Gridded distribution of total suspended particulate matter (TSP) and their chemical characterization over Delhi during winter

In the present study, total suspended particulate matter (TSP) samples were collected at 47 different sites (47 grids of 5 × 5 km2 area) of Delhi during winter (January–February 2019) in campaign mode. To understand the spatial variation of sources, TSP samples were analyzed for chemical compositions including carbonaceous species [organic carbon (OC), elemental carbon (EC), and water-soluble organic carbon (WSOC)], water-soluble total nitrogen (WSTN), water-soluble inorganic nitrogen (WSIN), polycyclic aromatic hydrocarbons (16 PAHs), water-soluble inorganic species (WSIS) (F−, Cl−, SO42−, NO2−, NO3−, PO43−, NH4+, Ca2+, Mg2+, Na+, and K+), and major and minor trace elements (B, Na, Mg, Al, P, S, Cl, K, Ca, Ti, Fe, Zn, Cr, Mn, Cu, As, Pd, F, and Ag). During the campaign, the maximum concentration of several components of TSP (996 μg/m3) was recorded at the Rana Pratap Bagh area, representing a pollution hotspot of Delhi. The maximum concentrations of PAHs were recorded at Udhyog Nagar, a region close to heavily loaded diesel vehicles, small rubber factories, and waste burning areas. Higher content of Cl− and Cl−/Na+ ratio (>1.7) suggests the presence of nonmarine anthropogenic sources of Cl− over Delhi. Minimum concentrations of OC, EC, WSOC, PAHs, and WSIS in TSP were observed at Kalkaji, representing the least polluted area in Delhi. Enrichment factor <5.0 at several locations and a significant correlation of Al with Mg, Fe, Ti, and Ca and C/N ratio indicated the abundance of mineral/crustal dust in TSP over Delhi. Principal component analysis (PCA) was also performed for the source apportionment of TSP, and extracted soil dust was found to be the major contributor to TSP, followed by biomass burning, open waste burning, secondary aerosol, and vehicular emissions.


Introduction
Globally, air pollution has become a serious problem, especially in metropolitan cities (Gurjar et al. 2004;Gerasopoulos et al. 2007;Balakrishnaiah et al. 2011;Bikkina et al. 2019). Total suspended particulate matter (TSP), the coarse mode pollutants, and their elemental compositions may cause serious effects on soil, vegetation, and crop and human health (Dockery et al. 1993;Ndiokwere 1984;David et al. 2017). TSP is dominated by dust and other vegetation activities (Air Quality Monitoring Network 2008); it includes airborne particles (solid and liquid) ranging from 0.005 to 100 μm in diameter (Khillare et al. 2004). Several studies on TSP have been conducted in various parts of the world to examine their influence on atmospheric chemistry as well as ambient air quality (Chow et al. 2002a;Christensen 2004;Salma et al. 2005;Arditsoglou and Samara 2005).
Developing countries like India experiences rapid growth in industrialization, urbanization, and human population which has increased the levels of pollutants in the atmosphere, declining its air quality and affecting human health and the environment (Goyal and Sidharth 2003;Li et al. 2019;Bond and Bergstrom 2006;Izhar et al. 2016). In India alone, half a million people are affected and die due to poor air quality (Jerrett 2015). With more persistence of atmospheric particulates and their capability to react with each other, it has become more hazardous (Goyal and Sidharth 2003). Delhi has experienced a growth in population at the rate of 3.3% and vehicles over 11 million in 2019 as per UN World Urbanization Prospects. Increasing urbanization over Delhi has made its people more vulnerable to aerosol air pollution. Delhi is now one of the most polluted cities in the world, and the total respirable suspended particulate reported in Delhi is highest among the other cities of India (Hsu and Zomer 2014). TSP concentration in Delhi has exceeded the Central pollution Control Board (CPCB), and World Health Organisation (WHO) prescribed limits (Guttikunda and Gurjar 2012).
TSP concentration varies according to the anthropogenic activities, natural sources, and meteorological conditions (relative humidity, solar radiation, wind speed, etc.) in the area. The influence of seasonal variations, mixing depth, and physical properties of Delhi plays a key role in determining the pollution load (Holzworth 1967;Dumka et al. 2016). The sources of the TSP decide the composition and its size distribution in the atmosphere (Contini et al. 2013). The suspended particulate has mineral dust, water-soluble organic, inorganic, and ionic species, heavy metals, organic and carbon, and sea salts compounds as its major components (Ram et al. 2012;Patel et al. 2020;Dumka et al. 2017). The identification of chemical constituents of TSP reflects the major sources and their associated processes affecting their physicochemical properties in the atmosphere (Perrino et al. 2009;Zhang et al. 2017). For source identification, the chemical characteristics of PM play an important role. Studies on potential sources of TSP in ambient air over Delhi were carried out by several researchers Jain 2007a, 2008;Hazarika et al. 2015;Shandilya et al. 2007); however, a comprehensive analysis of TSP is limited.
Delhi often faces uncontrolled pollution episodes during winter, and therefore, the regulatory bodies have intended to measure the criteria pollutants primarily in the categories area (residential, industrial, and traffic junction). Apart from this, due to increased population and commercial activities, areas of different types of sources of pollutants have been increased over Delhi. Although the number of observational sites over Delhi has been increased by different regulatory bodies, gridded data of pollutants over Delhi is missing. The studies of chemical characteristics of PM over the Delhi region are mostly based on single or few locations. Gridded data of pollutants over Delhi, which is distributed over 1484 km 2 , is limited.
In this study, the spatial variation in concentration of TSP has been estimated at 47 locations of Delhi (47 grids of 5×5 km 2 ) during January-February 2019. In addition, using the chemical components (OC, EC, WSOC, PAHs, WSIS, major and trace elements, etc.) of TSP, an effort has been made to determine location-specific sources of TSP in Delhi (Stewart et al. 2020). In the present study, principal component analysis (PCA) has been performed to investigate the potential source contributions to TSP in the ambient air of Delhi.

Study area
Delhi is one of the major metropolitan cities of India located between the Himalayas and Aravali range and situated in North India (28°12′-28°63′ N, 75°50′-77°23′ E) at the subtropical belt and altitude ranges between 213 and 305 m above sea level (asl). It is surrounded by Haryana on 3 sides and across the river Yamuna by Uttar Pradesh to the east and experiences a semi-arid climate. Delhi witnesses four distinct seasons (classified by Indian Meteorological Department, New Delhi): winter (January-February), summer or premonsoon (March-May), monsoon (June-September), and post-monsoon (October-December). The winter season with minimum temperatures of about 4°C is moderately cold and short. The smog phenomena in Delhi during winter have increased so much that it can illustrate the emissions and impacts of pollution not only in Delhi but also in other regions of the country by activities such as crop residue burning and wood-burning Kaskaoutis 2014). During winter, particulate matter (PM) affects the health of about 30% of the total population of Delhi than any other pollutant, causing serious respiratory and cardiovascular diseases (Gurjar et al. 2004). Emissions from different sources can be carcinogenic and hazardous for the atmosphere and human health. During the winter, atmospheric stability with conditions of low wind speed and temperature inversion lead to more atmospheric pollutants load in the lower layer of the atmosphere. Winter months are chilly (temperature:~2°C) and observe intense fog and haze. Summers are generally very hot and dry (temperature: 47°C) and observe frequent dust storms (soil and mineral dust). Due to low temperature, the high residence time of pollutant, low wind speed, precipitation, and low mixing height, the pollutant concentration is high during winters (Karar et al. 2006;Dumka et al. 2016).
Collection of samples TSP samples were collected using TSP samplers (highvolume sampler) on Pall quartz fiber filters (diameter: 11.0 cm) which were prebaked at 550°C for 6 h. Filters were desiccated for 24 h before and after the sampling for its conditioning. Quartz filters were pre-weighed using a microbalance (resolution: ±10 μg). Ambient air sampling was carried out at 1.5 m height above the ground level using a portable TSP sampler with a flow rate of 0.459 lpm for 4 h (daytime: 10:00-14:00 h) at 47 sites in Delhi during January-February 2019 (winter), in a campaign mode. After sampling, the filters were put in desiccators for 24 h and then weighed, and TSP concentration (in μg/m 3 ) was calculated. Quartz filters have been widely used to collect particulate matter for subsequent determination of carbon content by thermal optical analysis (Zhu et al. 2010(Zhu et al. , 2012Chen et al. 2010). However, the absorption of gaseous organics could occur during sampling because the quartz filters have a large surface area, which could lead to the overestimation of particulate organic carbon. On the other hand, the volatilization of particulate organic carbon from the filter would result in the underestimation of the particulate organic carbon. Nitrate can also give negative interference due to the dissociation of volatile ammonium nitrate. Semi-volatile organic compounds may also cause sample weight loss. To eliminate these uncertainties in OC/EC measurement and other chemical analyses, quartz back quartz (QBQ) filters were used during sampling apart from the field blank filters (Sharma et al. 2014).

Chemical analysis
Organic carbon and elemental carbon A small punch of 0.536 cm 2 of each filter was cut and analyzed for organic carbon (OC) and elemental carbon (EC) concentrations by using OC/EC carbon analyzer (Model: DRI 2001A, Make: Atmoslytic Inc., Calabasas, CA) using IMPROVE-A protocol (Chow et al. 2002b). OC/EC analyzer runs on a principle of preferential oxidation of OC and EC on which sample is heated to different temperature plateaus (140°C, 280°C, 480°C, 580°C, 740°C, and 840°C). OC is oxidized from the sample deposited on the filter in pure helium atmosphere at 140°C, 280°C, 480°C, and 580°C to estimate the OC components (OC1, OC2, OC3, and OC4); its function shows that organic carbon can be volatilized in a non-oxidizing helium atmosphere, whereas EC undergoes oxidation in a 98% helium and 2% oxygen at 580°C, 740°C, and 840°C to estimate the EC components (EC1, EC2, and EC3) that relies on the fact that elemental carbon must be combusted by an oxidizer. The main function of the optical component of the analyzer is for the correction of pyrolysis and carbonizes OC compounds into EC. Calibration of the OC/EC analyzer was performed by 4.8% of CO 2 + balance He gas along with known amounts of KHP (potassium hydrogen phthalate) and sucrose solution (Sharma et al. 2015). Each sample was analyzed in triplicate to estimate the repeatability error (estimated as 4-7%).

Water-soluble inorganic species
The sample filters were (2 × 2 cm 2 ) cut into smaller pieces and then extracted in deionized water (resistivity: 18.2 MΩ-cm) for 45 min, three times (45*3 = 135 min) in ultrasonicator. The extract was filtered using nylon membrane filters (size: ) and cations (Na + , NH 4 + , K + , Mg 2+ , and Ca 2+ ) were determined by using Ion Chromatograph (Dionex-ICS-3000, USA) under suppressed conductivity. Solution for mobile phase used for anions 20mM NaOH as an eluent and cations were analyzed by 5mM MSA (methane sulphonic acid) as eluent was prepared for the analysis. Anions were determined by using an IonPac-AS11-HC analytical column (4 × 250 mm; Dionex, USA) with a guard column (IonPac-AS11-HC, 4 × 50 mm; Dionex, USA), ASRS-300, 4 mm anion micro-membrane suppressor. Cations were determined by using a separation column of IonPac CS-17-HC (4 × 250 mm; Dionex, USA) with a guard column IonPac CS-17-HC (4 × 50 mm; Dionex, USA) with a suppressor CSRS-300 (4 mm; Dionex, USA). The calibration curve was displayed on the screen for each analytical run, and the chromatogram was processed by Chromeleon software. QA/QC of water-soluble inorganic species (WSIS) analysis of TSP samples has been assured by calibrating each ionic species with a known standard value (standard line ) were also analyzed, and concentration in each sample was corrected for their respective average blank concentrations (Sharma et al. 2014).
Water-soluble organic carbon and water-soluble organic nitrogen For WSOC, a circular punch of 7.065 cm 2 area from each sample was extracted in 20 ml of deionized water (specific resistance: 18.2 MΩ-cm) using an ultrasonicator for 30 min (10 × 3 = 30 min). The concentrations of water-soluble organic carbon (WSOC) and water-soluble total nitrogen (WSTN) were measured using a total organic carbon analyzer (TOC, Shimadzu, model TOC-L CPH) and total nitrogen monitor (Shimadzu, model TNM-L ROHS, assembled with TOC-L), respectively. The WSOC was measured as non-purgeable organic carbon (NPOC), where the samples were acidified by 5% HCl and spurge for 1.0 min) (zero air as sparging gas, flow: 80 ml/min) to remove the inorganic carbon (IC) fraction from the solution in the TOC analyzer (Rastogi et al. 2015). The IC-free solution was then transferred to the combustion column (heated at 680°C), where it was oxidized to form CO 2 via oxidation by Pt-Al 2 O 3 catalyst. This CO 2 was then detected by a non-dispersive infrared (NDIR) detector. The calibration curve was prepared using laboratory-made multiple standards from 1000 mg/L stock solution of potassium hydrogen phthalate (KHP). Patel et al. (2020) have described in detail the estimation of WSOC and WSON in particulates samples.
The nitrogenous compounds were decomposed and oxidized to nitrogen monoxide NO in the combustion tube (680°C) filled with Pt-Al 2 O 3 catalyst. The carrier gas carries NO to the TNM-L system, where it is measured for WSTN in the chemiluminescence detector. Furthermore, the 1000 mg/L stock solution of KNO 3 was diluted and run in the same way as the sample to obtain the calibration curve. Multiple blank filter punches were also analyzed for WSOC and WSTN concentration, and their respective concentration was subtracted from each sample. Furthermore, the concentration of WSON was estimated by taking the difference between WSTN and water-soluble inorganic nitrogen (WSIN). The N mass fraction from NH 4 + , NO 2 − , and NO 3 − was considered as WSIN There is no analytical method to measure directly the concentration of WSON. WSON is a comparatively small fraction of watersoluble total nitrogen (WSTN) (Rastogi et al. 2011).

Polar organic compounds
The sample filters were extracted twice with 15 ml dichloromethane (DCM) by using an ultrasonicator for 15 min. The extracted sample was then transferred to a rotatory flask to reduce the sample to 1 ml using a rotary evaporator at 30-40°C. The extracted samples were then filtered by a membrane filter (PVDF; size: 0.5 μm micro-syringe) and transferred into the vials. The sample was stored in a deep freezer at −20°C for further analysis. Extracted sample was then analyzed for organic compounds by using a gas chromatograph (GC) (Shimadzu, GC-2010 Plus) system which is equipped with an Rtx-5 molecular sieve column (0.25 μm thickness). The system maintained at 50°C for 5 min initially, then increased at the rate of 5°C/min-150°C held for 5 min, then 8°C/min-250°C held for 10 min and then 10°C/min-300°C for 5 min. The organic compounds were identified based on their retention time with standards. Three different classes of organic compounds were identified by the following standards: for n-alkanes

Trace elements
The quantitative analysis of elements (Na, Mg, Al, Fe, P, S, Cl, K, Ti, Ca, As, Pd, F, and Ag) in TSP samples was performed by using Wavelength Dispersive X-ray Fluorescence Spectrometer (WD-XRF); 2.0 cm diameter size of quartz filter (area: 3.14 cm 2 ) was used for the analysis of major and trace elements in TSP by WD-XRF (Rigaku ZSX Primus). The XRF spectrometer has components such as a sealed X-ray tube for excitation, scintillation counter (heavy element detector), flow proportional counter (light element detector), end window, and Rh target. The analysis was carried out at 36°C temperature with a 2.4 kW tube rating under vacuum conditions. ZSX software (Rigaku Corporation, Japan) was used for quantitative analysis and data procurement by parameter method. The instrument WD-XRF was calibrated periodically using micro-matter thin-film standards (Watson et al. 1999).
After the elemental analysis using X-ray fluorescence, the same filters were re-analyzed by inductively coupled plasmaoptical emission spectrometry (ICP-OES) for verification of concentrations of elements (B, Zn, Cr, Mn, Cu, Cd, Pb, Co, Ni, Bi, Li, Sr, and Ga) in TSP. Filters were extracted and transferred to a PTFE vessel filled with 5 ml concentration of HNO 3 , 10 ml concentration of 5% of HNO 3 , and 2 ml concentration of H 2 O 2 in the digestion tube for 50 min. After that, it was allowed to cool properly, and 10 ml of 5% of HNO 3 was added; then, it was allowed to stand for about 30 min. The solution was filtered by Whatman filter and syringe filter (0.42 μm) and makeup to a volume of 25 ml and analyzed using ICP-OES.

Principle component analysis
Principal component analysis (PCA), a multivariate statistical tool, can identify patterns in data and disclose information about differences and similarities. In PCA, assuming a linear relationship between the total mass concentration and the contributions of each species, PCA factorized the data in numerous steps (Song et al. 2006). PCA was performed for TSP data over Delhi to identify possible sources. Principle components to be retained were decided by examining the screed plot and % of the variance. The eigenvalue for extracted factor was taken more than 1. First, dimensionless standardized form has been transformed from chemical composition data.
where i = 1, … n samples; j = 1 … m elements; C ij is the concentration of elements j in sample i; and Cj and σ j are the arithmetic mean concentration and the standard deviation for element j, respectively. The PCA model is expressed as where k = 1, … p sources; g ik and h kj are the factor loading and the factor score, respectively. The equation helped in solving the eigenvector decomposition (Song et al. 2006). APCS method is used later based on factor loading scores of PCA, which estimates source contribution and source profile quantitatively (Thurson and Spengler 1985;Henry and Hidy 1979). Since PCA results are normalized for the data, thus the factor for true zero is derived as APCS is also known as rescaled scores and further linear regression can be derived from the following equation, where Mi is the measured mass concentration in sample i and ζ 0 denotes the mass contribution in the PCA made by the sources uncounted. APCS ki is the rotated absolute component score for source k in sample i, and ζkAPCS ki is the mass contribution in sample i made by source k. This follows the regression for the sample concentrations on the APCS which g e t s e a c h i d e n t i f i e d s o u r c e ' s e s t i m a t e d m a s s contribution (Song et al. 2006).

Results and discussion
The concentrations of TSP were estimated at the 47 sites of Delhi during January-February 2019, and their chemical components were also analyzed to assess the ambient air quality of Delhi as well as possible sources of TSP over the region. The results of these analyses are discussed in the subsequent subsection. . RP Bagh, a residential and nearby slum area with very high vehicular activities and a huge jagari industries (a product of sugarcane), may be a potential source of OC, whereas Karala village in the region of many food processing industries, different agriculture practices followed by open biomass burning, vehicular emissions, and dust resuspension. Mean OC mass has contributed approximately 9% to TSP with large variability (12-33%) over Delhi (see Supplementary Figure S3a). EC concentration over Delhi has varied from 5.8 to 47.1 μg m −3 (average: 15.8 ± 8.3 μg m −3 ) (Fig. 2b). The EC mass also showed a large spatial variation (2-8%) with a mean contribution of 5% to TSP. Ram and Sarin (2010) have reported higher EC concentration in the rural area (Jaduguda: 11.6 ± 2.0 μg m −3 ) than urban sites (Allahabad: 8.1 ± 1.7 μg m −3 ; Kanpur 6.2 ± 3.7 μg m −3 ; and Hisar: 8.  NCT-Delhi, in preparation). In addition to these two sites, some of the sites showed high variability, particularly in the north and east Delhi (E8 and F10 in Supplementary  Table S1). Very low concentration of EC was found at few locations of southwest districts (grid F5, F6, and G5 in Supplementary Table S1) and southeast districts (F3, G5, G3 in Supplementary Table S1) where OC concentration was also low. Using TSP collected from September 2006 to April 2007, Miyazaki et al. (2009) have estimated the mean OC concentration of the order of 68.4 ± 37.9 μg m −3 over Delhi. Figure 2c represents the spatial map of total carbonaceous aerosol (TCAs = 1.6 * OC + EC) over Delhi. TCAs have varied from 32.9 to 283.7 μg m −3 (average = 112.5 ± 56.0 μg m −3 ). Two potential hotspots are noticed at Rana Pratap Bagh (283.7 μg m −3 ) and Karala village (252.2 μg m −3 ), respectively. Ram and Sarin (2010) have shown that carbonaceous aerosol accounts for 30-35% of TSP at urban and rural sites of northern India, whereas in the present study, the mean contribution of total carbonaceous aerosol was 23-26% of TSP (see Supplementary Figure S3a). Sharma et al. (2018a, b) have shown a lower contribution of total carbonaceous aerosol to PM 2.5 (29%) and PM 10 (24%) except winter when the contribution was 29%. Similar studies Sharma et al. 2014;Bisht et al. 2015;Gupta et al. 2018;Sharma et al. 2018a, b;Gadi et al. 2019;Dumka et al. 2017) have reported a higher contribution of carbonaceous aerosol to PM. Spatial variation in carbonaceous aerosol over Delhi may be due to the variation of emission from the vehicle and local industries, biomass burning apart from common factors like lower boundary layer, advection of pollutants from Punjab and Haryana from agricultural residue burning site to the observed sites. The mass closer for organic matter (OM) alone accounts for approximately 18% which is higher than OC (9%) (see Supplementary Figure S4). Some of the OM comes from domestic heating and cooking purpose (Perrino et al. 2011) and fuel-burning (Chan et al. 1997). There can be a significant amount of primary biogenic particles in TSP, e.g., pollen, plant debris, and fungi. EC mainly comes from combustion sources including exhaust of vehicle-driven sources which contributed 2% of the total TSP concentration. Figure 2d represents the spatial map of WSOC over Delhi. Like OC and EC, WSOC also showed large variation (10fold) with the lowest concentration at Aram Bagh (4.5 μg m −3 ) and the highest at Rana Pratap Bagh (43.8 μg m −3 ). The mean concentration of WSOC over Delhi was 17.6 ± 8.1 μg m −3 . WSOC contributes 29% of OC over Delhi (see Supplementary Figure S5). When WSOC is less than 50% of the total OC, it means more insoluble carbon is dominant at all the locations of Delhi. More than 51% WSOC of OC concentration was only observed at Majnu Ka Tila, a tourist bus stand  Figure S6). Ram and Sarin (2010) have shown that the contribution of WSOC concentration to OC was more at a rural site (45%) than urban sites (35%) and stronger dependence of WSOC on OC (R 2 = 0.86) at urban sites. Miyazaki et al. (2009) reported a similar range of WSOC (8-55 μg m −3 ) in TSP samples over Delhi during winter, contributing 31 ± 11% of OC during the study period. Water-insoluble inorganic carbon (WSIC) (OC-WSOC) ranges from 12.5 to 104.1 μg m −3 with an average value of 42.8 ± 22.3 μg m −3 . WSIC normally comes from fossil fuel combustion, whereas the variability in the WSOC concentration at various points might be due to a combination of various factors like varying emission sources, sources strength, SOA formations, and meteorological conditions. WSOC may be used as a measure of SOA. Biomass burning also contributes to primary WSOC and secondary WSOC (by emitting VOCs forming SOA; Rastogi et al. 2015).

Mass concentration
In the present study, the OC/EC ratio has varied from 2.1 to 5.9 (average 3.9 ± 1.0) (in Supplementary Table S3). Spatial variation of OC/EC ratio has been given in Supplementary Figure S7a. Ram and Sarin (2010) have reported higher variability in OC/EC ratio (2.4-14.5) in TSP at urban sites of northern India, indicating dominant contribution from biomass burning sources (wood-fuel and agriculture waste). Sharma et al. (2014) have reported OC/EC ratio in the range of 3.8-5.8 (average 4.38 ± 2.36) in PM 10 over Delhi. A lower OC/EC ratio indicates either the absence of secondary organic aerosol (SOA) over Delhi or the dominance of fossil fuel combustion sources. Concentrations of OC and EC are the culmination of combustion of fossil fuel and biomass fuel; OC/EC ratio may indicate their competitive contributions. The highest OC/EC ratio at Karala village indicates the possibility of enhanced biomass burning as well as SOA formation. Bhowmik et al. (2020) have reported OC/EC ratio for the two in Delhi sites (5.9 ± 3.4, 5.9 ± 4.8) in PM 2.5 during winter 2018. An average OC/EC ratio of 6.6 was reported for biomass combustion by Saarikoski et al. (2008) and 7.3 for wood-burning emissions in particular (Sandradewi et al. 2008), whereas a low OC/EC ratio of 1.1 was found for traffic emissions (Sandradewi et al. 2008). Schauer et al. (2002) have also reported OC/EC values of 1.0-4.2 and 16.8-40.0 for diesel/gasoline-powered vehicular exhausts and wood combustion. In a new inventory for on-road vehicular emission, Jaiprakash and Habib (2018) have also proposed an OC/EC ratio (0.45-30) for different types of fuels (diesel, gasoline, CNG) used in vehicles of India. Venkataraman et al. (2005) have reported OC/BC ratio in the range 0.28-9.09 for few biofuels widely used in India. Saud et al. (2012) have reported OC/EC ratio in the range of 2.3-5.68 for different types of residential fuels (fuelwood: 2.3, crop residue: 3.68, and dung cake: 5.01) used over Delhi. The scatter plot between OC and EC concentrations over Delhi shows a significant correlation (R 2 = 0.74, p < 0.05), indicating their origin from common sources (see Supplementary Figure S6). Spatial variability of OC/EC ratio over Delhi indicates heterogeneity in the contribution of emission of biomass fuels and fossil fuels, biogenic emission, and/or secondary aerosol formation.
WSOC/OC ratio indicates the photochemical activity and/ or aging of the aerosols during the transport (Pio et al. 2007). WSOC is either produced by a gas-phase reaction from volatile organic carbon Pio et al. 2007;Weber et al. 2007) and biomass burning and/or vehicular emissions (Saarikoski et al. 2008). In the present study, WSOC/OC ratio has varied from 0.10 to 0.51 at the sites of Delhi (see in Supplementary Table S3 and Supplementary Figure S7b). The mean WSOC/OC ratio is 0.30 ± 0.10 and comparable to that reported by Miyazaki et al. (2009) over Delhi. They reported day (0.37 ± 0.09) and night (0.25 ± 0.09) variation of WSOC/OC ratios in TSP and found photochemical production SOA over Delhi. In the present study, frequency distribution of WSOC/OC ratio shows Gaussian spread, i.e., 0.10-0.20 (6 sites), 0.20-0.29 (14 sites), 0.29-0.39 (14 sites), 0.39-0.48 (11 sites), and 0.48-0.51 (2 sites) (see Supplementary Figure S8). Such a pattern suggests the variability in emission sources, their strength, and contribution from SOA at sites of Delhi. Lower WSOC/OC ratios could be due to the poor solubility of organic constituents from the combustion of liquid fuels (diesel, gasoline, etc.) in water. Ram and Sarin (2010) reported the variation of the average WSOC/OC ratio in TSP from 0.32 to 0.79 for the sampling sites in the IGP. The higher contribution of WSOC to OC can be explained by enhanced photochemical production of polar compounds. Higher OC concentration and lower WSOC/OC ratio at Rana Pratap Bagh indicate the presence of more advected aged aerosols in addition to local sources. In contrast, enhanced photochemical production of polar compounds in the vehicular emission at Majnu Ka Tila, a tourist bus stand along the NH-1, making the highest WSOC/OC ratio (0.51) and the possible formation of SOA (see Supplementary Figure S7b). Similar to the present study, Kondo et al. (2007) have also reported that 35% of OC was water-soluble at an urban location in Tokyo, Japan. Higher WSOC/OC ratio (see Supplementary Figure S7b) at Karala village supports the SOA formation as observed expected by higher OC/EC ratio. Jian et al. (2005) had reported that average WSOC/OC ratios were 0.30 at an urban location (Nanjing). Ram et al. (2012) have reported a large seasonal variation of WSOC/OC ratio of (0.21-0.70) at Kanpur. Saarikoski et al. (2008) have reported a value of 0.27 for vehicular emissions over an urban environment in Helsinki (northern Europe). An earlier study by Cheung et al. (2009) had reported the variation of WSOC/OC ratios from 0.06 to 0.19 in the diesel particles emitted from light-duty vehicles. Wu et al. (2019) have reported a higher WSOC/OC ratio (0.68 ± 0.40) at a Himalayan site of Nepal, indicating a higher water-soluble component of organic aerosol to OC and the possible formation of SOA in winter. From the above discussion, it has been found that fossil fuel burning and biomass fuels burning leading SOA have contributed to carboncontaining aerosols over Delhi.

Water-soluble ionic species in TSP
Water-soluble ionic species (WSIS) have contributed 19% of TSP (see Supplementary Figure S3). Most abundant watersoluble species are Ca 2+ (33%, , and NH 4 + (5%) in WSIS over Delhi (see Supplementary Figure S9a). Most of the watersoluble species showed maxima at Rana Pratap Bagh, the site of the highest TSP. The trend for the mean concentration of WSIS in Delhi was as follows: The concentration for cation species ranges from 33.1 μg m −3 at Vasant Kunj to 236.2 μg m −3 at Majnu a Tila. In an earlier section, it has been observed that a higher WSOC/OC (0.51) ratio at Majnu Ki Tila indicated the photochemical activity, leading to the formation of SOA; in contrast, a lower WSOC/ OC ratio at Rana Pratap Bagh supported advection of aged aerosol. A higher concentration of WSIS at Majnu Ki Tila supports the formation of SOA aerosol as compared to Rana Pratap Bagh. The lowest concentration of WSIS was found at several locations of Delhi (Surakhpur, Baprola, Kalkaji, Aram Bagh, and Bakhtawarpur). The ratios of the sum of equivalent concentrations of anions to cations (A/C) were used as an indicator to analyze the acidity of the different environments (Shen et al. 2009). During the study period, the A/C ratio varied between 0.28 and 0.78, with an average of 0.47 at our site, indicating an alkaline environment (see Supplementary  Figure 9b).
Primary WSIS (Na + , K + , Ca +2 , Mg +2 , and Cl − ) Figure 3a-e shows the spatial map of primary water-soluble ionic species over Delhi. Primary WSIS constitutes 67% of total water-soluble ionic species (see Supplementary Figure S12). Spatial variation of Cl − concentration was from 7.5 to 66.6 μg m −3 . Higher Cl − concentration at Rana Pratap Bagh (66.6 μg m −3 ) is followed by Manglapuri (49.3 μg m −3 ), Zafrabad metro (22.2 μg m −3 ), and Govindpuri (15.9 μg m −3 ) (Fig. 3a). These sites are characterized by highly anthropogenic activities (see Supplementary Table S3). Cl − is found in normally coarser particles, and its formation might be from the mechanisms, i.e., sea salt transformation and photochemical reaction (Chattopadhyay 2010), biomass burning , and use of fertilizers (Bakeer et al. 2016). In addition to open waste burning waste to energy (WTE) plants installed at Sukhdev Vihar-Okhla, Ghazipur, and Bawana, plastic uses municipal solid waste (MSW) as input. The polyvinyl chloride (PVC) plastics content of MSW is a significant contributor to emissions of chlorinated dibenzo-p-dioxins (dioxins) that are associated with MSW combustion (Shepherd 1993). Gunthe et al. (2021) have argued that the high local concentration of gas-phase hydrochloric acid, possibly emitted from plastic-contained waste burning and industry, has contributed to enhanced Cl − concentration over Delhi. Since sites are far from the sea, photochemical reaction, open waste burning could be the major sources along with the coal combustion in nearby slums area. Saxena et al. (2017) found 20.1 μg m −3 of Cl − in PM 10 ; in the present study, the mean concentration of Cl − in TSP was 27.2 μg m −3 .
Normally, the main source of K + is considered as soil, but fine particles of K + may be released into the atmosphere through the burning of plant materials (leaves, vegetable straw) (Cooper 1980;Saud et al. 2013;Sharma et al. 2016b). K + showed high concentration at Nizampur village (11.7 μg m −3 ), Rana Pratap Bagh (9.7 μg m −3 ), Trilokpuri (8.3 μg m −3 ), and Karala village (7.5 μg m −3 ), and the major sources might be responsible for biomass burning emissions and road dust (Fig. 3c). A high concentration of K + was found near the industrial region since Nizampur village is situated near Bahadurgarh which is an industrial region. The concentration for anion throughout Delhi ranges from 31.6 μg m −3 at Mahavir enclave to 217.3 μg m −3 at Majnu ka Tila. Higher K + /Na + (0.12-3.03) in the present study than seawater K + /Na + ratio (0.036) rules out the sea salt as a source over Delhi (see Supplementary Table S4). Correlation between K + and EC (R 2 : 0.27), K + and OC (R 2 : 0.4) confirms the role of biomass burning and fossil fuel burning over the study area (see Supplementary Figure S10 study by Goel et al. (2020) has given the possibility of NaCl and KCl as predominant compounds, i.e., crustal sources in the ambient air of Delhi (see Supplementary Figure S10). The nss-K + /EC ratio is indicative of fossil fuel burning, whereas K + /Cl − ratio and Cl − /Na + ratio provides the information of closure to the location of sources of Cl − . Spatial variation of nss-K + /EC ratio, K + /Cl − ratio, and Cl − /Na + ratio suggests the role of various location-specific sources of fossil fuel and primary sources, respectively (see Supplementary Figure S11). Mg 2+ showed high spatial variability (0.96-13.44 μg m −3 ) over Delhi. The highest concentration of Mg+ has been observed at Rana Pratap Bagh, followed by Dakshinpuri (11.1 μg m − 3 ) and Raghubir Nagar (10.1 μg m −3 ) (Fig. 3d). Construction dust is the major source of Mg + . Mg + /Na + ratio (0.22-1.54) higher than seawater Mg + /Na + ratio (0.12) rules out the role of sea salt in ambient air of Delhi (see Supplementary  Table S4). Correlation between Mg + with OC (R 2 : 0.5) suggests soil dust as a source (see Supplementary Figure S12). The high concentration for Ca 2+ commonly showed at Rana Pratap Bagh (153.9 μg m −3 ) (Fig. 3e) as we already know it exposed to high traffic congestion, dust, and municipality waste burning near the location. Correlation between Ca + and OC (R 2 : 0.3) suggests soil dust as a source. A strong correlation between Ca + and Mg + (R 2 : 0.61) suggests crustal as a common source. Heavy vehicular load, burning of coal nearby slum area, boilers in the small factories, and local biomass combustion might have contributed to F − concentration at Rana Pratap Bagh (1.1 μg m −3 ), followed by Shastri Bazaar (1.0 μg m −3 ) and Raghubir Nagar (0.7 μg m −3 ) (Fig. 3f Figure S9). Figure 4a-e shows the spatial map of secondary water-soluble inorganic ions over Delhi. SO 2 , NO x , NH 3 , and PO 4 3− are precursors of SO 4 2− , (NO 3 − + NO 2 − ), NH 4 + , and PO 4 3− , respectively, under suitable conditions.
In a megacity like Delhi, the possibility of dieseldriven automobiles, fossil fuel emissions, small and medium industries, and thermal power plants in and around Delhi-NCR (Suneja et al. 2020) might be responsible for the emission of SO 2 which might have been converted to SO 4 2− depending upon their prevailing meteorological conditions. SO 2 also release from conventional fuels like coal and furnace oils (Chattopadhyay 2010). Suneja et al. (2020) have reported the long-term trend of SO 2 in ambient air over Delhi, showing winter maxima and also concluded that the thermal power plants were one of the major local sources of SO 2 over the Delhi region. Under the colder atmospheric condition, other processes such as metal-catalyzed oxidation of SO 2 , the aqueous phase of H 2 O 2 /O 3 oxidation of SO 2 could be major pathways of SO 4 2− formation (Seinfeld and Pandis 2006;Collett et al. 1990). The combination of all the sources leads to the highest concentration of SO 4 2− (Fig. 4a) (Fig. 4c). The mass ratios of NO 3 − /nss-SO 4 2− higher than unity indicated that the sources of particles were mainly from mobile sources. Conversely, the mass ratios of NO 3 − /nss-SO 4 2− lower than unity suggested that the sources of particles mainly came from stationary sources (Li et al. 2016). In the present study, NO 3 − /nss-SO 4 2− ratio less than unity suggests the role of stationary sources. High spatial variation of NO 3 − /nss-SO 4 2− ratio more than unity in several locations of Delhi indicates the vehicular emission as one of the source TSP (see Supplementary Figure S13).
The highest concentration of NH 4 + was reported at Rana Pratap Bagh (12.6 μg m − 3 ), followed by Trilokpuri (12.5 μg m −3 Poothkhurd (12.4 μg m −3 ) and Aram Bagh (10.1 μg m −3 ) (Fig. 4d). Biomass burning, traffic congestion, and industrial emissions at all these three sites could be sources of NH 4 + . During the winter months, secondary organic aerosol formed from acidic gases mostly emitted from fossil fuel and biomass fuel combustion sources could be possible sources . The minimum concentration is found over South East Delhi. Soil dust, secondary inorganic formation, biomass burning, and fossil fuels burning could be possible sources of WSIS in TSP. The spatial distribution of PO 4 3− (Fig. 4e) shows the highest concentration at Mandawali (5.1 μg m −3 ).
A good correlation between WSIN and WSTN (R 2 = 0.64) suggests that in the significant number of samples, WSTN is evenly contributed by the sum of NO 3 − , NO 2 − , and NH 4 + (see Supplementary Figure S17). Asian soils contain 40% of nitrogen which might turn into water-soluble nitrate, as suggested by Kawamura et al. (2004). Therefore, the soil could be one of the major sources of water-soluble nitrate. In addition, sources of inorganic ionic species are mostly from gas precursors of the atmosphere (Seinfeld and Pandis 1998). Being a food basket of India, IGP uses a lot of nitrogenous fertilizers which are converted to NH 4 + . Vehicular emission contains a large amount of NH 3 which eventually converted to NH 4 + and contributes NH 4 + -N (Sharma et al. 2016a). A very poor correlation between TC and TN (R 2 < 0.09) resembles the observation of Hegde and Kawamura (2017) over the Thumba region of Trivandrum. Agnihotri et al. (2011) (Duan et al. 2009). WSON is originated from a variety of sources, including both natural and anthropogenic. Anthropogenic may include agricultural activities, fuel combustion, biomass burning, and industrial activity, and natural sources are mineral dust, algal blooms, and natural biomass burning (Mace et al. 2003;Münevver and Koçak 2018). Laskin et al. (2009) have suggested slow smoldering burning that leads to incomplete combustion of pyrolysis products as responsible for the presence of a large number of nitrogen-containing organic carbon species in the biomass burning aerosol samples.
WSON shows no correlation with OC when data of the 47 sites are considered (see Supplementary Figure S18). This could indicate that sources of WSON and OC are not common except a few. To identify which site represents a better correlation between WSON and OC, WSON/OC ratio has been calculated and it shows a large variation from 0.01 to 0.40. WSON/OC ratio has been divided into three ranges, i.e., 0.01-0.10, 0.10-0.20, and >0.20. Yu et al. (2017) have reported WSON/OC ratio between 0.30-0.51 over a megacity of south China. A stronger significant correlation (0.65) is observed between WSON and OC at 13 sites of Delhi (see Supplementary Table S5) when WSON/OC is within the range of 0.10-0.20 and as compared to the other two ranges 0.01-0.10 (R 2 = 0.40) and >0.20 (R 2 = 0.32). OC is mainly originated from the incomplete combustion of fuels (biomass, biofuel, municipality waste, etc.) and fossil fuels (coal, charcoal, etc.). Sites with a WSON/OC ratio of 0.10-0.20 have common sources which might be biomass burning and fossil fuel combustion. At other sites, sources of OC do not match with that of WSON.
Data of EC and WSON over 47 locations results in no good correlation between EC and WSON. EC and WSON at all the stations may not have common sources over Delhi except a few. WSON/EC ratio varied from 0.02 to 1.75 (see Supplementary Table S6). To identify which site represents a better correlation between WSON and EC, WSON/EC ratio has been considered and divided into three ranges, i.e., 0.02-0.29, 0.30-1.0, and above 1.0. Correlation (R 2 ) between WSON and EC, for WSON/EC ratios in the ranges of 0.30 to 1.0 and above 1.0 is 0.67 and 0.63, respectively (see Supplementary Figure S19). Only at those sites having a higher WSON/EC ratio (0.30-1.0 and above) have common sources of WSON and EC.
K + is an important tracer of biomass burning as OC. WSON does not show any good correlation with K+ when the whole data set of 47 sites are considered. To identify which site represents a better correlation between WSON and OC accordingly, data of K+ has been divided into three sets based on WSON/OC ratio. When 0.10<WSON/OC>0.20, correlation (R 2 = 0.22) between WSON and K + (see Supplementary Figure S20a Figure S20).
Total carbon and total nitrogen ratio have been calculated with available data. In the present study, the mean total C/N ratio was 17 ± 9 (4-40) (see Supplementary Table S7). The C/ N values are the important parameter for indicating soil or biological sources. Duan et al. (2009) reported the C/N ratio of the order of 31.5 (14.6-57.8). They have suggested that when TSP concentrations >400 μg m −3 and the C/N ra-tios<30, the soil would be the important factor resulting in heavy pollution in wintertime in Beijing. In the present study, apart from 4 sites, C/N values of <30 for 43 sites suggest soil organics such as humic acid as important sources of WSON. Soil dust, biomass fuels burning, and fossil burning are sources of nitrogen-containing aerosol in TSP.

Major and trace elements in TSP
Out of 27 trace species, Al is the most abundant species (25.4 ± 26 μg m −3 , 26%) followed by Ca (17.7 ± 18.3 μg m −3 ,18%), Fe (11.02 ± 7.09 μg m −3 , 11%), Mg (7.8 ± 7.9 μg m −3 , 8%), S (6.7 ± 5.5 μg m −3 , 7%), Na (6.7 ± 5.9 μg m −3 , 7%), K (5.4 ± 5.4 μg m −3 , 5%), and Cl (4.8 ± 5.1 μg m −3 , 5%) in total metal concentration in TSP (Fig. 6). These 8 metal species constitute 87% of total metal concentration; whereas, the other 19 metal species constitute 13% of metal concentration in TSP (see Supplementary Figure S22). The percentage contribution of trace metals is approximately 34% of TSP concentration, whereas Kumar et al. (2015) have reported large differences in the contribution of metal species to TSP at urban sites (31%) and open waste burning sites (8%). Kumar et al. (2018)   respectively. This indicates the spatial heterogeneity of sources of metals over Delhi. Other studies (Khemani et al. 1985;Mehra et al. 1998;Balachandran et al. 2000;Gadi et al. 2000;Anju and Banerjee 2003;Srivastav et al. 2003;Khillare et al. 2004;Monkkonen et al. 2004;Srivastava and Jain 2005;Yadav and Rajamani 2006;Srivastava and Jain 2007a, 2007b, 2007cSrivastava et al. 2009) have reported concentration of metal species at few selected sites of Delhi. The enrichment factors (EFs) determine the difference in natural (crustal) and anthropogenic sources. When EF is >5, sources are anthropogenic. The distribution pattern of EFs of metal species at 47 sites shows that maximum metals are mostly of crustal origin except a few, where EF of Zn, Cr, Cu, Pb, Co, and Li have crossed 5, indicating other anthropogenic sources (see Supplementary Figure S23a,b). Among all the metal species, B shows the highest EFs at most of the sites in Delhi. Duan and Tan (2013) have summarized the major sources of some of the heavy metals (Cu, Zn, Cr, Pb, etc.) in ambient air. Sharma et al. (2020) have reported a similar pattern at a station in central Delhi. In the present study, Ca, Mg, Ti, and Fe have a significant correlation with Al and indicate mineral dust as the major source. Ca shows a good correlation with Al (R 2 = 0.92), indicating crustal dust as the dominant source in the particulate matter. The significant linear correlation of Al with Fe (R 2 = 0.92), Al with Ti (R 2 = 0.49), and Al with Mg (R 2 = 0.95) indicated the dominant source of mineral dust in Delhi (see Supplementary Figure S24).
For the first time, a high enrichment factor of Li (16.5) at Aram Bagh is reported. Very large usage of lithium batteries in e-vehicles in Delhi might have triggered high enrichment in ambient air. Zhang and Zhou (2013) have reported electrolyte degradation, Li dendrite formation, and parasitic reactions with H 2 O and CO 2 which are all directly correlated to reversibility and cyclability of Li-air batteries when they are operated in ambient air. The lifetime of Li-air cell has sustained repeated cycling in ambient air for 100 cycles, i.e., 78 days. A very low concentration of .3 ng m −3 , average: 14.8 ± 12.0 ng m −3 ) is reported over Delhi, with the highest value at Aram Bagh. Xia and Gao (2011) have reported that sources of Cu are vehicle emission and smelting furnace burning. High EF is noticed at Aram Bagh (8.5), Jhilmil Colony (13.2), and Govindpuri (6.2), due to anthropogenic activities like heavy vehicular emission, construction activities (Govindpuri), industrial activities (Jhilmil colony), and traffic congestion, open waste burning (Aram Bagh) Steel, plastics, and pigments production (Li et al. 2012), contaminated soil (Sun et al. 2009), coal-fired boiler and furnace burning (Tian et al. 2010), and lead gasoline (Yang et al. 2003) are reported as sources of Pb. High EF of Pb was only noticed at Jhilmil Colony, where industrial activities could be a source of Pb. Several steps have been taken in removing Pb in gasoline. Low concentration of Pb (0.02-2.60 μg m −3 , average: 0.26 ± 0.41 μg m −3 ) as noticed over Delhi might be due to lower vehicular emission.
Coal and oil combustion (Tian et al. 2010;Pacyna et al. 2007) and waste incineration residue (Astrup et al. 2005;Pacyna et al. 2007;Font et al. 2015), paints, and the number of metalworking plants (Astrup et al. 2005;Unceta et al. 2010;Font et al. 2015) could be possible sources of Cr. In the present study, spatial variation of Cr concentration ranges from 0.06 to 6.41 μg m −3 (average: 1.26 ± 1.36 μg m −3 ) over Delhi. Kumar et al. (2018) have reported a lower concentration of 0.27 ± 0.07 μg m −3 at the NPL site and 0.14 ± 0.06 μg m −3 at the hospital site of Delhi in winter. Several sites of Delhi (Bhogal: 22.65; Rana Pratap Bagh: 18.7; Punjabi Bagh: 13.55; Jhilmil colony: 11.2; Nizampur village: 6.22; Ranhola and Neelwal village: 5.7) show EF of Cr higher than 5 attributing the anthropogenic activities as sources of Cr (see Supplementary Figure 23a). The most common sources of all these sources are vehicular emission, biomass burning, open waste burning, and industrial activities. Using the above discussion, it can be identified that crustal dust, biomass burning, and fossil fuels burning are possible sources of major and trace metals in TSP.
Mass closer for geological minerals (Al, K, Ca, Ti, Fe) analyzed by ED-XRF and trace elements (Na, Mg, P, S, K by ED-XRF and Zn, Cr, Mn, Cu, Cd, Pb, Co, Ni, Bi, Li, Sr, Ga by ICP-OES) accounted for 13.9% and 4.2% of TSP. These represent mineral oxides and metal oxides. The reconstructed mass of measured species is found around 69%. The remaining unidentified mass (32%) is possibly present as water, particle-bound water, insoluble acids, etc. (see Supplementary Figure S4).
The highest concentration of PAHs can be found in the urban area due to high vehicular activities and less dispersion of pollutants (Sharma et al. 2016a). The average concentration of PAHs was found at 667.73 ± 399.38 ng m −3 in Sharma et al. (2016b), and they also explained that the concentration of PAHs is higher during winter than monsoon and summer. Earlier studies also revealed the same (Halsall et al. 1994;Harrison et al. 1996;Panther et al. 1999;Park et al. 2002;Guo et al. 2003). Sharma et al. (2017) have also reported high values of PAHs (672 ng m −3 ) during winter.
Incomplete combustion of organic materials is the main source of emissions of PAHs by use of wood and biomass burning, oil, coal, and gas (WHO 2000;Wang et al. 2007;Tobiszewski and Namieśnik 2012;Alebić-Juretić 2015).
Identified PAHs are further classified into four classes based on their aromatic ring structure, 2-3 rings, 4 rings, 5 rings, and 6 rings. Five rings PAHs had contributed 52% to total PAHs concentration followed by six rings (42%), four rings (4%), and two-three rings (2%), respectively, over Delhi sampling locations (Fig. 7). It is clear from these values that 5-6 rings PAHs, which are considered to be carcinogenic, are dominant over the Delhi region.
High molecular weight compounds are more persistent and less volatile and can travel a long distance than other lower molecular weight compounds.  Table 2). Observance of the highest concentration of high molecular weight PAHs corroborated that carcinogenic PAHs were dominant in Delhi. 5-6 rings PAHs present in the least concentration were Naphthalene and Acenaphthylene, and their minimum and maximum ranges were 2.3-10.8 ng m − 3 and 1.4-7.8 ng m −3 , respectively.
For the identification of possible pollution sources of PAHs in the atmosphere, the molecular diagnostic ratios (MDRs) were calculated and compared with values available in the literature. In this study, for source identification, we have considered five major molecular diagnostic ratios, i.e., IP/(IP+ BghiP), BaP/(BaP+Chy), Phth (Phth+Anth), BaP/BghiP, and IP/BghiP (Table 2). All the diagnostic molecular ratios are indicative of different PAH origin sources. The benzopyrenes and chrysene are the major PAHs compounds associated with combustion sources (Kavouras et al. 1999). The mean IP/(IP+ BghiP) ratio (0.6 ± 0.2) strongly indicates coal and wood combustion sources. The BaP/(BaP+Chy) ratio (0.7 ± 0.2) indicates that PAHs are also derived from gasoline emissions. The IP/BghiP ratio (1.6 ± 0.9) shows that diesel emissions were dominant. The Phth/(Phth + Anth) ratio differentiates between biomass burning and fossil fuel emissions. The higher Phth/(Phth+Anth) ratio (0.6 ± 0.2) indicates biomass burning. This may be due to the biofuels burning in the villages and slum areas. The BaP/BghiP ratio (0.8 ± 0.2) indicates that PAHs are from traffic emissions. These source diagnostic ratios conveyed that PAHs have both vehicular as well as biomass burning emissions as dominant sources. Gadi et al. (2019) and Shivani et al. (2019) have reported similar results for fine ambient aerosols.
In a nutshell, qualitatively, it can be said that different types of sources, i.e., soil dust, biomass burning, fossil fuels burning (industrial activities, vehicular emission, open waste burning, etc.), and secondary organic and inorganic formation, have contributed to spatial variation in TSP concentration over Delhi based on their strength. For quantitative identification, a more robust statistical analysis is required. In the present , OC, EC, and WSOC) by the varimax-rotated factor matrix method of SPSS statistics viewer; 24 TSP constituents were used as variables in the dataset (Table 3), and some components (P, S, B, Ag, Pd, Ga, F, As, NO 2 − ) were excluded. Six components were extracted as principal components that have 82.1 % of the total variance of the data. Principle components to be retained were decided by examining the screen plot and percent of the variance. PCs that have an eigenvalue of >1 divided the data into groups and were considered for further analysis for source identification (Hazarika et al. 2015;Shivani et al. 2018).
Factor 1 (PC1) contained high loading of Mg, Fe, K, Al, Na, Cl, Ca, Ti, and OC with 32.4% of variance which indicates the major contribution of crustal/soil dust as these are soil-related crustal elements Srivastava and Jain 2007a;Pant and Harrison 2012;Hazarika et al. 2015) and occurrence of Fe with OC marked for road dust (Banerjee et al. 2015). Crustal/soil resuspension plays a major role in the elemental profile for coarse as well as in fine particles. In the winter season, soil dust and mining effect is the most dominant source in coarse particles like TSP due to the temperature inversion effect (Hazarika et al. 2015). Emissions of some of the elements such as Na, Cl, K, Al, and Zn were also linked with other anthropogenic activities (resuspension of soil particles, vehicular emissions, and metals used in industrial activities) (Hazarika and Srivastava 2016). Factor 2 (PC2) source was identified with high loading of OC, WSOC, and EC attributed to biomass burning with 16.9% of the variance. EC is marked extensively for combustion sources and also indicates minor chemical transformations (Song et al. 2006;Yin et al. 2010). On the other hand, OC is emitted directly from primary emission sources, i.e., combustion and vaporization of solvent (Turpin and Huntzicker 1991;Ho et al. 2003;Behera and Sharma 2010). Factor 3 (PC3) accounted for 12.3% of total variance with high loading of Pb, Zn, Bi, Cr with Cu indicating open waste burning, and Cu, Pb, and Zn are responsible for non-ferrous metal operations (Arditsoglou et al. 2005). Factor 4 (PC4) is responsible for 7.7% of total variance contained high loading of NO 3 − , SO 4 2− , and NH 4 + which indicates secondary inorganic aerosol (high loading of secondary nitrate and secondary sulphate). The product of secondary inorganic aerosol formed in the atmosphere can be emitted by natural or anthropogenic sources (Jain et al. 2020). Precursors of secondary NO 3 − are nitric oxide (NO) which is generated as a result of high-temperature combustion (Khan et al. 2010;Sharma et al. 2020). Factor 5 (PC5) is dominated by Cr and Zn with EC indicated the significant contribution to fuel-burning/vehicle exhaust, and it explained with 7.4% of the variance (Srivastav and Jain 2005). Factor 6 (PC6) is responsible for 5.4% of total variance with high loading of Co, Ti, and OC which is difficult to be explained and is assigned as unidentified. For a better understanding of possible sources in the mixed contribution, PCA was also performed for polar organics.

PCA of polar organics
PCA was performed for combined data of OC, EC, and 16 PAHs to identify major possible sources; 16 PAHs, OC, and EC constituents were used as variables in this data set. Table 4 summarizes the output of PCA for organics in TSP during winter over Delhi. PC1 has explained a total of 14% of the variance with high loading of Acth(82%), Fl(52%), Nph(37%), and IP(36%) which indicated the emission of PAHs from biomass burning in the ambient atmosphere ). PC2 has contained high loading of pyr(64%), BaF(56%), and BaP(64%) with 12% of the total variance. The dominance of these PAHs indicated the contribution of secondary organic carbon, vehicular activities (diesel dominated), and coal-burning to a high level over Delhi (Agarwal et al. 2009;Gupta et al. 2011). PC3 has contained high loading of Phth(84%), Flth(61%), BaA, Chy with 12% of the total variance and suggested the contribution from plastic and waste burning to observed PAHs over Delhi. Simoneit et al. (2005) also accounted for the contribution of open burning of roadside litter and landfill trash for total PAHs emissions. PC4 witnessed high loading of EC (80%), OC (78%), BkF, BbF with 12% of total variance which indicates the contribution of vehicular emissions and biofuel combustion (Guo et al. 2003;Gupta et al. 2011;Masih et al. 2012). BkF dominance indicated the emissions from the diesel vehicles relative to other PAHs (Venkataraman et al. 1994). The factor dominated by EC shows fossil fuel burning/vehicle exhaust emissions . PC5 explained high loading of DahA(80%), BghiP(78%), BaA(12%), BkF(13%) with 8% of total variance which indicates the contribution of cooking emissions, coal combustion, and vehicular emissions (Harrison et al. 1996;Gupta et al. 2011;Sarkar et al. 2010). PC6 showed high loading of Acy (77%), Fl (23%), Anth (13%) with 8% of the total variance, indicating space heating as the major reason. Space heating included the burning of wood and cow dung. Hazarika et al. (2015) have reported that the coal tar use in ongoing road repairing activity could be responsible for emissions of PAHs. In a nutshell, it can be considered that biomass burning (fuelwood and cow dung), secondary inorganic aerosol, vehicular emissions, plastic, and waste burning, cooking emissions,

Conclusion
Large variations of TSP at 47 sites of Delhi measured during winter (January-February 2019) suggest the variation in sources types and their strength. The chemical analysis (OC, EC, WSOC, TN, WSIS, and PAHs) of TSP and a robust statistical analysis (PCA) have provided the following qualitative and quantitative location-specific information of sources as given below.
& The maximum concentration of OC, EC, and WSOC was observed at Rana Pratap Bagh, a site with the highest TSP. The OC/EC ratio and good correlation between OC and EC suggest that sources are common and the mixture of biomass burning and fossil fuel burning. WSOC contributes 29% to OC, and a lower WSOC/OC ratio indicates the role of advected aged aerosols in addition to local sources. In contrast, a higher WSOC/OC ratio (0.51) at Majnu Ki Tila, a tourist bus stand, suggests the photochemical activity of freshly released organic compounds and the possible formation of SOA. Delhi being a megacity, vehicular emission might have contributed to fossil fuel combustion. & Primary WSIS (Cl − + Na + + K + + Ca +2 + Mg +2 + F − ) contributes 67% to WSIS with maximum concentration at Rana Pratap Bagh. Cl − /Na + ratio, K + /Na + , Mg + /Na + ratios rule out the presence of sea salt at the study sites and suggesting the role of biomass burning, open waste burning, and industrial emission. Secondary WSIS (SO 4 2− + NO 3 − + NO 2 − + NH 4 + +PO 4 3− ) contributes 33% to WSIS attributing lesser role of secondary inorganic formation. Minimum concentration was observed in southeast and southwest Delhi. Unprecedented higher concentration of Cl − all over Delhi suggests possible roles of plastic burning, municipality waste burning, and industry emission. & Nitrogen components contribute 2-3% of TSP; the highest concentration of WSIN and WSON was observed at Rawta Mor (21.7 μg m −3 ) and Majnu Ka Tila (21.6 μg m −3 ), situated on the highway with heavy traffic load in Delhi, respectively. The highest WSOC concentration was also observed at Majnu Ki Tila and strongly supported the hypothesis of SOA formation. A stronger significant correlation (0.65) is observed between WSON and OC at 13 sites of Delhi when WSON/OC is within the range of 0.10-0.20, attributing the sources of WSON and OC are common. When the C/N ratio is <30 and the TSP is >400 μg m −3 , then crustal soil is one of the sources. Out of 47 sites, 41 sites indicate the contribution of crustal to TSP. Perylene) with a minimum-maximum range of 5-130 ng m −3 , 4.9-115.1 ng m −3 , and 6.4-155.3 ng m −3 . High molecular weight compounds (5 rings and 6 rings) were observed at Udhognagar metro (240 ng m −3 and 299 ng m −3 ) due to high combustion activities and plastic factories around the location, and these species are highly dangerous even in nanogram quantity. & The analysis obtains the mass closer to approximately 68% of the total suspended particulates which determined the estimated sources: inorganics (25%), salts (9%), geological minerals (11%), trace elements (5%), and a sum of organic matter-EC (18%). The unidentified mass is about 31% which cannot be measured. OM represented the presence of biomass burning (cooking, space heating, and waste burning), exhaust gases, and oil refineries. During the winter period (study period), the formation of secondary reactions is enhanced which contributed to inorganic ions. & In a nutshell, qualitatively, it can be said that different types of sources, i.e., soil dust, biomass burning, fossil fuels burning (industrial activities, vehicular emission, open waste burning, etc.), secondary organic, and inorganic formation have contributed to spatial variation in TSP concentration over Delhi based on their strength. For quantitative identification, a more robust statistical analysis is required. In the present study, PCA has been performed using measured TSP concentration and its chemical properties. & PCA was performed for 24 constituents of TSP which identified four principal components with 82.1% of the total variance of the data. Soil dust/crustal with 32.38% of variance indicated major contribution followed by biomass burning (16.9%). Further analysis of PCA performed for the clarification of mixed sources, identified six principle components with 68% variance with biomass burning (14%), secondary organic carbon (12.5%), plastic and waste burning (12%), vehicular emissions (12%), cooking emissions (8.5%), and room heating (8.5%) are the major contributors over Delhi.
Our spatial gridded distribution of TSP concentration is among the first available with its chemical composition estimation. The significant outcome of this study is the identification of variation in source contribution at 5 × 5 km 2 grids covering entire Delhi, including rural and urban regions. Trajectory analysis suggests that apart from local sources, long-range transport from Haryana and Punjab carried along with the north-eastern low-level flow to the receptor sites could be another source during the winter.
Acknowledgements The authors are thankful to the Director, CSIR-NPL, New Delhi, and Head, Environmental Sciences and Biomedical Metrology Division, CSIR-NPL, New Delhi, India, for their constant encouragement and support. The authors are also thankful to the Ministry of Earth Sciences for the financial support (MoES/16/19/2017-APHH) (DelhiFlux).
Availability of data and materials All data generated or analyzed during this study are included in this published article and its supplementary information files Author contribution RJ has collected samples, done chemical analysis of organics, WSOC, metal using ED-XRF, and lead in the manuscript preparation. SA has collected samples and done chemical analysis of organics, WSOC, and nitrogen component and prepared extraction for ions. GK, RA, AM, LY, PY, NC, MR, RB, AR, and USS have collected samples over Delhi and contributed to the manuscript. NR and AP have assisted in WSOC and TN and contributed to the manuscript. Shivani and RG have assisted in the analysis of PAH and contributed to the manuscript. PS has analyzed metal using ICP-OEC and contributed to the manuscript. NV has assisted in metal analysis using ED-XRF and contributed to the manuscript. CS has contributed to the manuscript. SKS has analyzed OC, EC data and performed PCA and contributed to the draft of the manuscript. TKM has given conception and design of the study, guidance, analyzed the WSIS data, and contributed to the draft of the manuscript. All the authors read and approved the final manuscript.

Declarations
Ethics approval and consent to participate Not applicable

Consent for publication Not applicable
Competing interests The authors declare no competing interests