The surface fluxes in the wavelength range 280–2800 nm were measured during the pre-monsoon period, April–June 2003 along with the spectral distribution of aerosol optical depth (AOD) in the visible and near infrared wavelengths. The Ångström exponent alpha retrieved from the data showed abundance of desert aerosols over Delhi during this period. The aerosol composition constructed using the OPAC model indicated a typical mixture of two aerosol types: urban and desert. Due to this the aerosol mixture had a very low value of single scattering albedo ∼0.67. The average total radiative forcing efficiency observed at the surface in the broad wavelength band (280–2800 nm) was estimated and compared with the SBDART model calculated values.
 Delhi, one of the worst polluted mega cities of Asia, is also faced with a typical problem of desert aerosols every year during the pre-monsoon period i.e., April–June. These aerosols are brought by wind blown dusts from the Thar Desert in Rajasthan. The total suspended particulate matter (SPM) concentration during this period goes enormously high. Consequently, the visibility is reduced and the local radiative forcing is significantly affected. Although, some studies on air pollution [Goyal and Sidhartha, 2003; Nagdeve, 2004; Srivastava et al., 2005], green house gases [Padhy and Varshney, 2000] and their emission estimates and trends [Gurjar et al., 2004] over Delhi have been done in past, the desert aerosols and their effects on radiation flux have not been considered. However, it is already known that the SPM in Delhi's environment is not only contributed by vehicular and industrial activities but also significantly because of soil originated particles and re-suspended dust generated by strong winds and construction activities [Mönkkönen et al., 2004].
 The aim of this letter is to report the impact of wind blown desert dust aerosols on Ångström exponent and to quantify the average aerosol forcing due to this over Delhi. An attempt is also made to construct the average aerosol composition during the pre-monsoon sandstorm period using the OPAC (Optical Properties of Aerosols and Clouds) model of Hess et al. . We also present the results of column aerosol optical depths and flux measurements in the total broad radiation band (280–2800 nm) over Delhi during the same period. Finally, using these observations, the average direct radiative forcing was estimated and the results were compared with the SBDART radiation transfer model.
 Simultaneous ground based measurements of column aerosol optical depths (AOD), total solar radiation flux (direct plus diffuse) in the spectral range 280–2800 nm were done at National Physical Laboratory, New Delhi during April–June 2003 on the clear sky days. AOD measurements were taken using the Solar Light portable spectrometer MICROTOPS II instruments at 340, 500, 675, 870, and 1020 nm. The instrument was calibrated at Solar Light Company and the measurements were taken within a year of the procurement. More details of the instrument are available elsewhere [Morys et al., 2001]. The total global flux measurements were done using the Kipp and Zonen CM-21 pyranometer. This instrument has an absolute accuracy of about ±9 W m−2 but the error due to directional response (because detector responds differently to radiation depending on the incident angle) can be as much as ±10 W m−2. The monthly average SPM measurements can be obtained from Central Pollution Control Board (CPCB), Delhi from the website http://www.cpcb.nic.in.
3. Results and Discussion
 The CPCB observations for the pre-monsoon dust storm months April–June 2003 showed that the average monthly concentration of SPM over Delhi was enormously high. This enhancement was mainly due to the wind blown dusts brought to Delhi due to several dust storms from western desert region of Rajasthan and surrounding areas. The meteorological conditions over Delhi during these months showed that the winds were predominantly westerly or northwesterly and dry with relative humidity always below 50% and the average wind speed near surface less than 5 m s−1.
3.1. Optical Depth and Ångström Exponent
 Due to high SPM concentration the aerosol optical depth during April–June period was very high. The average AOD at 500 nm (τ500) was 1.17 ± 0.65. On several days in June the SPM concentrations were so high that the daily average τ500 reached 3.08 ± 0.55. The spectral variation of daily averaged AODs with wavelength for some of the representative days during this period is shown in Figure 1. The vertical bars at respective mean values indicate the standard deviation values. Although the AOD values at all the wavelengths were generally high (>0.50), the values on 6, 11, and 18 June 2003 was extremely high with large standard deviations, indicating high aerosol loading during these days. At Kanpur, a station about 350 km down south-southeast of Delhi, the CIMEL sky radiometer observations have also shown high aerosol values during pre-monsoon months [Dey et al., 2004; Singh et al., 2004].
 It is also noticed from the daily spectral variation of AOD that on several days the AOD variation does not follow logarithmic decrease with the increasing wavelength for the entire range of 340–1020 nm. Three kinds of spectral variations are noticed. (1) The AOD values decrease with increasing wavelengths: On normal days when the SPM level is low, atmosphere is devoid of any sand storm events and visibility is greater than 5km, mostly in April. In Figure 1 such plots are for April 16, 25, and May 7 2003. (2) The AOD values increase with increasing wavelength throughout the wavelength range: Those days when dust storms occur, SPM levels are very high and the visibility is less than 2.5 km, mainly during June (6, 11 and 18 June in Figure 1). (3) The AOD value initially decreases with wavelength till 675 nm and then starts increasing: On days following the dust-storm day when coarse particles are settling gradually but SPM levels are still high, visibility is in the range of 2.5–5 km (May 1, 30 and June 23 in Figure 1). This kind of variation suggests that the coarse sand and dust particles are added in the already polluted (vehicular and industrial origin) atmosphere of Delhi.
 As the AOD characterizes the aerosol loading the Ångström wavelength exponent α indicates the aerosol size distribution. It follows power law of Ångström given by τ(λ) = βλ−α [Ångström, 1964] and is estimated by a least square fit on a log-log plot scale of the experimental AOD versus wavelength measurements. In the present case α are estimated using AOD measurements at 340, 500, 675, 870 and 1020 nm. The variations of daily averaged AOD at 500 nm (τ500) and the corresponding α values during April–June are shown in Figure 2. The vertical bars indicate the standard deviation (±σ) values for AOD and the errors in the estimation of α. The figure clearly indicates that the large AOD values are associated with the small close to zero α values, a typical indicator of the desert aerosols [Cachorro et al., 2001]. Sometimes, when the sand storm over Delhi was very strong the high AOD values were coupled with the negative values of α (∼−0.06 on June 6, 11 and 18, 2003), strongly indicating the dominance of coarse particles. The average value of α during the entire pre-monsoon period April–June is about 0.328 ± 0.288. Similar behavior of high AOD associated with minimum α values have also been observed during Saharan dust experiment [Tanre et al., 2003] and during dust storm period in the Indo-Gangetic basin [Dey et al., 2004; Singh et al., 2004]. From the Persian Gulf measurements Smirnov et al.  have also reported a low Angstrom parameter ∼0.7 when dust aerosol dominated the atmosphere compared with the dust free cases when it was ∼1.2.
3.2. Aerosol Component Structure
 On the basis of Ångström exponent and the aerosol optical depth over Delhi during pre-monsoon dust period, a possible aerosol component structure has been determined using the OPAC model. The model, developed by Hess et al.  provided microphysical and optical properties of aerosols and clouds in the solar and terrestrial spectral range of atmospheric particulate matter. It assumes the aerosols in the atmosphere to be a mixture of different components like insoluble soil particles, water-soluble aerosols, sea salt, minerals, soot, etc. For the current study the software package OPAC3.1a was downloaded from the OPAC ftp-site. Hess et al.  provide a detailed description of the model.
 Our observations when compared with the model outputs showed that the Delhi aerosols during this period can be described only by a combination of two aerosol types: “urban” and “desert.” According to model the composition of aerosol that best represent the average aerosols characteristics over Delhi during this period are shown in Table 1. The densities of components are taken from Hess et al.  at 50% relative humidity. Delhi being driest during April–June (relative humidity <50%), the effect of humidity on aerosol size distribution has been neglected in this study. The concentrations of various components (Table 1) are adjusted so that the spectral optical depths, Ångström exponents and the visibility are in consistent with the observations described in previous section. A high value of soot particles (130,000 cm−3) indicates typical urban aerosol type and corresponds to 7.8 μg m−3. The only measured value of soot (BC) over Delhi is during December 2004 provided by A. Jayaraman (unpublished data, personal communication, 2005), indicates a high value in the range 10–50 μg m−3. This is probably because of comparatively low boundary layer height during December. High concentration of coarse mode particles over Delhi indicated by the model (Table 1) also seems to be a reality when compared with the Kanpur observations [Dey et al., 2004; Singh et al., 2004] which shows more than three fold increase in coarse mode particles during dust storm period. During strong sand storm days (6, 11,18 June, etc.) when the Ångström coefficient α is negative it is found that there is substantial amount of water insoluble soil dust particles in the atmosphere. The OPAC model in that case has to be used with slightly different composition with more water insoluble particles to match the observations.
Table 1. Average Aerosol Composition Over Delhi During Pre-Monsoon at 50% Relative Humidity
Density, g cm−3
Volume Mixing Ratio
Mass Mixing Ratio
1.272 × 10−02
4.982 × 10−03
2.300 × 10−02
1.280 × 10−02
Mineral (nuc. mode)
4.713 × 10−03
4.800 × 10−03
Mineral (acc. mode)
1.758 × 10−01
1.791 × 10−01
Mineral (coa. mode)
7.837 × 10−01
7.983 × 10−01
3.3. Aerosol Radiative Forcing
 The observed average aerosol radiative forcing was found by correlating the total instantaneous flux reaching the surface (measured by pyranometer) with the normalized τ500. The normalized τ500 is the AOD at 500 nm corrected for the air mass factor (1/μ, μ = cos θ, where, θ is the solar zenith angle (SZA)) to remove the influence of change in the slant air column length due to SZA variation [Jayaraman et al., 1998]. Figure 3 shows such a correlation and straight line fit between total flux and the normalized τ500 for the 280–2800 nm band for SZA within 30–60°. The slope of the straight-line fit gives the change in flux per unit increase in the AOD at 500 nm for the given range of SZA and τ500. The average radiative forcing efficiency given as the change in flux ΔF in Wm−2 for a 0.1 increase in the τ500 is observed to be −13.6 ± 1.4 for the 280–2800 nm band. It is to be noted that due to the unshaded pyranometer used here the uncertainty in total flux measurements due to SZA variation could not be ascertained. Further, a moderate value of correlation (R2 = .4627) in the plot indicates that the linear dependency between AOD and surface flux is affected due to multiple scattering when the AOD exceeds unity.
 The theoretical clear sky total (direct + diffuse) aerosol radiative forcing at the surface was found using the Discrete Ordinate Radiative Transfer model SBDART (Santa Barbara DISORT Atmospheric Radiative Transfer model). The model was developed at the University of California, Santa Barbara by Ricchiazzi et al. . The input for the model are the observed spectral variation of AOD along with single scattering albedo (SSA) and the asymmetry parameter at 500 nm. The SSA and the asymmetry parameter were obtained by running the OPAC model with the average aerosol composition described in Table 1. The values obtained are 0.672 and 0.790 respectively for 50% relative humidity. Although the value of SSA is small compared with the other observations [Dubovik et al., 2002] but it is found to increase with the wavelength which can be caused by the dominance of coarse particle mode (due to dust) in the aerosol distribution. A similar increase in the SSA with wavelength is also reported by Smirnov et al.  and attributed to the domination of the coarse particle mode. An SSA value of 0.73 was obtained by Babu et al.  for the continental urban location Bangalore which is more or less free from the desert aerosols. The SBDART model when used with the wavelength dependent SSA and asymmetry parameter in the present case shows a reduction in total downward flux at the surface by ∼8–13% for the zenith angle 30–60°.
 In order to compare the observations with the SBDART model results the total flux reaching the surface was calculated from the model in the band 280–2800 nm for the SZA 30–60° at an interval of every 5° and τ500 in the range of 0.6–2.0 at an increment of 0.2. To suit the surface albedo over Delhi during dust storm period the model is run with a composite albedo, which is a combination of sand and vegetation given in the model as standard ground reflectance. The ground level fluxes thus obtained from this model, as a function of AOD was then straight line fit with the normalized τ500 exactly in the same manner as the observed fluxes were fit. This method is more accurate for comparison with the current observation than obtaining the 24h average flux from the model [Xu et al., 2003; Babu et al., 2002] and then comparing with the observed forcing. The model calculated fluxes are shown in Figure 4 along with the straight line fit. The average radiative forcing efficiency from the model has a value of −16.1 ± 1.2 in the 280–2800 nm band. It indicates that the model downward forcing efficiency at the surface is about 18% higher than the observed one. This is partly because of the fact that the SBDART model over estimates the diffuse flux by almost 30% [Ricchiazzi et al., 1998]. Also, at higher AOD values, as in the present case, there is substantial contribution of multiple scattering and absorption of radiation due to dust aerosols that the pyranometer measurements do not take into account properly.
 The study of aerosols over Delhi during the pre-monsoon period (April–June) 2003 from observations and model calculations reveal the following:
 1. Due to high SPM concentration due to desert aerosols during this period the average Ångström α parameter has a very small value ∼0.328 which at times become negative (∼−0.06) when dust content is severely high. The average AOD at 500 nm (τ500) is high at 1.17 ± 0.65.
 2. The OPAC model suggests that the Delhi aerosol is best represented by a combination of urban and desert aerosol types which is the result of the coarse sand and dust particles being gradually added in the already polluted (vehicular and industrial origin) atmosphere of Delhi.
 3. Averaged over the range of measured SZA (∼300–600) and τ500 (1.17 ± 0.65) the total broadband (280–2800 nm) flux reaching the surface decreased by 13.6 ± 1.4 Wm−2 per 0.1 increase in the AOD during the pre-monsoon period over Delhi. The SBDART model estimate in the same range is however slightly higher at −16.1 ± 1.2.