Table 1 summarises the details of the whole air sample collection experiments for emissions from paddy stubble burning, garbage burning, busy traffic junctions, idling vehicular exhaust emissions and fuel evaporation. The paddy stubble burning samples (three flaming and three smouldering) were collected at an agricultural field in Kurari, Mohali (30.605∘ N, 76.744∘ E), on 4 November 2017 between 16:30 and 18:30 local time (LT). The garbage burning samples (five flaming and five smouldering) were collected at waste sorting and disposing stations in Mohali and surrounding villages (30.642–30.699∘ N, 76.713–76.729∘ E) between 7 and 17 February 2017. Figure S1 shows the flaming and smouldering fires, which were distinguished by a visual inspection of the presence of flame and white smoke as per previous studies (Chandra et al., 2017; Kumar et al., 2018). The fire in the flaming stage showed a clear flame with little smoke, while in the smouldering stage there was white smoke and no flame. The traffic samples were collected from three busy traffic junctions in Chandigarh and Mohali (Sohana Gurudwara Chowk, Mohali, 30.691∘ N, 76.698∘ E; Sector 79 and 80, Chowk, Mohali, 30.678∘ N, 76.721∘ E; and Transport Chowk, Chandigarh, 30.717∘ N, 76.812∘ E) from 3 to 15 March 2017. Although the vehicular emissions are known to be dependent upon several factors, their idling operation results in quite high emissions and fuel residues in the exhaust fumes (Yamada et al., 2011; Shancita et al., 2014). This is because, in idling operations, the engine does not work at its peak operating temperature and efficiency (Brodrick et al., 2002), resulting in incomplete fuel combustion (Rahman et al., 2013). In this study, prior to vehicular exhaust sampling, the engine was left running for about 5 min until it warmed up to normal working temperature (70–90 ∘C), and then the air was sampled directly from the mouth of the exhaust tailpipe, with the car in a stationary position and the engine running at idle speed. The idling vehicular exhaust samples were collected from 23 petrol vehicles (14 two wheelers and nine light-duty four wheelers), 33 diesel vehicles (six three wheelers, 12 light-duty four wheelers and 15 heavy-duty wheelers), nine LPG vehicles (three wheelers) and seven CNG vehicles (six three wheelers and one light-duty four wheeler) from March 2017 to October 2018 in Chandigarh and Mohali (30.660–30.750∘ N, 76.700–76.840∘ E). For a better representation in the results, the most common vehicle models on Indian roads were selected for this study based upon personal field observations and motor vehicle data provided by Ministry of Road Transport and Highways (MoRTH, 2017). The fuel evaporative emissions samples (10 each from the headspace of LPG, petrol and diesel) were collected in Mohali, Chandigarh and Panchkula on 13–14 August 2020. In India, the most commonly used LPG is of two types, namely domestic LPG for household cooking and commercial LPG for various commercial and industrial activities like hotels, restaurants, metallurgical applications, textiles, automotive, etc. Out of the total 10 samples of LPG evaporative emissions, five samples each were of domestic and commercial LPG. For a better representation, the samples of evaporative emissions were also collected from the most common brands of petrol, diesel and LPG fuels sold all over India and in Nepal, Bangladesh and Sri Lanka (namely Indian Oil, Hindustan Petroleum, Bharat Petroleum, Bharat Gas and Indane Gas). In addition, prior to the lighting of fires or turning on the engines, ambient air samples were also collected from the aforementioned sites to correct for ambient background concentrations.

The whole air was actively sampled in commercially available 6 L passivated SilcoCan air sampling steel canisters (Restek) and then analysed, using a thermal desorption gas chromatograph equipped with a flame ionisation detector (TD-GC-FID), within 1 d of the sample collection as per the collection procedure described in previous works (Chandra et al., 2017; Vettikkat et al., 2020). Stability tests of the compounds in the canisters were also conducted, which showed that all the measured compounds reported in this work, including alkenes and alkyne, remained stable for up to 3 d. The air was actively sampled into the canisters using a Teflon VOC pump (model N86 KT.45.18; KNF) operating at a flow rate of ∼ 5500 mL/min and pressurised up to 30 psi. The steel canisters were protected from dust and air particles using a Teflon membrane filter (pore size – 0.45 µm) in the sample inlet line. Prior to each sampling the canisters were cleaned and preconditioned as per EPA Method TO-15, using a TO-Clean canister cleaner (Wasson-ECE Instrumentation) and humidified nitrogen.

NMHCs in the sample air were measured using a gas chromatograph equipped with two flame ionisation detectors (GC-FID 7890B; Agilent Technologies). Sampling and pre-concentration was performed using a thermal desorption (CIA Advantage-HL and Unity 2; Markes International) unit coupled to the GC-FID system. Helium (99.999 % pure; Sigma Gases and Services) was used as the carrier gas. Hydrogen (99.9995 %; Precision Hydrogen 100 H2 Generator; Peak Scientific), synthetic air (99.999 %; Sigma Gases and Services) and nitrogen (99.9995 %; Precision Nitrogen trace 250 N2 Generator; Peak Scientific) were used as the FID gases (Table S1). Synthetic air (99.9995 %; Precision Zero Air 1.5 gas generator; Peak Scientific) was also used as the purge gas for the Markes thermal desorption unit.

Figure S2 shows the schematic representation of the TD-GC-FID instrument during a typical sample injection and chromatographic run. In the first stage, sample air was passed through a Nafion dryer (integrated into the CIA Advantage) to remove water (Badol et al., 2004; Gros et al., 2011). It was then preconcentrated at -30 ∘C (maintained by a Peltier cooling system) at 20 mL/min on an ozone precursor trap (U-T17O3P-2S; Markes International). The trap was a quartz tube (2 mm internal diameter; 60 mm long) containing Tenax TA, Carboxen 1003 and Carbosieve SIII as adsorbents. The preconcentrated trap was thermally desorbed by heating the trap rapidly to 325 ∘C, and held at this temperature for 20 min, so that all the preconcentrated NMHCs were thermally desorbed. Thermally desorbed NMHCs were then transferred via a heated inlet (130 ∘C) line onto the GC instrument consisting of two capillary columns (dimethyl polysiloxane, DB-1; 60 m × 0.25 mm, 1.00 µm film thickness; Alumina PLOT, Al2O3 deactivated with Na2SO4 and 50 m × 0.32 mm, 8 µm film thickness; Agilent Technologies). Table S1 lists the settings at which the flame ionisation detectors (FIDs) were operated and the oven temperature was ramped. Initially a temperature of 30 ∘C was maintained for 12 min, and thereafter it was increased at two subsequent rates of 5 ∘C/min (up to 170 ∘C) and 15 ∘C/min (up to 200 ∘C). The two columns were connected via a Dean's switch, which was turned on after 17 min of the chromatographic run. In these initial 17 min, the two columns were connected to each other in a series, and the eluents from the first column (DB-1) were directed onto the second column (Alumina PLOT). After 17 min, the series connection between both the columns was broken by turning on the Dean's switch, and the eluents from both columns were directed onto their respective FIDs. C6 and higher NMHCs were resolved on DB-1 column and detected on FID 1, while C2-C5 NMHCs were resolved on Alumina PLOT column and detected on FID 2. Thus, in a single run, C2-C10 compounds were measured simultaneously in two chromatograms.

Prior to the sampling, the instrument was calibrated by dynamic dilution with zero air at different mixing ratios (in the range of 2–200 parts per billion – ppb) using a standard gas calibration unit (GCU-s, v2.1; Ionimed, Ionicon Analytik GmbH). A NIST calibrated flow meter (Bios DryCal Definer 220) was used to measure the flows of both the standard gas and zero air mass flow controllers before and after the calibration experiments. Figure S3 shows the sensitivity and linearity of NMHCs obtained from the calibration experiments performed over a dynamic range of 2–200 ppb over two sets of calibrations, namely regular calibration of 2–20 ppb and a high mixing ratio calibration of 10–200 ppb. This covers a range of 2 orders of magnitude over which the instrument exhibited an excellent linearity (r2> 0.99) for all the 49 NMHCs. Figure S4 shows a typical chromatogram of the standard gas during the calibration experiment. Peak identification and quantification were performed using PC software (OpenLab CDS, ChemStation Edition, Rev. C.01.06(61); Agilent Technologies). Furthermore, Supplement 2 provides an example chromatogram for each source that was sampled. All the chromatograms were manually inspected to ensure correct peak identification, baseline determination and peak area calculation. The FID signal of a compound was recorded in the form of a current (picoampere – pA) by the instrument, and the area under the peak was calculated and expressed in units of pAs (picoampere seconds) and used to quantify the analyte. Individual peak areas (in pAs) were converted to parts per billion using the sensitivity factors obtained from calibration experiments. For highly concentrated samples, appropriate dilution was performed prior to the sample injection so that the measured concentrations were within the range of 5–30 ppb for most of the compounds. However, there were still a few compounds that were 50–200 ppb in some sources even after dilution (Table S2). The instrument linearity was therefore tested at high concentrations of up to 200 ppb, and excellent linearity (r2≥0.99) was observed for all the compounds (Fig. S3).

The Supplement 3 (in Excel file format) provides details of the measured mixing ratios for each individual sample measured by the TD-GC-FID system after dilution, mixing ratios of the compound in the actual sample after correcting for dilution along with uncertainty and the values in the corresponding background samples. For the major compounds determining the normalised source profiles (presented and discussed in Sect. 3.1), the sample values were significantly higher than the background values (even by an order of magnitude or more for smoke and vehicular exhaust source categories). Therefore, while the background values were used to calculate excess concentrations, they hardly played any role in the determination of the emission profiles. The peaks in the chromatograms of the emission sources were also well resolved and separated and were identified using the calibration gas standards. In case a shoulder peak was present, the parent peak was separately integrated, i.e., any interference from a shoulder peak was subtracted from the parent signal. In the calibration gas standard, some additional compounds were also present, namely 2,2,4-trimethylpentane, 2,3,4-trimethylpentane and methylcyclohexane, each of which had a well-resolved and separate peak during the calibration experiments. However, during the analysis of emission source samples, these compounds exhibited poor peak features like peak shape, several shoulder peaks, etc. Therefore, to remain consistent across all samples, these compounds were excluded from the analysis, and only those compounds that were well resolved were included. Table S3 lists the details of two VOC gas standards, namely the (1) gas standard (Chemtron Science Laboratories Private Limited) containing VOCs at a mixing ratio of circa 1 parts per million by volume (ppmv; stated accuracy of ±5 %) and the (2) gas standard (Apel Riemer Environmental, Inc.) containing VOCs at circa 500 ppb (stated accuracy better than 5 %) with which the instrument was calibrated. Instrumental sensitivities can change during a long run deployment owing to a change in settings and mechanical wear and tear, and therefore, regular calibrations are important to assess the instrumental stability. Table S4 shows the average sensitivity factors (pAs/ppb) and standard deviation derived from 13 calibrations performed regularly between December 2016 and October 2018, with no major changes (8 %–12 % for most of the measured compounds) observed in the instrumental sensitivities. A reasonable agreement (considering the maximum instrumental uncertainty error of < 15 %) was found for the average calibration factors between December 2016 and October 2018 and derived from the two different gas standards for the common compounds, such as isoprene (53.2 ± 4.9 and 55.6 ± 5.9 pAs ppb-1), benzene (67.8 ± 5.6 and 69.2 ± 5.5 pAs ppb-1) and toluene (74.6 ± 6.6 and 81.3 ± 7.7 pAs ppb-1). Table 2 lists the compound-specific precision errors, limit of detection (LOD) and total uncertainties. The precision of the instrument was evaluated under identical conditions, using the relative standard deviation of five individual measurements of 1 and 5 ppb of the standard gas mixture, and was in the range of 1 %–6 % for 1 ppb and 0.1 %–0.5 % for 5 ppb of the reported compounds. The limit of detection of the instrument was evaluated according to Eq. (1) at 5 % probability, using the standard deviation of eight zero/blank samples measurements under identical conditions (Penkett, 2007; ACTRIS, 2014). 1LOD=2tσ.

Here, σ is the standard deviation of eight blank measurements (manual integration of peaks in the blank sample; if peaks were missing then the integration of the baseline corresponding to the same retention time and average peak width was used), and t is the Student's t value for the 5 % probability and 7∘ of freedom. The instrumental LOD was in the range of 2–104 parts per trillion (ppt). The total uncertainties were calculated using the root mean square propagation of individual uncertainties, like the 5 % accuracy error inherent in the VOC gas standard concentration, error in the linear fit of the calibration curve, the error in the flow reproducibility of the two mass flow controllers and the precision error of the instrument. The overall uncertainties for all compounds were less than 15 %.

Figure 3a–d show the comparison of the percentage contribution of different chemical classes of NMHCs to the total mass concentrations, OH reactivity (s-1), normalised reactivity (gO3/gNMHC) and total BTEX loading (%) from various emission sources. The hydroxyl radical reactivity reflects the total pollutant loading of the air mass (Sinha et al., 2012) and was calculated using Eq. (2) as follows: 2Total NMHC OH reactivity=ΣkOH+NMHCi[NMHCi], where kOH+NMHCi is the first-order rate coefficient for the reaction of NMHCi with OH radicals (Atkinson et al., 1982, 2006), and [NMHCi] is the measured concentration of the NMHC.

The ozone formation potential (OFP) is used as a metric to measure the contribution of NMHCs to the total O3 formation potential in urban environments (Carter, 1994). Normalised reactivity R (gO3/gNMHCs emitted) is generally used to indicate OFP for NMHCs from emission sources using their source profiles and maximum incremental reactivity (MIR) values, using Eq. (3) as follows (Harley et al., 2000; Zhang et al., 2013): 3R=Σiωi×MIRi, where ωi are the weight percentage of NMHCi present in the emission source, and MIRi are the maximum incremental reactivity coefficients (Carter, 1994, 2009).

In order to ascertain any statistical difference between the average OFPs of the emission sources, we carried out Tukey's pairwise honestly significant difference test (which accounts for sample size), and the summary of the test results is provided in Table S5. Based on the statistical test, it could be concluded with more than 95 % confidence that CNG vehicular emissions and the fuel evaporative emissions had different OFPs compared to other emission sources. The averaged OFP for the emission sources was diesel vehicle exhaust (6.5 ± 0.6 gO3/gNMHC), smouldering paddy stubble fire (5.9 ± 0.2 gO3/gNMHC), LPG vehicle exhaust (5.7 ± 1.1 gO3/gNMHC), flaming paddy stubble fire (5.2 ± 0.9 gO3/gNMHC), flaming garbage fire (4.9 ± 1.1 gO3/gNMHC), smouldering garbage fire (4.4 ± 1.3 gO3/gNMHC), LPG evaporative emissions (4.5 ± 1.6 gO3/gNMHC), petrol vehicle exhaust (3.9 ± 0.7 gO3/gNMHC), diesel evaporative emissions (3.6 ± 0.9 gO3/gNMHC), petrol evaporative emissions (2.0 ± 0.4 gO3/gNMHC) and CNG vehicle exhaust (1.5 ± 0.8 gO3/gNMHC). Although, alkenes were not the largest emissions by mass, they were still the largest contributor to the OH reactivity (67 %–93 %) and OFP (70 %–83 %) in the fire and LPG evaporative emissions. In the paddy stubble and garbage fire emissions, alkenes and aromatics had the largest contribution to the total OH reactivity (∼ 90 % and 6 %–9 %, respectively) and OFP (70 %–82 % and 16 %–27 %, respectively). Alkanes have comparatively poor reactivity towards OH radical, and therefore, despite contributing 15 %–37 % to the total NMHC mass concentration, their contribution to the OH reactivity was very low (< 3 %).

In paddy stubble fires under flaming conditions, propene (33 %), and under smouldering conditions, isoprene (46 %), were the largest contributors to the total OH reactivity (details in Figure S5 and S6). These two NMHCs were also the largest contributors (∼ 40-50 % in total) to the OFP from paddy stubble fires. In garbage fires under both flaming and smouldering conditions, propene was the largest contributor to the OH reactivity (46 % and 42 %, respectively) and OFP (37 % and 30 %, respectively).

LPG evaporative and vehicular exhaust emissions comprised of 68 %–81 % and 56 % alkanes, respectively; however, > 90 % of total OH reactivity was contributed by the alkenes. Butenes were the largest contributors to total OH reactivity from domestic LPG evaporative, commercial LPG evaporative and LPG vehicular exhaust emissions (90 %, 79 % and 72 %, respectively) and OFP (71 %, 83 % and 59 %, respectively). A total of 81 % of NMHC emissions from CNG vehicular exhaust were C2–C3 alkanes, but the maximum contribution to the total OH reactivity and OFP was from ethene (47 % and 62 %, respectively).

In diesel evaporative emissions there was approximately equal contribution to the total OH reactivity from alkanes (36 %) and aromatics (44 %). This is because of the presence of larger fractions of heavier C5–C8 branched alkanes, which are generally more reactive towards OH radicals compared to the light C2–C4 alkanes. While the rate coefficient values of C2–C4 alkanes vary between (0.25–2.12) × 10-12 cm3 molecule-1 s-1 at 298 K, the rate coefficient values of C5–C8 alkanes are between (3.6-8.9) × 10-12 cm3 molecule-1 s-1 at 298 K. The largest contributors to the total OH reactivity were the trans-2-butene (10 %) and 1,2,4-trimethylbenzne (9 %), while 1,2,4-trimethylbenzene (13 %) and o-xylene (12 %) dominated the OFP from diesel evaporative emissions. OH reactivity from diesel vehicular exhaust emissions, however, were dominated by alkenes (> 75 %), and propene (32 %–39 %) and ethene (23 %–31 %) were the largest contributors in all the diesel vehicle categories. Both of these NMHCs also contributed > 50 % to the total OFP calculated from diesel vehicular exhaust. In petrol evaporative emissions, the largest contribution to the OH reactivity was i-pentane (∼ 30 %), pentene (22 %) and butene (20 %) isomers. However, in petrol vehicular exhaust both aromatics and alkenes became the dominant contributors (∼ 40 %–50 % each) to OH reactivity and the major contributors were propene (14 %–23 %), m/p-xylene (11 %–13 %), styrene (8 %–11 %), ethene (6 %–13 %) and toluene (7 %). For OFP from petrol vehicular exhausts, the largest contributing NMHCs were m/p-xylene (24 %–26 %) > toluene (14 %–16 %) > propene (5 %–9 %) > ethene (6 %–9 %).

The total OH reactivity from traffic emissions was dominated by alkenes (48 %) and aromatics (35 %). The NMHCs contributing the largest fractions to the total OH reactivity were styrene (9 %) > trans-2-butene (9 %) > isoprene (7 %) > 1-hexene (6 %) ∼ m/p-xylene (6 %) ∼ propene (6 %) and to OFP were m/p-xylene (14 %) > toluene (12 %) > 1,2,4-trimethylbenzene (7 %) ∼ ethene (7 %) > i-pentane (6 %). High contributions to OH reactivity from styrene, isoprene and 1-hexene are noteworthy. Even though these compounds were not the most abundant in the traffic samples by mass concentration, they are, however, very reactive with hydroxyl radicals in ambient air (isoprene – kOH= 10.0 × 10-11 cm3 molecule-1 s-1; styrene – kOH= 5.8 × 10-11 cm3 molecule-1 s-1; 1-hexene – kOH= 3.7 × 10-11 cm3 molecule-1 s-1 at 298 K; Atkinson et al., 1989; Atkinson, 1997). Isoprene, styrene and 1-hexene have been reported previously in various traffic and tunnel experiments across the world (Mugica et al., 1998; Borbon et al., 2001; Barletta et al., 2002; Ho et al., 2009; Zhang et al., 2018). Our traffic samples have comparable mixing ratios of isoprene observed from roadside ambient air measurements in Karachi (1.2 ± 0.9 ppb; Barletta et al., 2002), 43 Chinese cities (0.86 ± 0.83 ppb; Barletta et al., 2005) and Longchuan tunnel, Hefei (0.47 ± 0.20 ppb; Deng et al., 2018), but higher than Chapultepec Avenue tunnel, Mexico City (0.17 ± 0.02 ppb; Mugica et al., 1998), and the Fu Gui Mountain tunnel (0.14 ± 0.36 ppb; Zhang et al., 2018). In the Hong Kong tunnel experiment (Ho et al., 2009) and Taipei tunnel experiment (Hwa et al., 2002), isoprene was, however, undetectable. This variability in isoprene emissions from traffic/vehicular exhaust has been previously attributed to variable fuel types, vehicular engines and maintenance, driving patterns and sampling strategies.

The mixing ratios of styrene and 1-hexene measured in our traffic samples were higher than the Fu Gui Mountain tunnel (styrene – 0.08 ± 0.00 ppb; 1-hexene – 0.07 ± 0.00 ppb) but comparable to 1-hexene reported from Taiwan tunnels (Cross Harbour tunnel (0.99 ± 0.20 ppb), Chung-Bor tunnel (3.29 ± 2.36 ppb) and Chung-Cheng tunnel (2.49 ± 1.27 ppb; Chen et al., 2003). Though high mixing ratios of styrene are remarkable, it has been previously reported that styrene is one of the major VOCs emitted from diesel light-duty vehicles (LDVs), especially in cold transient mode (Tsai et al., 2012). Amongst our traffic samples, maximum mixing ratios of isoprene (1.11 ± 0.06 ppb), styrene (2.31 ± 0.16 ppb) and 1-hexene (2.38 ± 0.14 ppb) were observed in Transport Chowk (30.717∘ N, 76.812∘ E) which is one of the busiest traffic junction in Chandigarh during rush hours and witnesses a large vehicular fleet of diesel-run commercial LDVs.

In order to assess the health risks associated with these sources, we compared the fraction of BTEX compounds in each of the emission sources. Benzene is classified as a human carcinogen (IARC, 2012), the potential health risk assessments of which have already been elucidated in NW-IGP during the periods influenced by intense paddy stubble fires (Chandra and Sinha, 2016). Other benzenoids like toluene and xylenes have also been associated with adverse effects on human health (ATSDR, 2000, 2007) and are classified as group “D” carcinogens by the US Environmental Protection Agency (EPA). Using the BTEX fraction, which is a well-known metric (Słomińska et al., 2014), is useful for comparing the mass fractional BTEX content of the emission sources. The statistical differences in the average BTEX fraction between the different emission sources were ascertained by Tukey's pairwise honestly significant difference test, and the summary for this information is provided in Table S6. Based on the statistical test, it could be concluded with more than 95 % confidence that diesel and petrol evaporative emissions, diesel vehicles and smouldering paddy fires had different average BTEX fractions compared to other emission sources. Out of 28 possible pairwise comparisons, 14 pairs show statistically significant differences with ≥ 2σ confidence, three are only significant at 1σ level and the rest were not significant. The fraction of BTEX in the different emission sources was petrol vehicle exhaust (27 ± 5 %), smouldering garbage fire (26 ± 1 %), flaming garbage fire (24 ± 8 %), flaming paddy stubble fire (22 ± 5 %), diesel vehicle exhaust (19 ± 2 %), diesel evaporative emissions (17 ± 2 %), smouldering paddy stubble fire (13 ± 1 %) and petrol evaporative emissions (3 ± 1 %). LPG and CNG emission sources had < 1 % of benzene and, therefore, were identified as least harmful sources, while petrol vehicular exhaust, garbage fires and paddy stubble fires were the most toxic emissions which could severely impact human health considering their BTEX emission potential.

Inter-NMHC molar ratios (ppb/ppb) are very useful tools that can not only be used to distinguish between different emission sources but also constrain the identity of the sources affecting ambient mixing ratios in a complex environment (Barletta et al., 2005, 2017). This is because, for the NMHC species with similar chemical lifetimes, the molar ratios remain preserved during chemical oxidation and ambient dilution (Parrish et al., 1998; Jobson et al., 1999). Furthermore, NMHC molar ratios that remain similar across sources can also be employed to assess the photochemical age of air masses.

Table 3 lists the commonly used inter-NMHC molar emission ratios for the emission sources studied in this work. The toluene / benzene (T / B) ratio is a widely used ratio in identifying vehicular emission sources (Barletta et al., 2002, 2005). The T / B measured for traffic in this study was 3.54 ± 0.21 which is comparable to previous studies from busy traffic junctions in Karachi (2.2 ± 2.9; Barletta et al., 2002), Hong Kong (3.0 ± 0.4; Huang et al. 2015), Okhla, New Delhi (2.3 ± 1.7; Hoque et al. 2008), Antwerp, Belgium (3.5 ± 0.2; Buczynska et al. 2009), and Nanjing, China (2.6 ± 0.9; Wang et al. 2008). For the idling vehicular exhausts of different fuel types, this ratio varied between 0.38 and 10.9 and was 3.68 ± 0.58 for petrol vehicles, which is consistent with the previous works of Guo et al. (2011; 2.0–3.8). For the diesel vehicles, the T / B ratio in our study was 0.37 ± 0.20, which is similar to the average T / B ratio (0.37) from diesel vehicles in Australia (Anyon et al., 2003), Germany (0.56; Siegl et al., 1999) and Tokyo (0.3; Yamamoto et al., 2012). Furthermore, T / B ratios can also be useful for distinguishing the paddy stubble fire emissions in the flaming (0.38 ± 0.11) and smouldering stages (1.40 ± 0.10).

The i-butane / n-butane ratio (B / B) is another example of a widely used NMHC ratio for distinguishing between different fossil-fuel-related emission sources. However, in our study, we found that this ratio is not useful in a complex emission environment influenced by varied emission sources. This is because the ratio exhibits similar values (0.20–0.30) for paddy stubble fires, garbage fires, petrol evaporative, diesel evaporative and petrol vehicle exhaust and diesel vehicle exhaust emissions. Therefore, caution should be taken while using this ratio in complex emission environments where biomass burning, fossil fuel combustion and biogenic emission sources simultaneously occur on a significant scale and strength to contribute to the chemical composition of ambient air. i-pentane / n-pentane can, instead, be used as a more reliable ratio for distinguishing biomass burning emissions (0.06–1.46) from the petrol dominated traffic and fossil fuel emissions (2.83–4.13).