Aerosol climatology at Delhi in the western Indo-Gangetic Plain: Microphysics, long-term trends, and source strengths



[1] We present the climatology of aerosol microphysics, its trends, and impact of potential sources based on the long term measurements (for a period of 11.5 years from December 2001 to May 2012) of aerosol optical depths (AOD) in the spectral range 340–1020 nm from an urban center Delhi (28.6°N, 77.3°E, 238 m mean sea level) in the western Indo-Gangetic Plain (IGP). The study is the first ever long-term characterization of aerosols over the western IGP from the ground-based measurements. AODs are known to affect the air quality, visibility, radiative balance, and cloud microphysics of the region and IGP is one of the highest populated and polluted regions of the world. Our measurements show consistently high AOD during the entire period of observation. The seasonal variations of spectral AODs and Angstrom parameters are generally consistent every year. The AODs show a weak but statistically significant (in 95% confidence level) decreasing trend approximately −0.02/year at 500 nm, possibly, modulated by the pre-monsoon heavy dust loading during the first half of the observation period. The climatological monthly mean AOD at shorter wavelengths peaks twice, during June and November, while at longer wavelengths it shows only one peak in June. The annual variations of Angstrom exponent, α and its derivative, α′ suggest the prevalence of multi-modal aerosol size distributions at Delhi. The coarse-mode aerosols dominate during summer (March–June) and monsoon (July–September) seasons, whereas fine/accumulation mode enhances during post-monsoon (October–November) and winter (December–February) seasons. Potential advection pathways have been identified using concentration weighted trajectory (CWT) analysis of the 5 day isentropic air mass back trajectories at the observation site and their seasonal variations are discussed.

1 Introduction

[2] Atmospheric aerosols are known to affect the radiative balance of the earth at local, regional and global scale through direct and indirect effects [Intergovernmental Panel on Climate Change (IPCC), 2001, 2007]. Several manifestations of aerosol effects on atmosphere and climate have been recognized recently, viz., radiative forcing [Russell et al., 1999; Ramanathan et al., 2001; Satheesh and Moorthy, 2005], atmospheric warming [Hansen et al., 2000; Jacobson, 2001; Andreae et al., 2005; Lau et al., 2006], precipitation and cloud droplets [Twomey, 1977; Twomey et al., 1984; Breon et al., 2002; Rosenfeld et al., 2008], glaciers [Xu et al., 2009], and monsoons [Chung et al., 2002; Manoj et al., 2010]. Over the past two decades, observations of physical, chemical, and optical properties of atmospheric aerosols and their spatio-temporal variabilities have substantially improved our understanding and estimation of direct and indirect radiative effects. However, the uncertainties on their effects on the atmosphere and climate are still high and the level of understanding is far less than that of the greenhouse gases [IPCC, 2007]. One of the crucial factors contributing to this uncertainty is the large heterogeneity in aerosol optical and microphysical properties over spatial and temporal scales. Thus, for better understanding of the effects of aerosols on the Earth-atmosphere radiation budget and its implications on weather and climate, the accurate estimations of aerosol optical and microphysical properties are important [e.g., Haywood et al., 1999; Horvath et al., 2002; Meloni et al., 2005; Moorthy et al., 2009].

[3] Aerosol optical depth (AOD) is one of the most common parameters for characterizing the aerosols and doing radiative forcing estimations [Charlson et al., 1991; Twohy et al., 1995; Toon et al., 2000]. The columnar AOD can be measured directly using surface-based sun radiometry/photometry which yields the vertical column-integrated aerosol extinction coefficient measured from the surface to the top of the atmosphere, and using satellite by down-looking radiance measurements over a wide area. The vertical profile of extinction also is retrieved from LIDAR observations. The surface-based determination of AOD also serves as ground truth for satellites and are sometimes more suitable for evaluation of aerosol models, identifying specific sources and for monitoring temporal changes. AOD measurements at multiple wavelengths help to characterize the spatial and temporal distribution of aerosol loading and particle size distribution, anthropogenic and natural contribution, radiative forcing, visibility, and global dimming [Michalsky et al., 2010]. These aspects necessitate the long-term continuous ground-based measurements of aerosol parameters at distinct geographical locations to study their seasonal, inter-annual, and long-term variations. Even though, there have been such efforts on long-term characterization of aerosols globally [e.g., Holben et al., 2001; Smirnov et al., 2002; Kim et al., 2004, 2007, 2008, 2010; Eck et al., 2005, 2010; Kaskaoutis et al., 2007; Tomasi et al., 2007; Schafer et al., 2008; Michalsky et al., 2010; Gerasopoulos et al., 2011; Queface et al., 2011; Kudo et al., 2012] as well as over Indian regions [Jethva et al., 2005; Moorthy et al., 2007; Gogoi et al., 2009; Eck et al., 2010; Ram et al., 2010; Giles et al., 2011; Kaskaoutis et al., 2012; Dani et al., 2012], such studies are virtually non-existent over the western Indo-Gangetic Plain in the northern India. The Indo-Gangetic Plain (IGP) is the largest, heavily populated and polluted agriculture basin in the northern India [Prasad et al., 2005]. The urban mega-cities such as Delhi, Agra, Kanpur, and Varanasi are parts of this basin. The region suffers with severe fog, dense haze, and smog which are largely due to population growth and increasing urbanization/industrialization [Gautam et al., 2007]. The first-time dense pollution mapping over the IGP region by the POLDAR (POLarization and Directionality of the Earth's Reflectances) sensor on-board the ADES (ADvanced Earth Observation Satellite) [Goloub et al., 2001] attracted the attention of Indian investigators and this led to an in-depth aerosol characterization using various ground-based sun photometers/radiometers. Kanpur is the only known station in this region with long-term AOD measurements [Singh et al., 2004; Jethva et al., 2005; Dey et al., 2005; Gautam et al., 2007, 2009, 2010; Dey and Tripathi, 2008; Eck et al., 2010; Giles et al., 2011]. The present work is the first ever long-term study on aerosol microphysics from Delhi, the station in the western part of the Indo-Gangetic Plain.

[4] In this paper, we present the climatology of aerosol microphysics and its trends based upon the long-term measurements of spectral AODs, for a period of ~11.5 years, from December 2001 to May 2012 (~1580 clear sky days), using Microtops II Sun photometer deployed at the National Physical Laboratory (NPL), New Delhi. We also discuss in detail the seasonal variation of aerosol properties by identifying the potential advection pathways using the concentration weighted trajectory (CWT) Analysis of the 5 day isentropic air mass back trajectories.

2 Measurement Site, Instrument, and Data

[5] The experimental site Delhi (28.6°N and 77.3°E, 238 m mean sea level), one of the most densely populated cities in Asia (with about 10,400 persons km−2), is located in a semi-arid region at ~1100 km from the nearest coast on the North Arabian Sea and lies between the Indo-Gangetic Basin to the east and arid tracts of Thar and Margo desert at ~160 km to the west and south-west. The orography of the IGP is such that it forms a kind of natural channel sloping from the west to east with the tall Himalayan mountain ranges to the north and Aravalli, Vindhya, Satpura, and Bihar plateau to its south. This leads to the confinement of the aerosols into the region. The measurement site is in an urban setting with a semi-arid climate and is influenced by urban pollution as well as transported mineral dust. The climatic conditions are extreme in nature with very hot summers (the temperature rises even beyond 45°C) and cold winters with temperatures as low as 3°C. The monsoon season is from July to September. The year can be broadly divided into four distinct seasons, namely winter (December–February), summer (March–June), monsoon (July–September), and post-monsoon (October–November).

[6] For the present study, the aerosol optical depths (AODs) were measured over a long period of time (since December 2001 to May 2012) using handheld Microtops II Sun photometers (referred to as Microtops hereafter), manufactured by Solar Light Co., USA. The Microtops measures AOD at discrete wavelengths between 340 and 1020 nm with FWHM (full width at half maximum) bandwidth being ±2 nm at UV wavelengths and ±10 nm at visible and near-infrared wavelengths. The pointing accuracy toward the Sun for these instruments is better than 0.1° and the long-term stability of the filter is better than 0.1 nm yr−1. The narrow field of view of the collimators (2.5° full field of view) helps to reduce the stray light entering into the system. Generally, the Microtops for AOD measurements is equipped with five channels, each channel is fitted with a narrowband interference filter and a photodiode suitable for the particular wavelength range. The electrical signals from the five photo detectors are amplified and converted to digital form, and a self-contained microprocessor automatically calculates AOD at all the five wavelengths. The AOD calculation is based on Lambert-Beer's law. The maximum uncertainty in the optical depth estimations in each channel was kept approximately ±0.02 by strictly following the procedures for Microtops measurements [Morys et al., 2001; Porter et al., 2001; Ichoku et al., 2002, Smirnov et al., 2009]. More details about the instrument may be found elsewhere [Morys et al., 2001; Srivastava et al., 2006; Smirnov et al., 2009, 2011; Dani et al., 2012].

[7] We have used a pair of Microtops (denoted by Microtops1 and Microtops2), one with wavelengths 340, 500, 675, 870, and 1020 nm, and the other with wavelengths 340, 500, 870, 936, and 1020 nm. Both Microtops are periodically calibrated (once in a year) alternately so that the quality of data is maintained over a long period of time. The calibrations were done at Mauna Loa, using Langley technique, to the National Institute of Standards and Technology (NIST, USA) traceable standards. The calibration drift found in any wavelength channel is adjusted at the time of calibration by changing the irradiance calibration factors for that channel to maintain the accuracy. Moreover, the Microtops-measured spectral AODs have been found in very good agreement with the measurements using other radiometers such as Cimel [e.g., Evgenieva et al., 2008; Giles et al., 2011], Prede [e.g., Evgenieva et al., 2008], MFRSR (Multi-Filter Rotating Shadowband Radiometer), and multi-wavelength solar radiometer [Kompalli et al., 2010]. In the present study, the AOD data for only four wavelengths (340, 500, 870, and 1020 nm), which are common to both the Microtops, have been used. The consistency of the two Microtops data was checked by inter-comparing the measurements. The comparative study of AOD at 500 nm using the two Microtops, for all the periods (from year 2003 to 2011) in which the collocated measurements are available is shown in Figure 1. The scatter points denoted by different symbols constitute the AODs corresponding to each of the collocated measurements and the straight line denotes the linear least square fit to the entire data points. The correlation coefficients and the slopes of the least square fit to the individual data sets are given in Table 1. One can see that the two AODs are in excellent mutual correlation with a slope of ~1.0 and R2 value of >0.99 for all the years.

Figure 1.

Comparison of AODs at 500 nm between the two Microtops used in the study. The scatter points with different symbols represent the AOD values at each period of the collocated measurements and the straight line denotes the linear least square fit to the entire data points.

Table 1. The Details of the Comparative Study Between the Two Measurementsa
Period of StudySlopeR2
  1. a

    The correlation coefficients and slopes of the linear least square fit between the AODs (500 nm) using the two Microtops for each of collocated measurements.

2003 (Apr–Jun)0.9980.988
2004 (Nov–Dec)1.0010.992
2005 (Jan–Feb)0.9920.998
2006 (Jan–Feb)0.9990.998
2007 (Apr–May)0.9940.997
2008 (Jan–Jun)0.9951.000
2009 (Feb–May)0.9990.999

[8] The Microtops observations were made during the clear sky periods. Cloud contamination is strictly avoided in the measurements as the data collection was done only when the region of sky of ~10° around the Sun was free from clouds. Still, the sub-visible cirrus clouds (not observed by the instrument operator) are a potential contributor to the biases in the retrieved optical depth. In order to avoid this, the outliers lying beyond 2σ level (σ being the standard deviation) have been removed from individual day's observations. The data were taken at half hour intervals during daytime (about 09:00 to 16:00 h) with a minimum of three consecutive observations at a time (within a short span <20 s). Out of these three successive observations, the one with minimum AOD at 500 nm was used for further analysis, in order to ensure better pointing of Microtops toward the Sun, as the minimum AOD will correspond to maximum pointing accuracy [Morys et al., 2001; Porter et al., 2001; Ichoku et al., 2002; Singh et al., 2010]. Total number of such 14,651 quality-checked observations has been used in the present study. The daily average values were calculated only for those days which had at least six observations during the day. Such 1580 clear days of data obtained during 6 December 2001 to 11 May 2012 (which constitutes ~40% of the total days) have been used to study the seasonal variations and long-term trends in AOD at Delhi. From the daily averages, the corresponding monthly and seasonal average values were computed. The meteorological data were obtained using an Automatic Weather Station (AWS), co-located with the aerosol measurement site.

3 Prevailing Meteorology

[9] In addition to the dynamical processes such as convection and turbulence mixings within the boundary layer, the prevailing meteorological conditions and wind fields (mostly above the atmospheric boundary layer), have a great significance in modulating aerosol properties through convergence/divergence, removal, and size transformation processes [Aloysius et al., 2008; Saha and Moorthy, 2004, Kulmala et al., 2008]. The climatological pattern of the synoptic wind vectors derived during the study period from the reanalysis data of NCEP/NCAR (National Center for Environmental Prediction/National Center for Atmospheric Research) [Kalnay et al., 1996] at 850 hPa reveals strong seasonal variations, changing from strong westerlies to weak north-easterlies/south-easterlies, as the season changes from winter (December to February) to monsoon (July to September). During the transition periods of summer (March–June) and post-monsoon seasons (October–November), the winds were north-easterlies with moderate (~5 m s−1) speeds. The pattern is consistent with the long-term climatology of the region [Asnani, 1993]. These shifting synoptic conditions were found to influence the aerosol characteristics over the Indian region [Moorthy and Satheesh, 2000; Pillai and Moorthy, 2001; Gogoi et al., 2009]. In addition to the wind pattern, other two important meteorological parameters which directly affect the aerosol distribution are the precipitation and relative humidity [Radke and Hobbs, 1991; Hoppel et al., 1994; Eck et al., 2012]. The climatology of the seasonal average relative humidity (RH, in %) and seasonal total rainfall (in mm) has been obtained from the long-term database (2001–2012) at the station and the resulting pattern is shown in Figure 2. Seasonally, the highest value of relative humidity (Figure 2a) is observed during winter (~75%) followed closely by the monsoon (72%) and the lowest value is observed during the summer season. However, the mean seasonal total rainfall (Figure 2b) is highest during monsoon (~410 mm on an average), with a contribution of ~69% to the annual total. The summer rainfall constitutes ~24% of the total and the remaining 7% is during post-monsoon and winter seasons. Spectral AOD measurements were carried out in the background of these prevailing synoptic scale features.

Figure 2.

The climatology of the seasonal mean (Winter, DJF; Summer, MAMJ; Monsoon, JAS; and Post-monsoon, ON) pattern of near surface measuremnts of (a) relative humidity and (b) total rainfall at the measurement station Delhi.

4 Observations and Results

4.1 Daily, Monthly and Seasonal Variation of AOD and Angstrom Parameters

[10] The temporal variations of AOD at 500 nm (as a representative) for the entire period of study (from December 2001 to May 2012) are shown in Figure 3a. The figure reveals large scatter of points with values of AOD, ranging from as low as ~0.18 to as high as 3.0. Among the 14,651 point observations, ~70% of values lie above 0.6 (at 500 nm) implying the prevalence of highly turbid atmosphere over Delhi, throughout the year. These observations are in line with the earlier reported values of AOD, although for a relatively much shorter durations, from Delhi [Singh et al., 2005, 2010; Pandithurai et al., 2008], and are also comparable with the previous findings from other sites in the IGP [Jethva et al., 2005; Singh et al., 2006; Giles et al., 2011; Eck et al., 2012].

Figure 3.

The temporal variations of (a) AOD at 500 nm, (b) Angstrom exponent α, and (c) turbidity coefficient β during 6 December 2001 to 11 May 2012 at Delhi.

[11] The Angstrom parameters such as the Angstrom wavelength exponent, α and turbidity parameter, β have been estimated from the individual measurements, in the spectral range 340–1020 nm, using the Angstrom relation τ(λ) = βλα [Angstrom, 1964], where, τ(λ) denoted the AOD at wavelength λ (in µm). The time series of these parameters are shown in Figures 3b and 3c. The Angstrom exponent, α, which is an indicator of the dominance of fine/accumulation-mode aerosols in the size distribution, mostly varies between 0 and 1.0, except on few occasions when it lies beyond this range. During summer season, we find more frequent occurrence of near zero or slightly negative values of α, as the AOD at longer wavelengths exceeds that at shorter wavelengths, indicating the relative dominance of mineral dust aerosols in the atmosphere. Among the daily averages, the highest values of AOD (at 500 nm) are observed (~3.0) on 6 June 2003 and the lowest values on 1 March 2004 (~0.2). The Angstrom turbidity parameter β, which is an indicator of the columnar aerosol abundance or loading, shows similar variations as that of AOD. Significant seasonal variations are observed for all the three parameters. The highest values of AOD (500 nm) and β can be seen during the summer months of May–June, followed closely by post-monsoon season and the lowest values by March, when the air mass is in transition [Asnani, 1993]. A nearly opposite pattern for α is observed with winter/post-monsoon high and a summer low. The seasonal pattern is consistent for all the years. Previous studies from the Indo-Gangetic Plain also suggest that the size distribution of the aerosols undergoes strong seasonal variations with fine/accumulation-mode dominance (α > 1) during winter/post-monsoon seasons and coarse-mode dominance (α < 1) during summer/monsoon seasons [Singh et al., 2004].

[12] With a view to examine the seasonal, annual, and inter-annual variations clearly, the monthly mean values of spectral AODs and Angstrom parameters were estimated from the time series of the daily mean data. The contour map of the spectral AOD variations for the entire period of study is shown in Figure 4a. The monthly mean values are in the range of ~0.25–1.5 at 500 nm. It shows strong seasonal/annual variations in both spectral AODs and in spectral gradients, along with a weak but significant inter-annual variability. The seasonal changes in the spectral variation of AOD are clear from the figure with comparatively steeper spectra during winter/post-monsoon season and exceptionally flat spectra during summer. The steeper spectra are associated with the dominance of fine/accumulation-mode aerosols whereas the flatter spectra show the abundance of coarse-mode aerosols. It is interesting to observe a weak decreasing trend in the spectral AODs as the year progresses from 2001 to 2012. Among the Angstrom parameters (Figure 4b), the highest value of α (~1.1) is observed during January 2008 and the lowest value (~0.1) during July 2002. The turbidity parameter β shows opposite pattern as that of α. Similar seasonal variations for AOD and Angstrom parameters have also been reported in several earlier studies from other sites in the Indo-Gangetic Plain [Singh et al., 2004; Singh et al., 2005; Dey et al., 2005; Prasad and Singh, 2009; Gautam et al., 2010, 2011]. The extremely high value of β (~1.5) and higher values of spectral AODs, especially at longer wavelengths (Figure 4a), occurring during June 2003 is associated with intense dust storm activities. The anomalous and sustained dry conditions that prevailed over the entire north India during the period 2002–2003 was favorable for the occurrence of more frequent and intense dust storm episodes as well as for the accumulation of aerosols, leading to heavy aerosol loading over the IGP, which lies downwind to the dust outflow from the arid west Asia/northwest India [Kaskoutis et al., 2012]. Even though the seasonal pattern is strong in β, it is interesting to observe a gradual decrease in the amplitude of the peaks in summer from 2001 to 2012. The decrease in the peak is sometimes abrupt depending on the dry conditions and the frequency of dust storms. But in general, it indicates that the columnar aerosol loading undergoes a decreasing tendency from 2001 to 2012. This aspect has been examined later in the section 4.3.

Figure 4.

Monthly mean pattern of spectral AODs and Angstrom parameters, (a) the contour map showing the spectral variation of AOD and (b) time series of α and β.

4.2 Climatology of AOD and Angstrom Parameters

[13] As the measurements were spanning for a period of 11.5 years, the database is robust to generate the climatological pattern. The climatology of monthly and seasonal mean spectral AODs as well as the Angstrom parameters has been estimated. Figure 5 shows the monthly mean climatology of spectral AODs (Figure 5a) along with the Angstrom parameters (Figure 5b). The pattern brings out spectrally asymmetric seasonal variations. The variation of AOD at shorter (340 nm) and at mid-visible wavelengths (500 nm) is quite different from that observed at longer wavelengths (870 and 1020 nm). AODs at 340 and 500 nm depict highest values during the post-monsoon period of October–November (~1.1 at 340 nm), decrease throughout winter to reach the lowest values by March (~0.77 at 340 nm), increase continuously toward summer, and drop again in the monsoon season. Whereas, at longer wavelengths, AOD shows the lowest value during February (~0.35 at 1020 nm), increases gradually to reach the highest values by June (~0.75 at 1020 nm), and decreases thereafter. Similar climatological pattern is observed for Kanpur in the eastern IGP region [Singh et al., 2004] also. The climatological average values along with the number of observations, which constitute the mean values, are given in Table 2.

Figure 5.

The monthly mean climatology of (a) spectral AODs and (b) Angstrom parameters.

Table 2. Monthly Mean and Standard Deviation of AOD (500 nm) and Angstrom Parameters During 2001–2012 at Delhi
MonthsAOD500 (τ)αβNo. of Data Points
January0.82 ± 0.171.00 ± 0.050.40 ± 0.101400
February0.64 ± 0.100.92 ± 0.090.34 ± 0.061508
March0.61 ± 0.130.64 ± 0.130.39 ± 0.091604
April0.70 ± 0.070.49 ± 0.080.51 ± 0.061611
May0.88 ± 0.160.45 ± 0.120.67 ± 0.151434
June0.92 ± 0.220.37 ± 0.090.75 ± 0.20917
July0.88 ± 0.260.64 ± 0.300.61 ± 0.28429
August0.72 ± 0.220.66 ± 0.270.47 ± 0.15312
September0.64 ± 0.130.68 ± 0.110.39 ± 0.06947
October0.90 ± 0.230.89 ± 0.120.45 ± 0.091394
November0.91 ± 0.150.96 ± 0.050.44 ± 0.071350
December0.85 ± 0.100.99 ± 0.040.40 ± 0.061745

[14] In the present study, values of AODs remained much higher throughout the year, particularly when it is compared with the average AOD values observed at high-altitude Indian stations such as Leh and Hanle [Singh and Singh, 2004], Nainital [Sagar et al., 2004; Ram et al., 2010], Kullu [Guleria et al., 2012], and with the oceanic regions such as Arabian Sea [Satheesh et al., 2006a] and Bay of Bengal [Moorthy et al., 2003; Satheesh et al., 2006b; Beegum et al., 2012]. The AOD at Delhi is slightly higher compared to AODs at some of the other Indian stations like Ahmedabad [Ganguly et al., 2006], Pune [Pandithurai et al., 2007; Dani et al., 2012], Visakhapatnam [Niranjan et al., 2011], Bangalore [Babu et al., 2002; Vinoj et al., 2004], and Trivandrum [Moorthy et al., 2007], but quite comparable with the values at the IGP stations such as Kanpur [Singh et al., 2004], Kharagpur [Niranjan et al., 2006], and Bhubneshwar [Das et al., 2009]. More details of seasonal AOD values at various Indian stations are also given in Table 3. High AODs in the IGP region are attributed mainly to the widespread industrial activities, clustering of coal-based industries and thermal power plants [Girolamo et al., 2004; Nair et al., 2007], large density of population, and agricultural activities, along with the typical orography of the region.

Table 3. A Comparison of Seasonal Mean AOD and Angstrom Exponent (α) at Different Locations in India
LocationPeriodAOD (500 nm)αReference
Arabian Sea1995–20020.290.47  0.700.30  Satheesh et al. [2006a]
Bay of Bengal2000–2004 0.48 0.19 0.50  Satheesh et al. [2006b]
Port Blair2002–20080.310.26  0.931.1  Beegum et al. [2012]
Minicoy1995–19980.32   1.2   Satheesh et al. [2002]
Trivandrum2000–20030.430.400.290.381.00.850.321.20Moorthy et al. [2007]
Goa2000–20020.410.48  1.481.14  Suresh and Desa [2005]
Anantapur2005–20060.460.500.370.40    Kumar et al. [2009]
Hyderabad2008–20090.480.57  1.04 0.84 Sinha et al. [2012]
Pune2000–20040.380.42  1.40   Pandithurai et al. [2007]
Ahmedabad2002–20050.320.420.430.431. et al. [2006]
Kanpur2001–20030.570.540.660.631.260.600.661.12Singh et al. [2004]
Dibrugarh2001–20070.310.450. et al. [2009]
Manora Peak2005–20080.110.300.150.09    Ram et al. [2010]
Kullu2006–20090.200.340. et al. [2012]
Kharagpur20040.7  1.2    Niranjan et al. [2006]
Delhi2001–20120.770.780.740.910.970.490.660.93Present study

[15] The Angstrom turbidity parameter showed similar variations (both monthly and seasonal) as that of AOD at longer wavelengths (Figures 5b) with highest aerosol loading during June (summer) and lowest during February (winter). But, the Angstrom wavelength exponent, α varied between 0.35 and 1.0 (average ~0.75 ± 0.24) as season progressed from summer to winter. The highest values of α observed at this highly polluted measurement site is ~1.0 (during winter), which is comparable or even lower than those observed at other Indian locations [Moorthy et al., 2003, 2005, 2007; Vinoj et al., 2004; Sagar et al., 2004; Ramachandran, 2005; Niranjan et al., 2006; Beegum et al., 2008; Gogoi et al., 2009]. The seasonal mean values of α at different Indian locations are given in Table 3. Our results indicate that ~71% of the α (point observations) values lie between 0 and 1, and 28% of the values are in the range 1 and 2. Even during winter/post-monsoon season, ~45% of the values of α are less than 1. This suggests that the measurement site is generally dominated by relatively larger particles in comparison with the other urban cites in India.

4.3 Trends in Spectral AODs and Angstrom Parameters

[16] Notwithstanding the consistency in the seasonal pattern of aerosol distribution over the entire period, one can observe some weak but significant trends in AODs. Figure 6 presents the time series of anomalies in the mean monthly spectral AODs and Angstrom parameters during the study period after deseasonalizing the time series by removing the climatological monthly means from the respective monthly mean values. For trend detection, a linear trend model has been employed to the time series data, following Weatherhead et al. [1998]. According to this, the time series of any parameter Pt, at time t (in months), can be given as

display math(1)

where μ is a constant, ω is the magnitude of the trend per year, St is the seasonal component, and Nt represents the noise. The time series data have been subjected to deseasonalizing, and the resulting time series of the anomaly of the parameter (At) can be written as

display math(2)
Figure 6.

The deseasonalized pattern of monthly mean (a–d) spectral AODs and (e and f) Angstrom parameters from December 2001 to May 2012 at Delhi. The solid line represents the trend line.

[17] The noise variability Nt is assumed as the autoregressive of the order of one:

display math(3)

where ϕ is the autocorrelation function and εt is the independent random error. Here ϕ allows the noise factor at time t, Nt to be dependent of the noise at t − 1, which will be the case if the variability is not random and persists over time. The εt is assumed to be random with variance σε2. The error in the trend estimate, σω can be estimated from the autocorrelation ϕ, the number of years of data n, and error in the random variable σε:

display math(4)

[18] The above relation indicates that the precision of the trend is a function of the magnitude of the random variability in the data, the autocorrelation of the noise, and the data length [Weatherhead et al., 1998]. The number of years of data required to detect the trends described by equation (1) in 95% confidence level is estimated following Weatherhead et al. [1998] as

display math(5)

[19] The trend, ω has been estimated by applying generalized least square (GLS) fit model to equation (1).

[20] The values of trends observed in the columnar spectral AODs at Delhi are found to be statistically significant. The solid line in Figure 6 represents the trend line fitted using GLS estimator. The figure shows a decreasing trend in spectral AODs with approximately −0.017/year at 500 nm (Figure 6b), and the absolute magnitude of the trend increases slightly toward longer wavelengths to reach approximately −0.020/year at 1020 nm (Figure 6d). Similar analysis of the Angstrom parameters yielded a similar decreasing trend for β (−0.019/year) and an increasing trend for α (Figures 6e and 6f). The values of n* estimated using equation (5) were found to vary from 9 to 11 years, except for AOD at 340 and α, which is slightly less than the data length used for the present study. More details of statistical analysis pertaining to the trend estimation are given in Table 4. The observed decreasing trend in AODs with increasing magnitudes toward longer wavelengths and β (approximately −0.02/year) suggests a gradual decrease in the columnar aerosol loading resulting from the decrease in the concentrations of coarse-mode aerosols.

Table 4. The Statistical Details of Trends Analysis of Spectral AODs and Angstrom Parameters
ω (per year)−0.016−0.017−0.018−0.0200.012−0.019

[21] Analyses of trends have also been done in the yearly averaged data, after removing the bias by subtracting the climatological yearly mean values from each of the yearly mean values (not shown in figure). A gradual increase in the absolute values of the trends is observed in spectral AODs as the wavelength progresses from UV to near-IR range. At the shorter wavelength of 340 nm, the observed trend value is −0.018/year, decreasing slightly to −0.019/year at 500 nm. At 870 nm, the trend became substantially low (−0.02/year) and at 1020 nm, it reaches as low as −0.025/year. Similar analysis of the yearly anomaly of the Angstrom parameters also depicts an increasing trend in α (0.012/year) and a similar decreasing trend in β (−0.022/year). The result signifies that columnar aerosol properties are governed by the coarse-mode aerosols, mostly desert dust [e.g., Dey et al., 2004; Pandithurai et al., 2008; Gautam et al., 2009; Srivastava et al., 2012], and undergoes a decreasing trend during the study period. It is likely that this trend may be modulated by the pre-monsoon heavy dust loading, particularly during the first half of the observation, which declined during the later half.

[22] The variations of spectral AODs clearly indicate that the size spectrum of aerosols undergoes large deviations seasonally. At the same time, the simple Angstrom exponent, α does not convey this, as the values of α are more or less the same not only during winter and post-monsoon seasons but also during summer and monsoon seasons. In order to visualize the fine seasonal changes in the microphysical properties, we have examined the derivative of the Angstrom exponent in the next section.

4.4 Derivatives of the Angstrom Wavelength Exponent (α′)

4.4.1 Theory

[23] The curvature in the AOD spectra could be expressed as a second-order polynomial of the form as

display math(6)

where a0, a1, and a2 are the constants. The coefficient a2 accounts for the “çurvature” as the second derivative of equation (6) with respect to ln λ will yield “−2a2”. The second derivative of ln τ with respect to ln λ (denoted by α′) centered around 500 nm has been calculated numerically using the AODs at three wavelengths 340, 500, and 870 nm following Li et al. [1993] as

display math(7)

[24] The derivative of Angstrom exponent, α′ quantifies the deviation from linearity of the spectral AOD variation in logarithmic scale. The value of α′ close to zero indicates a constant slope for the AOD spectra with insignifciant curvature, while higher values (either positive or negative) suggest the presence of curvatures of either concave (negative values of α′) or convex (positive values of α′) types. While the Angstrom exponent is a mere indicator of the aerosol size distribution, its derivative gives a clearer picture of the possible type of aerosols and in turn the potential sources also. Positive values of α′ indicate the dominance of fine/accumulation-mode aerosol particles while negative values of α′ suggest the relative dominance of coarse-mode aerosol particles in the bimodal aerosol size distribution. More details on α′ can be found in several previous studies [Eck et al., 1999, 2001; Reid et al., 1999; O'Neill et al., 2001a, 2001b, 2002; Schuster et al., 2006; Kaskaoutis et al., 2006a, 2006b; Beegum et al., 2009; Soni et al., 2010].

4.4.2 Monthly and Seasonal Variations of α

[25] The derivative of α (in wavelength domain), denoted as α′, has been estimated using AODs at three consecutive wavelengths centered at 500 nm (340, 500, and 870 nm) employing equation (7). Figure 7a shows the time series of monthly mean pattern of α′ along with α. The pattern of α′ is more or less similar to α. Interestingly, one can observe more than one value of α′ (even with opposite sign) for a particular α value, especially during the months in which the prevailing air mass is in transition [e.g., Beegum et al., 2009]. These observations indicate that the changes in the microphysical properties are not accurately reflected in α. In addition, the climatological pattern of α′ (Figure 7b) is significantly different from that of α. Unlike α, highest positive values of α′ are observed during the post-monsoon month of November, and toward winter, the values reduce significantly. During summer, the reversal of the spectral curvature occurs with negative α′ values. Close to zero values of α′ are observed during monsoon season. These values of α′ indicate strong seasonal variations with fine-mode-dominated aerosol size distributions (mainly of anthropogenic, biomass burning, urban, and industrial origin) during post-monsoon/winter, and bimodal distributions with a strong coarse-mode during summer/monsoon seasons.

Figure 7.

The time series of monthly mean pattern of (a) the derivative of Angtrom exponent α′ and (b) its climatology.

5 Discussions

[26] The seasonal variations in aerosol characteristics are attributed to a combination of factors, viz., natural and anthropogenic sources (either local or remote) of different types, sinks (washout due to precipitation), local dynamics, and synoptic meteorology. The climatological pattern of AODs at Delhi (Figure 5a) indicates the prevalence of highly turbid atmosphere throughout the year (minimum AOD of ~0.6 at 500 nm) which suggests that the contribution of local sources (either situated locally or confined to northwest India and Indo-Gangetic Plain) plays a major role toward the columnar aerosol abundance [Dey et al., 2004; Singh et al., 2004; Girolamo et al., 2004; Gautam et al., 2007]. Superimposed on this high background level of AOD, the observed seasonal changes are contributed by the local dynamics and meteorology along with the advection from faraway locations [Aloysius et al., 2008]. Among the local sources, the vehicular emission and the frequently occurring dust storms, particularly in the northwest India including western IGP during summer are the major contributors. Although, there has been a significant reduction in the concentrations of PM10, SO2, sulfate aerosols, and CO concentrations at Delhi due to the implementation of metro transportation and conversion of diesel/petrol-based public vehicles to compressed natural gases, the effect has been negated by the large increase in the population of diesel-fueled cars [Narain and Krupnick, 2007]. Thus, vehicular emission is still a major contributor to the local pollution (~80%). Even though these anthropogenic emissions are independent of seasons, the impact is more during winter season due to the confinement of aerosols (poor horizontal and vertical advection) added with the insignificant removal mechanisms. This causes fine/accumulation-mode-dominated size distribution with positive values of α′ during winter. Toward summer, as the land gets dry due to increased solar heating, the conditions become conducive for picking up of dust from the arid and semiarid regions of central and western India. The dust storm episodes produce enormous amount of coarse-mode natural aerosols causing substantial reduction in visibility and radiation flux reaching the surface during summer [Singh et al., 2005; Prasad et al., 2007; Pandithurai et al., 2008; Gautam et al., 2010].

[27] The transported desert dust mainly from Southwest Asian arid regions including Iran, Afghanistan, Pakistan, Arabian Peninsula, and northwestern India (Thar Desert) is known to modulate the optical and radiative properties of aerosols over the IGP during summer [Grigoryev and Kondratyev, 1981; Middleton, 1986; Prospero et al., 2002; Dey et al., 2004; Singh et al., 2004; Singh et al., 2005; Jethva et al., 2005; Chinnam et al., 2006; Prasad and Singh, 2007; Pandithurai et al., 2008; Beegum et al., 2008; Gautam et al., 2009]. This is responsible for the highest values of AOD (especially at longer wavelengths) as well as β, and highly negative values of α′ observed at the measurement location during May–June. The intense spells of monsoon rainfall reduces the aerosol loading during the season, leading to a sharp decrease in AOD and subsequent increase in α. The further enhancement of fine/accumulation-mode aerosols during post-monsoon season is due to increased anthropogenic activities associated with festivals of Dushara and Diwali. During this time, a period of ~3 weeks from the second half of October to the first week of November, the extensive burning of crackers produces large amount of aerosols resulting in a smoky atmosphere [Attri et al., 2001, Singh et al., 2003]. This contributes significantly toward the observed highly positive values of α′ and highest value of AOD at shorter wavelengths during the post-monsoon season.

[28] The microphysical properties of aerosols at Delhi become clearer from Figure 8, which shows the variation of α′ as a function of AOD at 500 nm. Generally, more than 70% of the AOD spectra measured over Delhi have positive α′ values (e.g., α′ > 0) with convex curvature during winter and post-monsoon seasons, whereas during summer and even monsoon seasons, the AOD spectra are of concave types with negative or close to zero values of α′. In addition, the figure also shows that the effect of curvature is greater even at the conditions of high turbidity. This is clear from the scatter plots for different seasons. The curvatures are found to be significant (α′ > 1.0) even at high turbidity conditions (AOD > 1.0), especially during winter and post-monsoon seasons. During these seasons, the values of α′ are highly positive with the abundance of fine/accumulation-mode aerosols. Another interesting point to be noted, especially from the scatter plot for winter season is that, as the aerosol loading increases further (AOD > 2.0), α′ decreases. This points toward the presence of significant amount of larger particles in the vertical column even though the season is known for anthropogenic dominance. The reduction in the fine/accumulation-mode aerosol concentrations toward winter (indicated by the values of α′) might be due to some size transformation processes. At this juncture it is imperative to examine the pattern of the relative humidity observed at the site (shown in Figure 2a). It is interesting to observe higher values of relative humidity during winter season unlike the pattern observed at other parts of India. Even though the seasonal average value of relative humidity is ~75%, the values were much higher, reaching as high as 95%, during the peak winter time (second half of December to first half of January). This would lead to hygroscopic growth of aerosols [e.g., Jeong et al., 2007; Eck et al., 2012], which enables the particles in the fine/accumulation range to enter into the coarse-mode regime. Similar findings have also been reported from the eastern part of Indo-Gangetic Plain during winter season [Singh et al., 2004; Niranjan et al., 2006; Gautam et al., 2007].

Figure 8.

Scatter plots of AOD at 500 nm versus α′ (top) for the entire period from 06 December 2001 to 11 May 2012 and (bottom four figures) for different seasons.

[29] During summer season, coarse-mode dust aerosols dominate, which is also indicated by the negative values of α′. As the removal mechanisms being strong during monsoon season, the aerosol loading, particularly the coarser particle concentration, has been reduced considerably and this effect is reflected in the values of α′ too. The insignificant spectral curvature with close to zero values of α′ during the season might be due to the averaging of several positive and negative curvatures. This suggests to the addition of one or more modes to the size distribution, probably transported from distant source regions. Toward the end of monsoon season, the atmosphere is comparatively cleaner with reduced amount of coarse-mode aerosols. As discussed earlier, the locally produced anthropogenic aerosols from the burning of crackers alone is not sufficient for a persistent fine/accumulation-mode-dominated size spectrum with highest values of α′ during post-monsoon season, and this points toward the contribution by advection as well.

[30] s discussed earlier, the natural coarse-mode aerosols, mainly soil dust, play a major role in the columnar abundance at Delhi. The observed decreasing trend on long term could be either due to reduction in the production of coarse-mode aerosols or due to the efficient removal mechanisms. The time series of monthly total rainfall during the study period has been examined at the measurement site and absolutely no trend in precipitation is observed (figure not shown). Hence, the observed deceasing trend in aerosol loading might be either due to the reduction in the local production of dust or reduced advection or a combination of both. The more frequent occurrence of sudden peaks in the time series of AOD and β (Figure 3) during the first half of the study period indicates the more frequent occurrence of dust storm events in comparison with that during the second half. Recently, Dey and Girolamo [2011], based on 10 years of AOD data (2000–2010) retrieved from Multiangle Imaging Spectroradiometer, have reported a decreasing trend in the pre-monsoon and monsoon aerosol loading over the IGP region. Ramachandran et al. [2011] though reported a weak increasing trend in annual average AOD values (4.9%) over the Delhi region, based on the satellite data; they have also reported a decreasing trend in the summer AODs during 2000–2008.

5.1 Strength of Potential Sources and CWT Analysis

[31] The simple air mass back trajectory analysis can only identify the advection pathways of aerosols at the receptor site [Jose et al., 2005; Rousseau et al., 2004]. Hence, the potential source contribution function (PSCF) and the concentration weighted trajectory (CWT) analysis have been developed recently to quantify the regional contribution of each of advection pathways toward the measured aerosol parameters (either mass concentrations or column AOD). The CWT analysis is however more sophisticated method than PSCF as it gives the spatial pattern of the potential sources of aerosols arriving at a receptor location [Seibert et al., 1994; Hsu et al., 2003; Vinoj et al., 2010]. In this technique, weightings have been given to each of the trajectories reaching the receptor site on the basis of the mean value of the measured aerosol parameter (AOD in this case). Hence, each grid point gets a weighted value obtained by averaging the data (AOD) measured at the receptor site as the associated trajectory crosses the grid point (i,j) as

display math(8)

where Cij is the average weighted AOD in the grid cell (i,j), M is the total number of trajectories, l is the trajectory index, Cl is the measured AOD at the receptor location (here Delhi) associated with the trajectory l, and τijl is the residence time of the trajectory l in grid cell (i,j). The weighted AOD values at each grid thus obtained represent the AOD that can be expected at the receptor site, at any time, if the trajectory was passing through that spatial grid. For the present analysis, 5 day isentropic air mass back trajectories arriving at the receptor location at the height level of 1800 m from the mean sea level (above the atmospheric boundary layer) have been calculated for all the years, and these trajectories were segregated for each of the four seasons (winter, summer, monsoon, and post-monsoon).

5.1.1 Potential Source Regions

[32] Figure 9 shows the seasonal changes in the strength of various advection pathways based on CWT analysis. During winter season, the contribution by the cluster of local trajectories confined to the Indo-Gangetic Plain and northwest India is found to contribute more to the seasonal pattern (~0.8). In addition, contribution of the trajectories from the west of the measurement location, especially from Pakistan and Afghanistan is also significant. Even though, there were many trajectories from far north-west Asia, their contribution toward the measured AOD is insignificant (<0.2). Earlier investigators have already reported that during winter season, the entire North India including IGP is dominated by anthropogenic fine/accumulation-mode aerosols [Nair et al., 2007; Gautam et al., 2007; Verma et al., 2011]. Studies from the north-west Asian cities such as Lahore have also pointed out that those regions are heavily polluted with fine/accumulation-mode aerosols [e.g., Stone et al., 2010].

Figure 9.

The concentration weighted trajectory analysis showing the impact of potential sources and its seasonal variations at the measurement location Delhi.

[33] Toward summer season, the pattern changes significantly with higher contribution of advection from the entire northwest India, which includes the Thar Desert of Rajasthan and the eastern part of Pakistan with intense dust storm activities. This, along with the locally produced dust from the dry-arid land, would result in the abundance of coarse-mode natural aerosols at the measurement site leading to highest values of AOD (especially at longer wavelengths) and lowest values of α. Thus, the advection of dust through the higher levels would also contribute significantly to the observed pattern. During monsoon season, the highest contribution of advection is from the oceanic region of Arabian Sea (>0.4) and Bay of Bengal through the Indo-Gangetic Plain. These marine air mass would bring significant coarse-mode sea-salt aerosols to the measurement site [Tiwari et al., 2009]. In addition, significant contribution of long-range-transported mineral dust aerosols from the deserts of Arabian Peninsula has also been identified in modifying the aerosol properties over IGP during the season [Chinnam et al., 2006; Prasad and Singh, 2007; Gautam et al., 2010]. However, the accompanying rainfall would reduce the aerosol loading (mainly coarse-mode aerosols) and this would result in sharp decrease in AOD and subsequent increase in α.

[34] During post-monsoon season, advection of aerosols from the northern part of India (which includes the vast agricultural fields of Haryana, Punjab, Uttarakhand, and Himachal Pradesh) dominates as evident from the very high percentage contribution (>90% of the total) of the air mass back trajectories which originate and/or traverse through those regions. It is already reported in the literature that the burning of crop residues in the agricultural fields is the major source of anthropogenic aerosol loading, particularly from the north and northwest Indian states, during October–November period [Badarinath et al., 2009; Sharma et al., 2010]. Thus, the extremely high values of AOD, particularly at shorter wavelengths, and the sudden increase in the values of α are possibly due to the strong advection of the aerosols produced due to the burning of agricultural wastes. This, added with the local anthropogenic activities, results in the dominance of anthropogenic aerosols at the measurement location.

6 Conclusions

[35] Analysis of the long-term measurements of columnar spectral AOD (December 2001 to May 2012) over Delhi, an urban station in the western Indo-Gangetic Plain, unveiled certain interesting feature which are summarized below:

  1. [36] Long-term climatology of AOD from the surface measurements over Delhi shows quite turbid atmosphere throughout the year. The climatological monthly mean AODs at shorter wavelengths peak twice, during June and November, while at longer wavelengths it shows only one peak in June. The Angstrom exponent, α showed a near inverse relationship with a summer low and winter/post-monsoon high.

  2. [37] The AODs are consistent with other long-term measurements of aerosols in the IGP (e.g., MWR, AERONET) with similar variations in magnitudes and size, indicating variable concentrations and types of particles (such as dust, pollution, or mixture of them) at Delhi.

  3. [38] The most striking result from the present study at Delhi is the statistically significant decreasing trend observed in the columnar aerosol loading with a value of approximately −0.02/year at 500 nm. Similarly, a decreasing trend is also observed for β and a subsequent increase is observed for α (0.012/year). The observed trend might be largely modulated by enhanced dust activities during the first half of the time series data.

  4. [39] The derivative of the Angstrom exponent, α′ showed seasonally changing microphysics of aerosols over Delhi with positive values (convex spectral curvature, indicating dominance of coarse-mode natural aerosol) during post-monsoon and winter seasons and negative or close to zero values (concave spectral curvature, indicating fine/accumulation-mode abundance) during summer and monsoon seasons.

  5. [40] The contribution by advection from distant sources varies over the seasons, thereby producing significant seasonal variation in the columnar aerosol loading as well as in the size distributions at Delhi. While the cluster of local trajectories confined to the Indo-Gangetic Plain and northwest India modulates the aerosol properties during winter, the advection from the arid northwest India and west Asia modifies the aerosol properties during summer. The monsoon rainfall and the land wetness along with the advection from the oceanic regions of Arabian Sea and Bay of Bengal reduce the aerosol loading (mainly dust) during monsoon season. Toward post-monsoon season, the advected aerosols form the effluents from the burning of the crop residues from the agricultural fields in the north and northwest India along with the locally produced aerosols from the burning of crackers associated with the festivals lead to the dominance of anthropogenic aerosols at Delhi.


[41] One of the authors NKL is thankful to the Council of Scientific and Industrial Research (CSIR) for providing the senior research fellowship. We also thank the NOAA Air Resources Laboratory for the provision of the HYSPLIT transport and dispersion model and READY website ( A part of this work was also sponsored by ISRO GBP (Indian Space Research Organisation-Geosphere Biosphere Programme). Authors are thankful to Ritesh Gautam, NASA/GSFC, USA, and the other anonymous referees for their valuable comments and suggestions in improving this paper.