Spatial and Temporal Variations of Atmospheric CH₄ in Monsoon Asia Detected by Satellite Observations of GOSAT and TROPOMI

Abstract

Space-based measurements, such as the Greenhouse gases Observing SATellite (GOSAT) and the TROPOspheric Monitoring Instrument (TROPOMI) aboard the Sentinel-5 Precursor satellite, provide global observations of the column-averaged CH₄ concentration (XCH₄). Due to the irregular observations and data gaps in the retrievals, studies on the spatial and temporal variations of regional atmospheric CH₄ concentrations are limited. In this paper, we mapped XCH₄ data over monsoon Asia using GOSAT and TROPOMI observations from April 2009 to December 2021 and analyzed the spatial and temporal pattern of atmospheric CH₄ variations and emissions. The results show that atmospheric CH₄ concentrations over monsoon Asia have long-term increases with an annual growth rate of roughly 8.4 ppb. The spatial and temporal trends of XCH₄ data are significantly correlated with anthropogenic CH₄ emissions from the bottom-up emission inventory of EDGAR. The spatial pattern of gridded XCH₄ temporal variations in China presents a basically consistent distribution with the Heihe–Tengchong Line, which is mainly related to the difference in anthropogenic emissions in the eastern and western areas. Using the mapping of XCH₄ data from 2019 to 2021, this study further revealed the response of atmospheric CH₄ concentrations to anthropogenic emissions in different urban agglomerations. For the urban agglomerations, the triangle of Central China (TCC), the Chengdu–Chongqing City Group (CCG), and the Yangtze River Delta (YRD) show higher CH₄ concentrations and emissions than the Beijing–Tianjin–Hebei region and nearby areas (BTH). The results reveal the spatial and temporal distribution of CH4 concentrations and quantify the differences between urban agglomerations, which will support further studies on the drivers of methane emissions.

1. Introduction

Atmospheric CH₄ is the most important contributor to the anthropogenically enhanced greenhouse warming effect after CO₂ [1]. According to the IPCC Sixth Assessment Report (IPCC AR6), the global warming potential of CH₄ is 28 times greater than that of CO₂ in 100 years [2,3]. The oxidation of methane eventually leads to CO₂. As CH₄ decays, it also leads to the formation of tropospheric O₃ and stratospheric H₂O and changes in the concentrations of tropospheric hydroxyl, resulting in indirect radiative forcing [4]. The budget of CH₄ consists of a diverse mix of sources and sinks, with many sources overlapping spatially, so it is difficult to quantify the emissions [5]. Anthropogenic source emission types account for roughly 60% of atmospheric CH₄ emissions, such as agricultural activities, energy activities, and waste management [6,7,8]. Natural source emission types account for about 40%, including wetlands, oceans, termites, and clathrates [9]. In the background of tackling climate change and achieving a dual carbon target (aiming to have carbon dioxide emissions peak before 2030 and achieve carbon neutrality before 2060), monitoring regional CH₄ concentrations and their changes is important to help us better understand the driving factors of CH₄ variations and estimate CH₄ emissions.

Ground-based observations have become an effective method for revealing the relationship between greenhouse gases and anthropogenic emissions [10]. The long-term, high-quality surface observations of CH₄ and δ13C (CH₄) abundances by the WMO GAW community show that the global average atmospheric CH₄ has been increasing at an accelerating rate since 2007. Among them, the annual increases in 2020 and 2021 were the largest since the systematic record began in 1983 [11]. The main reason for the continuous increase in atmospheric CH₄ concentrations is the increase in anthropogenic emissions. Previous studies have suggested that this is likely to be caused by anthropogenic emissions from agricultural and fossil fuel sources in the tropics and the northern mid-latitudes and the decline in the atmospheric sink [3,12]. Although the atmospheric CH₄ concentration observed on the ground is highly accurate, it is difficult to identify and quantify the spatiotemporal changes in methane sources and sinks on a regional scale with sparsely distributed station observations.

Greenhouse gases Observing Satellite (GOSAT) and Copernicus Sentinel-5 Precursor (Sentinel-5P) observations provide important data sources for studying the spatiotemporal changes in the global atmospheric CH₄ concentration. GOSAT was successfully launched in January 2009 and is the first satellite dedicated to greenhouse gas observation. It was jointly developed by the Japan Aerospace Exploration Agency (JAXA), the National Institute of Environment (NIES), and the Ministry of Environment (MOE). Its main purpose is to obtain global atmospheric CO₂ and CH₄ concentration distribution information from space and estimate carbon sources and sinks on a continental scale, providing an important basis for formulating carbon emission reduction policies. The satellite observations of GOSAT cover the world within 3 days and have a spatial resolution of roughly 10.5 km at their nadir. The Copernicus Sentinel-5 Precursor mission consists of one satellite carrying the TROPOspheric Monitoring Instrument (TROPOMI) instrument launched by ESA on 13 October 2017. It can effectively observe trace gas components (such as CH₄, NO₂, CO, etc.) around the world on an urban scale. The width swath of TROPOM is approximately 2600 km on the ground, and the resolution is up to 7 km × 3.5 km, which greatly improves the accuracy of greenhouse gas observations with full daily global surface coverage. Previous studies have shown that these satellite observation data have the application potential of detecting emission sources and estimating anthropogenic emissions from regional or point sources [13,14,15]. However, due to the influence of cloud cover and sensor observation mode, the CH₄ concentration data observed by satellite remote sensing are irregularly distributed in time and space.

In order to obtain effective CH₄ concentration data with as much global spatiotemporal continuous distribution as possible, scholars use the kriging statistical method to interpolate the satellite observation data. Tomosada et al. [16] used the XCO₂ data product retrieved by GOSAT for kriging interpolation, calculated the variation function by separating the ocean and land, and considered the anisotropy of CO₂ concentration distribution to establish a semi-variation model. Bruhwiller et al. [6] used the general kriging method with a moving window to output the spatiotemporal continuous distribution data of global atmospheric CH₄ with a high temporal resolution. The observed mean mole fraction of CH₄ in the latitude zone was simulated, and the interannual variation of CH₄ emissions in the high latitude region (53°–90°N) was captured. Liu et al. [17] obtained the daily mean value of XCO₂ and XCH₄ in East Asia by using the L2 data products of XCO₂ and XCH₄ retrieved from the single-year GOSAT satellite observation, with the spatial kriging interpolation method. The overall result is consistent with the L3 product data released by the GOSAT satellite standard. With the accumulation of geographic spatiotemporal data and the development of the theory of spatiotemporal variation function, geostatistic methods have also been extended from space to the use of spatiotemporal interpolation. Liu and Li et al. [18,19] carried out global or regional XCH₄ mapping analysis using single satellite observation data based on spatiotemporal and geostatistical methods, and the results showed that the spatiotemporal distribution of regional CH₄ concentrations was correlated with the anthropogenic emission inventory, especially in eastern China, Bangladesh, the Indo-China Peninsula, and nearby regions. However, the observation time series and effective data points of a single satellite are limited in these studies. Using the fusion mapping results of multiple satellites can more effectively reveal the long-term dynamic changes in CH₄ concentrations.

The CH₄ concentration in the Asian monsoon region has continued to increase in recent years. However, due to the diversity and complex distribution of CH₄ emission sources in the Asian region, with a variety of climate types and land cover types, the response mechanism between the CH₄ emission and the space–time distribution of CH₄ concentrations in the region is still not clear [2], in which case, further analysis is needed. This study generates the regional long-term XCH₄ spatiotemporal continuous data set (Mapping-XCH₄) based on the spatiotemporal statistical method, which is able to utilize both time and space information simultaneously, using the XCH4 data observed by GOSAT and TROPOMI from 2009 to 2021. The spatiotemporal variation trend of XCH₄ in the Asian monsoon region is studied on different spatial scales. This paper aims to use XCH₄ spatiotemporal continuous data to detect and analyze the XCH₄ spatiotemporal changes in the Asian monsoon region at different spatial scales, reveal the application ability of satellite mapping data in the study of XCH₄ spatiotemporal characteristics, and provide effective support data for further research on the driving factors of atmospheric CH₄ concentration changes and carbon sources and sinks in the Asian monsoon region.

2. Materials and Methods

2.1. Materials

2.1.1. Satellite XCH₄ Observations

To analyze the spatial and temporal distribution characteristics of atmospheric CH₄ concentrations in the Asian monsoon region, we collected satellite-observed XCH₄ data from GOSAT and S5P spanning from April 2009 to December 2021.

Table 1. The products of CH₄ data from satellite observations.

Dataset	GOSAT-XCH4	TROPOMI-XCH4
Period	April 2009–December 2021.	November 2018–December 2021
Data Product	Bias-corrected FTS SWIR L2 CH4	S5P_OFFL_L2__CH4
Spatiotemporal Resolution	3 days, 10.5 km	1 day, ~7 km
Criteria of Data Screening	xCH4_quality_flag = 0, land_fraction > 90	qa_value > 0.75
Data Source	GDAS	Copernicus S5p Open Access Hub

gives a summary of the XCH₄ data products.

The XCH₄ data from GOSAT observations (GOSAT-XCH₄) are the bias-corrected FTS SWIR Level 2 CH₄ products (V02.95/V02.96) distributed by the National Institute for Environmental Studies (NIES), which are provided from the GOSAT Data Archive Service (GDAS) as monthly archive data (https://data2.gosat.nies.go.jp/GosatDataArchiveService/usr/download, (accessed on 1 March 2022)). GOSAT Project. Available online: www.gosat.nies.go.jp (accessed on 1 June 2020). According to the bias-correction processing results referred to in [20], the differences between the corrected SWIR L2 CH₄ product and the TCCON data from 2009 to 2018 are −0.140 ± 11.601 ppb for Gain H and 0.00 ± 18.428 ppb for Gain M over land.

The XCH₄ data observed by TROPOMI (TROPOMI-XCH₄) are from the Level 2 Offline (OFFL) timeliness data products that are available on the Copernicus S5p Open Access Hub (https://s5phub.copernicus.eu/dhus/#/home (accessed on 5 March 2022)). The detailed descriptions of CH₄ OFFL data versions, the respective dates, and the relevant improvements can be found in the Product Readme File (PRF) (https://sentinels.copernicus.eu/documents/247904/3541451/Sentinel-5P-Methane-Product-Readme-File.pdf/d7214038-25a9-416f-8deb-d5d6c766f92e?t=1658386700684 (accessed on 7 March 2022)). The XCH₄ retrieval is in good overall agreement with correlative ground-based measurements from TCCON and the Network for the Detection of Atmospheric Composition Change (NDACC) [21,22]. The systematic differences of the bias-corrected XCH₄ data with respect to TCCON data and NDACC data are, on average, −0.26 ± 0.56% and 0.57 ± 0.83%, respectively [23].

2.1.2. Ground-Based XCH₄ Data

Ground-based XCH₄ measurements of the Total Carbon Column Observing Network (TCCON) are used to validate XCH₄ data from Mapping-XCH₄ and individual satellite observations of GOSAT and S5P. TCCON is a network recording direct solar spectra in the NIR/SWIR spectral region using ground-based Fourier-transform spectrometers (FTSs). Accurate and precise column-averaged abundances of CH₄ retrieved from these high spectral (0.02 cm⁻¹) and temporal (~90 s) resolution spectra are the primary data for validating satellite XCH₄ data products.

There are four TCCON sites in the monsoon region of Asia, including Saga (Shiomi et al. [24]), Rikubetsu (Morino et al. [25]), Tsukuba (Morino et al. [26]), Xianghe (Yang et al. [27]), and Hefei (Liu et al. [28]). We used the GGG2020 version of the TCCON data products available from the TCCON data archive (https://tccondata.org/, accessed on 15 July 2022). In order to match the temporal scale of the Mapping-XCH₄ data, TCCON XCH₄ observations within 13:30 ± 2 h of the local standard time are selected to calculate the 3-day mean values.

2.1.3. CH₄ Emission Data from EDGAR

CH₄ emissions from the Emissions Database for Global Atmospheric Research (EDGAR) are used to compare with spatial XCH₄ patterns derived from satellite observations. EDGAR is a global greenhouse gas inventory developed by the European Commission’s Joint Research Centre (JRC) based on the bottom-up emissions calculation methodology using international activity data [1,29,30]. The global emissions are spatially allocated on 0.1° × 0.1° resolution grids using proxy data, such as point sources, population density, traffic networks, and nighttime lights [29,30,31]. The new version V7.0 of the EDGAR emission inventory provides global CH₄ estimates from 1970 to 2021, which are disaggregated to IPCC-relevant source-sector levels [32]. We used annual grid maps provided in netCDF format in which CH₄ emissions are expressed in kg substance/m²/s.

2.2. Methods

2.2.1. The XCH₄ Mapping Method Using GOSAT and TROPOMI Observations

We used the XCH₄ data observed from GOSAT and S5P to produce a spatiotemporally continuous mapping XCH₄ dataset (Mapping-XCH₄) in the monsoon region of Asia from April 2009 to December 2021. The dataset covers the Asian monsoon region that ranges from 10°N to 55°N and 61°E to 146°E, with a 1° grid interval in space and a 3-day interval in time. Figure 1 shows the flowchart of generation, which includes the steps of data preprocessing, spatiotemporal correlation analysis, and the mapping of gridded XCH₄ data.

(1)

Data preprocessing

To obtain valid XCH₄ data over land with good-quality observations, the XCH₄ data products were filtered using the criteria listed in Table 1. According to the geographic range of the study area, only XCH₄ data within the buffer zone from 5°N to 60°N and 55°E to 150°E were used in this paper.

As shown in Figure 2, these XCH₄ observations are irregularly distributed in space and time, with many gaps caused by impacts from geophysical factors such as aerosols and clouds. The spatial and temporal scales of GOSAT and TROPOMI observations are different, which results in the valid observations of GOSAT being much smaller than those of TROPOMI. Therefore, we converted the spatiotemporal scales of individual satellite XCH₄ data by calculating the average value XCH₄ within 3 days and 10 km. The equation is as follows:

$$\mathrm{X}\mathrm{C}{\mathrm{H}}_{4 \mathrm{u}\mathrm{n}\mathrm{i}}=\frac{{\sum }_{i=1}^n\mathrm{X}\mathrm{C}{\mathrm{H}}_{4 \mathrm{o}\mathrm{r}\mathrm{i}}}{\mathrm{N}}$$

(1)

where $\mathrm{X}\mathrm{C}{\mathrm{H}}_{4 \mathrm{o}\mathrm{r}\mathrm{i}}$ represents the XCH₄ data with original spatiotemporal units (10.5 km/3 days for GOSAT, 7 km/1 day for TROPOMI), and N is the amount of $\mathrm{X}\mathrm{C}{\mathrm{H}}_{4 \mathrm{o}\mathrm{r}\mathrm{i}}$ data within 3 days and a radius of 5 km. $\mathrm{X}\mathrm{C}{\mathrm{H}}_{4 \mathrm{u}\mathrm{n}\mathrm{i}}$ represents the XCH₄ data with the same space–time units (10 km/3 days).

This processing of converting spatiotemporal scales addresses the problem that GOSAT and TROPOMI observations have different sampling times at the same location while reducing the discrepancy between the XCH₄ data. Figure 3 is the scatter plot of GOSAT-XCH₄ and TROPOMI-XCH₄ during the overlap period of 2019 to 2021. Comparing the XCH₄ data at the original spatiotemporal scales, the overall difference between the GOSAT and S5P data is −0.63 ± 18.13 ppb, and the correlation coefficient (R) is 0.86. After converting the spatial and temporal scales, the data difference changes to a value of −0.23 ± 16.06 ppb, and the correlation is slightly improved (R = 0.87).

(2)

Spatiotemporal correlation analysis

The spatiotemporal correlation analysis aims to separate spatiotemporal deterministic trends and stochastic residual components from regional variables of XCH₄, and further characterize the spatial and temporal correlations of XCH₄ data in the study area [13,14,19]. In this paper, we applied the curve fitting method to long-term XCH₄ data in each 2° latitude band, which referred to Sheng et al. [13]. Figure 4a is the spatiotemporal distribution of XCH₄ deterministic trend, which shows annual increases, seasonal changes in time, and latitudinal gradient in space. The residual components of XCH4 are standard Normal distribution with a distribution range of ± 70 ppb.

Subsequently, we calculated the spatiotemporal variograms using 50 km and 3 days as the spatial and temporal steps, respectively. Figure 4b shows the spatiotemporal variograms model. In order to ensure the XCH₄ data within the large-scale study area would be temporally and spatially autocorrelated, the spatial and temporal ranges that the model flattens out were set to 2500 km and 150 days. In this model, the nugget/sill ratio of about 0.11 indicates a strong spatial and temporal correlation of XCH₄ data in the monsoon region in Asia. Compared to the previous studies, it outperforms the correlation calculated using only GOSAT XCH₄ data in the Eurasia region [19].

(3)

The mapping of gridded XCH₄ data

Based on the XCH₄ residual values and the variogram model, we mapped the gridded data using the spatiotemporal kriging method, which has been widely used for global and regional greenhouse gas (CO₂ and CH₄) interpolation [13,14,18,19,33,34]. The mapping method uses the known XCH₄ observations within a moving cylinder kriging neighborhood to predict the XCH₄ values at the center of each 1° grid. In this paper, the initial ranges of the cylinder were set to 200 km in space and 15 time units (45 days) in time. If the number of available data was less than 20, the spatiotemporal ranges increased by 10 km and 1 time unit, respectively. In the mapping results, the maximum ranges were basically less than 40 time units and 450 km for all grids.

2.2.2. Local Indicators of Spatial Association

In this paper, we used the local indicators of spatial association (LISA) to conduct a spatial analysis of the CH₄ concentration distribution in the study area to identify the spatial patterns of high-concentration and low-concentration CH₄ areas. The local Moran’s I coefficient was utilized to evaluate the significance of the statistical results of each grid, and a proportional relationship was established between the local statistical data and the corresponding global statistical data [35]. The calculation method is shown in Equation (2):

$$I_i=\frac{\left(y_i-\stackrel{-}{y}\right)}{\frac{1}{n}\sum {\left(y_i-\stackrel{-}{y}\right)}^2}{\sum }_{j\ne i}^n{w}_{ij}\left(y_j-\stackrel{-}{y}\right)$$

(2)

where $y_i$ and $y_j$ are the XCH₄ values of grid $i$ and grid $j$, $w_{ij}$ is the spatial weight value of XCH₄ between different grids calculated by the inverse distance squared method, n is the total number of all grids in the study area, and $I$ represents the local Moran index of the $i$th grid. $I_i>0$ indicates a positive spatial correlation; the larger value shows the significant spatial correlation. $I_i<0$ indicates a negative spatial correlation; the smaller value shows the larger spatial disparity [36].

2.2.3. Clustering Spatial Pattern of XCH₄ Temporal Changes

We studied the spatial aggregation characteristics of XCH₄ temporal changes by clustering the XCH₄ time series curves of each grid in the study area [13]. Particle swarm optimization (PSO) is used to optimize the clustering center of KNN (K-Nearest Neighbor) to obtain the optimal clustering center (30 categories) [37,38]. Within the same clustering category, the temporal variation characteristics of CH₄ for each grid are the same or similar.

3. Results

3.1. Spatiotemporal Characteristic of XCH₄ Variations in the Asian Monsoon Region from 2009 to 2021

The spatial and temporal variations of the Mapping-XCH₄ data in the Asian monsoon region are shown in Figure 5 and Figure 6. The temporal characteristics of XCH₄ variations in the study areas are extracted using the curve fitting method [13]. The long-term trend of XCH₄ from 2009 to 2021 increases from 1785.63 ppb to 1897.28 ppb with an annual growth rate of about 8.4 ppb. There is an obvious latitudinal gradient in the XCH₄, and the growth trends are similar in each latitudinal zone. Generally, the CH₄ concentrations reach the highest value in August–November, and the lowest value in April–June each year. In Figure 6, the seasonal cycle peak varies with the latitude, because of the wide range of climate zones in the regions in Asia. Both the growth rates and the seasonal fluctuations of XCH₄ from 2019 to 2021 are higher than those observed from 2009 to 2019, which may be related to different data sources used for mapping. Compared with the individual observations of GOSAT and TROPOMI, the mapping data fill the gap area in the satellite observations, especially for the period of GOSAT observations. The continuous XCH₄ data in space and time reveal the regional spatiotemporal variation characteristics more comprehensively.

Figure 7 shows the spatial distribution of multi-year average XCH₄ from 2010 to 2021. The low values are mainly distributed in China’s Qinghai–Tibet Plateau and the northern regions at latitudes above 40°N. The elevated CH₄ concentrations are mostly located in the North China Plain, southern China, South Asia, and Southeast Asia. These regions are dotted with wetlands and rice paddy fields, which are primary sources of atmospheric CH₄. Figure 8 shows that there are similar spatial distributions of the seasonally averaged XCH₄ over the period from 2009 to 2021. CH₄ concentrations in autumn and summer are generally higher than those in spring and winter, because of the stronger emissions from rice paddies and wetlands in the warmer and wetter seasons. The spatial pattern of XCH₄ is consistent with the mapping result using GOSAT data from 2009 to 2014, as described in Liu et al. [18]. However, the long-term XCH₄ data can reveal the more obvious seasonal variations of XCH₄, especially differentiating CH₄ concentrations between summer and autumn.

3.2. Spatial Pattern of XCH₄ Variations Corresponding to the Surface Emissions

Figure 9 is the spatial distribution of local Moran’s I coefficients calculated based on multi-year averages from 2010 to 2021, which represents the spatial associations of XCH₄ in the monsoon region of Asia. The dashed areas have a significant p-value of less than 0.05. The red and blue boxes correspond to the regions with low and high CH₄ concentrations in Figure 7, respectively. The results indicate that the gridded XCH₄ data for these regions have a significant spatial correlation. Compared to the nearby areas, the Sichuan Basin in China, northern India, and Bangladesh present extremely high XCH₄ values due to high anthropogenic and natural emissions, such as oil deposits, paddy fields, livestock, and topography [39,40,41,42,43,44,45].

The clustering results of the gridded XCH₄ time series are shown in Figure 10. The gridded XCH₄ changes in the same cluster have a similar temporal trend that consists of long-term growth trends and seasonal cycles. The clusters present intense spatial aggregation; the regions with high CH₄ concentrations are divided into clusters 4, 7, 10, 15, and 27, and those with low CH₄ concentrations are divided into clusters 21, 24, 25, and 29. In the clusters of China, we noted that a clear east–west boundary is roughly consistent with the Heihe–Tengchong Line, which reflects two roughly equal parts with contrasting population densities. The XCH₄ in the two parts has climbed steadily at a rate of ~8 ppb every year, but the concentrations in the eastern half are about 2% (25~45 ppb) higher than those in the west. The results indicate that the spatial pattern of CH₄ in China is significantly affected by anthropogenic and natural emissions.

Compared to the anthropogenic emissions from EDGAR in Figure 11, the spatial pattern of high CH₄ concentrations agrees well with the high-emission regions. At the same time, the CH₄ emissions in China also showed obvious differences between the east and west parts. We calculated the spatial correlation between EDGAR emissions and Mapping-XCH₄ based on the clusters to analyze the response of atmospheric CH₄ concentrations to CH₄ emissions. In Figure 11b, the XCH₄ data of each cluster region are subtracted from the annual average of China. There is an exponential linear relationship between XCH₄ and anthropogenic emissions in spatial distribution, and the correlation coefficients are greater than 0.7 for each year from 2010 to 2021. However, influenced by atmospheric transport, high anthropogenic emissions do not imply high CH₄ concentrations. The clusters of 4, 5, 10, and 27 show emission values above the fitting trend.

For the temporal correlation, we compared the annual time series of XCH₄ and anthropogenic emissions in these clusters. The regional CH₄ concentrations increase linearly with anthropogenic emissions (R > 0.8) except for clusters 3, 18, and 29, which have low emissions and high topography. This is because changes in CH₄ concentrations depend not only on surface emissions and sinks from chemical reactions but also on atmospheric transport caused by global and local circulations [44].

3.3. Regional Characteristics of XCH₄ Variations in Urban Agglomerations of China

To investigate the regional characteristics of spatiotemporal XCH₄ variations, we selected four urban agglomerations with high CH₄ concentrations in China as the study areas (Figure 12), including the Beijing–Tianjin–Hebei region and nearby areas (BTH), the Yangtze River Delta (YRD), the triangle of Central China (TCC), and the Chengdu–Chongqing City Group (CCG).

Figure 13 presents the monthly time series of the regional-averaged XCH₄ values calculated by the Mapping-XCH₄ data (red line) and compared with regional fitting values (Fitting-XCH₄) by the curve fitting method using raw satellite XCH₄ observations (gray line). In urban agglomerations with a large number of available observations (BTH, YRD, and TCC), the temporal variations of the monthly values of Mapping-XCH₄ and Fitting-XCH₄ are generally consistent. The large differences only occurred in the summer months, with more missing data during the GOSAT observation period. In CCG, where there are relatively few satellite observations, the Mapping-XCH₄ data also reflect the overall trend of XCH₄ changes. Comparing the monthly time series of the Mapping-XCH₄ and Fitting-XCH₄ data, the large deviations are during the summer months in YRD and TCC and around 2012 in CCG. This is because the curve fitting method is susceptible to extreme values, which result in the Fitting-XCH₄ data during the GOSAT observation period becoming more prone to outliers that are influenced by the spatial-temporal heterogeneity of the data. Therefore, the Mapping-XCH₄ data can be applied to provide effective information on regional XCH₄ changes.

We analyzed the variations in atmospheric CH₄ concentrations in different urban agglomerations by comparing the regional characteristics of XCH₄ and anthropogenic emissions from the EDGAR inventory (Figure 12). The CH₄ concentrations in YRD, TCC, and CCG are higher than those in BTH, which are about 6.74 ± 2.17 ppm, 7.80 ± 3.71 ppm, and 6.94 ± 2.55 ppm, respectively. As the land cover types are mainly farmland and shrubs, the urban agglomerations of YRD, TCC, and CCG have relatively high natural emissions. Additionally, their regional CH₄ sinks are affected by similar hydrothermal conditions. The high concentrations of CCG are also affected by the basin retention effect, though regional annual emissions are generally lower than those of YRD and TCC. Temporally, the increase in anthropogenic emissions in BTH from 2010 to 2021 was approximately twice that of YRD. Meanwhile, the annual XCH₄ values in BTH increased at a rate of 8.95 ppb, which is slightly higher than other regions (8.61 ppb for YRD, 8.46 ppb for TCC, and 8.27 ppb for CCG). These results indicated that the regional XCH₄ values calculated by the Mapping-XCH₄ data accurately quantified the differences among different urban agglomerations.

4. Discussion

In the spatiotemporal kriging interpolation algorithm, the kriging variances are calculated as the mapping error. Figure 14 shows the spatiotemporal distribution of mapping uncertainties expressed by the standard deviation. In terms of space, the uncertainties in other regions are less than 14 ppb on the whole, except for regions in the Indo-China Peninsula. In the article by Sheng et al. [46], it was clearly pointed out that the uncertainty of the mapping data largely depends on the number of available data points used for spatiotemporal mapping. This can also be well demonstrated by comparing Figure 2 and Figure 14a. For the same reason, the mapping uncertainty of the coverage period of TROPOMI data is smaller than that of GOSAT. However, we found that the seasonal variation amplitude of the TROPOMI period was higher than that of the GOSAT period in the analysis of different spatial scales. In the temporal trends for different latitude zones, the difference in the 20°–30° latitude zone is more obvious than in other latitude zones.

Compared with the TCCON data, the mean difference of all the stations is greater than 5.50 ppb, and the standard deviation is greater than 10 ppb (

Table 2. Information of Mapping-XCH₄ vs. TCCON data at different sites used in this paper.

Site	Latitude	Longitude	Start Time	End Time	N	ME	MAE	STD	R
Saga	33.24°N	130.29°E	28 July 2011	30 June 2021	747	8.0386	11.3856	13.3502	0.8970
Rikubetsu	43.46°N	143.77°E	24 June 2014	30 June 2021	439	10.8249	13.2031	12.7522	0.8516
Tsukuba	36.05°N	140.12°E	28 March 2014	28 June 2021	484	5.5391	9.0189	10.8386	0.8452
Xianghe	39.8°N	116.96°E	14 June 2018	30 November 2021	324	–9.3133	13.8883	15.3010	0.7261
Hefei	31.9°N	119.17°E	8 January 2016	31 December 2020	185	8.4948	13.3280	13.7827	0.8256

). The Xianghe station is located in the northern part of the Beijing–Tianjin–Hebei region of China. The concentration of CH₄ in the northwestern part of the station is significantly lower than that in the surrounding areas. Therefore, the mean deviation (ME) is negative, and the standard deviation is higher than at the other sites. Due to the spatial resolution of the coarse grid, the Mapping-XCH₄ data are not able to accurately monitor atmospheric CH₄ concentration changes at local point sources. However, for the time series, the mapping data can reflect the overall temporal variation trend of the area around the site, which can be seen in Figure 15.

5. Conclusions

Reducing CH₄ emissions is recognized as an effective option for rapid climate change mitigation, especially on decadal timescales, because of its shorter lifetime than CO₂. In order to fill the data gaps of individual satellite products, in this study, we mapped the temporal and spatial continuous XCH₄ data (Mapping-XCH₄) in monsoon regions in Asia from 2009 to 2021 using GOSAT and TROPOMI data observations. The dataset was used to analyze the long-term temporal and spatial distribution characteristics and changes.

Compared with the individual satellite observations of GOSAT and TROPOMI, the Mapping-XCH₄ data reveal the regional spatiotemporal variation characteristics more comprehensively. For the spatial distribution of the multi-year average XCH₄ from 2010 to 2021, there are elevated CH₄ concentrations in the North China Plain, southern China, South Asia, and Southeast Asia due to the primary CH₄ sources of wetlands and rice paddy fields. The low concentrations of CH₄ are mainly distributed in China’s Qinghai–Tibet Plateau and the northern regions at latitudes above 40°N. The long-term trend of XCH₄ from 2009 to 2021 increased from 1785.63 ppb to 1897.28 ppb, with an annual growth rate of about 8.4 ppb. In terms of seasonal changes, the CH₄ concentrations in autumn and summer are generally higher than those in spring and winter because of the stronger emissions from rice paddies and wetlands in the warmer and wetter seasons. Through a spatiotemporal correlation analysis, we found that the Mapping-XCH₄ data are positively correlated with emissions from the EDGAR inventory, showing a consistent spatial distribution.

These results demonstrate the potential of the mapping XCH₄ dataset for monitoring long-term CH₄ variations and provide new cases for further study on the driving factors of CH₄ changes and the responses of CH₄ concentrations observed by satellites to anthropogenic emissions at regional scales. In the future, combined with a large number of remote sensing parameters, such as surface temperature, land cover, and meteorological data, more refined interpolation will be carried out by using the deep learning method to increase the ability of the mapping data to identify point source emissions.