Cattle ruminant samples in South Asia were collected using a custom-built sampling instrument. Sample air was passed through magnesium perchlorate (CAS# 10034-81-8, Alfa Aesar) to remove moisture, into an electrically-powered membrane pump (KNF Neuberger N86), and out into two cylindrical 1000 mL borosilicate 3.3 glass flasks (Normag, Germany) with axial inlet and outlet, connected in series. The inlet and outlet of each flask were sealed with a Normag needle valve with high-diffusion-minimized sealing. Tubing was made of PTFE and Synflex^® and connections were Swagelok^® and UltraTorr™. The flasks were pre-conditioned with clean air (Strandmøllen, 20.9 % oxygen, and 79.180 % nitrogen, CnHm≤3ppm, CO2≤1ppm, CO≤1ppm, H2O≤3ppm) to eliminate contaminants. Before sampling, the flasks were conditioned in a 4-step protocol: Evacuated at high vacuum at 50 °C for 12 h, purged with nitrogen at 50 °C for 2 h, again evacuated at high vacuum at 50 °C, for 3 h, and finally filled with pre-conditioned clean air to a pressure of 1.3 bar (absolute). Sampling was conducted by positioning a funnel 2–5 cm from the cattle's mouths to capture their breath. The sample air was pumped through the flasks for 5 min, then closing the outlet valve and letting pressure build up to 1.7 bar (absolute), after which the flask valves were closed. Finally, flask in- and outlets were sealed with parafilm to prevent contamination from dust etc.

For combustion sources, we collected exhaust samples from agricultural crop residue burning in South Asian fields using the same custom-built instrument. Sampling was performed 3–15 cm from the burning rice paddies. A 0.45 µm inline gas filter was placed between the PTFE tubing and the metal tubing to remove aerosols. Each sampling session lasted 5 min, with the final flask pressure reaching 1.2 bar (absolute).

Samples were also collected to constrain the isotope fingerprints of aqueous microbial sources in South Asia, including rice paddies and wastewater. Rice paddy sampling involved dividing each paddy into four quadrants and taking one to three replicate samples from the center of each quadrant, totaling 4–12 samples per paddy field. For wastewater, three replicate samples were collected from sewage at each location. Before sampling, glass vials (VMR) were rinsed thrice with 125 mL of either rice paddy water or wastewater. Samples were then collected by submerging the vials to mid-depth (approximately 20 cm depth, the exact depth varied depending on the flooding conditions in individual paddy fields) for 20 s until bubbling ceased, followed by an additional five-second hold. The vials were then sealed with a bromine butyl rubber stopper (Apodan Nordic) attached to a string. After sampling, 0.5 mL of saturated ZnCl2 solution was added as a preservative, and the vials were crimp-sealed, labeled, and stored at 4 °C in the dark before and after being shipped to Stockholm University for further analysis.

Thus, we collected a substantial number of methane samples from the four sources: ruminants, biomass burning, rice paddies and wastewater (see Data S1 in the Supplement for details of each sample). Among them, ruminants and biomass burning represent two major sources of gaseous methane, while rice paddies and wastewater are significant atmospheric sources of aqueous, dissolved methane. The ruminant samples were obtained from 6 farms across South Asia, totaling 40 samples. For biomass burning, we conducted 4 sampling campaigns in different regions, collecting a total of 17 samples. Rice paddy samples were collected from 18 different rice-growing areas, amounting to 185 samples. Wastewater samples were gathered from 13 sewage treatment plants, totaling 38 samples. The sample distribution is illustrated in Fig. 1, with gaseous methane samples from biomass burning and ruminants primarily collected in Bangladesh, while aqueous methane samples from rice paddies and wastewater are distributed across Bangladesh and several densely populated regions of India. The background color of Fig. 1 represents total methane fluxes in 2023, sourced from EDGAR (Crippa et al., 2024), indicating significant methane emissions in South Asia.

Methane mixing ratios were measured using gas chromatography with flame ionization detection (GC-FID, Agilent Technologies 7890A). For gaseous source samples, methane was extracted from a glass flask using a syringe and injected directly into the instrument. For aqueous source samples, a portion of the liquid was extracted, and helium (He) was introduced. After equilibration for 2 h, a syringe was used to collect the headspace mixture of helium, methane and other dissolved gases for analysis. Three methane standards with multiple concentrations (1.6 ppm ±2%, Air liquid; 80.3ppm±2.0ppm, Linde; 250ppm±0.5%, Strandmøllen; 95 % confidence) in synthetic air were used for calibration.

The equilibrium between the gaseous and aqueous phases was evaluated using Henry's Law (Eq. 1): 1c=k×P where c is the concentration of dissolved methane (nmolL-1), k is Henry's law constant, and P is the partial pressure of methane. For the calculations: the water volume was 40 mL, the headspace volume was 10 mL, the headspace pressure was 1 atm, the equilibration temperature was 25 °C, the gas constant R was 0.08025 atmLmol-1K-1, and Henry's Law constant k for methane at 25 °C was 0.0014 molL-1atm-1.

Once the methane mixing ratios were determined, gaseous and aqueous source samples were analyzed for δ13C and δ2H using gas chromatography isotope ratio mass spectrometry (GC-IRMS; Delta V Plus, Thermo Fisher). Due to variable methane mixing ratios in source samples, two methods were used: pre-concentration (Precon) for diluted samples (Rice et al., 2001) and direct injection, using the GC injector, for concentrated samples. The Precon system was modified with custom-built components to improve isotopic analysis. In this configuration, only liquid nitrogen was used as the cryogen for all traps. CO2 and water vapor were first removed with chemical absorbents, followed by Trap 1 for additional purification. Trap 2 (a 1/8′′ stainless steel tube, 20 cm in length, packed with HayeSep D, mesh size 60–80) was then employed, with sufficient venting through the Precon six-port valve to remove most of the residual oxygen that could interfere with δ2H measurements. The sample was subsequently transferred to Trap 3 (a PoraPLOT capillary, 0.32 mm internal diameter), and final separation was performed on a 5m×0.32mm PoraPLOT column at -78°C (dry ice). This procedure ensured effective resolution of the methane peak from any remaining oxygen before conversion in the high-temperature reactor. Any interference by krypton (Kr) in the δ13C analysis was eliminated by post-column GC separation from the methane-derived carbon dioxide peak (PoraPLOT 7m×0.32mm; Schmitt et al., 2013). To match the relatively narrow detection range of the IRMS, syringe dilutions with He were applied. Isotopic values were corrected for instrumental drift and calibrated using standards.

Isotope values are reported in δ notation, representing the relative deviation of isotope abundance in a sample compared to international standards: Vienna Pee Dee Belemnite (V-PDB) for δ13C and Vienna Standard Mean Ocean Water (V-SMOW) for δ2H. For diluted samples, the two standards used were both 1.85 ppm, with δ13C values of -48.4±0.3‰ and -68.6±0.3‰, and δ2H values of -63±5‰ and -240±5‰. For concentrated samples, δ13C was measured directly using a 100 ppm standard with a δ13C value of -43.8‰, while δ2H was measured after pre-dilution and corrected using the same approach as for diluted samples. Analytical uncertainties of the reported isotopic composition are 0.09 ‰ for δ13C and 2.1 ‰ for δ2H. The here constrained isotopic data of the major methane sources in South Asia are summarized in Data S1.

To determine the isotopic values of the sources, we analyzed the isotopic data for all samples using the Keeling (Keeling, 1958; Pataki et al., 2003) and Miller–Tans (Miller and Tans, 2003) methods. These approaches follow the Eqs. (2) and (3): 2δobs=cbg×(δbg-δsource)×1cobs+δsource3δobs×cobs=δsource×cobs+cbg×(δbg-δsource) where c represents the CH4 mixing ratio, and the subscripts obs, bg, and source denote atmospheric observations, background levels, and source contributions, respectively. The Miller–Tans approach, which yielded narrow uncertainties, was used in the main text, while the Keeling plots are provided as additional information in Figs. S1–S4 in the Supplement.

Both Keeling and Miller–Tans approaches are fundamentally based on isotopic mass balance: 4δobs=cbg×δbg+csource×δsourcecbg+csource When the source concentration is much higher than the background: 5csource≫cbg Equation (4) simplifies to: 6δobs≈δsource This implies that when source concentrations are sufficiently high, the observed isotopic composition approaches that of the source. Therefore, high-concentration data points alone can provide a good approximation of the source isotopic signature, even without applying Keeling or Miller–Tans analyses.

We employed Kriging interpolation using the gstat package in R to evaluate the spatial distribution of isotopic values. This geostatistical method estimates values at unsampled locations based on the spatial autocorrelation of observed data, modeled through a fitted variogram. We applied this approach to interpolate δ2H values of global surface water and representative microbial methane sources (ruminants, wetlands and rice paddies, and waste) for comparative spatial analysis.

To calculate methane isotopic source signatures and integrate contributions from multiple sources, we used a combination of statistical approaches. Uncertainty propagation was quantified using Monte Carlo simulations (10 000 iterations), accounting for variability in both isotopic measurements and source fractions.

A comprehensive literature review was conducted to compile isotopic source signatures, which were further assessed for major global and regional methane sources (Data S2 in the Supplement). The review was carefully curated to minimize the influence of individual studies by selecting only a single representative value per region from each publication. Source-specific mathematical approaches were applied, as detailed in the following sections.

In the final section, we integrated the synthesized isotopic signatures with a range of top-down and bottom-up estimates to evaluate the discrepancies between current emission inventories and isotopic source constraints. Global isotopic data were compiled from our extensive literature review (Data S2). Isotopic values for microbial sources were calculated using Monte Carlo simulations, integrating our findings with estimates from Saunois et al. (2025), Ito et al. (2023), and the IPCC (Masson-Delmotte et al., 2021) assessment. For South Asia, we incorporated isotopic signatures of rice paddy methane, while for natural wetlands, we retained tropical values from the global review, as there is no evidence indicating significant methane oxidation in South Asian wetlands. Thermogenic methane isotopic values were sourced from extensive global (Sherwood et al., 2017) and European (Menoud et al., 2022) databases. The South Asian dataset focuses on methane sources across Afghanistan, Bangladesh, India and Pakistan. Thermogenic methane primarily originates from natural gas, coalbed methane, shale gas and other methane emissions associated with fossil fuel extraction, transportation and processing. This thermogenic category also includes minor contributions from biogenic methane present in various mineral deposits, incorporated to facilitate the source analysis of atmospheric methane.

The isotopic source signatures of methane from agricultural crop residue burning (C3 biomass) in South Asia was constrained and compared to measurements elsewhere (Fig. 2, Table 1, and Data S1–S2) to establish robust and representative source end-member values. The δ13C and δ2H values derived from Miller–Tans plots (Fig. 2A and B) were -30.9±2.2‰ and -201±18‰, respectively. Keeling plots yielded comparable δ13C values but slightly more enriched δ2H values (Fig. S1). Keeling and Miller–Tans plots are two formulations of the isotopic mass balance (Eq. 4), differing primarily in their treatment of background contributions. The Keeling approach (Eq. 2) derives the source signature from the intercept but is sensitive to background variability through its effect on the slope, which can distort linearity. In contrast, the Miller–Tans formulation (Eq. 3) derives the source signature from the slope, with background variability mainly affecting the intercept and increasing scatter while largely preserving linearity. As both methods rely on linear regression, increased scatter is generally less detrimental than distortion of linearity, making the Miller–Tans approach more robust in practical applications. Both approaches are most reliable when source-driven variability dominates over background variability. In our case, some high-concentration observations approach the condition csource≫cbg (Eq. 5), leading to δobs≈δsource (Eq. 6), such that the influence of atmospheric background variability becomes negligible. Further discussion of background effects is provided in Sect. S2 in the Supplement. In addition, the linear relationship between δ2H versus δ13C showed that the isotopic composition was influenced by mixing with atmospheric methane, with a gradient reflecting the transition from source to atmospheric background values (Fig. 2C).

To minimize bias from overrepresented datasets in specific regions, our global review consolidated data from each study and region into a single representative value (Fig. 2D and Data S2). There appeared to be a significant δ13C difference between methane emissions from C3 and C4 biomass combustion globally (Vernooij et al., 2022; Nisbet et al., 2022), presumably driven by the differing δ13C content of the feedstocks (Yao et al., 2022). By weighting δ13C values according to the global proportions of C3 and C4 vegetation (77 % and 23 %) (Still et al., 2003), we derived a global biomass-type-weighted mean δ13C value of -25.0±2.1‰. In contrast, the δ2H values of methane from C3 vs. C4 biomass burning did not exhibit a clear distinction (Fig. 2E), suggesting that δ2H was not strongly influenced by biomass type. The mean δ2H value for global biomass burning methane was -222±39‰. Some studies have shown that δ2H in surface water exhibits spatial (latitudinal) variability (Zakharov et al., 2004; IAEA/WMO, 2023), which would logically also influence δ2H signatures of biomass burning. However, available δ2H source signatures for methane remain limited, preventing further differentiation at present.

Given that δ13C variability in methane from biomass burning is influenced by the relative contributions of C3 and C4 biomass, these factors must be carefully considered when characterizing atmospheric-receptor isotopic signatures in specific regions. Based on our previous isotopic source apportionment of elemental carbon (EC) in South Asian atmospheric aerosols, C3 and C4 biomass combustion accounted for 90 % and 10 % of EC in winter, respectively (Dasari et al., 2020). Since EC and methane are co-emitted during combustion, a first approximation is that they may have the same proportional contributions. Using the isotopic values measured for C3 combustion in South Asia, the global mean for C4 combustion, and the regional C3 / C4 ratio, we derived a C3 / C4-weighted δ13C value of -29.5±2.0‰ for South Asia (Table 1). In contrast, δ2H was not influenced by C3 / C4 composition and does not require such adjustment. Overall, methane from biomass burning in South Asia was more depleted in δ13C and more enriched in δ2H than the global mean (-25.0±2.1‰).

Global wildfire-related methane emissions may be underestimated due to undetected small fires (Zhao et al., 2025), highlighting the need for top-down constraints of biomass burning emissions. Estimates of methane emissions from tropical biomass burning spanned a wide range of 14–34 Tgyr-1 (Kirschke et al., 2013), highlighting the importance of alternative approaches for methane assessment in South Asia. A recent study reported δ13C values of CH4 from tropical biomass burning, ranging from -12‰ to -16‰ for grassland fires and -16‰ to -28‰ for farmland fires (Nisbet et al., 2022), which align with global estimates. The relative proportions of C3 and C4 biomass remain a key determinant of isotopic signatures globally, while geographic variations have a minor influence. Additionally, combustion conditions and fuel moisture content can influence isotopic signatures, necessitating additional research to refine isotopic source characterization (Vernooij et al., 2022).

In South Asia, biomass burning is dominated by agricultural residue combustion and other fire types, such as wildfires and forest fires, and are expected to have similar methane isotopic signatures. Other combustion sources, such as traffic and coal combustion, contribute modestly to methane emissions but exhibit δ13C signatures of their raw materials similar to C3 biomass (Yao et al., 2022). Improved isotopic characterization of these sources can enhance source attribution. In South Asia, biomass burning emissions displayed more depleted δ13C and enriched δ2H values than global means reported from elsewhere, reflecting regional variations in fuel type and C3 / C4 biomass composition. Region-specific isotopic endmembers are therefore critical for accurate source apportionment.

The isotopic source signatures of ruminant methane from South Asia were constrained and compared with such measurements globally (Fig. 3, Table 2, Data S1–S2). The δ13C and δ2H values derived from Miller–Tans plots (Fig. 3A–B), yielded -68.7±0.5‰ (primarily reflecting C3 biomass) and -343±6‰, respectively. Keeling plots yielded comparable δ13C and δ2H values (Fig. S2). The relationship between δ2H and δ13C showed a clear gradient as the isotopic composition transitions from the source to the atmospheric background (Fig. 3C).

Methane isotopic values from global ruminant sources were summarized from the literature (Fig. 3D), revealing a notable δ13C difference between C3 and C4 diets, driven by the distinct δ13C content of these feedstocks. By weighting δ13C values according to the global proportions of C3 and C4 diets (70 % and 30 %) from a recent database study (Chang et al., 2019), we calculated a global C3 / C4 biomass-weighted mean δ13C value of -63.8±2.4‰. In contrast, the δ2H values for methane from ruminants globally showed no clear differentiation between C3 and C4 diets (Fig. 3E). The δ2H signature of methane is expected to be primarily derived from surface water, and thus may exhibit regional variability. The global mean δ2H value (-311±46‰) likely reflects this variability, which may arise from differences in the isotopic composition of environmental water as well as variations in rumination processes.

Methane emissions from C3-fed ruminants in South Asia (-68.7±0.5‰, Fig. 3A) were more depleted in δ13C than the global mean of C3-fed ruminants (-67.0±3.0‰, Fig. 3D). However, regional variability in C3 / C4 feed composition was an equally important factor that must be considered when determining the representative isotopic signature for South Asian ruminants. Based on a database study (Chang et al., 2019), ruminant diets in South Asia consisted of approximately 65 % C3 % and 35 % C4 plants. Using the isotopic values measured for C3 diet ruminants in South Asia, the global mean for C4 diet ruminants, and the regional C3 / C4 ratio, we derived a C3 / C4-weighted δ13C value of -63.3±1.1‰ for South Asia, which is comparable to the global C3 / C4-weighted mean (-63.8±2.4‰). In contrast, δ2H signatures showed a substantial discrepancy, with depletion exceeding by 32 ‰ in South Asia compared to the global mean, underscoring the importance of determining and using regionally-constrained source fingerprints in isotope-based source apportionment studies.

Recent studies have indicated that biogenic methane emissions have increased in the tropics, with considerable emissions from agricultural activities such as ruminant livestock farming and rice cultivation (Schaefer et al., 2016; Nisbet et al., 2025). South Asia, home to the world's largest ruminant stock, is potentially one of the major contributors to these emissions (Ganesan et al., 2017). Isotopic source fingerprinting to characterize ruminant methane emissions in the tropics and South Asia offers a promising approach to place quantitative constraints on the importance of ruminant and other sources. Isotopic source signatures must be carefully adjusted based on regional dietary compositions and environmental conditions, as the prevalence of C4 vegetation in tropical regions results in more enriched δ13C values in some areas, such as -57‰ in Kenya (Nisbet et al., 2022), -52‰ to -57‰ in Zimbabwe (Brownlow et al., 2017), -60‰ to -63‰ in Australia (Lu et al., 2021), and -65‰ in sub-Saharan Africa (Chang et al., 2019). Additionally, methane from ruminants is primarily produced in the rumen through enteric fermentation and then exhaled (Hook et al., 2010), but cattle are not the only ruminants contributing to methane emissions. While cattle are a major source, other domesticated species, including buffalo, sheep, and goats, as well as wild ruminants such as deer, also contribute substantially to methane emissions. Incorporating these additional ruminant sources may help develop a more comprehensive isotopic characterization. Ruminant methane showed similar δ13C source signatures globally but displayed distinct δ2H values in South Asia that deviate from the global mean (still within the uncertainty). Taken together, also for the ruminant releases, isotope-based source apportionment of atmospheric methane should employ region-specific endmember values.

The isotopic signatures of methane from South Asian rice paddies were quantified and compared with global values (Fig. 4, Table 3, and Data S1–S2). The δ13C and δ2H derived from Miller–Tans plots were -53.8±0.8‰ and -311±6‰, respectively (Fig. 4A and B). In contrast, the Keeling plots showed reduced linearity and more enriched δ13C and δ2H values (Fig. S3), reflecting the complexity of methane production and processing in rice paddies (Sect. S3.2 in the Supplement). The sample concentration range spanned several orders of magnitude and some high-concentration samples satisfied the condition csource≫cbg (Eq. 5), yielding δobs≈δsource (Eq. 6), and their isotopic values still exhibited noticeable variability, indicating the coexistence of multiple methane sources and/or the influence of in situ oxidation within the water column. Both Keeling and Miller–Tans methods are fundamentally designed for single-source perturbations; in multi-source systems, they tend to be biased toward the highest-concentration source (Monte Carlo mixing simulation in Sect. S3.1 in the Supplement), while weaker sources are suppressed or even negligible when concentration differences are large. In the Keeling method, background contributions are incorporated into the slope (Eq. 2). Under multi-source conditions, lower-concentration methane sources do not represent true background, but their influence becomes effectively indistinguishable from background variability within the Keeling framework. As a result, the combined variability of background and lower-concentration sources became significant in rice paddy samples, leading to deviations from linearity and reduced robustness. In contrast, the Miller–Tans formulation incorporates background into the intercept (Eq. 3); when concentration differences spanned several orders of magnitude for rice paddy methane, the slope was primarily controlled by the highest-concentration source, resulting in a more stable and interpretable relationship. Consistently, alternative statistical approaches (quantiles, arithmetic means, and concentration-weighted means) yielded more enriched isotopic signatures than the Miller–Tans method (Fig. 4D–E). Among them, the concentration-weighted mean (δ13C=-45.3±12.3‰, δ2H=-250±71‰) likely reflected methane dissolved in floodwater.

A significant linear relationship between δ13C and δ2H (Fig. 4C) further supports the presence of methane oxidation, consistent with isotopic enrichment associated with methanotrophic activity (Schaefer and Whiticar, 2008). In rice paddies, only 1 %–2 % of methane is emitted via diffusion through floodwater, whereas ∼90% is transported via plant-mediated pathways (aerenchyma) and 8 %–9 % through ebullition (Cicerone and Shetter, 1981; Schütz et al., 1989; Smartt et al., 2016; Li et al., 2025). Plant-mediated transport primarily transfers methane from subsurface anoxic layers and is therefore generally less affected by oxidation. However, due to mixing and circulation within the water column and the presence of oxic zones near roots, partially oxidized methane may also be entrained and transported through plant aerenchyma. Ebullition is also less affected by oxidation, while the diffusion pathway is more susceptible to isotopic enrichment through oxidation.

Previous studies have primarily relied on atmospheric sampling, whereas this study focuses on aquatic measurements, raising questions of representativeness. Key challenges include the presence of multiple sources, oxidation processes, and transport pathways. These issues are discussed in detail in Sect. S3.3–S3.4 in the Supplement. Briefly, both atmospheric and aquatic sampling may be subject to representativeness biases, as the Keeling and Miller–Tans methods are dominated by the highest-concentration source (Sect. S3.1), while contributions from lower-concentration sources may be indistinguishable from background variability within the fitting framework. Nevertheless, the Miller–Tans estimates are considered to best represent the isotopic signature of the dominant, minimally oxidized methane source and are therefore adopted as the most consistent metric across sampling approaches.

The global compilation of δ13C and δ2H values of methane emissions from rice paddies and wetlands revealed similar isotopic signatures of these two aquatic sources (Fig. 4F and G). The global mean δ13C and δ2H values for rice paddies were -59.8±5.3‰ and -324±18‰, respectively, while these for wetlands were -60.0±7.6‰ and -309±49‰. Both sources exhibited clear latitudinal trends, with more enriched isotopic signatures in tropical regions and more depleted values in boreal zones. These patterns were consistent with previous observations, which attributed the depletion in boreal wetland δ13C to reduced oxidation and the absence of C4 vegetation (France et al., 2022; Brownlow et al., 2017; Tyler et al., 1988; Fisher et al., 2017; Ganesan et al., 2018). In tropical and temperate zones, δ13C values for rice paddies and wetlands were nearly identical. However, due to the absence of rice paddies in boreal regions, the global mean δ13C value for rice paddy methane appeared slightly more enriched compared to that from wetlands. Conversely, global mean δ2H value was slightly more depleted, potentially reflecting data availability biases, as boreal wetlands exhibited the most depleted δ2H values. Methane from South Asian rice paddies (Miller–Tans values) was notably more enriched in δ13C compared to the global mean, while δ2H values are slightly enriched than the global mean. This enrichment was consistent with previous regional measurements (e.g., δ13C=-54.3‰ and -57.2‰; Rao et al., 2008) and might reflect the differences in precursor composition or pre-emission oxidation under South Asian field conditions.

Methane formation in rice paddies and wetlands primarily occurs via acetoclastic (acetate fermentation) and hydrogenotrophic (CO2 reduction with H2) pathways. The hydrogenotrophic pathway typically yields methane with more depleted δ13C values, whereas acetoclastic methanogenesis produces methane with relatively enriched δ13C values (Whiticar et al., 1986). The dominant pathway varies with substrate availability, temperature, and redox conditions across wetland and lake types. In wetlands, methane is also emitted through plant-mediated transport (∼30%-90%; more than 90 % in some studies), ebullition (up to ∼60%; more than 90 % in non-plant systems), and diffusion (up to ∼30%) (Van Der Nat and Middelburg, 1998; Ding et al., 2002; Jeffrey et al., 2019; Villa et al., 2020; Ma et al., 2017), similar to rice paddies but with varying pathway contributions. Both methane source pathways and oxidation processes influence the isotopic composition of these aquatic emissions, although the extent of these effects remains uncertain and requires further study. Given the broad spatial coverage of our dataset, the Miller–Tans values for rice paddy methane reflected minimally oxidized isotopic signatures and were considered regionally representative. In contrast, isotopic values for wetland methane require further evaluation; currently, literature-based values from tropical regions are recommended. Given that approximately half of global methane emissions originate from aquatic ecosystems (Rosentreter et al., 2021) and South Asia accounts for ∼20% of global rice production (Ganesan et al., 2017), applying region-specific isotopic source signatures and reducing the uncertainties are essential for accurately constraining methane emissions in South Asia.

The isotopic source signatures of methane were constrained from South Asian wastewater and compared with global wastewater sources (Fig. 5, Table 4, and Data S1–S2). The δ13C and δ2H values derived from Miller–Tans plots (Fig. 5A and B), yielded -46.4±1.2‰ and -355±5‰, respectively. Although Keeling plots exhibited moderate linearity and may be less reliable, they yielded similar δ13C and enriched δ2H values (Fig. S4). Methane oxidation would be expected to produce a characteristic co-enrichment trend in both isotopes. However, no clear relationship between δ13C and δ2H was observed for methane in wastewater (Fig. 5C). This lack of a systematic isotopic trend suggested minimal oxidation, indicating that degradation processes prior to release were limited for wastewater methane. The methane isotopic signatures were compared for isotopic quantiles, arithmetic means and concentration-weighted means (Fig. 5D and E). The median- and concentration-weighted means aligned closely with the values obtained from Miller–Tans plots, further supporting their reliability.

A global review of δ13C and δ2H values was conducted for methane emissions from waste sources (Fig. 5F and G), i.e., wastewater, landfills and other sources. The results indicated minor differences, suggesting that δ13C and δ2H signatures were not significantly distinct among various waste sources. Methane from global waste sources had mean δ13C and δ2H values of -54.0±5.4‰ and -295±18‰, respectively. Slight differences existed between methane emissions from wastewater and landfills, with wastewater showing more enriched δ13C and slightly more depleted δ2H values. Other sources, such as composting, biogas fermentation and other organic waste decomposition (Lu et al., 2021; Bakkaloglu et al., 2022), exhibited wider range of values. Nonetheless, our findings showed that methane isotopic signatures from waste sources were consistent globally, which facilitated isotopic source apportionment. This similarity may be attributed to similar methane production mechanisms across these sources. Additionally, the narrow range of δ13C values for global waste methane suggested minimal latitudinal variation, making further differentiation unnecessary. However, in South Asia, methane from wastewater was more enriched in δ13C and depleted in δ2H compared to the global mean values.

Methane emissions from waste sources were estimated to contribute approximately 12 % of global anthropogenic emissions (Saunois et al., 2025). In South Asia, landfill methane emissions were particularly significant (Chakraborty et al., 2011), and atmospheric data also suggested that the waste sector played a key role in regional methane emissions, as supported by δ13C constraints (Metya et al., 2022). Emissions from waste sources were also influenced by a range of factors, including microbial communities, temperature, pH, the CH4 / O2 ratio, nutrient levels and inhibitory chemicals (Polag et al., 2015; Nisbet et al., 2020; Woolley Maisch et al., 2025). Additionally, studies indicated that the operational status of landfills (active or closed) can influence the carbon isotopic signature (Bakkaloglu et al., 2022). However, our global review showed only minor distinctions among various waste sources, suggesting that the isotopic signatures we measured in South Asia should be representative for wastewater in the region. Further exploring other waste sources and various factors may improve our understanding of methane emissions from the waste sector. Although isotopic signatures of methane from waste sources showed limited variability globally, values in South Asia deviated significantly from the global mean. This highlights the need for region-specific isotopic endmembers also for waste sources in methane source apportionment studies.

There are geographic variations in methane isotopic compositions across the globe for any source class due to a combination of environmental factors and source materials. The isotopic signatures of microbial methane vary across regions due to multiple factors, including differences in raw materials, methanogenic pathways (Whiticar et al., 1986; Conrad, 2005), and the methane oxidation by methanotrophic bacteria. These factors are essential to consider and suggest that region-specific and sometimes system-specific isotope source fingerprinting are necessary to facilitate accurate isotope-based source apportionment. Previous studies identified correlations between methane isotopic values and regional environmental factors (Sherwood et al., 2017; Douglas et al., 2021). Building on our isotopic data and a comprehensive literature review, we investigated the geographic distribution of the isotopic signals of microbial methane in South Asia and worldwide.

The geographic distribution of methane isotopic signatures in South Asia was assessed for two microbial sources: rice paddies and wastewater (Fig. 6). Regional Miller–Tans-derived values for rice paddy methane showed substantial variability (Fig. 6A), with similar signatures in western India, the Indo-Gangetic Plain (IGP), and Bangladesh, but more depleted values in southern and eastern India. The enrichment in both δ13C and δ2H (Fig. 4C) suggested that multiple sources and/or pre-emission oxidation may drive the observed spatial variation. Given that rice cultivation was concentrated in the IGP and Bangladesh (Gumma, 2011), the production-weighted means of Miller–Tans values (δ13C=-45.5±2.5‰ and δ2H=-266±17‰) represented pre-oxidation signatures of floodwater methane, though partial oxidation and associated fractionation may still be present. More enriched production-weighted concentration-weighted means (δ13C=-41.7±7.5‰ and δ2H=-236±45‰) reflected the influence of oxidation. Although diffusion contributes only ∼1 %–2 % of rice paddy methane emissions, these fractionation processes may offer insights for wetlands, where diffusion accounts for a larger share (5 %–30 %). Nevertheless, the overall Miller–Tans values (δ13C=-53.8±0.8‰ and δ2H=-311±6‰; Fig. 4A) were minimally influenced by oxidation and best represented the unaltered, source-specific isotopic signature of rice paddy methane.

Wastewater methane isotopic signatures exhibited minimal regional variation, with India and Bangladesh showing similar δ13C values (Fig. 6B). Pre-emission oxidation of wastewater methane was negligible (Fig. 5C). To better represent regional emissions, we applied population-weighted averaging, assuming similar per capita methane production across South Asia, yielding δ13C=-45.0±2.4‰ and δ2H=-350±10‰.

Our global synthesis revealed pronounced latitudinal variations in the isotopic signatures of methane from wetlands and rice paddies (Fig. 4F and G). Beyond the effects of oxidation and vegetation type, regional water conditions may also influence the hydrogen isotopic composition of microbial methane. To investigate this, we compared the global distributions of δ2H in surface water (H2O) and microbial methane (Fig. 7). Surface water isotopic data were sourced from the literature (Nan et al., 2019; IAEA/WMO, 2023; Halder et al., 2015), and microbial methane δ2H values were derived from our dataset and the global review. Global microbial methane δ2H exhibited a moderate or weak correlation with surface water δ2H (Fig. 7; R2=0.549 for ruminants, 0.363 for rice paddies and wetlands, 0.217 for waste), reflecting similar regional patterns among surface water and microbial sources. This correlation was particularly pronounced in North America. Hydrogen atoms in surface water likely served as a source for microbial methane (Whiticar et al., 1986; Whiticar, 1999), contributing to the observed spatial similarities in isotopic signatures. Among microbial sources, δ2H values varied by source category: ruminants exhibited the most depleted isotopic values, followed by waste, while rice paddies and wetlands were relatively more enriched in isotopic composition. In tropical regions, microbial methane δ2H values were more depleted than the global mean values, potentially indicating unique microbial, environmental processes, and/or different surface water δ2H that require further investigation. Variations across microbial sources mainly stem from differences in methanogenesis, with each source maintaining internal consistency.

Latitudinal variations in aquatic methane δ2H (from rice paddies and wetlands) appeared to be influenced by both water isotopic composition and pre-emission oxidation. In South Asia, δ13C and δ2H enrichment in rice paddies methane (Fig. 4C) provided clear evidence of oxidation. Additionally, the latitudinal patterns of aquatic methane δ2H closely mirrored those of surface water δ2H (Figs. 7C, 4F and G), suggesting both factors may contribute. Similarly, ruminant methane exhibited parallel δ2H trends with surface water across latitudes but showed minimal oxidation, as reflected by depleted δ2H values (Fig. 7B) and a narrow δ2H range globally (Fig. 4G), likely due to direct atmospheric release. In contrast, waste sources showed minimal δ2H enrichment (Fig. 7D) and narrow δ13C and δ2H distributions globally (Fig. 5F and G), suggesting less influence from water sources and oxidation. In comparison, biomass burning methane exhibited a consistently narrow global δ2H range (Fig. 2E), as it was minimally influenced by surface water and was emitted directly into the atmosphere without oxidation.

Data scarcity in many regions limited the development of a comprehensive global distribution map (Fig. 7). Compared to the extensive observations and studies of δ13C (Nisbet et al., 2023), measurements and constraints based on δ2H remain much more limited, largely due to technical challenges associated with its analysis. However, a growing body of recent studies suggests that δ2H can provide valuable additional constraints on methane sources (Dasgupta et al., 2025; Riddell-Young et al., 2025). Nevertheless, other research indicated correlations between the δ2H of surface water (and precipitation) and the δ2H of aquatic methane sources in certain regions (Douglas et al., 2021). Our results indicated that δ2H followed predictable trends shaped by surface water isotopic composition and microbial processes. The correlation remained valid on a global scale (Fig. 7), though it was weaker, as numerous factors collectively influenced the isotopic signatures of each microbial source. Therefore, incorporating δ2H into isotopic source apportionment can enhance our understanding of the factors driving the rapid rise in global methane concentrations. In addition, previous studies have shown that the δ2H of H2 produced from biomass burning exhibits a latitudinal dependence (Röckmann et al., 2010). By analogy, the δ2H of CH4 from biomass burning may also be influenced by the isotopic composition of surface water and precipitation. However, as shown in Fig. 2E (Data S2), the currently available global dataset is too limited to resolve such variability. Despite progress, studies on methane isotopic source signatures remain incomplete, with significant data gaps across many regions. This study alleviated some of these gaps for South Asia, contributing to the required source fingerprint data for isotope-based source apportionment of airshed-receptor methane.