Aerosol radiative forcing during dust events over New Delhi, India



[1] We present results from sun/sky radiometer measurements of aerosol optical characteristics carried out in New Delhi during March–June, 2006, as part of the Indian Space Research Organization's Integrated Campaign for Aerosol Radiation Budget. For the first time at this site, derived are parameters such as aerosol optical depth (AOD), single scattering albedo (SSA), asymmetry parameter, Ångstrom exponent, and real and imaginary refractive indices in five spectral channels. During the campaign, a consistent increase in aerosol loading from March to June with monthly average AOD values at 0.5μm of 0.55, 0.75, 1.22 and 1.18, respectively, was observed. Ångstrom exponent gradually decreases from 1.28 (March) to 0.47 (June), indicating an increased abundance of coarse particles due to dust storms that transport desert dust from the Thar desert and adjoining regions. SSA at 0.5 μm is found to be in the range of 0.84 to 0.74 from March to June, indicating an increasing contribution from the mixture of anthropogenic and desert dust absorbing aerosols. Optical properties derived during the campaign are used in a radiative-transfer model to estimate aerosol radiative forcing at the surface and at the top-of-the atmosphere. A consistent increase in surface cooling is evident, ranging from −39 W m−2 (March) to −99 W m−2 (June) and an increase in heating of the atmosphere from 27 W m−2 (March) to 123 W m−2 (June). Heating rates in the lower atmosphere (up to 5 km) are 0.6, 1.3, 2.1, and 2.5K/d from March, April, May, and June 2006, respectively. Higher aerosol induced heating in the premonsoon period has been shown to have an impact on the regional monsoon climate.

1. Introduction

[2] Scattering and absorption due to atmospheric aerosols can cause surface cooling by blocking solar radiation reaching the surface, known as the dimming effect [Wild et al., 2005; Pinker et al., 2005]. In the Asian monsoon regions, the dimming effect is especially large owing to heavy pollution and frequent occurrence of dust storms [Wild et al., 2005; Lau et al., 2006]. Dust aerosols from neighboring deserts are transported over northern India between April and May. The presence of aerosols controls the cooling or heating of the Earth surface and the warming or cooling of the atmosphere. The increase in anthropogenic absorbing aerosols is found to be responsible for increase in rainfall in the southern part of China and drought in northern part of China [Menon et al., 2002]. According to Lau et al. [2006], heating in the lower atmosphere due to absorbing aerosols (mainly black carbon and dust) can act as an “elevated heat pump” that draws in warm moist air from below, leading to forced ascent, and enhanced convection, and increased summer rainfall over northern India. Aerosols have an indirect effect on clouds by increasing the cloud albedo and cloud lifetime [Penner et al., 2003; Twomey, 1977]. In contrast to this, some believe that radiatively absorbing aerosols warm the atmosphere, leading to suppression of precipitation due to evaporation of clouds. This is referred to as the aerosol semidirect effect [Ackerman et al., 2000; Johnson et al., 2004; Feingold et al., 2005]. According to the Intergovernmental Panel on Climate Change [2001] Fourth Assessment Report, the global average radiative forcing by aerosols is −1.2 W m−2, whereas it is about +2.6 W m−2 for greenhouse gases. Much attention has been paid to quantifying the radiative forcing by aerosols. The radiative heating/cooling by dust must also be taken into account to predict adequately the overall impact of aerosols on weather and climate because heating or cooling by dust alters atmospheric dynamics and thermodynamics [Quijano et al., 2000].

[3] Concentration of aerosols over the Indian region is found to be increasing, and studies report that the aerosol optical depth (AOD) over the northern part of India is higher as compared to the southern part [Singh et al., 2004; Sarkar et al., 2006; Gautam et al., 2007]. The AOD in the northern part of India shows an annual variability with higher aerosol loading during the dry season due to dust events. Moorthy et al. [2005] reported wintertime spatial characteristics of boundary layer aerosols over peninsular India based on a coordinated multiinstitutional road campaign experiment. Tripathi et al. [2006] reported characteristics of aerosols during foggy and hazy conditions over the Indo-Gangetic basin (IGB) during winter. Dust-storm events frequently occur in the IGB of northern India during premonsoon (March–June). On the basis of chemical composition and air mass trajectories [Chinnam et al., 2006; Prasad and Singh, 2007], the dust originates from three major sources: Oman, southwest Asian basins, and Thar Desert in Rajasthan. Chemical data from the above studies also indicate mixing of anthropogenic pollutants with the dust during transport. These dust storms apparently deposit silty materials in the downwind directions, as observed on the quartzite ridges in the Delhi area [Tripathi and Rajamani, 1999]. During the summer season, along with the dust, the wind carries heavy metals to the Indo-Gangetic basin [Yadav and Rajamani, 2003], causing severe air pollution and degradation in visibility. Middleton [1986] found higher frequency of dust event occurrence in the western part as compared to the eastern part of the IGB. New Delhi, one of the highly polluted mega cities in Asia, is situated in the western part of the IGB; it is faced with dust-storm events during the premonsoon season locally known as ‘Aandhi’ [Joseph et al., 1980]. During these events, wind-blown dust is brought in from the Thar Desert in Rajasthan and adjoining regions. Mineral dust aerosols enter the atmosphere through wind erosion of desert regions and disturbed soils [Miller et al., 2004; Tegen and Fung, 1994]. Dust mixes with anthropogenic aerosols in the polluted environment, and the resulting optical properties and the associated radiative impacts are regionally highly variable [Deepshikha et al., 2005]. Recently, Moorthy et al. [2007] reported that dust over the Great Indian Desert is more absorbent as compared to African dust and the inferred single scattering albedo (SSA) ranges between 0.88 and 0.94. Estimates of aerosol radiative forcing (ARF) over New Delhi are very sparse. Singh et al. [2005] reported on aerosol radiative forcing during the premonsoon months over New Delhi for the year 2003. However, their study is based only on AOD measurements, while the other parameters such as SSA and asymmetry parameter (ASP) are assigned according to the Optical Properties of Aerosols and Clouds (OPAC) as specified by Hess et al. [1998]. Ganguly et al. [2006] reported ARF over New Delhi during the winter season mainly as influenced by hazy and foggy conditions. Complete aerosol characterization over the IGB is also sparse and limited to the studies by Dey et al. [2004], Ramanathan and Ramana [2005], Ramachandran et al. [2006], and Singh et al. [2005]. In this study, we present aerosol optical properties over New Delhi, India, during the premonsoon months from March to June 2006 using a ground-based sun/sky radiometer data collected as part of the Indian Space Research Organization's Integrated Campaign for Aerosol Radiation Budget (ICARB).

2. Observations

[4] The instrument used in the present study is a sun/sky radiometer manufactured by PREDE, Ltd. It is an automatic sun-tracking sun/sky radiometer capable of measuring direct solar and diffuse sky radiance at five spectral channels. The filters are centered at wavelengths 0.4, 0.5, 0.675, 0.87 and 1.02 μm with a bandwidth of 0.01 μm. Column aerosol radiative parameters such as AOD, SSA, and ASP are derived from the sun/sky radiance measurements at five wavelengths from the visible to the near-infrared spectral regions using Version 4.2 of the Skyrad.Pack radiative transfer model [Nakajima et al., 1996]. Measurements of surface meteorological parameters (air temperature, wind speed and direction) were also made.

[5] The sun/sky radiometer can be calibrated onsite for solid view angle (Ω) and for absolute sensitivity (Vo). This is possible because of the speed of the tracking system and the use of a single detector for both direct and diffuse measurements. The sun/sky radiometer is operated once a month in the disc scan mode to estimate the solid view angles at different wavelengths as part of a recommended calibration procedure. Disk scan is performed by scanning the area of 2° × 2° around the solar disk from up to down and from left to right, with an angular resolution of 0.1°. Data from a disk scan performed on a clear day before the campaign were processed to determine Ω and were used in the analysis. The routine Langley method of calibration is based on measurements of direct radiation under the hypothesis that τ is constant during the calibration. Under most situations, such an assumption is not valid. It was demonstrated by Shaw [1976] that for a typical urban station, V0 deduced from the Langley method can have an error of ∼10%. Hence, an improved method of calibration based on both direct and diffuse radiation data [Nakajima et al., 1996] is used here. A retrieval with only R [R(θ) = VE(θ)/VmΩ] is performed for all available measurements, and the set of AOD (τ) values so obtained is used for deriving m0τ and then V0 by extrapolation along a V-m0τ graph. The difference between the standard Langley and the improved Langley methods for deriving the absolute calibration is shown in Figure 1. The calibration constant V0(λ) is estimated using the improved Langley plot technique [Soufflet et al., 1992] on selected clear sky days in January 2006 at Pune (18°32′N, 73°51E, 559 m AMSL) just before the ICARB campaign and is used in the retrieval analysis. Xiangao et al. [2004] also demonstrated the advantage of the modified/improved Langley plot method, showing that the calibration constant is stable for both forenoon and afternoon cases when this method was used instead of the normal Langley method. As seen from Figure 1, the improved Langley method shows much less scatter as compared to the normal Langley method.

Figure 1.

Langley and improved Langley plot method for deriving absolute calibration constant.

3. Experimental Site and Meteorology

[6] The sun/sky radiometer was operated on the campus of the Indian Institute of Tropical Meteorology, New Delhi (28.63°N, 77.17°E), an urban mega city and national capital of India. The selection of the campaign period was based on the following criteria: (1) a season during which atmospheric lifetime of aerosols is sufficiently long and (2) a season when probability of occurrence of severe weather phenomena like cyclones and deep depressions is low, so that the spatial distribution of the aerosol characteristics can be assumed to be temporally stationary.

[7] Figure 2 shows the variation in surface level temperature range and relative humidity during the campaign period. There was an overall increase in the surface air temperature from March becoming as high as 45°C in the month of May. In the first week of March, the temperature range (Tmax–Tmin) is found to be as high as 19°C and thereafter decreased to as low as 8°C not to exceed 16°C. In April, (Tmax–Tmin) is found to vary between 15° and 21°C and thereafter it is relatively low. Relative humidity was initially 80% in the beginning of the campaign and dropped to below 50% in the month of April and mid-May and thereafter showed an increased tendency.

Figure 2.

Daily average surface temperature range (Tmax–Tmin) and relative humidity observed at the experimental site during the campaign period.

4. Results and Discussion

4.1. Aerosol Optical Depth and Ångstrom Exponent

[8] Figure 3a shows time series of daily average aerosol optical depths observed at 0.5 μm wavelength during the ICARB campaign period of 1 March to 30 June 2006. The daily average aerosol optical depth shows consistent increase from March to June. Increase in AOD in April through June is due to dust events as reported by the India Meteorological Department, New Delhi, on the basis of visibility data (marked in Figure 3a). On these days, Total Ozone Mapping Spectrometer aerosol index (TOMS AI) was found to be greater than 3.0 (not all dust events are marked in Figure 3a). A similar trend is observed in the Moderate Resolution Imaging Spectroradiometer (MODIS) AOD and TOMS AI (Figure 3a). Monthly averaged AODs at 0.5 μm are 0.55 ± 0.19, 0.75 ± 0.27, 1.22 ± 0.26 and 1.18 ± 0.29, respectively, for March, April, May and June 2006 (Figure 3b). Comparison with AOD as observed at Kanpur (Aerosol Robotic Network (AERONET) site) in the eastern side of the Indo-Gangetic basin, about 400 km from New Delhi, is included in Figure 3b. In spite of the differences in the instrument used and data processing techniques, values were found to agree in March before the start of the dust events, but differ in April through June. Differences could be due to difference in geographical regions; New Delhi, which is on the western side of the IGP (closer to the Thar desert than Kanpur) is affected more by dust storms. MODIS-based monthly mean AODs over the site are 0.53 ± 0.19, 0.72 ± 0.25, 0.97 ± 0.22 and 0.97 ± 0.33 for March, April, May and June, respectively. Ground-based measurements of AOD compared well to MODIS-derived values in March and April, and there is a slight overestimate in May and June, months when dust events occurred. A plausible explanation for these differences: ground-based measurements are diurnally averaged, while the satellite overpass is at specific times and observations are spatially averaged. The corresponding monthly average TOMS AI over New Delhi is found to be 1.1 ± 0.6, 2.1 ± 0.9, 3.0 ± 0.9 and 2.0 ± 1.3 respectively. Higher TOMS AI values (>3.5) were observed on 9 days in May and 3 days in June during dust events [Prospero et al., 2002]. Dey et al. [2004] reported monthly average AODs of 0.82 and 0.84 during May 2001 and 2002, respectively, over Kanpur. Ångstrom exponent, α, is a good indicator of aerosol size distribution and can be derived from wavelength dependence of aerosol optical depth at five wavelengths by fitting a power law to the spectral AODs as given by

equation image
Figure 3.

(a) Time series of daily average AOD derived from Skyradiometer, MODIS, and TOMS aerosol index. (b) Monthly average spectral AOD variation over New Delhi (D) and AERONET Kanpur (K) observed during premonsoon 2006.

[9] For a higher Ångstrom exponent, the contribution from the fine-mode particles is increased. Over this site, Singh et al. [2005] reported an average AOD of 1.17 at 0.5 μm for several days during the premonsoon season of 2003, while values as high as 3.0 for the daily average were observed. They also reported an average value for the Ångstrom exponent during the premonsoon period to be about 0.328. Similar variation of high AOD and low ‘α’ values have been observed during a Saharan dust experiment [Tanré et al., 2003] and over the Indo-Gangetic basin [Dey et al., 2004]. In the premonsoon season, higher AODs have been observed in the months of April, May and June corresponding to the maximum decrease in α. Figure 4 depicts the time series of the daily and monthly average Ångstrom exponent for the campaign period. The Ångstrom exponent shows gradual decrease from 1.28 in the month of March to 0.47 in June, owing to the dominance of soil-derived coarse mode particles due to dust storms.

Figure 4.

(a) Time series of daily average Ångstrom exponent. (b) Monthly average Ångstrom exponent derived from Skyradiometer at New Delhi.

4.2. Volume Size Distribution

[10] Radiative impact of aerosols depends not only on aerosol concentration in space and time but also on their size and chemical composition. Compared to anthropogenic sulfate, desert dust is generally larger in size and more absorbing at solar and infrared wavelengths. This results in increased atmospheric heating along with decreased incident solar radiation at the ground and some greenhouse trapping of outgoing thermal radiation [Lubin et al., 2002]. The volume size distribution was retrieved from the direct solar and diffuse sky radiance measurements as discussed by Nakajima et al. [1996]. An improved version of the Skyrad.Pack radiative-transfer model is used (Version 4.2) in the present study. In the retrieval algorithm, it is assumed that the aerosols are composed of spherical and homogeneous particles. Scattering is simulated using Mie formulation, and multiple scattering effects are also taken into account. Figure 5 shows the ensemble of daily average aerosol volume size distribution retrieved from the sun/sky radiance data for March, April, May and June of 2006. It can be seen that the aerosol distribution is bimodal with fine-particle mode around 0.1–0.2 μm and coarse particle mode around 3–4 μm as represented in

equation image

where dV/dlnr is the volume size distribution, V0 is the column volume of particles per cross section of atmospheric column, r is the radius, rm is the modal radius, and σ is the standard deviation of the natural logarithm of the radii. In April, May and June the volume of coarse-particle fraction is larger than in March. In strong contrast to aerosols from biomass burning and urban aerosols (dominated by fine-mode accumulation particles) dust (dominated by coarse-mode particles) is composed of airborne desert soil material. Monthly averaged values of volume fraction and volume median radius for both fine and coarse modes are given in Table 1. Coarse volume fraction is found to increase by a factor of 2 to 3 during April–June when dust events are active. Monthly average size distribution parameters derived from the AERONET sky radiometer over Kanpur are also included in Table 1 and show increase in coarse fraction during April and May as compared to March. Using an in situ sampling analysis over New Delhi, Monkkonen et al. [2004] reported more fine particles in March than at other times, in agreement with the present study.

Figure 5.

Sun/sky radiance derived aerosol volume size distribution from March to June 2006.

Table 1. Monthly Average Values of Volume Size Distribution Parameters Over New Delhi and Kanpura
  • a

    VMR, volume median radius; SD, standard deviation of the lognormal size distribution; subscripts f and c, fine- and coarse-mode particles. First row corresponds to Delhi, and second row corresponds to AERONET Kanpur data (given in italics).

March 20060.1180.4594.3840.5740.0760.316
April 20060.1190.4774.120.5660.0690.645
May 20060.1320.4733.4360.6030.0860.903
June 20060.1390.5463.3830.6080.0940.915

4.3. Aerosol Radiative Properties

[11] Aerosol SSA, which is a fraction of scattering in the total extinction, is given by ωo = τscat/(τscat + τabs). It is unity for purely scattering aerosols (e.g., sulfate) and has low values for strongly absorbing aerosols (e.g., soot). Information on this parameter is needed for estimating aerosol radiative forcing. The sign at the top-of-the-atmosphere (TOA) forcing can change depending on the aerosol SSA [Takemura et al., 2002]. Ganguly et al. [2006] used scattering coefficients derived from Nephelometer and absorption coefficients from Aethalometer observations over New Delhi, to derive aerosol SSA at 0.5 μm during winter (December 2004). They were found to vary between 0.6 and 0.8 with an average value about 0.68 for December 2004. Figures 6a and 6b show the time series of daily average SSA at different wavelengths and the monthly average values for the campaign period. In the present study, SSA at 0.5 μm is found to vary between 0.71 and 0.95 with an average value of ∼0.79 during premonsoon. Monthly average SSA values are 0.84, 0.78, 0.75 and 0.74 for March, April, May and June 2006, respectively. AERONET-retrieved monthly average SSA values over Kanpur at wavelengths 0.47, 0.67, 0.87 and 1.02 μm are also included in Figure 6b. The lower SSA values observed over New Delhi are ascribed to higher air pollution and to being closer to the Thar Desert as compared to Kanpur. SSA is found to increase with wavelength mainly owing to abundance of coarse mode particles. Babu et al. [2002] used black carbon measurements combined with information from OPAC, to estimate SSA over a continental urban location at Bangalore and derived a SSA value of 0.73 at wavelength 0.5 μm. Ganguly et al. [2005] reported SSA from locations over the central Indian region to be in the range of 0.75–0.90. Singh et al. [2005] derived SSA from OPAC by describing the aerosols over New Delhi as a combination of two types: urban and desert [Hess et al., 1998] and determined it to be 0.672 at 50% relative humidity. Methods that use surface-based observations of chemical composition and BC in OPAC yield relatively lower values of SSA as compared to retrievals from surface-based remote sensing techniques that use sun/sky radiance which yield column-integrated values. Moorthy et al. [2007] reported that dust absorbing efficiency is higher (by nearly a factor of 2) over the “Great Indian Desert” and the adjoining regions, compared with west Asian/Arabian regions using the METEOSAT-derived Infrared Difference Dust Index [Tanré and Legrand, 1991; Legrand et al., 2001]. The dust absorption is found to be high during the April–May period. This is attributed to the presence of mixed aerosols from dust and black carbon over the Thar Desert region. Recent studies shows that significantly lower SSAs of aerosol mixtures could be due to possible mixing of dust and smoke aerosols [Jacobson, 2001]. Seinfeld et al. [2004] also demonstrate that aerosols in mixed state (adding black carbon and other aerosols to the mineral particles) can change dust aerosol radiative effects in many ways [Chandra et al., 2004]. Previous studies show that dust transported from East Asia to the Pacific does not absorb as much light as aerosols from South Asia or from the Sahara Desert [Seinfeld et al., 2004]. The Delhi region is known as one of the world's most polluted urbanized areas. Vehicular emissions, thermal power plants and other industrial units are the major polluting sources. Biomass burning is another important source, as wood is the most widely used fuel in this region in the dry season. Dust particles internally mixed with soot, sulfates, nitrates, or aqueous solutions can have drastically different properties from those that are evident at the dust source. The ability of dust particles to scatter and absorb solar and terrestrial radiation can be altered in different ways depending on the species that aggregate with dust particles.

Figure 6.

(a) Time series of daily average single scattering albedo. (b) Monthly average single scattering albedo at different wavelengths retrieved over New Delhi (D) and AERONET Kanpur (K).

[12] Certain optical properties of aerosols can be represented by the complex refractive index M = nreal − inimg. The imaginary part is associated with absorption and heating while the real part is associated with scattering and cooling. Retrieval of complex refractive index of aerosols is done with Version 4.2 of the Skyrad.pack inversion method [Nakajima et al., 1996] similar to that of Dubovik et al. [2000]. It allows global fitting of spectral and multiangle sun/sky radiances with a simultaneous search for the size distribution and complex refractive index. The monthly average values of nreal at 0.5 μm wavelength retrieved over New Delhi are 1.51, 1.53, 1.54 and 1.54 and nimag at 0.5 μm is −0.016, −0.019, −0.022 and −0.024 for March, April, May and June 2006, respectively. A gradual increase from March to June is evident. Singh et al. [2004] found a similar type of spectral variation during dust events over Kanpur in May 2002. Increase in the imaginary part of the refractive index indicates increasing absorption which is also reflected in SSA values from March to June.

[13] The asymmetry parameter (g) is a simple, single-valued representation of the angular scattering and is a key property controlling the aerosol contribution to forcing. It depends on the size and composition of the particles and is defined as the intensity-weighted average cosine of the scattering angle,

equation image

where θ is the angle between the transmitted and the scattered radiation and P(θ) is the phase function (angular distribution of scattered light). The value of g ranges between −1 for entirely backscattered light to +1 for entirely forward scattered light. Andrews et al. [2006] presented and compared different methods for deriving the aerosol asymmetry parameter from the phase function. The time series of daily average g is shown in Figure 7a. Figure 7b shows spectral variation of g for different months from March to June. There is a consistent increase from March with reduced spectral dependence following a similar variation as AOD. Values of g decrease with wavelength in the visible spectral region and slightly increase in the near-infrared region. The variation of SSA and g with Ångstrom exponent is examined and shown in Figure 8. Abundance of dust aerosols in the premonsoon season lowers the SSA and increases the absorbed radiation in the atmosphere (Figure 8b). The asymmetry parameter shows inverse relation with the Ångstrom exponent (Figure 8a). Five-day back-trajectories for each month are calculated using the HYSPLIT model shown in Figure 9, suggesting that air masses pass through desert regions in the months of April and May. With the advance of the summer, the west to east pressure gradient increases and the winds from west to southwest also strengthen reaching maximum of about 20 to 25 km/h during May and June. Surface winds raise a tremendous amount of dust from the loose sandy soil of the region, and the dust is transported eastward over the neighboring states of India. The probable anthropogenic sources in the Thar region are clay mines, gypsum quarries and copper mines [Yadav and Rajamani, 2006].

Figure 7.

(a) Time series of daily average Asymmetry parameter. (b) Monthly average asymmetry parameter at different wavelengths retrieved over New Delhi (D) and Kanpur (K).

Figure 8.

(a) Scatterplot between daily average asymmetry parameter and Ångstrom exponent. (b) Ångstrom exponent versus single scattering albedo.

Figure 9.

Five-day back-trajectories derived from HYSPLIT model for the experimental site for March–June 2006.

4.4. Changes in Aerosol Optical Characteristics due to Dust Events

[14] To compare the aerosol optical properties during non-dust and dust-event days, a set of relevant cases were selected and analyzed. Figure 10 shows the daytime-average aerosol optical characteristics such as (Figure 10a) AOD and Ångstrom exponent, (Figure 10b) volume size distribution and (Figure 10c) spectral variation of SSA for both dust and nondust cases. Higher AOD (as high as 1.5 at 0.5 μm) and lower α (as low as 0.3) were observed during dust event days. Volume size distribution retrievals show significant increase in coarse volume fraction (∼1 μg3/μg2) during dust event days as compared to nondust event days (∼0.2 μg3/μg2). Mode radii of coarse particles decreased from 4.3 to 3.4 μm during dusty days. A scatterplot between volume median radius (VMR) and aerosol optical depth for both fine- and coarse-mode particles for Delhi and Kanpur is included in the auxiliary material. VMR of fine-mode particles increases with increase in aerosol loading whereas VMR for coarse-mode particles decreases over Delhi. Over Kanpur VMRc decreases with increase in AOD but VMRf increases with increase in AOD. Dubovik et al. [2002] reported different VMR for dust aerosols from different geographic origins, and there is no dynamical response of volume median radius with increase in aerosol loading. Xiangao et al. [2004] reported increase in VMR for coarse particles with increase in AOD during dust outbreak days, whereas Pinker et al. [2001] reported decrease in VMRc with increase in AOD over sub-Sahel, West Africa, during dust outbreak conditions. Average SSA value at 0.5 μm is found to be 0.84 for nondust days and lowered to 0.75 during the dust event days.

Figure 10.

Comparison between aerosol optical characteristics on nondust and dust event days.

4.5. Error Analysis

[15] Version 4.2 of the Skyrad.Pack inversion algorithm is used to retrieve aerosol size distribution, complex refractive index, and SSA from spectral measurements of direct and diffuse radiation. The perturbation of the inversion resulting from random errors, instrumental offsets, and known uncertainties in the atmospheric radiation model are analyzed. Sun or sky channel calibration errors, inaccurate azimuth angle pointing during sky radiance measurements and inaccuracy in accounting for surface reflectance measurements are considered as error sources. An error analysis similar to one described by Dubovik et al. [2000] is performed by considering the following errors: (1) 5% error in absolute calibration constant V0(λ), (2) 5% change in solid view angle for each spectral channel, (3) 5% error in sky radiance, (4) 0.5° error in azimuth angle pointing, and (5) 50% change in surface reflectance. The inversion is performed for an almucantar data set corresponding to 54° solar zenith angle. An error of 5% in the absolute calibration constant V0(λ) can induce an error of 0.03 in retrieved AOD. The retrieved SSA for the above uncertainties is given in Figure 11a. Mean and maximum differences in retrieved SSA at all wavelengths are 0.004 and 0.02, except at 0.5 μm wherein the uncertainty ranges from 0.02 to 0.08. Retrieved size distributions with above uncertainties can also be seen in Figure 11b.

Figure 11.

(a) Single scattering albedo and (b) volume size distribution retrievals for biases in calibration constants, sky radiance, angular pointing, and surface reflectance.

4.6. Aerosol Radiative Forcing and Heating Rates

[16] Sun/sky radiometer-derived monthly average values of AOD, SSA and ASP are used with the Santa Barbara Discrete-ordinate Atmospheric Radiative Transfer code (SBDART) developed at University of California, Santa Barbara [Ricchiazzi et al., 1998] to estimate aerosol radiative forcing over New Delhi in the shortwave (SW) (0.3–3.0 μm) and long-wave (LW) (4.0–50.0 μm) spectral regions for March, April, May and June 2006. SBDART is a software tool that computes plane-parallel radiative transfer in clear and cloudy conditions. All important processes that affect the ultraviolet, visible, and infrared radiation fields are included. Sun/sky radiometer-retrieved AOD, SSA and ASP at 0.4, 0.5, 0.675, 0.87 and 1.02 μm wavelengths along with Ångstrom exponent derived from wavelength dependence of AOD are used as inputs to the model. AODs at other wavelengths are extrapolated using Ångstrom exponent; SSAs and ASPs at other wavelengths are scaled from by combining observations at the above wavelengths and continental polluted model of optical properties of aerosol and clouds (OPAC) [Hess et al., 1998]. Vertical profile of the temperature is constrained using radiosonde data available at New Delhi. MODIS-derived monthly average precipitable water content [King et al., 2003] and TOMS-derived total column ozone [Bhartia et al., 1993] are also used. Monthly average values of precipitable water content derived from MODIS over the site are 1.7, 2.0, 2.9 and 4.2 cm for the months March, April, May, and June 2006, respectively. TOMS-derived column ozone values are 282, 294, 289, and 292 Dobson units for the months March, April, May, and June 2006, respectively, are used as inputs in the radiative transfer model. Surface albedo over New Delhi is derived from MODIS bidirectional reflectance distribution function/Albedo products [Moody et al., 2005]. MODIS provides both black-sky albedo (BSA) (direct reflectance) and white-sky albedo (WSA) (bihemispherical reflectance) at seven spectral bands (0.47, 0.555, 0.659, 0.858, 1.240, 1.640 and 2.100 μm) as well as for three broad bands (0.30–0.87 μm, 0.70–5.0 μm and 0.30–5.0 μm). Actual albedo was calculated from BSA and WSA for March, April, May and June at different solar zenith angles, using the following relation:

equation image

where diffuse sky fraction (DSF) is a function aerosol loading and solar zenith angle. Observed average AOD is used to derive DSF using a look-up table provided by MODIS land team [Schaaf et al., 2002]. SW and LW fluxes were simulated with SBDART using sun/sky radiometer-derived monthly average aerosol optical depths, single scattering albedo, asymmetry parameter, MODIS-derived column precipitable water, spectral surface albedo, and TOMS-derived total column ozone. Diurnal average aerosol radiative forcing is estimated by computing the difference between net (down-up) radiative fluxes ‘with aerosols’ and aerosol free' conditions. Detailed methodology is explained in earlier publications [Pandithurai et al., 2004; Panicker et al., 2008].

[17] The estimated shortwave aerosol radiative forcing at the surface is −39, −64, −106 and −99 W m−2 for March, April, May and June 2006, respectively. Forcing values at the TOA are −12, −4, +5 and +24 W m−2, and hence the atmospheric forcing values are +27, +60, +111 and +123 W m−2 for March, April, May and June, respectively (Figure 12). Higher surface aerosol forcing in May is due to the higher aerosol loading (the AOD is 1.22 at 0.5 μm). Sign of TOA forcing changed to positive in May and June mainly owing to lower SSA and higher surface albedo. Increase in atmospheric absorption is mainly due to consistent decrease in SSA from March through June. Forcing efficiency at the surface is −71, −85, −87 and −84 W m−2 for March, April, May and June 2006, respectively. Higher surface forcing efficiency observed during April through June is mainly due to lower SSA and higher surface albedo values. In the LW, aerosol forcing values at the surface are +8.6, +15.7, +22.4 and +15.3 W m−2 for March, April, May and June, respectively. At the TOA, LW aerosol forcing is found to be +8.8, +12.3, +19.6 and +18.6 W m−2, and hence in the atmosphere they are found to be 0.2, −3.4, −2.8 and +3.3 W m−2 for March, April, May and June, respectively. Hence the net (SW+LW) aerosol forcing values at the surface are −30, −48, −84 and −84 W m−2 for March, April, May and June, respectively. At the TOA, the net forcing values are −3.2, +8, +25 and +43 W m−2, for March, April, May and June, respectively.

Figure 12.

Aerosol radiative forcing in shortwave (SW), longwave (LW), and net (SW + LW) at the surface and at the TOA over New Delhi during March–June 2006.

[18] The sensitivity analysis shows that an error of 0.02 in AOD and in SSA can result in an uncertainty of 1.5 and 3.0 W m−2 in Fsfc, respectively. The corresponding error at Ftoa can be 0.5 and 1.6 W m−2, respectively. Uncertainties in ARF may also occur owing to errors in surface albedo values used. Forcing estimates were made by using three sets of surface albedo values: (1) MODIS albedo, (2) vegetation albedo [Reeves et al., 1975], and (3) sand albedo [Staetter and Schroeder, 1978]. MODIS albedo values increase from March to June, whereas the model albedos of vegetation and sand are constant for all months. Difference between ARF at the surface using MODIS and vegetation albedos shows a difference of about 6–15%. The higher atmospheric heating in the months of May and June is mainly due to the observed lower SSA.

[19] The higher atmospheric absorption in the months of May and June leads to higher heating rates. The rate of change of temperature (dT/dt) in a layer due to radiative heating/cooling is called radiative heating/cooling rate, defined as

equation image

where Cp is the specific heat capacity and ρ is the density of air and dF/dZ is the radiative flux divergence. Differences in heating/cooling rates with and without aerosols were computed. Heating rates in the lower atmosphere (up to 5 km) are 0.6, 1.3, 2.1 and 2.5 K/d, respectively, for March, April, May and June 2006. Higher aerosol heating rates due to enhanced atmospheric absorption in dry seasons can have an impact on regional climate and monsoon circulation [Ramanathan et al., 2007; Pilewskie, 2007]. Detailed studies on the impacts of enhanced aerosol heating during premonsoon and their effect on regional climate are needed.

5. Conclusions

[20] 1. Over New Delhi during premonsoon, aerosol optical depth at 0.5 μm increases from 0.55 (March) to 1.2 (June). Ångstrom exponent shows gradual decrease from 1.28 (March) to 0.47 (June), indicating an increased abundance of coarse particles due to dust storms which raise dust and transport it from the Thar desert and adjoining regions.

[21] 2. The inferred SSA at 0.5 μm shows a decreasing trend from 0.84 in March to 0.74 in June, indicating stronger contribution from absorbing aerosols which are a mixture of both desert dust and anthropogenic (vehicular and industrial) aerosols.

[22] 3. Derived aerosol radiative forcing shows consistent increase in cooling at the surface from −39 W m−2 in March to −99 W m−2 in June. Heating in the atmosphere ranges from 27 W m−2 (March) to 123 W m−2 (June) since dust events lead to enhanced heating rates in the lower atmosphere.


[23] The authors would like to thank B. N. Goswami, IITM Director, for encouraging this work supported under ISRO-GBP Aerosol Radiation Budget Studies. Thanks are owed to T. Nakajima for providing the Skyrad.Pack version 4.2 code to retrieve aerosol optical characteristics from Sun/sky radiance data. Data used in this study are collected as part of the SKYNET network in collaboration with CEReS, Chiba University, Japan. Thanks are owed to R. P. Singh and B. N. Holben for AERONET Kanpur data used in this study. MODIS Terra data used in this study were acquired with the GES-DISC Interactive online visualization and analysis infrastructure. Thanks are owed to Calli Jenkerson, USGS, and Crystal Schaff, Boston University, for their help in retrieving MODIS surface albedo. The back-trajectories are computed from the FNL data archived at the NOAA Air Resources Laboratory using the HYSPLIT models. Thanks are also owed to colleagues at the lidar and radiation laboratories of IITM. The PREDE sun/sky radiometer was acquired under grant A2GCF001 from the National Space Development Agency of Japan (NASDA) (subsequently, Japan Aerospace Exploration Agency (JAXA)) to the University of Maryland in support of ADEOS II activities and partly supported by Global Earth Observation System of Systems (GEOSS), Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan. We are grateful to three anonymous reviewers for their constructive comments/suggestions, which helped to improve the manuscript.