The global demand for liquefied natural gas (LNG) reached 401 million t in 2023, with market growth driven largely by Asia, which accounts for 64 % of global LNG imports (GIIGNL, 2024; IGU, 2024). South Korea remains among the world's top three importers, with imports of approximately 45 million t in 2023 (GIIGNL, 2024; IGU, 2024). Due to its relatively low carbon intensity compared to coal or oil, LNG is often recognized as a “transition fuel” for decarbonization efforts (Al-Kuwari, 2023; IGU, 2024). The concept of LNG as a transition fuel has been supported by previous studies, highlighting its potential to facilitate a shift from more carbon-intensive sources and to support the intermittency of renewable energy (Al-Kuwari, 2023). However, this perspective is challenged by research showing that methane emissions associated with LNG supply chains and gas-fired power generation can significantly undermine this environmental benefit (Howarth, 2024), as methane has a much higher global warming potential than carbon dioxide over a shorter time period (IPCC, 2023).

In South Korea, the energy sector, including LNG gas-fired power plants, accounts for approximately 23 % of total national methane emissions (MOE, 2022). The number of LNG gas-fired power plants in the country has expanded in recent years to meet rising electricity demands, with plans to increase LNG-based power capacity from 43.3 GW in 2020 to 69.5 GW by 2038 under the 11th National Basic Plan for Electricity Supply and Demand in South Korea. This transition offers a cleaner alternative to traditional coal-fired facilities in terms of carbon dioxide emissions. However, the accurate quantification of methane emissions from LNG power plants has proven challenging, and significant uncertainties remain regarding the magnitude of fugitive emissions (Lyon et al., 2015). Most methane estimates for these facilities rely on bottom-up inventories, which often have difficulty capturing fugitive emissions from operating conditions (Alvarez et al., 2018; Brandt et al., 2014). These inventories, which apply generic emission factors to activity data, can underestimate actual emissions owing to their limitations in reflecting real-world variability (Howarth, 2024). In particular, recent studies on industrial methane emissions have consistently revealed a substantial discrepancy between bottom-up inventories and ground-based top-down atmospheric measurements. Rutherford et al. (2021) showed that bottom-up inventories frequently underestimate emissions from global oil and gas infrastructure because they depend on fixed factors that do not account for stochastic “super-emitter” events. Such discrepancies are well-documented in upstream production fields (Alvarez et al., 2018), but downstream industrial facilities, such as LNG power plants and distribution networks, are also prone to substantial fugitive emissions due to equipment leaks and operational venting (Weller et al., 2020; Plant et al., 2022). Despite their critical role in urban environments, these downstream sites lack facility-level monitoring.

To address the limitations of traditional bottom-up inventories, which often struggle to capture fugitive emissions from complex industrial operations, ground-based top-down measurement approaches utilizing mobile platforms have gained significant traction in recent years for quantifying urban methane emissions (IPCC, 2023; Alvarez et al., 2018; Brandt et al., 2014). Mobile measurements from vehicle-mounted sensors have proven to be highly effective for mapping and identifying fugitive methane emissions from urban natural gas distribution networks (Vogel et al., 2024). This provides a more accurate understanding of urban methane budgets. Mobile measurements identify more fugitive methane sources that are difficult to detect in bottom-up inventories and provide data for improved estimations of total emissions (Joo et al., 2024; Maazallahi et al., 2022; Maazallahi et al., 2020; Vogel et al., 2024; Mitchell et al., 2015; Jia et al., 2025). Vogel et al. (2024) conducted a study in 12 European cities and demonstrated that mobile measurements could effectively identify and quantify methane emissions from natural gas systems in diverse urban infrastructures. These measurements successfully detected methane emissions that traditional approaches have often missed. In Hamburg, Germany, Maazallahi et al. (2022) quantified urban natural gas emissions using a vehicle-based methane monitoring system and found that fugitive emissions accounted for approximately 15 % of the total estimated emissions, with individual leakage rates varying from 0.1 to 5 kgh-1. Joo et al. (2024) discovered significant missing fugitive methane emissions (approximately 573 tyr-1) from urban sewer networks in the Gwanak district of Seoul, South Korea, highlighting the limitations of relying solely on bottom-up inventories. Beyond urban infrastructure, mobile platforms are crucial for assessing fugitive methane emissions from industrial facilities. Studies in Texas and California, US, have quantified fugitive emissions and identified major sources by comparing ground-based top-down methane measurements and bottom-up inventories around oil and gas production and processing sites (Alvarez et al., 2018; Brandt et al., 2014; Lyon et al., 2015). Alvarez et al. (2018) evaluated methane emissions across the US oil and gas supply chain, estimating total emissions at 13 million tyr-1, a figure 60 % higher than the U.S. Environmental Protection Agency (EPA) estimates at the time. Furthermore, Brandt et al. (2014) quantified methane leaks in a North American natural gas system and reported leak rates ranging from 0.05 % to 8 % of the natural gas production, with substantial variation across different supply chain segments. Jia et al. (2025) quantified fugitive methane emissions from natural gas stations in China using on-site component-level measurements and identified “super-emitters,” which accounted for nearly 80 % of the total fugitive emissions detected at the LNG facilities, with methane concentrations exceeding 10 000 ppm. These studies highlighted the importance of direct measurement techniques for quantifying fugitive methane emissions and identifying key mitigation opportunities in the natural gas industry.

This study addresses the need for a comprehensive assessment of methane emissions from LNG gas-fired power plants, particularly those located within major urban infrastructures. This is one of the first studies to measure methane production at a major LNG power plant in metropolitan Seoul. Using a mobile greenhouse gas (GHG) measurement platform, we identified and quantified the fugitive methane sources within a large LNG gas-fired power plant. This study also compared ground-based top-down measurements with bottom-up inventory estimates of fugitive methane emissions. Using a mobile GHG measurement platform, this study aimed to provide a more accurate assessment of fugitive methane emissions, overcoming the limitations associated with bottom-up methods in real environments. A comparison between ground-based top-down measurements and bottom-up inventories will contribute to a better understanding of the gap between current emissions reporting and real environmental methane emissions of LNG-based power generation in urban areas. In this study, the term “fugitive emissions” is used broadly to encompass all methane emissions detected by the mobile GHG measurement platform including unintentional equipment leaks, operational venting, and potential incomplete combustion products.

Figure 4 illustrates three key methane emission hotspots (S1, S2, and S3) using a mobile GHG platform near an LNG gas-fired power plant. Methane enhancements from the mobile measurements indicate a maximum of 3795.7 ppb (S1) with an average of 1698.62 ppb, a maximum of 1188.94 ppb (S2) with an average of 466.22 ppb, and a significant maximum of 56 039.06 ppb (S3) with an average of 19 963.97 ppb, identifying them as specific emission sources associated with the LNG gas-fired power plant. S1 and S2 were located downwind of sections of the plant's natural gas pipelines and LNG power plant facilities, such as power generation units and smokestacks. S3 was located downwind of an exhaust pipe associated with internal processes and LNG power plant facilities. The mobile measurement strategy employed in this study was effective in monitoring the target area surrounding the access-restricted LNG power plant, enabling the identification and characterization of methane emission plumes from the facilities.

The contrasting methane emission characteristics of S1, S2, and S3 are shown in Figs. 4 and 5. Figure 5 shows the relatively constant methane emissions during the 10–15 mobile measurement trajectories at S1 and S2. In contrast, Fig. 6 highlights the frequent and pronounced spikes in the methane and ethane concentrations measured downwind of the exhaust source (S3). Methane and ethane concentration data can be used to distinguish between fossil fuels and microbial sources (Joo et al., 2024). Ethane is co-emitted with methane from fossil fuel sources, such as natural gas leaks, whereas microbial processes typically produce methane with negligible amounts of ethane (Maazallahi et al., 2022; Joo et al., 2024). Figure 6 shows that the ethane concentrations increase and decrease in a pattern similar to the methane concentrations at S3, strongly suggesting that the methane emissions measured at S3 originate from the LNG power plant.

Methane emission rates at S1 and S2 were quantified using the GPDM from the mobile measurement data. The average emission rates of these hotspots were 90 ± 9 kgh-1 for S1 and 18 ± 2 kgh-1 for S2 (Table 3 and Fig. 5). The emission rate at hotspot S3 was quantified using OTM-33A, which is appropriate for the higher concentrations captured in this area, with an average emission rate of 550 ± 58 kgh-1 (Table 3, Fig. 6). Notably, the emissions at S3 exhibited substantial temporal fluctuation, ranging from 64 ± 9 kgh-1 to 2053 ± 283 kgh-1 (31 May; Table 3). Quantification of the methane emission rates at the three hotspots revealed distinct characteristics for the natural gas pipeline-associated locations versus the exhaust pipes of the LNG power plant-associated locations in Table 3 and Figs. 5 and 6. The two hotspots located downwind of the natural gas pipelines and LNG power plant facilities, S1 and S2, exhibited relatively lower and more consistent methane emission rates. In contrast, hotspot S3, which was located downwind of the exhaust pipe and LNG power plant facilities, displayed significantly higher average emissions and pronounced variability. This extreme variability at S3 strongly suggests the occurrence of intermittent, high-magnitude emission events potentially linked to specific operational phases (such as startups, shutdowns, or load changes) or process inefficiencies (such as incomplete combustion), aligning with the concept of “super-emitter” behavior noted in other industrial emission studies (Alvarez et al., 2018; Brandt et al., 2014).

Table 3 shows the significant discrepancies observed between our emission quantifications derived from mobile measurements and fugitive methane estimates based on standard national emission factors and activity data from the target LNG power plant. To enable a direct comparison, the bottom-up annual emission estimates were converted into hourly emission rates (kgh-1) based on the monthly fuel consumption and operating hours corresponding to the measurement period. Although bottom-up inventories generally underestimated emissions, particularly during the significant peak event at S3, there were other instances, particularly at S1 and S2, where the inventories overestimated the measured emissions. Specifically, bottom-up estimates of this study showed a relatively narrow range of 180 ± 499 kgh-1. In contrast, the measured emissions exhibited significant variability across sites, ranging from 7 to 302 kgh-1 at S1, 2 to 47 kgh-1 at S2, and 5 to 2053 kgh-1 at S3. The maximum methane emission rate at S3 on 31 May, 15:00–16:00, was 2053 ± 283 kgh-1, which was significantly higher than the bottom-up estimate of 350 kgh-1. However, Table 3 also indicates that in several instances at S1 and S2, the bottom-up estimates were higher than the measured methane emissions. This highlights the issue of temporal representativeness and emission variability. Although S1 and S2 are presumed to be continuous emission sources, their actual emission rates fluctuate depending on specific operational and meteorological conditions. Because our mobile measurements are snapshot-based (e.g., n=5 for each source), they directly capture the methane emission rates under those specific conditions. Conversely, bottom-up inventories annualize emissions using static factors, failing to account for such operational variability and intermittency. Therefore, the lower measured rates at S1 and S2 strongly suggest that the statistical inventory approach overestimates the emissions by smoothing out these dynamic conditions, rather than our observation-based estimates underestimating them. Discrepancies between fugitive methane estimates from bottom-up inventories and methane emissions from mobile measurements are influenced by several critical factors in LNG power plants. Operational variability, such as startups and shutdowns, can lead to short-term spikes in methane emissions that are difficult to capture using bottom-up methane estimates (Brandt et al., 2014). Furthermore, undetected or slowly developing fugitive emissions from the LNG facilities such as pipelines and fittings can be identified as intermittent “super-emitter” events by ground-based top-down measurements rather than bottom-up methane estimates (Howarth, 2019; Alvarez et al., 2018; Karion et al., 2013). Despite these variations, the discrepancy at source S3 is especially significant. While bottom-up estimates may slightly overestimate emissions at S1 and S2 during normal operations, they completely underestimate the magnitude of emissions at S3. Ground-based top-down measurements at S3 exceeded bottom-up estimates by a factor of 6. Due to the “heavy-tailed” nature of methane emissions, in which a few super-emitters often dominate total emissions (Brandt et al., 2014; Zavala-Araiza et al., 2015), the significant underestimation at S3 outweighs the minor overestimations at S1 and S2. However, characterizing the total methane emissions requires accounting for the full distribution of sources, not just the extreme outliers. Williams et al. (2025) recently demonstrated that dispersed sources emitting less than 100 kgh-1 can collectively account for a majority of total methane emissions in oil and gas basins. In our study, while the super-emitter S3 dominates the total methane emissions, the routine leaks detected at S1 and S2 fall into this category of smaller, dispersed sources, highlighting the need for a comprehensive monitoring strategy that captures both heavy-tailed events and widely distributed smaller leaks.

These findings are consistent with previous research, highlighting the difficulties faced by bottom-up methods in fully accounting for fugitive emissions from LNG power plants (Alvarez et al., 2018; Brandt et al., 2014; Howarth, 2019; Karion et al., 2013; Zavala-Araiza et al., 2015). Alvarez et al. (2018) reported that the actual methane emissions from the US oil and natural gas supply chain were approximately 60 % higher than the estimates provided by EPA inventories. Similarly, our ground-based top-down measurements at the S3 hotspot revealed peak emissions of 2053 ± 283 kgh-1, which are approximately 6 times greater than our corresponding bottom-up estimate of 350 kgh-1. The substantial discrepancy between the measured emission rate and the bottom-up inventory stems directly from the limitations of the input data described in the methodology. While we utilized 5 min operational proxies (MW), the underlying activity data relied on “monthly aggregated LNG consumption.” This temporal resolution inevitably smooths out instantaneous high-emission events. Furthermore, the applied emission factor (87.5 tCH4PJ-6) accounts only for routine “post-meter leakage” under normal operating conditions and does not capture stochastic super-emitter events – such as incomplete combustion during start-ups or pressure relief venting – which likely caused the elevated plumes detected at the S3 hotspot. This level of discrepancy aligns with measurements of “super-emitter” phenomena in natural gas facilities. Mitchell et al. (2015) found that the top 30 % of natural gas gathering facilities contributed 80 % of the total emissions. One facility alone accounted for 10 % of all measured emissions from the gathering facilities. This indicates that a few sources can have a disproportionate impact on the overall emissions. Mitchell et al. (2015) reported a median throughput-normalized weighted average facility-level emissions rate of 0.079 % for processing plants. These plants differ from LNG power plants but have similar components. The magnitude of our peak S3 emission (2053 kgh-1) suggests a significant emission event that far exceeds typical operational estimates. This level of discrepancy aligns with “super-emitter” phenomena measurements in natural gas facilities. Jia et al. (2025) quantified fugitive methane emissions from typical natural gas infrastructure components (e.g., valves, flanges) using on-site measurement techniques. They reported that the total annual emissions from an entire LNG terminal were approximately 5202 kgyr-1 (averaging 0.6 kgh-1), with individual components emitting significantly less. In comparison, the methane emissions identified in our study at sources S1 (90 ± 9 kgh-1) and S2 (18 ± 2 kgh-1) are orders of magnitude higher than these typical component-level methane emissions. These results suggest that the emissions at S1 and S2 are caused by high-pressure fugitive leaks or compromised infrastructure, which is different from the minor component-level leaks that are typical of standard distribution stations. Furthermore, Zimmerle et al. (2015) reported comprehensive transmission and storage (T&S) sector measurements. Their study determined the mean methane emission rates to be approximately 670 Mgyr-1 (equivalent to ∼ 76.5 kgh-1) for transmission stations and 847 Mgyr-1 (equivalent to ∼ 96.7 kgh-1) for underground storage facilities. When placed in this context, the emission rate at S1 (90 ± 9 kgh-1) is comparable to the average emission load of an entire US transmission station, indicating a substantial facility-level emissions. In stark contrast, the maximum emission rate at S3 (2053 ± 283 kgh-1) is exceptionally high, exceeding the average emissions of these large-scale midstream facilities by a factor of over 20. While S1 and S2 represent significant localized leaks, S3 falls into the category of extreme super-emitters, exhibiting discharge rates far surpassing typical operational emissions found in standard natural gas infrastructure.

To further contextualize the magnitude of these emissions, we compared our findings with the official 2023 GHG inventory of Seoul (SCNSC, 2026). The reported annual methane emissions from the entire energy fugitive sector were approximately 4464 t, and the wastewater treatment sector accounted for approximately 1607 t. Although emissions at S3 hotspot are likely intermittent, extrapolating the average measured rate of 550 kgh-1 to an annual scale yields a potential magnitude of approximately 4818 t. Notably, this potential figure from a single hotspot exceeds the total reported fugitive emissions for the entire Seoul and is nearly three times higher than the emissions from the wastewater treatment sector alone. This comparison illustrates how current bottom-up inventories may significantly underestimate urban methane budgets by failing to account for the intensity of industrial super-emitters.

The fugitive methane emissions identified in this study also represent a significant economic loss. Based on the maximum emission rate at S3 (2053 ± 283 kgh-1), we estimated the potential annual financial loss. Using the lower heating value (LHV) of methane and the power generation charge rates from the Korea Gas Corporation (KOGAS, 2026), the estimated annual loss amounts to approximately USD 15.77 million. It is important to note the conceptual distinction between instantaneous emission rates and total annualized emissions. Meaningful extrapolation to total emissions is only robust when the temporal distribution, frequency, and duration of emission events are adequately characterized. Therefore, this calculation assumes that the maximum methane emission event occurs continuously, serving strictly as an upper-bound estimate of the potential revenue loss. This substantial economic cost underscores the necessity and cost-effectiveness of implementing stricter leak detection and repair (LDAR) protocols to minimize fugitive emissions. The annual financial loss was calculated using Eq. (3); 5Annual financial loss(USDyr-1)=QtCH4h-1×8760[hyr-1]×LHV[GJt-1]×Chargeyear[USDGJ-1]×f where, Q is the methane emission rate (tCH4h-1), LHV is the lower heating value of methane (50.1 GJt-1), the Charge is based on KOGAS power generation rates converted to USD (2023 average exchange rate, USD 1 = KRW 1 305 979), f is the continuous operation factor (set to f=1).

This study identified and quantified fugitive methane emissions from an urban LNG gas-fired power plant in Seoul using a mobile measurement platform. By integrating mobile measurements with GPDM and OTM-33A, we demonstrated that ground-based top-down approaches can effectively quantify emission sources even in complex urban environments where facility access is restricted. Our measurements revealed three distinct emission hotspots with varying characteristics. Sources S1 and S2 were identified as continuous fugitive leaks associated with underground pipeline infrastructure, exhibiting consistent emission rates of 90 ± 9 and 18 ± 2 kgh-1, respectively. In contrast, source S3 was characterized as a stochastic super-emitter, with a peak emission rate reaching 2053 ± 283 kgh-1. The magnitude and intermittency of S3 suggest it is driven by operational venting or incomplete combustion, events that are typically invisible to standard leak detection protocols. A critical finding of this research is the substantial discrepancy between our ground-based top-down measurements and bottom-up inventory estimates. The high temporal variability observed at S3 demonstrates that emission factors used in current inventories fail to capture transient, high-intensity events. This disparity underscores the inherent limitations of current greenhouse gas inventory calculation methods, which often fail to account for fugitive emissions arising from specific, variable operational conditions. Furthermore, the estimated potential financial loss of approximately USD 15.77 million yr-1 underscores the economic urgency of addressing these leaks.

In conclusion, relying solely on current inventories can lead to a significant underestimation of the methane emissions from urban energy infrastructure. We recommend implementing continuous methane monitoring systems and conducting regular mobile measurements to capture the full distribution of emissions, including those from heavy-tailed super-emitters. These transparent, measurement-based strategies are crucial for accurately assessing and effectively mitigating greenhouse gas emissions in the energy sector.