Methane Dynamics in Inner Mongolia: Unveiling Spatial and Temporal Variations and Driving Factors ^†

Abstract

Methane (CH₄), the second-largest greenhouse gas contributing to global warming, has a high warming potential despite its short atmospheric lifespan. Inner Mongolia, due to its high carbon and energy consumption industries, faces significant methane emission challenges. This study uses TROPOMI satellite data (February 2019 to December 2022) to analyze the long-term trends and spatial distribution of methane in Inner Mongolia. The results indicate significant spatial heterogeneity in the methane concentration distribution in Inner Mongolia, China. Higher methane concentrations are observed in the southeastern regions, whereas the central regions exhibit relatively lower concentrations. Temporally, the methane concentrations show an increasing trend with seasonal peaks from late August to early September. Using multiple stepwise regression and geographically weighted regression (GWR) methods, the study identifies the key factors influencing methane concentrations. Increased precipitation and soil temperature, along with intensified human activity, contribute to higher methane levels, while rising surface temperatures and increased vegetation suppress methane concentrations. The GWR model provides a better fit compared to the traditional methods, especially in regions with higher methane levels. This research offers insights for developing strategies to mitigate methane emissions and supports China’s emission control targets.

1. Introduction

In response to global climate change, China has committed to achieving a carbon peak by 2030 and carbon neutrality by 2060. The announcement of these goals has garnered significant attention domestically and internationally. The Chinese Ministry of Ecology and Environment recently released the ‘Action Plan for Methane Emission Control’, aiming to actively address climate change and contribute positively to global climate governance [1].

Methane (CH₄) ranks as the second-largest greenhouse gas contributing to global warming, accounting for 20% of the total greenhouse gas emissions and contributing 25% to global warming, second only to carbon dioxide (CO₂) [2,3]. Despite its lower atmospheric concentration compared to CO₂, methane serves as a critical climate driver, with a greenhouse effect that is 28 times more potent than CO₂ over a century [4]. Due to its relatively short atmospheric lifetime, controlling methane emissions has immediate and significant effects on achieving carbon neutrality, improving atmospheric quality, and mitigating global warming [5,6,7]. Inner Mongolia, located in northern China, primarily consists of the Mongolian Plateau, with a terrain characterized by plains, mountains, and hills. As a province with high carbon and energy consumption, the spatiotemporal evolution of methane in Inner Mongolia warrants attention.

Precise greenhouse gas monitoring improves emission management [8]. Scholars have studied methane emissions’ spatiotemporal patterns and sources, primarily in terms of agriculture and energy production [9,10]. The research includes agricultural emissions from rice paddies [11] and livestock [12]. Additionally, research has been conducted at the national level to estimate the methane emissions from major grain-producing and grain-consuming regions, depicting the spatiotemporal patterns of the agricultural methane emissions in China [13]. Scholars have also quantified the methane emissions from the natural gas production in Canada’s Montney Basin [14] and explored the relationship between oil, natural gas production, and methane emissions in the Utah Basin [15]. According to the International Energy Agency, China accounted for 15.65% of the global methane emissions in 2022, making their reduction a key academic research topic [16]. Wei et al. investigated the long-term trends, spatial differences, and relationships with climate variables of the methane column concentrations on the Qinghai–Tibet Plateau since December 2010 [17]. He et al. not only described the spatiotemporal distribution characteristics of methane in Hebei Province but also estimated the contribution of ruminant animals to the methane emissions in the province [18]. On a national scale, Zhang et al. identified the relationships between the methane emission trends and coal mining and rice cultivation considering China’s policy drivers [19].

Methane emission studies often use ground-based data and emission inventories, with limited past use of satellites like OMI and SCIAMACHY [20,21,22]. With advancements in satellite remote sensing technology, Sentinel-5P, equipped with the latest trace gas detection instrument TROPOMI (TROPOspheric Monitoring Instrument), has entered scholars’ radar. Launched by the European Space Agency on 13 October 2017 [23,24], Sentinel-5P offers a higher spatial resolution and more accurate precision. Launched by the European Space Agency in 2017 [23,24], the TROPOMI data have shown good agreement with the TCCON measurements [25] and other ground-based studies [26,27,28]. This study uses Google Earth Engine to analyze the methane concentrations and influencing factors in Inner Mongolia over the past four years, supporting atmospheric environmental management.

2. Materials and Methods

2.1. Study Area

Our study focuses on the Inner Mongolia Autonomous Region (Figure 1), located between 97°12′–126°04′ E and 37°24′–53°23′ N. Inner Mongolia, a crucial ecological barrier and a major agricultural and livestock production base in northern China, shares borders with Mongolia and several Chinese provinces, including Heilongjiang, Jilin, Liaoning, Hebei, Shanxi, and Ningxia. Spanning approximately 2400 km from east to west and over 1700 km from north to south, it covers an area of 1.183 million square kilometers. The region features diverse topography, with forests and plains in the east, deserts in the west, and plains in the central area, including the Hetao Plain and the Tumochuan Plain. Inner Mongolia has a temperate continental monsoon climate, with water and heat resources varying from semi-humid in the east to extremely arid in the west.

2.2. Data Sources

The TROPOMI daily offline Level 3 (OFF L3) CH4 observation data hosted on the Google Earth Engine (GEE) cloud platform were analyzed. Google Earth Engine is a cloud-based geospatial analysis platform. The processing of daily, monthly, and annual average TROPOMI CH4 observations was conducted using the JavaScript editor within the GEE framework. The data span from 2 February 2019 to 31 December 2022. Details of the datasets used to select influencing factors are provided in

Table 1. Sources of data on impact factors and their spatial and temporal resolutions.

Impact Factor	Product	Band Name	Spatial Resolution	Temporal Resolution
Human activity intensity	VIIRS Stray Light Corrected Nighttime Day/Night Band Composites	avg_rad	500m	monthly
Vegetation coverage	MOD13A2.006 Terra Vegetation Indices 16-Day Global 1 km	NDVI	1km	16 day
Soil Temperature	ERA5-Land Daily Aggregated—ECMWF Climate Reanalysis	soil_temperature level_1	0.1°	daily
Precipitation	GPM: Global Precipitation Measurement	precipitaioncal	0.1°	daily
Land Surface Temperature	MOD11A2.006 Terra Land Surface Temperature and Emissivity 8-Day	LST_Day_1km	1km	8 day

. To assess the temporal and spatial consistency between these factors and CH₄ concentrations, annual mean synthesis and resampling operations were applied, achieving a spatial resolution of 10 km post-resampling.

2.3. Methods

2.3.1. Spatial Autocorrelation Theory

Spatial autocorrelation examines the correlation between a variable or geographical phenomenon and its spatial location, assessing whether there is significant correlation between the attribute values of adjacent spatial elements. In this study, Moran’s I index was employed to quantify the clustering of methane concentrations in the Inner Mongolia region. Moran’s I ranges from −1 to 1; values less than 0 denote negative correlation, 0 signifies statistical non-correlation, and values greater than 0 indicate positive correlation. Moran’s I is calculated using the following formula:

$$I={\displaystyle \frac{n}{{\sum }_{i-1}^n{\sum }_{j-1}^n{\omega }_{ij}}}\cdot {\displaystyle \frac{{\sum }_{i-1}^n{\sum }_{j-1}^n{\omega }_{ij}\left(Z_i-\overline{Z}\right) \left(Z_j-\overline{Z}\right)}{{\sum }_{i-1}^n{\left(Z_i-\overline{Z}\right)}^2}},$$

(1)

where $n$ is the number of observed units, $Z_i$ is the attribute value of the ith observed unit, $\overline{Z}$ is the average attribute value of all observed units, and ${\omega }_{ij}$ is the spatial weight matrix. In this study, for the analysis of Moran’s I, the average value data of methane within the time scale from February 2019 to December 2022 are used, which are calculated through Google Earth Engine (GEE).

2.3.2. Cold and Hot Spot Analysis

Cold and hot spot analysis is employed to identify spatial clustering characteristics of high (hot spots) and low (cold spots) methane concentrations within local regions. This study utilized cold and hot spot analysis to identify methane concentration patterns across each region. The formula used is as follows:

$$G_i^{*}={\displaystyle \frac{{\sum }_{j-1}^n{\omega }_{ij}\left(Z_j-\overline{Z}\right)}{{\sum }_{j-1}^n{\omega }_{ij}}}\cdot {\displaystyle \frac{n{\sum }_{j-1}^n{\omega }_{ij}^2-{{\sum }_{j-1}^n{\omega }_{ij}^2}^2}{{\sum }_{j-1}^n{\left(Z_i-\overline{Z}\right)}^2(n-1)}},$$

(2)

where $G_i^{*}$ is the Z score, which is used to indicate the degree of clustering of each observed unit. A higher Z score indicates a hot spot area; a lower one indicates a cold spot area. For the cold and hot spot analysis, this study conducts mean synthesis through GEE and uses the average value data of methane within the time scale from February 2019 to December 2022.

2.3.3. Spatial Regression Analysis of Influencing Factors

Regression analysis is a widely used quantitative analysis method that primarily investigates statistical relationships and quantitative changes between variables. It describes and reflects these relationships in the form of regression equations, thereby better understanding the scope of variables influenced by other variables and providing a scientific basis for control and prediction.

In this study, for the regression analysis, the data used are the annual average values of methane for each year from 2019 to 2022. Regarding other influencing factors, their annual average values are also calculated for each year. In addition to being uniformly resampled to 10 km, as previously mentioned, these factors are processed in a consistent manner in terms of time scale.

• Stepwise Multiple Regression

Multiple linear regression is a commonly used statistical analysis method for studying the dependent relationship between the dependent variable and multiple independent variables. It explores correlations between variables using experimental data, constructs approximate mathematical models and expressions based on data and experience, and predicts and controls corresponding variables accordingly. The computational steps of multiple linear regression are as follows:

• Introduce variables one by one to construct a linear regression model of the dependent variable with i independent variables:

$$y={\beta }_0{+\beta }_p{x}_p+\epsilon ,$$

(3)

Here, y is the dependent variable, corresponding to CH₄ concentration in grid units in this paper. ${\beta }_0$ is the intercept constant, ${\beta }_p$ is the regression coefficient affecting factor x_p, and x_p represents p independent variables. $\epsilon$ is the variability in y that is included in the model but not explained by the linear relationship with p independent variables; it is a random variable known as the error term.

2.

For each variable added to the model, use hypothesis testing to determine whether it significantly improves the model fit. Generally, t-statistics or F-statistics are used to test the significance of coefficients. If the result is significant, the variable is retained in the model; otherwise, it is removed.

3.

Repeat the above steps to gradually introduce or eliminate variables until no further variables can be added or removed, ultimately obtaining the optimal model, which is as follows:

$$y={\beta }_0+{\beta }_1{x}_1{+\beta }_2{x}_2{+\beta }_3{x}_3{+\beta }_4{x}_4+\dots {+\beta }_p{x}_p+\epsilon ,$$

(4)

where $y$ is the dependent variable, corresponding to the CH₄ concentration in grid cells. The term ${\beta }_0$ is the intercept constant, ${\beta }_1{,\beta }_2{,\beta }_3{,\dots \beta }_p$ are the regression coefficients for the influencing factors x₀, x₁, x₂, x₃…x_p, and $\epsilon$ is the error term.

• Geographically Weighted Regression (GWR)

Geographically weighted regression (GWR) is a regression analysis method tailored for spatial data, aiming to locally model the relationship between variables at different spatial locations. This approach explores the variation between independent and dependent variables of the study subject at a specific spatial scale. In contrast to traditional OLS regression, GWR incorporates spatial adjacency matrices to illustrate spatial proximity, thereby constructing spatial dependency relationships among variables. This enables a better reflection of spatial heterogeneity and non-stationarity. The computational formula is as follows:

$$Y={\rho }_{1}Y+X+u, u={\lambda }_2·\epsilon +\mu$$

(5)

where Y represents the dependent variable, while X denotes the independent variable. $\beta$ represents the spatial regression coefficients of the dependent variable. u signifies the error term that varies with spatial changes, and μ denotes white noise. W₁ is the spatial weight matrix reflecting the spatial trend of the dependent variable itself, while W₂ is the spatial weight matrix reflecting the spatial trend of the residuals. ρ represents the coefficient of spatial lag, and λ represents the spatial error coefficient.

3. Results

3.1. Spatial Distribution of CH₄

To understand the distribution characteristics of the methane concentration in Inner Mongolia, two methods were employed: spatial autocorrelation analysis and cold and hot spot analysis. The spatial autocorrelation analysis yielded a Moran’s I value of 0.873362, a z-score of 96.427915, and a p-value of less than 0.01, indicating a high positive spatial correlation. This suggests that the observed clustering is highly unlikely to be due to chance (with a probability of less than 1%). This implies that the methane concentration tends to cluster in Inner Mongolia, with the geographically adjacent regions exhibiting similar properties in the distribution grid. Moreover, the Moran’s I value not only demonstrates a significant positive spatial correlation in the overall methane concentration distribution in Inner Mongolia but also provides clues for in-depth analysis. To more precisely observe the spatial aggregation of the methane concentrations in different areas, a cold and hot spot analysis was conducted using ArcGIS 10.2 for visualization. Figure 2 displays the aggregation hot spots and cold spots of the methane concentration, aiding in a more accurate understanding of its spatial distribution patterns in Inner Mongolia. As shown in Figure 2, with 95% confidence, Chifeng and Tongliao in southeastern Inner Mongolia and Bayannur in western Inner Mongolia can be identified as methane emission hot spots. Conversely, the central part of Inner Mongolia is identified as a cold spot for methane emissions.

Furthermore, Figure 3 illustrates the distribution of the annual average CH₄ concentration in the Inner Mongolia Autonomous Region from 2019 to 2022. According to Figure 3, the methane concentration increased annually, with the values ranging from 1760 ppb to 1970 ppb. In terms of the spatial distribution, the pattern has remained largely consistent over the past four years, showing a trend of lower concentrations in the east and higher concentrations in the west, and lower concentrations in the north and higher concentrations in the south. Specifically, the northeast region, the west of Hohhot, Chifeng, and Tongliao in the southeast, and Bayannur, Ordos, and Hohhot in the west are areas with high methane concentrations. In contrast, the central regions, including Xilin Gol League and Ulanqab, exhibit relatively low methane concentrations.

3.1.1. Temporal Distribution of CH₄

Figure 4 illustrates the change in the monthly mean methane concentration in Inner Mongolia from 2019 to 2022. It is evident that the methane concentrations have been increasing year by year and exhibit distinct seasonal characteristics. Additionally, using the latitudes and longitudes of twelve cities in the Inner Mongolia Autonomous Region (Hohhot, Baotou, Wuhai, Chifeng, Tongliao, Erdos, Hulunbuir, Bayan Nur, Ulanqab, Hinggan League, Xilingol League, and Alxa League), the monthly average methane concentrations were obtained. Figure 4 depicts the temporal trends of the methane concentration distributions across various cities in Inner Mongolia.

Over the past 47 months, the monthly mean change in methane concentration shows significant seasonal differences and periodic fluctuations in the Inner Mongolia Autonomous Region. From a seasonal perspective, the methane concentration exhibits a clear pattern of being “high in summer and low in spring”. Specifically, methane concentrations typically drop to their lowest levels between March and April and rise rapidly to their annual peaks between June and August. Methane concentrations are also generally high in winter, sometimes reaching a peak in December.

On an urban scale (Figure 5), the methane concentrations are high in Wuhai and Tongliao while relatively low in Baotou and Hulunbuir. According to the literature, methane emissions are related to fossil energy production, rice cultivation, and animal rumination. Wuhai City and Alxa League, located in the southwest of the Inner Mongolia Autonomous Region, have significant coal deposits and less vegetation, resulting in substantial methane emissions. Tongliao, situated in the southeast in the hinterland of Horqin Grassland, has developed agriculture and animal husbandry and is a major cattle and sheep production base, leading to high methane emissions.

3.1.2. Driving Factor Analysis of CH₄

With the advancement in urbanization and industrialization, methane emissions have shown a rapid growth trend, prompting extensive research by scholars both domestically and internationally on the driving factors. Therefore, considering the existing remote sensing image resources available in Google Earth Engine (GEE), this paper selects several factors that are suitable for analysis using GEE. The indicators used in the analysis and their meanings are detailed in

Table 2. Indicator definitions for spatial analysis.

Indicator	Definition
Day/Night Band (DNB)	representing nocturnal light intensity as a proxy for human activity
Normalized Difference Vegetation Index (NDVI)	used for assessing vegetation coverage and growth status
Soil Temperature (SOIL_TEMP)	indicating the temperature of the soil
Precipitation (PRE)	representing the amount of rainfall
Land Surface Temperature (LST)	indicating the temperature of Earth’s surface

.

Based on the area of the Inner Mongolia Autonomous Region, a 10 km by 10 km grid was constructed in this study. The mean CH₄ concentration values and related influencing factors were extracted for each grid unit. All the data were standardized, and a total of 10,893 grid cells were selected for the experiments.

• Stepwise Multiple Regression Analysis

Using stepwise regression analysis, the grid cell data were analyzed with SPSS statistics 25 software to quantify the relationship between the influencing factors and methane concentrations from 2019 to 2022. The results are presented in

Table 3. Stepwise multiple regression results.

Impact Factor	2019	2020	2021	2022
PRE	0.168 *	0.470 *	0.320 *	0.280 *
NDVI	−0.315 *	−0.546 *	−0.398 *	−0.206 *
LST	-	−0.308 *	−0.184 *	−0.120 *
SOIL_TEMP	0.231 *	0.372 *	0.345 *	0.189 *
DNB	-	0.184 *	0.124 *	0.139 *
R2	0.579	0.496 *	0.494 *	0.505 *

* indicates p < 0.01; - indicates non-significant.

.

In 2019, only the DNB and LST variables showed significance, indicating that these drivers had a stable and significant impact on the methane concentration at the 10 km scale compared to the other four drivers: PRE, NDVI, SOIL_TEMP, and DNB. The corrected goodness of fit of the model, represented by R², indicates that the selected variables explain approximately 50% to 60% of the methane concentration variability in the Inner Mongolia Autonomous Region. For precipitation (PRE), a positive effect was observed from 2019 to 2022, with this effect initially strengthening and then weakening. The NDVI vegetation index demonstrated strong inhibitory and negative effects, also showing an initial strengthening followed by weakening. The LST surface temperature exhibited a continuous negative effect on methane from 2020 to 2022. SOIL_TEMP had a consistent positive effect with fluctuations, and DNB showed a weak promoting effect from 2020 to 2022.

In conclusion, precipitation and soil temperature have a strong positive effect on the methane concentration in Inner Mongolia, whereas the vegetation index, surface temperature, and human activity intensity have a negative inhibitory effect. Climate variables such as PRE, LST, and SOIL_TEMP have profound effects on methane sources and sinks. Precipitation regulates methane emissions and absorption by influencing vegetation succession and soil moisture [29]. Wet soil provides an ideal anaerobic environment for methanogens, promoting methane production [30]. Additionally, temperature affects the substrate supply for methanogens, influencing their methane production rate, with warming enhancing the methane production processes [29]. The NDVI vegetation index is negatively correlated with methane concentration because deserts and wastelands predominate in Inner Mongolia, resulting in rare plant growth, low biomass density, and minimal plant methane emissions. Furthermore, reduced vegetation causes soil exposure and drought, promoting the soil oxidation of the atmospheric methane [31]. In Inner Mongolia, human activities, particularly grazing, significantly affect the methane concentration. The rumination of cattle and sheep is a major methane source; however, a study in the Eurasian steppe found that heavy grazing significantly reduces methane absorption, whereas moderate and mild grazing do not have significant effects [32].

Furthermore, the regression diagnostic parameter Jarque–Bera was significant, indicating that the model errors exhibited significant non-normal characteristics. Additionally, the p-value of the Koenker (BP) test was significant at the 1% level, suggesting the presence of statistically significant heteroscedasticity and instability in the model. Given these findings, it is more appropriate to analyze the data using a geographically weighted regression (GWR) model, which can better account for the spatial variability and unsteady state characteristics of the dataset.

• Geographically Weighted Regression (GWR)

As mentioned above, the methane concentration distribution exhibits significant spatial heterogeneity, suggesting that a geographically weighted regression (GWR) model may provide a better fit by accounting for spatial scale variations. The GWR model can assess the impact of various drivers on methane concentrations across different spatial locations, thereby enabling a deeper exploration of the spatial differentiation of the methane concentration in the Inner Mongolia Autonomous Region.

Through a comparative analysis, the average overall goodness of fit for the GWR model from 2019 to 2022 was superior to that of the linear stepwise regression analysis. The R² values obtained from the GWR model for each year were categorized into five groups using the natural break classification method, and the corresponding spatial distribution patterns were illustrated, as shown in Figure 6.

It can be observed that there are spatial differences in the R² distribution of the GWR model. Specifically, the average goodness of fit in 2019 was 0.781, with the regions of high R² values primarily concentrated in the northeast and west. In 2020, the average goodness of fit was 0.719, with a similar R² distribution pattern to 2019 and the high-fit regions remaining in the northeast and west. In 2021, the average goodness of fit was 0.769, with high R² values again in the northeast and west. However, in 2022, the average goodness of fit decreased to 0.680, with prominent R² values in the southeast, northeast, and central regions. This indicates a strong connection between the fitting performance of the GWR model and the spatial distribution of the methane concentrations. The regions with higher methane concentrations, such as the southeast and western parts of the Inner Mongolia Autonomous Region, showed stronger correlations between the drivers (PRE, NDVI, LST, SOIL_TEMP, and DNB) and methane concentrations, resulting in better model fitting.

Furthermore, the GWR model, which incorporates local parameter estimation, demonstrated significant advantages in elucidating the relationship between the explanatory variables and the dependent variable compared to the stepwise regression method. The GWR model explained 70% to 80% of the variance in the dependent variable, which is approximately 30% higher than that explained by the stepwise regression model. This advantage is primarily due to the stepwise regression’s assumption of spatial distribution homogeneity for methane concentrations and independent variables, while, in reality, a significant spatial autocorrelation exists within the study area. Therefore, GWR models, which account for spatial scale variations, more accurately capture the complex relationship between the methane concentration distribution and its drivers, making them more suitable for modeling analysis.

4. Conclusions

In this study, spatial autocorrelation analysis and cold and hot spot analysis were utilized to investigate the temporal evolution pattern of methane in the Inner Mongolia Autonomous Region. Five driving factors—precipitation, soil temperature, surface temperature, normalized vegetation index, and human activity intensity—were selected to quantitatively analyze their influence on the methane concentration in Inner Mongolia using multiple stepwise regression and geographically weighted regression. The following conclusions were drawn:

• From 2019 to 2022, the methane concentration in Inner Mongolia exhibited significant spatial clustering characteristics, with high-value areas primarily concentrated in the southeast and low-value areas in the central region. Overall, the methane concentration in Inner Mongolia showed an increasing trend year by year, maintaining a relatively stable spatial distribution pattern characterized by lower concentrations in the east and north and higher concentrations in the west and south. Temporally, the methane levels in Inner Mongolia displayed clear seasonal fluctuations, being higher in summer and lower in spring, with a cyclical pattern of change. Typically, the concentrations decreased in March to April and rose rapidly from June to August. Notably, Tongliao City, with its developed agriculture and animal husbandry, and Wuhai City and Alxa League, with sparse vegetation and rich coal resources, exhibited relatively high methane concentrations, while Baotou and Hulunbuir had relatively low concentrations.
• The effects of the five drivers on the methane concentration in Inner Mongolia were differentiated. The multiple stepwise regression analysis showed that precipitation (PRE) and soil temperature (SOIL_TEMP) had continuous positive effects, while human activity intensity (DNB) had a weak effect, and vegetation index (NDVI) and surface temperature (LST) had inhibitory effects. There is a reciprocal relationship between methane emissions and climate change. Precipitation, temperature, and the vegetation index influence methane emissions by regulating soil moisture and the living environment of methanogens. Additionally, human activities, particularly grazing, significantly affected the methane absorption and emission processes. Notably, the impact of different drivers on the methane concentrations exhibited distinct spatial heterogeneity. After considering the spatial location factors, the overall goodness of fit of the geographically weighted regression model was significantly improved compared to the multiple stepwise regression model, especially in areas with higher methane concentrations.

This paper explored the spatial and temporal dynamics of methane in Inner Mongolia and quantitatively analyzed the effects of five types of drivers on the methane concentrations. However, the specific mechanisms behind these effects require further investigation. The results of this study provide an important scientific reference for developing targeted methane emission reduction strategies.