The sampling site (26∘51′55.4′′ N, 81∘0′17.5′′ E) is marked as a red triangle in Fig. 1. It is located in the city of Lucknow at a height of ∼ 12 m above the ground of the Uttar Pradesh Pollution Control Board (UPPCB) office building in Gomti Nagar. Residential buildings, office complexes, schools, big parks, and commercial spaces surround the sampling site. The industrial and manufacturing plants within and around the city are related to steel metal components and fabrication, automobile parts, chemical industries, and food and agro-based and handicraft sectors (chikankari, zardozi, bone craft). The industrial map of Lucknow and the nearby districts with major/mini-industrial areas, large-/medium-scale industries, sewage treatment plants, solvent-based industries, sugar mills, pharmaceutical industries, and power plants is also shown in Fig. 1. The measurements of NMVOCs were conducted using a PTR-ToF-MS (Ionicon Analytik GmbH) from 18 December 2020 to 5 May 2021, covering the winter and summer seasons. The study period is divided into two seasons according to the classification by the IMD (Indian Meteorological Department) as winter (December–February) and summer (March–May). The gaps in the sampling period from 3–8 January and 21 March–9 April were due to maintenance and calibration of the instrument. The average daily temperature was ∼ 28 ∘C over the whole study period in the city. The mean daily temperature during winters (December–February) was around ∼ 25 ± 2.5 ∘C and during summers (March–May) around 32 ± 3 ∘C. The relative humidity ranged from 64 ± 14 % during winters to 42 ± 11 % during summers. The comparison of temperature and relative humidity changes during both seasons are shown as box plots in Supplement Fig. S1. These values are based on the days when NMVOC measurements exist. The pre-dominant wind direction was south-south-east during the colder periods and south-west during the warmer periods, as shown in Supplement Fig. S1. The wind speed is relatively calmer during the winters than in the summers. All the instruments were placed inside a temperature-controlled laboratory during the campaign. Detailed descriptions of the instruments can be found in the subsequent sections.

The PTR-ToF-MS is widely used for measuring NMVOCs with high mass resolution and sensitivity. A detailed description of the instrument can be found in other studies (Jordan et al., 2009; Graus et al., 2010; Tripathi and Sahu, 2020; Tripathi et al., 2022), while a brief description is given here. The PTR-ToF-MS is based on the chemical ionisation method, facilitated by proton-transfer reactions with hydronium (H3O+) ions as the primary reactant ion, which causes much less fragmentation of organic molecules in the sampled air. The natural components of air (nitrogen, oxygen, hydrogen, carbon dioxide, argon) have a lower proton affinity than water molecules. They thus do not react with H3O+, while most VOCs have a higher proton reactivity than water, facilitating non-dissociative proton transfer. These H3O+ ions are generated with high efficiency (∼ 99.5 %) through a hollow cathode discharge source, and then these reactant ions enter the adjacent drift tube section. The sampled air is also injected into the drift tube section, where proton-transfer reactions between hydronium ions [H3O+] and neutral VOCs [RiHj] occur to form protonated product VOC ions [RiHj+1] and water molecules [H2O] as shown in Eq. (1). These [RiHj+1] then enter the orthogonal acceleration reflectron time-of-flight mass spectrometer via a specially designed transfer lens system (Jordan et al., 2009). 1H3O++RiHj→RiHj+1+H2O The parameters of the drift tube of the instrument were maintained at 2.2–2.4 mb, 60 ∘C, 600 V, 130 Td for pressure, temperature, voltage, and electric field (E/N, where E is the electric field strength and N is the gas number density), respectively, and operated with a time resolution of 30 s. Typically, these or similar values have been observed as most suitable for ambient air measurements of NMVOCs (Blake et al., 2004, 2009). During the study, the PTR-ToF-MS instrument's inlet was connected to a Teflon PFA (perfluoroalkoxy) tube (1.5 m in length) for drawing air samples at a flow rate of 60 mL min-1. The inner diameter of the tube was 0.75 mm, and the residence time of the air in the inlet was less than 1 s. The PTR-ToF-MS can identify hydrocarbons (HCs) and oxygenated VOCs at sub-ppb levels within 1 s (Graus et al., 2010; Müller et al., 2012). In this study, the PTR-ToF-MS measured 173 NMVOCs (m/z 31.018 to 197.216) at the sampling site. The reaction rates (k) of the ions were applied from the literature (Cappellin et al., 2012). A rate constant of 2 × 10-9 cm3 s-1 was assumed for all ions whose reaction rates (k) were not available in the literature (Smith and Spanel, 2005). The overall uncertainties were in the range of 8 %–13 % in the calculations of the mixing ratios of VOCs present in the standard mixture. The cause of uncertainties in calculating VOC mixing ratios includes the uncertainties in the mass flow controllers (MFCs) of a gas calibration unit (GCU) and the standard mixture (±5 %–6 %). The reaction rates (k) of the ion were applied from the literature (Cappellin et al., 2012). A rate constant of 2 × 10-9 cm3 s-1 was assumed for all ions for which reaction rates (k) were not available in the literature. Hansel et al. (1999) and Steinbacher et al. (2004) reported up to 30 % uncertainty in the calculations of the mixing ratios of VOCs due to the k reaction rate. The calibration of the instrument was performed at the start, middle, and end of the campaign using a certified standard gas mixture (L5388, Ionicon Analytik GmbH Innsbruck, with a stated accuracy of better than 8 %) containing ∼ 1.0 ppmv of VOCs. A detailed description of the calibration set-up, including zero measurements, is given in the previous studies (Jain et al., 2022; Tripathi et al., 2022) and in Supplement Fig. S2. For method detection limits (MDLs), we have calculated the MDL using 3σ (standard deviation) of the zero air of 20 min time duration data points. The exact mass-identified chemical formula and the family of species of the observed NMVOCs (173 in number) are given in Table S1 in the Supplement. The mixing ratios of these measured 173 NMVOCs are averaged over the study period and compared in the box plots as shown in Fig. S3. The three most-abundant NMVOC species are observed to be acetaldehyde, acetone, and acetic acid.

A HR-ToF-AMS (Aerodyne Research Inc., USA) was also deployed for campaign measurements. A HR-ToF-MS (Decarlo et al., 2006) measures size-resolved mass spectra of non-refractory PM2.5 (NR-PM2.5) with a high time resolution (2 min). A detailed description of the instrument can be found in other studies (Lalchandani et al., 2021; Shukla et al., 2021) and is explained briefly here. The ambient aerosol particles were sampled through the PM2.5 cyclone (BGI, Mesa Labs, Inc.), which gets transmitted through stainless-steel tubing (∼ 8 mm inner diameter and ∼ 10 mm outer diameter) with a maintained flow (0.08 Lpm). This set-up is further connected to a Nafion dryer (MD-110-144P-4: Perma Pure, Halma, UK) to reduce moisture content and is then connected to the instrument's sampling inlet. The ambient aerosols enter the aerodynamic lens through a sampling inlet (100 µm-diameter critical orifice) and focus on a narrow beam. This particle beam then enters a sizing chamber, where it can be sorted based on its size. This size-resolved beam enters the vaporisation chamber, and the non-refractory part of the particles (NR-PM2.5) vaporises at 600 ∘C and ∼ 10-7 Torr. These gaseous molecules are then ionised and detected by a ToF-MS, depending on their m/z ratio. The HR-ToF-AMS was operated in the high-sensitivity V mode for two cycles of 60 s (total 2 min), regularly switching between MS and PToF mode for 30 s each. During the study period, the particle-free air was provided for 1–2 h every week to check and correct the fragmentation table at m/zs 12, 16, 18, 29, 33, 40, and 44. The IE (ionisation efficiency) calibrations were performed at the start, middle, and end of the campaign study following the mass-based method (Jayne et al., 1998, 2000) with an SMPS (scanning mobility particle sizer) unit (TSI Inc.). The raw data from the HR-ToF-AMS were analysed for unit mass resolution (UMR) and high resolution (HR) using the SQUIRREL (version 1.59) and PIKA (version 1.19) toolkits in the Igor Pro software (version 6.37). The NR-PM2.5 is chemically categorised into organics (Org), nitrates (NO3), sulfates (SO4), and chlorides (Cl). The organic aerosol mass spectra obtained from HR analysis and UMR analysis were combined from m/z 12 to 300 (∼ 422 ions) to make the input matrix for PMF analysis. The PMF analysis was performed using the ME-2 engine implemented in SoFi Pro (Source Finder, Datalystica Ltd., Switzerland) (Canonaco et al., 2013, 2021) in the graphical interface software Igor Pro version 6.37 (Wavemetrics, Inc., Portland). The detailed analysis and results of the PMF of NR-PM2.5 are given in other studies (Lalchandani et al., 2021; Tobler et al., 2020; Talukdar et al., 2021), beyond the scope of this paper. In brief, the organic aerosol (OA) mass spectra from the HR-ToF-AMS were explained by five factors, consisting of one hydrocarbon-like organic aerosol factor (AMS_HOA), two solid fuel combustion factors (AMS_SFC/BB and AMS_SFC/OA), one more-oxidised oxygenated OA (AMS_MO-OOA), and one low-oxidised oxygenated OA (AMS_LO-OOA).

An Aethalometer (Magee Scientific, model AE-33) was also deployed at the campaign site to measure the real-time black carbon (BC) mass concentrations. It collects the aerosol particle samples on the quartz filter tape and quantifies the optical attenuation at seven different wavelengths (370, 470, 520, 590, 660, 880, and 950 mm) with high temporal resolution (1 min). It is based on a dual-spot technique for loading corrections (Drinovec et al., 2015). The change in optical attenuation measurements in the selected time interval at 880 nm is converted to equivalent BC (eBC) measurements using the mass absorption cross section (MAC) of 7.77 m2 g-1 (Drinovec et al., 2017, 2015). Using the enhanced absorption of biomass burning aerosols in the near-ultraviolet and blue wavelength ranges, the Aethalometer's multi-wavelength BC data may be apportioned into biomass burning and traffic combustion sources (Sandradewi et al., 2008; Zotter et al., 2017). The model employs an absorption Ångström exponent (AAE) value that corresponds to both vehicular and biomass combustion as the primary sources of light-absorbing particles. In this study, the AAE values of 0.9 for traffic and 1.5 for biomass burning emissions are based on previous studies (Tobler et al., 2020; Lalchandani et al., 2021). More details about the instrument can be found in the previous studies (Lalchandani et al., 2021; Shukla et al., 2021). The sampling site (building) is a part of the national Central Ambient Air Quality Monitoring Stations (CAAQMS). The meteorological parameters (temperature, relative humidity, wind parameters) and concentrations of trace gases (NO2, SO2, ozone) are downloaded from the CAAQMS dashboard (https://app.cpcbccr.com/ccr/#/caaqm-dashboard-all/caaqm-landing, last access: 15 July 2022) managed by the Central Pollution Control Board (CPCB), the government of India for Gomti Nagar station, Lucknow.

Numerous receptor models have been used to analyse the dynamic behaviour of ambient aerosol measurements and relate it to physical sources. One of the recently developed algorithms, PMF (Paatero and Tapper, 1994), has been explored by numerous studies to apportion the measured bulk composition and temporal variation of aerosols (Zhang et al., 2011; Talukdar et al., 2021). The PMF algorithm is a non-negative, symmetrical factor analytic technique that produces unique factorisation by iterative re-weighting of individual data values and unique solutions. It solves the common bilinear equation, given as (Eq. 2): 2Xmxn=GmxpFpxn+E, where X represents the measured matrix, G and F are unknown matrices, and E is the error/residual matrix. The m and n represent the time series and individual mass dimensions, and p is the number of factors. The calculated quantities G and F represent time series and profiles of the specific factor of the model solution, respectively. The ME-2 solver decreases the rotational ambiguity and fits the G and F entries to minimise the uncertainty in quantity “Q”. This Q is the sum of the squared residuals weighted by their respective uncertainties, as given in Eq. (3). From the equation, it can be inferred as a normalised chi-square metric, where eij represents the residual matrix of E and σij represents measured data uncertainties. 3Q=∑i=1m∑j=1neijσij2Qexp⁡=nm-pm+n Another quantity, Qexp⁡, the degree of freedom, depends on the dimensions of the matrix and the number of factors. In an ideal case, the ratio of Q/Qexp⁡ is expected to be 1, with all the elements of the measured matrix and uncertainties well-defined. However, it has been noticed in earlier studies that the absolute value of the ratio, Q/Qexp⁡, is not always equal to 1 due to errors in measured data uncertainties, transient sources, and unknown model residuals. It is recommended to use the relative change in this ratio and characteristics of the physical source while choosing the optimum factor solution (Paatero and Tapper, 1993, 1994). In this study, this algorithm is applied over the measured NMVOC mass spectra using ME-2 (a multi-linear engine) (Paatero, 1999) over SoFi Pro (Source Finder, Datalystica Ltd., Switzerland) (Canonaco et al., 2013) in a graphical interface software Igor Pro version 6.37 (Wavemetrics, Inc., Portland). Earlier studies applied a similar PMF algorithm over mass spectra of 90 NMVOCs in Delhi (Jain et al., 2022; L. Wang et al., 2020) and 101 NMVOCs in Beijing (L. Wang et al., 2021). In this study, for the first time, we have included 170 NMVOCs measured by the PTR-ToF-MS from m/z 42.034 to m/z 197.216. The input and the residual error matrix for the PMF analysis were prepared using time series of mass spectra and calculated individual errors for each data point, as explained in the previous study (Jain et al., 2022). After incorporating the calibration factors, the uncertainties or the residual error matrix are estimated by multiplying the peak area by the correction matrix. The total uncertainties vary in the range of 8 %–12 % during calculations of the mixing ratios of NMVOCs. The three most-abundant NNMVOCs are not included in the PMF analysis due to their high signal-to-noise ratios and relatively higher (about 5–15 times) concentrations than other NMVOCs, as shown in Fig. S3. The pre-treatment of the input matrix also includes applying a minimum error threshold. The weak variables, with a signal-to-noise ratio of < 2, and bad variables, with a signal-to-noise ratio of < 0.2, are down-weighted by 2 and 10, respectively (Paatero and Hopke, 2003; Ulbrich et al., 2009).

The PMF algorithm calculates factor profiles, unlike the chemical mass balance (CMB) receptor model. The most crucial decision for the interpretation of the findings of the PMF is selecting the optimum modelled number of factor solutions. This is achieved by applying several mathematical metrics, correlating with external measurements, and interpreting the physical sources. The ratio of Q/Qexp⁡ is first examined for every factor solution. The factor solution with an absolute value of the Q/Qexp⁡ ratio near 1 indicates an accurate estimation of errors, and it should be selected but not observed for real observations. The Q/Qexp⁡≫1 and ≪1 indicate underestimation and overestimation of errors or variability in the factor solution, respectively. It is anticipated that Q will drop with each addition of the number of factors, as this introduces extra degrees of freedom to improve the fit of the data. Another important metric is the evaluation of scaled residuals in the time series and mass spectra. The scaled residuals ±3 for each data point in the time series are considered evidence of a good solution (Paatero and Hopke, 2003; Canonaco et al., 2021). Figure S4 shows the scaled residuals over the time series and diurnal cycle for the 3–10-factor solution. In the present study, the Q/Qexp⁡ does not lie near 1, but the high percentage change in the Q/Qexp⁡ is observed while examining three- to five-factor solutions, as shown in Fig. 2. The total scaled residual of all species is calculated and plotted for different factors in Fig. 2. The changes in the residuals and the drops in Q/Qexp⁡ indicate that the five-factor solution is an optimum solution. This solution is further analysed regarding their mass spectral features, time series, and correlation with external tracers (Org, NO3, SO4, Cl from NR-PM2.5, organic resolved factors, gases (O3, NO, NO2, NOx, SO2), temperature, relative humidity (RH), wind direction (WD), wind speed (WS), and BC concentrations).

The optimum factor solution from the PMF analysis was further refined by self-constraining the secondary volatile organic compound (SVOC) factor with random values varying from 0.1 to 1 with delta a=0.1. Finally, a=0.3 was chosen as the optimum solution after examining the temporal and diurnal variation of the factor. More details about the constraint of the solution are explained in Text ST1 in the Supplement. Further, the uncertainty of the selected solution is quantitatively addressed by bootstrap analysis (Davison and Hinkley, 1997; Paatero et al., 2014), a module available in SoFi Pro (Canonaco et al., 2021), as explained in Text ST2. Previous studies have also followed this methodology for uncertainty estimation of organic aerosol source apportionment (SA) results (Lalchandani et al., 2021; Tobler et al., 2020; Shukla et al., 2021; Lalchandani et al., 2022), elemental aerosol SA results (Shukla et al., 2021), and VOC SA results (L. Wang et al., 2021; Jain et al., 2022; Stewart et al., 2021c). The uncertainty or PMFerror is observed as 1 % or less for all the factors (Fig. S5). This implies that the five-factor solution is a statistically robust solution with rather low uncertainty.

Ozone formation potential (OFP) is a reactivity-based estimation technique to assess the sensitivity of the VOCs for ozone formation (Carter, 2010, 1994). Numerous VOCs are emitted into the atmosphere from various sources, followed by distinct reaction pathways, and have different OFPs. The calculated reactivities of VOCs have been investigated in multiple modelling studies depending on the environmental conditions (Carter, 1994a). This approach is based on calculating OFPs using maximum incremental reactivity (MIR) values for individual VOC species, reported and updated by Carter (2010). MIR values are calculated as the change in the ozone formed by adding a VOC to the base case in a scenario with adjusted NOx concentrations. The OFP of individual VOCs is estimated using Eq. (4). Here, this equation is adopted (Carter, 2010, 1994) and modified for this study to calculate the ozone formation potential for each of the factors, resolved from PMF analysis as given below (Eq. 5): 4OFPj=[VOCj]×Cj×MIRj,5OFPi=∑j=0n[VOCj]×Cj×RCji×MIRj, where OFP(j) and OFP(i) represent the ozone formation potential for an individual VOC(j) and a factor number (i), respectively, expressed in micrograms per cubic metre. [VOCj] represents the mixing ratio (ppbv) of a given VOC ion (j), Cj is the number of carbon atoms present in each VOC ion (j), and RCji is the relative contribution of VOC ion (j) to the factor (i). MIRj is the maximum incremental reactivity of a VOC ion (j). The MIRj values are adopted from Carter (2008, 2010, 1994a). The above equation is used to compute the OFP for each factor and to determine which source factor contributes the most to ozone generation, as explained later. The MIR values are available for a limited number (40) of NMVOCs given in Table S2. The NMVOCs without reported MIR values are not considered for OFP estimation.

The chemical pathways and reaction products involved in SOA formation from NMVOCs are poorly understood or even unknown. Estimation of SOA formation (SOA yield) has been largely constrained to indirect methods due to the complexity of the chemical matrix of organic aerosols and the lack of direct chemical analysis methods. Numerous studies have estimated SOA yield from different species involving computer modelling and chamber experiments (Zhang et al., 2017). The smog-chamber studies help estimate the value for SOA yield as more reliable as they mimic the actual scenarios. These parameters also helped in improving model parameterisation and SOA mitigation strategies. For the current study, the contribution of an individual NMVOC species to SOA is estimated by multiplying the SOA yield by the concentration of the NMVOC species in the atmosphere (amount available for the reaction), as shown in Eq. (6). The SOA yields YSOA(j) reported by Bruns et al. (2016) were used for this analysis. The compounds for which SOA yields values are not available from the literature directly, and it is estimated that compounds with more than six carbon atoms (C >6) will have the same SOA yield of 0.32 (Bruns et al., 2016). Based on their structure, the compounds (C >6) are considered to contribute significantly to SOA (Bruns et al., 2016). The average value (0.32) of the published SOA yield of 18 compounds (C > 6) is used. In this study, the individual SOA yield values considered are given in Table S3. The contribution of the individual factor to SOA formation is also estimated using Eqs. (6) and (7) as given here. 6CSOA(j)=YSOA(j)×VOCj7CSOAi=∑j=0nVOCj×RCji×YSOA(j) CSOA(j) and CSOA(i) represent the contribution to SOA formation for an individual VOC(j) and a factor number (i), respectively, expressed in micrograms per cubic metre. VOCj represents the concentration (µg m-3) of a given VOC ion (j), and RCji is the relative contribution of VOC ion (j) to the factor (i). YSOA(j) is the SOA yield of VOC ion (j). This analysis represents the estimated OFP and SOA formation potential of the air mass composition at the sampling site, not the actual OFP and SOA formation potential from various sources. This means that air masses dominated by fresh emissions (e.g. traffic) will have different OFP and SOA formation potentials than those in aged air masses (e.g. long-range transport of BB plumes) or any other source.

Concentrated-weighted back-trajectory (CWT) analysis determines the originating source and transport of air parcels at the receptor location within a specific period (Seibert et al., 1994; Draxler et al., 1998). The HYSPLIT model (v4.1 Hybrid Single-Particle Lagrangian Integrated Trajectory) was used to perform the CWT analysis (Draxler et al., 2018; Stein et al., 2015). The 72 h back trajectories with a 3 h time interval at 100 m of arrival height above the ground were calculated using monthly GDAS (Global Data Assimilation System) files (ftp://arlftp.arlhq.noaa.gov/pub/archives/gdas1, last access: 15 July 2022) with a 1∘ × 1∘ resolution. The estimated back trajectories (BTs) were weighted with VOC factor time series and averaged over 3 h intervals using a CWT model to locate air masses based on their concentrations. ZeFir (Petit et al., 2017), an IGOR-based interface, was used to construct the CWT graphs, as shown in Fig. S6.

The average daily concentration of measured NMVOCs during the study period was 125.5 ± 37.5 ppbv. Figure 3 shows the daily time series and monthly mean concentrations of NMVOCs, inorganic and organic fractions of NR-PM2.5, O3, NOx, SO2, temperature, relative humidity, wind speed, and wind direction. Out of 173 detected NMVOCs, the levels of three major species (acetaldehyde, acetone, and acetic acid) were present 5–15 times higher than for other species, as shown in Fig. S3. The monthly averaged concentrations of NMVOCs were higher during the winter months from December (193.7 ppbv) to January (110.2 ppbv) till February (109.7 ppbv) than during the summer months March (101.2 ppbv), April (137.8 ppbv), and May (150.8 ppbv). The averaged concentrations of NMVOCs (127 ± 40 ppbv), as well as NR-PM2.5 (inorganics and organics) (102.8 ± 51 µg m-3), were higher during the winter months. The calm conditions and relatively lower planetary boundary layer during winters have slowed down the dispersion of the pollutants. In contrast, NR-PM2.5 (39.8 ± 20 µg m-3) decreased drastically during the summer months, but NMVOC concentrations (122 ± 32 ppbv) were similar to winters. This may be due to high temperatures during warmer periods leading to more photo-oxidation of primary VOCs (Sahu et al., 2017), production of biogenic VOCs (Sahu et al., 2017; Baudic et al., 2016), and evaporation of volatile household products (M. Qin et al., 2021), while aerosol particles managed to disperse in the atmosphere due to the high planetary boundary layer and windy conditions. The difference in the emission sources' characteristics during both seasons may have also played an important role. During the winter, the PM2.5 is exceeded on more days than the National Ambient Air Quality Standards (NAAQS). PM2.5 exceeds standards more frequently than ozone. On approximately 80 % of the days during the whole study period, PM2.5 exceeds the NAAQS in the city, as shown in Fig. 3.

The three most-abundant NMVOCs were not considered in the PMF analysis, as explained in Sect. 2.5. The remaining 170 NMVOCs, considered for the PMF analysis, varied from m/z 42.034 to m/z 197.216. The average concentration of these 170 NMVOCs was 79.3 ± 30.6 ppbv. The average concentration during winters was 86.7 ± 35 ppbv, a bit more than during summers at 68.3 ± 17.2 ppbv. These NMVOCs belong to different families based on their chemical composition. They are categorised as aromatics (Ar_CxHy), simple non-aromatics (N_CxHy), furans, phenols, and oxygenates: first-order (CHO1), second-order (CHO2), and third-order (CHO3), nitrogen-containing compounds (CxHyNz and CxHyNzOn), and others. The others include high-order oxygenates (CHO4) and some hydrocarbons (CxHy). The degree of unsaturation (i.e. the number of rings and/or double bonds) of more than 4 distinguishes aromatics (ArCxHy) from the CxHy family. This allowed us to identify important VOC markers, their families, and their role in their atmospheric chemistry. Overall, during the study period, the highest contributing family belongs to oxygenates and aromatics. The CHO1, CHO2, and CHO3 families were 28.8 % (∼ 20.1 ppbv), 16.8 % (11.7 ppbv), and 2 % (1.4 ppbv) of total NMVOC concentrations. The contributions from Ar_CxHy and N_CxHy were about 21.5 % (∼ 15 ppbv) and 10.6 % (∼ 7.4 ppbv), respectively. Nitrogen-containing compounds were relatively less present (5.6 % CxHyNz and 1.2 % CxHyNzOn); 6.3 % (∼ 4.4 ppbv) and 3.7 % (∼ 2.6 ppbv) were contributed by furans and phenols at the site, and the rest was included in the others (3.4 %). The CPCB notified the annual NAAQS only for benzene of 5 µg m-3 (∼ 1.6 ppbv), while the World Health Organization (WHO) recommended no safe level of exposure to benzene. The mean mixing ratio of benzene during the study period was found to be 2.9 ± 1.9 ppbv, which is around 2 times higher than the standard guidelines. Prolonged exposure or high short-term exposure to benzene adversely affects the health of citizens of the city due to its haematotoxic, genotoxic, and carcinogenic properties.

All three abundant NMVOCs present at m/z 45.034 (C2H5O, acetaldehyde), 59.049 (C3H7O, acetone), and 61.028 (C2H5O2, acetic acid) are OVOCs. The sources of these OVOCs could be direct emissions from biogenic and anthropogenic activities and the secondary/photochemical processes. Diurnal variations of secondary formation, anthropogenic emission levels, meteorological conditions, and planetary boundary layer heights influence OVOC / benzene ratios (Sahu et al., 2017; Tripathi et al., 2022; Sahu et al., 2016) to some extent. The diurnal patterns of the acetaldehyde / benzene, acetone / benzene, and acetic acid / benzene ratios are plotted to check the influence of biogenic and secondary sources (see Fig. 4a–c). All the OVOC / benzene ratios are observed to increase during the daytime (10:00–18:00), similarly to temperature variation. This implies the influence of these compounds' photochemical formation and/or biogenic emissions. The elevated OVOC concentrations during the night confirm the influence of anthropogenic emissions.

Acetone and acetaldehyde are formed during photo-oxidation and ozonolysis of various terpenes and aromatic compounds emitted from multiple biogenic and anthropogenic sources (Lee et al., 2006a, b; S. Wang et al., 2020). Acetone plays a major role in ozone production. It can be transported to remote areas due to its long lifetime in the troposphere (∼ 15 d) (Seco et al., 2007). The average concentrations of acetone during winters, late winters, and summers were observed to be 13.6 ± 4.5, 15.3 ± 5.4, and 34.9 ± 10.3 ppbv, respectively. The observed concentrations of acetone in Lucknow are on the higher side of the range of measured concentrations in other Indian cities. The reported average concentrations of acetone in the present study are comparable to Delhi (whole year) at ∼ 16.7 ppbv (13–15 ppbv during winters) (Jain et al., 2022) but higher than Ahmedabad at 5.35 ± 1 ppbv during late winters (Sahu et al., 2016) and Mohali at 5.9 ± 3.7 ppbv during summers (Sinha et al., 2014). This shows the presence of more OVOCs in cities within the IGB region than in other cities of India.

Figures 4a and 3b show that the diurnal variation of acetaldehyde and acetone, respectively, starts increasing from 09:00 in the morning to 16:00 in the afternoon. The acetaldehyde and acetone had their morning maxima at around 10:00 and 11:00 LT, respectively, and later during the morning rush hours of vehicular emissions (08:00–10:00). This trend is similar to a previous study in Ahmedabad (Sahu and Saxena, 2015). It indicates the secondary formation of acetone and acetaldehyde from terpenes and aromatics emitted from vehicles. The OH reaction rate constant of the hydroxyl radical with acetaldehyde (15 × 10-12 cm3 molecule-1 s-1) is significantly higher than the reaction rate constant of the hydroxyl radical with acetone (0.17 × 10-12 cm3 molecule-1 s-1), indicating faster degradation of acetaldehyde than acetone (Atkinson and Arey, 2003). Figure 4c shows the diurnal variation of acetic acid and the acetic acid / benzene ratio. Acetic acid is one of the most abundant VOC species in the atmosphere globally, with a half-life of more than 1 d. It contributes to atmospheric acidity (Chebbi and Charlie, 1996) and is responsible for 30 % of the acidity of the wet deposition in polluted urban areas (Seco et al., 2007; Peña et al., 2002). This compound also has toxic effects on human health. Residential wood combustion is one of the critical sources of acetic acid (Bruns et al., 2017). High levels of acetic acid have also been reported from aged open biomass (hay and straw) burning plumes (Brilli et al., 2014), a variety of biomass fuels (Stockwell et al., 2015), and natural gas (Gilman et al., 2013). The average concentration of acetic acid during the whole study period is about 10.3 ± 4.1 ppbv, highest during winters at 15.2 ± 3.5 ppbv and lower during summers at 6.9 ± 2.1 ppbv. The observed increased concentrations of acetic acid in the winter and lower concentrations in the summer may demonstrate that its production is influenced by open biomass burning crops in nearby fields and residential wood combustion for heating and cooking purposes in Lucknow. The diurnal pattern of acetic acid concentrations shows high concentrations during the night, which implies its accumulation.

This section includes a discussion of the selection of the source apportionment solution and its interpretation. The NMVOC factors are identified based on their mass spectra, diurnal and temporal variation, and correlation with external tracers. For the first time, we have included mass spectra of 170 NMVOCs from m/z 42.034 to m/z 197.216 in the PMF analysis. The three abundant NMVOCs (compounds below m/z 42) detected by PTR-ToF-MS, acetaldehyde, acetone, and acetic acid, are not included in the PMF analysis. Including these NMVOCs in the PMF analysis resulted in biased solutions where only these ions are well-explained. Additionally, a few small alkane and alkene (C1–C4) compounds, which are not detected by PTR-ToF-MS, are excluded from the PMF analysis. However, previous studies have found that these ions are minor contributors to SOA formation. Included compounds (above m/z 42) are major contributors to SOA formation and dominant markers of various sources. As explained in Sect. 2.4, the optimum solution after the PMF analysis chosen is a five-factor solution. This selected five-factor PMF solution exhibits distinct mass spectral characteristics related to different sources and atmospheric processes. Figure 5 shows the intricate plots of the profile and the diurnal variation of the five-factor solution. The five factors are traffic, SFC 1 (solid fuel combustion), SVOCs, SFC 2, and VCPs (volatile chemical products) after thoroughly investigating markers, chemical species and their families, diurnal variation, and the relation to meteorological parameters and external measurements. The diurnal variation of the factors for two seasons (winter and summer) was compared, as shown in Fig. S7. The time series of the five factors resolved from NMVOC mass spectra are correlated with external measurements such as OOAs, BC concentrations, and CAAQMS data (WD, WS, RH, temperature, NO, NO2, NOx, SO2, O3), as given in Fig. 6.

Based on the method explained in Sect. 2.5, the ozone formation potential was calculated for each factor after considering the MIR values of NMVOC species as given in Table S2. The relative contribution of each NMVOC to the individual factor after PMF analysis is multiplied by the averaged individual concentration of the NMVOC species. The highest contributor species to the ozone formation potential is toluene, followed by xylene, isoprene, and methyl cyclohexene. The distribution of individual sources of OFP is shown in Fig. 10a. Toluene, xylene, and isoprene were found to be the highest contributors in terms of OFP in other Asian cities, including Guangzhou and Beijing (Zheng et al., 2009; Zhu et al., 2016; Zhan et al., 2021; Duan et al., 2008). In the previous study in Delhi, it was also noticed that the contributions of aromatics (xylene and toluene) have a substantial effect on the ozone formation potential (Tripathi et al., 2022). The traffic factor contributes the maximum to the OFP among all the factors with 34.6 %, followed by SFC 1 (23.9 %) and then SFC 2 (14.5 %), SVOCs (13.5 %), and VCPs (13.5 %). Similarly, the contribution to the formation of SOA is also estimated for each factor with the SOA yield of individual NMVOCs, as given in Table S3. The overall SOA yield is influenced by toluene, benzene, phenol, naphthalene, xylene, methyl furan, and trimethyl benzene. These compounds mostly belong to the aromatics and first-order oxygenate families. The measured SOA from the HR-ToF-AMS may be considered the sum of a more-oxidised oxygenated OA factor (AMS_MO-OOA) and one low-oxidised oxygenated OA (AMS_LO-OOA) factor (Lalchandani et al., 2021). The five highest contributors to SOA formation potential were correlated with the measured SOA in Fig. S10, and the high-resolution time series shows the co-occurrence of high and low peaks of benzene, toluene, and xylene with measured SOA during the day and night hours. This shows the significant role of aromatic NMVOC species in the formation of SOA. The primary factors, traffic and SFC 1, are the highest contributors to the SOA formation, with 28 % and 27 %, respectively, as shown in Fig. 10b. These factors are ridden with the highest SOA-formation-contributing NMVOC species. Previous studies have also found that aromatic hydrocarbons contributed more than 95 % to the SOA formation potential in other Asian cities (J. Qin et al., 2021; Zhan et al., 2021). It was observed that the sources related to vehicular emissions (diesel and petrol-driven vehicles), paddy stubble fire, and garbage fire emissions were the most contributing factors to ozone formation potential in Mohali (Kumar et al., 2020). In the present study, the SVOC factor contributes 22 % to the SOA formation, with the maximum contribution from high-molecular oxygenated species. SFC 2 and the VCPs are contributing less to the SOA formation.

In contrast, this sequence is not similar to the relative contributions of the sources according to their concentration (Fig. 10c). The source contributing to the highest concentration of NMVOCs is SFC 1, followed by traffic, SVOCs, and VCPs. The lowest contributor is SFC 2. This comparison shows the importance of the source of NMVOCs to SOA and ozone chemistry. The factor contributing the most to the concentration of NMVOCs may not necessarily influence the ozone and SOA formation similarly. These values estimate the potential for ozone and SOA formation and do not indicate the actual yields of ozone and SOA. This estimation method represents the complex behaviour of NMVOCs, NOx, and solar radiation for producing tropospheric ozone and SOA. There are many NMVOC species with unknown ozone and SOA yield values. One needs to understand the chemical fates and pathways of many NMVOCs by mimicking a real-time atmosphere in smog-chamber studies or through computational modelling studies. More research on this section is needed. Nonetheless, other parameters, including solar radiation and concentrations of oxides of nitrogen, also play a key role in the formation of ozone in the troposphere. In reality, OFP and SOA do not provide complete information about how VOCs influence O3, and organic aerosol chemistry zone formation in Lucknow is more sensitive to NMVOC concentrations than NOx, similar to other Asian cities. So, decreasing the VOC / NOx ratio would also help reduce the secondary pollutants (O3 and SOA). It is observed that vehicular emissions were the main source of aromatics (benzene, toluene, xylene). Therefore, vehicular emission control strategies should be implemented to reduce aromatic BTEX. Stringent implementation of policies and fuel-efficient standards related to vehicular emissions in Japan and South Korea have primarily improved the air quality (13 %–17 % reduction in NMVOCs) (Wang et al., 2014). In the present study, one of the key observations was that toluene is the main contributor to SOA and ozone production potential. This illustrates that targeting other sources of some NMVOCs (toluene and xylene) will also enhance their control. For example, paint solvents (sources of ethylbenzene and xylene) and printing products (sources of toluene) were targeted in a city, Hong Kong, where the VOC content of 172 types of consumer products was prescribed by the respective government (Lyu et al., 2017). In the present study, other potential contributor species are methyl cyclohexene (for ozone) and naphthalene (for SOA). These compounds are related to volatile chemical products, as found in the PMF analysis in Lucknow. This implies that stringent policies related to solvent-based industries such as textile, automobile, paint, and disinfectant are needed. Regulation and control of NMVOC content in the manufacturing and use of solvent-based products such as paints, disinfectants, fungicides, and insecticides should also be implemented. In China, various industries implemented end-of-pipe measures to control NMVOCs, such as refineries, plant oil extraction, gasoline storage and service stations, pharmacies, and crude oil storage and distribution (Wang et al., 2014). It is also estimated that China's end-of-pipe technologies and new energy-saving policies will help decrease about one-third of NMVOC emissions (Zhang et al., 2020). Phenols and furans were observed as some of the highest contributors to SOA formation potential related to solid fuel combustion. This suggests controlling solid fuel use for residential energy and crop-residue burning in the fields within and around the city of Lucknow. Firewood burning during the heating period and domestic in-field straw burning have substantially reduced emissions from biomass burning in China (Wu et al., 2020). Derwent et al. (2007) reported that reactivity-based VOC control measures might be more effective than mass-based regulations in controlling ozone and secondary organic aerosol formation. The present study also suggests that the reduction in VOCs, especially from vehicular emission, is needed for the abatement of O3 and SOA formation in urban areas.