Spatiotemporal Characteristic of XCO₂ and Its Changing Contribution Rate from Different Influencing Indicators in Mongolian Plateau of Central Asia

Abstract

The Mongolian Plateau plays a crucial role in global carbon cycling, but the spatiotemporal characteristics of XCO₂ concentration and its driving mechanism remain insufficiently explored. To solve this scientific issue, the synergistic methodology of mathematical statistics—the Pearson correlation and random forest model—was established using the main source of Orbiting Carbon Observatory 2 (OCO-2) satellite data. Results indicate the following: (1) Average XCO₂ concentration of the Mongolian Plateau was 412 ppm, with an annual growth rate of 2.29 ppm/a from 2018 to 2022, along with higher values in the south and lower values in the north. The seasonal change displayed a clear temporal feature, in the order of spring (414.83 ppm) > winter (413.4 ppm) > autumn (411.3 ppm) > summer (409.12 ppm). The spatial distributions in spring, autumn, and winter were relatively consistent, all showing higher XCO₂ concentrations in the east and lower concentrations in the west, whereas summer exhibited the opposite pattern. (2) From the perspective of the natural environment, XCO₂ change was negatively correlated with the normalized difference vegetation index (NDVI), precipitation (PRE), and temperature (TEMP). Temporal analysis further revealed that this negative correlation was most pronounced in the eastern region, in which these three elements were all relatively high. (3) According to the random forest model, the influence of both single and interactive factors on the plateau’s XCO₂ varied significantly. A comparison of driving factors revealed that the NDVI had the highest contribution rate (0.35), followed by fossil fuel combustion emissions (ODIAC), wind direction (WD), and wind speed (WS). As for interaction effects, the combination of NDVI and ODIAC showed the highest contribution rate (over 0.25), indicating a strong joint influence on XCO₂. Other important interactions included WS and WD, ODIAC and WS, and NDVI and WS (all above 0.05). These findings provide valuable insights into the driving mechanisms of XCO₂ on the Mongolian Plateau, offering a reference for regional carbon emission reduction policies.

1. Introduction

Global climate change is profoundly impacting the Earth’s system, with global warming—driven by the rapid rise in greenhouse gas concentrations such as carbon dioxide and methane—posing a significant threat to ecosystems [1]. The Sixth Assessment Report of the Intergovernmental Panel on Climate Change highlights that over the past century, the overall increase in global temperatures has been primarily caused by fossil fuel combustion and emissions of other greenhouse gases, with the current global average temperature now approximately 1.1 °C higher than pre-industrial levels [2]. The instability of the global climate system and ongoing warming trends have led to more frequent extreme weather events, bringing risks to human life and production activities [3]. As a key ecological region of the Eurasian continent, the Mongolian Plateau is highly sensitive to global climate change, making it an ideal location for studying CO₂ variations and their driving mechanisms [4]. Understanding the spatiotemporal characteristics of CO₂ changes in this region and their influencing factors will not only deepen our knowledge of regional carbon cycle processes, but also provide valuable scientific insights for global climate change research.

The Global Atmosphere Watch (GAW) and the Total Carbon Column Observing Network (TCCON) are two major ground-based observation networks that provide high-precision data on surface and column-averaged CO₂ concentrations [5]. However, the spatial coverage of ground-based observation stations is limited, making it challenging to comprehensively depict the global distribution of CO₂. In contrast, satellite observations utilizing solar or terrestrial radiation in the near-infrared/shortwave infrared (NIR/SWIR) and thermal infrared (TIR) bands can provide large-scale CO₂ data, significantly expanding the spatial coverage of observations [6]. Currently, multiple satellites, including the Atmospheric Infrared Sounder (AIRS), the Scanning Imaging Absorption Spectrometer for Atmospheric CH₄ (SCIAMACHY), the Infrared Atmospheric Sounding Interferometer (IASI), and Sentinel-5P, are widely used for global greenhouse gas monitoring [7,8,9,10,11]. In recent years, satellites specifically designed for CO₂ observation, such as TanSat, GOSAT, and OCO-2, have further improved measurement accuracy and spatial resolution [12]. Among them, OCO-2, launched by NASA on 2 July 2014, is dedicated to global atmospheric CO₂ monitoring. OCO-2 employs shortwave infrared retrieval techniques to provide high-precision, high-temporal-resolution data on the column-averaged dry-air mole fraction of CO₂. This study utilizes OCO-2 observational data to analyze the spatiotemporal variations of XCO₂ over the Mongolian Plateau and its influencing factors.

Currently, research on atmospheric CO₂ variations over the Mongolian Plateau remains relatively limited, particularly in terms of long-term XCO₂ observations and analyses of its driving factors. Existing studies primarily focus on the spatiotemporal variations of near-surface CO₂ over this region. Specifically, the annual average concentration of near-surface CO₂ over the Mongolian Plateau has shown a gradual increase, ranging from 2.19 to 2.38 ppm/a [13]. Furthermore, studies on near-surface CO₂ and its influencing factors in Mongolia indicate that, based on Köppen climate classification analysis, CO₂ concentrations are negatively correlated with the NDVI [14]. While these studies provide valuable insights into the carbon cycle processes of the Mongolian Plateau, they are largely constrained to near-surface observations and lack a systematic assessment of XCO₂ and its environmental drivers.

As can be seen from the above, spatiotemporal monitoring and driving research on XCO₂ has become a current academic focus. Regarding this scientific issue, the research was executed using high-resolution OCO-2 satellite observations in this study. Our research objective is (1) to provide the spatiotemporal distribution and seasonal evolution pattern of XCO₂ over the Mongolian Plateau (i.e., Inner Mongolia of China and Mongolia Country); (2) to analyze the changes in XCO₂ from the perspectives of natural environmental factors and human activities, and further to quantify the contributions of different environmental factors and human activities to XCO₂ variations and identify key driving forces, including the factors of temperature, precipitation, wind speed, vegetation, and fossil fuel combustion; and (3) to explore the interactions among major environmental factors and their influence on XCO₂ dynamics. In the process of achieving this goal, we innovatively employed the random forest model to quantify the contributions of different factors to XCO₂ changes. To establish a better approach system to comprehensively analyze the research content, the synergistic methodology of mathematical statistics—Pearson correlation and random forest model—was established. This research contributes to the quantitative assessment of CO₂ changes in the region and provides a scientific basis for the fields of industrial carbon emissions, climate warming, and extreme weather events.

2. Data and Methods

2.1. Overview of the Study Area

The Mongolian Plateau is located in Central Asia, roughly spanning 37°–52° N and 87°–122° E. As shown in Figure 1, the region is characterized by a relatively high elevation, predominantly ranging between 800 and 1500 m [15]. It features an arid to semi-arid climate, with low annual temperatures and scarce precipitation. The ecosystem is primarily composed of grasslands, deserts, and forests. In recent years, climate change and human activities have increasingly threatened the region’s ecological stability, exacerbating issues such as vegetation degradation and desertification, which in turn have weakened its carbon sequestration capacity while intensifying carbon emission pressures [16]. As a typical inland region, the Mongolian Plateau is not directly influenced by oceanic carbon sinks, and its XCO₂ variations are primarily regulated by terrestrial ecosystems and anthropogenic emissions. Its unique geographical location and climatic conditions make it a key area for climate change research.

2.2. Data Collection

2.2.1. OCO-2 Data

This study primarily utilized the OCO-2 Level 2 Lite Full Physics Version 11.1r satellite data products relevant to the study region. OCO-2 is specifically designed for global atmospheric CO₂ concentration measurements. It operates in a sun-synchronous orbit (approximately 705 km above the Earth’s surface), completing one orbit around the Earth every 99 min, with a local overpass time of 13:36 [17]. It is equipped with a hyperspectral imaging spectrometer, which provides ground-level XCO₂ observation data with a spatial resolution of approximately 2.25 km, with an expected precision ranging from 0.5 to 1 ppm [18]. To ensure high-quality and reliable data, region-specific observations were filtered using a quality flag value of 0, although some temporal and spatial data gaps may still exist due to orbital and atmospheric limitations.

2.2.2. MODIS Data

The normalized difference vegetation index is derived from the MOD13A3 dataset of the MODIS satellite and is widely used to assess vegetation growth and health. This dataset provides NDVI values with an 8-day temporal resolution and a 500-m spatial resolution, enabling precise monitoring of vegetation seasonality and growth dynamics [19]. Since vegetation influences CO₂ concentrations through both photosynthesis and respiration, the NDVI is chosen as a representative vegetation index because it not only directly reflects vegetation coverage but also provides long-term continuous observations, making it suitable for large-scale assessments of vegetation’s impact on XCO₂ variations.

2.2.3. ERA5-Land Data

ERA5-Land is a high-resolution reanalysis dataset provided by the European Centre for Medium-Range Weather Forecasts, featuring a spatial resolution of 0.1° and an hourly temporal resolution, making it suitable for regional climate change analysis. This study utilizes 2-m air temperature, 10-m wind direction, and wind speed from ERA5-Land [20]. Air temperature influences vegetation growth and soil respiration, thereby regulating regional CO₂ exchange, while wind direction and wind speed play a crucial role in atmospheric CO₂ transport, shaping the spatial distribution of XCO₂ [21]. To facilitate analysis, this study applies a conversion formula to transform U and V wind components into wind speed and wind direction data for evaluating their impact on the spatial distribution of XCO₂.

2.2.4. CHIRPS Data

Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) is a high-resolution precipitation dataset that integrates satellite observations with ground-based meteorological station data, providing precipitation estimates at a 0.05° spatial resolution and daily temporal resolution. This study utilizes CHIRPS precipitation data (PRE), as precipitation is a key factor influencing vegetation growth and soil moisture. Additionally, it plays a regulatory role in soil microbial activity and carbon release processes [22].

Table 1. Summary of data used in this study, including source and spatiotemporal resolution.

Data Name	Data Source	Spatiotemporal Resolution
Atmospheric CO2 column concentration	OCO-2 Satellite Observation Data	16-day/1.29 × 2.25 km
Normalized difference vegetation index	MODIS/MOD13A3	16-day/1 km
Temperature and precipitation	ERA5-LAND	Hourly/0.1° × 0.1°
Wind direction and wind speed	ERA5-LAND	Hourly/0.1° × 0.1°
Precipitation	CHIRPS	Daily/0.05° × 0.05°
Fossil fuel combustion emissions	ODIAC	Monthly/1 km
Digital Elevation Model	Shuttle Radar Topography Mission	Year/30 m

presents detailed information on the datasets used in this study, including OCO-2, MODIS, ERA5-Land, ODIAC and CHIRPS.

2.3. Research Methodology

2.3.1. Preprocessing of OCO-2 Satellite Data

To facilitate the analysis of the spatiotemporal distribution of XCO₂, spatial interpolation is applied to convert discrete point data into a continuous spatial distribution map. However, the high density of OCO-2 satellite observation points may lead to overfitting if interpolated directly. To reduce data redundancy and mitigate this issue, this study applies an averaging process using a 0.1° × 0.1° grid to resample the XCO₂ data. This approach enhances the smoothness of the interpolation results and improves spatial visualization.

2.3.2. Factor Selection and Calculation

For time-series data integration and spatial correlation analysis, XCO₂ data are overlaid with the monthly mean raster data of each influencing factor to generate a multi-band time-series image. Correlation analysis tools are then used to compute the pixel-wise correlation between XCO₂ and the multi-band images of influencing factors. This study employs the Pearson correlation coefficient, which ranges from –1 to 1 and quantifies the strength and direction of a linear relationship between two variables: –1 indicates a perfect negative correlation, 0 indicates no linear correlation, and 1 indicates a perfect positive correlation. The formula used to obtain this value is given below:

$$r={\displaystyle \frac{\sum (X-\overline{X})(Y-\overline{Y})}{\sqrt{\sum {(X-\overline{X})}^2\sum {(Y-\overline{Y})}^2}}}$$

(1)

Here, $X$ and $Y$ are two variables representing the XCO₂ values and the values of each influencing factor, respectively, and $\overline{X}$ and $\overline{Y}$ are the mean values of these two variables [23,24].

2.3.3. Variance Inflation Factor Analysis

To quantitatively assess collinearity among independent variables, the variance inflation factor (VIF) was calculated. The VIF quantifies the extent to which a given variable is explained by other independent variables, and the thresholds are commonly defined as follows: VIF < 5: Weak collinearity, negligible impact; 5 ≤ VIF < 10: Moderate collinearity, acceptable influence; VIF ≥ 10: Severe collinearity, which may affect model stability and requires corrective measures. The formula for the VIF is as follows:

$${VIF}_i={\displaystyle \frac{1}{1-R_i^2}}$$

Here, $R_i^2$ represents the coefficient of determination when the $i$th explanatory variable is regressed on all other explanatory variables.

2.3.4. ODIAC Data

The Open-source Data Inventory for Anthropogenic CO₂ is a dataset providing high-resolution global CO₂ emission data. It uses a 1 × 1 km grid to record CO₂ emissions from human activities in detail, covering major sources such as fossil fuel combustion, cement production, and natural gas burning [23,24]. The ODIAC dataset has been developed by an international research team to provide accurate and accessible information on global CO₂ emissions in support of in-depth research on climate change and environmental management.

As shown in Figure 2, ODIAC data indicate that CO₂ emissions from fossil fuel combustion are primarily concentrated in Inner Mongolia, particularly in the Hohhot–Baotou–Ordos region. This area has a relatively high level of industrialization and a dense transportation network, forming distinct CO₂ emission hotspots. In contrast, CO₂ emissions in Mongolia are relatively sparse and mainly concentrated in and around the capital, Ulaanbaatar. Specifically, the Hohhot–Baotou–Ordos region hosts 1219 heavy industry enterprises, accounting for 42% of all heavy industry enterprises in Inner Mongolia’s 12 leagues and cities, highlighting the region’s significant contribution to CO₂ emissions.

It was determined that fossil fuel combustion on the Mongolian Plateau resulted in approximately 1114 Mt of CO₂ emissions. Statistical analysis of CO₂ emissions across leagues and cities revealed significant differences between Inner Mongolia and Mongolia. As shown in Figure 3, red represents emissions from Mongolia, while blue represents the average monthly emissions from Inner Mongolia. The results indicate that Mongolia accounts for only 6% of total emissions (approximately 62 Mt), whereas Inner Mongolia contributes 94% (approximately 1052 Mt).

Further analysis identified four leagues and cities with monthly CO₂ emissions exceeding 100 Mt, namely, Ordos, Hohhot, Hulunbuir, and Baotou. Among these, Ordos displays the highest emissions, reaching approximately 200 Mt. In fact, Ordos produced 7.787 billion tons of coal in 2022, accounting for one-sixth of China’s total coal production, earning it the title of China’s largest coal-producing city. In contrast, coal combustion in Mongolia is primarily concentrated in Ulaanbaatar, with the rest of the country collectively emitting approximately 219 Mt.

2.3.5. Random Forest Model

This study’s random forest model was an ensemble learning method based on decision trees aimed at analyzing the contribution of various driving factors to the spatial distribution and changes of XCO₂ on the Mongolian Plateau. Here, XCO₂ was treated as the target variable, while six driving factors, NDVI, TEMP, PRE, WD, WS, and ODIAC, were treated as feature variables aimed at constructing a regression model. Random forest performs its regression task by training multiple decision trees; each tree’s prediction contributes to the final predicted value of the target variable. During training, each decision tree randomly selects a subset of the training data; at each splitting node, a random subset of features is chosen. Within this subset, the best split point is subsequently selected. The training continues until the maximum depth of a given tree is reached or the number of samples in a leaf node becomes smaller than the predefined minimum sample value. For each input sample, all decision trees provide a prediction; the final predicted value is the average of all the trees’ predictions. The formula used to obtain this value is given below:

$$\widehat{y}={\displaystyle \frac{1}{N}}\sum _{i=1}^N{y}_i$$

(2)

Here, $\widehat{y}$ represents the final predicted value, $y_i$ is the prediction from the ith tree, and N is the total number of decision trees [25].

The random forest model also measures the contribution of each feature to the final prediction by calculating the former’s importance across all trees. A common method for calculating feature importance involves assessing the reduction in mean squared error (MSE) for each tree [26]. Specifically, the contribution of each feature to the reduction in model error at each split point is calculated. Thereafter, the contribution value of each feature is averaged across all decision trees, which results in the final feature importance score being obtained. The formula regarding the final feature importance is given as follows:

$${Importance}_j={\displaystyle \frac{1}{N}}\sum _{i=1}^N({MSE}_{before}-{MSE}_{after})$$

(3)

Here, ${Importance}_j$ represents the final importance of feature $j$, while ${MSE}_{before}$ and ${MSE}_{after}$ represent the MSEs before and after the split (using the same feature), respectively. Thus, the random forest model can not only quantify the impact of each feature on the prediction of XCO₂ variations but also reveal the importance of different driving factors in the spatial distribution and spatiotemporal changes of XCO₂ [27].

3. Results and Discussion

3.1. Analysis of Overall Distribution and Seasonal Evolution Pattern of XCO₂

To analyze the spatial distribution characteristics of XCO₂ in different regions of the Mongolian Plateau, this study performs statistical analysis on the XCO₂ data from 2018 to 2022. As shown in Figure 4, the average XCO₂ concentration in the study area is 412 ppm. XCO₂ exhibits an overall increasing trend, with a total increase of 11.44 ppm, corresponding to an average annual growth rate of 2.29 ppm/a, which is 2.18 higher than the global average growth rate [28]. In different regions, there is a significant difference in XCO₂ between the regions of Inner Mongolia of China and Mongolia, with the spatiotemporal distribution of XCO₂ on the Mongolian Plateau showing higher values in the south and lower values in the north. The regions with higher concentrations are mainly located in the central region of Inner Mongolia, such as the cities of Hohhot and Ordos.

According to the interaction between sources and sinks, the atmospheric XCO₂ concentration in a region exhibits temporal variations. For the Northern Hemisphere, the concentration typically reaches its peak in April and its lowest point in August, with seasonal average values showing consistent patterns. Thus, this study divides the year into four seasons based on months, namely, spring (March–May, MAM), summer (June–August, JJA), autumn (September–November, SON), and winter (December–February, DJF). As shown in this figure, the seasonal averages are as follows: spring (414.83 ppm) > winter (413.4 ppm) > autumn (411.3 ppm) > summer (409.12 ppm). The distribution patterns in spring, winter, and autumn are relatively consistent, with the eastern region showing higher concentrations than the western region (Figure 5a–c). But for summer, the distribution pattern reverses compared to the other three seasons—that is, the eastern region has lower XCO₂ concentrations than the western region (Figure 5d).

3.2. Analysis of the XCO₂ Changing Correlations from the Perspectives of Natural Environments

The Mongolian Plateau, located in a high-latitude inland region, exhibits distinct climatic differences influenced by latitudinal zonality. The southern part of the plateau is relatively warmer, while moisture primarily originates from the Pacific and Arctic Oceans, leading to higher precipitation levels in the eastern and northern regions. This study employs Pearson correlation analysis to examine the linear relationships between these natural factors and XCO₂.

As illustrated in Figure 6, XCO₂ is significantly negatively correlated with the NDVI, precipitation, and temperature across most of the plateau, indicating that areas with a higher NDVI, greater precipitation, and moderate temperatures generally correspond to lower XCO₂ levels. Temporal correlation analysis further reveals that this negative correlation is most pronounced in the eastern regions, where vegetation cover is higher, precipitation is more abundant, and temperatures are moderate. These findings suggest a strong association between XCO₂ variations and natural factors across the Mongolian Plateau.

3.2.1. Collinearity Issue Is Reliable in Mongolian Plateau for Analyzing the XCO₂ Changing Correlations from the Perspectives of Natural Environments

In this study, temporal Pearson correlation coefficients were calculated between XCO₂ and three environmental factors, NDVI, precipitation, and temperature, based on monthly data from 2018 to 2022 (Figure 7). The results indicate that all three factors are significantly negatively correlated with XCO₂ (correlation coefficients of −0.48, −0.46, and −0.33, respectively), suggesting that vegetation activity, precipitation, and temperature may have a suppressing effect on XCO₂. This finding is consistent with previous analyses. However, further temporal correlation analysis among NDVI, PRE, and TEMP revealed strong intercorrelations (correlation coefficients of 0.89, 0.89, and 0.80, respectively), indicating similar temporal variation trends among these environmental factors. Therefore, it is necessary to further evaluate the issue of multicollinearity among these environmental factors.

The VIF values calculated in this study are as follows: NDVI = 4.87, PRE = 6.99, and TEMP = 2.35. While PRE’s VIF slightly exceeds 5, all variables remain below the critical threshold of 10, indicating that collinearity is acceptable.

3.2.2. Analysis of the Effect of Wind Field on XCO₂

The Mongolian Plateau, located in northern Asia, serves as a crucial pathway for atmospheric circulation. It is primarily influenced by the westerly wind belt and the monsoon system, both of which play key roles in the transport of regional air masses. This study utilizes wind field data from all twelve months of 2018 to analyze the characteristics of wind speed and direction in the region, and to explore their potential impact on the spatial distribution of XCO₂. As shown in Figure 8, the annual mean wind speed over the plateau is 1.63 m/s. Higher wind speeds, exceeding 2 m/s, are observed from November to February of the following year, mainly concentrated in the central and eastern parts of the region. From March onward, wind speeds gradually decrease, reaching their lowest point in August. Southerly winds dominate for most of the year, while winds in the east–west direction show no clear pattern.

We employed the Pearson correlation coefficient to analyze the impact of wind direction and wind speed on XCO₂ concentrations over the Mongolian Plateau. As shown in Figure 9, considering that wind direction is a circular rather than a linear variable, the wind direction and speed data were converted into two components, U (zonal wind) and V (meridional wind), for analysis. The results indicate that the average correlation coefficient between XCO₂ and the U component is 0.32, showing an overall positive correlation. The figure illustrates that only a small portion of the western region exhibits a negative correlation, suggesting that an increase in westerly winds tends to be associated with a rise in regional XCO₂ concentrations. The average correlation coefficient between the V component and XCO₂ is −0.23, indicating that stronger northerly winds are also associated with increased XCO₂ concentrations. Conversely, enhanced southerly winds correspond to a decrease in XCO₂ levels. This pattern is particularly evident in the southern and eastern parts of the Mongolian Plateau.

3.3. Analysis of the Contributions of Single and Interactive Factors to XCO₂ Changes

3.3.1. Analysis of Contribution Rates of Different Influencing Factors on XCO₂ Changes

The contribution rates of different influencing factors on XCO₂ changes were analyzed using a random forest model. To assess the model’s prediction performance, we calculated MSE and RMSE. The calculation results indicated that the model’s MSE was 18.8 and its RMSE was 4.3, suggesting that the average prediction error was 4.2 ppm, accounting for approximately 1% of the plateau’s total XCO₂. The random forest model effectively captured the variation trends of XCO₂ on the plateau and demonstrated high accuracy. Figure 10 illustrates the contribution rates of the primary driving factors to the spatiotemporal variation of XCO₂ on the Mongolian Plateau. Evidently, the influence of each factor on the plateau’s XCO₂ varied significantly, indicating that changes in XCO₂ on the Mongolian Plateau were driven by multiple environmental and anthropogenic factors. Among these, the NDVI, as the primary carbon sink indicator, was deemed the most important, with a contribution rate of 0.35, underscoring the key role of vegetation in carbon absorption. Fossil fuel combustion emissions, with a contribution rate of 0.2, revealed a significant impact of human activities on CO₂ concentrations on the plateau, particularly in areas with intensive industrial activities. Additionally, WD and WS each had an importance value that exceeded 0.1, revealing their moderate influence on XCO₂ variation in the plateau region. In contrast, PRE and TEMP had a low contribution rate. This reflected the complexity of XCO₂ distribution pattern on the Mongolian Plateau. Therefore, vegetation and fossil fuel combustion emissions were the two key factors that predominantly influenced the spatiotemporal variation of XCO₂ on the plateau, while other factors influenced XCO₂ distribution and concentration to varying degrees.

3.3.2. Analysis of Contribution Rate of Interactive Factors on XCO₂ Changes

This study employs a random forest model to assess the interactive effects of six environmental factors on XCO₂ concentrations over the Mongolian Plateau. As shown in Figure 11, the NDVI and ODIAC emerge as the most significant interaction terms, with a contribution rate of over 0.25, indicating that the coupling between vegetation and fossil fuel emissions plays the most important role in the regional carbon cycle. Additionally, WS and WD, ODIAC and WS, and NDVI and WS also exhibit a high contribution rate of over 0.05, suggesting that wind direction, by influencing spatial transport patterns, interacts with other factors to determine atmospheric XCO₂ distribution. Further analysis reveals that climate-related interactions, such as NDVI and PRE, TEMP and PRE, and TEMP and WD, are also significant, reflecting the combined effects of environmental factors on carbon exchange. These findings further underscore the critical role of vegetation in regulating XCO₂ variations over the Mongolian Plateau and highlight its sensitivity to anthropogenic emission changes.

4. Conclusions

This study analyzed the spatial–temporal characteristics of XCO₂ over the Mongolian Plateau from 2018 to 2022 and investigated the key driving factors influencing its variations. The main conclusions are as follows:

(1)

For overall distribution and seasonal changes of XCO₂, the average XCO₂ concentration is 412 ppm, with an annual growth rate of 2.29 ppm/a from 2018 to 2022. In different regions, XCO₂ concentrations exhibit significant spatial heterogeneity, with higher values in the southern regions and lower values in the north. The regions with higher concentrations are primarily located in the central region of Inner Mongolia, particularly in cities such as Hohhot and Ordos. Meanwhile, the seasonal analysis reveals a clear temporal pattern, with the following order: spring (414.83 ppm) > winter (413.4 ppm) > autumn (411.3 ppm) > summer (409.12 ppm). The XCO₂ concentration patterns in spring, winter, and autumn are relatively consistent, with higher values in eastern regions and lower in western regions, but the pattern is the opposite in summer.

(2)

From the perspectives of natural environments on XCO₂ changes, XCO₂ is significantly negatively correlated with the NDVI, precipitation, and temperature across most of the plateau, indicating that areas with a higher NDVI, greater precipitation, and moderate temperatures generally correspond to lower XCO₂ levels. Temporal correlation analysis further reveals that this negative correlation is most pronounced in the eastern regions, where vegetation cover is higher, precipitation is more abundant, and temperatures are moderate.

(3)

As for the contributions of single and interactive factors to XCO₂ changes, evidently, the influence of each factor on the plateau’s XCO₂ varied. The NDVI, as the primary carbon sink indicator, was deemed the most important, with a contribution rate of 0.35, followed by fossil fuel combustion emissions, WD, and WS, but PRE and TEMP displayed a low contribution rate. Meanwhile, for interactive factors, the NDVI and ODIAC showed the highest contribution rate (over 0.25). WS and WD, ODIAC and WS, and NDVI and WS also exhibited a high contribution rate (over 0.05), while the contributions of other factors were lower.