Isotopic signatures of methane emission from oil and natural gas plants in southwestern China

Current evaluation of methane (CH₄) emissions to the atmosphere from global oil and gas (ONG) sector is subject to considerable uncertainty. Particularly for China, few measurements have been conducted, making it difficult to quantify emissions and conduct mitigation measures. Recently, the isotopic composition of CH₄ (δ¹³C) has been used for evaluating the contributions of fossil fuel sources to global budgets, providing a more effective approach across both regional and global scales. Here, we present a field study of CH₄ mixing ratios and δ¹³C based on UAV sampling and ground monitoring across 11 ONG sites located in southwestern China. We found that the values of δ¹³C-CH₄ provide a solid basis for identifying the CH₄ leakage and characterizing source distributions at the ONG site-levels, despite that the meteorological and site conditions as well as the surrounding environment could exert influence on the signal strengths. With the Keeling plot approach, we determined that the mean δ¹³C source isotopic signatures of CH₄ emission from these ONG sites were −25.66 ‰, heavier than previously found for other sites globally. This also indicates that they were mainly thermogenic sources. Finally, by incorporating the updated source isotopic signatures determined for China, we conducted a back-of-envelope assessment to qualitatively infer the global CH₄ emission from the fossil fuel sources. The result suggests an overestimation of global CH₄ emission from the fossil fuel inventory by 2.86 Tg CH₄ yr⁻¹, and an underestimation from the microbial sources. This study highlights the necessity of determining CH₄ isotopes on ONG site-levels, providing an indirect but important reference to quantifying CH₄ inventories among global industrial sectors.

1 Introduction

Methane (CH₄) is a major greenhouse gas (GHG) in the atmosphere, with a global warming potential 82.5 times greater than carbon dioxide (CO₂) over a 20-year timeline, and 29.8 times greater over a 100-year period (Intergovernmental Panel on Climate Change, 2021). The mixing ratios of CH₄ in the atmosphere has increased by 150 % since the industrial revolution, primarily driven by human activities (Hmiel et al., 2020; Tian et al., 2016; Skeie et al., 2023; Saunois et al., 2025; Nisbet et al., 2025). However, for the past decades, the major paradox of the CH₄ research community lie within the nonlinear trend of CH₄ mixing ratios in the global atmosphere (Schwietzke et al., 2016; Montzka et al., 2011). Debates have been raised regarding the contribution sources of CH₄ (the drivers of the atmospheric CH₄ growth) and the high uncertainty of sector-wise CH₄ emission inventories (Rice et al., 2016; Tibrewal et al., 2024; Saunois et al., 2025; Nisbet et al., 2025; Michel et al., 2024; Zhao et al., 2025). Therefore, the identification of CH₄ sources and the quantification of their contributions are essential for solving such puzzle.

The oil and natural gas (ONG) industry is one of the major contributors to anthropogenic CH₄ emissions accounting for approximately 15 % of global emissions (Lauvaux et al., 2022). China's ONG industry emerges as potent CH₄ emission sources, which are estimated to contribute 3216.61 Gg in 2024 (International Energy Agency, 2024). However, national-scale estimates remain highly variable and uncertain (Zhang et al., 2014; Sun et al., 2022), largely due to limited observation and poor estimation based on generalized emission factors. Further, the evaluation of CH₄ emission from ONG activities in China is complexed by other strong anthropogenic sources such as agricultural production (e.g., ruminants, rice paddies), wetlands, landfills, and wastewater (Skeie et al., 2023).

Since conventional measurements of CH₄ mixing ratios cannot differentiate among various emission sources, recent studies have applied stable isotope tools for source attribution, based on the distinct isotopic signatures (e.g., δ¹³C-CH₄) of microbial and thermogenic CH₄ (Suzuki, 2021; Peng et al., 2024; Leitner et al., 2020; Dlugokencky et al., 2011). For example, by analyzing the atmospheric δ¹³C-CH₄ data, Basu et al. (2022) found that microbial sources accounted for 85 % of the growth in methane emissions between 2007 and 2016. Another isotope work indicates that CH₄ emissions from the fossil fuel sector remained largely unchanged at the 1980s and 1990s levels, but increased significantly between 2000 and 2009 (Rice et al., 2016). Therefore, isotopic measurements from the atmosphere provide important benchmarks for validating CH₄ contributions from particular sources, and could assist in evaluate CH₄ emission budgets (Michel et al., 2024; Kuhlmann et al., 2025).

For ONG sources, global observations of CH₄ isotopic signatures have not yet reached agreement, resulting in large deviations in estimating the source contributions and emission trends (Schwietzke et al., 2016). While the isotope measurement of the background atmosphere sheds light to the global mean source signatures, local measurements of the ground or the lower atmosphere could be more direct in reflecting specific sources such as ONG production or processing. One study has estimated CH₄ emissions from the abandoned ONG wells in the United States, indicating a major contribution from the coalbed and natural gas sources (Townsend-Small et al., 2016). Another research conducted the CH₄ isotopic measurements in the atmosphere over the ONG fields in Romania, confirming the signature from the ONG sources despite a large variation range for δ¹³C (Menoud et al., 2022). A recent study revealed that the isotopic characteristics of CH₄ enable differentiation of ONG sources by region and depth, reflecting both the spatial and vertical variability of fossil fuel extraction (Ars et al., 2024). Several researchers used mobile monitoring in Kuwait and identified distinct isotopic signatures from microbial, ONG, and vehicular sources, finding that microbial emissions were the predominant contributor while ONG sources played a relatively minor role (Al-Shalan et al., 2022).

Recent advancements in UAV technology have facilitated novel approaches to monitor and quantify CH₄ emissions, particularly in industrial sites with limited access (Shaw et al., 2021). UAV-based active AirCore systems enable high-resolution sampling and quantification of CH₄ emissions from point sources such as mine ventilation shafts (Andersen et al., 2023). Combined with inverse Gaussian and mass balance approaches, this method allows effective upscaling to regional emission estimates and has shown high accuracy in the Upper Silesian Coal Basin (R² = 0.7–0.9). So far, few studies have deployed the UAV method for isotopic determination of methane from ONG site-level (Leitner et al., 2023), and there is a knowledge gap of CH₄ isotopic measurement at the site level considering various factors such as source types (Zhang and Zhu, 2008; Schoell, 1980), processing (e.g., purification or production of light hydrocarbon), meteorological condition, sampling method, size of the site and so on (Liu et al., 2019).

Therefore, to address the CH₄ isotope signatures from Chinese ONG plants and to examine the potential of UAV-based sampling methods, we conducted a field sampling campaign in Sichuan Province, SW China, covering 11 production/processing sites. Vertical profiles (50–300 m) of the air above ONG plants as well as the ground air had been sampled for CH₄ isotopic analyses, in order to provide semi-quantitative understanding of the emission source distributions and to reveal the isotope signatures from local Chinese ONG production and processing activities.

2 Method

2.1 Study sites

The study area is located in Sichuan Basin, Southwest China, where about 19 % of the country's total natural gas reserves have been discovered (The People's Government of Sichuan Province, 2025). Until 2022, the region has about 77 000 km gas pipelines. Between 2013 and 2023, natural gas production in this region increased from 21.31 × 10⁹ to 59.48 × 10⁹ m³ (Sichuan Provincial Bureau of Statistics, 2025), with an average annual growth rate of about 11 %. In 2020, ONG production in Sichuan accounted for 24 % of China's total ONG production.

We monitored CH₄ mixing ratios and sampled air for isotope measurements across 11 ONG processing or transportation stations in the central Sichuan Basin. The study region is characterized by a humid subtropical climate, with consistently warm and humid conditions throughout the year. The areas of these stations vary from 2000 to 300 000 m², while the production activities also vary, including natural gas purification plants, gas gathering stations, light hydrocarbon plants, pigging stations, pressurization stations, etc. (Table 1). Most ONG sites are located in remote areas, where the surrounding environment is farmland (paddy fields), bodies of water, forests, and scattered residences, with no large-scale cattle farms, pig manure sites, or landfill sites nearby. In addition, during our observational campaign, no biomass or coal burning was spotted in the surroundings. For the reasons of privacy and confidentiality, the accurate locations and contours of the ONG stations cannot be disclosed in this paper.

Table 1 Background information of the studied production/processing sites for oil and natural gas.

2.2 Sampling methods

From 13 to 19 April 2023, we monitored and collected samples at 11 ONG production stations in the central Sichuan Basin, obtaining a total of 74 air samples, including 28 from ground and 46 from air overhead. Ground air samples were collected at heights ranging from 0.5 to 2 m above the group. Sampling locations were chosen in open areas of each station, including areas near pipelines and production equipment. The criteria for selecting the open-ground area includes: no significant elevation in CH₄ concentration compared to the ambient background (Chen et al., 2024), as measured by a portable TDLAS methane analyzer; a minimum distance of 20 m from any facility; and the absence of visible pipelines or valves. Sampling in the area of pipelines and production equipment was performed at locations that showed abnormal mixing ratios after ground monitoring; sampling was performed in the central area if no apparent CH₄ emission was detected. For large sites (> 10 000 m²), multiple sampling points were established, while for small sites (< 10 000 m²), 1–2 sampling points were established in facility areas, and the sampling time for each sample was about 45–50 s.

Air sampling was performed by an Unmanned Aerial Vehicle (UAV) equipped with an automatic sampling pump (Fig. 1). The UAV model was a DJI-T10 (upgraded version), and the sampling pump model was KVP04-1.1-12V (1.25 L min⁻¹). Taking into account the altitude ranges in previous studies (Kim et al., 2025; Han et al., 2024; Chen et al., 2024; Liu et al., 2021; Liu, 2018; Ali et al., 2017) and the drone's battery life, the monitoring altitudes were defined at 50, 100, 200 and 300 m, respectively. Initially, a ground sampling spot was identified, typically within the pipeline vicinity of the plant. Subsequently, a UAV equipped with an automatic sampling pump and air collection bags were lifted to 300 m above the ground sampling site. The UAV then sequentially descended to altitudes of 200, 100, and 50 m, respectively, dedicating 45 to 60 s at each elevation for collecting air samples. This systematic approach ensures a comprehensive and stratified sampling strategy, facilitating the assessment of atmospheric constituents at varying heights.

Figure 1 Site demonstration of the UAV automatic sampling system and the schematic diagram of the sampling device.

The volume of each air sample was approximately 1 L, stored in HOONPO Teflon gas bags (1 or 2 L). Air was sampled at varying altitudes from all sites except for S1, which was sampled specifically at 50 and 100 m. Our sampling was conducted in the daytime between 11:00 a.m. and 02:00 p.m.(UTC+8). Air sampling and UAV cruising were synchronized. For comparison, we had sampled a production well with significant emissions (built in the 1980s), which was short of maintenance for long. This was meant for directly exploring source signals (2 samples from the leakage point of the well, and one sample the open area of the site). In addition, we sampled (2 from the riverbank, 2 from the park) ambient air near an urban park and an urban river to analyze the ambient CH₄ mixing ratios and isotope values. The sample list and results are provided in Table S1 (Supplement).

The influence of meteorological conditions on the CH₄ mixing ratios and isotopes at the ONG plants was also considered. Therefore, a portable meteorological station was deployed at each station during the sampling periods. It was equipped with a three-dimensional ultrasonic wind speed and direction sensor (model: M307200), which recorded the wind speed (horizontal and vertical) and direction (horizontal and vertical) near the ground (3 to 10 m according to field conditions). The sampling frequency was 32 Hz with a resolution of 0.1 m s⁻¹ for wind speed and 0.1° for wind direction, and the precision of wind direction and speed is 2° and 0.2 m s⁻¹, respectively. We also obtained air pressure, solar radiation, temperature, and relative humidity from weather station.

2.3 Measurement methods

Gas samples were analyzed within one month after on-site sampling. Picarro G2132-i was used to detect the isotope and mixing ratio of CH₄, which is based on Cavity Ring-Down Spectroscopy (CRDS). This system (in similar models) has been previously used by other studies of in situ CH₄ source characterization (Rella et al., 2015; Geum et al., 2024). For each sample measurement, the total analysis time is over 180 s through the Picarro G2132-i instrument, and the average of the last 120 s of CH₄ isotope and mixing ratio data was integrated and exported for raw sample data. We used two international primary standards (VPDB-referenced; Std1 and Std2, −68.6 ‰ and −40.0 ‰, respectively) and one secondary standard (Hstd, −46.89 ‰), all of which had been cross-calibrated (more details could be found in the Supplement, Sect. 3). During each measurement sequence (Supplement, Fig. S1), all three standards were measured together with the samples (five samples per sequence). Std1 was used for calibration and for correcting sequence-related drift, Std2 served as a quality control check, and Hstd was used to constrain long-term drift. All measurements were completed within one week after sampling. In addition, repeated measurements of the same isotope standard (Std1) over a year have demonstrated the excellent stability of the instrument (Supplement, Fig. S2). The δ¹³C detection precision (1σ, 1 h window) of the instrument is as follows: when the mixing ratio of CH₄ is greater than 1.8 ppm, the precision of 5 min mean value is better than 0.8 ‰, when the mixing ratio of CH₄ exceeds 10 ppm, the precision is better than 0.4 ‰.

2.4 Calculation of source isotopic signatures

Based on the sample detection data, the Keeling plot method was used to determine the CH₄ source (Keeling, 1958; Pataki et al., 2003) for each field station, the source isotopic signatures were obtained using a linear regression method based on the least squares approach (France et al., 2016; Akritas and Bershady, 1996). As shown in Eq. (1):

$$\begin{array}{}\text{(1)}& {\mathit{\delta }}_{\left(\mathrm{a}\right)}=\left[{\mathrm{CH}}_{\mathrm{4}\left(\mathrm{b}\right)}\right]\times ({\mathit{\delta }}_{\left(\mathrm{b}\right)}-{\mathit{\delta }}_{\left(\mathrm{s}\right)})\times (\mathrm{1}/[{\mathrm{CH}}_{\mathrm{4}\left(\mathrm{a}\right)}\left]\right)+{\mathit{\delta }}_{\left(\mathrm{s}\right)}\end{array}$$

Where δ_(a), δ_(b), and δ_(s) represent the δ¹³C values of the sample, the background air and the average source, respectively. [CH_4(a)] and [CH_4(b)] represent the CH₄ mole fractions of the sample and the background air, respectively. The intercept (δ_(s)) of the fit line is the isotope value of the CH₄ source present in the mixed sample. In linear regression, $\mathrm{1}/\left[{\mathrm{CH}}_{\mathrm{4}\left(\mathrm{a}\right)}\right]$ and δ_(a) represent independent (x-axis) and dependent (y-axis) variables, respectively. This method is suitable for carbon dioxide, methane (Thom et al., 1993), water vapor (Moreira et al., 1997), and other gases, but each gas has its specific considerations (Pataki et al., 2003). The gas samples from each station were collected within 30 min, during which the atmospheric background values (isotope and mole fraction of CH₄) did not change, fulfilling the application conditions of this method (Lu et al., 2021).

2.5 HYSPLIT model

The Hybrid Single-Particle Lagrange Integrated Trajectory (HYSPLIT) model developed by the National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory, is a widely used public platform for different atmospheric scales and supports online modules (Pereira et al., 2019). The model has been used to calculate the air mass transfer trajectories at different altitudes (Shan et al., 2009; McGowan and Clark, 2008; Stein et al., 2015). In the air above ONG plants, ground-based metrological station cannot capture the dynamics of wind directions and speeds. Thus, we deployed the HYSPLIT model to analyze the influence of meteorological conditions on CH₄ mixing ratios and isotopes particularly for the UAV-based sampling and measurements. The time resolution of the model could reach 1 h and the height resolution was 1 m. In this study, 24 h backward trajectories were calculated at each site for five altitudes: ground level, 50, 100, 200, and 300 m. The input data included the longitude and latitude of the site from field measurements and sampling time, while the output information were wind direction and speed at different heights. Additionally, we evaluated the stability indexes based on the HYSPLIT model outputs. It shows that the Pasquill stability classes during UAV sampling were mostly C in our sampling campaigns (Slightly unstable conditions).

2.6 Source partitioning with end-member mixing method

End-member mixing method is a common method for identifying major sources and quantifying contributions of multiple sources (Bugaets et al., 2023; Barthold et al., 2011). The end-member mixing model is conducted with the mass balance of air mixing as well as composition of the tracers, based the following assumptions: (1) the ambient atmospheric constituents are constant, (2) the tracer is conservative, (3) and the source mixing ratios differ from the background. Here, we used CH₄ mixing ratios and isotopes as tracers to investigate the contribution of atmospheric background, open surface area, and facility area to the sampled air above various ONG plants.

2.7 Statistics

Data analysis and graphing were performed using Origin 2024 software for Windows. Linear fitting was based on the principle of the least square method, indicating the 0.95 confidence intervals. A value of P<0.05 was considered significant for statistical analysis, and the fitting results are expressed as fitting mean and standard deviation. Maximum, minimum, mean, median, outliers, and 25 %–75 % range values were also analyzed and reported in the figures or tables.

3 Results

3.1 Variations of CH₄ mixing ratios and isotopes across ONG plants

The CH₄ mixing ratios and δ¹³C-CH₄ values from the studied 11 ONG sites ranged from 1.88 to 3.66 ppm and from −48.14 ‰ to −30.41 ‰, respectively (Supplement, Table S1). The maximum and minimum values of CH₄ isotopes were observed at sites S2 and S6 (100 m), respectively. The variation of the CH₄ mixing ratios and isotopic values at stations S2, S4, and S7 is significantly greater than that observed at other stations (Fig. 2). The results of direct emissions from the production well were $-\mathrm{16.19}\pm \mathrm{5.53}$ ‰ and 118.98±0.52 ppm, respectively. By comparison, the CH₄ isotope and mixing ratio of the urban samples were $-\mathrm{46.20}\pm \mathrm{0.47}$ ‰ and 2.04±0.07 ppm, respectively.

More specifically, in the near-ground air, we found higher CH₄ mixing ratios and isotopic values in areas close to production equipment than those in the open areas (Fig. 2), although the mixing ratios and isotopic values at S6 and S9 were very similar to the atmospheric background. Overall, the CH₄ isotopic values observed from the ground air at the field stations ranged from −47.68 ‰ to −16.19 ‰ ($-\mathrm{41.99}\pm \mathrm{7.10}$ ‰ on average), while the isotopic values from the air over plants were more constrained in a small range of −48.14 ‰ to −44.11 ‰.

Figure 2 Box whisker plots of CH₄ mixing ratios (a) and isotopic values (b) for air sampled near the plant ground (Ground data) and air sampled above the plants (mean, median, 25 %–75 % range, and 1 SD are indicated in the figures; the red lines refer to CH₄ mixing ratios (a, 1.9 ppm) and isotopic values (b, −47 ‰) from the atmospheric background.

3.2 Vertical profiles of CH₄ mixing ratios and isotopes and source partitioning

The vertical distribution of the CH₄ mixing ratios and isotopes differed from site to site. For instance, the mean CH₄ mixing ratios were higher at 100 m or above than at 50 m, yet the isotopic values (δ¹³C) were somewhat lower (Fig. 3). From the perspective of each ONG site, the patterns were similar yet more complex (Supplement, Fig. S3). Some stations exhibited consistent trends (S1, S3, S6, S7, S9, S11), while others displayed different trends (S2, S4, S5, S8, S10). For instance, the CH₄ mixing ratio and isotopic values at 100 and 200 m altitude of station S8 were inversely proportional. As the altitude increased from the ground to 300 m, the CH₄ isotopic values of stations S4 and S1 exhibited a decline, ranging from −45.47 ‰ to −47.00 ‰ (ground to 300 m) and from −42.17 ‰ to −47.12 ‰ (ground to 100 m), respectively (Supplement, Fig. S3). The CH₄ isotopic values of stations S6 initially decreased with increasing altitude and subsequently increased, reaching a minimum at 100 m altitude (−48.14 ‰). The variation of the CH₄ isotope vertical profile at station S8 was analogous to that observed at sites S6, with the exception that the CH₄ isotopic minimum value reached −45.95 ‰ at 200 m altitude. The variation of CH₄ isotopic values with altitude at station S9 was complex, exhibiting a decrease followed by an increase, which then decreased again, reaching minimum and maximum values at 50 m (−47.60 ‰) and 200 m (−44.11 ‰), respectively.

Figure 3 Box whisker plots showing the variations of CH₄ mixing ratios (a) and isotopic values (b) at different heights (from ground to 300 m at all sites); include mean, 25 %–75 % range, and 1 SD; “n” represents the number of samples. The red lines refer to CH₄ mixing ratios (a, 1.9 ppm) and isotopic values (b, −47 ‰) from the atmospheric background.

3.3 Source partitioning for CH₄ emission from the ONG plants

The end-member mixing method is a commonly employed technique for calculating isotope mixing by various sources of GHGs (Bugaets et al., 2023). In this study, we determined the contribution fractions of CH₄ from the atmospheric background, surface, and facility areas to the air over the sites (more details could be found in Supplement: Sect. 1 and Table S4). The results indicate that atmospheric background is the predominant source of methane, with larger contribution compared with those determined for ground and facility area (Fig. 4). Nevertheless, at altitudes of 100 m, the ground and facility-derived contributions became slightly more important, indicating that these two altitudes may be more representative of the whole-plant emission signals. In particular, at higher altitudes, the signal of CH₄ emissions from ONG plants below may be dampened by mixing of background air or even biological sources in the vicinity of the sites. Despite so, we didn't find any strong emission signals from surrounding environment outside the regions of the ONG plants (no apparent elevation of CH₄ mixing ratios along the vertical profile in the atmosphere).

Figure 4 The fractional contributions from ambient background, surface, and facility areas contribute to the air sampling above ONG plants. (a) the proportion of contributions to all heights of all stations with 1 SD; (b) the proportion of contributions to different heights with standard error.

3.4 Characteristics of source isotopes

The Keeling plot method was employed to determine the isotopic signatures (δ¹³C) of CH₄ sources at each station, as presented in Fig. 5. The range of the CH₄ source isotopic signatures varied from $-\mathrm{52.71}\pm \mathrm{6.06}$ ‰ to $-\mathrm{11.88}\pm \mathrm{2.32}$ ‰ (mean value of −25.66 ‰), indicating that they were mainly thermogenic sources (associated with oil production) (Menoud et al., 2022; Sherwood Lollar et al., 2002). Globally, the range of CH₄ isotopic values from fossil fuels is −75 ‰ to −25 ‰ (Defratyka et al., 2021). Our results mostly align with this large range, but were at the higher end.

The source isotopic signature for station S5 was $-\mathrm{52.71}\pm \mathrm{6.06}$ ‰, which was lower than the atmospheric background value. However, the data fitting for this station was poor, indicating large uncertainty (R²=0.08). On the other hand, the direct measurements of emission from wells indicated that source ¹³C signature was $-\mathrm{19.42}\pm \mathrm{2.19}$ ‰, which is close to the result of the ground-air sample collected at the site ($-\mathrm{16.19}\pm \mathrm{5.53}$ ‰) (Supplement, Table S1). We conducted continuous monitoring of the natural gas production wells where elevated methane mixing ratios were consistently observed. Given that the sampling was carried out in close proximity to the leakage sources (wells) surrounded by high safety walls, we believe that the derived isotopic signatures could represent the emission source from ONG plants rather than being affected by the surrounding environment.

Figure 5 The CH₄ source isotopic signatures of 11 field stations determined based on the Keeling plot method. The blue area represents the 95 % confidence interval, and the red line is the result of linear regression posterior mean fit; The samples in different positions are distinguished by different colors. The intercept and R² are given, which means the source isotope signal value and the fitting degree, respectively. Linear regression was performed using the least squares method.

4 Discussion

4.1 CH₄ emission signals as revealed by the isotopic measurements from both ground and atmosphere

During our sampling campaign, the mean δ¹³C-CH₄ at most ONG plants was higher than the atmospheric background ($-\mathrm{47.0}\pm \mathrm{0.3}$ ‰) (Tyler, 1986), although several sites (S1, S3, and S5) had values close to the atmospheric background. In the meanwhile, the average values of CH₄ mixing ratios were significantly higher than the atmospheric background (1.9 ppm) at all sites (Skeie et al., 2023). This supports apparent CH₄ leakage from most sites during our study. Further, as referred to the Keeling plot approach, the correlation between CH₄ mixing ratios and isotopes at the ONG sites was significant (R²=0.91) (Fig. 5). Besides, the ground exhibited an even stronger correlation (R²=0.95) than the air (R²=0.31) (Supplement, Fig. S4). These findings indicated that the CH₄ sources at these ONG plants were generally consistent, reflecting single-source contribution. The δ¹³C source isotopic signatures from each site may vary, but the major range was consistent with the fossil-fuel sources (Defratyka et al., 2021). In addition, an investigation of the potential sources of CH₄ in the vicinity of the ONG sites revealed that the primary source of CH₄ at the station was from ONG, with other sources exerting less impact (Table 1).

Further, the variations of δ¹³C-CH₄ across sites may reflect the influence from sources and local facilities. As shown in Fig. 5, the source isotopic signatures determined with the Keeling plot method were distinctively higher in the purification and production sites (S1–S2 and S7–S8) than in the gas gathering and transport lines (S3–S6) (Fig. 5). Studies have indicated that infrastructure, including components such as dehydrators, valves, compressors, and pipelines, represents a significant source of CH₄ emissions from the ONG system. Infrastructure is particularly vulnerable to CH₄ leakage due to corrosion and wear (Anifowose et al., 2014; Fernandez et al., 2005; Burnham et al., 2012; Anifowose and Odubela, 2015).

On the other hand, sampling locations appear to influence the observational results, particularly when comparing measurements near ground with those made in the air above (Figs. 2 and 3). We discovered that a majority of the ONG production stations exhibited both higher δ¹³C and CH₄ mixing ratios in the ground measurement than in the air at 50–200 m aboveground (Fig. 3). This suggests that ground-based measurements are likely to exert more significant source signals when CH₄ leakage occurs nearby. However, this may only apply to small ONG sites. Alternatively, measurements of air above the plants could provide information on the site-level emission, of which 100 m (on average) seems to perform best in representing emission signals as confirmed by the end-member analyses (Fig. 4) according to the local settings. By contrast, the mixing ratios and δ¹³C of CH₄ at S6 were lower in the near-ground air than in the air overhead. This discrepancy may be attributed to the uncertainty associated with the small plant size (5167 m² with a processing capacity of 630 × 10⁴ m³ d⁻¹) and thus large impact from surrounding environment sources of CH₄ in the air. Therefore, the surrounding emission sources, together with the meteorological conditions, are likely to exert an impact on the UAV-based measurements considering local conditions of the ONG plants.

4.2 Factors of drone-based isotope measurements in the atmosphere

As shown in Figs. 2, 3 and 4, a wide range was found for the CH₄ mixing ratios and isotopic values determined at the studied ONG plants, with particular variabilities observed at varying altitudes above ground. This could be attributed to a number of factors, including the presence of other CH₄ sources in the vicinity and the influence of meteorological conditions (Kavitha and Nair, 2016). Around all ONG sites, the dominant land-use types are rural roads and scattered paddy fields, through which small ditches or streams flow. No livestock farm or landfill was observed in these areas, and no biomass burning was observed during our sampling period (April), which aligns with the legal prohibition of such activities in China. Taking this evidence into consideration, we can rule out the possibility of significant contributions from biological sources from the surrounding environment. The dampened signals of both CH₄ mixing ratios and isotopes in the air-borne measurements may be partly due to convection with air carrying biological sources outside the range of the ONG plants, but the overall strength is small as confirmed with the Keeling plots (Fig. 5). Further, at station S6 where the 200–300 m-measurement exhibited high CH₄ levels, we also found higher δ¹³C, which does not support the importance of biological sources which is usually characterized with lower δ¹³C (Fisher et al., 2011, 2017).

Importantly, our UAV sampling points were mostly located near the center or at least away from the edges of the ONG sites, which are significantly larger in area compared to the scattered paddy fields nearby. The Pasquill stability class during the sampling indicated slightly unstable atmospheric conditions, suggesting predominant vertical mixing over horizontal transport. This enhances the reliability of isotopic measurements in reflecting methane emissions from the station itself, with the ONG site being the primary source influencing the isotopic signatures. Furthermore, in a parallel study conducted by our team in the same region (Chen et al., 2024), several ambient air samples were collected for δ¹³C-CH₄ analysis near paddy fields (1.5 m above the surface, 10–20 m from field boundaries). The isotopic values (δ¹³C-CH₄ = $-\mathrm{47.2}\pm \mathrm{0.2}$ ‰; unpublished data) were quite close to the global background, indicating minimal influence from the paddy fields. This could be attributed to the relatively low methane emissions during the sampling period, as April is typically dry in the Sichuan region.

The influence of meteorological conditions is significant and complex, and challenging to analyze. Wind direction and speed in the air above plants were obtained using the HYSPLIT model (Supplement, Table S3), and the results for near-ground air were cross-validated with those obtained from the meteorological station (Supplement, Fig. S5). The correlation analysis between wind speed and CH₄ isotope results revealed an exponential relationship with a R² value of 0.33 (Supplement, Fig. S6). This indicates that as wind speed increases, the impact of CH₄ diffusion and dilution becomes more significant. Wind direction plays a role in the uncertainty of CH₄ distribution, as it has a significant influence on CH₄ transport near the surface, resulting in a non-uniform distribution of CH₄ and typically higher mixing ratios downwind from the emission source. Furthermore, upwind CH₄ sources can have a notable impact on CH₄ levels over the station. The utilization of HYSPLIT model serves a crucial function in this regard (Supplement, Fig. S7 for a detailed example of S7 site).

Moreover, the local conditions of the ONG plants are among the primary determining factors of the air-borne measurements, encompassing factors such as the size of the sites, the treatment processes employed, the processing capacity, and the timing and location of sampling. Typically, a larger site size is likely to produce greater signals of CH₄ in the atmosphere above head and to be less affected by other biological sources from surrounding environment (Omara et al., 2016). A clear positive correlation was observed between the site area and the isotope results for site areas below 10 000 m² (Supplement, Fig. S8). This is interesting and probably also reasonable, as extra-large industrial sites may encompass more complexed influence from both meteorological conditions and ground source distributions. Besides, the intermittent nature of emissions from the site facilities introduces an element of uncertainty with regard to the sampling time and locations (Omara et al., 2016). Overall, by conducting a simple Principal Component Analysis (PCA), we identified a weak relationship among wind direction, wind speed and isotopes, and a strong correlation between the size and capacity of the sites with CH₄ isotopes (Supplement, Fig. S9). This means that, during our sampling campaign, ONG plant-related factors are the major players in determining the measured CH₄ isotope results.

4.3 Global source isotopic signatures of ONG-derived CH₄

While no studies have specifically focused on the isotopic characteristics of CH₄ leakage from ONG plants in the Sichuan region of China, several studies have investigated the characteristics of CH₄ isotopes in Chinese ONG production regions, across the Sichuan Basin, Xinjiang, Northeastern China, and the Ordos Basin (Cai et al., 2013; Huang et al., 2017; Wang et al., 2018; Zhang et al., 2018; Zou et al., 2007; Zhu et al., 2014; Liu et al., 2019; Dai et al., 1985, 1992). The reported values of CH₄ isotopes cover a wide range from −54.9 ‰ to −17.4 ‰ (Supplement, Table S2). The difference in ¹³C-CH₄ reflects the origin of natural gas such as biogenic and abiogenic gases (Sherwood Lollar et al., 2002; Dai et al., 2005), of which biogenic gases include coal- and oil-type gases, respectively. The Sichuan Basin in our study has a complex geological environment and many gas-production layers, such as Cambrian, Ordovician, Carboniferous, Jurassic and so on (Zhang et al., 2018; Cai et al., 2013); also, CH₄ from different geological layers can have variable isotopic characteristics. In comparison with the findings of other researchers on CH₄ isotopes in the Sichuan Basin (Supplement, Table S2), our results of ¹³C-CH₄ isotope signatures spanned more widely and appeared to be generally heavier (Fig. 5).

Among the global observations, Menoud et al. (2022) examined isotopic signatures of CH₄ from an ONG extraction plant in Romania. Their methodology aligns closely with ours, and their findings indicate a range of δ¹³C values from $-\mathrm{67.8}\pm \mathrm{1.2}$ ‰ to $-\mathrm{22.4}\pm \mathrm{0.04}$ ‰ (Menoud et al., 2022). In Kuwait, methane from the southern Burgan field had a δ¹³C-CH₄ of $-\mathrm{48.9}\pm \mathrm{0.2}$ ‰, and the signals were slighter lower by measuring downwind of coastal refineries ($-\mathrm{51.6}\pm \mathrm{0.5}$ ‰) (Al-Shalan et al., 2022). A study on CH₄ isotopes in the Western Canadian Sedimentary Basin showed that the δ¹³C signatures of CH₄ from ONG sources ranged between −71 ‰ and −29.3 ‰, and the isotopic composition was primarily controlled by geological structure rather than the type of hydrocarbon (e.g., ONG sources) (Ars et al., 2024). In an Arctic study, Fisher et al. (Fisher et al., 2011) found that CH₄ in the Arctic atmosphere during summer is mainly of microbial origin, whereas in spring and winter, the dominant source is likely CH₄ emissions from natural gas fields, with a source δ¹³C-CH4 signature of $-\mathrm{52.6}\pm \mathrm{6.4}$ ‰. In the UK, δ¹³C–CH₄ signatures from ONG sources were well constrained, showing consistent values of approximately $-\mathrm{36}\pm \mathrm{2}$ ‰, reflecting homogenized North Sea gas (Zazzeri et al., 2015). Compared with those previous work on ONG-related CH₄ sources, our results overlap with the wide ranges as reported; the overall mean values are heavier and well above the global mean of fossil fuel CH₄ isotope ($-\mathrm{44.0}\pm \mathrm{0.7}$ ‰) (Michel et al., 2024). This discrepancy can be attributed to a number of factors, including thermogenic origin, geographical differences (Menoud et al., 2022), the treatment processing of natural gas, and likely uncertainty due to limited sample size.

The methane isotopic signatures observed in this study were compared with those reported from major ONG production regions worldwide (Fig. 6). The δ¹³C of CH₄ was found to be lighter in the United States and Canada, but heavier in China. Regional variations in δ¹³C values were observed, even within the same region (e.g. in China). Our results exhibited a significantly heavier δ¹³C than those of other studies. This is most likely to be attributed to differences in the origin of CH₄ (Zhang and Zhu, 2008; Wang et al., 2018; Defratyka et al., 2021; Schoell, 1980), as we have excluded the possibility of strong influence from biological sources or biomass burning in the surrounding environment as discussed above.

Figure 6 Comparison of CH4 isotopic signatures in oil and natural gas production regions across major hydrocarbon-producing countries; data from both literature and this study are included. The number of data points is shown at the bottom of the box chart, and carry on the error analysis. “n” represents the number of data points.

The δ¹³C of CH₄ represents a valuable indicator for constraining and estimating CH₄ emissions particularly from anthropogenic sources of the globe (Milkov et al., 2020). As a sum-up, the mean δ¹³C signatures of CH₄ sources as indicated from measurements of atmospheric background integrated the collective contributions from various sources of CH₄. Hence, with the updated isotopic signatures for specific sources such as ONG industry, the previous conclusions on global contribution/flux of CH₄ from ONG industry may need to be revised (Schwietzke et al., 2016). In comparison with previous studies, the δ¹³C values from ONG industry in our work (−25.66 ‰, mean values of the 11 stations) are higher, especially different from the global flux-weighted averaged by Schwietzke et al. (Schwietzke et al., 2016). By incorporation of flux contribution from Chinese ONG industry, isotope signatures as well as global datasets utilized in the previous work (Schwietzke et al., 2016), we conducted a sensitivity analysis, examining the effect on diverse source contributions (in flux) when updated the δ¹³C-CH₄ from Chinese ONG industry (Supplement, Sect. 2).

Our finding suggests that, the updated ¹³C isotope signature based on field observation from China would elevate global fossil fuel-derived CH₄ isotopes signature by about 0.46 ‰; as a consequence, the new result would lead to a smaller contribution from global ONG industry (corresponds to an overestimation of emissions by 2.86 Tg CH₄ yr⁻¹) but a larger contribution from microbial sources. This finding is consistent with some recent research findings, such as Chandra et al. (2024), who reported that CH₄ emissions decreased in fossil fuel sources, while increasing in microbial sources during 1990–2020. In Australia, CH₄ emissions from agricultural ponds, which are microbial in origin, have been underestimated in national greenhouse gas inventories, indicating that actual CH₄ emissions may be higher than officially reported (Malerba et al., 2022).

4.4 Feasibility and limitations

Atmosphere CH₄ isotopic research has shown its power in distinguishing between microbial and fossil sources of global atmospheric CH₄ trends (Basu et al., 2022; Bruhwiler et al., 2017). However, due to scarcity of observational evidence of various CH₄ source signatures, large uncertainties still exist for such estimations. The objective of our research was to distinguish sources of CH₄ as well as to indicate CH₄ leakage strength at site-level, providing basic but convincing information for constraining CH₄ leakage. With both ground- and air-based approaches, our study has demonstrated the feasibility of our sampling method and research design in studying the characteristics of CH₄ sources and their influencing factors at ONG stations in SW China. Nevertheless, it is necessary to point out, that the impact of meteorological conditions and site conditions on the dampening/masking of CH₄ isotope signatures in the atmosphere cannot be neglected, particularly considering the limited sampling sizes during our campaign. Therefore, the reconciliation between ground-based and atmospheric measurements as well as source partitioning remain to be further validated, given more sampling coverage both spatially and temporally. In addition, more sampling at different locations or different ONG plants would be greatly beneficial to better confirm the CH₄ source isotope signature from fossil fuel industry in China.

Recently, the debates on the global atmospheric trends of CH₄ levels and the driving sources are continuing (Saunois et al., 2025; Kirschke et al., 2013; Rice et al., 2016; Tibrewal et al., 2024). Overall, the decline of global mean CH₄ isotopic signals seems to slightly speed up in recent years, likely supporting the importance of microbial emissions (Fisher et al., 2017). The atmospheric chemists have recently clarified the important link of air pollutants with atmospheric CH₄ sink strengths, further complicating the story. For the ONG industry, previous studies have reported improvements in technology, equipment, and management practices which will assist in reducing CH₄ leakage (US Environmental Protection Agency, 2012). Therefore, the contribution from ONG sources to the global CH₄ budget is likely to decline. With more field observation and more up-to-date database being established, we would be able to pursue a more realistic evaluation of ONG-contributed CH₄ emission inventory, providing guidance on further mitigation measures.

5 Summary

In this study, we examined the δ¹³C isotopic characteristics of air samples collected from ONG stations in the central Sichuan Basin, China. The CH₄ isotopes were measured near the ground and in the air along a vertical profiles of altitudes. By comparing isotopic results across sites and among sampling locations, we found that source distributions by industrial facilities as well as processing capacity/site sizes act as the major driving factors of CH₄ mixing ratios and isotopic signals, while the influence from meteorological conditions and other sources from the surrounding environment may exert less impact. Based on the Keeling plot method, we determined that the source δ¹³C signature of CH₄ from the ONG sites ranged was −25.66 ‰, indicating a heavy δ¹³C of fossil fuel. In comparison with the CH₄ isotopic values from the global ONG sources, our study reported generally heavier isotopic signatures. By updating the isotope signatures of Chinese ONG with our observation, we conducted a weighted calculation of flux and isotope for the global CH₄ budget, suggesting that the global CH₄ emissions from microbial sources may be underestimated, while those from fossil fuel sources may be overestimated. Overall, our study confirms the effectiveness of isotope method in distinguishing industrial CH₄ sources and sheds light to the global estimation of CH₄ budgets utilizing the isotope geochemistry approach.