Satellite-Derived Approaches for Coal Mine Methane Estimation: A Review

Highlights

What are the main findings?

• Satellite observations have advanced from coarse (~10 km) to fine (25 m) resolution, greatly improving plume detection and narrowing gaps between bottom-up and top-down methane estimates.
• Significant uncertainties persist due to wind variability, terrain heterogeneity, retrieval-algorithm limitations, and limited temporal sampling, leading to inconsistent quantification, especially for diffuse surface coal-mine emissions.

What is the implication of the main finding?

• Reliable site-level methane estimates require integrating satellites with ground and aerial validation, improved atmospheric modelling, and machine-learning methods that address diffuse and variable coal mine emissions.
• Reducing these uncertainties is essential for accurate reporting and effective mitigation in coal mining regions, thereby strengthening global climate action.

Abstract

Methane emissions from coal mines, especially surface operations, are spatially diffuse, presenting significant challenges for accurate quantification. Satellites such as TROPOMI, GHGSat, PRISMA, GaoFen-5, and GOSAT have been extensively used for detecting methane emissions at various scales, from individual point sources to regional and global assessments. Despite various advancements, methane quantification via satellite observations remains subject to several challenges. Various quantification methods for the same observation can produce variable results. Also, meteorological conditions, terrain complexity, and surface heterogeneity introduce uncertainties in emission estimates. The selection of wind speed and direction, along with retrieval-algorithm limitations, can lead to significant discrepancies in reported emissions. Additionally, satellite-based observations capture emissions only at specific overpass times, which may introduce temporal uncertainties compared to inventories derived from continuous emission estimations. This study provides a comprehensive review of satellite-based coal mine methane (CMM) monitoring, evaluating current methodologies, their limitations, and recent technological advancements. We discussed the potential of emerging machine-learning techniques, improved atmospheric modelling, and integrated observational approaches to enhance methane emission quantification. By refining satellite-based monitoring techniques and addressing existing challenges, this research will support the development of more accurate emission inventories and effective mitigation strategies for the coal mining sector.

1. Introduction

Globally, greenhouse gas (GHG) emissions have continued to rise due to human-induced activities [1,2,3,4,5,6]. Carbon dioxide (CO₂) and Methane (CH₄) are two major GHG contributors to global warming and climate change [7]. Methane, though accounting for 18% of greenhouse gas emissions, has a warming potential 86 times greater than CO₂ over a 20-year period; however, it has a relatively short atmospheric lifetime of 12 years [8]. It is responsible for 30% of the current rise in global temperature [9]. Therefore, mitigation of methane emissions will help limit the atmospheric warming as well as improve air quality, because methane also contributes to surface ozone formation, a key factor in air-quality deterioration [10]. In response to growing concerns, the Global Carbon Project (GCP) was established in 2001 as a core research initiative of Future Earth and a partner of the World Climate Research Programme [11]. Its objective is to advance a comprehensive understanding of the global carbon cycle, integrating biophysical, biogeochemical, and human dimensions to support policies aimed at stabilizing atmospheric GHG concentrations. The Intergovernmental Panel on Climate Change (IPCC) has raised serious concerns regarding the rise in global GHG emissions, as the past decade experienced a temperature increase of 1.1 °C [8]. Warming beyond 2 °C could have devastating consequences for ecosystems worldwide [12]; hence, the international goal is to limit global warming to well below 2 °C, preferably 1.5 °C [13]. To achieve this, the Paris Agreement set a target to reduce global GHG emissions by 43% by 2030 and reach net zero by 2050 [13]. Building on this framework, during COP26 (2021), the United States and the European Union, along with more than 150 participating countries, launched the Global Methane Pledge (GMP), a commitment to reduce anthropogenic methane emissions by 30% from 2020 levels by 2030 [14]. Further strengthening global observation and mitigation efforts, the World Meteorological Congress (WMC) introduced the Global Greenhouse Gas Watch (G3W), an integrated global initiative combining satellite and ground-based observations with modelling systems [15]. G3W aims to deliver timely, high-resolution CH₄ flux data to support World Meteorological Organization (WMO) Members in monitoring progress and guiding mitigation efforts under the Paris Agreement. Yet, despite these ambitious initiatives, current mitigation efforts remain insufficient to keep global temperatures within the targeted thresholds [16,17]. This highlights the critical need to intensify GHG reduction strategies, particularly through innovation in detection, quantification, and monitoring technologies.

Various natural and anthropogenic sources impact the global methane budget; however, energy and agriculture-related emissions together cause more than 60% of the total methane budget [18,19]. Fugitive emissions from the energy sector alone contribute approximately 32% (ranging from 22% to 42%) of total global methane emissions, stemming from oil and gas operations, coal mining, and other sources [8]. The fugitive emissions originate from a variety of sources, ranging from small-point sources such as leaks and flaring to broader-area sources such as surface coal mines and landfills exhibiting highly variable emission rates, from a few kilograms per hour (kg/h) to several tonnes per hour (t/h). Saunois et al. [4] presented the most recent global methane budget for 2010–2019, integrating bottom-up inventories and top-down inversion estimates based on GOSAT satellite observations. The global CH₄ emissions were estimated at 669 Tg CH₄ yr⁻¹ (bottom-up) and 575 Tg CH₄ yr⁻¹ (top-down), with an atmospheric growth rate of 20.9 Tg CH₄ yr⁻¹. Fossil fuel–related CH₄ emissions were estimated at 105 Tg CH₄ yr⁻¹ during 2000–2009 (bottom-up: 97–123 Tg yr⁻¹; top-down: 88–115 Tg yr⁻¹), increasing to 120 Tg CH₄ yr⁻¹ (bottom-up: 117–125 Tg yr⁻¹) and 115 Tg CH₄ yr⁻¹ (top-down: 100–124 Tg yr⁻¹) during 2010–2019. An additional enhancement of 7.5 Tg CH₄ yr⁻¹ was observed in 2020. Among fossil fuel sources, coal mining contributed 30, 40, and 41 Tg CH₄ yr⁻¹ during 2000–2009, 2010–2019, and 2020, respectively, making it the second-largest contributor after the oil and gas sector [4]. According to the 2024 publication by the IEA, coal contributed approximately 6.85% to worldwide methane (CH₄) output [19].

Coal is a major source of energy, as it is still the highest contributor among all other conventional energy sources and hence caused 33% of global CO₂ emissions by 2019 [4,8]. During the coalification process, methane is generated primarily due to biological activity at early stages (biogenic methane) and thermal decomposition of organic matter at later stages (thermogenic methane) and becomes trapped in the coal seams [20]. Methane is pre-drained from boreholes drilled into the seam prior to mining to reduce in situ gas pressure and enhance operational safety. The member states of the Organization for Economic Co-operation and Development (OECD) regulated the coal-bed methane extraction for a period of 10 years before the start of the mining [4]. The amount of extractable methane depends on factors such as coal type, depth of the seam, and geological conditions [21]. Mining activities further facilitate the release of methane through diffuse emissions generally referred to as CMM. The gas content increases with the depth of the coal seam, leading to generally higher methane emissions from deep-seated coal deposits [21]. To mitigate explosion risks during mining operations in an underground mine, since methane-air mixtures become explosive at concentrations of 5%, ventilation systems are used to maintain methane concentrations in the range of 0.1–0.5% [4,22]. Ventilation air methane (VAM) is commonly released into the atmosphere or incinerated, despite ongoing efforts under the UNFCCC’s Clean Development Mechanism [23] to improve coal mine gas recovery. In recent years, China has taken steps to regulate these emissions. As outlined in the Methane Emissions Control Action Plan dated 7 November 2023, the country introduced policies aimed at boosting VAM utilization to 6 billion cubic meters by 2025, with a goal of aligning with international benchmarks by 2030 [24]. VAM behaves as a point source and has a higher possibility to be detected as well as quantified using existing techniques [4]. However, methane emissions from surface mines present significant challenges due to the diffuse and multifaceted nature of their sources. In surface mining, activities such as drilling, blasting, excavation, transportation, washing, and stockpiling, as well as emissions from the exposed coal in the pits themselves, contribute to CMM. The diverse and dispersed nature of these sources makes quantification and mitigation particularly complex [25].

In recent years, a diverse range of observation technologies has rapidly advanced for monitoring methane emissions, particularly remote sensing systems offering higher spatial resolution. Methane is now observed using satellite, aerial, UAV, and ground-based sensors, each characterised by distinct detection limits and operational constraints. Key ground and aerial technologies include infrared imaging [26], laser absorption spectroscopy methods [27,28], metal-oxide-based point sensors [29], cavity ring-down spectroscopy [30], non-dispersive infrared (NDIR) sensors [31], ground- and aerial-based passive SWIR spectroscopy [32], and Fourier-transform infrared spectroscopy (FTIRS) [33,34,35]. Ground observation methods often have higher observational accuracy with better temporal coverage but limited spatial coverage. Arial and UAV-based instruments, on the other hand, have higher spatial coverage with point-to-area emissions observations [36]. Satellite remote sensing of methane, particularly, is a cost-effective approach with higher spatial coverage and more mature in terms of plume modelling [36]. Most satellites estimate atmospheric methane using passive remote sensing in the shortwave infrared (SWIR) spectrum, where methane exhibits strong absorption features [37,38,39,40]. However, the mid-wave infrared (MWIR) spectrum also shows sensitivity to atmospheric methane and can be used under certain conditions [41]. Figure 1 displays various current and future satellite missions for dedicated methane observation, and a few tailored solutions due to their observations in the SWIR band. GOSAT, SCHAMECY, TROPOMI, MethaneSAT, Carbon Mapper’s Tanager, and GHGSat are the current dedicated methane observation satellites; however, EMIT, EnMAP, PRISMA, Sentinel-2, Landsat, WorldView-3, and GaoFen-5 are the hyperspectral missions that can be utilized for methane plume detection. These satellites can also be further categories as point-source imagers and Area Flux mapper based on the swath of the satellite. The flux detection limits of the point-source satellites are better than the area mapper; for example, GHGSat can detect an emission rate close to 100 kg/h [42]. However, the area mapper, such as TROPOMI, can scan a large area and compromise the minimum detection limits but provide daily global coverage. MethaneSAT, on the other hand, was designed to estimate the dispersed emission above 500 kg/h with a wider view, which is often limited by high-precision satellites [43]. However, on 20 June 2025, MethaneSAT lost contact with ground mission controllers, completing its duty for a period of almost a year [43]. Tanager-1 Satellite, the mission initiated by Carbon Mapper, claimed to achieve a minimum emission rate in the range of 66–144 kg/h of methane with 30 m spatial resolution and 19 km swath, and a constellation of satellites for a higher revisit rate [44].

Several studies have been conducted to estimate regional and global methane budgets [4,5,6,45,46,47,48,49,50,51,52,53,54]. These studies employed approaches ranging from bottom-up inventories to satellite-based observations to assess global methane trends across multiple sectors, with some specifically focusing on CMM. Between 2015 and 2024, 24 review studies were published addressing CMM, of which eight specifically examined methane emission observation, quantification, and sensing technologies. Additionally, four new review studies were published between 2024 and 2025, highlighting the growing research interest in this area.

Dreger and Kędzior [55] reviewed two decades of coal production and associated methane emissions in the Upper Silesian Coal Basin, Poland, using IPCC reporting frameworks. Gao et al. [50] analysed China’s CMM emissions using bottom-up inventory approaches. Kai et al. [56] presented the first comprehensive review of advancements in remote sensing technologies and their application to the coal-mining industry, focusing primarily on satellite-based observations over China and highlighting the limitations of current satellite missions in detecting coal mine emissions. Kim et al. [57] compared recent progress in satellite-based methane monitoring but did not address the challenges related to emission quantification. Karacan et al. [52] assessed the status of CMM abatement policies and their implementation. Zieba and Smoliński [58] evaluated methane emissions from coal mines across the European Union. Mohammadimanesh et al. [59] systematically reviewed satellite-based methane observation and quantification techniques, identifying key limitations in detecting point-source emissions and noting that only 77 relevant studies were found across all emission sectors. Ke et al. [60] examined recent developments in metal-oxide sensor technologies for methane detection. Collectively, these studies reveal a clear gap in the literature regarding the discussion and analysis of the quantification of CMM using satellite-based remote sensing and discuss their limitations. While Kai et al. [56] discussed various satellite remote sensing techniques, the focus of the present review is to extend beyond observational aspects and emphasise quantification approaches specifically applicable to coal mining emissions.

While several reviews have examined CMM emissions from policy, inventory, and sensor technology perspectives, a major gap remains in the integration of satellite-based observation and quantification approaches. In contrast to oil and gas infrastructure, where strong economic incentives have driven rapid advances in detection and quantification technologies, coal mining sites, particularly surface operations, present distinct technical challenges for both methane detection and quantification. Addressing this gap, the present review focuses on studies from 2015 to September 2025 that apply satellite-based remote sensing to CMM detection and quantification. Methane emissions from coal mines exhibit wide variability, ranging from diffuse, low-level releases linked to mining activities to concentrated VAM that can reach “super-emitter” levels of several tons per hour [61]. Emissions from surface mines, though generally weaker (a few kilograms per hour), remain difficult to detect from space. Nonetheless, multiple studies demonstrate that satellite observations can effectively capture basin-scale methane distribution patterns, offering valuable insights into regional emission dynamics. Accordingly, this review evaluates the capabilities of current satellite missions, compares emission estimation methods, examines observational limitations, and outlines future research strategies to enhance the role of satellite-based monitoring in supporting mitigation and policy efforts.

2. Materials and Methods

Based on the scope of the current review, we did a literature survey based on the Scopus database using the keyword “coal + mine + methane”. Other search criteria are given in

Table 1. The search criteria for the current manuscript.

Item	Source
Database	Scopus
Seach Title	coal AND mine AND emission
Focus group	Satellite Observations
Time range	2015 to 2026
Document Type	Article, Conference Paper, Review, Book Chapter, Letter, Editorial, Data paper
Language	English

. In the past 12 years, satellite remote sensing technologies have been developed at a rapid pace. Technology has evolved from SCIAMACHY and GOSAT-like sensors with a revisit rate of several days to GHGSat constellations like point observers with pinpointing of leaks from gas pipelines. This review compiles satellite-based studies on CMM emissions conducted between 2015 and 2026, highlighting observational efforts over the past decade. As per the SCOPUS database, a total of 2431 articles were published in the past 12 years with the three key words (CMM). A large number of research papers focused on the detection of methane from a mine safety perspective, not for the emission quantification. More than 136 studies have been conducted to quantify CMM using various ground, aerial, and satellite-based observation techniques. In the present manuscript, we focus exclusively on satellite-based observations (

Table 2. List of satellites used for the CMM emission observations and their specifications.

S. No.	Sensor Name	Revisit Rate	Type of Sensor	Methane Band	Resolution	Min Emission Rate	Swath Width
1	TROPOMI	1 Day	Hyperspectral Imaging	2.2–2.4 µm	5.5 × 7 km	~10 t/h	2600 km
2	GaoFen-5	4–5 Days	Hyperspectral Imaging	2.11–2.45 μm	30 × 30 m	~1 t/h	60 km
3	PRISMA	2–3 Days	Hyperspectral Imaging	2.10–2.45 μm	30 × 30 m	~0.5 t/h	30 km
4	GHGSat-D	1 Day	Remote Sensing (Optical)	1.63–1.675 μm	25 × 25 m	~0.1 t/h	12 km
5	GOSAT	3 Days	Atmospheric Monitoring	1.65 μm	10 × 10 km2	~50 t/h	100 km
6	IASI (METOP)	12 h	Infrared Atmospheric Sounding	3.7–15 μm	5.5 × 7 km2	~0.1 Tg CH4/year	2200 km
7	EnMAP	27 Days	Hyperspectral Imaging	2.10–2.45 µm	30 × 30 m	~1 t/h	30 km
8	Ziyuan (ZY-01, 02)	3 Days	Hyperspectral Imaging	2.10–2.45 µm	30 × 30 m	~1 t/h	115 km
9	EMIT	Variable	Hyperspectral Imaging	2.10–2.45 µm	60 × 60 m	~1 t/h	75 km

), as they provide consistent, large-scale spatial and temporal coverage that is not feasible with ground or aerial methods. This focused approach allows for a more coherent comparison of methodologies and results within a single observational framework, avoiding inconsistencies that may arise from integrating fundamentally different measurement platforms.

For CMM emission observations using satellite, there are only 23 research articles (

Table 3. Summary of Satellite-Based Methane Emission Quantification Studies across Different Regions.

S. No.	Title	Region	Satellite	Quantification Methods	Identified Emission Rates and Available Uncertainties
1	Merging TROPOMI and eddy-covariance observations to quantify 5 years of daily CH4 emissions over a coal-mine dominated region	China	TROPOMI	Model-free mass balance	Estimated CH4 emissions over Shanxi, China, at 126 ± 58.8 µg m−2 s−1 using TROPOMI data integrated with high-frequency eddy-covariance fluxes, revealing strong daily variability and reduced bias through high-frequency observations.
2	High-resolution satellite estimates of coal mine methane emissions from local to regional scales in Shanxi, China	China	EMIT, EnMAP, GaoFen-5B, ZY1-02D	Integrated mass enhancement method	Estimated CH4 emissions over Shanxi, China, during 2019 to 2023 at 8.9 ± 0.5 Tg yr−1 using High-resolution satellites. Reported a decline of 20 to 30 ppb in mean ΔXCH4 using TROPOMI in Shanxi, China.
3	COCCON Measurements of XCO2, XCH4 and XCO over Coal Mine Aggregation Areas in Shanxi, China, and Comparison to TROPOMI and CAMS Datasets	China	TROPOMI	No quantification, only concentration analysis	First ground-based XCH4 estimates with EM27/SUNs in Shanxi, China, showed good agreement with TROPOMI observations, with a mean bias of 7.15 ± 9.49 ppb. No emission rate estimation.
4	Seasonal and trend variation of methane concentration over two provinces of South Africa using Sentinel-5p data	South Africa	TROPOMI	No quantification, only concentration analysis	Used satellite observations for seasonal analysis of methane in coal mine provinces of South Africa and observed an increasing trend of XCH4 from 2019 to 2023.
5	Unveiling Unprecedented Methane Hotspots in China’s Leading Coal Production Hub: A Satellite Mapping Revelation	China	GaoFen 5B	Integrated mass enhancement method	Using GaoFen-5B observations, detected 138 plumes over 82 facilities in Shanxi, China, and estimated an emission of 1.2 (+0.24/−0.20) Mt CH4/y. Reported the intermittent characteristics of methane on the activity level.
6	A survey of methane point-source emissions from coal mines in Shanxi province of China using AHSI on board GaoFen-5B	China	GaoFen 5B	Integrated mass enhancement method	Using GaoFen-5B observations, detected 93 plumes over 32 facilities in Shanxi, China. Discussed the necessity of careful plume detections under variable surface conditions.
7	Quantifying CH4 emissions from coal mine aggregation areas in Shanxi, China, using TROPOMI observations and the wind-assigned anomaly method	China	TROPOMI	Wind-assign anomaly method coupled with the cone-plume model	Estimated the CMM emission in three regions of Shanxi, China, with emission rates of 0.706 Tg yr−1 ± 25%, 1.176 Tg yr−1 ± 20% and 0.412 Tg yr−1 ± 21% and reported a discrepancy of 64 to 176% with EDGARv7.0.
8	Exploiting the entire near-infrared spectral range to improve the detection of methane plumes with high-resolution imaging spectrometers	Poland, China	PRISMA	Integrated mass enhancement method	The study checked the match-filter method of plume detection for PRISMA over Chinese coal mines.
9	High-resolution assessment of coal mining methane emissions by satellite in Shanxi, China	China	TROPOMI	HYSPLIT Model	Using the HYSPLIT model with TROPOMI observations, CMM emissions in Shanxi were estimated to be 8.5 ± 0.6 Tg CH4 yr−1 (2019) and 8.6 ± 0.6 Tg CH4 yr−1 (2020).
10	Automated detection and monitoring of methane super-emitters using satellite data	Global	GHGSat-C, PRISMA, Sentinel-2, EMIT TROPOMI,	Integrated mass enhancement method	The study focused on a CNN-based plume-detection approach using multi-satellite data across the globe.
11	Huge CH4, NO2, and CO Emissions from Coal Mines in the Kuznetsk Basin (Russia) Detected by Sentinel-5P	Russia	TROPOMI	No quantification, only concentration analysis	The study found a total of 339 events using TOPOMI observations. However, no flux estimation was conducted.
12	Observed changes in China’s methane emissions linked to policy drivers	China	GOSAT	GOES-Chem-based Inverse modelling	Estimated the long-term trends of Chinese coal mines using satellite-based model emissions in China.
13	Quantifying CH4 emissions in hard coal mines from TROPOMI and IASI observations using the wind-assigned anomaly method	Poland	TROPOMI + IASI (METOP)	Wind-assign anomaly method coupled with the cone-plume model	The USCB annual CH4 emissions were estimated to be 496 kt/y from TROPOMI and 437 kt/y from TROPOMI–IASI observations. These estimates closely match the E-PRTR (448 kt yr−1) and CoMet (555 kt yr−1) inventories.
14	Methane Emissions from Superemitting Coal Mines in Australia Quantified Using TROPOMI Satellite Observations	Australia	TROPOMI	Cross-sectional Flux method	Estimated the TROPOMI-based flux for the Australian coal mines and compared it with IPCC-based emission estimates.
15	Mapping methane point emissions with the PRISMA spaceborne imaging spectrometer	China	PRISMA	Integrated mass enhancement method	Tested the PRISMA capabilities in the flux observation for various point sources in coal mines and other sources.
16	Sustained methane emissions from China after 2012 despite declining coal production and rice-cultivated area	China	GOSAT	NAME model emission estimations	Estimated the long-term trends of Chinese coal mines using satellite-based model emissions in China.
17	China’s coal mine methane regulations have not curbed growing emissions	China	GOSAT	GOES-Chem and Bayesian emission estimations	Estimated the long-term trends of Chinese coal mines using satellite-based model emissions in China.
18	From data to actionable insight: Monitoring fugitive methane emissions at oil and gas facilities using satellites	Global	GHGSat-D	No quantification, only technological discussions	Discussed the technological aspect of the GHGSat in various methane emission detection and quantification.
19	Quantifying Time-Averaged Methane Emissions from Individual Coal Mine Vents with GHGSat-D Satellite Observations	China, USA, Australia	GHGSat-D	Integrated mass enhancement and the Cross-sectional flux method	Estimated the coal mine emission in the USA, China, and Australia using high-resolution GHGSat-D observation and tested both IME and CSF methods.
20	The added value of satellite observations of methane for understanding the contemporary methane budget	Australia	TROPOMI	Model-free mass balance	Estimated the global CH4 budget using satellite observations along with satellite observation-based flux estimation of Australian mines.
21	Temporal and spatial comparison of coal mine ventilation methane emissions and mitigation quantified using PRISMA satellite data and on-site measurements	USA	PRISMA	Integrated mass enhancement	Ventilation air methane (VAM) emissions in Virginia, USA (2020–2023), were estimated using PRISMA satellite observations combined with ground-based measurements. ERA-5–based top-down estimates were higher than those using GOES-FP inputs, with in situ measurements indicating median emission rates of ~3800 kg h−1 (VS12) and ~2000 kg h−1 (VS16).
22	High-Resolution Satellite Reveals the Methane Emissions from China’s Coal Mines	China	EMIT	Integrated mass enhancement	Estimated emission intensities of 0.48 kg GJ−1 from CMM, derived from 88 plumes observed by EMIT and available through the Carbon Mapper portal between January 2023 and July 2024, were higher than those reported by EDGAR v8.0 (0.24 kg GJ−1) and GFEI v2 (0.18 kg GJ−1).
23	Global Methane Retrieval, Monitoring, and Quantification in Hotspot Regions Based on AHSI/ZY-1 Satellite	China	ZY1-02D	Integrated mass enhancement	High-resolution ZY-02D satellite observations were used to estimate CMM emission rates in China between July and December 2023. Three distinct plumes detected over coal mining sites exhibited emission rates of 6029 ± 1033 kg h−1, 4298 ± 762 kg h−1, and 6146 ± 1126 kg h−1, respectively.

) that were published. In our current review of methane emissions from coal mines, we have structured our analysis into key categories: types of sensors and satellites, data extraction and synthesis methods, evaluation of sensor capabilities, accuracy assessment, geographic distribution of studies, validation of the results and future aspect. This framework allows for a comprehensive examination of the available literature on CMM quantification using satellite observations. This review aims to provide valuable insights for both the scientific community and policymakers to enhance their understanding of the satellite-based CMM quantification approaches.

3. Satellite Platforms Used for CMM Observations

This section discusses remote sensing satellites used for space-based observations of CMM emissions, along with various models and methods for quantifying emission rates. It also explores the techniques employed, their limitations, and the overall effectiveness of current observational approaches.

Methane exhibits two absorption bands in the shortwave infrared (SWIR) region: a weaker band near 1700 nm and a stronger band around 2300 nm (Figure 2; [41]). Spaceborne shortwave infrared (SWIR) sensors with higher spectral resolution help in the characterisation of the Earth’s surface, biosphere, and atmospheric and estimation of chemical and physical properties [62,63,64]. Within the realm of spaceborne remote sensing, this improved spectral resolution has been leveraged across a variety of Earth observation fields from LULC changes to air quality. Atmospheric methane retrievals are performed in spectrally resolved observations of solar irradiation reflected off the Earth’s atmosphere within the SWIR spectrum, approximately 1.6 to 2.5 µm, and the MWIR spectrum, around 3.5 and 5 µm [65].

These absorption features are instrumental in satellite-based retrievals, enabling the estimation of column methane concentrations through atmospheric radiative transfer models (RTMs). SCIAMACHY and GOSAT were the first two atmospheric methane observation missions, and with continuous development in sensor technologies, various missions have been launched (Figure 1). Although satellites like Sentinel-2, WorldView-3, GaoFen-5, Landsat, EMIT, PRISMA, and EnMAP were not specifically designed for methane observations, their SWIR measurements have been successfully utilized for detecting and quantifying methane emissions [61,66,67,68]. For CMM quantification, the most utilised satellite observations include those from TROPOMI, GaoFen-5, PRISMA, GHGSat-D, GOSAT, and IASI. In addition to these, observations from EMIT, EnMAP, and ZY1-02D have also been applied for CMM emission observation. The specifications and capabilities of these satellite platforms relevant to CMM monitoring are summarised in Table 2 and discussed in the following sections.

3.1. TROPOMI

The TROPOspheric Monitoring Instrument (TROPOMI) was co-funded by ESA and the Netherlands. Key organizations from the Netherlands include KNMI (Royal Netherlands Meteorological Institute), SRON (Space Research Organization Netherlands), TNO (Netherlands Organization for Applied Scientific Research), and Airbus DS-NL, on behalf of NSO (Netherlands Space Office). KNMI and SRON are responsible for the development of Level 1B and some Level 2 products of satellite observation (e.g., Methane, Nitrogen Dioxide). With a swath width of 2600 km and spatial resolution of 5.5 × 7 km², it provides daily global coverage [40,69,70]. It carries out the observation of methane in the SWIR band at 2.3 µm (2314–2382 nm) [40]). The local overpass time is around 1:30 PM for TROPOMI. Global methane observations from TROPOMI have been widely used to study CMM emissions in regions such as Poland, Australia, China, South Africa, and other parts of the world [68,70,71,72,73,74,75,76,77,78,79].

3.2. GaoFen-5

GaoFen-5 (GF 5) is part of China’s civilian Earth observation satellite series under the China High-Definition Earth Observation System (CHEOS) program. It is designed for advanced remote sensing applications to support environmental monitoring, resource management, and other state-sponsored initiatives. GF 5 or GF 5A or GF 5-01 was the first satellite of the GaoFen-5 series, and GF 5-02 or GF-5B was the second satellite. The first satellite was launched on 08 May 2018, and the second satellite was launched on 7 September 2021. The GaoFen-5 mission has several observation sensors from Hyperspectral imagers, Visible sensors, Atmospheric Infrared Ultra spectrometer, Directional Polarisation Camera, Environmental and GHG observation instruments. The Advanced Hyperspectral Imager (AHSI) on GaoFen-5 can capture images in a wide band ranging from 400 nm to 2500 nm, with a spatial swath of 60 km and a spatial resolution of 30 m [80,81]. He et al. [82], Han et al. [83], and Bai et al. [66] utilized its 2100 to 2450 nm spectral window for column mass methane observations, focusing on CH₄ absorption features near 2300 nm. The local overpass time for Gaofan-5 is around 1:30 PM.

3.3. PRISMA

PRISMA (PRecursore IperSpettrale della Missione Applicativa) is a dedicated hyperspectral mission launched by the Italian space agency for Earth observations. The satellite can obtain the hyperspectral images within ~400–2500 nm band with a spatial resolution of 30 m and a swath size of 30 × 30 km² [84]. The spectral resolution for PRIMA is not uniform, and it varies from 9 nm to 15 nm. The band near 2300 nm with a spectral resolution of 10 nm can be used for the methane column mass inversions [85]. The inversion of the hyperspectral image for methane observations can be carried out in the wavelength band from 2280 nm to 2380 nm, as it is affected most by methane and water vapor absorption for PRISMA. The PRISMA satellite, operating in a sun-synchronous orbit, has a local overpass time of approximately 10:30 AM. PRISMA observations have been utilised for CMM estimation in Poland [41], China [41,85], and the rest of the world [68].

3.4. GHGSat-C/D

GHGSat-C/D are part of a constellation of small satellites designed for global methane observations. They were designed and operated by GHGSat Inc. for high-resolution methane observation, capable of detecting emissions as low as ~100 kg/h with high precision [37,86]. With a higher spatial resolution of 25 m × 25 m and a 12 × 12 km² swath, GHGSat provides high-resolution methane inversion results [36]. Using solar backscatter observations in the spectral range 1630–1675 nm of the SWIR band, GHGSat-C/D can measure the atmospheric column methane. GHGSat-D has a mean return time of 2 weeks with an overpass time of 10:00 AM local solar time on the ground. However, GHGSat-C has an overpass time of 09:30 AM. With a constellation of satellites by the GHGSat team, they can capture daily observations of a target source on a daily basis [87]. The detailed technical discussion about the GHGSat satellites is discussed by Varon et al. [36]. GHGSat observations have been used for CMM estimation in the USA [61], Australia [61], and China [26,61], as well as for global plume detection [68].

3.5. GOSAT

GOSAT is a JAXA mission within Japan’s GCOM (Global Change Observation Mission) programme designed to monitor the global distribution of carbon dioxide and methane. GOSAT (Greenhouse gases Observing SATellite) is an environment-monitoring satellite developed by Japan, which launched on 23 January 2009 and remains operational. It is part of ESA’s Third-Party Missions Programme, in which ESA has an agreement with JAXA to distribute data products from the mission. Thermal and Near-Infrared Sensor for Carbon Observation (TANSO)—onboard GOSAT consists of two instruments: the TANSO-FTS that observes the greenhouse gases, and the other is the TANSO-CAI that senses clouds and aerosols. The satellite has a sun-synchronous polar orbit with a local pass time of 01:00 PM. GOSAT-1 performs nadir measurements of solar backscatter in the SWIR spectrum to retrieve methane column densities (1.65 μm), with high sensitivity in the troposphere and weaker sensitivity in the stratosphere [88]. The satellite observes three circular pixels, each 10 km in diameter, spaced 260 km apart along the orbit track. It samples the same locations every three days with global coverage. GOSAT observation helped in the development of methane emission observation and trend estimation. Miller et al. [89], Sheng et al. [90], and Zhang et al. [91] investigated the methane emission trends in China using GOSAT observations.

3.6. IASI (METOP)

The Infrared Atmospheric Sounding Interferometer (IASI) is a nadir-viewing Fourier-transform spectrometer (FTS) installed on METOP satellites (A, B, and C). Operating within the MWIR band (3.7–15 μm), IASI exhibits high sensitivity, particularly in the middle troposphere and stratosphere. It offers a spatial resolution of approximately 12 km and covers a swath width of 2200 km, with an equatorial overpass occurring at 09:30 and 21:30 local time. With nearly 14 orbits per day, IASI data have been extensively utilised for trace gas profiling [92,93,94,95]. Tu et al. [78] presented the observation of tropospheric methane using IASI observations coupled with TROPOMI.

3.7. Other Point Observations Satellites

EMIT (Earth Surface Mineral Dust Source Investigation), EnMAP (Environmental Mapping and Analysis Program), and Ziyuan-1 02D are hyperspectral imaging missions not originally designed for methane detection but have recently been utilised for methane observations due to their high spectral resolution in the shortwave infrared (SWIR) range. EMIT, operated by NASA aboard the International Space Station, captures data in the 980–2130 nm range with a spatial resolution of ~60 m and has been used to detect strong methane plumes. EnMAP, operated by DLR, covers 420–2450 nm with 30 m spatial resolution and a swath of 30 km, offering daytime overpasses at around 11:00 AM local time. It has demonstrated utility in capturing methane enhancements over coal mines under suitable conditions. ZY1-02D, launched by China with the Advanced Hyperspectral Imager (AHSI), provides 30 m resolution over a 60 km swath in the 400–2500 nm range and is used in methane analysis, particularly in a study over Shanxi, China [66]. While these missions are not dedicated to methane monitoring, their SWIR observations are being exploited for high-resolution CMM detection and analysis.

4. Methods Used for CMM Quantifications

The quantification of methane emissions, particularly flux estimation, involves multiple steps (Figure 3). Satellites observe spectrally resolved solar radiation reflected from the Earth’s surface and atmosphere in the shortwave infrared (SWIR) spectrum, which is processed as Level 1 radiance data. Using radiative transfer models (RTMs), this spectral information is inverted to retrieve the column-averaged dry-air mole fraction of methane (XCH₄), typically referred to as Level 2 data. Subsequently, XCH₄ data are used to estimate emission rates or fluxes through various approaches, such as mass balance, plume inversion, or data assimilation techniques. This section covers the methodologies used in the studies so far for estimating CMM fluxes using satellite-based observations.

4.1. Inversion Methods

The inversion of raw satellite radiances (Level 1) to column-averaged dry-air mole fraction of methane (XCH₄) or Level 2 data involves various inversion techniques. A clear-sky radiative transfer model is integrated within an inverse modelling framework to retrieve methane concentrations from satellite-based imaging spectroscopy data. These methods primarily rely on radiative transfer processes occurring between the atmosphere and the observing instrument. RemoTeC full-physics algorithm was first developed for GOSAT [96,97,98] and later was also utilised for TROPOMI observations for column mass concentration of methane [78]. For CMM observations, RemoTeC retrieved dry-air column mass methane (XCH₄) observations have been applied in 13 studies for TROPOMI and GOSAT inversions [68,71,72,73,74,75,76,77,78,79,89,90,91]. For TROPOMI, the model atmosphere is divided into 36 uniform vertical layers, with input profiles of trace gases and meteorological parameters sourced from ECMWF data. The retrieval algorithm relies on spectral bands, specifically the Near-Infrared (NIR) and Short-Wave Infrared (SWIR), to extract aerosol information. In the NIR band, O₂ absorption features are used, while in the SWIR band, absorption by CH₄ and H₂O provides sensitivity to aerosol scattering effects. These spectral regions allow for the retrieval of aerosol amount, size, and height, which are simultaneously estimated along with methane columns to account for aerosol-induced light path modifications. GHGSat applies an inversion algorithm built on a simplified radiative transfer equation, incorporating high-resolution absorption line data from the HITRAN database and using vertical profiles based on the U.S. Standard Atmosphere, discretised into 100 equally spaced atmospheric layers for radiative transfer modelling [37].

GHGSat inversion models exclude thermal emission and molecular scattering effects in their atmospheric radiative transfer modelling, as these contributions are minimal within the instrument’s SWIR spectral bandpass. This simplified GHGSat inversion model is referred to as radiative transfer modelling (RTM). Data-driven match-filter technique [85,99,100] was utilised to convert the spectra directly into a plume-enhanced figure, and this technique is utilised for GaoFan-5B and PRISMA observations by Bai et al. [66], Han et al. [83], He et al. [82], and Roger et al. [101]. The matched-filter retrieval approach is a data-driven technique that offers various key benefits over traditional full-physics and simplified radiative models. One of its primary strengths is its robustness against radiometric and spectral distortions, such as vertical striping caused by detector inconsistencies, which often affect satellite observations. By operating on a pixel- or column-level basis, matched-filter methods effectively manage these systematic errors without the need for complex correction schemes [85]. Moreover, they facilitate the direct retrieval of methane concentration anomalies (ΔXCH₄), whereas full-physics approaches typically require additional steps, including background estimation, to derive similar outputs. Matched-filter retrievals are also significantly more computationally efficient, making them well-suited for large-scale processing. In contrast, full-physics models, while offering detailed atmospheric characterisation, are computationally demanding and more susceptible to uncertainties in model inputs. However, matched-filter approaches face limitations due to their reliance on basic linear signal frameworks, which fail to encapsulate the intricate physical mechanisms involved in detecting gases. Consequently, they can lead to substantial retrieval inaccuracies and tend to produce numerous false positives [85].

4.2. Wind Observations

Wind is one of the key parameters for plume detection and emission rate estimation. Equation (1) displays a simple mass balance approach for the estimation of the source rate or emission rate of methane for a satellite observation:

$${Flux rate}_{{CH}_4}= {\int }_{-y}^{+y}{\int }_{-x}^{+x}\mathsf{\Delta }\mathrm{X}{CH}_4\times U_{\mathrm{e}\mathrm{f}\mathrm{f}} dx dy$$

(1)

Here, ΔXCH₄ is the enhancement in dry-air column methane (or the anomaly) calculated by subtracting the background or upwind dry-air column methane concentration from the concentration in the downwind pixel, U_eff is the perpendicular effective wind speed, and x and y are the dimensions of the plumes. For oil and gas sources, plume lifetime is typically around 5 min, whereas for coal mines, plumes can persist for up to 1 h [61]. Therefore, effective wind speed is calculated by averaging wind data over a specific time depending on the type of source, plume size, and duration [61]. Most studies employing the mass balance approach for emission rate estimation determine effective wind based on the methodology outlined by Varon et al. [61]. It is also worth noting that the altitude of the wind is a key parameter. Varon et al. [61] calculated the effective wind by averaging the wind at 10 m. Wind information is also helpful to validate the plume by comparing the plume alignment with the available wind direction. For global wind data, satellite observations depend on the global meteorological reanalysis products. NASA Goddard Earth Observing System-Fast Processing (GEOS-FP) reanalysis and European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5), which is the fifth generation ECMWF reanalysis product [102], are the two main reanalysis products that were used for emission estimations.

To obtain effective wind in the mass balance equation, 10 m wind was most used by GOES-FP in 6 studies [61,66,68,83,85,101]. However, ERA5 wind at 10 m, 100 m, 300 m, 1000 m, and 1500 m was used for emission rate estimates by Schuit et al. [68], Hu et al. [71], Tu et al. [75,78], Sadavarte et al. [79], and He et al. [82]. The high-resolution satellite observations, Varon et al. [61], Bai et al. [66], Schuit et al. [68], Han et al. [83], He et al. [82], Roger et al. [101], Guanter et al. [85] and Ayasse et al. [103] used 10 m winds for effective wind analysis and flux estimates.

Palmer et al. [73] used 10 m wind while applying a simple mass balance approach for emission quantification for the Australian coal mine region and did not consider the variable elevation of the coal mine region and the wide swath of TROPOMI. However, others consider these limitations while carrying out the effective wind and use 100 m to 1500 m winds. Tu et al. [78] used 330 m winds from ERA5 while quantifying emissions in the Upper Silesian Coal Basin (USCB) region in Poland, which has a mean altitude of 300 m. They found that flux estimation with 10 m and 500 m wind caused variation by −25% and 13% in emission rate, as 10 m wind was 20% lower and 500 m wind was 32% higher in comparison to the 330 m wind [78]. Hu et al. [71] used wind at 850 hPa, considering that Shanxi, China, has many mountains and only 16% of its land surface is below 1500 m, while 17% exceeds this elevation. Tu et al. [75] used 100 m wind for source rate/emission rate estimations for CMM observation in China. Similarly, in Australia, Sadavarte et al. [79] used boundary layer winds (~ 1000 m) for TROPOMI observations. Peng et al. [76] used a unique approach by incorporation the GFS and GDAS meteorological data. Therefore, for high-resolution satellite observations that primarily capture localised and high emission rates, using 10 m wind data helps reduce wind-related uncertainties. In contrast, for area-integrated mapping approaches, winds at higher altitudes provide more accurate flux estimations.

4.3. Plume Detection

Emission estimation of any methane source first requires careful plume detection to minimise false positives and estimation errors. The methane plume detection starts with the methane column enhancement by subtracting the upwind methane mass or the background mass from the observed scene, or directly calculating the methane enhancement through the match-filter technique. This way, the methane plume images are obtained. However, there are various false positives [61], so further inspection is required, and for this purpose, manual checking is mostly preferred. Varon et al. [61], while studying the CMM, showed the false positives over the San Juan mine in New Mexico. Similarly, other studies also used the manual plume selection method to keep any observation error as low as possible [66,71,79,82,83,85,101]. Another technique for plume detection is the cone-plume model (CPM) to assess methane dispersion from localized sources. Tu et al. [78] developed this model for coal mine plumes in Poland, and later Tu et al. [75] used similar approaches in Shanxi, China. Peng et al. [76] used the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model for the plume shape and dispersion for coal mine emission estimation in China. While CPM is based on observational data, the HYSPLIT model is a dispersion model that simulates plume behaviour assuming steady-state emissions. All these methods need manual verification and have a significant number of human resources for plume modelling. In the case of a coal basin with multiple coal mines emitting daily plumes, manual detection demands significant human effort. To address this, Schuit et al. [68] developed an automated methane plume detection and monitoring system using a two-step machine-learning approach. This method employs a convolutional neural network (CNN) to identify plume-like structures in methane data, followed by a support vector classifier to differentiate actual emission plumes from retrieval artifacts. Schuit et al. [68] reported the detection of 581 plumes over coal mine facilities using this approach.

4.4. Flux Observation and Estimation Techniques

The final and key step in studying any methane source is the application of the mass balance equation for emission rate or flux quantification. Different approaches have been employed to estimate methane emissions, each with its own strengths and limitations. These methods range from simple mass balance techniques to complex inverse modelling frameworks [42]. Comparing various source estimation processes helps in understanding the reliability of emission estimates, the impact of observational constraints, and the uncertainties associated with different methodologies. Integrated mass enhancement (IME) is the most used method for the emission rate quantification for coal mine emission observations, and seven studies were conducted based on the IME method [61,66,68,82,83,85,101]. Only two studies were conducted with the cross-sectional flux (CSF) method [61,79] and the wind-assign anomaly (WAS) method [75,78]. Two were studied using a model-free mass balance approach [71,73], and one used the HYSPLIT model approach [76]. For GOSAT observations, NAME Model and GOES-Chem with Bayesian outcome were utilised to obtain the emission rate estimates [89,90]. To further smooth out most of the observation artifacts, wind rotation plume and time averaging methods were also employed by Varon et al. [61] and Sadavarte et al. [79]. Varon et al. [61] carried out emission rate quantifications using both IME and CSF methods for the same mines and provided an opportunity to compare the two methods to discuss various uncertainties and limitations for coal mine emissions. The IME and CSF methods show notable differences in their measured values across, while estimating three coal mine emissions by Varion et al. [61], and the IME method tends to produce higher estimates in certain cases, while the CSF method remains relatively stable considering the difference in the estimation approaches. These differences arise because the IME method is more sensitive to background concentration uncertainties and plume dispersion assumptions, while the CSF method directly integrates flux across the plume’s cross-sectional area, making it less sensitive to total plume extent but potentially affected by alignment with wind direction.

4.5. Emerging Technologies

Quantifying methane emissions remains a challenging task, requiring precise background information, continuous satellite observation, and plume detections. The limitation of the spatial and temporal coverage can be overcome with the help of multi-satellite observations [66,68].

Most previous studies have relied on manual plume detection, which demands significant human effort. However, automated detection can significantly enhance quantification. Schuit et al. [68] introduced a two-step machine-learning algorithm using a convolutional neural network (CNN) to autodetect methane plumes and distinguish them from artifacts. Their model, trained on TROPOMI data (pre-2021), detected 2974 plumes, of which 20% (581) were associated with coal mines. Between 2018 and 2021, they trained the model on 828 plumes, identifying 195 plumes over mining regions. These plumes were detected across China (269), Poland (7), South Africa (50), Russia (64), Australia (46), India (55), and the USA (52). To classify the sources of methane plumes, Schuit et al. [28] utilized three high-spatial-resolution satellite instruments: GHGSat, PRISMA, and Sentinel-2. Their study estimated that coal mine emissions contribute approximately 4.7% (2.1 Tg yr⁻¹) of the total global methane flux. Using the IME method, they reported a mean emission rate of 44 t/h, with individual detections ranging from 8 to 122 t/h. However, because TROPOMI has a relatively large pixel size, a single detected plume often includes emissions from multiple coal mines. In fact, high-resolution GHGSat observations showed that up to ten distinct coal mine plumes could be encompassed within a single TROPOMI detection.

Coal mine emissions also exhibit strong spatiotemporal variability [71]. Satellite sensors capture emissions at specific overpass times, and extrapolation is often required to estimate long-term fluxes. Such estimates are typically compared with bottom-up inventories and IPCC Tier 1 and Tier 2 quantifications, which cannot fully represent the short-term variability of mine emissions [66,71,73,75,79]. Quantification uncertainties can be further reduced by incorporating real-time ground-based observations. Two studies were conducted in the available literature. One study was conducted in Shanxi, China, where CH₄ fluxes from ventilation air methane (VAM) were measured using the eddy-covariance method over a 12-day period [71], and the other study was conducted in Virginia, USA, where VAM was observed at two ventilation shafts and compared with PRISMA observations.

5. Methane Monitoring Across Different Coal Basins and Related Uncertainties

Various satellite observations have been utilised for CMM monitoring and quantification across the globe. Shanxi, China, has been the most studied region due to the large number of coal mines and China’s significant contribution to CMM emissions [82]. Australia, Poland, the USA, Russia, and South Africa are other major regions that have been the focus of satellite-based CMM emission observations. Figure 4 shows that from satellite observation to quantification of methane has a number of variable paths and hence may be the reason for the variability of this final quantification. Most studies have focused on a single country or basin; however, Varon et al. [61], Schuit et al. [68], and Roger et al. [101] examined coal mine emissions across multiple countries. In this section, we provide a comprehensive review of studies categorised by country, followed by a discussion on global-scale assessments of CMM emissions. The number of studies conducted in each country is shown in Figure 5.

5.1. China

China’s 12th Five-Year Plan set a target of utilising 8.4 billion cubic meters (5.6 Tg) of CMM by 2015 [89]. To achieve this, various policies were implemented along with incentives for CMM utilisation. However, between 2005 and 2012, CMM utilisation increased only from 0.6 to 2.3 Tg CH₄ (equivalent to 0.9–3.5 billion cubic meters) [104], falling significantly short of the 2015 goal. Miller et al. [89] conducted a total emission and trend analysis of methane from 2010 to 2015 using GOSAT data at a global scale, with a primary focus on China. The study aimed to assess the impact of policy implementation on methane emissions from coal mines in China. It also incorporated global model estimates and atmospheric inversions for source estimations. Miller et al. [89] reported an increasing methane trend in China of 1.1 ± 0.4 Tg CH₄ yr⁻¹, while for India, the increase was 0.7 ± 0.5 Tg CH₄ yr⁻¹ during 2010–2015. The methane emissions trend from the coal sector assessed by Miller et al. [89] showed a rising trend between 2010 and 2015, prompting a reassessment of both satellite observation data and CMM policies. Sheng et al. [90] repeated the methane trend analysis for China using an extended observational record and an updated modelling approach with GOSAT data, aiming to verify the findings of Miller et al. [89]. They used the UK Met Office’s NAME (Numerical Atmospheric Dispersion Modelling Environment), a Lagrangian particle dispersion model [105], to perform the inversions. Sheng et al. [90] reported a slow increasing trend of 0.36 ± 0.04 Tg CH₄/year during 2010–2017 and 0.5 ± 0.2 Tg CH₄ yr⁻¹ during 2010–2015, whereas Miller et al. [89] had reported 1.1 ± 0.4 Tg CH₄ yr⁻¹ for the same period. Sheng et al. [90] observed an increasing trend of 0.7 ± 0.3 Tg CH₄ yr⁻¹ from 2010 to 2012 due to CMM emissions, followed by a flattening trend with a smaller growth of 0.1 ± 0.06 Tg CH₄ yr⁻¹ from 2012 to 2017. Sheng et al. [90] also noted the coarser resolution of EDGAR v4.2, which attributed 85% of emissions to coal and mixed sources, whereas a higher-resolution national inventory by Sheng et al. [106] found that only 28% of emissions were dominated by coal and other sources.

Furthermore, using PRISMA images, Guanter et al. [85] quantified methane emissions in the Shanxi Basin, China, using an advanced matched-filter method. Four plumes were identified with emission rates (Q_c) of 5900 ± 2400 kg/h, 7700 ± 3100 kg/h, 8700 ± 3500 kg/h, and 9600 ± 3800 kg/h. No information was provided regarding the type of coal mine by Guanter et al. [85]; however, based on the plume size and emission rate, the emissions appear to originate from the ventilation shafts of underground mines. As per the analysis by Guanter et al. [85], PRISMA observations were unable to detect emissions below 500 kg/h. Additionally, the observations showed a strong dependency on surface type, requiring special care in plume selection.

Zhang et al. [91] extended GOSAT-based methane emission estimates by assimilating high-quality surface methane measurements with previously available satellite and ground-based data for 2010–2017. The study reported an annual mean total emission rate of 54 Tg CH₄ yr⁻¹ for China, with 50 Tg CH₄ yr⁻¹ attributed to anthropogenic sources, closely aligning with China’s official UNFCCC report (54 Tg CH₄ yr⁻¹ for 2014). Zhang et al. [91] estimated a linear trend of 0.73 Tg CH₄ yr⁻², with an ensemble range of 0.56–0.85 Tg CH₄ yr⁻², which was higher than Sheng et al.’s [90] estimate of 0.36 ± 0.04 Tg CH₄ yr⁻² but lower than Miller et al.’s [89] estimate of 1.1 ± 0.4 Tg CH₄ yr⁻². The study emphasized the importance of continuous observations, as satellite-only inversion estimated an annual mean methane emission of 59 Tg CH₄ yr⁻¹ with a positive trend of 0.16 Tg CH₄ yr⁻². The study also analysed spatial variability in CMM emissions across China, finding a positive trend in Shanxi, Inner Mongolia, and northern Shanxi, which coincided with increased coal production. Conversely, a negative trend was observed in Henan and southern Shanxi, where mine closures and coal-bed methane (CBM) capture and utilisation were more effective.

Peng et al. [76] estimated methane emissions in Shanxi, China, from 2019 to 2020 by assimilating TROPOMI column concentrations with plume simulations using the HYSPLIT model. Shanxi accounts for 15% of global coal production, with 239 mines producing more than 0.5 million tonnes of coal annually in 2019 [76]. A total of 112 images were processed, revealing seasonal variations, including a 14% (0.1 Tg CH₄ month⁻¹) reduction in emissions during the 2019 Spring Festival. The mean total flux observed was 8.55 ± 0.6 Tg CH₄ yr⁻¹ (8.5 ± 0.6 Tg CH₄ yr⁻¹ in 2019 and 8.6 ± 0.6 Tg CH₄ yr⁻¹ in 2020). TROPOMI-based emissions were higher than PKU-CH₄ v2 (5.8 ± 0.5 Tg CH₄ yr⁻¹) and GFEI v2 (7.3 ± 2.0 Tg CH₄ yr⁻¹) but closely aligned with EDGAR v6.0 (8.8 Tg CH₄ yr⁻¹).

Hu et al. [71] also estimated the methane emission in Shanxi, China, from 2018 to 2022 using the TROPOMI observation using mass balance approaches coupled with the high-frequency eddy-covariance flux observations. They reported a 5-year mean emission of 126 ± 58.8 µg m⁻² s⁻¹, which was slightly higher than EDGAR reported (120 µg m⁻² s⁻¹). Tu et al. [75] quantified methane emissions in Shanxi Province, China, from May 2018 to May 2023 using TROPOMI observations, wind anomaly methods, IPCC Tier 2 data, and comparisons with EDGAR v7.0 and CAMS-GLOB-ANT inventories. Contrary to previous estimates, they reported that Shanxi exceeded one billion tonnes of coal production in 2021, accounting for 12% of global output. The study identified 600 coal mines in Shanxi and grouped them into three regions—Yangquan, Changzhi, and Jincheng—reporting lower emissions than EDGAR v6.0.

The coal mines in the Shanxi Province of China are mostly underground mines. Therefore, the high-resolution satellite observations were used for the detection of the coal mine plumes by He et al. [82], Han et al. [83], and Bai et al. [66]. He et al. [82] used the spectral match filter for the retrieval of the ΔXCH₄. Using the GaoFen-5 data, they were able to detect a total of 93 plumes for a total of 32 methane sources. The emissions from these point sources showed a diverse range from a minimum of 761.78 ± 185.00 kg/h to a maximum of 12,729.12 ± 4658.13 kg/h and a mean emission rate of approximately 4040.30 kg/h. This study provided information regarding various challenges associated with satellite remote sensing in China.

Han et al. [83] conducted an extensive and detailed investigation of CMM sources across Shanxi Province, China, utilising the Advanced Hyperspectral Imager (AHSI) aboard GaoFen-5B between 2021 and 2023. They identified 138 intermittent emission events across 82 sites, collectively estimated to release 1.20 (+0.24/−0.20, 95% confidence interval) million tonnes of methane annually. The study also highlights discrepancies among EDGAR, GMM, and satellite observations, especially in northern China, due to the exclusion of CMM exploitation in recent years. It was further emphasised that high-resolution satellite observations were helpful in determining emissions from underground mines with highly concentrated emissions and supported the variable emission rate findings presented by He et al. [82]. Han et al. [83] also reported, for the first time, the heavy-tail characteristics of coal emissions, a feature previously observed only in oil and gas emissions. Varon et al. [61] used high-resolution satellite (GHGSat-D) observations to estimate CMM emissions from the Bulianta mine in China. Applying both IME and CSF inversion approaches, they calculated emissions of 2410 ± 1000 kg/h and 2450 ± 970 kg/h, respectively. In contrast, the Chinese State Administration of Coal Mine Safety [107] reported only 170 kg/h during a safety assessment, a value markedly lower than the satellite-derived estimates.

Furthermore, Bai et al. [66] conducted a detailed survey of CMM in China using multiple high-resolution satellites from 2019 to 2023. The study involved observations from seven high-resolution satellites, including GHGSat and six hyperspectral missions: GaoFen-5 01 and 02, Ziyuan-1 02D, PRISMA, EnMAP, and EMIT. They later interpolated these emissions for Shanxi, China, and compared them with datasets such as PKU-CH₄ v2, GFEI, EDGAR v8.0 (2019–2022), TI 2019, INVTRO (2019–2020), and GCMT alongside observational results of the 2019–2023 mean ΔCH₄ from TROPOMI. The combined spatial interpolation of GCMT and satellite observations aligned well with the hotspots detected using TROPOMI-based ΔCH₄ observations. However, eastern Shuozhou and central Datong were underestimated by GCMT and EDGAR v8.0 (2022), while TI 2019 showed underestimation in central Shuozhou and Xinzhou. When comparing the observations presented by Bai et al. [66] for 2019–2023 with those by Han et al. [83], it is worth noting that the latter either failed to detect emissions in eastern and central Shuozhou, central Datong, and central Xinzhou, or stricter emission regulations led to a reduction in CMM. Additionally, both studies observed a decline of 20 to 30 ppb in mean ΔXCH₄ using TROPOMI between 2021 and 2023, clearly indicating the impact of the “CBM extraction first, coal mining second” policy. Lu et al. [108] used high-resolution Ziyuan-02D observations to estimate the coal mine emission rates in China during July to December 2023. Three plumes at coal mines in China were reported with emission rates of 6029 ± 1033, 4298 ± 762, and 6146 ± 1126 kg/h. This study was conducted only to demonstrate the feasibility of Ziyuan-02D observations in Coal mines and other point sources. Li et al. [109] used the EMIT satellite observations to study the CMM emissions in China. The plume images and emission rate data were collected through the Carbon Mapper open-access portal. The study was conducted for a total of 88 plumes reported in China, and they estimated an average of 4.36 t/h emission rate from January 2023 to the end of July 2024. The study reported that the mean emission rate of 32 mines based on the EMIT observations (0.48 kg/GJ) was significantly higher than the emission intensities reported by EDGAR v8.0 (0.24 kg/GJ) and GFEI v2 (0.18 kg/GJ).

5.2. Australia

Varon et al. [61] quantified CMM emissions at the Appin mine, estimating an emission rate of 5850 ± 2360 kg/h using the IME method and 4980 ± 2100 kg/h using the CSF method, based on high-resolution satellite observations from GHGSat-D during the period from August 2016 to December 2018. Extrapolating the mean hourly emissions to annual emissions for the Appin coal mine resulted in an estimated annual emission of 51.19 ± 20.67 Gg/y and 43.62 ± 18.40 Gg/y based on IME and CSF methods. For Appin, Cardno [110] estimated emissions at ~5200 kg/h based on coal production activity, whereas Ong et al. [111] reported a higher flux of 10,800–12,600 kg/h based on ventilation flow rates. The emission rates derived from these two observational methods show a significant difference due to the difference in the basic assumptions and quantification mechanisms of both methods. Furthermore, Sadavarte et al. [79] estimated coal mine emissions in the Bowen Basin, Australia, using two years of TROPOMI observations. By analysing 125 clear-sky observations over three major sources, methane emissions were quantified using the cross-sectional flux method. The estimated annual methane emissions for the three mines, which included both surface and underground operations, were 230 ± 50 Gg y⁻¹, 190 ± 60 Gg y⁻¹, and 150 ± 63 Gg y⁻¹ for the period 2018–2019. Notably, the surface mine exhibited significantly higher emissions despite lower production levels (7.7 Mt in 2018–2019 and 5.8 Mt in 2019–2020). Palmer et al. [73] also reported the coal mine emission Bowen Basin, Australia, using the TROPOMI observations during 2019. Palmer et al. [73] estimated methane emissions of 3.1 ± 1.5 Mt CH₄ y⁻¹ and 3.3 ± 1.5 Mt CH₄ y⁻¹ for Capcoal (underground and surface) and Moranbah North/Broadmeadow (underground), respectively. For Coppabella (surface) and Hail Creek (surface), the estimated emissions were 0.9 ± 0.4 Mt CH₄ y⁻¹ and 1.2 ± 0.6 Mt CH₄ y⁻¹. Notably, the emissions reported for Hail Creek in this study were lower than those estimated by Sadavarte et al. [79], highlighting that satellite-based observations and emission quantifications can vary significantly depending on the methodology applied.

Schuit et al. [68] detected 46 coal mine-related plumes in Australia using the TROPOMI satellite. They were able to identify plumes with a minimum emission rate of 4000 ± 1000 kg/h and a maximum of 72,000 ± 28,000 kg/h, with the mean emission rate being 22,826.09 ± 9630.43 kg/h. This translates to a mean annual emission rate of 200.0 ± 84.6 Gg CH₄ y⁻¹. However, satellites such as TROPOMI, with their broad spatial coverage, are not able to distinguish point sources effectively and often capture the emissions from multiple sources simultaneously. This limitation reduces its capability in coal mine emission estimation and needs a more careful study.

5.3. Poland

The Upper Silesian Coal Basin (USCB) in Poland is one of the major coal-producing regions in Europe. Tu et al. [78] quantified methane emissions from the USCB between November 2017 and December 2020 using TROPOMI observations, and spatial variability was examined with methane data from the Copernicus Atmosphere Monitoring Service (CAMS), including both analysis and forecast data. The plumes were analysed with a self-developed simple cone-plume model to reduce the wind-related uncertainties. The estimated annual methane emissions from the USCB basin were 496 ± 17 kt CH₄ y⁻¹, which closely aligned with the European Pollutant Release and Transfer Register (E-PRTR) estimate of 448 kt CH₄ y⁻¹. The study further estimated the uncertainty introduced by wind conditions, which contributed to a 13% increase in emission rates. Tu et al. [78] further leveraged IASI observations to estimate tropospheric methane concentrations while analysing emissions over the USCB region in Poland. Their study calculated a tropospheric flux of 437 ± 27 kt CH₄ y⁻¹ using combined TROPOMI and IASI datasets. However, their findings highlighted increased uncertainty in TROPOMI + IASI inversions due to complexities associated with CH₄ vertical distribution. The reported methane estimates were 40% lower than those derived from the CAMS model and the CAMS-GLOB-ANT inventory, as the latter accounts for emissions from all sectors. The results of Tu et al. [78] aligned closely with the E-PRTR inventory estimate of 448 kt CH₄ y⁻¹ and showed reasonable agreement with the CoMet inventory (555 kt CH₄ y⁻¹). Additionally, they were comparable to prior assessments over the USCB region, which reported emission estimates varying between 9 and 79 kt/y for specific mining shafts [112] and up to 477 kt CH₄ y⁻¹ based on airborne measurements [113]. Furthermore, using a CNN-based approach, Schuit et al. [68] detected only seven plumes in Poland, all located in the USCB region. The observed emission rates ranged from a maximum of 41,000 ± 20,000 kg/h to a minimum of 10,000 ± 4000 kg/h. This corresponds to a mean annual emission rate of 254.04 ± 98.59 Gg/y.

5.4. Russia

Trenchev et al. [77] presented emission estimates for the Kemerovo region in Russia from May 2018 to December 2022 based on TROPOMI observations. This region contains a total of 86 coal mines spanning an area of 26,000 km². Before assessing the emission rate, the study conducted an error analysis and removed pixels that fell outside ±3σ around the median (µ), ultimately reporting a total of 339 emission events. They observed the periodic occurrence of high concentration clusters of methane over the coal mine areas. The study demonstrated spatial variations in emissions; however, the authors did not carry out flux emission calculations. They also highlighted the limitations of satellite observations, noting that the high density of mines in the area prevented satellites from resolving mine-specific emissions. Schuit et al. [68] estimated methane emissions in coal mining regions of Russia, detecting plumes (64 plumes were detected related to coal) with emission rates ranging from 0.2 ± 0.1 t/h to 2.4 ± 1.1 t/h using GHGSat observations, while underground coal mine vents were found to emit up to 8.8 t/h. They quantified a mean source rate of 40.4 ± 16.9 t/h using TROPOMI data. However, they also reported that a single TROPOMI-based target represents contributions from up to 10 different point sources that may cause overestimation compared to estimation from a higher-resolution GHGSat observation from a coal mine area. In Russia, two outlier emission rates were observed, while three plumes exhibited unusually high uncertainty based on the Schuit et al. [68] plume data.

5.5. South Africa

Sibiya et al. [74] conducted an analysis of the spatiotemporal variation of total column methane over Mpumalanga, a coal mining province, and the Eastern Cape using TROPOMI observations. The XCH₄ shows strong seasonal variation, with a declining trend from March to June and an increasing trend in the later months. During the period from 2019 to 2024, the mean concentration ranged between ~45 to 50 ppb. The study also discusses how a large coal mine in the Mpumalanga region acts as a significant source of methane emissions, which results in a higher mean XCH₄ in Mpumalanga compared to the Eastern Cape. However, the authors did not carry out any emission rate analysis of coal mine-related emissions. In contrast, Schuit et al. [68] detected 50 plumes associated with coal in South Africa. The mean emission rate in South Africa was 23.4 ± 9.7 t/h, based on CNN-driven plume information using TROPOMI data. The minimum and maximum emission rates observed were 7 ± 3 t/h and 59 ± 21 t/h, respectively, with a median of 21 ± 10 t/h during 2021. Only one plume was found to be outside the typical range, marking it as an outlier with an exceptionally high emission rate relative to the bulk of the observations.

5.6. USA

For the United States, Varon et al. [61] used high-resolution GHGSat-D satellite observations between August 2016 and December 2018 to quantify methane emissions from the San Juan coal mine. The IME method yielded an average flux of 2320 ± 1050 kg/h, while the CSF method produced a slightly higher value of 2390 ± 1070 kg/h, about 2–3% greater than the IME estimate. The reported 1σ uncertainties (40–45%) accounted for factors such as wind speed and direction errors, model uncertainties, retrieval noise, and source variability. Comparisons with previous studies showed that aerial measurements reported lower emissions, ranging from 360 to 2800 kg/h [114] and 1446 kg/h [115], whereas vent emission rate estimates suggested 2585 kg/h [116]. Recently, Karacan et al. [117] estimated ventilation air methane (VAM) emissions using PRISMA satellite data combined with ground-based observations in Virginia, USA, for the period 2020–2023. Their study compared bottom-up and top-down emission estimates using both GOES-FP and ERA-5 meteorological datasets. The results indicated overall lower emission estimates, largely influenced by surface heterogeneity and associated variability. Notably, ERA-5–based estimates were found to be significantly higher than those derived from GOES-FP inputs. Furthermore, comparison within situ gas sensor measurements at ventilation shafts revealed that Ventilation Shaft (VS)-12 exhibited an interquartile range of approximately 3800 kg h⁻¹, while VS-16 showed around 2000 kg h⁻¹. The implementation of VAMOX Regenerative Thermal Oxidizer (RTO) mitigation systems led to a marked reduction in emission rates; however, discrepancies remained between the top-down and bottom-up estimates.

5.7. Comparative Analysis of Remote Sensing Data

Satellite observations offer advantages in terms of spatial coverage; however, they face persistent challenges in retrieval accuracy and validation. Despite technological progress, only a limited number of studies have directly validated satellite-derived methane estimates using in situ measurements in coal mines. For example, Hu et al. [71] performed high-frequency methane monitoring near ventilation shafts in Shanxi, China, employing the eddy-covariance technique to quantify emissions. Their findings demonstrated the critical role of surface flux measurements in constraining and verifying satellite-derived CH₄ retrievals. On a five-year mean basis, TROPOMI-derived emissions (126 ± 58.8 kt yr⁻¹) were slightly higher than those reported by the EDGAR inventory (mean: 120 kt yr⁻¹; Q3: 167 kt yr⁻¹). While EDGAR suggested a steady year-on-year increase in CMM emissions, TROPOMI observations exhibited no clear temporal trend, underscoring the necessity of continuous surface observations to better capture temporal variability and validate satellite estimates. Karacan et al. [117] further advanced this comparison by analysing ground-based Ventilation Air Methane (VAM) measurements against PRISMA satellite data. Their results revealed that satellite-based estimates were lower than ground observations, with top-down inversions extremely sensitive to meteorological factors such as wind speed and direction.

6. Technical Challenges

Methane quantification involves three main steps: observation, plume detection or modelling, and quantification. These steps can further be categorised into six sub-components: type of satellite, satellite inversion method, wind model, wind altitude, plume-detection method, and quantification method (Figure 4). Each combination of these factors carries its own strengths, limitations, and validation requirements, ultimately influencing the accuracy of emission estimates.

Starting with satellite observations, most methane products are derived from the inversion of passive SWIR measurements, typically using the 1.6 µm absorption band of methane. However, these measurements are constrained by factors such as spectral bandwidth, signal-to-noise ratio (SNR), and atmospheric interferences (e.g., aerosols, clouds, and water vapor). Moreover, retrieval accuracy is often affected by surface reflectivity and the limited spatial resolution of current instruments, which can restrict their ability to detect small or diffuse emission sources. Satellite observations have inherent limitations in spatial and temporal coverage, as well as sensing capabilities. The blind experiments carried out by Sherwin et al. [118] highlighted the constraints of satellite observations, as identical plumes assessed by different teams showed considerable variation, with 55% of the average estimates lying within ±50% of the measured values from the control release. These challenges become even more pronounced when monitoring coal mine emissions, which are dispersed, continuous, and highly variable, making accurate quantification particularly complex.

The first-generation SWIR satellite GOSAT observes circular pixels of 10 km diameter spaced 260 km apart along the orbit track, resulting in sparse spatial coverage that requires model interpolation and introduces significant uncertainties. In recent years, higher-resolution satellite observations have become available, but various limitations persist. Using high-spatial-resolution observations of GaoFen-5B, methane plumes with emission rates reaching up to 0.116 Tg CH₄ y⁻¹ were observed in Shanxi, China [82]. However, these estimates could diverge by at least two orders of magnitude from those obtained using bottom-up and, indicating that such plume detections cannot be directly scaled to annual emissions [82]. TROPOMI, on the other hand, provides daily and broader spatial coverage but has a high detection limit of 10,000–25,000 kg/h, which can bias its flux estimates towards lower emission events. GaoFan-5B observation-based emissions estimates at the local scale, under idealised conditions, by Bai et al. [66] performed well at the small spatial scale while compared to EDGAR data but struggled at the regional scale. They also highlighted the issue of data availability in both time and space and hence raised concern towards the continuous monitoring and robust trend analysis. However, satellite observations capture only instantaneous plumes, whereas coal mine emissions are continuous and highly variable in space and time, especially in terms of surface mines. Consequently, extrapolating short-term plume detections to derive annual emission estimates introduces significant uncertainty and raises concerns about accuracy [66,75].

Methane retrievals require clear-sky conditions and low aerosol loads because clouds and aerosols strongly absorb and scatter SWIR radiation, reducing measurement accuracy. Therefore, any observation with cloud or high aerosols (Aerosols Optical Depth > 0.3) is often discarded. For example, between 2018 and 2019, TROPOMI captured only 124 clear-sky methane column observations out of roughly 500 measurements in Australia [79]. Therefore, even a dedicated satellite with a one-day revisit rate suffers from insufficient daily coverage for annual estimates, highlighting the essential role of continuous/more frequent ground observations in reducing extrapolation uncertainties. Further challenges in satellite inversions arise due to atmospheric aerosol concentrations. For TROPOMI observations, RemoTeC, a full-physics radiative transfer model (RTM), was used to convert satellite measurements into column concentrations. In contrast, GHGSat employs a simpler RTM and relies on a proxy method for methane inversions. While RemoTeC explicitly accounts for atmospheric aerosols, proxy-based methods do not, introducing additional uncertainties in the retrieved methane concentrations.

Surface albedo is also a crucial factor affecting the remote sensing of methane emissions. High albedo (>0.7) and low solar zenith angles (~0) can cause radiance levels to exceed satellite specifications, introducing biases [61]. Conversely, low-reflection surfaces (albedo <0.05), such as dark mine surfaces, pose challenges for detection. Coal mines often exhibit low albedo, reducing satellite sensitivity [61,82]. Additionally, surface heterogeneity complicates SWIR remote sensing, as various surfaces strongly absorb in SWIR bands, requiring careful corrections, particularly in high-resolution spectroscopy. False positives in satellite methane inversion observations were reported in GaoFen-5B data due to large solar panel arrays, greenhouses, buildings, water bodies, and moist cultivated lands [82]. Coal mines, particularly surface mines, feature highly variable albedo and topography, leading to observation artifacts due to strong SWIR absorption [61,82].

Wind plays a critical role in the quantification of methane emissions from plume detection to final emission estimates. Accurate plume detection is often complicated by retrieval artifacts such as stripping noise, surface reflectance variations, and stray light, which can be comparable in magnitude to the methane signal itself [61]. To address these challenges, observed plumes are reoriented to a common wind direction, minimising background noise, reducing wind-related errors, and improving the accuracy of time-averaged methane enhancements [61,82]. The plume-reorientation method has been proposed to reduce such observational artifacts [61,79]. The effectiveness of plume identification also relies on an optimal wind speed range; too low, and the plume may not develop sufficiently; too high, and the plume disperses, reducing detectability [82]. Additionally, complex topography further complicates retrievals. In China, wind at 10 m, 100 m, and 1500 m scales were included, which further creates inconsistency while comparing these results together and with other observational studies and inventories data. The source of wind data also proved to be a critical factor, as different models were employed to generate the wind information, and model winds show significant variation with respect to ground truth. For station-level winds varied significantly from 0.5 m/s to 8 m/s during GaoFen-5B overpasses, the ERA-5 model winds remained relatively stable [82]. Consequently, satellite observations indicated strong variability in the plume for the same point source, and the reported difference was 10,204.71 kg/h between the minimum and maximum [82]. This suggests that averaging the plume based on limited observations may lead to significant over- or underestimation.

Notably, limited studies were conducted for the satellite validation for coal mines, and only IPCC estimates were used. Although aerial and UAV-based surveys provide valuable spatial information, they cannot serve as definitive ground truth since their emission quantification also depends on dispersion modelling and meteorological assumptions. Controlled blind release experiments have also been instrumental in evaluating the performance and uncertainty of satellite retrievals [118]. However, the current experimental site does not fully capture the temporal dynamics and spatial heterogeneity of real-world coal mine emissions. They are typically more representative of localised or underground point sources rather than diffuse surface emissions.

Finally, a major challenge in satellite-based methane observations is the inability to measure emissions at night, as all the methane observation satellites are passive sensors that rely on solar scattered or reflected radiation. The geometry, topography, and geology of a mine strongly influence emission rates, while micro-meteorological conditions can cause significant variability from one source to another, so the extrapolation of emissions based on an instantaneous plume may lead to bias. Additionally, fixed satellite overpass times can introduce bias, as emissions captured at a specific time may not represent typical activity, potentially leading to overestimation or underestimation. These challenges highlight the need for a robust, continuous/more frequent methane observation methodology for coal mines and underscore the importance of integrating satellite data with ground-based measurements for reliable emission quantification.

7. Future Directions

Future research on CMM monitoring should aim to overcome the temporal, spatial, and methodological constraints inherent in current satellite observation systems. Satellite missions provide valuable large-scale observations but remain limited by fixed overpass times, inability to capture nighttime emissions, and data gaps caused by cloud cover or complex terrain. These temporal irregularities highlight the need for integrated, multi-platform observation frameworks that combine satellite data with continuous ground-based monitoring networks, UAVs, and aerial surveys [112,113]. Such integration can provide high-frequency datasets and could fill diurnal and seasonal observation gaps that satellites alone cannot resolve. Next-generation satellite missions such as MethaneSAT (launched in March 2024) [115], the Tanager satellite constellation [119], Sentinel-5 (launched in August 2025), and the upcoming MERLIN, TANGO, and CO2M missions promise significant advancements in spatial resolution, revisit frequency, and detection sensitivity, enabling more accurate and timely monitoring of atmospheric methane. Although MethaneSAT’s mission was interrupted in mid-2025, its design philosophy has paved the way for future instruments capable of regional-to-facility scale methane quantification. Tanager and similar high-resolution satellites are expected to enable multiple daily overpasses and enhance the detection.

To address wind-related uncertainties, one of the major sources of bias in top-down methane quantification, future efforts should emphasize improved in situ wind observations integrated with physical and machine-learning-based models [120] to enhance the accuracy and reliability of emission estimates. Machine-learning algorithms can also facilitate automated plume attribution [68], cloud screening, and bias correction [121], thereby improving the robustness of emission retrievals. These innovations, combined with coordinated ground validation efforts, can help create standardized reference datasets essential for inter-satellite calibration and uncertainty reduction. Data validation methods also require further upgrades to effectively support uncertainty reduction. Recently, the UN Environment Programme’s International Methane Emissions Observatory (IMEO), supported by Australia, has announced a research initiative to enhance the precision of methane emissions data from coal mines. This study aims to use a simulated open-cut coal mine to assess the performance of advanced measurement technologies, ranging from ground-based sensors to aircraft and satellites [122].

Integrating these multi-scale observations into national and industrial monitoring systems can bridge the current gap between research and real-world implementation, ensuring that satellite-derived insights translate into actionable mitigation strategies and compliance mechanisms. National Greenhouse and Energy Reporting (NGER) Australia also raised this issue and emphasized for more observation to enhance the accuracy of lower-order methods [123]. Ultimately, the path forward lies in synergizing space-based, airborne, and ground-based methane observations with AI-enabled analytics and policy-aligned data assimilation. This multi-platform, multi-disciplinary approach will enable more accurate, timely, and policy-relevant methane emission assessments, paving the way for improved monitoring, reporting, and mitigation of CMM emissions worldwide.

8. Conclusions

The rise in global methane abundance and its upward trend are well acknowledged by the scientific community. However, major challenges persist in accurately quantifying methane source and sink fluxes. Coal mines remain a significant contributor to both energy production and greenhouse gas (GHG) emissions. With the growing global demand for coal, CMM emissions have also increased. Effective CMM detection and mitigation depend on precise emission accounting, and for much of the past decade, satellite observations served as the primary tool while other technologies were still under development.

Over these past 10 years, various observation and quantification approaches have been developed, with satellite remote sensing. Satellites such as TROPOMI, PRISMA, GaoFen-5, GOSAT, and GHGSat-D are commonly used for global methane monitoring, and other satellites, IASI, EMIT, EnMAP, and Ziyuan-1 02D, have also contributed to methane emission quantification. The improvement in satellite resolution from GOSAT (10 × 10 km²) to GHGSat-D (25 × 25 m²) has led to reduced uncertainty in emission estimates. Flux estimation, a multi-step process involving plume detection, reorientation, and dispersion analysis, has advanced from manual detection towards automated approaches enabled by convolutional neural networks (CNNs). The minimum flux estimates are also improved from ~50 t/h of GOSAT to ~0.1 t/h for GHGSat-D. With the availability of higher-resolution satellite observations, the gap between bottom-up and top-down estimates is narrowing and will further improve with the most recent and upcoming satellite missions, such as Tanagar-1 by Carbon Mapper and MERLIN.

Wind conditions constitute a major source of uncertainty in emission estimates. Given the complex terrain and highly heterogeneous surface conditions of mines, the choice of wind data remains a critical factor. Even observations from the same satellite over the same location often yield markedly different emission estimates, largely due to variations in wind conditions. Differences in inversion methodologies further contribute to variability in emission estimates. While satellite sensors can detect concentrated plumes typically associated with underground mines, they often struggle to accurately capture the more diffuse and spatially dispersed emissions from surface operations. The concern of daily overpass at a fixed time further causes concern for the activity-induced over- or under-estimations and affects the extrapolated mean emission rate. Due to these limited, various global satellite studies have shown variable results.

China, the world’s largest producer of coal [124], owing to its extensive coal mining activities, has been a central focus of methane emission studies. Satellite platforms such as GOSAT, TROPOMI, GaoFen-5, GHGSat, PRISMA, and other hyperspectral missions have been employed to monitor CMM emissions across the region. The quantified emissions show both increasing [89] and decreasing [90] trends with satellite-based inversion. However, with high-resolution satellite observations, emission estimates for most undermine emissions were improved; however, emission estimates reported discrepancies with Global inventories [83]. Australia and Poland have also gained attention for their CMM emissions. The limited top-down estimates have reported differences between the measured and reported emissions using TROPOMI and GHGSat-D observations. High-resolution PRISMA satellite observations, on the other hand, were consistent with the European Pollutant Release and Transfer Register (E-PRTR) and other inventories [39]. The ground validation of the satellite data is available only for underground mines and at a very limited scale, and there is no such study for surface mines.

Nevertheless, satellite-based observations offer broad spatial coverage and unmatched capabilities to identify super-emitters. Given the existing limitations of current approaches, effective CMM mitigation requires the development of new observational strategies that integrate high-resolution satellite data with ground-based and aerial validation methods. Future efforts should focus on improving satellite retrieval algorithms, reducing wind-related uncertainties, and enhancing the synergy between multiple observational platforms. A comprehensive and multi-sensor approach will be essential for achieving more accurate methane emission assessments and implementing effective mitigation strategies.