1 Introduction

Particulate matter (PM2.5, PM10, PM10–2.5) and gaseous air pollutants in urban environments play a vital role in air quality, visibility degradation, light extinction processes, etc. (Foyo-Moreno et al. 2014; Pateraki et al. 2014; Valenzuela et al. 2015; Khalil and Shaffie 2016). Especially over densely populated areas in the South and East Asia, visibility has significantly declined due to a dramatic increase in atmospheric particulate pollution (Wang and Shi 2010; Tiwari et al. 2014; Wang et al. 2015). Atmospheric light extinction is defined as the fractional loss of intensity in a solar light beam per unit distance due to extinction (scattering and absorption) by aerosol particles and gasses (Pitchford et al. 2007). The extinction aerosol coefficient (bext) is highly related to particle size and chemical composition (Badarinath et al. 2008; Huang et al. 2014) and can be measured by in-situ (e.g. Nephelometer, Aethalometer) and remote sensing (e.g. LIDAR) techniques or estimated based on aerosol mass and extinction efficiencies of various chemical constituents by assuming them as externally mixed particles (Cao et al. 2012; Wang et al. 2012; Zhang et al. 2013a; Tao et al. 2014; Shen et al. 2014).

Carbonaceous particles (CP) constitute a large fraction of the atmospheric aerosol with serious concerns about climate change, air pollution, visibility impairment and human health (e.g., Singh et al. 2014; Reddington et al. 2015). CP are composed of organic carbon (OC) and elemental carbon (EC), also known as equivalent black carbon (eBC). EC is a primary pollutant emitted from incomplete combustions of carbon-contained materials (Viana et al. 2007; Paraskevopoulou et al. 2014). It constitutes an important driver for global warming due to its strong absorbing efficiency (Novakov et al. 2005; Bond et al. 2013). In contrast, OC can be either released directly into the atmosphere (primary OC: POC) or produced from gas-to-particle reactions (secondary OC: SOC) (Bougiatioti et al. 2013). It is mostly scattering in nature with an important role in the formation of urban haze (Gautam et al. 2007). However, it can contain significant amounts of light absorbing species termed Brown Carbon (Andreae and Gelencsér 2006). OC represents a mixture of many organic compounds (including polycyclic aromatic hydrocarbons, PAHs), some of which are mutagenic and/or carcinogenic (Smith et al. 2007; Li et al. 2008). A significant fraction of OC is water soluble (known as WSOC), with potential influence on chemical reactions and aerosol-cloud interactions (Decesari et al. 2000; Ram and Sarin 2011).

Due to their serious impact on atmospheric chemistry, monsoon circulation and climate, CP and ionic species have been of particular interest over the Indo-Gangetic Plains (IGP) during the last decade (Rengarajan et al. 2007; Behera and Sharma 2010a, 2010b; Ram et al. 2010a, b, Ram et al. 2012; Ram and Sarin 2011; Tiwari et al. 2013a, 2013b; Rastogi et al. 2014; Sharma et al. 2014; Srinivas and Sarin 2014; Bisht et al. 2015; Panda et al. 2016). These studies measured the chemical characterization of the PM2.5, diagnostic ratios for identification of the aerosol source apportionment, and primary and secondary aerosol formation mechanisms. Other works have dealt with aerosol emission inventories (Streets et al. 2004; Lu et al. 2011) from different sources helping to understand the role of biomass burning. Venkataraman et al. (2005) examined the relative contributions of fossil-fuel combustion, biomass and biofuel burning to BC mass, which were estimated to ~25 %, 33 % and 42 %, respectively, whereas their contributions to OC were found to be ~13 %, 43 %, and 44 %, respectively. As a consequence, different kinds of burning, along with atmospheric mixing and long-range transport, are responsible for changing atmospheric composition over India, affecting aerosol-cloud interactions, monsoon circulation, and climate (Ganguly et al. 2012; Nair et al. 2012).

Many studies summarized by Bisht et al. (2015), analyzed the CP, PM2.5, and PM10 concentrations and examined their influence on air quality over Delhi. Bisht et al. (2015) using PM2.5 chemical analyses for CP, SO4 2− and NO3 − examined the seasonal evolution of their daytime and night-time concentrations, the contributions to PM2.5 mass and their association with the meteorological conditions and long-range transport over Delhi during January to December 2012. The current work aims to further contribute to the findings by Bisht et al. (2015) by estimating the bext for Delhi based on chemical analysis of PM2.5 samples by means of the original and revised IMPROVE algorithms (Pitchford et al. 2007; IMPROVE 2011). PM2.5 samples were collected from December 2011 to November 2012 covering periods with different atmospheric circulation patterns, emission rates, and aerosol characteristics. The main objectives of the study are: (i) to analyse the temporal variability of PM2.5 and CP in comparison with Bisht et al. (2015), (ii), to assess the primary and secondary aerosol emissions, (iii) to investigate the temporal evolution of bext and the contributions of individual components of PM2.5 on it, (iv) to examine the correspondence of bext to visibility degradation over Delhi and, (v) to compare the results obtained from the original and revised IMPROVE algorithms.

2 Experimental details and methodology

2.1 Site description

PM2.5 samples were collected on the rooftop of a building (15 m above ground level) on the premises of the Indian Institute of Tropical Meteorology (IITM), New Delhi during December 2011 to November 2012. The sampling site is surrounded by heavy roadside traffic, small, medium and large-scale industries and agricultural fields towards the south, which along with long-range transported and road re-suspended dust are the major sources of aerosols and pollutants over Delhi (Srivastava et al. 2012; Tiwari et al. 2013a, 2013b). Air masses from northwest directions dominate during post-monsoon (October – November) and winter (December–February) shifting to the west during pre-monsoon (March–June) and then to southeast during the rainy summer monsoon (July–September), thus controlling the PM concentrations and characteristics across Delhi (Lodhi et al. 2013; Bisht et al. 2015).

2.2 Instrumentation and chemical analysis of PM2.5

For the current study, 75 PM2.5 samples (~5–7 samples per month with duration of 10–12 h per day during the daytime) were collected from December 2011 to November 2012. Although the study period is similar to that (January – December 2012) used in Bisht et al. (2015), the sampling days and duration are different allowing us for a direct comparison of the PM2.5 and CP concentrations between the two studies. PM2.5 sampling was conducted over 10–12 h per day and only during the daytime. Thus, there can be concerns about their representativeness and this approach could also mean that some aerosol/pollution events may not have been caught in the presented time series. The PM2.5 samples were collected on quartz fiber filters using APM 550 samplers (Envirotech Pvt. Ltd., India; http://www.envirotechindia.com/). The filters were subjected to 24-h desiccation before and after the sampling for the removal of moisture. Then, they were weighted using an electronic microbalance (Model GR202, A&D Company Ltd. Japan) with high resolution (0.001 mg) and accuracy. The particle concentrations were determined gravimetrically by the difference in their weights before and after the sampling.

The exposed filters were analyzed for CP (OC and EC) mass concentrations by a thermal/optical OC/EC analyzer (Sunset Laboratory Inc. Model 4 L; http://www.sunlab.com) using the NIOSH 5040 (National Institute for Occupational Safety and Health) thermo-optical transmittance protocol (Birch and Cary 1986). A punch from the filter (2.1 cm2) is stepwise heated in an inert (100 % He) atmosphere, followed by heating in an oxidizing medium (~10 % O2 and 90 % He). The carbon evolved during the processes is oxidized to CO2 and converted to methane (CH4) that is measured via a suitable detector. The detector transmits a 678 nm laser beam through the sampled filter. The return of the transmitted signal to its initial value on the thermograph is taken as a split line between OC and EC. The amount of carbon fraction evolved before the split line is defined as OC and after the split line as EC. A similar process has been followed using a blank filter to determine the detection limit of OC and EC in order to correct for the contribution of OC from the quartz filters. More details about the analysis of OC and EC mass concentrations are presented in previous papers (Tiwari et al. 2012, 2014; Bisht et al. 2015).

The concentrations of several ionic species (SO4 2−, NO3 −, F− and NH4 +) are necessary for the estimations of bext via the IMPROVE algorithm. They were analyzed using an Ion Chromatograph (DIONEX-2000, RFIC, USA). The anions (SO4 2−, NO3 − and F−) were analyzed by the Ion Chromatograph (model IC-2000) using an IonPac-AS15 analytical column with an ASRS ultra II 2 mm micro-membrane suppressor. The eluent was 38 mM potassium hydroxide/1.7 mM sodium bicarbonate. Triple deionized water was used as the regenerator. The system has a detection limit of ∼0.02 ppm. Field blanks were also analyzed using similar procedures as filter samples and found to be below the detection limits. The glassware used in the extraction and analysis were soaked in diluted nitric acid overnight and washed thoroughly with tripled distilled and deionized water to remove any adhered impurities. Due to technical problems with the instrument, cations were not analyzed in the present study. The NH4 + concentration was measured by the indophenols-blue method using a Spectronic-20D (Thermospectronic) spectrometer. Further details about the analysis of water-soluble inorganic ions are presented in previous papers (Ram et al. 2010a; Tiwari et al. 2009, 2012; Bisht et al. 2015).

A Chemiluminescence NO-NO2-NOx analyzer (Model 42i Thermo Electron Corporation, USA) was used for NOX measurements. Continuous PM10 measurements were also made at IITM using a beta attenuation monitor (C14 BETA; Thermo Andersen, USA) (Tiwari et al. 2013a). The PM10 concentrations were averaged to the same time intervals as the integrated PM2.5 samples so that the coarse particle mass (PM10–2.5) could be calculated for later use in the bext estimations.

Meteorological parameters (e.g. temperature, relative humidity (RH), wind speed, and direction) and horizontal visibility records were obtained from the India Meteorological Department (IMD), New Delhi, located in close proximity (~500 m) to the sampling site. Additional information about the mixing height (MH) over Delhi was obtained from the HYSPLIT model (http://www.arl.noaa.gov/ready.html; Draxler and Rolph 2016). The PM2.5 concentrations, CP, ionic species and bext are compared with visibility and MH, while the detailed investigation of the influence of the meteorological parameters, air masses and boundary-layer dynamics on the CP and ionic species concentrations was presented by Bisht et al. (2015).

2.3 Estimation of extinction coefficient

An initial estimation of bext was made using the original version of the Interagency Monitoring of Protected Visual Environment (IMPROVE) algorithm (Pitchford et al. 2007; IMPROVE 2011). The IMPROVE algorithm was initially adopted by the U.S. Environmental Protection Agency (EPA) for solar-light extinction estimations and to set a measure of the haze levels to provide a basis for their reduction (Pitchford et al. 2007). The algorithm has been widely used over the globe (Cao et al. 2012; Zhang et al. 2013a; Tao et al. 2014; Shen et al. 2014). In the present work, the original IMPROVE algorithm was initially used after some small modifications as:

$$ {b}_{ext}=3\times f(RH)\times \left\{\left[AS\right]+\left[ AN\right]\right\}+4\times \left[OM\right]+10\times \left[ LAC\right]+0.6\times \left[CM\right]+0.161\times \left[ Nitrogen\ Dioxide\right]+\left[ Fine\ Soil\right]+10 $$
(1)

where f(RH) is the hygroscopic growth term as a function of RH (Malm and Day 2001); daily-averaged RH values during the sampling periods were used to determine f(RH), which may underestimate the aerosol light scattering. The [AS], [AN], [OM], [LAC], [CM] and [Fine Soil] are the mass concentrations of ammonium sulfate (AS = (0.944) × [NH4 +] + (1.02) × [SO4 2−]), ammonium nitrate (AN = (1.29) × [NO3 −]), organic matter, light absorbing carbon, coarse mass, and soil, respectively, while the value of 10 corresponds to the Rayleigh scattering, which is considered independent from the site elevation and meteorological conditions. Eq. (1) also assumes zero absorption from the atmospheric gasses (i.e. O2, N2) and particles such as sulfate, nitrate, and OM, while the light extinction contributed by any individual component can be estimated as separate term considering the particles as externally mixed (Pitchford et al. 2007). The unit of bext is in Mm−1 (= 10−6 m−1), the mass concentrations of the chemical species are expressed in μg m−3, while f(RH) is unitless. Bext is estimated at 550 nm, which is traditionally used for visibility protection applications to characterize the light extinction (Pitchford et al. 2007). The measured SO4 2− and NO3 − ionic concentrations are associated with the ammonium salts as (NH4)2SO4 and NH4NO3, respectively in Eq. (1).

The LAC corresponds to the measured EC concentration, while the CM to the coarse mode (PM10–2.5) concentration as described in section 2.2. Based on the previous OM vs OC relationships over IGP (Ram and Sarin 2015 and references therein), the mass concentration of OM is estimated as OC × 1.6, correcting the OC mass for other elements (hydrogen, oxygen, nitrogen) associated with the OC molecules. Therefore, Eq. (1) uses the factor 1.6 instead of 1.4 in the original IMPROVE algorithm for the mean ratio of OM to measured OC. Furthermore, Eq. (1) involves the absorption by NO2, which is neglected in the original IMPROVE algorithm. The soil fraction was estimated from the F− mass assuming a global crustal abundance of 3.5 % (Taylor and McLennan 1985) using the equation:

$$ \left[ Soil\right]=\left(\frac{1}{0.035}\right)\times \left[{F}^{-}\right]\kern0.5em =(28.57)\times \left[{F}^{-}\right] $$
(2)

The F− concentration for soil mass estimations was used because of unavailability of Al+, Ca2+, Mg+2 or Fe2+, which are much more abundant than F− in the soil samples (Guinot et al. 2007), this assumption introduces some additional uncertainty into the calculation of the reconstructed fine mass (RCFM) and in the estimations of bext, taking also into account that F− can have anthropogenic sources as well (Kumar et al. 2010; Saxena et al. 2014).

In addition, bext estimations across Delhi were also performed via the revised IMPROVE algorithm, which is well documented in Pitchford et al. (2007), as:

$$ {b}_{ext}=2.2\times {f}_s(RH)\times \left[ Small\ Sulfate\right]+4.8\times {f}_L(RH)\times \left[ Large\ Sulfate\right]+2.4\times {f}_s(RH)\times \left[ Small\ Nitrate\right]+5.1\times {f}_L(RH)\times \left[ Large\ Nitrate\right]+2.8\times \left[ Small\ Organic\ Mass\right]+6.1\times \left[ Large\ Organic e\ Mass\right]+10\times \left[ Elemental\ Carbon\right]+1\times \left[ Fine\ Soil\right]+1.7\times {f}_{ss}(RH)\times \left[ Sea\ Salt=1.8*{Cl}^{-}\right]+0.6\times \left[ Coarse\ Mass\right]+ Rayleigh\ Scattering\ \left( Site\ Specific\right)+0.33\times \left[{NO}_2(ppb)\right] $$
(3)

Compared to Eq. (1), the revised IMPROVE algorithm takes into consideration three components (i.e. sulfate, nitrate and organic mass) that are split into small and large particle-size modes, while for sulfate and nitrate the algorithm includes different growth factor terms for the large and small modes for less biased bext estimations as:

$$ \left[ Large\ Sulfate\right]=\frac{{\left[ Total\ Sulfate\right]}^2}{20},for\ \left[ Total\ Sulfate\right]<20\ \mu g\ {m}^{-3} $$
(4)
$$ \left[ Large\ Sulfate\right]=\left[ Total\ Sulfate\right],for\ \left[ Total\ Sulfate\right]\ge 20\ \mu g\ {m}^{-3} $$
(5)
$$ \left[ Small\ Sulfate\right]=\left[ Total\ Sulfate\right]-\left[ Large\ Sulfate\right] $$
(6)

Similar differentiation and a threshold of 20 μg m−3 exists for the nitrate and OM. The revised algorithm considers variable dry-mass extinction efficiencies that are smaller (larger) than the original values for the small (large) particle size modes. This change was made to account for the growth of dry particles results in size ranges with higher scattering efficiencies under more turbid conditions (Pitchford et al. 2007). Furthermore, the Rayleigh scattering component was calculated for each daily sample taking into consideration the elevation of the site and the annual average temperature. The sea-salt component was neglected in the present study due to unavailability of Cl− measurements for its consideration. However, this deletion is expected to produce little bias since sea salt is mostly important in coastal environments. Behera and Sharma (2010b) found negligible sea-salt amount over Kanpur in the central IGP. The ratio between the OM and measured OC was fixed at 1.6 in the initial calculation. It has been modified to 1.8 in the revised IMPROVE algorithm. The fS(RH) and fL(RH) were taken from Pitchford et al. (2007), for the averaged RH values during the sampling periods, while the remaining components, i.e., EC, fine soil and coarse mass were calculated identically in the revised algorithm compared to the original calculations.

3 Results and discussion

3.1 PM2.5 and carbonaceous aerosol concentrations

Before examining the primary and secondary aerosol emissions and the estimations of the bext via chemical analysis of PM2.5, the temporal variability of PM2.5 and CP concentrations is analyzed in this section. The mean (December 2011 to November 2012) PM2.5 mass concentration at the measuring site was 153.6 ± 59.8 μg m−3 varying from ~65.0 to 289.7 μg m−3 compared to 171.6 ± 51.6 μg m−3 obtained from 90 PM2.5 samples during January – December 2012 (Bisht et al. 2015). This difference is attributed to the fact that the current work uses only daytime PM2.5 samples, which are generally lower than the nighttime ones, especially during post-monsoon and winter (Bisht et al. 2015). Table 1 summarizes the seasonal-mean mass concentrations of PM2.5 and CP, while the daily concentrations are presented in Fig. 1.

Table 1 Seasonal-mean concentrations of PM2.5, carbonaceous aerosols (OC, OM, and EC) and coarse particles (PM10–2.5) over Delhi during December 2011 – November 2012. [Organic Matter: OC*1.6; total carbonaceous particles: TCP = OM + EC]
Full size table
Fig. 1
figure 1

Daily variability of PM2.5, OC and EC mass concentrations on the 75 days with PM2.5 samples in Delhi during December 2011 – November 2012

Full size image

The PM2.5 annual and seasonal means exceed the National Ambient Air Quality Standard (NAAQS; http://www.cpcb.nic.in/National-Ambient-Air-Quality-Standards.php) of 40 μg m−3 and are far beyond the annual average limits of 12 μg m−3 established by the United States Environmental Protection Agency (USEPA; http://www.epa.gov/air/criteria.html) and the 24-h limits (25 μg m−3 not to be exceeded for more than 35 days in a year) of the European Union legislation (Directive 2008/50/EC 20). On a monthly mean basis, 241.9 ± 41.6 μg m−3 was observed in November 2012 due to substantial agricultural crop residue burning in northwest India (Kaskaoutis et al. 2014). Minimum PM2.5 (84.6 ± 13.6 μg m−3) was observed in August 2012 because of washout similar to Bisht et al. (2015). The highest PM2.5 concentrations during post-monsoon and winter (Table 1) are attributed to increased agricultural burning, biofuel and solid waste burning for heating purposes as well as the lower mixing layer height trapping the pollutants near the ground (Hÿvarinen et al. 2011; Rastogi et al. 2014; Bisht et al. 2015).

Prominent peaks in PM2.5 were mostly observed during the pre-monsoon (Fig. 1), associated with transported dust plumes from the Thar Desert, Arabian Peninsula, and other areas in western Asia that affect ground-level PM2.5 concentrations (Tiwari et al. 2013b). High PM2.5 concentrations have been reported over the whole IGP [e.g. Agra: 90.2 ± 7.2 μg m−3 (Pipal et al. 2011); Delhi: 148.4 ± 67.0 μg m−3 (Dey et al. 2012), 97 ± 56 μg m−3 (Tiwari et al. 2009), 122.3 ± 90.7 μg m−3, (Tiwari et al. 2013a), 123 ± 87 μg m−3, (Guttikunda and Calori 2013); Lucknow: 101.1 ± 22.5 μg m−3, (Pandey et al. 2012); Kanpur: 95 μg m−3 (Sharma and Maloo 2005); Kharagpur: 89.7 ± 33.7 μg m−3, (Srinivas and Sarin 2014), among many others], which are comparable in magnitude with the current concentrations.

A mean OC concentration of 33.5 ± 15.9 μg m−3 was found on an annual basis, varying from ~12.1 to 69.8 μg m−3 (Fig. 1). OC is the major contributor (~22 %) to PM2.5 mass, comparable to results at urban European sites (Putaud et al. 2010). The EC concentration ranges between 2.5 and 19.0 μg m−3 (mean of 6.9 ± 3.9 μg m−3) corresponding to ~5 % of the PM2.5 mass. Bisht et al. (2015) reported very similar values for OC (37.7 ± 14.3 μg m−3) and EC (7.8 ± 3.7 μg m−3), with larger concentrations during night-time in all seasons. OC exhibits its highest concentration (59.6 ± 5.3 μg m−3) in January and lowest (19.3 ± 5.4 μg m−3) in August, whereas EC presents a November high (14.2 ± 2.9 μg m−3) and an August low (3.9 ± 0.9 μg m−3). Prior work in Delhi has found average OC in PM10 of 25.6 ± 12.5 μg m−3 and EC of 12.1 ± 6.8 μg m−3 in 86 samples collected between January 2013 and June 2014 with similar seasonal patterns to those observed in this study (Sharma et al. 2016). In a previous study in 2010, Sharma et al. (2014) reported average OC of 26.7 ± 9.2 μg m−3, and EC of 6.1 ± 3.9 μg m−3.

A near-zero intercept in the regression of EC against OC [EC = (0.21 ± 0.02)*OC – (0.06 ± 0.59), R2 = 0.70] suggests negligible OC emissions from primary non-combustion sources, such as road re-suspended soil carbonates (Kamani et al. 2015), biogenic emissions, evaporation of fuel and solvents. This result differs from Kanpur, where an intercept of 5.76 was found (Behera and Sharma 2010b). The high concentrations of CP during late post-monsoon and winter (Table 1) are attributed to calm, stable atmospheric conditions and lower MH (Bisht et al. 2015), along more extensive fossil fuel and biofuel combustion for heating purposes, influence of transported smoke from agricultural burning, and fireworks during Diwali festival (Tiwari et al. 2012).

The OC/EC ratio is commonly used as an identification factor for the relative dominance between biomass/biofuel burning and fossil-fuel combustion (Turpin and Huntzicker 1995; Ram et al. 2010a; b; Pio et al. 2011; Panda et al. 2016). In the present study, the OC/EC varies from 2.45 to 9.26 (average: 5.18 ± 1.47) compared to 5.86 ± 0.99 in Bisht et al. (2015). It exhibited only a small seasonal variability from 4.24 ± 0.83 in the post-monsoon to 5.66 ± 1.68 in the monsoon (Table 1). These values are indicative of mixing of various carbonaceous sources, such as coal combustion, motor vehicle exhausts, agricultural burning, solid waste burning, industrial and thermal power plants emissions. More comparisons between OC, EC concentrations as well as OC/EC over Delhi and other cities in Asia can be seen in Bisht et al. (2015). Panda et al. (2016) report a winter only OC/EC ratio of 1.37 ± 0.16. However, their OC/EC analysis was performed with the IMPROVE protocol so that their EC values are likely to be a factor of 2 higher than with the NIOSH protocol (Chow et al. 2004). Thus, their OC/EC ratio is similar to the minimum value observed in this study. Sharma et al. (2014) found that OC/EC ratios varied from 3.8 to 5.8 in Delhi during 2010. Both of these studies examined PM10 using the IMPROVE protocol that typically measures more EC than the NIOSH protocol used in the present study.

During the study period, the total CP (TCP = 1.6 x OC + EC) varied from 22.0 to 126.1 μg m−3 (60.6 ± 28.8 μg m−3) contributing ~39.6 % to the PM2.5 mass, almost similar to that found in Athens (Mantas et al. 2014). Despite the large seasonality in the TCP concentrations (Table 1), the contributions to the PM2.5 mass are similar (42–43.5 %) in all seasons, except in pre-monsoon when it is lower (35.5 %) due to enhanced contribution of dust (Lodhi et al. 2013), highlighting the serious role of CP in emission rates and atmospheric chemistry over Delhi.

3.2 Estimation of primary and secondary emissions

Secondary organic aerosol (SOA) is an important component of haze formation, visibility degradation, climate and health-related issues (Hoyle et al. 2011; Ram and Sarin 2011). However, the estimation of SOA is complicated because it is difficult to distinguish between SOA and oxidized POC (Robinson et al. 2007; Jimenez et al. 2009). An approach to estimating SOA is through the estimation of secondary organic carbon (SOC) using the EC tracer method (Turpin and Huntzicker 1995; Castro et al. 1999; Ram et al. 2008, Ram et al. 2010a, b; Behera and Sharma 2010b; Panda et al. 2016). The EC tracer method for POC, SOC, POA, and SOA estimates are as follows (Feng et al. 2009; Ram and Sarin 2011):

$$ POC={(EC)}_{total}\times {\left(\frac{OC}{EC}\right)}_{primary} $$
(7)
$$ SOC={(OC)}_{total}-POC $$
(8)
$$ POA=1.6\times POC $$
(9)
$$ SOA=1.6\times SOC $$
(10)

One of the major uncertainties in the estimations of POC and SOC is the estimation of (OC/EC)primary (Behera and Sharma 2010b). In order to determine the (OC/EC)primary, we assume that the minimum OC/EC (2.45) represents the aerosol having negligible SOC (Castro et al. 1999; Panda et al. 2016). This value is reasonable based on the limited other results available for Delhi (Sharma et al. 2014; Panda et al. 2016). Using this approach, the annual average concentrations of POC and SOC were estimated to be 17.0 ± 9.7 μg m−3 and 16.7 ± 9.2 μg m−3 (~50.6 % and 49.4 % of OC mass), respectively, indicating equal contributions from POC and SOC to the total OC mass over Delhi (Table 2).

Table 2 Seasonal-mean concentrations of estimated POC, SOC and SOA, along with fractions of SOC/OC and SOA/PM2.5 over Delhi during December 2001–November 2012
Full size table

The cumulative contributions of POC and SOC to the daily PM2.5 samples along with the SOC/OC (%) ratios are shown in Fig. 2a, while Fig. 2b shows the daily variation of NO2 and SOA concentration and its contribution to PM2.5 from December 2011 to November 2012. The POC emissions dominate during post-monsoon due to frequent and extended agricultural burning in NW IGP, while SOC and SOA formations and % contributions to OC and PM2.5 dominate during winter (Figs. 2a, b, Table 2). Feng et al. (2009) found SOC/OC of the order of 18–19 % for urban and 23–25 % for rural sites in China. These values are much lower than the SOC formation observed in Delhi. Similarly, high POA emissions during post-monsoon and winter are due to agricultural crop residue burning, domestic biofuel and waste burning for heating (Chen et al. 2006).

Fig. 2
figure 2

Daily variability of (a) primary (POC), secondary (SOC) organic carbon emissions and SOC/OC and, (b) secondary organic aerosol (SOA), NO2 concentrations and SOA/PM2.5 for the 75 PM2.5 samples in Delhi during December 2011–November 2012

Full size image

The relative (%) contributions of POC, SOC, and EC to the total carbon (TC = EC + OC) are illustrated in Fig. 3 on a seasonal basis, revealing a significant SOC contribution ranging from 31.1 % in post-monsoon to 46.4 % in monsoon. The POC accounts for 48.9 % of the TC during post-monsoon, also associated with the largest fraction (20 %) of EC due to large-scale agricultural burning in Punjab resulting in transported smoke plumes over Delhi during October to November 2012 (Kaskaoutis et al. 2014). Significant fractions of POC (43.8 %) and EC (17.9 %) are also found during pre-monsoon attributed to wheat crop residue burning in NW IGP and forest fires in Himalayan foothills (Kumar et al. 2011; Vadrevu et al. 2012) although with lower TC concentrations (Fig. 3). The decrease in combustion activities and open fires during monsoon, the dilution of carbonaceous aerosols due to the increasing MH (Bisht et al. 2015) and the precipitation washout lead to the lowest TC concentration (27.0 ± 9.1 μg m−3), EC and POC % contributions (Fig. 3). On an annual average, POC, SOC, and EC account for 42 %, 41 %, and 17 %, respectively of the Delhi TC.

Fig. 3
figure 3

Relative (%) contribution of POC, SOC and EC emissions to the total carbon (OC + EC) on seasonal basis. The seasonal-mean values of TC are also given

Full size image

3.3 Inorganic species

The concentrations of the ionic species (SO4 2−, NO3 −, F− and NH4 +) are plotted in Fig. 4 (box and whisker charts) along with PM2.5, CP, and NO2 for a direct comparison between them. The average concentrations of SO4 2−, NO3 −, NH4 + and F− are 19.1 ± 12.5, 8.9 ± 8.6, 5.8 ± 3.0 and 1.2 ± 1.1 μg m−3, respectively indicating large contributions of sulphate and nitrate to PM in Delhi, although their concentrations are lower than those reported by Bisht et al. (2015). On a monthly basis, the highest concentrations of SO4 2− and NO3 − were found to be 39.4 ± 13.8 μg m−3 and 29.6 ± 7.2 μg m−3, respectively, in November, while the lowest values were in May (SO4 2− = 9.7 ± 5.9 μg m−3) and July (NO3 − = 2.5 ± 1.4 μg m−3). Ammonium and fluoride maximize in February (9.8 ± 2.8 μg m−3, 2.6 ± 1.3 μg m−3, respectively). The seasonality in SO4 2− and NO3 − for both daytime and nighttime concentrations is documented in Bisht et al. (2015).

Fig. 4
figure 4

Annual-mean concentrations (box and whisker plots) of PM2.5, OC, EC, NO2 and ionic species in Delhi during December 2011 – November 2012. The mean value stands for circle, the median with line (50 %), and vertical bars correspond to lower 5 % and upper 95 % of the values, while the asterisks represent the min and max values at each case

Full size image

The ionic species and OC are light scattering-type aerosols, while EC is a strong absorber of solar light (Novakov et al. 2005; Bond et al. 2013). Beyond their climate-forcing response, the ratios between organic and inorganic species are commonly used for source identification and mixing processes in the atmosphere (Behera and Sharma 2010a; Srinivas and Sarin 2014). A linear regression analysis between PM2.5, organic and inorganic components was performed and the Pearson correlation coefficients are summarized in Table 3. Statistically significant (95 % confidence level) positive correlations are found among the OC, EC, NO3 − and SO4 2− with PM2.5, suggesting that all species and, more specifically OC and SO4 2−, contribute significantly to PM2.5. SO4 2− is highly correlated (r = 0.79) with NO3 − indicating similarity in the emission sources or reaction pathways, as also shown between OC and EC (r = 0.84) for vehicular and bio-fuel combustion processes. The NO3 −/SO4 2− ratio was calculated to be 0.46 varying between 0.25 (May) and 0.75 (November), similar to values seen by Sharma et al. (2016). Significant correlations were found between NO3 − and OC (r = 0.66) and with EC (r = 0.64). EC is an indicator of vehicular emissions in urban environments as well as biomass burning while some of the nitrate and OC can be attributed to motor-vehicle emissions. SO4 2− was correlated with OC (r = 0.67) and EC (r = 0.60) likely result from the condensation of OC on the sulfate particles (Tiwari et al., 2015). The sulfate-EC correlation is commonly observed in urban settings and likely to result from coagulation of EC emissions with the secondary sulfate particles. In contrast, NH4 + and F− exhibit moderate-to-low correlations with the other components. Almost all of the chemical components exhibit significant negative correlations with the MH indicating that boundary-layer dynamics play an important role in air pollution levels in Delhi, as documented in Bisht et al. (2015). Therefore, the pollution levels maximize in the post-monsoon and winter and during the early morning and late evening hours when the MH is low trapping the pollutants near the surface (Bisht et al. 2015).

Table 3 Correlation coefficient (r) between PM2.5, carbonaceous aerosols and ionic species over Delhi from December 2011 to November 2012. [MH: Mixing height]. Values in bold (bold-asterisk) correspond to statistically significant correlations at 95 % (99 %) confidence level
Full size table

3.4 Estimation of light extinction coefficient by measured chemical species

The light extinction coefficient (bext) was estimated across Delhi using the chemical composition data of the PM2.5 samples using the original and revised IMPROVE algorithms (Pitchford et al. 2007). Before the bext estimations, PM2.5 was reconstructed (simulated) based on the measured chemical constituents. The RCFM (reconstructed fine mass) is computed as:

$$ RCFM=\left[AS\right]+\left[ AN\right]+\left[OM\right]+\left[ LAC\right]+\left[ Fine\ Soil\right] $$
(11)

where [AS] and [AN] are the concentrations of ammonium sulfate and ammonium nitrate, respectively, [OM] is the summary of primary and secondary organic matter, [LAC] is [EC], and fine soil is calculated by Eq. 2.

The measured and reconstructed PM2.5 concentrations exhibit a satisfactory agreement (R2 = 0.69) between the 75 samples (Fig. 5). In the most cases, RCFM underestimated the measured PM2.5 with the largest underestimations reaching 60 %. In about 20 cases the measured and simulated PM2.5 exhibit an excellent match with Root Mean Square Error (RMSE) <5 %, while the mean bias difference was found to be 23 %. The consideration of F− as a representative of the soil component introduces large uncertainties in the calculation of RCFM and contributes to the underestimations shown in Fig. 5, since major soil components like Ca2+, Mg2+, Al+ as well as K+, Na+, Cl− are not considered, whose summary was found to have an important fraction in the ionic composition over IGP (Ram et al. 2010a; Tiwari et al. 2016). Therefore, the large scatter of some data points and depart from the x = y line may be attributed to these issues.

Fig. 5
figure 5

Correlation between measured and reconstructed PM2.5 concentrations in Delhi during December 2011 to November 2012. The dashed lines show the lower and upper 95 % confidence levels of the linear regression. The dotted line represents the 1–1

Full size image

The percentage (%) contribution of each component to the reconstructed PM2.5 on a seasonal basis is shown in Fig. 6, revealing a clear dominance of OM (~40–44 %), followed by (NH4)2SO4 (~18–21 %). The soil represents a significant fraction in winter, pre-monsoon and monsoon (~25–30 %), while its % contribution minimizes in post-monsoon (12 %) due to increasing in components associated with biomass burning. The large contribution of soil component during pre-monsoon and monsoon is attributed to enhanced dust activity and transported dust plumes (Lodhi et al. 2013), while the large winter fraction may be mostly associated with road dust resuspension due to dry weather conditions (Behera and Sharma 2010b). F− that was used for soil estimations (Eq. 2), may also have anthropogenic sources other than soil. The mean fractional contribution of EC remains nearly constant at 4.7 to 5.6 %, with the largest value (6.7 %) in the post-monsoon. During post-monsoon, NH4NO3 contributes ~17 % to the reconstructed PM2.5 compared to 5–8 % in the other seasons (Fig. 6). Since Na+ and Cl− ions were not measured, the sea-salt mass contribution to PM2.5 was neglected that can only contribute a small fraction of the underestimation of the RCFM (Fig. 5) since prior work found low concentrations of sea salt over central IGP (Behera and Sharma 2010b).

Fig. 6
figure 6

Percentage (%) contribution of organic matter (OM), elemental carbon (EC), ammonium sulfate (NH4)2SO4, ammonium nitrate NH4NO3 and soil in the PM2.5 reconstructed mass on seasonal basis

Full size image

The daily variation of the estimated bext using the original and revised IMPROVE algorithms is plotted in Fig. 7a, b, respectively, as the cumulative absolute contributions of each chemical species. On a daily basis, bext varies from 196 to 1345 Mm−1 (mean of 545 ± 263 Mm−1) for the original IMPROVE algorithm, and from 256 to 2189 Mm−1 (mean of 696 ± 380 Mm−1) for the revised IMPROVE calculation. On a seasonal basis, bext is highest (910 ± 280 Mm−1) in the post-monsoon, followed by winter (877 ± 180 Mm−1), pre-monsoon (442 ± 109 Mm−1) and monsoon (363 ± 110 Mm−1) using the original IMPROVE, while the seasonality is slightly modified in the case of the revised IMPROVE with the highest bext occurring in winter (1222 ± 371 Mm−1), followed by post-monsoon (1123 ± 308 Mm−1), pre-monsoon (535 ± 145 Mm−1) and monsoon (457 ± 133 Mm−1). These bext values are significantly larger than those reported by Komppula et al. (2012) at Gual Pahari, about 20 km south of Delhi. However, Komppula et al. (2012) estimated the bext via LIDAR profiles as averaged values between 1 to 3 km from March 2008 to March 2009 and their results showed increasing bext within the first 1-km. Misra et al. (2012) reported bext values near the surface of 0.2–0.4 km−1 from May 2009 to September 2010 in Kanpur, which is much lower than the estimated bext over Delhi. However, the current estimates of bext are comparable to those found for aerosol-pollution episodes (bext between ~0.6 and ~1.1 km−1 near the surface) over Kanpur (Kaskaoutis et al. 2013), indicating high pollution concentrations in Delhi during the measurement period. High bext values of the same magnitude as those found in Delhi were estimated over the semi-arid urban site of Xi’an, China (Cao et al. 2012). The bext values exhibited an annual mean of 1102 ± 1020 Mm−1, ranging from 711 ± 499 Mm−1 in spring (MAM) to 1607 ± 1361 Mm−1 in winter (Dec, Jan., and mid-Feb). Furthermore, seasonal-mean bext values from 193 ± 95 in spring to 319 ± 207 in autumn were estimated in rural Guangzhou, southern China (Chen et al. 2016). Seasonally-changing meteorological conditions, rainy washout, changes in mixing-layer height and boundary-layer dynamics, variations in emission rates and influence of long-range transported plumes of biomass smoke and/or desert dust are factors that result in the large daily and seasonal variation in bext (Hÿvarinen et al. 2011).

Fig. 7
figure 7

Daily variability of the horizontal visibility and the cumulative contribution of each chemical component to the extinction coefficient obtained from the original IMPROVE (a) and from the revised IMPROVE (b) over Delhi from December 2011 to November 2012

Full size image

The daily bext estimations (Fig. 7a, b), as well as the seasonal means computed from the revised IMPROVE algorithm, are 22 % higher on an annual basis than those computed with the original IMPROVE algorithm. Figure 8 shows the correlation between the bext values computed from the original and the revised IMPROVE algorithms. A high correlation of R2 = 0.95 is observed between the two sets of bext values showing very little scatter but a significant departure from the x = y line, especially for the larger bext values. The results show that the estimations of the two algorithms are very similar under relative clean to moderately polluted environments, but under turbid atmospheres, the revised IMPROVE algorithm computes significantly larger bext values. This result shows the influence of different hygroscopic growth factor values and the split of AS, AN, and OM into small and large modes concludes to overestimation of the bext especially under turbid environments (Pitchford et al. 2007). Unfortunately, measured bext values are not available for the study period so it is not possible to assess the suitability of the two algorithm versions in representing the light extinction over Delhi.

Fig. 8
figure 8

Correlation between the bext values obtained from the original and revised versions of the IMPROVE algorithm over Dehi during December 2011–November 2012. The dotted line shows the linear regression and the pink area the limits of the linear regression at 95 % confidence level. The dotted line represents the 1–1

Full size image

The relative (%) contribution of the individual components to estimate bext on a seasonal basis is shown in Fig. 9a, b for the original and revised IMPROVE algorithms, respectively. As far as the original IMPROVE is concerned, OM has the largest contribution to bext (~40 %, on an annual basis) followed by (NH4)2SO4 (~21 %), EC (~13 %), coarse-mode particles (~10 %), NH4NO3 (~9 %) and soil (~6 %), while the contribution of NO2 is negligible (0.1–0.2 %). Alternatively, the revised IMPROVE calculations showed that on annual basis, OM contributes to ~48 % to the bext, followed by (NH4)2SO4 (~24 %), EC (~10 %), coarse mode (~8 %), NH4NO3 (~5 %) and soil (~5 %). The contribution of the chemical components to bext follows the same order for both IMPROVE versions, but the revised IMPROVE estimates a higher contribution from OM by ~8 %, with small (±2–3 %) differences in the other components. This difference is because the measured OM belongs mostly in the large mode, which is associated with the higher coefficient (6.1) in the Eq. (3) compared to coefficient 4 in Eq. (1). This calculation results in much larger bext values for the OM estimated from the revised algorithm. However, the large contributions of EC remain despite its low fraction in the reconstructed (4.7–6.7 %) and measured (4.5 %) PM2.5. This result highlights the significant impact of EC on solar-light extinction due to its strong absorbing capability (Bond et al. 2007, 2013).

Fig. 9
figure 9

Percentage (%) contribution of the chemical constituents to light extinction coefficient over Delhi for both the original (a) and the revised (b) IMPROVE algorithms on seasonal basis from during December 2011 – November 2012

Full size image

Despite the much lower OM concentrations in pre-monsoon and monsoon (Table 1), the OM contribution to bext in these seasons is comparable or even higher to that of winter (Fig. 9a, b) underlying the significant role that CP play in atmospheric chemistry and radiation attenuation over Delhi throughout the year. In general, the contributions of EC and NH4NO3 to bext are higher during post-monsoon and winter from both algorithms due to the larger impact of agricultural and biofuel burning. Soil contributes from 2.5 % (post-monsoon) to 7.7 % (winter) for the original IMPROVE and from 2.1 % (post-monsoon) to 5.9 % (pre-monsoon) for the revised IMPROVE, exhibiting much lower contributions to bext compared to PM2.5 (12.0–29.7 %) (Fig. 6). (NH4)2SO4 contributes significantly (~15–30 %) all year long due to continuously high emissions of SO2 and NH3 over the urban environment (Behera and Sharma 2010b), while the revised IMPROVE algorithm estimates larger (NH4)2SO4 contributions for the monsoon, post-monsoon, and winter and lower for pre-monsoon. Both algorithms agree that the highest contributions of coarse-mode particles to bext is in pre-monsoon (15.8 % and 13 %) due to much larger dust influence (Lodhi et al. 2013). The coarse mode contributions in winter are smaller (5.2 % and 3.7 %), suggesting that the vast majority of the particles are below 2.5 μm, similar to Komppula et al. (2012). The coarse-mode aerosol contribution to bext is also important during post-monsoon (11.3 % and 9.1 %), suggesting that the smoke particles may also be found in coarse mode after aging, condensation and coagulation processes during their transportation (Dumka et al. 2015).

A similar approach for estimation of bext via chemical analysis of PM2.5 in China (Cao et al. 2012) revealed largest contribution (~40 %) of (NH4)2SO4, followed by OM (~24 %), NH4NO3 (~23 %) and EC (~9 %), with minor contributions from soil dust (~3 %) and NO2 (~1 %). The dominance of CP (OM: 39.3 % and EC: ~20 %) on bext during winter was also found at Pearl River Delta, China (Zhang et al. 2013b) followed by (NH4)2SO4 (16 %), coarse mode (13 %) and NH4NO3 (11.8 %).

Light extinction is strongly linked with visibility impairment (Cao et al. 2012). The recorded daily horizontal visibility in Delhi is plotted against the estimated bext values from both algorithm versions in Fig. 10. Strong negative correlations (P < 0.05) with visibility, associated with 51 % and 54 % of the variance (R2), are calculated for the original and revised IMPROVE estimates of bext, indicating a direct impact of aerosol mass concentration on visibility degradation, as expected. In both cases, the scatter of the data points increases for the lower bext values, while for visibility values below 1.5 km, which correspond to very turbid atmospheres (dust storms or heavy haze/fog conditions), the bext estimates from the revised algorithm are significantly higher. However, changes in RH affect bext (see Eqs. 1 and 3) and its relationship with visibility due to the formation of low-level fog during winter and decreasing of visibility due to humid atmospheres during monsoon, resulting in weakening the correlation between bext and horizontal visibility. Furthermore, the visibility values were classified into five groups (1-km interval from 0 to 5 km), that are compared to the bext values estimated via both IMPROVE algorithms (Fig. 11a, b). During the study period, the mean visibility was found to be 2.8 ± 0.9 km, varying from 1.7 ± 0.80 km (post-monsoon) to 3.2 ± 0.5 km (pre-monsoon). Decreased visibility is associated with an increase in bext from 404 ± 89 Mm−1 for visibility 4–5 km to 500 ± 191 Mm−1 (vis =2–3 km) and 1120 ± 165 Mm−1 (vis < 1 km), according to the original IMPROVE estimations. The revised IMPROVE increases further the bext for the lowest visibility ranges, up to ~1600 Mm−1 for visibility <1 km. This strong negative correlation supports the use of the current methodology for the estimations of bext across Delhi. Similar atmospheric conditions of haze/fog formation are usually observed over polluted sites in China, where the high concentrations of secondary inorganic constituents (SO4 2− and NO3 −) were found to have a major impact on visibility degradation (Cao et al. 2012).

Fig. 10
figure 10

Correlation between the visibility and bext values estimated from the original and the revised IMPROVE algorithms over Delhi during December 2011–November 2012

Full size image
Fig. 11
figure 11

Box and whisker plots (as in Fig. 4) of the extinction coefficient for 5 groups of horizontal visibility for both original IMPROVE (a) and revised IMPROVE (b) estimates

Full size image

Figure 12 examines the influence of anthropogenic [OM + EC + NH4NO3 + (NH4)2SO4] and natural aerosols (soil + coarse mode) on bext estimates from both algorithms. Strong relationships (R2 = 0.94 and R2 = 0.97) are found between bext and anthropogenic components (Fig. 12a, c), suggesting that the urban/anthropogenic emissions play the major role in light extinction and visibility impairment in Delhi. The natural-origin particles also contribute to bext, but both the correlations are associated with large scatter (Fig. 12b, d) indicating that their role in the extinction of solar light is much lesser than that of anthropogenic aerosols. When the concentrations of the natural component are below 100 μg m−3, the relationships with bext are rather neutral. However, it should be noted that the assumption that the soil and coarse-mode particles are exclusively of natural origin is a first approximation since anthropogenic activities and farming practices also emit soil particles as well as the re-suspension of soil dust by road traffic. In contrast, F− may also indicate non-soil particles (Kumar et al. 2010), while a part of the OM, which is considered as anthropogenic in Figs. 12a, c may be derived from biogenic volatile organic compounds (VOC’s). However, these fractions are very difficult to quantify with detailed organic speciation and, therefore, the classification of anthropogenic and natural particles is subjected to bias.

Fig. 12
figure 12

Correlation between anthropogenic [OM + EC + NH4NO3 + (NH4)2SO4] (a, c) and natural (soil + coarse mode) (b, d) concentrations with extinction coefficient as function of visibility for the 75 PM2.5 samples in Delhi during December 2011–November 2012. The dotted lines show the lower and upper 95 % confidence levels of the linear regressions. The upper graphs (a, b) are for original-IMPROVE bext retrievals and the lower ones (c, d) for bext retrievals from the revised IMPROVE

Full size image

4 Conclusions

Chemical analysis for carbonaceous constituents (OC, EC) and ionic species (SO4 2−, NO3 −, NH4 + and F−) was performed on 75 PM2.5 samples collected in Delhi from December 2011 to November 2012. Additional measurements of PM10 mass concentrations, meteorological parameters, and horizontal visibility were made. Analyses focused on examining the seasonality of the particulate chemical components and their contribution to the extinction coefficient (bext) and visibility degradation. The annual-averaged PM2.5 mass concentration was 153.6 ± 59.8 μg m−3, from which the OC (33.5 ± 15.9 μg m−3, 22 % of PM2.5) and the ionic species (~34.9 ± 21.5 μg m−3, ~23 %) hold the largest fractions, followed by EC (6.9 ± 3.9 μg m−3, ~5 %). The PM2.5, OC and EC concentrations exhibited significant daily and seasonal variations peaking during post-monsoon and winter. The average OC/EC of 5.18 ± 1.47 (range from 2.45 to 9.26) indicated various types of carbonaceous aerosol emission sources (combustion of diesel, gasoline, coal, burning of biofuels/biomass). Primary and secondary OC were found to be of the same magnitude (17.0 and 16.7 μg m−3, respectively), exhibiting significant seasonal variability with larger concentrations in post-monsoon and winter. Secondary organic aerosol formation accounted for ~18 % of PM2.5, revealing a mean concentration of 26.7 ± 14.8 μg m−3 . The highest concentrations of SO4 2− (39.4 μg m−3) and NO3 − (29.6 μg m−3) were observed in November 2012, associated with major agricultural biomass burning in northwestern India. Estimates of bext were performed via PM2.5 chemical analysis using the original and the revised IMPROVE algorithms. Very large daily variability was found between the estimated bext values, which on the seasonal basis were found to maximize during post-monsoon and winter and to be strongly related (r = −0.72) with visibility impairment over Delhi. The percentage contributions of the chemical species to bext were seasonally dependent, following the order of OM (~40 %, 48 %), (NH4)2SO4 (~21 %, 24 %), EC (~13 %, 10 %), coarse-mode aerosols (~10 %, 8 %), NH4NO3 (~9 %, 5 %), soil (~6 %, 5 %) and NO2 (~0.2 %, 0.3 %), on annual basis for the original and the revised IMPROVE algorithms. The revised IMPROVE estimated significantly higher values of the bext for more turbid atmospheres and for visibility ranges below 1 km compared to the original algorithm; however, the bext values from both algorithms were highly correlated (R2 = 0.95). Future works exploring the suitability of the IMPROVE algorithms in simulating the measured bext using the larger dataset of PM2.5 samples will contribute significantly in monitoring the pollution/haze episodes over Delhi and helping the policy makers for mitigating efforts.

The current work is the first attempt to estimate the bext for the Delhi polluted atmosphere using chemical analysis of PM2.5 samples using the original and revised versions of the IMPROVE algorithm. In general, the estimated bext values follow the seasonality revealed from previous studies over Delhi and surroundings using in-situ (Nephelometer/Aethalometer) measurements and/or LIDAR profiles, giving support to the current estimates, especially due to an absence of validation procedure since measured bext is not available. Comparison with measurements at 21 sites in the USA showed that the original algorithm underestimated the highest bext values and overestimated the lowest ones. The revised algorithm reduced the under-prediction in bext under severe haze conditions and the over-prediction during low-haze periods compared with the original algorithm (Pitchford et al. 2007). Future work using a larger and more complete chemically characterized dataset of PM2.5 samples for more representative analysis regarding the seasonality of PM2.5, carbonaceous aerosols, and inorganic species, along with validation of the bext estimations against ground-based measurements will allow further use of chemical compositions for estimating extinction coefficients and visibility degradation across Delhi.