Regional Atmospheric CO₂ Response to Ecosystem CO₂ Budgets in China

Abstract

The distribution of atmospheric CO₂ is not homogenous, primarily due to variations in the CO₂ budgets of regional terrestrial ecosystems. To formulate a comprehensive strategy to combat the increasing global CO₂ levels and associated warming, it is crucial to consider both the distribution of atmospheric CO₂ and the CO₂ budgets of ecosystems. This study focused on analyzing the relationship between regional atmospheric CO₂ and CO₂ budgets in China from 2010 to 2017. Initially, a robust estimation model of net ecosystem CO₂ exchange was developed to calculate CO₂ budgets using collected emission data. Subsequently, Pearson correlation, redundancy analysis, and geographically weighted regression techniques were employed to examine the link between atmospheric CO₂ levels, CO₂ budgets, and other meteorological factors at various spatial and temporal scales. The findings from the redundancy analysis and geographically weighted regression indicated that the atmospheric CO₂ content of each province could not be solely determined by the regional CO₂ budgets. However, a significant and positive correlation between atmospheric CO₂ levels and CO₂ budgets was observed in non-coastal provinces for a period of six months (R² ranging from 0.46 to 0.83). Consequently, it is essential to promote a balance between CO₂ emissions and the CO₂ uptake capacity of regional ecosystems. This balance would minimize positive CO₂ budgets and effectively mitigate the increase in atmospheric CO₂ levels.

1. Introduction

In order to combat global warming and fulfill the goals outlined in the Paris Agreement, several countries, including China, have announced their commitment to carbon neutrality or net-zero carbon dioxide emissions [1]. The Chinese government has set targets to peak carbon emissions by 2030 and achieve carbon neutrality by 2060. One of the policy measures employed is the carbon emission permit allocation and trade system, which aims to gradually control and reduce carbon emissions from enterprises. In designing such an allocation strategy, various factors such as economic growth, industrial structure adjustment, energy structure optimization, and coordinated control of other air pollutant emissions are commonly considered. Currently, carbon emission permits are primarily allocated among provinces in China based on emission generation and transfer data, as well as local economic levels, to ensure fairness and efficiency [2,3,4]. However, for a more comprehensive allocation strategy, it is essential to consider ecosystem factors related to carbon emissions and their mitigation.

Carbon emissions encompass the total CO₂ equivalent of all greenhouse gases [5]. According to the Intergovernmental Panel on Climate Change (IPCC) report in 2022, CO₂ accounted for 75% of global greenhouse gas emissions, followed by methane (18%), nitrous oxide (4%), and fluorinated gases (2%) in 2019. Different terrestrial and aquatic ecosystems can counteract carbon emissions by sequestering carbon, primarily in the form of CO₂, through the photosynthesis of plants and algae [6,7]. On the other hand, the depletion of methane, nitrous oxide, and fluorinated gases mainly occurs through physiochemical reactions (e.g., oxidation, photolysis, reaction with chlorine, and precipitation) in the atmosphere [8,9,10,11], rather than being influenced by land and aquatic ecosystems. While soil can also act as a sink for nitrous oxides through microorganism activities, its overall importance at a global scale is considered to be minimal [12]. Hence, the carbon sequestration function of different ecosystems primarily affects variations in atmospheric CO₂ concentrations.

Despite the cycling of CO₂ in the atmosphere, its global distribution is not uniform. By examining the global patterns of carbon dioxide in the mid-troposphere observed by NASA from 1 May to 31 May 2013 (https://climate.nasa.gov/vital-signs/carbon-dioxide/, accessed on 1 December 2022), it was evident that high concentrations of CO₂ were found in the Northern Hemisphere, while lower concentrations were observed in the Southern Hemisphere. This disparity was mainly attributed to the limited CO₂ uptake by plants in the Northern Hemisphere during that period. In China, Fu et al. (2018) [13] found that mid-tropospheric CO₂ concentrations were higher in northern China compared with southern China, with four high-concentration centers located in the southwest of northeast China, west Inner Mongolia, east and west Xinjiang, and lower concentrations observed in Yunnan and the Tibetan area. This distribution is also closely related to vegetation’s capacity for CO₂ absorption. Additionally, for a specific small area, Zhang et al. (2022) [14] discovered that changes in CO₂ source and sink characteristics jointly contributed to a decrease in atmospheric CO₂ concentration over three years in the Nanling area of China, as determined through in situ atmospheric CO₂ measurements that excluded the impact of weather conditions. These findings raise an intriguing question: if the distribution of carbon emissions or carbon emission permit allocation is not balanced with the distribution of ecosystem carbon absorption capacity, could regional CO₂ distribution become more uneven?

Answering this question requires the quantification of ecosystem carbon absorption capacity and determining the extent to which regional atmospheric CO₂ concentrations are sensitive to regional CO₂ budgets. In our previous study, we developed a robust estimation model for net ecosystem CO₂ exchange (NEE) to determine NEE values for different regions in China [15]. Building upon this, and by collecting atmospheric CO₂ concentration and CO₂ emission data while determining NEE values for various regions, this study aims to analyze the sensitivity of regional atmospheric CO₂ concentrations to regional CO₂ budgets at different spatial and temporal scales through correlation and regression analyses. Furthermore, we have employed redundancy analysis (RDA) to compare the contributions of regional CO₂ budgets and climate factors to variations in atmospheric CO₂. The insights gained from this study will shed light on the extent to which regional CO₂ emissions can impact atmospheric CO₂ levels in China, and whether the uneven distribution of CO₂ poses a potential risk to regional ecosystems.

2. Materials and Methods

2.1. Data Collection

The data collection process for NEE estimation was extensively described in our previous study [15].

To calculate the regional CO₂ budget, monthly CO₂ emission data (comprising emissions from power generation, industry, residential sources, transportation, and agriculture) for each province in China from 2008 to 2017 were collected from the Multi-resolution Emission Inventory for China (MEIC) database, maintained by the Department of Earth System Science at Tsinghua University (version v2.0). Additionally, monthly atmospheric CO₂ concentrations spanning the period of 2010 to 2018 were obtained from the AIRS3C2M database, hosted by the Goddard Earth Sciences Data and Information Services Center (DOI: 10.5067/Aqua/AIRS/DATA336).

Climate parameters that have the potential to influence the movement of atmospheric CO₂ were also gathered. These included the eastward components of the 100 m and 10 m winds (u100 and u10), the northward components of the 100 m and 10 m winds (v100 and v10), east turbulence surface stress (mmtss), north turbulence surface stress (mntss), forecast surface roughness (fsr), convective available potential energy (cape), boundary layer height and dissipation (blh and bld), and the angle of sub-grid scale orography (anor). The climate parameter data were obtained from the ERA5 products provided by the Copernicus Climate Change Service (C3S) Climate Data Store (CDS) (DOI: 10.24381/cds.f17050d7). The monthly data for each grid were used to calculate the average values for each province in China over six-month periods.

Considering the variability of wind directions, which can change significantly or even reverse within a matter of days, we utilized the average values of u100, v100, u10, and v10 to estimate the average wind strength (w100 and w10) in each province over six-month intervals. These values represented the overall average horizontal movements of atmospheric CO₂. Similarly, the average mmtss and mntss values were employed to determine the average mean turbulence surface stress (mtss), disregarding orientation. The calculation of mtss was carried out as follows:

$${\stackrel{-}{w100}}_{6 months}=\sqrt{{\stackrel{-}{u100}}_{6 months}^2+{\stackrel{-}{v100}}_{6 months}^2}$$

(1)

$${\stackrel{-}{w10}}_{6 months}=\sqrt{{\stackrel{-}{u10}}_{6 months}^2+{\stackrel{-}{v10}}_{6 months}^2}$$

(2)

$${\stackrel{-}{mtss}}_{6 months}=\sqrt{{\stackrel{-}{mmtss}}_{6 months}^2+{\stackrel{-}{mntss}}_{6 months}^2}$$

(3)

2.2. NEE Estimation Model Construction

The estimation of regional NEE was carried out using the Random Forest model, with the decision tree number (N) set to 100 and the minimum leaf size (M) set to 5. The dataset, consisting of 16,920 collected data points, was randomly divided into 14,186 training samples and 2734 validation samples. The randomly selected training samples were then utilized in a supervised learning process conducted in Matlab. The model’s performance was assessed by comparing the calculated NEE values with the observed values from 1000 validation samples. The goodness of correspondence between the calculated and observed values was evaluated using the coefficient of determination (R²) and the root mean square errors (RMSE), computed as follows:

$$RMSE=\sqrt{\frac{{\sum }_{i=1}^n{(y_{cal,i}-y_{obs,i})}^2}{n}}$$

(4)

where y_cal,i and y_obs,i are the calculated and observed NEE values and n represents the number of observed-calculated NEE pairs.

2.3. Sensitivity Analysis of Regional Atmospheric CO₂ to Regional CO₂ Budget

The sensitivity of regional atmospheric CO₂ to regional CO₂ budgets was examined through Pearson correlation analysis, which involved assessing the relationship between variations in atmospheric CO₂ concentrations and CO₂ budgets for each province in China. It is important to note that the correlation analysis was limited to the available data from the period of 2010–2017. The changes in atmospheric CO₂ concentrations for each province were calculated on a monthly basis, as well as for two-month, four-month, six-month, eight-month, ten-month, and twelve-month intervals. The CO₂ budgets were determined by summing the CO₂ emissions and regional NEEs. However, it should be emphasized that the calculation of CO₂ budgets did not take into account CO₂ cycling, resulting in what can be referred to as theoretical static CO2 budgets. The Pearson correlation analysis of regional atmospheric CO₂ and regional CO₂ budgets was conducted using Proc Corr in SAS 9.4 (SAS Institute, Madison, WI, USA).

To analyze the multiple correlations between regional CO₂ budgets, climate factors, and regional atmospheric CO₂ and its variations, redundancy analysis (RDA) was performed. The RDA was implemented using the “vegan” package in R (version 4.2.2) [16].

Furthermore, geographically weighted regression was employed to investigate the relationship between regional atmospheric CO₂ variations and regional CO₂ budgets, considering spatial relationships and the non-stationarity of CO₂ budgets across different provinces [17]. The geographically weighted regression was conducted using MATLAB software, analyzing the regional atmospheric CO₂ variations and regional CO₂ budgets for each province over six-month intervals.

The entire process of data collection, model construction, and data analysis is summarized in the flowchart depicted in Figure A1.

3. Results and Discussion

3.1. Regional Terrestrial NEE, CO₂ Emissions, and Atmospheric CO₂ Content

The estimated net ecosystem CO₂ exchange (NEE) values for the 31 provinces in China are presented in Figure 1a and Figure A3a. The NEE values for each province exhibited periodic fluctuations, with the greatest negative values (indicating the highest carbon uptake capacity from the atmosphere) occurring during the summer season. Provinces such as Xinjiang, Qinghai, Sichuan, Yunnan, Guangxi, and Tibet displayed particularly large negative NEE values, partially attributed to their expansive land areas. Monthly NEE estimations revealed that most provinces in China exhibited negative annual NEE values, except for Tianjin, Shanghai, and Jiangsu, which had positive values (averaging 0.57, 0.73, and 10.55 Mt CO₂ yr⁻¹, respectively), indicating a tendency for their terrestrial ecosystems to release CO₂ into the atmosphere. These NEE values were derived from the robust NEE model, which demonstrated substantial performance in estimating NEE for various land types, including arable lands (R² = 0.63), forests (R² = 0.75), and grasslands (R² = 0.75). However, the model performed less effectively for smaller land features such as water bodies, ice, tundra, and urban areas (R² = 0.46), owing to significant variations in carbon absorption capacity across different land types (Figure A2). A similar NEE estimation using the Random Forest model was conducted by Huang et al. (2021) [18], which exhibited good performance for various forest, grassland, wetland, and cropland types (R² ranging from 0.57 to 0.91). In a previous study, we estimated the annual NEE for China to be approximately −1130 Mt C yr⁻¹ using a different database and a different Random Forest model [15]. This estimate aligns with the NEE estimation in the current study (averaging −4219 Mt CO₂ yr⁻¹, equivalent to −1151 Mt C yr⁻¹), further affirming the robustness of our NEE estimation.

Analysis of the Multi-resolution Emission Inventory for China (MEIC) data revealed that CO₂ emissions across all provinces displayed periodic variations, with higher emissions occurring primarily during the winter season (Figure 1b and Figure A3b). Among the provinces, Shandong, Jiangsu, and Hebei consistently ranked as the top three emitters of CO₂. The periodic fluctuations in both net ecosystem CO₂ exchange (NEE) and CO₂ emissions resulted in corresponding variations in regional CO₂ budgets for each province (Figure 2). Specifically, Shandong, Jiangsu, and Hebei had the highest positive CO₂ budgets (80.47, 60.43, and 59.60 Mt CO₂ per month, respectively), while Tibet, Yunnan, and Qinghai had the highest negative CO₂ budgets (−48.28, −21.73, and −19.66 Mt CO₂ per month, respectively) (Figure 2a and Figure A3c). When examining CO₂ budget density, Shanghai exhibited significantly higher CO₂ budget levels (2515.08 t CO₂ km⁻² per month) compared with other provinces, followed by Tianjin (1155.57 t CO₂ km⁻² per month) (Figure 2b and Figure A3d). As most provinces demonstrated positive CO₂ budgets, atmospheric CO₂ concentrations displayed an increasing trend with periodic variations each year (Figure 3 and Figure A3e). Notably, provinces such as Tibet, Yunnan, and Qinghai, which exhibited negative CO₂ budgets, generally displayed lower atmospheric CO₂ concentrations. However, these provinces still showed an upward trend in atmospheric CO₂ levels over time. This observation suggests that the long-term effects of atmospheric CO₂ cycling can contribute to the homogenization of CO₂ concentrations, thereby influencing the heterogeneity of regional atmospheric CO₂ sensitivity to regional CO₂ budgets.

3.2. Regional Terrestrial NEE, CO₂ Emissions, and Atmospheric CO₂ Content

Pearson correlation analysis was conducted to examine the relationship between regional CO₂ budgets and changes in regional atmospheric CO₂ concentrations for various time intervals, ranging from monthly to twelve months, during the period of 2010–2017. It was anticipated that positive CO₂ budgets would correspond to an increase in atmospheric CO₂ concentrations. As such, the Pearson correlation coefficients were predominantly positive across most provinces, with the exception of Guangdong (Figure 4). Over time, a similar trend emerged among the provinces, with the Pearson coefficients initially increasing with longer time intervals and subsequently decreasing. Generally, the highest Pearson coefficients were observed with a six-month interval. However, in the coastal provinces of Fujian, Shanghai, Hainan, and Guangdong, weak correlations were found between changes in atmospheric CO₂ concentrations and CO₂ budgets, with corresponding Pearson coefficients of 0.242, 0.275, 0.117, and −0.268, respectively, for the six-month interval. It is worth noting that these four provinces are all coastal regions. Additionally, the coastal provinces of Zhejiang and Jiangsu exhibited relatively lower Pearson coefficients for the six-month interval (0.543 and 0.539, respectively) compared with the other provinces, which ranged between 0.60 and 0.80. These findings suggest that the interactions between CO₂ budgets and CO₂ transportation in coastal areas may influence atmospheric CO₂ dynamics. In the short term, atmospheric CO₂ transportation can disperse emitted CO₂ and increase atmospheric CO₂ concentrations. However, over a longer time frame, atmospheric CO₂ transportation tends to homogenize regional atmospheric CO₂ concentrations with those of other regions, thereby mitigating the impact of CO₂ budgets on atmospheric CO₂ levels.

While CO₂ budgets played a significant role in explaining the variations in regional atmospheric CO₂ concentrations, they alone were insufficient to fully account for these variations. Analysis of the collected data revealed that the atmospheric CO₂ variations across all provinces did not exceed 10 ppm over a six-month period. However, the CO₂ budgets for the same period in the 31 provinces resulted in atmospheric CO₂ variations ranging from −17.32 ppm (Yunnan) to 768.12 ppm (Shanghai), with an average value of 78.24 ppm. These values take into consideration the accumulation of most CO₂ within the troposphere, at altitudes ranging from 10 to 16 km, in China. Consequently, it is evident that other factors contribute significantly to the observed atmospheric CO₂ variations.

Numerous studies have investigated the factors influencing regional atmospheric CO₂ concentrations. For instance, Zhou et al. (2022) [19] identified monthly mean daily maximum global radiation, monthly effective accumulated temperature, monthly mean daily maximum vapor pressure deficit, and monthly precipitation as the key meteorological variables influencing atmospheric CO₂ in forest systems. Yang et al. (2020) [20] found that soil temperature, air temperature, photosynthetically active radiation (PAR), below-canopy CO₂ concentration, vapor pressure deficit, and soil water content at 50 cm were the main meteorological factors influencing CO₂ exchange on daily and monthly time scales. Other meteorological factors, such as wind patterns, can also play a crucial role in atmospheric CO₂ dynamics by affecting its transport [21,22]. Although multiple factors can influence atmospheric CO₂, they can be broadly categorized as CO₂ budgets and CO₂ transport within the atmosphere.

In the previous section, we developed an NEE estimation model that incorporated potential factors affecting atmospheric CO₂, such as surface temperature, soil temperature, surface net solar radiation, precipitation, evaporation, soil water content, NDVI index, EVI index, canopy height, and forest age, to calculate CO₂ budgets. To explore the interactive effects of CO₂ budgets and CO₂ transport, including winds, surface turbulence stress, forecast surface roughness, convective available potential energy, boundary layer height, boundary layer dissipation, and sub-grid scale orography, on atmospheric CO₂ variations over a six-month period, a redundancy analysis (RDA) was conducted.

RDA biplots were generated to illustrate the contributions of the considered factors to the variations in atmospheric CO₂ (deltaCO2) and atmospheric CO₂ content (CO2). The projection of each factor’s arrow onto the arrow of the focused variables indicated the direction and intensity of its effect. For example, Figure 5 depicts that, in the case of Beijing, the factors of budgetCO2, w10, w100, mtss, bld, and fsr all positively contributed to deltaCO2, while only parameters blh and cape had negative contributions. Conversely, in Tianjin, only budgetCO2 had a positive contribution to deltaCO2, while the other factors had negative contributions. Furthermore, for both Beijing and Tianjin, the considered factors appeared to have minimal effects on atmospheric CO₂ content (CO2) since their arrows were nearly orthogonal to the CO2 arrow. RDA biplots for the other 29 provinces can be found in Figure A4. The RDA results indicated that CO₂ budgets and climatic transport parameters directly influenced the variation in atmospheric CO₂ (deltaCO2), but not atmospheric CO₂ content (CO2) (Figure A3). In most cases, deltaCO2 showed a strong positive correlation with CO₂ budgets, except for the coastal provinces of Shanghai, Fujian, Guangdong, and Hainan, which aligned with the results of the correlation analysis. Additionally, deltaCO2 tended to be predominantly influenced by parameters such as fsr (21/31 provinces), blh (26/31 provinces), and cape (31/31 provinces), resulting in negative effects. The boundary layer, where emitted pollutants mix [23,24], is likely to have a similar effect on emitted CO₂. Thus, a higher boundary layer favors the diffusion of emitted CO₂ to the free atmospheric layer, reducing local CO₂ accumulation. Similarly, convective available potential energy (cape), which represents the integrated work that positive buoyancy forces would perform on air parcels rising vertically through the atmosphere, inhibits the uplift of pollutants when it is negative [25]. Consequently, higher cape values contributed to the atmospheric diffusion of emitted CO₂. In comparison to other parameters, mtss and bld had relatively less impact on deltaCO2.

Furthermore, the parameters of boundary layer height (blh) and convective available potential energy (cape) primarily affected the vertical movements of atmospheric CO₂, whereas its horizontal movement was driven by winds. The RDA analysis also revealed that wind strength at 100 m and 10 m heights (w100 and w10) were significant factors influencing deltaCO2. However, their relationship with deltaCO2 varied, with both positive and negative associations observed.

For the provinces of Tianjin, Liaoning, Shanghai, Jiangsu, Shandong, Henan, Guizhou, Shaanxi, and Xinjiang, w100 and w10 exerted a negative influence on deltaCO2. As w100 and w10 represent cumulative winds for each province, this suggests that winds tend to transport CO₂ out of these provinces. It is important to note that this is not necessarily due to higher atmospheric CO₂ contents in these provinces (Figure A5a), but rather their CO₂ budgets. Some of these provinces exhibited the highest annual CO₂ budgets in China, such as Shanghai, Tianjin, Jiangsu, and Shandong, with values of 20,120.66, 9244.60, 5023.94, and 4185.66 t CO₂ km⁻², respectively, ranking among the top four provinces (calculated from Figure 2b). Additionally, other provinces showed significantly higher annual CO₂ budgets compared with their neighboring provinces, such as Liaoning and Henan (Figure A5b). These findings indicate that the atmospheric CO₂ variations in a particular province are influenced not only by its own CO₂ budget but also by the budgets of other provinces.

Geographically weighted regression results demonstrated a weak linear relationship between regional atmospheric CO₂ variations and their own CO₂ budgets over a six-month period, as indicated by several negative slope factors (Figure 6a) and R² values below 0.4 (Figure 6b).

In summary, the contents of atmospheric CO₂ were influenced by a combination of CO₂ budgets and atmospheric transportation. While our findings indicate that regional CO₂ budgets alone may not fully account for the variations observed in regional atmospheric CO₂, they still exert significant impacts and demonstrate meaningful positive correlations with atmospheric CO₂ variations over a relatively short-term period of six months.

3.3. Emission Allocation Policies Related to CO₂ Budgets

Unlike the findings of previous studies, such as Zhang et al. (2022) [14], which suggested that changes in CO₂ source and sink characteristics played a joint role in decreasing atmospheric CO₂ concentration over a three-year period in the Nanling area of China, our results indicate that regional CO₂ budgets alone do not exert dominant influence on altering regional atmospheric CO₂ at the provincial scale. One possible explanation for this disparity could be the presence of point sources emitting CO₂, whereby the increase in atmospheric CO₂ resulting from positive CO₂ budgets might be more pronounced in specific localized areas within each province. Therefore, it may be necessary to analyze the relationship between CO₂ budgets and atmospheric CO₂ at a smaller scale, focusing on sub-regions or localized areas. To support this notion, we conducted a similar correlation analysis by aggregating data from provinces into larger regions of China, namely northeast China, north China, central China, south China, east China, southwest China, and northwest China. The results revealed that the correlation between CO₂ budgets and changes in atmospheric CO₂ content weakened at the regional and national scales compared with the provincial level (Figure A6). Consequently, regional CO₂ budgets may have a more significant impact on atmospheric CO₂ dynamics in smaller areas, contributing to both positive and negative effects on specific microcosms such as temperature increase [26], growth pressure on vegetation [27], and CO₂ fertilization effects [28].

Furthermore, the regional sensitivity of atmospheric CO₂ to CO₂ budgets can lead to an imbalanced distribution of atmospheric CO₂ and variations in warming patterns. This imbalance can potentially lead to overestimation or underestimation of the costs associated with warming in specific areas, making it challenging to develop comprehensive carbon emission allocation policies that ensure fairness and efficiency. In our view, an optimized emission allocation strategy should aim to achieve a balance between carbon emissions and uptake in each region, such as provinces or even cities, to establish an acceptable CO₂ budget. This approach would enable effective control of regional atmospheric CO₂ variations and minimize potential damages to regional ecosystems. To implement such a carbon allocation strategy, detailed information on regional emissions, the CO₂ uptake capacity of ecosystems, and CO₂ transport flux needs to be obtained. Numerous efforts have already been made to estimate gridded CO₂ emissions [29,30], carbon budgets of terrestrial ecosystems [31,32], CO₂ diffusion flux [33], as well as advancements in techniques for processing remote sensing air data [34,35]. Encouraging further research in these areas is crucial for gaining a comprehensive understanding of regional carbon dynamics.

4. Conclusions

This study demonstrates that CO₂ budgets have a considerable influence, though not a dominant one, on the alteration of atmospheric CO₂ over a six-month period in most provinces of China. Moreover, the findings highlight the potential effects of different regional CO₂ budgets on the uneven distribution of atmospheric CO₂, particularly at smaller scales below the provincial level. Therefore, it is essential not to overlook these effects and instead promote further monitoring research on gridded CO₂ emissions and uptake, as well as studies focusing on regional CO₂ variations and their impact on local ecosystem functions. These endeavors will contribute to the development of a more comprehensive CO₂ emission allocation strategy that takes into account both economic and ecological factors related to CO₂ generation and uptake. By doing so, potential ecological risks can be mitigated, ensuring a balanced approach to carbon management.