1 Introduction

Carbonaceous aerosols contribute to climate change by disturbing energy budget of the Earth (Xu et al. 2013). Elemental carbon (EC) has the potential to absorb solar radiation, resulting in positive radiative forcing in the atmosphere (Xu et al. 2013). On the other hand, organic carbon (OC) causes negative radiative forcing by scattering the Sun’s radiation (Jones et al. 2005). OC is potent enough to alter surface properties of aerosols and microphysical properties of clouds as well (Sjogren et al. 2007). Besides the climate impacts, OC is hazardous to health (Winquist et al. 2015). It consists of many carcinogenic and mutagenic species (Li et al. 2008) which diminish the longevity of humans (Harrison and Yin 2000).

Incomplete combustion of biofuels, fossil fuels and open biomass burning (Bond et al. 2007; Saarikoski et al. 2008) is the major anthropogenic sources of EC and OC. Except this, OC also has secondary sources such as chemical transformation of gaseous pollutants. Researchers describe South Asia as a major source of pollutants in the world today (Lawrence and Lelieveld 2010), and India being the most populous developing country of South Asia, energy consumption from coal and other biomass and fossil fuel is on rise which in turn increases the BC (black carbon) and OC concentrations (Bond et al. 2013). Black carbon or EC is emitted as particles due to incomplete combustion of fossil fuels, biofuel and biomass. Again, results derived from the Indian Ocean Experiment (INDOEX) (Ramanathan et al. 2002) revealed a thick layer of carbon-rich absorbing aerosols spreading over a large area from northern India to the intertropical convergence zone (ITCZ), known as Atmospheric Brown Cloud (ABC), which is capable enough to perturb the energy budget of India and so also South Asia. Besides, studies by Rehman et al. (2011), Bond et al. (2013) and Saud et al. (2013) also suggest the necessity of EC and OC measurement in Indo-Gangetic Plain (IGP) of India.

In the present study, focus is on contribution of OC and EC to PM2.5, because Seinfeld and Pandis (2006) observed that in urban locations contribution of carbonaceous aerosols to PM2.5 may be 40 %. Previous observations taken by Mahapatra et al. (2014a) and Sharma et al. (2014a) over Bhubaneswar (an eastern coastal rapidly urbanizing site in India) and Delhi (megacity in IGP), respectively, claimed high BC and particulate matter (PM) concentrations in winter as compared to any other seasons. Backward air trajectory analysis done by Norman et al. (2001) revealed that in winter, air masses reach Bhubaneswar through the IGP region and Western Asia, considered to be the most polluted and populated regions. Bhubaneswar being situated at the eastern coastal plain, near to Bay of Bengal (~60 km), pollutants from Pakistan, West Asia and IGP enter the Bay of Bengal through Bhubaneswar (Sen et al. 2014; Kulshrestha and Kumar 2014). Therefore, for the very first time, an attempt has been made to evaluate the existence of any similarities in PM, EC and OC characteristics of Bhubaneswar and Delhi during winter months of 2012–2013. This evaluation would provide some crucial information regarding possible source of pollutants for both the locations and potential effect of polluted IGP on eastern coastal plains.

2 Study area

Measurement of PM2.5 at Bhubaneswar was taken at the roof top of CSIR-Institute of Minerals and Materials Technology (20°29′N and 85°83′E), an urban location (altitude of 45 m above sea level) situated in the eastern coastal plains of Odisha, India. Bhubaneswar being one of the fastest developing urban areas, there is sporadic growth of different industries like fertilizer, iron and steel, agro, paper and cement in the vicinity. The monthly 24-h average temperature in winter is 23 °C with stable atmospheric conditions. More details regarding the study site and regional meteorology have been suggested elsewhere (Mahapatra et al. 2014b).

PM2.5 samples were collected at sampling site of CSIR-National Physical Laboratory (28°38′N, 77°10E, 218 m above sea level), Delhi. Surrounding the sampling site, vast traffic density (~100 m) and agricultural fields (~500 m towards southwest) are present. According to the records of Delhi statistical handbook 2012, a total number of registered vehicles in the city were 7.77 million in 2012–2013. Besides industrial emissions, vehicular emission, secondary aerosol, biomass burning and dust storms could also increase particulate matter concentrations (Sharma et al. 2014a). Detail description of air flow pattern and other meteorological conditions over the site has been described elsewhere (Sharma et al. 2014b). The monthly 24-h average temperature in winter is 15 °C at Delhi.

3 Methodology

3.1 Sampling

PM2.5 sampling at both the locations was carried out on pre-baked (at 550 °C for 6 h in muffle furnace) quartz microfibre (QMA) filters (47 mm diameter). Sampling time was 24 h for Delhi, whereas for Bhubaneswar, it was 8 h (from 1000 IST to 1800 IST). Pre-baked filters were desiccated for 24 h before measuring the initial (before sampling) and final weight (after sampling). The desiccated filter papers were weighed using Sartorius semi-micro balance CPA225D with 0.01 mg readability to measure the initial and final weights. Fine particulate sampler (M/s. Envirotech; Model: APM 550) and low-flow air sampler (M/s. Polltech India Ltd) were used for PM2.5 sampling at Delhi and Bhubaneswar, respectively. PM2.5 samples were collected at a constant flow rate of 16.7 lpm through the respective samplers. Sample was collected at both the locations on different dates between 20 December 2012 and 26 February 2013. After collecting samples, filters were stored under dry condition at −20 °C in the deep freezer prior to analysis.

3.2 Analysis of OC and EC

Following USEPA methods, IMPROVE protocol with negative pyrolysis areas zeroed (Sharma et al. 2014a), PM2.5 samples collected from the said locations were analysed for OC and EC concentrations using OC–EC carbon analyser (Model: DRI 2001A, M/s. Atmoslytic Inc., Calabasas, CA, USA) at CSIR-NPL, Delhi. According to the principle of OC–EC analyser, ~0.536 cm2 punched area of QM-A filter was heated to 140, 280, 480 and 580 °C in pure helium to have OC1, OC2, OC3 and OC4, respectively (Chow et al. 2004). Again, for EC1, EC2 and EC3, the same punched area was heated to 580, 740 and 840 °C in 98 % helium and 2 % oxygen (Chow et al. 2004; Saud et al. 2012). Each filter was analysed in triplicate with several blank runs (OC–EC analyser runs without filters to eradicate the pollutants/impurities) to get the representative estimation of OC and EC mass in PM2.5. To perform OC and EC analysis, QM-A filters have been used by several researchers (Chen et al. 2004; Zhu et al. 2010). Moreover, to remove uncertainties in OC calculations due to volatilization of particulate organic carbon and over estimation of OC by absorption of gaseous organics, blank filters were analysed. Total OC (OCtot), EC and TC were calculated using the following formulae (Chow et al. 2004; Turpin and Lim 2001; Gu et al. 2010),

$${\text{OC}}_{\text{tot}} = {\text{OC}}_{1} + {\text{OC}}_{2} + {\text{OC}}_{3} + {\text{OC}}_{4} + {\text{OP}}$$
$${\text{EC}} = {\text{EC}}_{1} + {\text{EC}}_{2} + {\text{EC}}_{3} - {\text{OP}}$$
$${\text{TC}} = {\text{OC}} + {\text{EC}}$$
$${\text{TCA}} = {\text{OC}} \times 1.6 + {\text{EC}}$$

where OP is the maximum pyrolysed carbon concentrations and TCA is the total carbonaceous aerosol in urban location.

3.3 Semi-empirical EC-tracer method

As suggested by (Castro et al. 1999), in brief,

$${\text{OC}}_{\text{pri}} = {\text{EC}} \times \left( {\frac{\text{OC}}{\text{EC}}} \right)_{ \hbox{min} }$$
$${\text{And}}\;{\text{OC}}_{\text{tot}} = {\text{OC}}_{ \sec } + {\text{OC}}_{\text{pri}}$$
$${\Rightarrow}\, {\text{OC}}_{ \sec } = {\text{OC}}_{\text{tot}} - {\text{OC}}_{\text{pri}}$$

where \({\text{OC}}_{\text{pri}}\) is the primary OC concentration, \({\text{OC}}_{\text{tot }}\) is the total OC measured, \({\text{OC}}_{\text{sec }}\) is the secondary organic carbon \(( {\text{OC}}_{\text{sec }} )\) to be calculated, \(\left( {\frac{\text{OC}}{\text{EC}}} \right)_{ \hbox{min} }\) is the minimum OC/EC ratio observed throughout the experiment.

3.4 Analysis of K+

One-fourth fraction of each exposed filter paper was taken in a 50-ml stoppered test tube and extracted in 15 ml of ultrapure de-ionized water (conductivity >18.2 MΩ) in an ultrasonic bath. The filter was left overnight in the prepared solution to ensure complete solubility of the ions. The soluble components were then separated using centrifugation technique (at 2500 rpm over 10 min). The supernatant solution was filtered twice using nylon and hydrophilized poly (tetrafluoroethylene) membrane filters of 0.45 and 0.2 μm pore sizes, respectively (Das et al. 2011) in order to ensure high purity of the sample solution. The filtrate was then analysed using a Dionex ion chromatograph. The analysis of K+ present in PM2.5 samples collected at Bhubaneswar has been carried out by Dionex ion chromatographs (Model IC-1000), consisting of two systems for analysis of both anions and cations simultaneously along with an autosampler for accuracy and repeatability of sampling procedure. The system was fitted with appropriate guard, separation columns, suppressors and conductivity detectors. For cation analysis, CS14 columns with CSRS suppressor were used. K+ analysis of PM2.5 collected at sampling site of Delhi has been carried out by ion chromatograph (Model: Dionex ICS-3000, USA) with CS14 column. Calibration standards have been prepared by National Institute of Standards and Technology (NIST, USA). The analytical error (repeatability) was estimated to be 3 % based on triplicate (n = 3).

3.5 Meteorological data

At Bhubaneswar, meteorological data were collected through automatic weather station (AWS) of Rainwise Inc (CC-3000) installed on the roof of CSIR-IMMT at a height of 15 m from ground level. Meteorological parameters (average temperature, relative humidity and wind speed) used in this paper were daily averages of those obtained from the AWS at 15-min frequency. The meteorological parameters (temperature, RH, wind speed, wind direction and pressure, etc.) at CSIR-NPL, Delhi, were measured using sensors of meteorological tower (five stages tower of 30 m height, 100 m away from the observational site). The tower takes observations at five different layers. We use the meteorological data available at 10 m height (i.e. temperature, RH, wind speed and wind direction) during the study days over Delhi.

3.6 MACCity emission data

MACCity emission inventory is an extension of the Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP), a historical emissions project, and was developed as a part of two projects funded by the European Union, namely, monitoring atmospheric composition and climate (MAAC) and CityZen developed in order to provide a consistent emission inventory through 1850–2100. Initially, historical emissions data set had been developed on a decadal basis for the period 1850–2000, and later the ACCMIP and the Representation Concentration Pathways (RCP) 8.5 emission data sets have been extended on a yearly basis for the period 1960–2020 for the anthropogenic emissions and 1960–2008 for the biomass burning emissions.

4 Results and discussion

4.1 OC and EC fractions in PM2.5

Respective mean mass concentrations of PM2.5, OCtot, EC and TC observed at Bhubaneswar were 60.72 ± 20.1 µg/m3 (<97.87 to >36.34 µg/m3), 11.16 ± 4.28 µg/m3 (<20.86 to >7.66 µg/m3), 6.00 ± 2.26 µg/m3 (<10.97 >3.42 µg/m3) and 17.15 ± 6.48 µg/m3 (<31.83 to >11.52 µg/m3). On the other hand, mean mass concentrations of PM2.5, OCtot, EC and TC observed at Delhi were 186.25 ± 47.46 µg/m3 (<256.02 to >108.41 µg/m3), 16.46 ± 6.61 µg/m3 (<26.47 to >9.60 µg/m3), 12.04 ± 4.43 µg/m3 (<18.21 to >5.76 µg/m3) and 28.50 ± 10.98 µg/m3 (<44.68 >15.36 µg/m3), respectively (Table 1). The average TCA, determined by multiplying a factor of 1.6 to OCtot (Turpin and Lim 2001), was observed to be 17.15 ± 6.48 and 30.1 ± 10.18 µg/m3 at Bhubaneswar and Delhi, respectively (Table 1).

Table 1 OCtot, EC and TC variation from PM2.5 fraction on respective sampling days at Bhubaneswar and New Delhi
Full size table

Ram and Sarin (2010) have reported ~30 to 35 % contribution of TC in total suspended particulate (TSP) mass at urban and rural sites of northern India. In agreement to this, percentage contribution of TC in PM2.5 at Bhubaneswar was evaluated to be ~30.38 %. However, at Delhi, the percentage of TC in PM2.5 was ~14.97 % which is in accordance with the reports of Sharma et al. (2014a, b) where ~18.4 % contribution of TC was observed in PM10 mass during the year 2010. In a recent study, Mandal et al. (2014) have also reported very high annual average concentrations of OC (93.0 ± 44.7 µg/m3), EC (27.3 ± 13.4 µg/m3) and TC (176.1 ± 84.7 µg/m3; ~66 % of PM10 mass) in PM10 (280.7 ± 126.1 µg/m3) at an industrial area in Delhi. Therefore, location-specific measurements also had an effect on the percentage contribution of TC to the PM concentration. Difference in percentage contribution of TC to PM2.5 over Delhi and Bhubaneswar might be due to different emission sources.

Due to urbanization and rapid growth of pollution sources, PM2.5 concentrations of Bhubaneswar and Delhi were 1.01 ± 0.34 and 3.10 ± 0.79 times higher than the National Ambient Air Quality Standards (NAAQS) guideline which is 60 µg/m3 during the winter season. Again, OCtot/EC ratio over the observation sites are 1.88 ± 0.24 (Bhubaneswar) and 1.37 ± 0.16 (Delhi) as presented in Table 1. It has been observed (Table 1) that the average OCtot concentrations in PM2.5 were 11.16 and 16.46 µg/m3 and the average EC concentrations were 6.0 and 12.04 µg/m3 at Bhubaneswar and Delhi, respectively. Therefore, it is obvious that a high EC concentration at Delhi leads to a comparatively lower OCtot/EC ratio. These observations further supported pre-dominant contribution of fossil fuel combustion along with crustal sources to the overall PM2.5 concentration at both the study sites. Various locations in India, like Ahmedabad, Allahabad, Agra, Kanpur, Hisar, Mt. Abu and Manora Peak, revealed a high OC/EC ratios (Rengarajan et al. 2007, Rastogi and Sarin 2009; Ram and Sarin 2010; Ram et al. 2012; Pachauri et al. 2013) basically due to biomass burning sources, while the ratio at Pune (2.9 ± 0.5) and Mumbai (2.0 ± 0.3) was comparatively lower presenting fossil fuel combustion and crustal source (Venkataraman et al. 2006; Safai et al. 2014). Table 2 shows OC/EC ratios of various study area in India and their emission sources. Saarikoski et al. (2008) found an OC/EC ratio of 0.71 and 6.6 representing vehicular and biomass burning sources, respectively. However, according to Chow et al. (1996), OC/EC ratio higher than 2 indicates the presence of secondary organic carbon (OCsec). The partially elevated level of OC/EC ratio at Bhubaneswar in comparison with Delhi indicates the probability of OCsec formation at the former site. For the present study, sources of OCtot, EC and OCsec contributions are described latter.

Table 2 OC/EC ratio over various locations of India in winter season
Full size table

The percentage of EC, OCtot and TC in PM2.5 varied significantly at both the sites during the period of study (Table 3). Though the average OC1 fraction (Table 4) over the observation site in Delhi indicated substantial contribution of biomass burning to the carbonaceous aerosol mass, the average percentage contribution of OCtot to PM2.5 at Delhi was 8.65 ± 1.55 where as at Bhubaneswar it was 19.75 ± 9.34. Similarly, the average percentage contribution of EC and TC at Bhubaneswar was 10.63 ± 5.06 and 30.38 ± 14.28, respectively, where as at Delhi average percentage contribution of EC and TC was 6.32 ± 0.97 and 14.97 ± 2.40, respectively. It can be observed that the percentage contribution of OCtot, EC and TC to PM2.5, at Bhubaneswar was almost double than that of Delhi. This observation prudently indicates that the pollution sources at Delhi were myriad with carbonaceous aerosols contributing ~15 % to the total PM2.5 concentration, whereas at Bhubaneswar, the carbonaceous aerosols contributed almost 30 % to the total PM2.5 concentration during the study period. This observation can be justified by the two 6-day special observation periods during November 2009 and March 2010 by Perrino et al. (2011) who reported that combustion sources contributed to 6–7 % of the total aerosol mass, whereas the rest came from soil, inorganic secondary compounds and other organic species formed in the atmosphere.

Table 3 Percentage contribution of OC, EC and TC to PM2.5 at Bhubaneswar and Delhi
Full size table
Table 4 Percentage of eight carbon fractions over Bhubaneswar and Delhi
Full size table

A good correlation was observed between OCtot and EC at both Bhubaneswar (R = 0.92) and Delhi (R = 0.95) indicating their common sources of emission. Further, the OCtot/EC ratio at Bhubaneswar (1.88) and Delhi (1.37) indicates the dominance of motor vehicle exhaust, fossil fuel combustion and road dust during the study period. This is further supported by higher concentration of EC1 and OC2 fractions at both the sites (described in latter section) indicating the influence of vehicular exhaust and fossil fuel combustion.

OC concentration in Delhi reported during winter months by Srivastava et al. (2014) (38.1 ± 17.9 µg/m3), Tiwari et al. (2013) (54 ± 39 µg/m3), Pant et al. (2015) (104.4 µg/m3) was significantly higher than the current study where as reports by Sharma et al. (2013, 2014a, b) are more or less similar to the OCtot concentration reported here. These observations support a significant spatial variation of ambient aerosols depending on the sampling location in a megacity like Delhi.

To get a clear picture of seasonality of anthropogenically emitted OC at Delhi and Bhubaneswar, anthropogenic emissions of OC derived from MACCity emission inventory (Granier et al. 2011; Diehl et al. 2012) have been used in this study (Fig. 1). The OC emissions shown in the figure are average emission of 4 years (2010–2013) over geographical regions around Delhi (28.25°–29.75°N) and Bhubaneswar (19.25°–20.75°E). OC concentrations, derived from anthropogenic emissions, were observed more in Delhi as compared to Bhubaneswar which justifies the findings of the present study. Saud et al. (2012) observed that in IGP, dung cake is the highest contributor (1.74–4.64 g/Kg) to OC and IGP alone has 45 % contribution to total OC emissions of the country.

Fig. 1
figure 1

Seasonal variation in anthropogenic OC emissions over geographical regions around Delhi (28.25°–29.75°N) and Bhubaneswar (19.25°–20.75°E) over the period of 4 years

Full size image

These results indicate the dominance of carbonaceous aerosols from fossil fuel combustion like coal, in both the locations. Therefore, further analysis was done to support this observation.

4.2 Characterization of carbon fractions

IMPROVE protocol analyses EC and OC into eight carbon fractions. These carbon fractions represent different emission sources. According to Cao et al. (2005) and Gu et al. (2010), OC1 is the representative of biomass burning, while OC2 of coal OC3 and OC4 represents road dust emissions. Similarly, EC1 is representative of motor vehicular emissions, while EC2 and EC3 are mixtures of coal combustion and vehicular exhausts. In Bhubaneswar, OC2, OC3, OC4, and EC1 contribute 16.62 ± 2.51, 15.98 ± 0.46, 7.66 ± 3.11 and 26.62 ± 2.28 % (average) to TC concentrations which suggest dominance of coal combustions and vehicular emissions as sources of EC and OC (Table 4). Similarly, average variations in these carbon fractions observed in Delhi were 20.80 ± 9.51 (OC1), 31.56 ± 7.27 (OC2), 33.93 ± 4.48 (OC3), 13.53 ± 8.65 (OC4) and 103.45 ± 11.83 % (EC1) in TC. It can be observed that at Delhi, EC1, OC2 and OC3 made the maximum contribution to TC followed by OC1 and OC3. This indicates motor vehicle exhaust is the predominant contributed to TC followed by coal combustion, road dust and biomass burning sources at Delhi. Percentage of all the eight carbon fractions over Delhi were higher than that detected at Bhubaneswar indicating a very high pollution scenario at the national capital. Average percentage variation of OC1 for Bhubaneswar was 0.13 ± 0.12 % (in TC) even reaching zero value on some dates indicating least contribution of biomass burning source. However, in Delhi, the period being peak winter season, maximum use of biomass for heating purposes usually contributes to a high OC1 concentration in comparison with Bhubaneswar. Apart from biomass burning, there is also a chance of condensation of semi-volatile organic matter over the pre-existing particulate matter at lower-temperature conditions. Again, OC2 and OC3 abundance was relatively higher at Bhubaneswar suggesting contributions from coal combustion and road dust along with OCsec production at the site (Gu et al. 2010; Cao et al. 2004). Pyrolysed carbon (OP) percentage was more in Delhi (Table 4) in comparison with Bhubaneswar. Yang and Yu (2002) observed a direct relation between OP and water-soluble organic carbon (WOCSEC). They found that WOCSEC comprises up to 13–66 % of OP.

4.3 Secondary organic carbon (OCsec)

Organic carbon has both primary and secondary emission sources (Ram et al. 2008; Cabada et al. 2004; Canonaco et al. 2013). Primary organic carbon (POC) is emitted from biomass and fossil fuel burning (Ram et al. 2008), while condensations of semi-volatile organic carbons (VOCs) and atmospheric transformations of biogenic and aromatic hydrocarbons contribute to OC (Rengarajan et al. 2007) collectively, known as secondary organic carbons (OCsec). As direct measurement of OCsec is not viable, using EC-tracer method, OCsec was calculated at the locations (Turpin and Lim 2001). \(\left( {\frac{\text{OC}}{\text{EC}}} \right)_{ \hbox{min} }\) was taken to be 1.63 and 1.21 for Bhubaneswar and Delhi, respectively. The minimum values were determined by considering the average value of lowest three (OC/EC) ratios as suggested by Rengarajan et al. (2011).

During the sampling period, average OCsec concentration at Bhubaneswar and Delhi was 1.77 and 2.57 µg/m3, respectively, whereas for Bhubaneswar, average percentage contribution of OCsec to OCtot was 15.76 ± 8.41 %, and for Delhi, it was 14.65 ± 7.46 % of OCtot (Table 1). Therefore, it is clear that for Bhubaneswar, OC concentrations were equally dependent on primary as well as secondary sources. Slightly higher percentage contribution of OCsec at Bhubaneswar in comparison with Delhi indicates local and fresh emissions (Cao et al. 2004) from vehicular traffic, industrial emission and coal combustion from three coal-based thermal power plants surrounding the observational site of Delhi during winter (Sharma et al. 2014a). Besides the emission sources, local meteorology of a particular location plays a major role in building OCsec of that area. Researchers suggest that the higher the intensity of solar radiation, the more the photochemical activity and suitable conditions for OCsec formation (Cao et al. 2004). In the present context, though the surface air temperatures of Bhubaneswar (average temperature for the study period being 22.6 ± 2.22 °C) were higher due to higher intensity of solar radiation than that of Delhi (average temperature for the study period being 12.88 ± 3.52 °C), during the sampling days, OCsec concentration over the Delhi site was 1.45 times higher than that of Bhubaneswar pertaining to the myriad pollutants from various sources at the megacity Delhi. Further investigation is required to bring more clarity to such observations.

4.4 Tracing the source of EC and OC

4.4.1 Water-soluble potassium as marker of biomass burning

Various researchers (Khare and Baruah 2010; Ram and Sarin 2010; Satsangi et al. 2012) proposed water-soluble K+ as elemental marker for biomass/wood combustion. For the first time, Andreae in 1983 proposed K+/OC and K+/EC ratios as indicator to emissions from biomass and fossil fuel burning, respectively. During Savanna burning (Echalar and Gaudichet 1995), K+/OC ratio was observed to be 0.08–0.10, while it was recorded to be 0.04–0.13 for agricultural residue burning (Andreae and Merlet 2001). However, K+/EC ratio exhibited relatively low value, 0.13 ± 0.04 at Jaduguda due to fossil fuel emissions (Ram and Sarin 2010). As suggested by Andreae (1983), K+/EC ranging from 0.21 to 0.46 and 0.025 to 0.029 represents biomass and fossil fuel combustion, respectively. Table 5 represents ratio of K+/EC and K+/OCtot for all measurement dates at Bhubaneswar and Delhi. Average ratio of K+/EC and K+/OCtot over Bhubaneswar is 0.018 ± 0.013 and 0.006 ± 0.004 and at Delhi is 0.321 ± 0.122 and 0.230 ± 0.066, respectively. This observation further indicates that agricultural residue and fossil fuel combustion were the major contributors to OC and EC concentrations at Bhubaneswar. However, at Delhi, high K+ concentrations were determined in the samples (Table 5) suggesting the role of biomass burning sources as a probable contribution to EC and OC fractions of PM2.5. Therefore, at Delhi, biomass burning sources were supposed to be one of the contributors to carbonaceous aerosols along with fossil fuel combustion, motor vehicle exhaust and particles from crustal origin (Sharma et al. 2014a). Though measurement of levoglucosan a biomarker could have been a better parameter for confirming the role of biomass burning (Pant et al. 2015; Nirmalkar et al. 2015), it is out of the scope of the present study.

Table 5 K+ concentrations in PM2.5 at Bhubaneswar and Delhi
Full size table

4.4.2 Effect of long-range transport and wind pattern

HYSPLIT back-trajectory model (Draxler and Rolph 2012) was used to determine the source of pollutants through long-range transport at the receptor sites (Fig. 2a). Using this on-line model, 120-h backward wind trajectories reaching Bhubaneswar and Delhi were estimated at 500 m AGL and vertical velocity mode during the observation period. Within the boundary layer, 500 m height was chosen to indicate the air mass flow. Meteorological fields for this analysis were used from the global data assimilation system (GDAS). This analysis suggests that during the observation period, air mass reaching Bhubaneswar travels through IGP and parts of western India, whereas air masses reaching Delhi travel from parts of western India, Asia, Europe and even few regions of Africa suggesting respective source regions. Analysis of eight carbon fraction, discussed earlier, suggests the presence of coal combustion as one of the major sources of carbonaceous aerosols. Hence, a map of coal-based thermal power plant spread across western, northern and eastern India along with their capacity was generated (Fig. 2b) in order to determine the role of such industries and their contribution to regional pollution through long-range transport. From the figure, it can be observed that a number of such power plants are located at a close proximity to both the observation sites. This indicates that emissions from such coal-based thermal power plants could also impact the carbonaceous aerosol concentration at both the sites through long-range transport mechanism. Further average wind speed (m/s) and wind direction during winter season over the Indian subcontinent at mean pressure level of 1000, 925 and 850 millibar were determined using NCEP/NCAR reanalysis data (Fig. 2c). The wind circulation pattern across India during winter season indicates accumulation of pollutants from the nearby regions at both the sites and lower wind speed over Delhi in comparison with Bhubaneswar. This might aid in accumulation of pollutants over Delhi in comparison with Bhubaneswar leading to higher concentration of pollutants over Delhi. Hence, the prevailing wind pattern and long-range transport aid in aerosol transport from nearby regions to the receptor site.

Fig. 2
figure 2

a HYSPLIT back trajectory wind flow pattern over Bhubaneswar and Delhi. b Location of various coal-based thermal power plants which are considered to be the major contributors to carbonaceous aerosols at both the measurement sites. c Average wind speed (m/s) and wind direction during winter season over the Indian subcontinent at mean pressure level of 1000, 925 and 850 millibar using NCEP/NCAR reanalysis data

Full size image

4.5 Role of meteorology

Meteorology has been observed to play a vital role in variation of PM2.5 concentration (Tai et al. 2012), EC and OC variation (Lin et al. 2009). Developing a correlation matrix is traditional way of grouping few variables concerned into a matrix form and to visualize the correlation among them (Satsangi et al. 2014). Correlation matrix for this study was developed using SPSS Statistics 20 software. The relationship between various meteorological parameters (average temperature, relative humidity and wind speed), PM2.5, EC and OC was analysed using correlation matrix (Tables 6, 7) over both the locations. At Bhubaneswar, wind speed (WS) had a weak positive but non-significant correlation with PM2.5 (r = 0.200), OCtot (r = 0.386) and EC (r = 0.192). This suggests that WS might not be the predominant controlling factors and more number of measurements as well as chemical characterization of PM would give a clear picture to the existing correlation (Tai et al. 2010). However, at Delhi, wind speed had a negative correlation with PM2.5 (r = −0.232), OCtot (r = −0.064) and EC (r = −0.050) suggesting accumulation of pollutant and hence a rise in concentration of PM2.5. Temperature had a negative correlation with PM2.5, EC and OCtot at Bhubaneswar (r = −0.23, r = −0.65, r = −0.63) which indicates accumulation of particulates due to compressed boundary layer conditions due to low-temperature conditions. Similarly, at Delhi, PM2.5 (r = −0.76), EC (r = −0.77) and OCtot (r = −0.78) were negatively correlated with temperature indicating the profound role of low temperature in increasing particulate concentration. More biomass burning in homes with drop of temperature could also be a probable cause of increasing particulate matter load. At Delhi, relative humidity had a weak and non-significant negative association with PM2.5 (r = −0.005) and component EC (r = −0.13) and OCtot (r = −0.03); however, at Bhubaneswar, relative humidity had a non-significant but weak positive association with PM2.5 (r = 0.208) and OCtot (r = 0.027) but negative with EC (r = −0.07). This suggests that OC increased with higher relative humidity at Bhubaneswar.

Table 6 Correlation matrix with meteorological parameters and rotated component at Bhubaneswar
Full size table
Table 7 Correlation matrix with meteorological parameters and rotated component matrix at Delhi
Full size table

A strong positive correlation between EC and OCtot is attributed to primary emission sources of the two maintaining a fixed ratio over a particular location depending on the sources (Na et al. 2004). Though a correlation study between various meteorological parameters has been attempted with PM, EC and OCtot, their quantification is beyond the scope of the manuscript as the primary focus of the paper is to undertake a comparative study of sources at both the locations. Also a fewer number of samples limit the scope of quantifying meteorological impacts on the species mentioned above. However, the results of the study are indicative that meteorology plays a vital role in variation of PM, OCtot and EC, but long-term measurements would be fruitful in quantifying the same.

4.6 Role of atmospheric stability

Various studies suggest that stable atmospheric conditions play a vital role in increasing the concentration of PM and vice versa (Zhou et al. 2015; Aryal et al. 2008). Inversion conditions being an indicator of stability, an attempt was made to get an understanding of stability conditions at Bhubaneswar and Delhi on the measurement days. The same upper air data were obtained from the University of Wyoming website at 0000 GMT/0530 IST. Thermodynamic variables such as equivalent potential temperature \((\theta_{e} )\) were plotted against pressure (THETAPLOT diagram) where minima in \(\theta_{e}\) were identified as characteristics of inversion (Kumar et al. 2010). THETAPLOT diagram suggested in Fig. 3a for Bhubaneswar shows lowest equivalent potential temperature at low pressure levels ranging from 800 to 600 hPa on the sampling day, except 08 Jan 2013 (1000 hPa). Again, low pressure levels represent higher mixing height leading to vertical dispersion of pollutants. However, in Delhi, the pressure ranged between 800 and 1000 hPa (Fig. 3b) indicating a stronger inversion and more stable atmospheric condition in comparison with Bhubaneswar. Therefore, in comparison with Bhubaneswar, particulate matter trapping in air parcel under strong inversion cap could be possible which in turn increased mass of PM2.5 as well as components EC and OCtot.

Fig. 3
figure 3

Inversion graph over Bhubaneswar (19.25°–20.75°E) (a) and Delhi (28.25°–29.75°N) (b)

Full size image

5 Conclusion

Winter average concentrations of PM2.5 at Bhubaneswar and Delhi were 60.72 ± 20.10 and 186.25 ± 47.46 µg/m3, respectively. Percentage contributions of total carbon (TC) concentrations to PM2.5 were higher at Bhubaneswar (~31 %) than at Delhi (~20 %) indicating almost 80 % contribution to PM2.5 from a myriad source of pollution at the later site. The OCtot/EC ratio at Bhubaneswar (1.88) and Delhi (1.37) supported the dominance of motor vehicle exhaust, fossil fuel combustion along with road dust at both the locations. Corresponding K+ ion contribution to PM2.5, K+/OCtot and K+/EC ratios of both the locations depicted that biomass burning could contribute as one of the sources to OCtot and EC concentrations at Delhi, whereas fossil fuel burning could be the predominant source at Bhubaneswar. Percentage of all the eight carbon fractions were higher at Delhi indicating a very polluted atmosphere in the megacity. OC2, OC3 and EC1 were the dominant species (>13 %) at Bhubaneswar among all the eight carbon fractions which indicates that coal combustion, road dust and motor vehicle exhaust are the major contributors to OCtot and EC fractions. At Delhi, however, OC1, OC2, OC3, OC4, OP and EC1 were the most abundant species (>13 %) indicating that primarily motor vehicle exhaust, coal combustion and biomass burning were contributing to the high eight carbon fractions. These observations were further supported by the wind pattern that circulated in close proximity of various major coal-based thermal power plants to both the sites. MACCity-derived anthropogenic data also indicated localized sources of OCtot and EC at Delhi. On the other hand, sources of OCtot and EC at Bhubaneswar were attributed to long-range transport, vehicular exhaust and fossil fuel combustion. Besides primary sources, secondary sources of organic carbon were also present at the locations; however, OCsec concentration was 1.45 times more in Delhi as compared to Bhubaneswar. As this is an initial study dealing with OCtot and EC concentrations in PM2.5 over Bhubaneswar, more in-depth analysis is required to identify the exact emission sources of the carbonaceous aerosols.