Introduction

Atmospheric particulate matter (PM) is a complex mixture of organic and inorganic substances suspended in the air and has significant impacts on the environment, and in turn on human health1,2,3,4,5,6,7. Therefore, evaluation and understanding of PM sources are essential for the development of effective strategies to mitigate their adverse effects. Carbonaceous aerosols (CA) are an important component of PM playing a major role in perturbing Earth’s radiative balance and thereby affecting atmospheric chemistry and climate at both long- and short-term scales8,9,10,11,12. Carbonaceous aerosols can constitute a major fraction (~ 30–70%) of total aerosol mass depending on the location and emission sources and therefore can significantly affect the local climate and air quality13. These aerosols exhibit varying chemical, physical and optical properties, and are primarily present in continental and urban tropospheric environments. Identifying and understanding their sources remains a significant challenge14,15,16,17. Globally, approximately 75% of CA is contributed by South and East Asian regions because of an increase in anthropogenic emissions over the past few decades related to rapid development18,19,20,21,22. Particularly, the concentration of CA in the Indian subcontinent in the last few decades has increased significantly due to a variety of factors related to economic growth23,24,25, which in turn has affected atmospheric chemistry and become a major concern of deteriorating human health in the region26,27,28. The typical sources of CA in this region include biomass/bio-fuel combustion, vehicular emissions, construction/road dust, secondary PM formation, and industrial emissions18,19. Compared to other regions, emissions from biomass (post-harvest crop residue) burning and coal combustion (thermal power plants and brick kilns) are the major contributors of CA in the Indian subcontinent19,27. The subcomponents of CA are organic carbon (OC) and elemental carbon (EC) or black carbon (BC) which together constitute the total carbon (TC) fraction of ambient aerosols; EC or BC nomenclature is based on the method of determination by thermo-optical or optical properties, respectively. These components (OC and EC) possess distinct light absorption properties and have the potential to affect the Earth’s radiative balance29,30,31,32. The EC fraction typically originates due to incomplete combustion of fossil fuel or biomass. The TC fraction also contains numerous polar and non-polar compounds such as phenols, isoprene, humic acids, polycyclic aromatic hydrocarbons (PaHs), and carbonyls etc., which play different roles in the ambient atmospheric chemistry depending on existing meteorological conditions33,34,35. However, the real impact of carbonaceous aerosols on climate and health is still ambiguous due to the existing uncertainties related to their sources, chemical composition and properties, and formation mechanisms.

Traditional chemical analyses performed on filter-based PM, such as ionic species abundances determined by ion chromatography, OC and EC concentrations by carbon analyzer, and PM mass by gravimetry provide information on pollution levels and likely sources36,37. In the past decade, several studies have demonstrated the potential of stable carbon (δ13C) and nitrogen (δ15N) isotope composition of aerosols to trace the nature of pollution sources and their contribution to PM, and to understand the reaction mechanisms related to the secondary formation and atmospheric aging and their control on the atmospheric chemistry, etc38,39,40,41,42,43,44,45,46,47,48. The aerosol research community has extensively used the δ13C of PM to shed light on the sources and transformations of emissions49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64. Several authors have attempted to use carbon isotopic composition for the characterization of biomass-burning products65,66,67,68,69,70,71,72,73. Furthermore, δ13C of various phases (gas and particulate) has been used to understand transformation mechanisms (gas to particle phase) and coupled with 14C/12C ratio measurements (Δ14C; modern biomass/biofuel sources have a value > 0‰ due to bomb-spike-origin 14C in late 1950s because of nuclear weapon testing, whereas fossil fuel sources have a value of -1000‰) to quantify the contribution of fossil and non-fossil sources to the ambient PM19,56,74,75,76,77,78. The aerosol transformation processes from the original source might lead to isotopic fractionation (kinetic or equilibrium) resulting in a slightly different isotopic composition of the emitted aerosols than its original source39,43,55,79,80 (Fig. 1). For example, chamber-based experiments by Irei et al. (2006) showed that secondary organic aerosol (SOA) generated by OH-radical reactions with toluene resulted in about 0.6 ± 0.2‰ lower δ13C value of SOA than the gas phase products. On the other hand, combustion typically results in about 1.3 ± 0.5‰ lower δ13C in the exhaust gases relative to the liquid fossil fuels whereas δ13C value does not change during coal burning38. During long-range-transport of aerosols, photochemical aging could result in up to 1.5-3‰ enrichment in 13C in remaining aerosols compared to their sources40,53. Nevertheless, the final isotopic composition of the sampled aerosol carries information on the primary sources as well as the transformation processes such as secondary formation and aging47,81,82,83. However, the deconvolution of the final isotopic composition of PM to determine the primary sources requires that the isotopic composition of sources must be distinct, and the associated isotopic fractionations are well known.

Fig. 1
figure 1

Schematic showing stable carbon isotopic fractionation and related change in the δ13C value of aerosols (↑ indicate 13C-enrichment and ↓ indicate 13C-depletion) through various stages of aerosol emission, formation, and atmospheric processing38,39,40,45,93,97,100.

Full size image

In this work, we have evaluated the carbon isotopic variability in heavily polluted (urban) and pristine sites (background) in India and assessed the use of δ13C in characterizing and understanding the origin of carbonaceous aerosols. We determined δ13C variations in PM2.5 aerosols from eight strategic sampling locations of the National Carbonaceous Aerosols Program – CarbOnaceous AerosoL Emissions, Source apportionment & ClimatE Impacts (NCAP-COALESCE) network of the government of India. In this network, the sampling stations have been selected to assess the influence of anthropogenic pollutants, particularly carbonaceous aerosols, on climate change24. One of our primary objectives is to evaluate whether δ13C of aerosols can be used as a standalone proxy for quantitative source apportion estimates so that users are aware of the advantages and pitfalls of this proxy, which has been increasingly utilized in characterizing source signature of CA. The novelty of our study lies in utilizing seasonal δ13C datasets from multiple sites in India characterized by diverse local climate/environment and the Keeling plot to evaluate the site-specific source dominance.

Procedure and methods

Sampling sites

Eight sampling sites spanning across eight states of India were selected based on different criteria such as emission inventory, wind trajectories, dispersion modeling results, and cross-correlation (r) and coefficient of divergence (COD) analyses. These analyses aided in identifying sampling locations for overall regionally representative aerosols, i.e., without the influence of local pollution sources84. The sampling locations are (Fig. 2): Kashmir (KMR, northern Himalayan region in Jammu and Kashmir, 34.13 ⁰N, 74.84 ⁰E), Mohali (MOH, upwind Indo-Gangetic Plains in Punjab, 30.67 ⁰N, 76.73 ⁰E), Rohtak (RTK, upwind Indo-Gangetic Plains in Haryana, 28.88 ⁰N, 76.62 ⁰E), Bikaner (BKN, western arid desert region in Rajasthan (28.03 ⁰N, 73.26 ⁰E), Bhopal (BPL, central India in Madhya Pradesh, 23.28 ⁰N, 77.27 ⁰E), Shyamnagar (SHY, downwind Indo-Gangetic Plains in West Bengal, 22.50 ⁰N, 88.23 ⁰E), Jorhat (JOR, North-East India in Assam, 26.73 ⁰N, 94.15 ⁰E), and Mysuru (MYS, background site in southern India in Karnataka, 12.31⁰N, 76.61 ⁰E). Figure 2 presents the sampling locations along with population density, and the presence of power plants and cement industries, which are likely sources of anthropogenic CA emissions. The KMR sampling site is located at 1500 amsl in a semi-urban setting within the University of Kashmir campus in the Kashmir Himalaya region, characterized by a temperate climate. The MOH sampling site is located in an urban environment within the IISER Mohali campus in the southern part of Chandigarh city in Punjab, north-west India, dominated by monsoonal climate. The BKN sampling site is located within the Maharaja Ganga Singh University campus, which is in the Thar desert region (dry and arid) of Rajasthan in northwestern India. The SHY site was within the Bose Institute campus in Shyamnagar, West Bengal, India, an urban setting with a prevailing humid subtropical climate. The JOR site is located at CSIR-Northeast Institute of Science and Technology, Council of Scientific and Industrial Research (CSIR-NIEST) campus, Assam near the Brahmaputra River in a rural environment. The MYS site is in Karnataka, south-west India, characterized by a tropical savanna climate, and considered a background site due to minimal influence from emission-based sources. The RTK site is in the state of Haryana about 70 km north-west of New Delhi, characterized by a subtropical monsoon climate. The BPL sampling site is within the IISER Bhopal campus, Madhya Pradesh, Central India, having a subtropical climate characterized by hot summer and mild winter.

Fig. 2
figure 2

Map of India showing the NCAP-COALESCE sampling locations (KMR, MOH, RTK, BKN, JOR, BKN, SHY, and MYS). Also shown are the areas likely to have higher anthropogenic emissions of CA, (1) population distribution per square kilometer in greyscale with densely populated regions in red (data from: (1) https://data.worldpop.org/GIS/Population_Density, accessed on January 15, 2025), (2) major powerplants consuming biomass, coal, gas, and oil114, (3) cement industries115.

Full size image

Sampling details

Aerosol samples (PM2.5) were collected on (prebaked at 550 ⁰C) Quartz-fiber filters (47 mm diameter) from January to December 2019 at eight sampling sites in India on every alternate day. A speciation air sampling system (SASS, Met One Instruments Inc., USA) with five channels was utilized for collecting 24-h time-integrated samples using a mass flow controller unit operated at ~ 6.7 m3 h− 1 flow rate. Every two weeks, field blank samples were also collected to account for sampling artifacts and perform background correction. Sampling details have been provided in previous studies from the NCAP-COALESCE network85,86,87,88,89,90,91,92.

Analytical procedure

The stable carbon isotopic composition (δ13C value) of aerosols was determined at the stable isotope analytical facility of the Indian Institute of Technology Kanpur (IITK) using an elemental analyzer (Flash EA 2000, Thermo Scientific) coupled with a stable isotope ratio mass spectrometer (IRMS; Thermo Scientific Delta V Plus) in a continuous-flow mode. The detailed analytical procedure was given in our previous studies45,48,93. Briefly, about 500–2000 µg of the sample was taken from a quartz filter, packed in air-free tight tin capsules and dropped into an elemental analyzer, resulting in the combustion of the sample and generation of CO2 in stoichiometrically equivalent amount of carbon present in the sample. The generated CO2, after online purification and removal of moisture, was transferred to the IRMS via a ConFlow IV interface, and its 13C/12C ratio was measured. All isotopic data are represented in δ notation on the VPDB scale (Vienna Pee Dee Belemnite) as:

$$\delta ^{{13}} C_{{sample}} \left( {\permille;{\text{ }}VPDB} \right){\text{ }} = {\text{ }}[\left( {^{{13}} C/^{{12}} C} \right)_{{sample~~}} /{\text{ }}\left( {^{{13}} C/^{{12}} C} \right)_{{VPDB}} - 1]{\text{ }}x{\text{ }}10^{3}$$

A set of carbon isotope reference standards (LSVEC: ‒46.6‰; NBS-19: 1.95‰; and IAEA-600: ‒27.711‰) from the International Atomic Energy Agency (IAEA) were analyzed along with each set of unknown samples to convert raw carbon isotope ratios to the VPDB scale using the normalization procedure described in Paul et al.94. Analytical precision based on repeat analyses of a check standard (IAEA-601: ‒28.81‰) was ‒28.83 ± 0.04‰ (n = 28). In total 1120 samples have been analyzed along with an additional 32 repeat analyses. The analytical procedure for the measurement of total carbon (TC = OC + EC) is described in the supplementary section.

Spatial distribution maps of δ13C

δ13C values obtained from this study (see Table 1) and reported by others (see Table 2) for different Indian sites were used to create aerosol δ13C spatial variability maps for each season, using GIS technique. Spatial interpolation was performed using the Inverse Distance Weighted (IDW) method available in the Geostatistical Analyst toolbar of ArcGIS®. IDW interpolation, a commonly employed technique for mapping variables, offers an exact and convex method that fits continuous spatial variation models. It relies on distances to weight known locations and uses these values to estimate those at new points95,96.

Table 1 δ13C values in different seasons at various NCAP locations.
Full size table
Table 2 δ13C values in different seasons at various sites in India from previous studies.
Full size table

Results and discussion

Annual variation of δ 13C at the sampling sites

The δ13C and TC of PM2.5 exhibit significant variability among NCAP-COALESCE sampling sites along with showing seasonal variability at specific sites (Figs. 3 and 4, and S1, Table 1 and Table S1). The highest and lowest δ13C values of −16.0‰ (14th July 2019) and −29.4‰ (6th September 2019) were recorded at the Bikaner and Shyamnagar sites, having a low TC content of ~ 0.8 µg m− 3 at both the sites (Fig. 3, Fig. S1). The overall average δ13C for samples analyzed in this study is −26.6 ± 0.9‰ (1σ). The overall variabilities at individual sites, that is the difference between the maximum and minimum δ13C values in increasing order are: Jorhat (1.9‰), Mysuru (2.5‰), Kashmir (2.9‰), Bhopal (3.4‰), Rohtak (3.7‰), Mohali (3.8‰), Shyamnagar (4.0‰), and Bikaner (12.2‰). Jorhat (Assam) located in North-East India showed the least variation in δ13C (< 2‰) among all the sites, whereas Bikaner showed the largest variation of ~ 12‰ (Fig. 4). Barring Kashmir and Jorhat, every other sampling site shows relatively 13C-rich (higher δ13C values) signature in the winter season (Table 1). Figure 5 shows distinct seasonal δ13C spatial variability. The overall average TC content in our samples is 14.8 ± 17.4 µg m− 3, with the highest TC content of ~ 75 µg m− 3 in post-monsoon at the RTK site and the lowest value of ~ 1.5 µg m− 3 in monsoon at BKN and MYS background sites (Table S1). TC contents in our samples strongly correlate with the aerosol mass concentrations. Seasonal variations in the TC content are significant and generally are higher during winter across most sites with RTK and KMR recording elevated levels (Fig. 3b, Fig. S1; Table S1). Interestingly, Rohtak and Mohali located in upwind IGP have significantly higher TC during post-monsoon, which is dominated by biomass burning in this region (Figs. S1 and S2). Monsoon season consistently shows the lowest TC values, likely due to enhanced wet deposition. Site-wise discussion of δ13C in PM2.5 is provided next.

Fig.3
figure 3

(a) Annual variation of δ13C in carbonaceous (PM2.5) aerosols at different NCAP-COALESCE sites. Horizontal dotted line is the average δ13C and the shaded region (light blue) represents the standard deviation. 3(b) Annual variation of TC in PM2.5 aerosols at different NCAP-COALESCE sites. RTK: Rohtak; MYS: Mysuru; MOH: Mohali; BPL: Bhopal; BKN: Bikaner; JOR: Jorhat; KMR: Kashmir; SHY: Shyamnagar.

Full size image
Fig. 4
figure 4

Plot showing annual median (box plot of variabilities) δ13C at the NCAP sampling locations along with δ13C of sources from literature: (a) range of fossil fuel combustion from −22.6‰ to −27.2‰100, and (b) C3 biomass combustion from −24.2‰ to −30.1‰45,93. RTK: Rohtak; MYS: Mysuru; MOH: Mohali; BPL: Bhopal; BKN: Bikaner; JOR: Jorhat; KMR: Kashmir; SHY: Shyamnagar.

Full size image
Fig. 5
figure 5

Spatial distribution map showing δ13C variability in India in different seasons. Data generated in this study (Table 1) and from the literature (Table 2) were used to prepare these maps.

Full size image

Bhopal

The δ13C of PM2.5 in these samples ranges from −28.1‰ to −24.7‰ (−26.6 ± 0.6‰) (Figs. 3a and 4)88. An average δ13C of ~ −27.0‰ in pre-monsoon and monsoon (Table 1) indicates the contribution of C3 biomass and/or wood combustion; δ13C of C3 plants from India, like globally (−20‰ to −32‰)97, varies from −24.2‰ to −30.1‰92. δ13C of PM in pre-monsoon and monsoon at Bhopal are similar to those observed in the same seasons at other Indian locations [e.g., Beas (pre-monsoon, PM2.5): −27.4 ± 0.5‰45; Jodhpur (summer and monsoon, PM10): −27.5‰98 (Table 2). On the other hand, the δ13C in the winter season varies from −27.0‰ to −24.7‰ (−26.0 ± 0.5‰). The average δ13C in winter is very similar to those observed in another IGP location in Kolkata in winter (δ13C of PM10 of −26.0‰99, Table 2). δ13C variation in winter at Bhopal suggests an additional contribution from a 13C-rich source, such as the fossil fuel sources including gasoline/diesel/biodiesel vehicular exhaust (−22.6 to −27.293,100). Photochemical aging in aerosols during long-range transport also results in 13C enrichment leading to a higher δ13C value (up to ~ 3.3‰ for water-soluble organic carbon52) in remaining aerosols47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101. Yadav et al. (2022)88 interpreted the isotopic variability in conjunction with the secondary organic carbon (SOC) contributions in both seasons and identified two separate mechanisms. According to them, high SOC/OC ratios in monsoon suggest an influence of SOC formation in the aqueous phase. Through aqueous phase chemistry, the highly water-soluble volatile organic compounds (VOCs) are transformed to the particle phase SOC with lower δ13C than precursors because of kinetic fractionation39,92,102. While in post-monsoon and winter, a relative increase in δ13C of ~ 1‰ (t-test, p < 0.0001) relative to the monsoon has been related to the biomass-burning aerosols originating from upwind north-west IGP region dominated by post-harvest biomass burning. However, as discussed before, photochemical oxidation could also lead to this ~ 1‰ enrichment. Therefore, to deduce source characteristics, other parameters should be used together with the δ13C value.

Bikaner

The δ13C of Bikaner site varies from −28.3‰ to −16.0‰ (−25.8 ± 1.8‰), indicating contribution from different sources (Table 1; Figs. 3a and 4). The average δ13C at Bikaner is 1.7‰ higher than that (−27.5‰ in PM10, Table 2) observed at a similar nearby location in Jodhpur, Rajasthan98. In this desert region, resuspended dust is a major source of ambient aerosols103, which is not the case for most of the other Indian locations. As a result, even in the monsoon, it has a characteristic δ13C signature (−25.1 ± 3.2‰; Table 1) with wide variability, compared to other sites (Figs. 5 and 6). Roy et al. (2023)103 analyzed the microphysical, optical, and chemical properties of fine aerosol collected from Bikaner City from October 2020 to January 2021, and reported a significant contribution of mineral dust to the ambient aerosol concentrations in this desert region. They reported an 11% contribution of mineral dust to total PM2.5 mass, with fine dust concentrations increasing from 9.3 ± 5.6 to 13.6 ± 6.7 µg m− 3 across clean to high pollution periods, which is not typical in other urban or rural regions. Others have argued that biomass burning is not a significant contributor to ambient pollution in this region but transboundary pollution from biomass burning, dominant in the Punjab and Haryana regions, can have an impact in this region103.

Fig. 6
figure 6

Box plot depicting seasonal variabilities in δ13C at eight NCAP sites.

Full size image

Jorhat

Jorhat has a typical tropical monsoonal climate characterized by high humidity and heavy rainfall. The δ13C at the Jorhat (rural) site is homogeneous and varies little from −28.2‰ to −26.4‰ (−27.4 ± 0.3‰, Figs. 3a and 4, Fig. S1), indicating contribution from the same source(s) or different sources with the same composition throughout the year. A slightly higher δ13C (t-test, p < 0.0001) of PM2.5 for pre-monsoon (−27.2 ± 0.3‰) and monsoon (−27.3 ± 0.3‰) than winter (−27.7 ± 0.4‰) and post-monsoon (−27.6 ± 0.3‰) is observed (Table 1). δ13C at Jorhat is much lower compared to other nearby locations of the Bay of Bengal (− 26.5 ± 0.8‰ in pre-monsoon PM10) and Kolkata having −26.0‰ in winter PM10 (Table 2)99,104. Although δ13C values indicate the source to be dominantly C3 biomass origin, information on atmospheric processes cannot be deduced solely from this parameter. Qadri et al. (2022)86 performed cluster analysis and potential source contribution function (PSCF) modeling considering total carbon (TC) in conjunction with δ13C data and identified C3 biomass/biofuel combustion to be the primary source of aerosols. Their modeling results also identified that a slightly higher δ13C during the pre-monsoon (−27.2 ± 0.3‰) and monsoon (−27.3 ± 0.3‰) relative to the winter (−27.7 ± 0.4‰) and post-monsoon (−27.6 ± 0.3‰; Table 1) was due to contribution from aged aerosols and enhanced secondary aerosols derived from photo-oxidation during winter and post-monsoon. Such information cannot be obtained solely from δ13C of aerosols.

Kashmir

The δ13C of aerosols ranges from −28.8‰ to −26.0‰ (−27.3 ± 0.5‰, Figs. 3a and 4, Fig. S1), suggesting a minor change in the sources of aerosols. The isotopic compositions strongly suggest a source dominated by C3 biomass/biofuel combustion and negligible influence of industrial pollution. Whereas the post-monsoon (‒27.7 ± 0.5‰), winter (‒27.6 ± 0.5‰), and pre-monsoon (‒27.2 ± 0.3‰) seasons show similar average δ13C values, that for the monsoon season is slightly 13C rich (‒26.8 ± 0.4‰, t-test, p < 0.0001; Table 1). The higher δ13C in monsoon could be a result of the influence of a low volatile fraction of organic carbon (LVOC) which makes up a refractory fraction of organic carbon due to photochemical processing/oxidation105. Qadri et al. (2022)86 showed that higher LVOC corresponded to higher δ13C, which suggests more influence from aged organic carbon fractions. Lower δ13C in pre-monsoon at Kashmir is comparable to Beas (‒27.4 ± 0.5‰, PM2.5)45 and Jodhpur (‒27.5‰, PM10) (Table 2), and could be attributed to enhanced secondary organic aerosol production45. The seasonal variation of δ13C at Kashmir was more or less similar to the Jorhat site located in NE India in a similar Himalayan setting (Fig. 5).

Mohali

δ13C at the MOH site ranges from −27.8‰ to −24.1‰ (Figs. 3a and 4, Fig. S1). The δ13C value during the post-monsoon (‒26.5 ± 0.4‰) was ~ 1‰ less than that of the winter season (‒25.6 ± 0.6‰) (Table 1). The seasonal (post-monsoon and winter) δ13C at Mohali was very similar to those observed at other IGP locations such as Beas, Kanpur, Delhi, and Varanasi (Table 2). The winter in this area is significantly dominated by post-harvest open biomass burning (Fig. S2). Using δ13C of carbonaceous aerosols collected during October 2016 (post-monsoon) and May 2018 (pre-monsoon) from a site in this region (Beas, Punjab) as an input parameter in an isotope mixing model, Singh et al. (2021)45 concluded that biomass/bio-fuel combustion is a significant source of air pollution in every season in this region. Singh et al. (2021)45 also utilized organic carbon (OC), elemental carbon (EC) measurements and their sub-fractions such as OC1, OC2, OC3, OC4, EC1, EC2, EC3 of aerosols/sources and air mass back trajectory (AMBT) analysis to strengthen their conclusions. However, fossil fuel-based sources (gasoline: −26.1‰; diesel: −26.3‰; bio-diesel: −26.5‰)93 contribute more to the pollution episodes during the monsoon and post-monsoonal periods (Fig. 6). δ13C observed in monsoon (−26.1 ± 0.6‰) and post-monsoon (−26.5 ± 0.4‰) in Mohali is similar to that observed in Beas (monsoon: −26.4‰ and post-monsoon: −26.6‰; Table 2)45. Moreover, the formation of secondary aerosols through photochemical oxidation (gas-to-particle partitioning) caused during the aging of aerosols results in an enrichment of 13C in the particle phase47,52. Such enrichment could explain the ~ 1‰ higher δ13C values in winter and pre-monsoon compared to that of monsoon and post-monsoon in Mohali (Figs. 4 and 6)45.

Rohtak

The δ13C of PM2.5 ranged from −28.3‰ to −24.6‰ (Figs. 3a and 4). The δ13C values of pre-monsoon (−26.4 ± 0.8‰), monsoon (−26.4 ± 0.3‰), and post-monsoon (−26.8 ± 0.7‰) are similar, but the winter season PM was slightly 13C-rich (−25.9‰ ± 0.6‰; Table 2). The seasonal δ13C was comparable with IGP locations Beas (−26.5 ± 0.4‰ for post-monsoon and −26.4 ± 0.5‰ for Monsoon PM2.5), Delhi (−26.4‰ for post-monsoon PM2.5), and Kolkata (−26.0‰ for winter PM10) (Table 2). In winter, the δ13C of PM2.5 at Rohtak (−25.9‰) was lower than those observed at a nearby location in Hisar (−23.9 ± 0.5‰19; Table 2). In this region, winter, pre-monsoon, and post-monsoon are dominated by post-harvest combustion of paddy and wheat residue (C3 biomass). These seasons also exhibit an overall δ13C variability of up to 0.8‰ relative to that of the monsoon season (0.3‰; Table 1). Enriched values in winter could result from aqueous phase processing of aerosols during which the aerosols encounter water droplets in clouds or fog and trigger chemical reactions and transformations106. During these interactions, the lighter 12C isotope is typically removed or oxidized more readily, resulting in enrichment in 13C (higher δ13C) in the remaining aerosols80. Using the United States Environmental Protection Agency’s IsoSource linear mixing model, Singh et al.45 inferred significant influence from fossil fuel-based vehicular sources in this region (gasoline: −26.1‰; diesel: −26.3‰; bio-diesel; −26.5‰) during pre-monsoon and monsoon.

Mysuru

The annual δ13C in Mysuru varied between −27.2‰ to −24.7‰ (Figs. 3a and 4, Fig. S1). The δ13C in winter (‒25.5 ± 0.4‰) was ~ 1‰ enriched in 13C compared to other seasons (pre-monsoon, monsoon, and post-monsoon: average δ13C of −26.4 ± 0.5‰) (Table 1). Yadav et al. (2022)88 utilized δ13C along with OC, EC, and water-soluble inorganic ions in the aerosols and argued that proximal vehicular exhaust emissions contributed carbonaceous aerosols during pre-monsoon, monsoon, and post-monsoon. The variation of δ13C in winter was between −26.1‰ to −24.5‰ and when plotted against the reciprocal of PM2.5 mass concentration (Keeling Plot)107, indicated contributions from traffic (petrol and diesel) and coal-based combustion sources (see Fig. 2 of Yadav et al.88). This conclusion was also supported by a MixSIAR model (Bayesian mixing model) derived proportional contribution of 71% from fossil-based sources.

Shyamnagar

The δ13C of PM2.5 at SHY varied between −29.3‰ to −25.3‰ (Figs. 3a and 4, Fig. S1). Wintertime average δ13C at Shyamnagar was −26.4 ± 0.3‰ and post-monsoon average was −26.4 ± 0.5‰, which was marginally higher than pre-monsoon and monsoon averages of −27.0 ± 0.4‰ and −27.4 ± 0.4‰, respectively (Table 1). δ13C in winter at Shyamnagar is similar to the neighboring Kolkata (−26.0‰, PM10), as observed in a previous study99 (Table 2). The δ13C in pre-monsoon is slightly lower than that observed in the Bay of Bengal (−26.5 ± 0.8‰) reported by Agnihotri et al. (2011)104. Utilizing δ13C versus 1/TC relationship (Keeling plot) in the aerosols from this sampling site, Singh et al. (2023)92 identified a significant contribution from vehicular exhaust in the post-monsoon and winter seasons, which have nearly similar δ13C values (− 26.4‰). Also, the results of air trajectory and PSCF analyses suggested that the atmospheric aging of anthropogenic aerosols arriving from upwind IGP could be responsible for the relatively 13C-rich values observed in the winter and post-monsoon seasons. In contrast to winter and post-monsoon, most samples from the pre-monsoon and monsoon were 13C-depleted (about − 1‰ lower) and had the lowest average TC contents of ~ 4 to 8 µg m− 3 (Fig. S1; Table S1). Singh et al.92 argued that the influence of SOC (38-72%) in winter and post-monsoon resulted in 13C enrichment, compared to pre-monsoon and monsoon. As an IGP outflow site, Shyamnagar receives aerosols from upwind IGP regions with significant anthropogenic sources during the winter and post-monsoon.

Comparison with δ 13C of aerosols at other sites globally

The average seasonal δ13C values of carbonaceous aerosols measured at Indian NCAP sites range from ‒27.7‰ to ‒25.1‰ and exhibit spatial and seasonal variability (Table 1). We have compared our data (average isotopic compositions) with those from other global sites representing rural/forest, urban, and marine environments, presented in Table 3. Isotopic variability in our sites broadly aligns with global observations, that is higher δ13C values (13C-enriched signature) are typical of urban and combustion-dominated regions, whereas forest and marine environments are characterized with 13C-depleted signature (lower δ13C values), excluding the marine site in Okinawa, Japan119. For instance, Indian locations like MYS, KMR, and JOR (‒26 to ‒27.7‰), located in less polluted regions, exhibit δ13C values similar to those in the Arctic Ocean (‒27.4 ± 1.3‰) and forest and marine sites in Lithuania (‒27.2 to ‒27.9‰) during summer time (Table 3), indicative of biogenic carbon (C3 plants) sources. In contrast, urban sites such as MOH and RTK are relatively 13C-enriched (up to ‒25.5‰), which is comparable to urban settings like Mexico City (‒25.4 ± 1.2‰), Prague (‒25.5 to ‒27.2‰), and urban centers in Poland (‒25.3 to ‒26.7‰) and Beijing, China (‒23.8 to ‒24.9‰), highlighting contributions from fossil fuels and vehicular emissions. Interestingly, the isotopic composition of Indian sites is more homogenous compared to those of the global sites that show a few sites having more 13C-rich signatures (e.g., ‒22.3‰ at Okinawa, Japan, Table 3). We would like to note that δ13C variability of aerosols at a specific site is largely controlled by the regional/local combustion sources and anthropogenic emissions, with urban sites showing strong influence from fossil fuel emissions/biomass burning, whereas background and remote sites are more influenced by biogenic or long-range transported carbon.

Table 3 δ13C values of aerosols at selected sites globally, representing urban, rural, and marine locations.
Full size table

Seasonality of δ 13C variability at NCAP locations

Winter

Our data show seasonal δ13C variations at all the sampling sites (Figs. 5 and 6), which can be attributed to a plethora of factors such as the variation of the sources, meteorology and related atmospheric chemistry (Fig. 1). The special distribution map of isotopic variability shows that Kashmir (KMR) and Jorhat (JOR) sites, located in the North and North-East India in the Himalayan foothills, have the lowest average δ13C values (by 2 to 3‰) compared to other sites during the winter season (Fig. 5). Temperature in winter season in the North and North-East India, especially in the Himalayan regions, becomes much lower compared to other parts of the country. Low rainfall, low wind speed, and low boundary layer height create stagnant conditions in this region. Thus, the particles emitted from the sources do not undergo significant dilution/dispersion and are minimally affected by transformation processes during travel by air masses. Therefore, δ13C of PM2.5 [−27.6 ± 0.5‰ (KMR) and −27.7 ± 0.3‰ (JOR); Table 1] likely represents the composition of its source, which are dominantly C3 biofuel/biomass or fossil fuel-based sources. The TC contents at KMR and JOR attain the highest values in winter (Table S1). These two locations are also relatively less urbanized and therefore the contribution from other anthropogenic activities such as vehicular/industrial sources are negligible. On the other hand, other Indian locations (RTK, MYS, MOH, BPL, BKN, SHY) show higher δ13C values in winter (Figs. 5 and 6, and S1), which could be due to the presence of 13C-enriched emission sources and/or aging of particles/SOA formation during transport that could result in 13C enrichment. Chamber-based experiments have demonstrated ~ 2–3‰ 13C enrichment in oxalic acid (a dominant organic aerosol species) through photocatalytic reactions in the aqueous phase101. The usual sources of oxalic acid in aerosols could be from direct emissions such as burning biomass and fossil fuels, or secondary production during oxidation of organic precursors such as VOCs. The VOCs, NOx, and oxidants predominantly originate from vehicular emissions. Aided by unstable air circulation patterns such as turbulence and inversions, these NOx and VOC precursors could promote secondary production of oxalic acid in urban and polluted areas. These locations are also more urbanized and populated compared to Jorhat and Kashmir with more influence from anthropogenic sources.

Pre-monsoon

The average δ13C in pre-monsoon exhibits relatively lower values and more overall variability at all locations (except KMR and JOR), compared to the winter season (Figs. 5 and 6). In pre-monsoon, an increase in temperature and sunlight intensity, along with changes in wind speed/direction, results in strong convective mixing of airmasses, causing dispersion of pollutants. Atmospheric processes could result in an increase in δ13C of aerosols undergoing photochemical aging during long-range transport47,80. On the other hand, enhanced SOA formation due to photochemical oxidation in pre-monsoon/summertime would result in 13C depletion in the oxidation products and 13C enrichment in the precursors (Fig. 1). Kashmir (KMR) in the Upper Himalayan Region and Jorhat (JOR) in the foothills of lower Himalaya exhibited an increase in δ13C in pre-monsoon compared to the winter season (Fig. 6). It has been observed that the products in the aerosol phase are more enriched in 13C than in the gaseous phase if equilibrium separation between phases takes place79. It is possible that PM2.5 in KMR and JOR undergoes a transformation process due to unique atmospheric conditions in which equilibrium fractionation results in 13C enrichment of the product phase. In Kashmir, cooler temperatures enhance the condensation of semi-volatile organic compounds (SVOCs) onto aerosols, enriching heavier isotopes (13C) in the particle phase due to partitioning. In Jorhat, high humidity from rising temperatures and pre-monsoon rains might promote aqueous-phase reactions in droplets, where equilibrium effects preferentially incorporate lighter isotopes (12C) into the liquid phase, leaving particles enriched in 13C (higher δ13C). These contrasting conditions may amplify isotopic enrichment at these locations compared to other locations.

Monsoon

The δ13C variability (average: −26.6 ± 1.2‰) in monsoon, considering all studied locations, shows an interesting result. The average δ13C and the magnitude of variability for each location is typically the lowest amongst all seasons, except for the BKN location (Figs. 6). The most 13C-depleted signature of monsoon aerosols could be attributed to the frequent removal of emitted aerosols from the ambient air due to washout/scavenging by rainfall and high humidity, resulting in a signature representative of the isotopic composition of their primary sources. The frequent washout of the aerosols nullifies the isotope fractionation effect of the long-range transport of aerosols and SOA formation. However, δ13C of BKN shows the largest variability ranging from −16‰ to −28‰ (Figs. 3 and 6), has a more 13C-rich signature (average δ13C of − 25.1), and is seen as a distinct hotspot in Fig. 5. We attribute this distinct signature to additional influence from marine aerosols. Sudheer et al. (2016)98 reported δ13C (av. ‒27.5‰: ‒29.6‰ to ‒25.8‰) of PM10 carbonaceous aerosol collected during summer and monsoonal season (May-September) from Jodhpur, located ~ 240 km south of our site BKN. Based on δ13C and other chemical proxies (WSOC, OC, EC, and ionic species), Sudheer et al.98 identified a major contribution from biogenic primary emissions and minimal contribution from combustion emissions, SOA formation, or resuspended soil dust and long-range transport. According to Ceburnis et al.108, the δ13C of marine aerosols varies from −20.3‰ to −25.3‰. Bikkina et al.47 reported δ13C in total suspended particulates (TSP) sampled over the Arabian Sea in winter, from ‒25.1‰ to ‒22.9‰. Kirillova et al.40 showed that the marine aerosols reaching Sinhagad, India (near the west coast) during monsoon were about 2‰ 13C-enriched compared to that of TOC. It is possible that the distinct 13C-rich signature at BKN during monsoon could be influenced by contributions from marine aerosols. However, further tracer studies (such as sodium, chloride, magnesium, calcium ions, and methanesulfonic acid (MSA)) are needed to decipher the quantitative mass contribution of various sources.

Post-monsoon

The δ13C values during post-monsoon become slightly higher compared to the monsoon and resemble the signature observed for pre-monsoon (Figs. 5 and 6; Table 1). The JOR and KMR sites show lower δ13C values (average of ‒27.7‰) relative to the other locations. Fire maps for different seasons during our sampling duration show most fire events originating in the upwind IGP region (Fig. S2), which is related to large-scale post-harvest open biomass (paddy; primarily of C3 origin) burning in the post-monsoon. A decrease in temperature and boundary layer height (particularly at nighttime) in North India in the post-monsoon results in a stagnant condition45. The longer time spent by the particles in these conditions and the associated equilibrium isotopic fractionation between gas and particle result in 13C enrichment in the product aerosols45. These conditions are amplified in the winter seasons at most locations causing higher δ13C in aerosols, distinctly observed in Fig. 5. Further, the WSOC fraction tends to increase the δ13C of the total carbon which further increases during long-range transport40,49,80,105, and likely play a major role during post-monsoon and winter in transporting aerosols originating from upwind IGP to Bay of Bengal (BOB) region, which explains relatively higher δ13C values in winter and post-monsoon aerosols from our study sites.

Use of stable carbon isotope ratios of CA as a stand-alone proxy for source characterization

δ13C is a valuable tool for investigating several aspects related to the sources of carbonaceous aerosols, but our analysis of data from eight Indian sites implies that its utilization as a standalone proxy for source apportionment purposes has certain limitations. The uncertainty in the source apportionment can arise due to the large range of δ13C values of sources, causing overlapping isotopic signatures observed for biomass burning, fossil fuels, dust, or industrial sources. For example, Fig. 4 shows that there could be several sources (e.g. C3 plants and fossil fuels) having overlapping δ13C values dominating the pollution at a particular location. The isotopic composition of all probable sources, locally or regionally, should be determined to identify characteristic (distinct vs. overlapping) source signatures. Moreover, the degree of atmospheric processing and the isotopic fractionation incurred during transport along with the mixing of multiple sources also complicates the quantification of primary sources (Fig. 1). Atmospheric transformation processes such as different SOA formation mechanisms (photochemical oxidation) depend on the availability of precursors, oxidants, temperature, humidity, and atmospheric stability, etc., often characteristics of a specific location49,80. The principal factor controlling the initial isotope fractionation is the kinetic isotope effect (KIE), which primarily arises during unidirectional (irreversible) reactions of organic substances83. Depending on the KIE, the reaction products may go through additional reactions that result in additional stepwise δ13C changes, and therefore knowledge of the reaction mechanism and the associated KIE must be known to deduce the primary isotopic composition of the sources.

Nevertheless, δ13C could be used in conjunction with other chemical tracers such as temperature-resolved carbon fractions OC/EC, WSOC, water-soluble ions (e.g. K+ BB), organic compounds (e.g. levoglucosan), and Δ14C values to identify and quantify aerosol sources, particularly, identify combustion sources and secondary sources (OC/EC), the contribution of WSOC, proportion of biogenic vs. fossil-derived carbon (using Δ14C values) and identify biomass burning sources (K+ BB).

For example, others have used δ13C versus 1/TC (the so-called Keeling plot) for source characterization. The y-intercept of the linear regression in a Keeling plot represents the isotopic signature of the dominant source107. The δ13C versus 1/TC variations in our data and the y-intercepts for our studied sites are given in Fig. 7. The δ13Cintercept for BPL (−26.4 ± 0.0‰) indicates a major contribution from vehicular exhaust88. δ13Cintercept for BKN (−25.8 ± 0.2‰) indicate more 13C-enriched sources dominating this site. A few higher 1/TC values at this site indicate low aerosol concentrations and are more 13C-rich ( ~ −21‰). As discussed before in the previous section and supported by previous studies40,47,108, this 13C-rich signature at BKN is attributed to contributions from marine aerosols. δ13Cintercept for MYS (− 25.8‰) is also lower with the lowest overall TC content and indicates contributions from traffic (petrol and diesel) and fossil fuel combustion sources88. At the Mysuru site, periods with elevated 1/TC values (> 1, Fig. 7) indicate that δ13C is predominantly influenced by contributions from vehicular exhaust, reflecting the dominance of local combustion-related emissions. Sites JOR (rural) and KMR (semi-urban Himalayan region) exhibit the lowest δ13Cintercept of −27.5‰, suggesting a dominant contribution from C3 biomass combustion (e.g., wood, crop residue) and less fossil fuel influence. TC values in these regions during the winter are among the highest, consistent with seasonal wood combustion for heating and cooking (Fig. 3b; Table S2). Sites MOH, RTK (and SHY) show similar δ13C signatures (δ13Cintercept ≈ ‒26.4‰), suggesting contributions from vehicular emissions and C3 plant combustion, especially during post-harvest seasons.

Fig. 7
figure 7

δ13C vs. 1/TC of PM2.5 aerosols at our NCAP sites. Dashed lines and solid lines indicate average and range of δ13C of various sources. See Fig. S3 for individual plots.

Full size image

Air Mass Back Trajectories (AMBTS) can also be used in tandem with δ13C data for identification of the geographical origin of long- and short-range sources19,56,92. Similarly, analyzing forest fire data (e.g., MODIS data) in the region, and identifying fire hotspots could help interpret the carbon isotope composition of aerosols (e.g., Qadri et al.,86). Moreover, for improving the accuracy of source apportionments, isotope mixing statistical models such as IsoSource, MixSIAR, and PMF can be utilized alongside other indicators such as OC/EC, ionic species, and metals56,109,110,111,112. Therefore, correlations between δ13C and other proxies discussed above, or using these as input parameters in statistical models would reduce the uncertainties in isotopic end member values and improve the model resolution, giving a robust and comprehensive source apportionment of carbonaceous aerosols.

Conclusion

Determination of stable carbon isotope (δ13C) measurements is an advanced analytical technique that is utilized for deciphering the ambient pollution sources and understanding particle formation and transformation processes of carbonaceous aerosols. This work utilizes δ13C variability in carbonaceous aerosols sampled in a nationwide NCAP-COALESCE network in India and discusses how good a tracer δ13C value of aerosols is in tracing (and quantifying) the emission sources. Our data show a distinct δ13C variability in Indian aerosols both in space and time (seasonal). Out of the eight locations continuously monitored, Jorhat (northeast India) and Kashmir (northern Himalayan region) sites exhibit the most depleted δ13C values during all seasons. Mohali and Rohtak, located in the upwind IGP, showed similar variations in δ13C in all four seasons, indicating the dominance of similar sources and atmospheric processes. On the other hand, the Bikaner (arid and dry) site shows more positive δ13C values during pre-monsoon and monsoon, depicting the influence of resuspended dust (due to high wind speed). In winter and post-monsoon, the δ13C (at Shyamnagar, Bhopal, and Rohtak) shows the influence of transboundary pollution from other states. Background site Mysuru (southern India) showed more positive δ13C in winter and post-monsoon season compared to other sites, suggesting the influence of aerosol aging. While other seasons suggested an influence from more negative biogenic sources. Using the y-intercept in a δ13C versus 1/TC relationship (Keeling plot), we assessed the dominant sources across our study sites. These results indicate that emissions from vehicles/fossil fuels dominate at BPL and MYS sites, whereas C3 biomass combustion dominates at rural and semi-urban locations like JOR and KMR. On the other hand, MOH, RTK, and SHY locations are influenced by multiple sources, such as post-harvest biomass burning and vehicular combustion. Due to the complex and varied nature of aerosol sources and the potential for isotopic fractionation involved during aerosol formation/transport, δ13C is not an ideal standalone proxy for source apportionment of atmospheric aerosols. Further, the interpretation of δ13C can be difficult if sources have overlapping isotopic compositions. However, δ13C could be used with other proxies such as radiocarbon (Δ14C), OC/EC, K+ BB, AMBTs, trace elements/organic species, and aerosol size distribution, etc., to differentiate between fossil and biogenic sources and perform robust source apportionment estimates. Specific biomarkers like levoglucosan can also be utilized with δ13C to identify contributions from biomass burning. Isotope mixing statistical models that combine relevant proxies and δ13C can provide a more accurate estimate of sources and a comprehensive understanding of aerosol composition, their environmental impacts, and climate change.