Long-term and interannual variations of atmospheric methane observed by the NIES and collaborative observation networks

Effective action for climate change mitigation requires an accurate understanding of global greenhouse gas budgets, including those of methane (CH₄). Atmospheric measurement data provide key constraints for estimating the magnitudes and distributions of sources and sinks and are utilized in atmospheric chemistry transport modeling studies. Long-term atmospheric measurement networks have revealed decadal, interannual, and seasonal variations in atmospheric CH₄. In 2020, a record-breaking annual CH₄ increase was recorded, but its cause is still unknown. This study analyzes atmospheric CH₄ variations using data from the National Institute for Environmental Studies (NIES) and its collaborative observation networks. Datasets from ground, mobile, and satellite platforms, employing diverse measurement techniques, confirmed past episodes, recent remarkable increases, and spatial distributions of atmospheric CH₄. Our data clearly showed a sustained CH₄ increase from 2020 to 2022, with the highest annual increase in 2021. The atmospheric CH₄ increase was pronounced in the northern mid-to-high latitudes in 2020, but the enhancement shifted south in 2021 and 2022. This study demonstrates the capability of observational data from the NIES and collaborative networks in accurately characterizing spatiotemporal variations in atmospheric CH₄ regularly, supporting the improvement of our estimates of the global CH₄ budget.

Methane (CH₄) is an important greenhouse gas with emissions strongly linked to energy consumption, agriculture, waste management, and various natural sources (e.g., Bousquet et al. 2006; Dlugokencky et al. 2011; Saunois et al. 2020; Chandra et al. 2021, 2024; Ito et al. 2023). CH₄ is primarily destroyed by reaction with hydroxyl radicals (OH), and its atmospheric lifetime is approximately a decade (i.e., shorter than the residence time of CO₂ and many other greenhouse gases); therefore, the reduction of CH₄ emissions is considered an effective method to contribute to near-term mitigation of climate change to reach the goal of the Paris Agreement (Collins et al. 2018). The Global Methane Pledge also calls for further action to reduce anthropogenic emissions. Therefore, an accurate quantitative understanding of anthropogenic and natural emissions in response to human activity and climate variability is required. Of particular concern is the possible massive release of CH₄ from regions in the northern high latitudes, which are prone to global warming (Yokohata et al. 2020). The global budget of CH₄ is periodically estimated by diverse studies, such as surface and satellite observations, atmospheric chemistry transport models, and emission inventories (Saunois et al. 2020). Despite extensive efforts, the estimates of the magnitude and geographical distribution of CH₄ emissions from various sectors remain uncertain (Saunois et al. 2020; Chandra et al. 2021, 2024; Wang et al. 2022).

Atmospheric CH₄ abundance (mole fraction) is controlled by the balance of its sources and sinks. Since ground-based systematic measurements began in the 1980s, atmospheric CH₄ has shown distinct growth rate episodes. For example, the eruption of Mt. Pinatubo from 1991 to 1992 (Dlugokencky et al. 1996; Wang et al. 2004; Bândǎ et al. 2013; Chandra et al. 2021) and the increased biomass-burning emissions induced by the 1997–1998 El Niño (Dlugokencky et al. 2001; Bousquet et al. 2006; Chandra et al. 2021). The growth rate of atmospheric CH₄ plateaued from 1999 to 2006 but then began to increase (Rigby et al. 2008; Dlugokencky et al. 2009), followed by a further increase in 2014 (Nisbet et al. 2021); however, the driving factors of these variations are still debated (Schaefer et al. 2016; Rigby et al. 2017; Turner et al. 2017; Worden et al. 2017; Morimoto et al. 2017; Fujita et al. 2020; Chandra et al. 2021). Furthermore, annual anomalies and decadal emission changes have contributed to variations in the long-term trends of atmospheric CH₄ (Dlugokencky et al. 2003; Jackson et al. 2020; Chandra et al. 2021, 2024). As reported by global atmospheric measurement networks, such as those of the National Oceanic and Atmospheric Administration (NOAA) and Advanced Global Atmospheric Gases Experiment (AGAGE), atmospheric CH₄ exhibited a record-breaking growth rate globally for 2020, followed by a sustained increase in the following years (Lan et al. 2025). Recent studies have investigated the possible causes of this pattern from inverse analysis (the optimization of sources/sink magnitudes with atmospheric observations) (Qu et al. 2022; Peng et al. 2022; Feng et al. 2023; Drinkwater et al. 2023). In particular, the COVID-19 worldwide lockdown measures caused a reduction in OH due to the global reduction in NO_x emissions (Miyazaki et al. 2021; Laughner et al. 2021), and the reduced CH₄ removal by OH was likely responsible for a substantial proportion of the anomalously high growth rate in 2020 (Peng et al. 2022; Qu et al. 2022; Feng et al. 2023).

In this study, we describe atmospheric CH₄ measurement data acquired from the National Institute for Environmental Studies (NIES) and cooperative observation networks, including in situ/flask measurements and ground/satellite-based remote sensing observations. The present study combines and analyzes previously published datasets with newly evaluated and reported data. By demonstrating the applicability of our dataset for characterizing spatiotemporal variations in atmospheric CH₄ over large areas of the world, we aim to encourage readers to further utilize observational data toward a better understanding of atmospheric CH₄ variations and an improved global CH₄ budget.

The representative locations of the data collected in this study are shown in Fig. 1. Various CH₄ mole fractions and column-averaged dry-air CH₄ mole fraction (XCH₄) data were obtained based on the different platforms and methodologies. See Table 1 for a summary of the site/platform information for the datasets.

2.1 In situ and flask measurements

2.1.1 Measurement scale

The CH₄ mole fractions obtained from the in situ and flask measurements in this study were reported on the NIES 94 CH₄ scale, which was established based on a series of gravimetrically prepared standard gases (primary standards) (Sasakawa et al. 2025). Sample analysis was performed relative to on-site and in-house working standard gases for in situ and flask measurements. The working standard gases underwent a validation process of measurements against secondary standard gases, whose mole fractions were traceable to those of the primary standards, using a gas chromatograph equipped with a flame ionization detector (GC-FID) system (Nomura et al. 2021). The Round Robin Comparison Experiments 5 (2009–2012) and 6 (2014–2015) demonstrated that the NIES 94 CH₄ scale was consistently higher by 3.0–5.5 nmol mol⁻¹ (hereafter referred to as ppb) compared to the WMO scale (X2004A) for the mole fraction range of approximately 1730–1940 ppb (http://www.esrl.noaa.gov/gmd/ccgg/wmorr/, last access: 1 August 2024). We also participated in the ‘Sausage Flask’ Inter-comparison program for 2005–2021. In this program, sets of flasks were filled with dried natural air and sent worldwide for interlaboratory comparisons. The results indicated that the NIES 94 CH₄ scale was approximately 5 ppb higher than the WMO scale (Jordan et al. 2022), which was consistent with the results of the round-robin comparison experiments.

2.1.2 Ground-site measurements

Atmospheric CH₄ mole fractions at Hateruma (HAT, 24.0607°N, 123.8094°E) and Ochiishi (COI, 43.1603°N, 145.4974°E) monitoring stations were measured semi-continuously using GC-FID (Tohjima et al. 2002). Sample air drawn from an air intake at the top of the tower (37/47 m and 52/94 m above ground/sea level for HAT and COI, respectively) was introduced by a diaphragm pump into a drying unit, which consisted of a Nafion® dryer and a cold trap (− 40°C). Next, CH₄ in an aliquot of the dried air sample was sequentially determined using GC-FID at intervals of 10 min. The GC-FID measurements were calibrated against three standard gases with known CH₄ mole fractions every two hours. Therefore, the system repeated the two-hour cycle of measurements (three standard gases and nine air samples), continuously. The data were processed to produce hourly mean values, which were calculated from three to six sample air measurements, for analysis (Tohjima 2016a, 2016b). The analytical precision of the GC-FID was approximately 2 ppb.

Atmospheric measurements of CH₄ mole fractions were also taken at Mt. Happo Observatory (HPO, 36.6966°N, 137.7981°E, 1850 m.a.s.l.) located in the mountainous area near the Sea of Japan coast since July 2013 (for more logistical information, see Okamoto et al. 2018). The site is operated by the Ministry of the Environment of Japan, as part of Japan’s national ambient air pollution monitoring stations, contributing to the Acid Deposition Monitoring Network in East Asia (EANET; http://www.eanet.asia/) program. Sample air drawn from an air intake at the rooftop of the station (6 m above ground level) was introduced into a wavelength-scanned cavity ring-down spectroscopy (WS-CRDS) instrument (Picarro Inc., Santa Clara, CA, USA, model G2401), which simultaneously measures CO₂, CH₄ and H₂O, after passing through the sample-drying unit to prevent the water interference (Nara et al. 2014).

Atmospheric CH₄ mole fractions were observed from whole air samples collected in flasks from four sites in Asia: Nainital, India (NTL, 29.36°N, 79.46°E, 1940 m.a.s.l.), Comilla, Bangladesh (CLA, 22.43°N, 91.18°E, 30 m.a.s.l.); Danum Valley, Malaysia (DMV, 4.98°N, 117.84°E, 426 m.a.s.l.); and the top of Mt. Fuji, Japan (MFJ, 35.21°N, 138.43°E, 3776 m.a.s.l.). The CH₄ mole fractions in the air samples were analyzed using GC-FID, as described by Nomura et al. (2021). NTL is located in the Himalayan mountain range and is influenced mainly by easterlies passing through the Indo-Gangetic Plain during the monsoon period and by westerlies during other seasons. The air samples were collected at the Aryabhatta Research Institute of Observational Sciences in Nainital. CLA is surrounded by paddy fields and is influenced by human activities in the neighboring town. Air samples were collected by the University of Dhaka and Bangladesh Meteorological Department. At NTL and CLA, air samples were compressed into 1.5-L glass flasks at + 0.15 MPa pressure from ambient conditions once a week around 14:00 local time (LT). The sample air was dehumidified by passing through a cold trap (− 30 °C). Further details are described in Nomura et al. (2021), and the data are publicly available (Terao et al. 2022a, 2022b). DMV is surrounded by undisturbed tropical rainforest, and the NIES and Malaysian Meteorological Department have performed weekly flask sampling since 2010. Air samples from an intake on the top of 100 m tower drawn by a diaphragm pump were collected into 1.5-L glass flasks at a pressure of + 0.15 MPa after passing through a cold trap (− 20 °C) at 22:00 LT. At MFJ, sampling has been conducted monthly since 2017, in addition to continuous CO₂ measurements from 2009. The air samples at MFJ represent the free troposphere in the middle-latitude Asian region (Nomura et al. 2017) and were collected by a diaphragm pump into 3.3-L stainless-steel flasks at a pressure of + 0.2 MPa at 22:00 LT. As the MFJ observatory is only accessible in the summer, during the summer, we retrieved the flask samples from August of the previous year to July of the present year and replaced them with a new set of flasks for the next year. As MFJ was closed in the summer of 2020 owing to the COVID-19 outbreak, air sampling was intermittent from July 2020 to June 2021.

2.1.3 Tower and aircraft measurements in Siberia

The Japan-Russia Siberian Tall Tower Inland Observation Network (JR-STATION) comprises nine (currently six) tower sites in Siberia, where simultaneous multi-point semi-continuous observations of CH₄ have been made (Sasakawa et al. 2010, 2025). At each site, ambient air was sampled from inlets positioned at two distinct heights of the tower and dehumidified for subsequent analyses. The CH₄ mole fraction was measured using a modified SnO₂ semiconductor sensor based on a natural gas leak detector (Suto and Inoue 2010). Because of the switching of flow paths, data were available for three minutes per hour at each height. We studied CH₄ variations at four measurement sites in the network: Demyanskoe (DEM), Karasevoe (KRS), Berezorechka (BRZ), and Azovo (AZV), where semi-continuous CH₄ measurements have been maintained to a reasonable extent. BRZ is in the middle of a boreal forest (taiga). There is a small village with a population of dozens of people near the tower. However, there is no large-scale agriculture or industry in the vicinity. The closest large city is Tomsk (60 km northeast), with a population of approximately 0.5 million. KRS is located on the shore of a 5-km-diameter marshy lake in the middle of the taiga. DEM is located in the middle of the taiga and is surrounded by extensive wetlands. AZV is located adjacent to a small town in a steppe region. The closest large city is Omsk (30 km northeast), which has a population of approximately one million. These datasets are publicly available (Sasakawa et al. 2023a, 2023b, 2023c, 2023d).

We also present CH₄ variations from air samples collected using aircraft over two locations in Siberia. Within a 130-km radius from Surgut (SUR, 61°N, 73°E), air sampling was conducted approximately once a month (typically around midday or in the afternoon, LT) using a chartered Antonov An-24 aircraft at altitudes of 0.5–7.0 km, commencing in July 1993. During sampling, an air sample from an inlet positioned ahead of the engine exhaust was pressurized into a Pyrex glass flask using a diaphragm pump. The sampling flights were halted in 2017 but resumed in 2018 using a Cessna aircraft with lower sampling altitudes of 0.5–4 km. Unless otherwise noted, air sampling over Novosibirsk (described below) was performed similarly. In the pine forest region approximately 150 km southwest of Novosibirsk (NOV, 55°N, 83°E), air sampling was conducted approximately once a month by using an Antonov An-30 research plane operated by the Institute of Atmospheric Optics (Antokhin et al. 2012), which began in July 1997. A leased Tupolev Tu-134 aircraft was utilized from March 2011 to June 2017, switching to a leased Yakovlev Yak-40 aircraft in October 2017. All air samples were sent to the NIES for GC-FID analysis of CH₄. Further details are provided by Sasakawa et al. (2017, 2024b, Sasakawa and Machida 2024a). In addition, Umezawa et al. (2012a) presented isotope measurements of CH₄ from air samples over SUR.

2.1.4 Voluntary observation ship (VOS) measurements

The NIES Voluntary Observation Ships (VOS) program has been implemented to observe atmospheric greenhouse gases and related gases in the western Pacific since 1994, the North Pacific since 1995, and Southeast Asia since 2007 (Nara et al. 2011, 2014, 2017; Terao et al. 2011; Yamagishi et al. 2012; Hoshina et al. 2018; Tohjima et al. 2005, 2024). Thirteen commercial cargo vessels have been used in the program. In this study, we analyzed data obtained from the western and northern Pacific regions. The air samples were introduced from an air intake on the top deck of each ship (approximately 30 m above sea level), which was away from the smokestack at the stern, into the observation room of the cargo vessel. A metal bellows pump was used for flask sampling and a diaphragm pump was used for WS-CRDS.

The air sampling and subsequent analytical methods, as well as the historical use of VOS, for 1994–2010, are described by Terao et al. (2011). Since 2010, air sampling has been conducted onboard the TRANS FUTURE 5 of Toyofuji Shipping Co. Ltd. (TF5; November 2005 to the present) during north–south cruises in the western Pacific and onboard the PYXIS of Toyofuji Shipping Co., Ltd. (PX; November 2001 to April 2013) and NEW CENTURY 2 of Kagoshima Senpaku Kaisha, Ltd. (NC2; June 2014 to the present) during east–west cruises in the north Pacific.

Atmospheric CH₄ measurements have been taken onboard the TRANS FUTURE 5 using a WS-CRDS instrument (Picarro Inc., Santa Clara, CA, USA, model G1202) since September 2010. The instrument used is a prototype Picarro G1301, which simultaneously measures CO₂, CH_4, and H₂O. This study analyzed air samples introduced into the instrument after passing through the sample-drying unit to prevent water interference during CH₄ measurements. Exhaust-contaminated samples were excluded when the dry-air mole fractions of CO₂ and O₃ abruptly increased and decreased, respectively. Further details are provided by Nara et al. (2014).

2.1.5 Commercial aircraft measurements (CONTRAIL)

Comprehensive Observation Network for TRace gases by AIrLiner (CONTRAIL) is an ongoing program that measures atmospheric traces gases using commercial aircraft of Japan Airlines (JAL) (Machida et al. 2008; Matsueda et al. 2008). Air samples were collected in December 2005 using Automatic air Sampling Equipment (ASE) or Manual air Sampling Equipment (MSE) along international and domestic flight tracks (https://cger.nies.go.jp/contrail/ase_mse/; last access: 1 August 2024). Air samples were analyzed for CH₄ using a GC-FID system at NIES. Before the CONTRAIL program, the first phase of the JAL Airliner Observation Project was conducted. The ASE prototype collected air samples from the upper troposphere over the western Pacific between 1993 and 2005 for CO₂ and CH₄ measurements (Matsueda and Inoue 1996). CH₄ analysis was performed using a GC-FID system at the Meteorological Research Institute (MRI) on the MRI CH₄ scale (Tsuboi et al. 2016). This study analyzed CH₄ data obtained from the first phase of the JAL project (1993–2005) and current CONTRAIL program (after 2005). We applied the NIES-MRI scale difference of + 2 ppb based on inter-comparison exercises (Tsuboi et al. 2017) and analyzed the data using the NIES 94 CH₄ scale. The current CONTRAIL dataset (Machida et al. 2023) does not include CH₄ data from the first phase of the JAL project but this will be updated shortly. Past studies have analyzed CONTRAIL CH₄ data to study the upper troposphere over the Pacific region (Umezawa et al. 2012b; Schuck et al. 2012) and the upper troposphere/lowermost stratosphere over northern high latitudes (Sawa et al. 2015).

2.2 Total column measurements

2.2.1 Ground-based measurements

The Total Carbon Column Observing Network (TCCON) was established in 2004 to obtain precise and accurate measurements of column-averaged dry-air mole fractions of CO₂, CO, CH₄, and N₂O (denoted as Xgas, e.g., XCO₂). XCH₄ is the ratio of the total number of CH₄ molecules to the total number of dry-air molecules in the air column. TCCON stations operate a high-resolution Fourier transform spectrometer (FTS), the IFS 125 HR (or IFS 120/5 HR, upgraded from IFS 120 HR) manufactured by Bruker Optics, which records direct solar spectra in the near-infrared spectral region. For XCH₄, the achieved precision (1σ) is 3.5 ppb (Wunch et al. 2010). The total uncertainties (including accuracy) were determined to be below ~ 9 ppb for GGG2014 (Wunch et al. 2015) and ~ 7 ppb for GGG2020 (Laughner et al. 2024). As of January 2025, the global network consists of 28 operational sites (https://tccon-wiki.caltech.edu/Main/TCCONSites); however, TCCON stations are land-based and are mainly distributed across North America, Europe, and East Asia in the Northern Hemisphere. To fill observational gaps, the COllaborative Carbon Column Observing Network (COCCON) emerged in 2016, based on the low-resolution (0.5 cm⁻¹) EM27/SUN FTS, developed by the Karlsruhe Institute of Technology in cooperation with Bruker Optics (Gisi et al. 2012; Hase et al. 2016). EM27/SUN delivers a total uncertainty (precision and accuracy) comparable to that of TCCON, with values of ~ 4 ppb, assuming careful calibration of each spectrometer (Frey et al. 2015, 2019; Hedelius et al. 2016; Sha et al. 2020). In addition, portable EM27/SUN spectrometers have allowed their use in various field campaigns (e.g., Hase et al. 2015; Ohyama et al. 2023) as well as the deployment of customized versions on mobile platforms such as ships (Butz et al. 2022; Klappenbach et al. 2015; Knapp et al. 2021).

TCCON data are widely used for evaluating greenhouse gas (GHG) satellite data including the Greenhouse Gases Observing Satellite (GOSAT) and GOSAT-2 (e.g., Yoshida et al. 2013, 2023; Someya et al. 2023), the Orbiting Carbon Observatory-2 (OCO-2) and OCO-3 (O'Dell et al. 2018; Taylor et al. 2023), the Copernicus Sentinel-5 Precursor (S5P) (Sha et al. 2021), and the Chinese Carbon Dioxide Observation Satellite (TanSat) (Yang et al. 2020). Therefore, the TCCON and COCCON networks are currently the most important sources of reference data for ongoing satellite GHG missions.

This study presents XCH₄ results from three NIES-operated TCCON sites Rikubetsu (RJ, 43.46°N, 143.77°E, Hokkaido and Tsukuba (TK, 36.06°N, 140.12°E, Honshu), Japan, and Burgos (BU, 18.53°N, 120.65°E, Ilocos Norte), Philippines and one NIES-operated COCCON site: Tsukuba (TKB, 36.06°N, 140.12°E). Rikubetsu is a town with a population of approximately two thousand, located in a mountainous area covered with boreal forests in the eastern part of Hokkaido. It is approximately 200 km east of Sapporo, the prefectural capital of Hokkaido, Japan. The Rikubetsu TCCON FTS station (https://tccon-wiki.caltech.edu/Main/Rikubetsu, last access: 1 August 2024) is situated inside the Rikubetsu Space Earth Science Museum (Galaxy Forest Observatory) and is part of the Network for the Detection of Atmospheric Composition Change (NDACC) FTS stations (https://ndacc.larc.nasa.gov/stations/rikubetsu-japan, last access: 1 August 2024). Tsukuba is a city with a population of approximately two hundred and fifty thousand located in the southern part of Ibaraki Prefecture. It is located approximately 50 km northeast of Tokyo. The surrounding area is covered by a complex of rural and urban land use. The Tsukuba TCCON FTS station (https://tccon-wiki.caltech.edu/Main/Tsukuba, last access: 1 August 2024) located inside a NIES building, with the COCCON EM27/SUN FTS is also operated on the roof of the same building (https://evdc.esa.int/publications/coccon-version-1-dataset-from-atmospheric-observatory-of-tsukuba-available-at-the-evdc-data-handling-facilities-covering-start-date-jan-1st-2020-to-end-date-dec-25th-2020/, last access: 1 August 2024). The TCCON Tsukuba site is also part of NDACC (https://ndacc.larc.nasa.gov/stations/tsukuba-japan; last access: 1 August 2024). Burgos is a town with a population of approximately ten thousand, located in the northernmost part of Luzon Island of the Philippines (18.52°N, 120.65°E). It is approximately 55 km from Laoag City, the capital of the Ilocos Norte Province. Ilocos Norte is a "coal-free province," in which coal-fired power-generating facilities are banned (https://www.pressreader.com/philippines/manila-bulletin/20161001/281573765188700, last access: 1 August 2024). The TCCON FTS station (https://tccon-wiki.caltech.edu/Main/Burgos, last access: 1 August 2024) is situated in the substation area of the EDC Burgos Wind Project and is surrounded by wind turbines in protected natural vegetation (shrubs, trees, and grass). Local biomass burning occurs after the harvest and before the local monsoon season (late March to April).

2.2.2 Satellite-based measurements (GOSAT and GOSAT-2)

GOSAT and its successor GOSAT-2 are Japanese Earth observation satellites that monitor the global distribution of CO₂ and CH₄ from space and are promoted by the Ministry of the Environment Government of Japan, the Japan Aerospace Exploration Agency, and NIES. As of August 2024, GOSAT and GOSAT-2 have been in operation since their launches on January 23, 2009, and October 29, 2018, respectively. The data are available to the public on the NIES websites: the GOSAT Data Archive Service (https://data2.gosat.nies.go.jp/index_en.html, last access: July 14, 2024) and the GOSAT-2 Product Archive (https://prdct.gosat-2.nies.go.jp/, last access: July 14, 2024).

GOSAT is in a sun-synchronous sub-recurrent orbit at 666-km altitude with 3-day recurrence cycle and a descending node at approximately 12:48 LT (Kuze et al. 2009; Yokota et al. 2009). Its main sensor, the Thermal And Near-infrared Sensor for carbon Observation Fourier Transform Spectrometer (TANSO-FTS), observes the short-wavelength infrared (SWIR) spectrum reflected from the Earth’s surface. TANSO-FTS has a pointing mechanism to observe the off-nadir direction within the pointing range of ± 35 deg in the cross-track direction and ± 20 deg in the along-track direction. In addition, TANSO-FTS SWIR can change its gain to improve the dynamic range; a medium gain is used over bright surfaces such as the Sahara and central Australia, whereas a high gain is used elsewhere. TANSO-FTS acquires observational data every 4.6 s, with a footprint size of approximately 10.5 km in diameter at sea level.

XCH₄ data were retrieved from the SWIR spectra (Yoshida et al. 2011, 2013). Atmospheric light scattering is a major source of error in gas retrieval from space-based measurements of reflected sunlight. Light-path modification due to atmospheric light scattering shows large uncertainty because it strongly depends on the optical properties, vertical distributions of scattering particles, and surface albedo. Therefore, observational data with clouds or heavy aerosols within the instantaneous field-of-view of TANSO-FTS were excluded from the retrieval analysis. Based on comparison with the TCCON data (GGG2020, Laughner et al. 2024), retrieved XCH₄ had accuracies of − 0.14, 0.00, and 0.00 ppb for land with high gain, land with medium gain, and ocean with high gain, respectively, with precisions of 11.6, 18.4, and 11.9 ppb, respectively (NIES GOSAT Project 2023). Note that the differences in accuracy and precision with gain are due to the nonlinear response of TANSO-FTS (see Suto et al. 2013 for more details).

We used XCH₄ data from the NIES GOSAT FTS SWIR Level 2 product versions 02.95/02.96 for General Users to derive regional changes in observed XCH₄ from 2010 to 2022 for 16 land regions. The geographical regions for which the GOSAT data were analyzed were the same as those in Niwa et al. (2024) but the two regions of Southeast Asia (North and South) were merged to increase the amount of GOSAT data over the region. Data with a land fraction of 100% and high gain were used to analyze the annual increase. Given the relatively short GOSAT-2 observation period (five years), regional long-term trends were analyzed only for GOSAT L2 data.

The NIES GOSAT project estimates whole-atmosphere monthly mean CH₄ concentrations from XCH₄ data (GOSAT FTS SWIR Level 2 products) by filling temporal and spatial data gaps with simulated CH₄ concentrations (GOSAT Level 4 B product) and reports the global mean XCH₄ (see the NIES GOSAT project website (https://www.gosat.nies.go.jp/en/recent-global-ch4.html, last access: July 14, 2024) for more details. We used whole-atmosphere monthly mean XCH₄ based on SWIR Level 2 version 02.90/02.91, with a bias correction applied.

GOSAT-2 is also in a sun-synchronous sub-recurrent orbit but at 613-km altitude with a 6-day recurrence cycle and descending node at around 13:00 LT. The satellite carries TANSO-FTS-2 to monitor XCO₂ and XCH₄ with a nadir footprint diameter of about 9.7 km and has updated functions such as an intelligent pointing system to automatically point the cloud-free region to avoid clouds and a wider along-track pointing angle of ± 40 degrees (Suto et al. 2021; Yoshida et al. 2023).

The whole-atmosphere monthly mean CH₄ concentrations were estimated based on XCH₄ using the full-physics method in the GOSAT-2 TANSO-FTS-2 SWIR L2 Column-averaged Dry-air Mole Fraction Product in the NIES GOSAT-2 project. Bias correction was first applied to the original XCH₄ data, which were processed in the same manner as the GOSAT whole-atmosphere mean CH₄ concentrations. We used the GOSAT-2 whole-atmosphere monthly mean CH₄ concentrations based on version 02.00 of the SWIR L2 product (Yoshida and Oshio 2022).

2.3 Time-series analysis

This study was conducted through the collective efforts of contributing researchers, with different datasets processed for time-series analysis using the methods of Thoning et al. (1989) and Nakazawa et al. (1997). Although the two methods differ in functions and procedures of analysis (e.g., Yashiro et al. 2009; Pickers and Manning 2015), both are digital-filtering methods that compute a best-fit curve as a combination of a long-term trend, seasonal cycle, and short-term variations from discrete data which are not equally spaced in time. Each method was applied to two selected datasets (COI and HAT) for comparison. For both COI and HAT data, the trend patterns in annual averages and annual increases deduced from both methods agreed well, while the values from both methods showed some degree of variation from year to year (1996–2022); for annual averages, 89% of the values (24 of 27 years) agreed within ± 3 ppb for both COI and HAT; for annual increases, 70% and 74% of the values (19 and 20 of 27 years) for COI and HAT agreed within ± 3 ppb, respectively. Considering these potential differences, in the analyses below, we did not compute a value by combining the analytical values from different methods but rather characterized the overall site-to-site and year-to-year variations from our observation network (see Sect. 4). A more detailed comparison of the two methods is beyond the scope of this study. In the analyses below, the method of Thoning et al. (1989) was applied to the datasets from the NIES monitoring stations (COI and HAT), JR-STATION (DEM, KRS, BRZ and AZV), Siberian aircraft (SUR and NOV), CONTRAIL, and TCCON/COCCON, while the method of Nakazawa et al. (1997) was used for the datasets from Asian flask sampling (CLA, NTL, DMV and MFJ), ships (VOS), and GOSAT and GOSAT-2. The uncertainties of the trend curve were estimated as standard deviations from a bootstrap method in which 100 datasets were prepared by random resampling from the original data.

Accordingly, the original time-series datasets were analyzed to deduce the variation components at different timescales, including long-term trends and seasonal cycles. The characteristics of temporal variations in individual datasets from selected sites/platforms are described in Sect. 3. For a more integrated discussion (Sect. 4), we derived annual values from all available datasets. In this study, the annual average (ppb) refers to the value averaged over the target year based on the long-term trend, with the seasonal cycle removed. The growth rate (ppb yr⁻¹) refers to the instantaneous rate of increase and is calculated as a derivative of the long-term trend. The annual increase (ppb) indicates the amount of increase within the target year and was calculated as the difference in the values of the long-term trend on 1 January of the target year and the following year. Uncertainties of the annual averages and annual increases were estimated as standard deviations from the aforementioned bootstrap calculations. Note that the bootstrap uncertainty does not include the measurement uncertainty described above. The annual averages and annual increases are tabulated in Supplementary Table 1.

3.1 In situ and flask measurements

3.1.1 Ground sites

The time series of the monthly mean atmospheric CH₄ mole fractions observed at COI and HAT are shown in Fig. 2. In general, the CH₄ mole fraction at COI was larger than that at HAT, whereas the amplitude of the seasonal variation was much larger at HAT than at COI (Tohjima et al. 2002). The dominant origins of the air masses at HAT varied seasonally between summer maritime air from lower latitudes and winter continental air from higher latitudes. This clear seasonal regime in the air mass origins increased the magnitude of the seasonal variation at HAT. The long-term trend curves for both HAT and COI (Fig. 2) showed persistent increases, especially after 2006–2007. Figure 2b shows the growth rates calculated from the trend curves. Consistent with previous reports on global trends (e.g., Nisbet et al. 2019; Saunois et al. 2020), the growth rates at HAT and COI decreased gradually until 2006, reached almost zero in 2004–2005, and turned positive in 2006–2007. After 2007, the growth rate increased, with year-to-year variations. The recent annual increases for COI and HAT were 17.4 ± 1.4 and 11.3 ± 2.3 ppb for 2020, 16.9 ± 1.1 and 16.8 ± 2.5 ppb for 2021, and 11.2 ± 1.9 and 16.4 ± 2.2 ppb for 2022, respectively.

3.1.2 Towers and aircraft in Siberia

In this study, daytime averages from the JR-STATION tower data were analyzed. In general, the CH₄ mole fractions from the tower sites showed an increasing trend at all sites over the last 20 years (Fig. 3). At DEM, substantial CH₄ increases were observed in 2014–2015. At KRS, significant CH₄ increases were observed in 2014–2016 and 2020. In the BRZ, although the data gap was relatively large, the CH₄ increase was pronounced in 2020. At AZV, the CH₄ mole fraction showed the highest level in 2017–2018, followed by a slight decrease in 2019 and increase in 2020.

For the aircraft data, altitude averages observed only after 10:00 LT were analyzed. The CH₄ mole fraction over SUR at an altitude of 0.5 km showed large interannual variability before 2000 (Fig. 4a). Subsequently, an increase was evident in 2002–2003, followed by no apparent increase until 2011. From 2012 onwards, the CH₄ mole fraction showed a clear increase rate (~ 11 ppb yr⁻¹) pronounced in 2019. However, at the highest altitude of 7.0 km, only a slight increase was observed from 1998 to 2007 after a sharp rise in 1998. Then, the increase was again small until 2013 (~ 5 ppb yr⁻¹) but with a marked annual increase in 2008. Observations at altitudes above 4 km have been halted since 2015 (see Sect. 2.1.3). Over NOV, the CH₄ mole fraction at 0.5 km altitude showed large interannual variations from 1998 to 2006 (Fig. 4b). Significant annual increases were observed in 2007 and 2008, after which the CH₄ mole fraction was almost constant until 2011. The CH₄ data at 7.0 km showed no increasing trend from 1998 to 2005 but gradually increased (~ 7 ppb yr⁻¹) from 2006 to 2018. The CH₄ mole fraction at 7.0 km exhibited an apparent annual decrease in 2019, followed by a remarkable increase, with a large annual increase in 2020.

3.1.3 VOS

Figure 5 shows the time series of the CH₄ mole fraction, their long-term trends, and growth rates derived from the VOS flask measurements at 16 latitudes (from 50°N to 35°S) from October 1994 to March 2023. Agreeing with the results of Terao et al. (2011), the CH₄ mole fraction was larger in the northern high latitudes and decreased gradually to the south. The amplitude of the seasonal variation was largest at 20–30°N, plausibly due to the seasonal replacement of air masses of continental and oceanic origins, as observed at HAT (Sect. 3.1.1 and Tohjima et al. 2002). The long-term trend curves showed persistent increases at all latitudes between 50°N and 35°S. The growth rates showed an overall decrease from 1994 to 2006, except for a large increase from 1997 to 1998, and then increased from 2006 onwards with superimposed interannual variations. The recent annual increases differed by latitude and year. For 2020, the annual increase was the largest at 18.2 ± 2.0 ppb at 25°N but lower in the Southern Hemisphere. However, after 2021, we observed larger annual increases in the Southern Hemisphere than in the Northern Hemisphere. For example, the annual increases at 25°N, 10°N, 5°S, 20°S, and 35°S were 18.2 ± 2.6, 13.8 ± 1.2, 17.8 ± 1.2, 18.5 ± 0.5, and 16.5 ± 0.7 ppb in 2021 and 13.2 ± 3.1, 13.3 ± 1.4, 17.1 ± 1.2, 20.2 ± 0.9 and 18.3 ± 0.7 ppb in 2022, respectively.

3.1.4 CONTRAIL

The CH₄ mole fractions in the upper troposphere observed by the CONTRAIL commercial aircraft, including those in the first phase of the JAL project, are shown in Fig. 6. The plotted data were obtained during the cruising flights above 8 km altitude and grouped into latitude bands of 10° intervals (± 5°) without longitudinal data selection. Air samples with low nitrous oxide (N₂O) mole fractions were considered influenced by stratospheric air (Ishijima et al. 2010; Sawa et al. 2015; Umezawa et al. 2015) and thus excluded from the analysis. The data selection was made in the same manner as Umezawa et al. (2015), using the difference in the N₂O mole fraction from the trend curve observed at Mauna Loa (MLO), Hawaii (Hall et al. 2007). In this study, we considered the samples with > 2.6 ppb lower than the MLO trend as influenced by the stratospheric air. Note the lower threshold than that in Umezawa et al. (2015) (1.3 ppb below the MLO trend), because of the rather scattered nature of the N₂O data in the CONTRAIL observation. In Fig. 6, even in the upper troposphere, the CH₄ mole fraction increased in all latitudinal bands. The CH₄ mole fraction was larger in the northern middle- to high-latitude regions than in the Southern Hemisphere. Seasonal maxima in summer were observed from 30°N to 40°N. The high summertime CH₄ can be explained by air mass transport from the source regions of the Asian continent (Umezawa et al. 2012b, 2018). In contrast, the CH₄ mole fractions in the equatorial region showed minima in the boreal summer.

3.2 Total column measurements

3.2.1 TCCON/COCCON

Temporal variations in the daily averaged XCH₄ data retrieved from the three TCCON stations (Rikubetsu, Tsukuba, and Burgos) and the COCCON station in Tsukuba are presented in Fig. 7. The Japanese stations show similar seasonal variations, with minima from February to March and maxima between August and October. In contrast, the minima in Burgos shifted by three months toward summer (June), while the maxima occurred in August–October, in phase with those observed in Japan.

At Tsukuba, TCCON and COCCON showed comparable XCH₄ values; however, the annual average XCH₄ of the TCCON Rikubetsu station was 17.1 ± 2.9 (8.3 ± 3.4) ppb smaller than that at the TCCON (COCCON) Tsukuba station for 2014–2019 (2016–2019). From 2017 to 2019, Burgos exhibited similar annual average XCH₄ values than those at Tsukuba (TCCON and COCCON). A substantial increase in XCH₄ of 17.1 ± 1.7 ppb was observed in 2020 at Rikubetsu. In contrast, Tsukuba TCCON (COCCON) and Burgos experienced smaller increases of 12.7 ± 1.7 (10.4 ± 2.3) and 10.0 ± 0.8 ppb, respectively. In 2021, the trend reversed, with Burgos and Tsukuba TCCON (COCCON) showing significant increases of 18.0 ± 1.0 and 17.0 ± 1.6 (21.6 ± 1.5) ppb, respectively, compared to only 9.0 ± 1.6 ppb at Rikubetsu. The characteristics of larger annual increases in 2021 in the south were consistent with the annual increasing trends found in the VOS flask measurements (Sect. 3.1.3).

3.2.2 GOSAT

Figure 8 shows the time series of whole-atmosphere monthly mean XCH₄ obtained from GOSAT and GOSAT-2 observations and their fitting curves, trends, and growth rates. XCH₄ increased steadily during the analysis periods of April 2009–March 2023 (GOSAT) and August 2019–October 2023 (GOSAT-2), with clear seasonal variations. The growth rates of the GOSAT XCH₄ were mostly smaller than 10 ppb yr⁻¹ before 2020 (except for 2011) but showed a large increase of ~ 20 ppb yr⁻¹ in 2020. GOSAT-2 XCH₄ also indicated high growth rates of above 10 ppb yr⁻¹ for most of the observation period. The annual increase in GOSAT was the largest in 2021 (16.7 ± 0.6 ppb), followed by 2020 (15.6 ± 0.5 ppb). The annual averages of GOSAT and GOSAT-2 XCH₄ were in good agreement, with only a 0.3% difference. The annual increase in GOSAT-2 XCH₄ exhibited a trend similar to that of GOSAT, although GOSAT-2 data were available only after 2020.

Despite the good agreements of the annual averages and increases between GOSAT and GOSAT-2 data, there are some differences in their seasonal variations in the fitted curves and phases of the growth rates between the whole-atmosphere concentrations. One of the possible causes of the seasonal differences is relatively larger XCH₄ values over land in the Northern Hemisphere, particularly from February to June, in the original GOSAT-2 L2 product, compared to GOSAT XCH₄ values, which may be due to smaller aerosol optical thickness simultaneously retrieved from GOSAT-2 observations than that from GOSAT (Yoshida et al. 2023). This aerosol influence in GOSAT-2 XCH₄ data is to be reduced in the future version of the L2 product. It is also noted that the TCCON data versions used for data bias corrections in producing the whole-atmosphere concentrations are different, that is, GGG2014 and GGG2020 for GOSAT and GOSAT-2 whole-atmosphere concentrations, respectively, which might also cause the difference in the bias-corrected whole-atmosphere concentrations. The relatively shorter data record of 5 years for GOSAT-2 may also influence the trend calculation that uses a cutoff period of 24 months in the digital filter. The accumulation of GOSAT-2 data and steady efforts of updating retrieval algorithms and bias correction methods for both GOSAT and GOSAT-2 data processing will lead to more consistent global XCH₄ data sets from the satellite observations.

We also analyzed the variations in the daily mean XCH₄ in the GOSAT L2 full-physics product averaged over 16 subcontinental land regions (Fig. 9). The results are plotted in Fig. 10 for April 2009–March 2023. XCH₄ showed persistent increasing trends with regional differences in seasonal cycles. In general, the annual averages were larger over regions at northern middle latitudes (Regions 2, 6, 12, and 14) and low latitudes (Regions 3, 4, 7, 8, 13, and 15), whereas they were smaller over regions at northern high latitudes (Regions 1, 10, and 11) and in the Southern Hemisphere (Regions 5, 9, and 16). This contrasts with the latitudinal gradient observed at the surface background sites, where the CH₄ mole fraction was larger in the northern middle-to-high latitudes (see Sect. 4). XCH₄ was similar over lower latitudes and northern middle latitudes.

With the recent CH₄ increase, rapid growths of XCH₄ were observed in some regions. The growth rates in Boreal North America (Region 1) and Southeast Asia (Region 15) peaked from 2019 to 2020. Likewise, maximum growth rates were observed in 2020 in Tropical South America (region 4), Europe (region 6), Northern Africa (region 7), Central Africa (region 8), Southern Africa (region 9), South Asia (region 13), East Asia (region 14), and Oceania (region 16). In most regions, the growth rate remained large in 2021 (over 10 ppb yr⁻¹). Compared to the average annual increase of XCH₄ of 7.6 ± 0.7 ppb over all regions from 2010 to 2019, the annual increases in 2020, 2021, and 2022 were substantially large (15.6 ± 0.5, 16.7 ± 0.6, 13.0 ± 0.6 ppb, respectively). The annual increase in 2020 was the largest in South Asia (Region 13), at 23.4 ± 3.5 ppb, followed by Tropical South America (Region 4; 21.0 ± 2.9 ppb), Central Africa (Region 8; 19.2 ± 2.3 ppb), and Southeast Asia (Region 15; 17.8 ± 3.1 ppb). In 2021, the annual increases were 18.1 ± 1.2 ppb in Southern Africa (Region 9), 17.4 ± 1.0 ppb in Temperate South America (Region 5), 16.9 ± 0.6 ppb in Temperate North America (Region 2), and over 10 ppb in all the other regions.

The XCH₄ retrieval algorithms from GOSAT and GOSAT-2 observations use a priori CH₄ concentrations (Yoshida et al. 2013, 2023). To avoid a strong dependency of the retrieved XCH₄ on the a priori concentrations, the L2 products used in this study stored selected data that satisfied the Degree of Freedom for Signal (DFS) to be 1 or greater. The growth rates of the a priori concentrations that were produced by climatological or periodic CH₄ fluxes showed less variability through the period and smaller growth rates after the year 2020 than those from the retrieved XCH₄. These imply that the calculated growth rates of GOSAT XCH₄ in Fig. 10 are produced by information from the GOSAT observations. The retrieval processing of the GOSAT data was made under cloud-free conditions, which might affect the calculated growth rates over some regions in cloudy periods, such as tropical rainy seasons. To minimize regional and seasonal influence on the calculated growth rates, the 16 regions in this study are set to be wide at subcontinent scales, and the growth rates are calculated from trend lines that are smoothed by the 26th-order Butterworth filter with a cutoff period of 24 months.

Figure 11 presents the annual averages of CH₄ and XCH₄ from the datasets analyzed in this study. The earliest data began in 1994 (VOS) and the annual averages showed long-term CH₄ variations at various locations. The atmospheric CH₄ observed in our measurement networks generally increased over the entire period, which is consistent with the global annual averages reported by NOAA (gray dotted lines in panels b–d, Lan et al. 2025). Although the measurements were initiated later, the XCH₄ data from GOSAT and TCCON/COCCON showed a monotonically increasing trend after 2010 (Fig. 11a). Note that, in Fig. 11, the NOAA data are shifted by 5 ppb for comparison (see Sect. 2.1.1) and that the XCH₄ data are based on validation independent from the in situ and flask data (see Sect. 2.2).

The annual CH₄ averages from the ground sites exhibited large site-to-site variability (Fig. 11b). The largest CH₄ mole fraction was found at CLA (Bangladesh). As described in a previous study (Nomura et al. 2021), the site is surrounded by various CH₄ sources, including wetlands, rice paddies, and compressed natural gas stations. Such local and regional emissions would contribute to persistently large and highly variable CH₄ levels. Other sites with large CH₄ mole fractions were in Western Siberia (JR-STATION), where the tower sites (DEM, KRS, BRZ, and AZV) were in boreal forests with vast expanses of wetlands (Sasakawa et al. 2010). The CH₄ data from NTL (India) were as large as those from Western Siberia, implying a strong influence of CH₄ sources in the Indo-Gangetic Plain (Nomura et al. 2021). The other ground sites (COI, HAT, MFJ, and DMV) generally represented baseline CH₄ levels and delineated the latitudinal gradient in the Northern Hemisphere (e.g., Dlugokencky et al. 1994; Umezawa et al. 2012b; Chandra et al. 2021). It is noted that all these sites are in the Northern Hemisphere and their CH₄ mole fractions are larger than the global averages reported by NOAA (gray dotted lines in Fig. 11b).

The aircraft (CONTRAIL)- and ship (VOS)-based measurements collected data mainly in the western Pacific region (Fig. 1). The datasets covered the middle latitude to middle latitude of both hemispheres and exhibited clear latitudinal gradients, with larger CH₄ in the Northern Hemisphere than in the Southern Hemisphere (Fig. 11c and d). The NOAA global CH₄ averages closely align with the ~ 10°N data from both CONTRAIL and VOS.

Figure 11c and d also shows that the latitudinal gradient was larger in the lower troposphere (ship) than in the upper troposphere (aircraft). To illustrate the latitudinal gradient, the annual average CH₄ mole fractions are plotted in Fig. 12b for five selected years (2012–2016). As previously reported (Umezawa et al. 2012b), the CH₄ mole fraction was larger in the lower troposphere (VOS) than in the upper troposphere (CONTRAIL) at the northern middle latitudes, whereas the opposite was true at the southern middle latitudes. The CH₄ mole fractions from both platforms were in agreement in the tropics, indicating a small vertical gradient in the region. Such latitudinal and altitudinal gradients have been explained in terms of the interhemispheric air exchange through the upper troposphere (Nakazawa et al. 1991; Miyazaki et al. 2008; Sawa et al. 2012; Umezawa et al. 2012b; Saito et al. 2013; Bisht et al. 2021; Belikov et al. 2022). Figure 12b shows the unique characteristics of the individual sites. We employed the NOAA Marine Boundary Layer (MBL) reference to indicate the zonally averaged baseline CH₄ mole fractions (https://gml.noaa.gov/ccgg/mbl/; last access: 2 August 2024). The CH₄ mole fractions from the VOS were consistently aligned with those from the NOAA MBL reference, indicating that our western Pacific measurements approximate the zonal baseline values. The NIES monitoring stations (HAT and COI) also represent baseline air, whereas other surface sites signify different degrees of influence from regional CH₄ sources.

The latitudinal distributions of XCH₄ from GOSAT and TCCON/COCCON datasets are shown in Fig. 12a. Note that the selected five years (2017–2021) are different from those in Fig. 12b because of data availability. It is also noted that the latitudinal distributions for GOSAT XCH₄ were from the data over land (see Sect. 2.3), whereas those for VOS/CONTRAIL were from the data over the Pacific. XCH₄ exhibited latitudinal gradients with maxima at northern low latitudes and poleward decreases (Fig. 12a), clearly differing from those depicted from direct measurements of the CH₄ mole fraction in the lower-to-upper troposphere (Fig. 12b). This indicates that tropospheric CH₄ constitutes a limited weight of XCH₄ because XCH₄ varies with not only the tropospheric CH₄ mole fraction but also the tropopause height and vertical profile above (Saeki et al. 2013; Inoue et al. 2014; Müller et al. 2024). The latitudinal maxima in the northern tropics may be attributed to the vertical transport of surface CH₄ via deep convection over regions with high CH₄ emissions (Patra et al. 2009; Stanevich et al. 2021; Chandra et al. 2017). XCH₄ showed a north–south gradient of ~ 60 ppb between the middle latitudes of both hemispheres, which is approximately half of that observed at the surface (VOS).

To infer temporal changes in the latitudinal gradient, we calculated the differences in the CH₄ mole fraction at different latitudes in the Southern Hemisphere (inter-latitude difference). At respective latitude bands of the VOS data, as well as the NIES stations (COI and HAT), differences in the annual CH₄ average with respect to 20°S (VOS data) were calculated (Fig. 13a). As the CH₄ mole fraction was relatively uniform to the south of 10°S (Fig. 12b), the 20°S data were considered to represent CH₄ in the extratropical Southern Hemisphere. The VOS data at 40°N and 50°N were excluded due to sparseness of cruises during the analysis period. For comparison, those from the NOAA MBL reference were similarly calculated in the same manner for corresponding latitudes (dotted lines in Fig. 13a). As shown in Fig. 13a, the inter-latitudinal difference with respect to the Southern Hemisphere generally increased both in the Pacific region (VOS and COI/HAT) and zonally (NOAA MBL reference). This implies that the long-term global CH₄ trend is largely attributable to variations in the extratropical Northern Hemisphere. As the inter-latitude differences increased and decreased from the late 1990s and the early 2000s, we calculated the trends of the inter-latitude gradient for the period 2007–2022 (i.e., the phase of the steady CH₄ increase), as shown in Fig. 13b. The measurements in the Asia–Pacific region exhibited greater trends at northern middle latitudes than at low latitudes (shaded bars). In contrast, the NOAA MBL reference data exhibited greater trends at lower latitudes (open bars). This difference could be ascribed to differences in the spatial representativeness of the data, such as our measurements under the larger influence of the CH₄ emission increase in East Asia after the 2000s (Ito et al. 2019, 2023; Chandra et al. 2021, 2024; Niwa et al. 2024). It is also interesting to note the interannual variations in inter-latitude differences. There was a pronounced decrease in the inter-latitude difference between 1998 and 2002 for all observation latitudes (Fig. 13a). This can be explained by the decrease in anthropogenic emissions at the northern middle-to-high latitudes (Chandra et al. 2021).

Figure 14 depicts the annual increases in atmospheric CH₄ and XCH₄ calculated from the individual datasets for 2010–2022. To characterize the variations uniquely observed in our measurements, the results were compared to the globally averaged annual increase reported by NOAA (gray lines in all panels, Lan et al. 2025). First, the VOS-based annual increase (black lines in all panes), calculated as the average for all observation latitudes from the VOS datasets, showed interannual variations that were in good agreement with the global annual increase by NOAA, although the magnitudes were slightly different. For example, an increase in CH₄ occurred in 2014 both globally (gray) and in the Pacific region (black), with the increase being smaller in the latter region. Given the absence of strong CH₄ sources in the Pacific region, the primary cause of the 2014 increase likely occurred outside this region. In 2014, large annual increases were observed over Siberia (DEM, KRS, SUR, and NOV) (Fig. 14c and d), which may imply contributions from enhanced wetland emissions in the region. In this regard, an increasing trend in surface temperature anomalies was evident in summer (June-July–August), which was calculated from NCEP Reanalysis data (https://psl.noaa.gov/data/gridded, last access: 1 August 2024), whereas a terrestrial ecosystem model indicated no clear enhancement in emissions from Siberian wetlands (Ito 2021). The annual increases in COI and HAT exceeded the VOS-based value (black line) in 2014 (Fig. 14c), which was indicative of the larger increase occurring in the northern region.

Next, we addressed the annual increase observed after 2020. The annual increases calculated from the VOS data were 12.8 ± 1.6, 16.4 ± 1.4, and 15.8 ± 1.8 ppb in 2020, 2021, and 2022, respectively. The largest global annual increase in 2021 is similarly reported by NOAA (14.8 ± 0.4, 17.6 ± 0.4, and 13.3 ± 0.3 ppb, respectively; Lan et al. 2025). In 2020, larger annual increases in XCH₄ occurred at northern latitudes, being larger at Rikubetsu than at Tsukuba and Burgos (Fig. 14a); at JR-STATION sites and at COI than at other sites of the northern lower latitudes (Fig. 14c, except CLA with exceptionally large variation); and from 20 to 30°N for CONTRAIL (Fig. 14e). Such latitudinal trends in the annual increase in 2020 were not observed in the GOSAT XCH₄ data (Figs. 9 and 14b). In 2021, greater annual XCH₄ increases were observed at Tsukuba and Burgos than at Rikubetsu (Fig. 14a) and the annual CH₄ increase at COI was comparable to that at the lower latitude sites HAT and DMV. While the annual increases in VOS showed a small latitudinal gradient in 2020, larger values were observed in the south in 2021 and 2022 (Fig. 14f). These results suggest that, while the large CH₄ increase was sustained for the three years, the annual increases in the respective years could be ascribed to different causes, at least in their latitudinal distribution. However, a similar southward annual increase was not clear in the GOSAT XCH₄ data (Fig. 14b). The temporal and spatial patterns of XCH₄ are governed not only by variations in surface CH₄ fluxes, OH sinks through the troposphere, and atmospheric transport but also by the total chemical sink in the stratosphere and changes in tropopause height (e.g., Washenfelder et al. 2003; Saeki et al. 2013). In addition, the XCH₄ analyses in Figs. 9 and 14b are collective results at subcontinental scales that contain the information in the inland areas over the CH₄ sources. This may have caused the difference in the observed annual increases between CH₄ at the surface sites and GOSAT XCH₄. The TCCON sites of Rikubetsu, Tsukuba, and Burgos were validation sites for the GOSAT L2 product used in this study (NIES GOSAT project 2023). The observed TCCON XCH₄ and GOSAT XCH₄ correlated well, with small biases on smaller temporal and spatial scales for the GOSAT validation process than those for the analysis in this study, that is, the GOSAT data are selected within a ± 2° latitude–longitude box centered on each TCCON sites, and the TCCON data is averaged within ± 30 min of the GOSAT observation time.

Combining analyses of emission datasets, wetland emission models, and atmospheric inversions, Peng et al. (2022) attributed the 2020 CH₄ increase to decreased OH destruction and increased wetland emissions, with roughly half the contribution from each. They also reported that the annual increase in 2020 observed by the NOAA network was larger in the Northern Hemisphere than in the Southern Hemisphere, which is consistent with our measurements. Their analyses suggested the dominance of the northern extratropics in enhanced wetland emissions, that is, North America and Siberia. In contrast, two other inversion analysis studies suggested that the 2020 CH₄ increase was predominantly attributable to enhanced emissions in eastern Africa, although a change in OH also played a significant role (Qu et al. 2022; Feng et al. 2023). As indicated by Feng et al. (2023), the differences in the inversion analyses among the above studies may be related to the use of different datasets. Peng et al. (2022) incorporated surface in situ and flask CH₄ data into their inversion system, whereas Qu et al. (2022) and Feng et al. (2023) relied on the GOSAT XCH₄ data released from the University of Leicester (Parker and Boesch 2020). In general, surface measurement networks are sparse in tropical regions, while satellite data have deficits for high-latitude regions. Collective data analyses combined with an inversion system, as recently examined by Niwa et al. (2024), are required to clarify the possible data-driven biases in interpreting atmospheric CH₄ variations in terms of emission and sink changes. Incorporating data from different platforms, Niwa et al. (2024) suggested that emission increases in the tropics and northern low-latitude regions contributed to the 2020–2022 increase of atmospheric CH₄. They also highlighted the significant role of Asian sources with the strong constraints provided by the datasets presented in this study.

We presented atmospheric measurements of the mole fraction and column-averaged mole fraction of CH₄ from various platforms (ground, ship, aircraft, and satellite) of the NIES and collaborative long-term observation networks, including datasets previously reported and newly reported in this study. These datasets were comprehensively analyzed for the first time to characterize spatiotemporal variations in atmospheric CH₄ over large regions of the world. Our analysis confirmed that CH₄ underwent a pattern of slow growth in the early 2000s, regrowth in the late 2000s, and accelerated growth onwards, as reported by other global observation networks. The year-to-year variations were consistent between the CH₄ and column-averaged CH₄ datasets for the period after 2010, when GOSAT data became available. Our measurements onboard ships in the Pacific Ocean (VOS) and monitoring stations in Japan (COI and HAT) generally represented variations in baseline air at corresponding latitudes, whereas other ground and aircraft measurements exhibited unique characteristics, including regional to local source signals. A recent CH₄ increase from 2020 to 2022, with the largest annual increase in 2021, was well characterized by our observation data. Our data suggested larger annual CH₄ increases at northern middle-to-high latitudes in 2020, with more pronounced increases in the south in the following years. Monitoring these spatiotemporal variations in atmospheric CH₄ is crucial for understanding the variations in contributing sources and sinks. To support estimation of the global CH₄ budget, the developed datasets are available for scientific and collaborative studies.

The availability of the dataset presented in this study is shown in Table 1. The digital-filtering program developed by Nakazawa et al. (1997) is available on the PyPI website (https://pypi.org/project/n97fit/; last access: 2 August 2024). The curve-fitting program developed by Thoning et al. (1989) is available on the NOAA website (https://gml.noaa.gov/ccgg/mbl/crvfit/crvfit.html; last access: 2 August 2024).

AZV:

Azovo, Russia

BRZ:

Berezorechka, Russia

CLA:

Comilla, Bangladesh

COCCON:

COllaborative Carbon Column Observing Network

COI:

Ochiishi, Japan

DEM:

Demyanskoe, Russia

DMV:

Danum Valley, Malaysia

CONTRAIL:

Comprehensive Observation Network for Trace gases by AirLiner

FTS:

Fourier Transform Spectrometer

GOSAT:

Greenhouse gases Observing SATellite

HAT:

Hateruma, Japan

JR-STATION:

Japan-Russia Siberian Tall Tower Inland Observation Network

KRS:

Karasevoe, Russia

MFJ:

Mt. Fuji, Japan

MRI:

Meteorological Research Institute

NIES:

National Institute for Environmental Studies

NOAA:

National Oceanic and Atmospheric Administration

NOV:

Novosibirsk, Russia

NTL:

Nainital, India

SUR:

Surgut, Russia

TCCON:

Total Carbon Column Observing Network

VOS:

Voluntary Observation Ship

We appreciate the generous cooperation of Toyofuji Shipping Co. and Kagoshima Senpaku Co. in the NIES VOS program. We thank the captains and crews of the M/S PYXIS, M/S NEW CENTURY 2, and M/S TRANS FUTURE 5. We thank Shigeru Kariya and Tomoyasu Yamada of the Global Environmental Forum for their assistance with data collection for the VOS program. Flask air sampling at NTL was conducted by Manish Naja of the Aryabhatta Research Institute of Observational Sciences and that at CLA was conducted by Md. Kawser Ahmed of the University of Dhaka and those at the DMV were Ahmad Fairudz Jamaluddin of the Malaysian Meteorological Department. Calculations of some products and reference datasets of GOSAT and GOSAT-2 were performed using the NIES supercomputer system (NEC SX-Aurora TSUBASA and HPE Apollo 2000). We acknowledge the update of the digital-filtering program (Nakazawa et al. 1997) and its public release by Hisashi Yashiro (NIES). We are grateful to the NOAA for making their measurement data publicly available, which enabled comparisons with our data.

JR-STATION, CONTRAIL, VOS was financially supported by the Global Environmental Research Coordination System from the Ministry of the Environment of Japan (E0653, E0752, E1151, E1253, E1254, E1451, E1652, E1752, E1851, E1951, E2151, E2251, E2351, E2452). Aircraft observations over Siberia and onboard the VOS were supported by a fund for global environmental monitoring from the Center for Global Environmental Research (CGER), NIES. The flask-based observations at the Asian sites (NTL, CLA, and DMV), CONTRAIL, and the associated data analysis were supported by the Environment Research and Technology Development Fund of the Environmental Restoration and Conservation Agency (JPMEERF20142001, JPMEERF20152002, JPMEERF20172001, JPMEERF20182002, JPMEERF20182003, JPMEERF20222001, JPMEERF21S20800, JPMEERF21S20810, and JPMEERF24S12200). CONTRAIL was also supported by the GRENE Arctic Climate Change Research Project, the Arctic Challenge for Sustainability Project, and the Arctic Challenge for Sustainability II Project. The TCCON sites at Tsukuba, Rikubetsu, and Burgos, and the COCCON sites at Tsukuba were supported in part by the GOSAT Series project. The Burgos site was supported in part by the Energy Development Corp. Philippines. The GOSAT and GOSAT-2 missions were promoted by the Ministry of the Environment, Government of Japan, the Japan Aerospace Exploration Agency, and NIES.

Authors and Affiliations

1. National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba, Ibaraki, 305-8506, Japan

   Taku Umezawa, Yasunori Tohjima, Yukio Terao, Motoki Sasakawa, Astrid Müller, Tazu Saeki, Toshinobu Machida, Shin-Ichiro Nakaoka, Hideki Nara, Shohei Nomura, Masahide Nishihashi, Hitoshi Mukai, Matthias Max Frey, Isamu Morino, Hirofumi Ohyama, Yukio Yoshida, Jiye Zeng, Hibiki Noda, Makoto Saito, Tsuneo Matsunaga, Takafumi Sugita, Hiroshi Tanimoto, Yosuke Niwa, Akihiko Ito, Yousuke Yamashita & Tomoko Shirai

2. Meteorological Research Institute, Japan Meteorological Agency, 1-1 Nagamine, Tsukuba, Ibaraki, 305-0052, Japan

   Masahide Nishihashi, Kentaro Ishijima, Kazuhiro Tsuboi & Yousuke Sawa

3. Graduate School of Agricultural and Life Sciences, The University of Tokyo, Tokyo, 113-8657, Japan

   Akihiko Ito

4. Climate Research Division, Environment and Climate Change Canada, 4905 Dufferin Street, Toronto, ON, M3H 5T4, Canada

   Misa Ishizawa

5. Dokkyo University, Soka, 340-0042, Japan

   Hidekazu Matsueda

Contributions

TU designed the study, led the data analyses, and wrote the manuscript, with contributions from all authors. YTo conducted COI/HAT measurements and analyzed the data. YTe, SNo, MN, and HMu conducted air sampling from Asian sites onboard the VOS and analyzed the data. SNa, HNa, and HT conducted measurements onboard the VOS and analyzed the data. MSas and TMac maintained the measurement scale, conducted tower and aircraft measurements in Siberia, and analyzed the data. TSh and MI also analyzed the data from Siberia. TMac, YN, KI, KT, YS, and HMa conducted air sampling from the CONTRAIL aircraft and analyzed the data. IM, HO, AM, MMF, TSu, and HT conducted the TCCON/COCCON measurements and analyzed the data. TMat, IM, YYo, TSa, JZ, HNo, and MSai constructed GOSAT/GOSAT-2 data products and analyzed the data. AI provided insights from the emission data. YYa and TSh processed the data for an open database (Global Environmental Database). All the authors have read and approved the final version of the manuscript.

Corresponding author

Correspondence to Taku Umezawa.

Competing interests

The authors declare that they have no competing interests.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions