India because of fewer sources of anthropogenic emissions relative to local sources of dust emissions. fs is also lower than 0.55 over the Tibetan Plateau (Table 2), thus highlighting the fact that the Tibetan Plateau is the only region in the Indian subcontinent where coarse dust particles, raised locally from dry soil [Xia et al., 2008] by strong surface winds (Figure 4), are the predominant contributors to AOD in the winter season. Large fs (>0.8) in the high AOD (>0.3) zones over the northern Bay of Bengal and southern Arabian Sea is attributed to high anthropogenic fraction in the outflow from the IGB and central India and from south India, respectively. Higher fs (∼0.74) over the north Indian Ocean in the winter season relative to the other seasons is consistent with the seasonal cycle observed in previous studies [Ramana and Ramanathan, 2006]. The fraction of spherical particles to AOD over the IGB, central India, western India and south India is more than 0.85, whereas it is less than 0.8 over the Tibetan Plateau and the oceanic regions (Table 2). A peak in crop waste burning during the winter harvest season in central and south India results in a relatively higher fs and AE relative to the IGB and western India. AODnsp is slightly higher over the IGB and the oceanic regions (>0.03) than all other regions (AODnsp < 0.03). AODnsp over the Tibetan Plateau looks similar to the other parts of the subcontinent, although fs (and hence AE) is much lower. This may be attributed to low sensitivity of particle property retrieval in this region because of low AOD (mean value of 0.09) values; however, fnsp is still highest (∼0.22, Table 2) over the Tibetan Plateau in this season among the land regions. Nonspherical particles account for <20% of the AOD over the landmass elsewhere.

[27] We have developed an index (Table 3) based on a combination of the optical depth segregated by size and shape of the particles to infer the relative changes in seasonal mean aerosol properties from the preceding season in terms of anthropogenic and natural aerosols. Increments in AODl and AODnsp in the winter season over the preceding postmonsoon season are assumed to be due to an increase in dominantly natural aerosols, whereas increments in AODs and AODsp are assumed to be due to an increase in dominantly anthropogenic aerosols and vice versa. Figure 9 summarizes the spatial distribution of this index for each season, compared to the previous one. The spectral variation of AOD (i.e., AE) depends on the relative fractions of small and large particles. If both the natural and anthropogenic aerosols increase in the winter season over the preceding postmonsoon season, ΔAODl, ΔAODnsp, ΔAODs and ΔAODsp (Δ represents the difference between the seasonal mean values in the winter and postmonsoon seasons) will be positive, but ΔAE will be negative [positive] for a larger (Index 1) [smaller (Index 8)] increase in natural aerosols compared to anthropogenic aerosols. Similarly, if both anthropogenic and natural aerosols decrease in the winter season relative to the preceding postmonsoon season (i.e., negative ΔAODl, ΔAODnsp, ΔAODs and ΔAODsp), a larger decrease in natural aerosols over anthropogenic aerosols will lead to positive ΔAE (Index 7), whereas a larger decrease in anthropogenic aerosols over natural aerosols will result in a negative ΔAE (Index 2). Index 3 [Index 6] represents the regions where natural aerosols increase [decrease] and anthropogenic aerosols decrease [increase] in any particular season relative to its preceding season, resulting in a larger decrease [increase] in AE. However, it is possible that a simultaneous increase in large and nonspherical particles and decrease in small and spherical particles is not large enough to influence any detectable seasonal change in spectral optical properties (Index 4). For regions where AODs, AODl and AODnsp increase, while AODsp decrease, Index 5 is assigned. Index 0 is assigned to the area where the changes in the optical properties cannot be categorized based on the criteria discussed above, while ∼4% of the area (shown in white in Figure 9) has no successful retrieval to perform this test.

Figure 9.

Spatial distribution of the index (for details see Table 3) characterizing the changes in seasonal mean aerosol properties compared to the preceding season. For example, Indices 6 and 7 in winter and Index 6 in postmonsoon represent increasing anthropogenic particle fraction over the ocean because of transport of aerosols from the mainland; in premonsoon, Index 1 represents increasing natural particle fraction because of transport of dust, and Index 8 over the land represents increasing anthropogenic particle fraction because of seasonal peak in biomass burning; and in monsoon, Index 3 represents increasing natural particle fraction over the ocean because of persisting influence of dust transport and enhanced production of maritime aerosols. White represents no data.

Table 3. Indices That Characterize the Changes in Seasonal Mean Aerosol Properties Compared to the Preceding Season in Terms of Anthropogenic and Natural Aerosolsa
IndexΔAODsΔAODspΔAODAnthropogenicΔAODlΔAODnspΔAODNaturalΔAEΔAODsp/ΔAODnspΔAODAnthropogenic/ΔAODNatural
  • a

    Anthropogenic aerosols are defined as dominantly small and spherical particles; Natural aerosols are defined as dominantly large and nonspherical particles. Respective increase (decrease) of any given particle property in one season from the previous season is indicated by plus (dash) sign. No change detected and/or change uncertain for any given particle property are marked as cross sign, and larger increase (decrease) in natural compared to anthropogenic particles and vice versa is denoted by double plus sign (double dash).

1++++++ +
2− −
3+++
4+++×
5+×++++×
6++++++
7− −+++
8+++ +++++++
0All Other×
WhiteNo data (4% of total)

[28] For winter, Indices 2 and 0 dominate over India and Bangladesh, Index 7 dominates mainly over Pakistan and the Arabian Sea, Index 6 dominates over the Bay of Bengal, and the Tibetan Plateau and north Indian Ocean show no clear pattern (Figure 9). A reduction of AOD in the winter season compared to the preceding postmonsoon season over the western part of IGB, western and central India is attributed to reduction of both natural and anthropogenic aerosols, but a larger decrease of natural aerosols is reflected by Index 7 in the western part of the subcontinent, while a larger decrease in anthropogenic aerosols is reflected by Index 2 in the western IGB, western India and central India. The high AOD zone in the eastern IGB is characterized by Index 0, because AODl and AODsp increase and AODnsp decreases, while AODs remains similar in this region. Relative humidity (not shown here) in this region is higher in the postmonsoon season compared to the winter season, hence hygroscopic growth seems unlikely to explain the observed increase in AODl and AODsp during the winter season; rather drier conditions in the winter compared to the postmonsoon season, facilitated by stronger subsidence may lead to larger accumulation of local soil dust and a reduction in the amount of desert dust from distant sources, the net effect of which leads to a decrease of fs and AE. AOD decreases over the Arabian Sea in the winter season compared to the postmonsoon season, but the decrease is higher for natural over anthropogenic aerosols (Index 7). Larger blow out of aerosols from the IGB by stronger wind in the winter season compared to the postmonsoon season results in Index 8 over the coastal Bay of Bengal, but as the aerosols are transported deeper into the Bay of Bengal, stronger subsidence in the winter season (Figure 4) results in Index 6. No clear pattern over the Tibetan Plateau is attributed to low sensitivity in retrieved properties in both the postmonsoon and winter seasons because of low AOD.

4.2. Premonsoon Season

[29] In the premonsoon season, aerosol spectral optical properties change significantly from the preceding winter season because of enhancement in dust loading [Dey et al., 2004; Moorthy et al., 2007b; Prasad and Singh, 2007; Ramachandran and Cherian, 2008]. In general, AOD increases by >25% over the Indian subcontinent, with the largest increase (nearly three times) observed over the Tibetan Plateau (Figure 8). AOD is still lower (in the range 0.2–0.3) over the western Ghats and the Tibetan Plateau than the rest of the landmass. The gradient in AOD from the coast to deeper ocean over the Arabian Sea and Bay of Bengal is strongest among all seasons, as also observed by ship-based measurements during ICARB [Satheesh et al., 2009]. AOD is higher over the northern Bay of Bengal along the east coast of India than the northern Arabian Sea (Figure 8) because, in addition to dust, emission from open biomass burning in the IGB, central India and south India are transported to the Bay of Bengal by lower tropospheric westerly winds and its dispersal is inhibited by strong subsidence along the east coast (Figure 4). Aerosol mass concentration near the surface has been found to be much higher over the Bay of Bengal than the Arabian Sea during ICARB in the 2006 premonsoon season [Nair et al., 2008], but the MISR 9 year climatology reveals similar mean (0.33 ± 0.05) columnar AOD over these two regions (Table 2). However, the spatial coverage of ICARB in situ measurements and our analysis are not exactly same, and ICARB measurements might not represent multiyear average conditions. Aerosol characteristics are very different in these two oceanic regions as discussed below.

[30] AE reduces in this season because of an increasing dominance of large particles. The decrease is highest close to the dust source regions namely the northern Arabian Sea and western India. AE over the Mumbai metropolitan region still remains higher than 1.2 because of a higher relative influence of small particles from large anthropogenic emissions over large natural particles. Although mean AOD is similar over the Arabian Sea and Bay of Bengal, mean (±SD) AE (Table 2) is much higher over the Bay of Bengal (1.02 ± 0.12) than the Arabian Sea (0.80 ± 0.13). In the high AOD (>0.4) zone of the northern Arabian Sea, AE is found to be lower than 0.6. Air mass back trajectories (not shown here) at 850 mb originating from these high AOD zones in the Arabian Sea and Bay of Bengal using the NOAA HYSPLIT model [Draxler and Hess, 1998] support the Great Indian Desert and the Arabian Peninsula as the dust source regions. However, the relative influence of dust on the aerosol characteristics is stronger over the Arabian Sea compared to the Bay of Bengal because of the proximity to the source regions and lower influence of biomass burning emissions transported to the ocean, as supported by the lower mean values of AE and fs (Table 2). Higher relative influence of coarse particles on spectral AOD over the Arabian Sea compared to the Bay of Bengal was also observed by the ship-based measurements during ICARB [Kedia and Ramachandran, 2009]. The gradient in AE and fs is also sharper from north to south over the Arabian Sea compared to the Bay of Bengal. AE does not show any significant change over the Tibetan Plateau in this season from the preceding winter season, however, fs and AODnsp do change. One year (August 2006–July 2007) measurements of aerosol properties at Nam Co (Figure 1) AERONET station in the Tibetan Plateau have shown that mean AE remains lower than 0.5 during February to July [Cong et al., 2009]. Even in other months, mean AE remains lower than 1. Mean (±SD) annual AERONET-retrieved AE over the Tibetan Plateau of 0.42 (±0.27) suggests that coarse dust particles are a large component of the aerosol field throughout the year [Cong et al., 2009]. Mean seasonal variation of AE from MISR over the Tibetan Plateau lie in the narrow range of 0.83 and 0.88 (Table 2), which is consistent with that of Cong et al. [2009] given the high bias in MISR-retrieved AE in heavy dusty conditions (Figure 2b). fs decreases over most of the Indian subcontinent in this season because of dominant influence of large dust particles. The Great Indian Desert has the lowest fs (<0.55). While fl (i.e., 1-fs) increases over the western and central parts of the IGB, fs (and hence AE) increases over the eastern IGB relative to the winter season. Mean premonsoon AE over the IGB is almost similar to the mean wintertime AE because of this spatial contrast. Another striking feature is a high gradient of fs and AE from the IGB to the foothills of the Himalayas (Figure 8), suggesting an increasing relative influence of small particles over the foothills than over the basin. This is further supported by observations of higher AE (0.82 ± 0.22) derived by sunphotometer measurements over Nainital (see the location in Figure 1) in the Himalayan foothills during the premonsoon season of 2002 [Sagar et al., 2004] compared to AE (0.56 ± 0.26) retrieved by AERONET at Kanpur [Singh et al., 2004]. We also examined AERONET measurements at these two locations in the years 2008 and 2009 that show lower mean (±SD) AE over Kanpur in both the years (0.65 ± 0.33 in the year 2008 and 0.68 ± 0.14 in the year 2009) compared to Nainital (0.94 ± 0.04 in the year 2008 and 0.72 ± 0.02 in the year 2009). This gradient may be attributed to larger fs because of higher biogenic emission from denser vegetation coverage and higher emission from forest burning, supported by MODIS fire counts, over the Himalayan foothills region compared to the IGB. Pollutants with large anthropogenic fraction from the valley lifted upward along the Himalayan foothills through mesoscale circulation [Dumka et al., 2008] may also contribute to the observed gradient in AE and fs. Large emission of small particles from open biomass burning overcompensates the relative influence of dust on spectral AOD in the eastern part of the IGB as indicated by an increase of AE and fs in this season compared to the preceding winter season. This also leads to an overall increase in AOD in that region compared to the winter season (Figure 8). South India, particularly the coastal region, has the least relative influence of dust on the aerosol optical properties in this season as suggested by an increase in fs and AE from the preceding winter season.

[31] Nonspherical particles are found to increase from the winter season substantially over the IGB, western India, Taklamakan Desert and the oceanic regions, where AODnsp is greater than 0.09, contributing >20% (Table 2) to AOD. AODnsp shows a spatial gradient from high values in the western to low values in the eastern part of the IGB owing to settling of dust along the IGB during its long-range transport from Arabian Peninsula and Great Indian Desert. AODnsp shows an increase in this season from the winter season over some parts of the Tibetan plateau, where AOD rises over 0.2. The most striking feature in the spatial distribution of AODnsp, however, is the very high value (>0.12) over the Arabian Sea and Bay of Bengal along the coasts of India and a rapid decrease over land. The relative contribution of the nonspherical dust particles to AOD (i.e., fnsp) is more than 40% in these coastal water regions. The accumulation of dust transported from the Arabian Peninsula over the Arabian Sea along the west coast of India is facilitated by the strong subsiding air (Figure 4) in the region. However, the dust over the Bay of Bengal has two different dominant pathways of transportation. The first one is from the Great Indian Desert via the IGB settling down in the northern Bay of Bengal in the presence of subsiding air. In the second pathway, dust from the Arabian Peninsula transported to the western Ghats along the west coast of India, coupled with the cyclonic flow centered on south central India (Figure 4), is further transported around the southern tip of India into the Bay of Bengal where strong subsidence remains. This also restricts transportation of dust to south India, where the regional mean AE is the highest (1.11 ± 0.17) among all other regions in this season (Table 2 and Figure 8). The strong gradient in nonspherical particles does not seem to be an algorithm issue, because AODnsp is low (<0.06) over the north Indian Ocean. It also does not reduce sharply across the land-ocean boundary; rather it remains high over the low-lying coastal area along the west coast, dropping off only over the topographically high regions (Figure 1), where it encounters ascending air (Figure 4).

[32] It is evident from the previous discussion that the premonsoon season shows a marked change in the aerosol spectral characteristics over the entire Indian subcontinent compared to the preceding winter season. But how much is the change attributed only to dust? Is the seasonal change in relative contribution of small and spherical particles to the overall aerosol characteristics negligible compared to the influence of dust? Indices 1 and 8 (Table 3) dominate in the Tibetan Plateau (Figure 9) with no spatial pattern because of strong interannual variability in aerosol loading from local dust storms and dust transport from the Taklamakan Desert [Xia et al., 2008]. On the other hand, a definite spatial pattern emerges over the Indian landmass south of the Himalayas. The wedge-shaped region in the western part of the Indian subcontinent with a larger increase in natural over anthropogenic aerosols (Index 1) is close to the dust source regions. However, the increment in AODs over the Arabian Sea and western India is low (<0.05), and thus Index 1 in this region is mainly due to an increase in large size spherical particles. A large fraction of spherical particles in these regions seems to be natural, because the lower tropospheric wind (Figure 4) is mostly westerly and thus, we cannot use spherical versus nonspherical properties to distinguish anthropogenic from natural aerosols in this region. This is also supported by lower CO (an atmospheric tracer of anthropogenic emission) mixing ratio in the premonsoon season compared to the winter season over these regions [Kar et al., 2008]. The area with Index 8 becomes larger as we move along the IGB from the west to the east implying reduced relative influence of natural aerosols on aerosol optical properties. The effect of natural aerosols is least over the Gangetic West Bengal and Bangladesh, close to the northern Bay of Bengal. A separate pattern (Index 5) has been observed over the central to eastern IGB in this season. One plausible explanation for this observation may be that a large fraction of nonspherical dust particles is smaller than 0.7 μm (the cutoff size between small and large particles in the MISR retrieval). Mixing of spherical particles with nonspherical particles across the entire size range leading to an increase in AODnsp and decrease in AODsp may be another possible explanation. Mixing of dust with anthropogenic particles, particularly black carbon, has been found to be a probable mixing state over Kanpur [Dey et al., 2008], which is situated in the westernmost part (Figure 1) of the Index 5 zone. However, there is a lack of in situ data from the eastern part of the IGB to support this observation.

[33] Earlier in situ [Dey et al., 2004; Ganguly et al., 2006a; Prasad and Singh, 2007; Satheesh et al., 2008] and satellite observations [Jethva et al., 2005; Moorthy et al., 2007b; Ramachandran and Cherian, 2008] have attributed the changes in premonsoon aerosol characteristics to dust loading only. But, these results show that anthropogenic aerosols also increase over a large part of the subcontinent and the increment is large enough to influence the overall aerosol properties over certain regions. These observations are consistent with the fact that open biomass burning is at its peak in the premonsoon season in the IGB, central India and south India [Venkataraman et al., 2006], while other emission sources do not change significantly. A larger increase of anthropogenic aerosols compared to natural aerosols (Index 8) in this season has been found over the Himalayan foothills, northeast India, Myanmar, east coast of India and Indian landmass south of 15°N latitude. Low-lying topography, favorable meteorology (Figure 4) and increased emission of biomass burning facilitate a buildup of aerosols along the narrow fringe of the east coast in south India, where AE and fs increase in the premonsoon season from the winter season (Figure 8) leading to Index 6. In contrast to these four indices (Indices 1, 5, 6, and 8), other indices are mostly found over the oceans. The oceanic region receives substantial anthropogenic pollution from the landmass in the winter season (see Ramanathan et al. [2001] and subsequent studies), but the anthropogenic fraction is reduced by a change in wind direction in the premonsoon season. The effect is more pronounced in regions far away from the land. The high AOD and AODnsp zone over the Arabian Sea and Bay of Bengal has been characterized by Index 1, but AE and fs distributions reveal higher relative influence of the anthropogenic aerosols on aerosol properties in the Bay of Bengal than the Arabian Sea. Index 3 dominates in the north Indian Ocean because transportation of anthropogenic aerosols over the oceans from the subcontinent is reduced because of a reversal in wind direction (Figure 4), and natural aerosols increase because of dust transport. However, no clear pattern is observed in the southern Bay of Bengal and eastern Indian Ocean, which is attributed to larger interannual variation in aerosol properties as shown by AODσ in Figure 8.

4.3. Monsoon Season

[34] During the monsoon season, stronger westerly winds (Figure 4) transport greater components of dust from the Arabian Peninsula to the Indian subcontinent. In general, the spatial distribution of AOD (Figure 8) in this season is influenced by monsoon precipitation (Figure 5a). For example, very high AOD (>0.6) is observed over the low-precipitation zones in western India (mainly over the Great Indian Desert and neighboring arid regions) and the northern Arabian Sea. However, monsoon precipitation is not spatially and temporally uniform (Figures 5a and 5b). Suppressed precipitation in the monsoon break phase allows for a rapid build up of aerosols in the high anthropogenic source regions (e.g., IGB, Figure 6), while particles are being washed out by the precipitation in the active monsoon phase. This also leads to very high intraseasonal and interannual variability in aerosol characteristics as seen from the large AODσ values (Figure 8) and wider frequency distributions of particle properties (Figure 7). AODσ is higher than 0.12 over the Arabian Sea, northern and eastern Bay of Bengal and some parts of the IGB, where the interannual variation in monsoon precipitation (Figure 5b) is highest (>250 mm), thus further establishing the fact that AOD distribution is strongly influenced by monsoon rainfall. Further investigation on the influence of monsoon rainfall on AOD was conducted by Ravi Kiran et al. [2009], who have reported very high and statistically significant differences in AOD over India during active and break phases of the monsoon using MODIS data. Even with monsoon precipitation, AOD remains high over many regions. However, we note that the chance of successful retrievals of aerosol properties (either by MISR or MODIS) is higher during the break monsoon phases than the active monsoon phases because of higher cloud coverage in the active phases. Hence, the monsoon climatology reported here (and in the literature) is biased toward the observations during the monsoon break phases.

[35] Mean (±SD) AOD is highest in the monsoon season among all the seasons over western India (0.53 ± 0.10), Arabian Sea (0.47 ± 0.10) and Bay of Bengal (0.35 ± 0.06), and is similar to the premonsoon mean values over the IGB (0.45 ± 0.08) and central India (0.36 ± 0.05) (Table 2). Emission of anthropogenic aerosols continues to be high over the IGB [Habib et al., 2006] and the west coast industrial sector surrounding the Mumbai metropolitan area. Anthropogenic fraction is much higher in the Mumbai industrialized region than IGB as suggested by higher AE and fs. AOD spatial gradient over the Bay of Bengal is lower in the monsoon season compared to the premonsoon season. The high AOD (>0.5) over the northern Arabian Sea in the monsoon season has been detected earlier using MODIS data [Satheesh et al., 2005]. A large increase in AOD (>30%) in this season from the premonsoon season over the western part of the subcontinent is attributed to an increase in mainly small and spherical particles transported from north and northwest Asia (Figure 4), as supported by an increase in fs and AE (Figure 8). This is reflected in the characterization of this region by Index 6 (Figure 9), showing an increase in anthropogenic aerosols and a decrease in natural aerosols compared to the premonsoon season. Enhanced production of maritime aerosols by strong surface wind (Figure 4) also contributes to the observed increase of AOD over the Arabian Sea in this season along with dust, thus showing low AE (<0.6) compared to over the land. Ground-based measurements of spectral AOD by multiwavelength radiometer at Minicoy (Figure 1) also show AE of ∼0.4 during this season [Vinoj et al., 2008]. Unfortunately, data from Port Blair in the Bay of Bengal (Figure 1) are not available in the literature for comparison in this season. Mean (±SD) AE over the Bay of Bengal (0.73 ± 0.17) is ∼38% higher than the Arabian Sea (0.53 ± 0.10). Except for the western to central IGB, Great Indian Desert and parts of central India, AE is greater than 1.0 and fs is greater than 0.7 over most of the Indian landmass because of low relative influence of large dust particles on spectral AOD. fs is mostly <0.6 over the Arabian Sea and in the range 0.55–0.65 over the Bay of Bengal and north Indian Ocean except the coastal northern Bay of Bengal, where higher fs may be attributed to larger removal of coarse particles than small particles by high (>1500 mm) precipitation (Figure 5a).

[36] AODnsp is highest (mean value of 0.23) over the Arabian Sea, where it contributes 51% to the AOD, whereas it contributes ∼37% to the AOD over the Bay of Bengal. High AODnsp over the oceanic regions indicates a persistent transportation of dust in the monsoon season. High (>900 mm) precipitation along the west coast of India contributes to a sharp decrease in AOD (as well as AODnsp) from the Arabian Sea to the western Ghats, while strong westerly winds continue to transport natural particles across the Bay of Bengal. By combining MODIS-derived AOD and NCEP-derived wind data, Satheesh et al. [2006d] have shown that during the monsoon season, the relative contribution of maritime aerosols to the AOD over the northern Arabian Sea is lower than that over the southern Arabian Sea. The air mass to the northern Arabian Sea has been found to come from west Asia and Africa [Satheesh et al., 2006d], whereas the air mass to the southern Arabian Sea comes from the open oceans [Vinoj et al., 2008]. We have also looked into the spatial gradient of the absorbing aerosol index (AAI) over the Arabian Sea (not shown here) from Total Ozone Mapping Spectrometer (TOMS, for data up to the year 2005) and Ozone Monitoring Instrument (OMI, after the year 2005). We found that AAI decreases from north to south and becomes very low (∼0) south of 10°N latitude (up to which the high AODnsp zone exists). This further supports the dominance of dust to the AOD (Table 2) in the Arabian Sea during this season. An additional factor to consider is the potential for thin cirrus contamination, since the occurrence of thin cirrus is highest during the monsoon season [e.g., Wang et al., 1996]. Thin cirrus contamination would produce artificially large AODnsp, which we would be attributing to dust in our discussion. However, satellite-based lidar and solar occultation observations have shown that thin cirrus occurrence in the monsoon season is much larger over the Bay of Bengal compared to the Arabian Sea [Wang et al., 1996; Nazaryan et al., 2008], suggesting that if thin cirrus contamination were prevalent in the MISR retrievals, then AODnsp would be largest over the Bay of Bengal. Since this is not observed in Figure 8, we conclude that thin cirrus contamination is unlikely to dominate the spatial patterns of AODnsp observed in Figure 8. The sharp reduction in AODnsp from the Arabian Sea to the west coast of India may also be affected by abrupt reduction in sample density (Figure 3) over the west coast of India because of large cloud cover.

[37] As discussed above, changes in the aerosol properties over most of the oceanic regions in the monsoon season compared to the premonsoon season are characterized by Index 3 (Figure 9). However, the northern Arabian Sea and the central Bay of Bengal are characterized by Index 0 instead of Index 3 observed over most of the oceanic region. Surface wind speed over these regions is very high (Figure 4), which may increase emission of spherical maritime particles. Transportation of anthropogenic aerosols from north and northwest Asia may also contribute to the observed increase in AODsp over the northern Arabian Sea, but Index 0 over the central Bay of Bengal seems to be a local feature. A decrease in AOD over the northern Bay of Bengal near the coast of Bangladesh is attributed to a larger decrease in anthropogenic than natural aerosols, which is characterized by Index 2. Index 6 over the western IGB and central India is attributed to a reduction in dust transportation as observed in AODnsp spatial distribution, which also results in an increase of fs and AE (Figure 8).

4.4. Postmonsoon Season

[38] Aerosol regional mean climatology in the postmonsoon season is very similar to that for the winter season (compare the mean values in Table 2 and the frequency distributions in Figure 7), but the spatial distribution differs in several regions. For example, the wintertime high AOD zone in the IGB shows a larger spread and higher interannual variability (AODσ > 0.08) across the basin in this season, owing to a stronger peak in crop waste burning in the western part of IGB than the eastern part [Venkataraman et al., 2006] and weaker subsidence in the eastern part of IGB compared to the winter season (Figure 4). The spread of the aerosol outflow from the IGB to the Bay of Bengal, where AOD > 0.3, is smaller in this season because of smaller subsidence area compared to the winter season, while the spread of the outflow to the Arabian Sea is larger compared to the winter season. A similar spatial gradient in AOD over the Bay of Bengal was also observed during ship-based measurements of spectral AOD and black carbon concentration during the 2003 postmonsoon season by Sumanth et al. [2004]. AOD decreases below 0.1 over most of the Tibetan Plateau, while it lies in the range 0.2–0.3 in central India and most of south India, except for the western and eastern Ghats mountain regions on the west and east coasts of India, where AOD is smaller than 0.2.

[39] The spatial distribution of the indices characterizing the seasonal changes of aerosol properties is very different from the other seasons (Figure 9). Mean AOD reduces by >25% over most of the geographic regions compared to the monsoon season (Table 2), primarily because of a reduction in emission of maritime and transportation of dust particles. But AODs and AODsp also decrease over part of the landmass at a lower rate than natural aerosols and thus these regions are characterized by Index 7. However, this is not the case in all years. Some years have shown an increase in anthropogenic aerosols in this season (in that case the regions would have been characterized by Index 6) and the other years have shown a decrease (Index 7) compared to the monsoon season, but the mean difference in AODs and AODsp between the postmonsoon and monsoon seasons of all the years leads to the characterization by Index 7. Anthropogenic aerosols are transported to the north Indian Ocean as suggested by a rise in fs above 0.75 and as the natural aerosols reduce compared to the monsoon season, north Indian Ocean is characterized by Index 6. Over some parts of the Arabian Sea and Bay of Bengal, AODs, AODsp and AODl increases and AODnsp decreases compared to the monsoon season, thus it is shown as Index 0. Transition in aerosol characteristics during this season [Ramana and Ramanathan, 2006] makes it difficult to distinguish natural from anthropogenic particles by our criteria based on MISR's sensitivity.

[40] A larger decrease in natural compared to anthropogenic aerosols (Index 7) over most parts of the Indian subcontinent as we transition to this season from the monsoon season leads to an overall increase in AE and fs (Figure 8). AE is mostly >1.0 over the landmass except some parts of IGB and the Tibetan Plateau, where AODnsp is >0.06. AE almost doubles (mean AE = 1.04) in the postmonsoon season from the preceding monsoon season over the Arabian Sea. The spatial distribution of AE is fairly uniform over the Arabian Sea (Figure 8), as aerosols with large anthropogenic fraction, as suggested by high fs (>0.75), are being transported from the Indian landmass (Figure 4). AE shows a gradient with higher value (AE > 1.0) in the northern Bay of Bengal compared to the southern Bay of Bengal. fs is >0.8 over a large part of the northern Bay of Bengal because of transportation of aerosols with high anthropogenic fraction from IGB and east Asia aided by subsidence (Figure 4). The air mass over the southern Bay of Bengal has a larger maritime component as suggested by lower fs (in the range 0.7–0.8). AODnsp (Figure 8) and fnsp (Table 2) are found to decrease over the Arabian Sea compared to the premonsoon and monsoon seasons, but still remain higher compared to the winter season. Although AODnsp is much lower compared to the premonsoon season over the Bay of Bengal, fnsp is similar in these two seasons (Table 2). The overall decrease of AODnsp in this season relative to the monsoon season may be influenced by a reduction in cirrus contamination, but in situ observations also support the fact that dust transport reduces in the postmonsoon season [Dey and Tripathi, 2008; Ganguly et al., 2006a], as shown by AODnsp distribution.

5. Discussion, Summary, and Conclusions

[41] In view of the ongoing efforts to characterize aerosol properties over the Indian subcontinent, we have presented an extensive satellite-based climatology of aerosol optical and microphysical properties, namely AOD, fs, fl, fsp, fnsp at 558 nm wavelength and AE, using 9 years (2000–2008) of data from MISR. No previous study has discussed the climatology of particle properties other than AOD and fine mode fraction (to some extent) with more than five years of data. Here, the aerosol climatology has been discussed in view of influencing parameters, viz. meteorology, emission inventories and topography, within the known limitations of the MISR retrieval algorithm. Our evaluation of the MISR aerosol product against AERONET (section 2) points to the similar conclusions about the quality of MISR aerosol retrieval (e.g., low bias for high AOD condition and high bias for AE in heavy dusty days), that is described and explained by Kahn et al. [2009]. Furthermore, our ability to explain the spatial and temporal variability of the MISR aerosol properties based on meteorology, emission sources and topography, and numerous corroborations of the seasonal and regional variability of MISR aerosol microphysical and optical properties against available in situ observations support our interpretations of the MISR aerosol data.

[42] The mean climatological values of various particle properties over the major geographical regions of the Indian subcontinent (Figure 1) are summarized in Table 2 for easy comparison. The seasonal variability of aerosol properties reveals that AOD peaks during the premonsoon and monsoon season because of enhanced emission of natural aerosols. Large and nonspherical dust particles dominate the aerosol optical properties in the Tibetan Plateau throughout the year, while they contribute significantly to AOD in the premonsoon and monsoon seasons in other regions. Xia et al. [2008] have also discussed the seasonal and interannual variability of AOD over the Tibetan Plateau, which is quite similar to our results. However, the mean climatological values reported here are slightly different from their study, because they used older versions of MISR data for seven years (February 2000–May 2007) period compared to our use of 9 years of Version 22 MISR data. The broad regional differences of AOD depend in part on the emission sources, but the variation within a particular region depends significantly on meteorology and topography. For example, wintertime AOD in the IGB is higher than any other region because of higher emissions of anthropogenic aerosols [Streets et al., 2003; Bond et al., 2004; Venkataraman et al., 2005, 2006; Habib et al., 2006]. However, higher wintertime AOD in the eastern part than the western part of the IGB can be better explained by meteorology and topography [Di Girolamo et al., 2004].

[43] Although we have presented the first climatology of aerosol optical and microphysical properties, specifically AOD, fs, fl, fsp, fnsp at 558 nm wavelength and AE, in both spatial and temporal contexts, one component missing in our analysis is the climatology of aerosol absorption (i.e., SSA). Unfortunately, the existing satellite sensors are not able to retrieve SSA accurately. MISR overestimates SSA because of the absence of mixtures containing carbonaceous and dust particles in the current algorithm. TOMS/OMI derived AAI is able to provide some information on the presence of absorbing aerosols during the premonsoon season [Lau et al., 2006], but AAI tends to be insensitive to the boundary layer aerosols [Mahowald and Dufresne, 2004]. As the absorbing carbonaceous aerosols are mostly confined within the shallow boundary layer in the postmonsoon and winter seasons over the Indian subcontinent [Tripathi et al., 2005b; Ganguly et al., 2006b; Nair et al., 2007], it is difficult to use AAI to characterize absorbing aerosols in these seasons. The high AOD zone in the eastern IGB has high aerosol absorption as shown by a recent chemical transport model simulation [Adhikary et al., 2008]. Recent observations during ICARB have also suggested that the aerosols (predominantly dust) above the marine boundary layer in the premonsoon season are absorbing in nature and contribute to large atmospheric warming [Satheesh et al., 2008]. The issue of mixing of dust with anthropogenic components and their radiative effect because of enhanced absorption becomes increasingly relevant in view of high anthropogenic and dust particles in the premonsoon and monsoon seasons as revealed by MISR data. The high aerosol loading over the Indian subcontinent as observed by MISR, a fraction of which is absorbing black carbon [Ramanathan and Carmichael, 2008], has several likely climatic implications of significance, e.g., enhancement in precipitation through an elevated heat pump [Lau et al., 2006], weakening of the Indian monsoon circulation, solar dimming and the retreat of Himalayan glaciers [Ramanathan and Carmichael, 2008; Ramanathan and Feng, 2008].

[44] Nonspherical particle shapes can result in an overestimation of TOA aerosol cooling by >10% and atmospheric warming by ∼6% as compared to spherical particle with similar composition [Mishra et al., 2008]. Error in radiative transfer simulations because of approximating nonspherical particles as spherical particles is of equal magnitude to the error stemming from uncertainty in aerosol refractive index [Kahnert et al., 2007]. High climatological AODnsp and fnsp values as observed by MISR in the premonsoon and monsoon seasons, particularly over the IGB, Tibetan Plateau and the oceanic regions where fnsp is >20%, imply that particle nonsphericity must be accounted for to improve the estimates of aerosol direct radiative forcing, because so far, available estimates in the Indian subcontinent were based on spherical particle models only. The effect of nonspherical particle model on the simulated aerosol optical properties for dust-anthropogenic particle mixing also needs future attention.

[45] Long-term exposure to particles smaller than 2.5 μm (i.e., PM2.5) causes various respiratory and cardiovascular diseases. The very high climatological AOD over the densely populated regions (e.g., IGB, Mumbai metropolitan area) must be having a tremendous impact on the health of millions of people. However, there is a lack of a robust data set on surface aerosol concentration to study the health effect of outdoor aerosol in the Indian subcontinent (the rate of premature deaths in India because of indoor air pollution was examined by Smith [2000]). MISR data have been used in conjunction with global chemical transport model to derive surface PM2.5 concentrations successfully over the United States by Liu et al. [2004, 2007]. A global analysis by van Donkelaar et al. [2006] has already shown that mean annual surface PM2.5 over the IGB during 2001–2002 is in the range 40–50 μg m−3, four to five times larger than the World Health Organization air quality guideline of 10 μg m−3. Similar efforts should be made for the entire region by taking advantage of the long-term data set of size-segregated MISR particle properties in conjunction with measurements by the ground-based network of the Central Pollution Control Board of India for validation.

[46] Recent global aerosol models have improved because of the assimilation of satellite and in situ observations [e.g., Stier et al., 2005; Adhikary et al., 2007, 2008; Zhang et al., 2008; Lee and Adams, 2009], but many continue to fail in capturing the strong spatial variability of AOD [e.g., Stier et al., 2005, Figure 7a; Chin et al., 2009, Figure 8], as well as an overall underprediction of AOD [e.g., Lee and Adams, 2009; Chin et al., 2009] over India as compared to the satellite observations. MISR's ability to retrieve particle optical and microphysical properties, even over bright surfaces, provides an opportunity to both evaluate global aerosol models in further detail and to improve the simulations through data assimilation. The MISR-observed climatology of fs, fl, fsp and fnsp presented here and their spectral variations (normally characterized by AE) may also be used in constraining the regional estimation of aerosol forcing derived by extrapolating forcing efficiency [Chen et al., 2009], that has been simulated from optically equivalent aerosol models based on in situ measurements at a few isolated locations in the Indian subcontinent [Ramanathan and Ramana, 2005; Dey and Tripathi, 2007]. However, quantitative interpretation of such results should be done cautiously, to account for the known MISR aerosol property biases.

[47] The strong spatial heterogeneity in the aerosol climatology presented here suggests that the number and monitoring capabilities of existing ground-based aerosol measurement sites are insufficient to derive robust statistics of aerosol optical and microphysical properties for the entire Indian subcontinent. For example, the eastern part of IGB, with its dense rural population, has very few in situ observations despite this region's high aerosol loading and temporally varying microphysical properties. Central India also lacks coordinated in situ observations. Even the oceanic regions surrounding the subcontinent landmass show strong spatial and temporal variability in aerosol properties, yet continuous in situ measurements are available at only three locations (Figure 1), Port Blair in the Bay of Bengal, Minicoy in the Arabian Sea, and Hanimaadhoo in the north Indian Ocean. Ship-borne measurements of aerosol characteristics are almost negligible during the monsoon season, when the relative influence of dust on aerosol properties is highest (fnsp > 0.35) among all seasons. In recent years, CIMEL sunphotometers were deployed as part of AERONET at a few more sites in the IGB and foothills of the Himalaya (see the AERONET home page, http://aeronet.gsfc.nasa.gov) and more ground-based observatories are being set up as part of ISRO-GBP and ABC (not shown in Figure 1), which will add valuable information to the ongoing efforts in understanding the aerosol direct and indirect effects in the region. The MISR 9 year aerosol climatology presented in this paper can help in planning future initiatives for in situ observations in the Indian subcontinent.

[48] In summary, the major conclusions of this work are as follows:

[49] 1. The MISR aerosol products, except SSA, provide a good picture of regional and temporal variability of aerosol characteristics over the Indian subcontinent, as corroborated by many in situ studies, and tying logically to meteorology, topography, and emission source behavior. The Indian subcontinent displays strong seasonal and spatial heterogeneity in all aerosol properties studied here. AOD is not only high over the urban/industrialized sectors, but also high over the densely populated rural areas, thus stressing the importance of considering these rural regions for future in situ measurements.

[50] 2. In the winter season (December–February), high AOD (>0.3) is observed over the IGB, the industrialized region surrounding the mega city Mumbai, and their outflow to adjacent oceans, in part because of high anthropogenic emission sources. Favorable meteorological conditions allow a buildup of aerosols over the eastern part of IGB, where coarse dust particles add to higher AOD (AOD > 0.4) resulting in a lower AE (<0.8) compared to the rest of the landmass. Mean AOD over the IGB is >50% higher than any other region. Coarse dust particles dominate the aerosol optical and microphysical properties (fnsp = 0.22) over the Tibetan Plateau. fs is mostly greater than 0.75 over the surrounding oceanic regions because of the transportation of aerosol with a large anthropogenic fraction from the subcontinent, consistent with the previous findings.

[51] 3. Enhanced dust activity, as suggested by larger AODnsp compared to the winter season, results in an increase in AOD by >25% and a decrease in AE over a major fraction of the Indian subcontinent in the premonsoon season (March–May). The largest increase (nearly threefold) in AOD is observed over the Tibetan Plateau. However, along the foothills of the Himalaya, AE (>1.2) and fs (>0.75) are large relative to the Tibetan Plateau to the north and IGB to the south. This may be due to higher biogenic emission and emission from open biomass burning in the Himalayan foothills relative to the other regions. The AOD spatial gradient from the coast to deeper ocean is highest in this season. A rapid decrease in AODnsp from the Arabian Sea and Bay of Bengal to the coastal inland is attributed to a combination of topographic and meteorological effects. Climatologically, the fraction of natural aerosols to the AOD is lower over the Bay of Bengal compared to the Arabian Sea resulting in a much higher AE (1.02 ± 0.12 compared to 0.80 ± 0.13).

[52] 4. The influence of transported dust on the aerosol optical properties persists into the monsoon season (June–September), especially over the IGB, western India and northern Arabian Sea, where AOD is >0.5. However, small and spherical particles transported from the north and northwest Asia contribute significantly to the large increase in AOD observed over the western part of IGB. The highest land-ocean contrast in AE is observed in this season, when enhanced production of coarse maritime particles by strong surface winds along with dust results in low AE (<0.8) and fs (<0.6) over the oceans. The large particle fraction is much higher over the Arabian Sea compared to the Bay of Bengal because of proximity to dust source regions. AODnsp is highest (>0.15) in the monsoon season over the Arabian Sea and Bay of Bengal contributing >35% to the AOD. The intraseasonal and interannual variability of AOD is highest in this season as suggested by higher AODσ compared to the other seasons.

[53] 5. The regional mean aerosol climatology in the postmonsoon season (October–November) is similar to the wintertime mean aerosol climatology, but notable difference has been observed in the spatial distribution of the aerosol optical and microphysical properties. For example, high AOD (>0.4) is uniformly spread across the entire IGB and AE is much higher over the eastern IGB compared to the winter season, probably because of a weaker subsidence in the eastern IGB and higher crop waste burning in the western IGB in the postmonsoon season. fs and AE are higher over central and south India because of higher open biomass burning compared to the winter season. AE shows a gradient over the Bay of Bengal because of a larger relative influence of anthropogenic fraction on the AOD over the northern part compared to the southern part. AE almost doubles over the Arabian Sea compared to the preceding monsoon season because of a transportation of aerosols with a high anthropogenic fraction from the subcontinent landmass by a reversal of winds. The AODnsp decreases over the Indian subcontinent from the monsoon season, but remains larger compared to the winter season.

[54] 6. An index is developed to characterize the changes of mean seasonal aerosol properties compared to the preceding season in terms of the relative influence of natural and anthropogenic aerosols on aerosol optical and microphysical properties (Table 3 and Figure 9). Both natural and anthropogenic aerosols decrease in large parts of the landmass in the winter compared to the postmonsoon (Index 2) and in the postmonsoon compared to the monsoon season (Indices 7 and 2), but the Bay of Bengal is characterized in the winter season by a decrease in natural and an increase in anthropogenic aerosols compared to the postmonsoon season (Index 6). The high AOD zone in the eastern part of the IGB shows an increase in large and spherical particles, possibly emitted locally from rural activities and results in a lower AE compared to the rest of the IGB, and thus is characterized by Index 0. The dominance of Indices 1 and 8 in the premonsoon season suggests an increase in small and spherical particles simultaneously with an increase in natural aerosols, but a large fraction of small and spherical particles over the ocean may be of natural sources. An increase in maritime aerosols along with dust in the monsoon season over the oceanic region is reflected by Index 3 over the north Indian Ocean.

Acknowledgments

[55] Partial support from the Jet Propulsion Laboratory of the California Institute of Technology under contract 1260125 is gratefully acknowledged. The MISR aerosol data are distributed by the NASA Langley Atmospheric Sciences Data Center. The meteorological data are obtained from the NCEP/NCAR reanalysis data set. PIs of Kanpur and Nainital AERONET sites, Brent Holben, S. N. Tripathi, R. P. Singh, and P. Pant are acknowledged for their efforts in maintaining the operation of the instruments. We thank the reviewers for their comments, which helped us to improve the manuscript.