Brownness of organics in anthropogenic biomass burning aerosols over South Asia

In South Asia, biomass is burned for energy and waste disposal, producing brown carbon (BrC) aerosols whose climatic impacts are highly uncertain. To assess these impacts, a real-world understanding of BrC's physio-optical properties is essential. For this region, the order-of-magnitude variability in BrC's spectral refractive index as a function of particle volatility distribution is poorly understood. This leads to oversimplified model parameterization and subsequent uncertainty in regional radiative forcing. Here we used the field-collected aerosol samples from major anthropogenic biomass activities to examine the methanol-soluble BrC optical properties. We show a strong relation between the absorption strength, wavelength dependence, and thermo-optical fractions of carbonaceous aerosols. Our observations show strongly absorbing BrC near the Himalayan foothills that may accelerate glacier melt, further highlighting the limitations of climate models where variable BrC properties are not considered. These findings provide crucial inputs for refining climate models and developing effective regional strategies to mitigate BrC emissions.

1 Introduction

Carbonaceous aerosols, such as black and organic carbon, make up most fine particulate matter (PM_2.5) emissions globally (McDuffie et al., 2020; Roy et al., 2023; Kurokawa and Ohara, 2020; Crippa et al., 2018) and ∼ 40 % over South Asia (Tibrewal et al., 2024; Pandey et al., 2014; Sadavarte et al., 2019). Anthropogenic biomass usage for residential cooking and heating (Pandey et al., 2014; Habib et al., 2023; Navinya et al., 2023), residue burning for agricultural waste disposal (Kapoor et al., 2023b; Azhar et al., 2019), and biomass-fired brick kilns (Weyant et al., 2014; Tibrewal et al., 2023) are the common sources of these carbonaceous aerosols across South Asia (Tibrewal et al., 2024; Pandey et al., 2014; Sadavarte et al., 2019; Ohara et al., 2007) and many other developing countries (Bonjour et al., 2013; Yevich and Logan, 2003; McDuffie et al., 2020). These aerosols perturb the Earth's energy balance, depending on their mixing state, size distribution, wavelength dependence of optical properties, and absorption strength (Zhang et al., 2020; Neyestani and Saleh, 2022; Brown et al., 2018; Arola et al., 2015; Bond and Bergstrom, 2006). However, the extent of this perturbation remains uncertain (Szopa et al., 2021; Gliß et al., 2021). Over the last 2 decades, extensive research has focused on the climate impact of highly absorbing black carbon (BC) (Bond et al., 2013). In contrast, the climate implications of light-absorbing organic carbon (OC), termed brown carbon (BrC), have received relatively little attention and are thus less certain (Saleh et al., 2018; Brown et al., 2018; Saleh, 2020). The chemical composition of BrC varies significantly, and consequently its optical properties, as reported across previous studies, span orders of magnitude in the imaginary refractive index (k) values that determine its light-absorbing strength (Chakrabarty et al., 2023; Choudhary et al., 2021, 2018, 2017; Dey et al., 2021; Kapoor et al., 2023a; Kirillova et al., 2016; Rana et al., 2020; Rathod et al., 2017; Saleh et al., 2018, 2014; Srinivas and Sarin, 2014). Previous studies have often measured aged ambient BrC that is weakly absorbing (k_BrC,550 < 0.01) due to photobleaching (Sumlin et al., 2017); hence some climate impact assessment studies have regarded BrC as a weakly absorbing or non-absorbing particle (Lee et al., 2010; Sand et al., 2021; Zhang et al., 2020). However, this underestimates the impact of freshly emitted BrC that has high absorption strength (k_BrC,550 > 0.1) and resists photobleaching, resulting in an extended atmospheric lifetime (Chakrabarty et al., 2023). Furthermore, the formation of light-absorbing secondary BrC and the enhancement of BC absorption due to OC coating (Rastogi et al., 2021; Bhowmik et al., 2024; Kapoor et al., 2022) add complexity to radiative transfer models.

BrC has a wide range of absorption strengths; studies show k_BrC,550 varying from ∼ 0.007 (Islam et al., 2022) to ∼ 0.2 (Chakrabarty et al., 2023). In addition to the different methods used to derive BrC optical information, such variation is associated with the different combustion conditions (Saleh et al., 2018), aging of BrC (Sumlin et al., 2017; Dasari et al., 2019; Romonosky et al., 2016; Chen et al., 2021), and secondary reactions (Wang et al., 2020; Kroll et al., 2007; Kroll and Seinfeld, 2008; Hecobian et al., 2010). An experimental study explained that the progressive transformation of BC precursors to BC results from different combustion conditions, which create the BrC–BC light absorption continuum (Saleh et al., 2018). This continuum shows an increase in the absorption strength of carbonaceous aerosols that is associated with a decrease in wavelength dependence (w), solubility, and volatility (Saleh, 2020). Recent studies have also observed such a relationship but for a smaller range of k_BrC values (< 0.01) (Devaprasad et al., 2024; Luo et al., 2022). However, information about real-world source-specific BrC absorption and its position in the BrC–BC continuum is lacking. Understanding this light absorption continuum alongside carbonaceous aerosol emissions aids BrC parameterization in climate models (Zhang et al., 2020; Saleh et al., 2014). Presently, because source-specific BrC information is absent from emission inventories, many climate models inadequately account for BrC. Studies have used the BrC-to-BC ratio along with k_BrC to understand its direct radiative effect (Park et al., 2010; Feng et al., 2013). Furthermore, other studies (Zhang et al., 2020; Neyestani and Saleh, 2022; Brown et al., 2018) have employed BrC parameterization schemes based on laboratory-generated data to address the climate impact of BrC, but this approach might not adequately represent real-world biomass burning conditions (Saleh et al., 2014; Lu et al., 2015). Hence, regions with high OC emissions and stronger BrC (S-BrC), also known as dark BrC (k_BrC,550 > 0.1), could have a high climate impact caused by persistent BrC, which is possibly underestimated in the absence of regional source-specific BrC data.

The recent Carbonaceous Aerosol Emissions, Source Apportionment and Climate Impacts (COALESCE) field emission measurement campaigns and questionnaire surveys in India (Navinya et al., 2023; Kapoor et al., 2023b; Tibrewal et al., 2023; Habib et al., 2023) have prepared a comprehensive inventory encompassing both formal (transportation, industries, and power generation) and informal (residential, agricultural residue burning, and brick production) emission sectors (Venkataraman et al., 2020; Tibrewal et al., 2024). The emission estimates show the substantial contribution of anthropogenic PM_2.5 in India from biomass fuel burning practices for residential cooking and agricultural residue burning (Kapoor et al., 2023b; Tibrewal et al., 2024; Habib et al., 2023). Recent studies have highlighted considerable biomass consumption for residential heating and brick production (Tibrewal et al., 2023; Navinya et al., 2023). Figure 1 shows that 91 % of the OC emissions (3 Tg yr⁻¹) over India are from three sources: residential cooking (COOK), heating (HEAT), and agricultural residue burning (AGRI), with most emissions from the Indo-Gangetic Plain (∼ 50 %) (Tibrewal et al., 2024). The unexplored climate impacts of OC emitted from these biomass-based sources make the Indian subcontinent particularly prone to environmental challenges.

Figure 1 The spatial distribution of OC emissions (Mg yr⁻¹ per pixel) from three major sources – agricultural residue burning (AGRI), residential cooking (COOK), and residential heating (HEAT) – covers ∼ 91 % of the total OC emissions (3.3 Tg yr⁻¹) over India. OC emissions are taken from the Carbonaceous Aerosol Emissions, Source Apportionment and Climate Impacts (COALESCE) Speciated Multipollutant Generator for India (SMoG-India) emission inventory for the year 2019 (Tibrewal et al., 2024). Here, the pixel size is 5 km × 5 km. The pie chart represents the shares of anthropogenic biomass burning sources in the total OC emissions over India. Other sources (OTHR) of OC include brick production, transportation, industries, and power generation. Publisher's remark: please note that the above figure contains disputed territories.

This study leverages samples of aerosol particle emissions collected on filter substrates during the COALESCE field campaign to evaluate BrC–BC light absorption continuum behavior in real-world biomass burning emissions. Using a UV–Vis spectrophotometer, we examine BrC derived from major biomass fuel sources such as cooking, heating, agricultural residue burning, and brick production. The study aims to connect BrC with the thermo-optically resolved carbon fractions to parameterize BrC absorption over South Asia. Further, it endeavors to couple source-specific BrC properties with the BC-to-organic aerosol (OA) ratio to explore the spatial variability in the absorption properties of BrC emitted across India.

2 Data and methods

2.1 Data collection

A field-based emission measurement campaign (Fig. S2 in the Supplement) was conducted from October 2021 to April 2022 in rural parts of Gujarat and Maharashtra, two western Indian states. These locations were selected based on their representativeness of the fuels and devices commonly used in South Asia based on previous studies (Navinya et al., 2023; Kapoor et al., 2023b; Tibrewal et al., 2023; Habib et al., 2023). The primary aim of this campaign was to capture physical, chemical, and optical information about the emissions from biomass sources: agricultural residue burning, brick production from clamps, cooking, and heating. The source emission sampling system, as described by Kumari et al. (2024) and Venkataraman et al. (2020), consists of a multi-arm inlet design adapted from Roden et al. (2006) to function as an area plume sampler, positioned 1 to 1.5 m above the emission source (Fig. S2). The system comprises eight arms that aspirate aerosols, which are then combined in a mixing plenum to ensure representative sampling of the smoke plume. Aerosols drawn through the inlet pass through a 2.5 µm cutoff cyclone and are subsequently divided into two streams for real-time and for time-integrated filter-based measurements. Aerosols from the latter airstream were collected on quartz filter substrates for offline laboratory analysis over the entire duration of the experiment, encompassing the ignition, flaming, and smoldering phases, in order to obtain a sample representative of the complete combustion cycle. The temperatures of the emitted plumes were diluted by the surrounding air, reaching levels close to the temperature of ambient air before entering the multi-arm sampler. This ensured that the emissions had undergone gas-to-particle partitioning, corresponding to the properties of emissions used in climate models. In this study, we utilized aerosol-laden quartz filter substrates from 14 different fuel and source combinations (Table S1 in the Supplement) to understand soluble BrC absorption (Mm⁻¹ = 10⁶ m⁻¹) and total OC concentration (µg m⁻³).

2.2 Estimation of BrC properties

We used 4.5 mL of methanol solvent and dissolved two 0.25 in. diameter punches of quartz filters in the solvent. After 1 h of sonication, the extracted solvent was passed through a 0.22 µm polytetrafluoroethylene membrane syringe filter (Fisherbrand™) to remove insoluble debris. The absorption of this methanol-soluble OC (considered BrC absorption) was estimated using a UV–Vis spectrophotometer (LAMBDA 35, PerkinElmer) with a working range of 300 to 900 nm and a spectral resolution of 1 nm. Equation (1) was used to estimate the absorption coefficient at any given wavelength (Chakrabarty et al., 2023; Sarkar et al., 2019; Satish and Rastogi, 2019; Srinivas and Sarin, 2013, 2014; Bikkina et al., 2020; Boreddy et al., 2021; Choudhary et al., 2017, 2018, 2021, 2022; Dasari et al., 2019; Dey et al., 2021; Kirillova et al., 2016; Mukherjee et al., 2020; Rajeev et al., 2022; Rastogi et al., 2021; Rathod et al., 2017; Rana et al., 2020; Shamjad et al., 2016, 2018).

$$\begin{array}{}\text{(1)}& b_{\text{abs},\mathrm{BrC},\mathit{\lambda }}={\displaystyle \frac{(A_{\mathit{\lambda }}-A_{\mathrm{700}})\times V_{\text{Extract}}\times \ln \left(\mathrm{10}\right)}{V_{\text{Sampled}}\times L\times f_{\text{filter\hspace{0.17em}\hspace{0.17em}area}}}}\end{array}$$

In Eq. (1), A_λ is absorbance at wavelength λ, V_Extract is the volume of solvent extract used (4.5 mL in this study), V_Sampled is the volume of air sampled, f_{filter  area} is the fraction of filter area used for the analysis, and L is the optical path length (0.01 m). Given that soluble BrC does not absorb at wavelengths of 700 nm and longer or, at best, absorbs very little, the absorption at 700 nm (A₇₀₀) was used to normalize absorbance to account for signal drift within the instrument, which is a limitation of this method. In this study, the estimated BrC only includes the methanol-soluble component and may not fully represent total BrC, including its insoluble components. The estimated BrC absorption could be underestimated due to excluded insoluble BrC and tarball structures, which possess high absorption strength (Corbin et al., 2019; Chakrabarty et al., 2023, 2010). The underestimation may be more pronounced as particle light absorption strength increases, i.e., closer to the dark-BrC region, since particle solubility is inversely proportional to light absorption strength (Saleh, 2020). In brief, Saleh (2020, and references therein) reviewed and categorized different BrC classes based on their volatility, using UV–Vis spectrometry, optical closure (Aethalometer, cavity ring-down spectroscopy, and photoacoustic spectroscopy), and electron energy loss spectroscopy techniques. While UV–Vis spectrometry misses out insoluble particles, optical closure techniques consider absorption by particles regardless of their solubility. However, they have uncertainties associated with separating BrC light absorption from the total aerosol light absorption. In this study only two data points, observed marginally in the dark-BrC region, might be affected.

Quartz filters were examined using a Magee Scientific DRI multi-wavelength thermo-optical carbon analyzer with the IMPROVE_A protocol to estimate the elemental carbon (EC) and organic carbon (OC) concentrations (Chow et al., 2007). Thermo-optically resolved carbon fractions (OC1, OC2, OC3, OC4, EC1, EC2, and EC3) were used after pyrolytic correction to reconstruct the total organic carbon and total elemental carbon fractions (Chow et al., 2007). For the purpose of representation in Fig. 3, pyrolytic carbon was assigned to OC4. These fractions are associated with the volatility of the OC (Kapoor et al., 2023a; Shetty et al., 2023; Tohidi et al., 2022; Vodička et al., 2015; Soleimanian et al., 2019; Ma et al., 2016), as these OC fractions are measured under increasing temperature peaks (140, 280, 480, and 580 °C) during thermo-optical analysis. Hence, OC1 exhibits relatively high volatility compared to OC2, while OC2 is more volatile than OC3, and similarly, OC3 shows more volatility than OC4. In this study, pyrolysis-corrected EC was treated as a proxy for BC to facilitate the comparison with other studies. The uncertainties associated with OC and EC measurements are 5 % and 10 %, respectively (Cheng et al., 2021; DRI Manual, 2015). Cheng et al. (2021) reported an overall uncertainty of approximately 10 % for methanol-soluble k_BrC determined through UV–Vis spectrophotometry. When accounting for the 5 % manufacturer-reported uncertainty in OC concentration, the corresponding uncertainty in the absorption coefficient is estimated to be around 10 %.

Furthermore, OC concentration and $b_{\text{abs},\mathrm{BrC},\mathit{\lambda }}$ were used to calculate the mass absorption coefficient (MAC_BrC,λ). The imaginary refractive index of BrC (k_BrC,λ) was estimated by considering the density (ρ) of freshly emitted OC to be 1500 kg m⁻³ (Liu et al., 2013; Shamjad et al., 2016), using the following relation (Jennings et al., 1979):

$$\begin{array}{}\text{(2)}& k_{\mathit{\lambda }}={\displaystyle \frac{\mathit{\rho }\times \mathit{\lambda }\times {\text{MAC}}_{\mathit{\lambda }}}{\mathrm{4}\mathit{\pi }}}.\end{array}$$

The same equation has been used in many previous studies, some of which cover the same geographic region (Shamjad et al., 2018; Bikkina and Sarin, 2019; Shamjad et al., 2016; Rana et al., 2020; Liu et al., 2013; Zhang et al., 2020). In addition, an absorption Ångström exponent (AAE) between 365 and 550 nm (${\text{AAE}}_{\mathrm{365}/\mathrm{550}}$ = $-\ln (b_{\text{abs},\mathrm{BrC},\mathrm{365}}/b_{\text{abs},\mathrm{BrC},\mathrm{550}})/\ln (\mathrm{365}/\mathrm{550})$) was also estimated to understand the spectral dependence of the BrC absorption coefficient. Similarly, w (AAE-1) indicates the spectral dependence of the imaginary refractive index between 365 nm (a commonly used wavelength for studying BrC absorption) and 550 nm (the peak of solar radiation intensity). In this study, we have used w and k for ease of comparison with previous studies (Saleh et al., 2014; Lu et al., 2015; Luo et al., 2022; Saleh et al., 2018). However, AAE and MAC can also be used alternatively.

2.3 Spatial variation in BrC absorption

The relationship between fuel- and source-averaged k_BrC,550 and the BC-to-OA ratio (k_BrC,550 = $\mathrm{0.0365}(\pm \mathrm{0.006})\times (\mathrm{BC}/\mathrm{OA})+\mathrm{0.0047}(\pm \mathrm{0.0037})$, R² = 0.93) was established using field-collected fuel samples. Similarly, w was also calculated as a function of the BC-to-OA ratio ($w=\mathrm{5.355}(\pm \mathrm{0.50})\times \exp (-\mathrm{0.428}(\pm \mathrm{0.25})\times (\mathrm{BC}/\mathrm{OA})$; R² = 0.60). Here, OA was derived by multiplying OC by a factor of 1.8, a methodology consistent with previous studies (Turpin and Lim, 2001; Chow et al., 2015; Navinya et al., 2020; Provençal et al., 2017; Kumar et al., 2023) and aligned with the OA density considered (Kuwata et al., 2012). Although this factor does not impact the R squared (R²) of the relationship, it facilitates comparisons with other studies that have utilized the BC-to-OA ratio to derive k_BrC,550. The spatial distribution of BC and OC emissions from the SMoG-India emission inventory (Tibrewal et al., 2024) was integrated into the equation, after converting OC into OA using the same factor, to calculate the nationwide k_BrC,550 and w for the major (∼ 90 %) OC-emitting sources: AGRI, COOK, and HEAT. Additionally, we derived overall k_BrC and w values through a weighted averaging approach, incorporating OC emissions (Fig. S5 in the Supplement) as weights along with source-specific information (Fig. S3 in the Supplement). BRICK (brick production) was omitted because field-based samples were limited to clamp kilns and not available for other major brick production technologies, including Bull's trench kilns and vertical shaft brick kilns (Weyant et al., 2014; Tibrewal et al., 2024, 2023).

3 Results and discussion

3.1 BrC–BC absorption continuum

The measured k_BrC,550 values varied from 0.0007 to 0.1199, while w ranged from 7.52 to 1.00, highlighting the inverse dependence of k_BrC on w (Fig. 2). A previous study using synthetic fuels under different combustion conditions reported a similar observation based on experimental measurements (Saleh et al., 2018). Relative to the present study, different field-collected sources and fuels reflected real-world variations in burning practices. An equation fitted to the data ($w=\mathrm{0.1917}/(k_{\mathrm{BrC},\mathrm{550}}+\mathrm{0.02886})$) has an R² value of 0.58, and an extension of this curve with 95 % prediction bounds overlaps the BC absorption region (k₅₅₀ = 0.6–0.8 and w = ∼ 0–0.2) (Bond and Bergstrom, 2006; Saleh et al., 2018; Liu et al., 2018; Gyawali et al., 2013). The range of k_BrC,550 and w values observed in this study spans three broad classes of BrC (weak, moderate, and strong) suggested by Saleh (2020) for different combustion conditions. Saleh (2020) suggests that while combustion processes emit particles containing a mix of different BrC classes, smoldering biomass emissions are skewed more toward weakly absorbing BrC (W-BrC), while high-temperature biomass combustion emissions are skewed more toward moderately and strongly absorbing BrC (M-BrC and S-BrC). In the present work, some data points, mainly from cooking and heating, exhibit greater spectral variation (larger w) than that suggested for M-BrC, while falling within its k_BrC,550 range. Changing combustion conditions were observed during several experiments, where both flaming and smoldering combustion phases occurred, while particles were collected as a time-averaged filter sample. Here, the greater spectral dependence in M-BrC measurements implies that these samples would exert stronger light absorption in the near-UV range than typical M-BrC would. The thermo-optically resolved carbon fractions show a decline in the total OC fraction, mainly in OC1 and OC2 (relatively high volatility fractions), with increasing BrC absorption strength from weak to moderate (Fig. 3a). A simultaneous increase in EC highlights the dominance of BC absorption as the strength of BrC absorption increases, as also reported previously (Saleh et al., 2014; Chakrabarty et al., 2023). Relationships between BC, OC, and BrC properties, reported by Saleh et al. (2014), are useful in parameterizing BrC absorption in radiative and climate models (Brown et al., 2018; Neyestani and Saleh, 2022; Wang et al., 2018).

Figure 2 BrC–BC light absorption continuum on a semi-log scale, showing the wavelength dependence (w = AAE-1) of the imaginary part of the refractive index (k₅₅₀) versus the imaginary part of the refractive index at 550 nm (k₅₅₀), $w=\mathrm{0.1917}(\pm \mathrm{0.074})/(k_{\mathrm{BrC},\mathrm{550}}+\mathrm{0.02886}(\pm \mathrm{0.014}\left)\right)$. Here, the BC region lies between k₅₅₀ = 0.6–0.8 and w = 0–0.2. The BrC classes are defined per Saleh (2020): strongly absorbing BrC (S-BrC), moderately absorbing BrC (M-BrC), and weakly absorbing BrC (W-BrC). The arrow at the right indicates the reduction in the solubility and volatility with an increase in k₅₅₀ (Saleh, 2020). The shaded grey area represents the continuum reported by previous studies (Saleh et al., 2014; Lu et al., 2015; Luo et al., 2022; Saleh et al., 2018), and the equations for the shaded area are given in Sect. S1 and Fig. S1 in the Supplement. The right axis displays the range of wavelength dependencies for the three BrC classes.

Figure 3 (a) k_BrC,550 distribution (right) and thermo-optically resolved carbon fractions (left) with varying BrC classes based on wavelength dependence. Here, the strongly absorbing BrC (S-BrC), moderately absorbing BrC (M-BrC), and weakly absorbing BrC (W-BrC) ranges are < 1.5, 1.5–4, and > 4, respectively, in terms of wavelength dependence (Saleh, 2020). (b) Source-specific k_BrC,550 distribution (right) and thermo-optically resolved carbon fractions (left). The distributions of violin plots show the kernel density. The arrow near the legend indicates the reduction in the relative volatility from OC1 to OC4 (Ma et al., 2016). For the purpose of representation, pyrolytic carbon was assigned to OC4.

3.2 Source-specific BrC

We observed that the variability in source-specific BrC properties is larger within a source category than among different source categories. Figure 3b shows no significant changes in k_BrC,550 among different source categories. However, there are much larger differences among individual data points in a source category because of varying fuels, meteorology, and burning practices. The k_BrC,550 means from agricultural residue burning, brick production, cooking, and heating are 0.026 (± 0.035), 0.015 (± 0.026), 0.015 (± 0.003), and 0.010 (± 0.006), respectively (Fig. 3b). A large variation in k_BrC,550 was observed during agricultural residue burning, with banana, which has a high moisture content (Tock et al., 2010) showing a k_BrC,550 of 0.008, and pigeon pea (an oilseed legume), which has a k_BrC,550 of 0.082. In comparison, k_BrC,550 varies from 0.006 (final stage) to 0.022 (initial stage) during brick kiln operation and from 0.002 (crop residue) to 0.013 (firewood) during residential heating. This contrasts with cooking, where deliberate efforts are made to ensure efficient burning of fuel for meal preparation. Hence, BrC properties in cooking emissions do not vary much (k_BrC,550 = 0.015 ± 0.001). Our study observed k_BrC,365 of ∼ 0.1 (± 0.01) for cooking, which is higher than lab-measured values (0.014–0.054) for the same fuels at 350 nm (Rathod et al., 2017). We observed that MAC_BrC,365 stayed between 1.5–2.5 m² g⁻¹ for all source–fuel combinations, except for pigeon pea residue burning (MAC_BrC,365 = 4.01 m² g⁻¹). The current findings are comparable with the MAC_BrC,365 value of 2 (± 0.5) m² g⁻¹ from Indian air masses influenced by agricultural residue burning (Satish et al., 2020). The values reported in our study are in the upper range of ambient MAC_BrC,365 (0.62–2.3 m² g⁻¹) reported previously over India (Sarkar et al., 2019; Shamjad et al., 2018; Satish et al., 2020; Rastogi et al., 2021; Rana et al., 2020; Kirillova et al., 2016; Dey et al., 2021), which could be due to photobleaching of ambient BrC that decreases MAC. However, our estimation of MAC_BrC,365 aligns well with the previously reported source-specific values (1.09–2.53) (Pandey et al., 2020; Debbarma et al., 2024; Rathod et al., 2017). The observed AAE_BrC (∼ 5.23 ± 1.51, range 2–8.5; see Table S1) is comparable with previous observations (∼ 5.31 ± 1.67, range 2.3–6.8) for biomass burning over India (Islam et al., 2022; Pandey et al., 2020; Rathod et al., 2017; Satish et al., 2020). In agricultural residue burning, banana residue shows the lowest k_BrC,550 (0.008) and BC-to-OA ratio (0.030) (Table 2 in the Supplement). In contrast, pigeon pea residue burning has the highest k_BrC,550 (0.082) and BC-to-OA ratio (2.054). A similar relationship between k_BrC,550 and BC-to-OA ratio has also been observed in other source–fuel combinations and has been used to parameterize k_BrC,550 and w (Fig. 4).

Figure 4 (a) Mean K_BrC,550 versus the BC-to-OA ratio (k_BrC,550 = $\mathrm{0.0365}(\pm \mathrm{0.006})\times (\mathrm{BC}/\mathrm{OA})+\mathrm{0.0047}(\pm \mathrm{0.0037})$; R² = 0.93) and (b) w versus the BC-to-OA ratio ($w=\mathrm{5.355}(\pm \mathrm{0.50})\times \exp (-\mathrm{0.428}(\pm \mathrm{0.25})\times (\mathrm{BC}/\mathrm{OA}\left)\right)$; R² = 0.60). Here, OA = OC × 1.8: the factor 1.8 is widely used to convert OC into OA (Turpin and Lim, 2001; Chow et al., 2015; Navinya et al., 2020; Provençal et al., 2017; Kumar et al., 2023). The shaded grey area represents the relationship reported by previous studies (Saleh et al., 2014; Lu et al., 2015; Luo et al., 2022), and the equations for the shaded area are given in Sects. S2 and S3 and Fig. S1). The right axis in (b) displays the ranges of wavelength dependencies for the three BrC classes.

3.3 Parameterization of k_BrC and w

We leveraged the significant correlation (p value < 0.01) between the BC-to-OA ratio and the BrC properties (k_BrC,550, R² = 0.93; w, R² = 0.60) to build a relationship between these quantities. Despite the variety of fuel burning technologies used, such as traditional stoves, open residue burning, and brick clamps, k_BrC,550 variability is explained (R² = 0.93) by the BC-to-OA ratio. We observed that k_BrC,550 varies linearly from 0.006 to 0.74 for BC-to-OA ratios of 0 to 20 (Fig. 4a). Similarly, we explain w using the BC-to-OA ratio to provide an approximation of the BrC absorption over the different wavelengths. We observed an exponential relation between w and the BC-to-OA ratio with an R² of 0.60 (w varies from 5 to ∼ 0 for BC-to-OA ratios of 0 to 20, respectively) (Fig. 4b). Relatively to the present studies, the relationship used in climate modeling studies (Zhang et al., 2020; Neyestani and Saleh, 2022; Brown et al., 2018) given by Saleh et al. (2014) would overestimate the k_BrC,550 over South Asia (Fig. S7 in the Supplement). In contrast, previous studies (Saleh et al., 2014; Lu et al., 2015; Luo et al., 2022) underestimate the range of w values observed in this study, which may result in an underestimation of k_BrC,365 (Fig. S7). Such an underestimation would propagate uncertainties to radiative forcing calculations, especially over South Asia.

3.4 Spatial differences in k_BrC,365 and w

Several studies have reported ambient BrC absorption in the South Asian region (Dey et al., 2023; Srinivas and Sarin, 2013, 2014; Bikkina et al., 2020; Boreddy et al., 2021; Choudhary et al., 2017, 2018, 2022; Dasari et al., 2019; Dey et al., 2021; Kirillova et al., 2016; Mukherjee et al., 2020; Rajeev et al., 2022; Rastogi et al., 2021; Rana et al., 2020; Shamjad et al., 2016, 2018), while most climate models continue to consider weakly absorbing BrC absorption (Sand et al., 2021; Feng et al., 2013), regardless of sources and combustion conditions. Feng et al. (2013) simulated global BrC absorption using k_BrC values that are 2- to 5-fold weaker than those observed in our study, and they noted underestimation of BrC absorption efficiency over South Asia owing to the presence of strongly absorbing BrC. Other studies (Brown et al., 2018; Zhang et al., 2020) have used k_BrC,550 values (Saleh et al., 2014; Mcmeeking, 2008) that are 2- to 3-fold higher than those observed in this study to simulate the global radiative impact of BrC. Hence, neglecting the spatial variability in k_BrC could lead to bias in understanding its radiative impact. Thus, we calculated emission-weighted BrC optical properties across the Indian region to demonstrate their spatial heterogeneity in this region. The relationships shown in Fig. 4 were used to make a spatial map of k_BrC,550, $k_{\mathrm{BrC},\mathrm{365},}$ and w, with emission strength from the COALESCE SMoG-India emission inventory (Tibrewal et al., 2024). SMoG-India is a multi-sectoral, multi-pollutant data set available at a 5 km grid resolution, developed under the COALESCE network (Venkataraman et al., 2020), which also facilitated the collection of samples used in the present study.

Figure 5 The spatial distribution of (a) k_BrC,365 and (b) wavelength dependence (w). The BC-to-OA ratio is taken for the year 2019 from the COALESCE SMoG-India emission inventory (Tibrewal et al., 2024). Publisher's remark: please note that the above figure contains disputed territories.

Figure 1 shows the large OC emissions over the Indo-Gangetic Plain, with annual emissions ranging from 50–70 Mg yr⁻¹ per pixel (pixel size is 5 km × 5 km), while other regions emit ∼ 10–20 Mg yr⁻¹ per pixel. Emission-weighted spatial information about w (range 4.3–5.3) and k_BrC,550 (0.006–0.023) aids in the estimation of k_BrC,365. Figure 5a shows k_BrC,365 ranges from 0.05 to 0.14, indicating strong absorption in the UV–Vis wavelengths. The Himalayan foothills show large k_BrC values compared to other parts of India, mainly due to high BC-to-OA emissions from the predominant heating activity. A recent study highlighted the low photobleaching rate of BrC near Himalayan regions due to the low ambient temperatures (Choudhary et al., 2022). The coincidence of dark-BrC particle emissions in this study, along with their reported extended lifetimes, could result in snow darkening upon deposition along with accelerated snowmelt and glacier melt (Chelluboyina et al., 2024). The northwestern region of India exhibits the highest OC emissions from agricultural residue burning (Fig. S5), primarily from straw residue burning (Kapoor et al., 2023b), which has a relatively low BC-to-OA ratio. Consequently, the k_BrC remains lower compared to its values in other regions, such as Maharashtra and Andhra Pradesh, where oilseed crop burning is prevalent (Kapoor et al., 2023b), resulting in a higher BC-to-OA ratio and higher k_BrC values. Heating activities are particularly intense in the colder areas, especially in the Himalayan foothills, with higher use of firewood in the eastern India (Navinya et al., 2023), leading to significantly higher BC-to-OA ratios and elevated k_BrC in the northern and eastern regions (Fig. S3). In the central Indo-Gangetic Plain, particularly in Uttar Pradesh and Bihar, dung cake is more commonly used for heating (Navinya et al., 2023), which contributes to very low k_BrC values. The variation in the BC-to-OA ratio across India due to cooking activities is minimal (0.075–0.125) compared to that from agricultural residue burning (0.025–0.2) and heating (0.025–0.25), resulting in substantially low spatial variation in k_BrC,365 (0.06–0.08) from cooking (Fig. S3). The k_BrC,550 values of combustion aerosol emissions from India vary from 0.006 to 0.023 (Fig. S6 in the Supplement), with some hotspots scattered across the country. These numbers highlight the order-of-magnitude increase in k_BrC,365 compared to k_BrC,550, with higher values over eastern and northern India. An earlier investigation also noted elevated modeled BrC absorption in the eastern regions of India (Zhu et al., 2021). The substantial emissions of BrC across the country, coupled with the high k_BrC values observed in certain other regions, suggest that BrC particles may have significant radiative impacts over the region.

4 Implications

The variability in k_BrC,near-UV across modeling studies, ranging from 0.045 (Zhang et al., 2020) to 0.168 (Lin et al., 2014), arises from methodological, fuel, and burning condition disparities in the studies reporting BrC absorption properties from lab-based biomass combustion (Kirchstetter et al., 2004; Chen and Bond, 2010; Lack et al., 2012). However, our study, using field measurements of a variety of sources, introduces source- and fuel-specific k_BrC values, enhancing modeling capabilities for a more nuanced understanding of the radiative and climate impacts of BrC. Additionally, the observed varying wavelength dependence (w), linked with the BC-to-OA ratio in this research, amplifies uncertainty when it is assumed to be constant in models (Zhang et al., 2020). Compared to the findings of this study, typical BrC parameterization schemes (Saleh et al., 2014; Lu et al., 2015; Luo et al., 2022) in climate models tend to overestimate k_BrC,550 while substantially underestimating wavelength dependence, which may misrepresent near-UV BrC absorption in world regions with biomass combustion emissions resembling those in South Asia. Additionally, this study's findings aid in pinpointing biomass fuels and activities, including burning of some agricultural residues and residential space heating, that are both prone to emitting more strongly absorbing BrC (k_BrC,550 > 0.1) and prevalent across developing nations. These variations in k_BrC with sources and fuels lead to spatial variations in emitted BrC properties. In the Himalayan foothills, residential space heating produces more strongly absorbing (and more persistent) BrC emissions, and the deposition of these emissions increases the potential risks of increased snow darkening and accelerated glacier melting. Leveraging this information with emission inventories enables the identification and potential interventional targeting of these biomass fuels and activities, with the goal of reducing both their local health impacts and their global climate impacts.