Assessing the Spatial Distribution of Carbon Emissions and Influencing Factors in the Yellow River Basin

Abstract

The spatial distribution and trend of carbon emissions in the Yellow River Basin—an important ecological barrier and economic belt in China—directly affect the stability of the ecosystem and the sustainable development of the regional economy. Based on the data for carbon emissions in China’s counties from 1997 to 2017, this paper utilizes standard deviation ellipses, Theil index nested decomposition, and geographic detector models to make a comprehensive description of the spatial and temporal distribution and dynamic evolution characteristics of carbon emissions in the Yellow River Basin. Factors influencing carbon emissions are also analyzed from multiple dimensions. According to the findings, (1) carbon emissions at the county level show a clear upward trend without reaching a peak, exhibiting a spatial distribution of higher emissions in the east and lower in the west and higher in the south and lower in the north, with the mid-lower reaches being the center. The junction of the Shandong, Shaanxi, and Gansu provinces further exhibits a significant expansion, forming two core areas of carbon emissions. (2) Carbon emissions at the county level in the Yellow River Basin are influenced by both economic and geographic factors, exhibiting a significant high carbon spillover effect and a low carbon lock-in effect. The gravity center of the distribution has shifted towards the mid-lower reaches, with the upper reaches displaying dispersion tendencies. (3) Intra-regional disparities are the main source of the overall spatial differences in carbon emissions, with the largest disparities being observed in the upper reaches, followed by the middle reaches, and the smallest disparities being observed in the lower reaches. Further analysis shows that the level of economic development is the primary factor influencing the spatial variation of carbon emissions, and the combined effects of population size and industrial agglomeration are the key drivers of the annual growth in carbon emissions.

1. Introduction

Global warming has become a significant issue threatening human beings’ survival and sustainable development. According to data from the National Energy Administration and the International Energy Agency, China’s carbon emissions in 2023 amounted to 12.6 billion tons, making it the country with the highest carbon emissions in the world. To address global climate change, China has actively embraced the Paris Agreement by setting a “dual-carbon” target. Under the dual pressures of international responsibility and domestic development transformation, regions within China are tasked with differentiated emission reduction objectives [1]. Among them, the Yellow River Basin (YRB) stands out as an important energy supplying base for China. However, it is facing an evident dilemma between the comprehensive management of the basin and economic development [2]. The traditional development model has placed the YRB in a dilemma where environmental protection conflicts with economic and social progress. Recognizing this challenge, the Chinese government has proposed promoting high-quality development in the YRB. Therefore, it is significant to investigate and monitor the evolution of carbon emissions in the YRB in a scientific manner and analyze distinctive characteristics in its spatial structure of carbon emissions. This article aims to explore carbon emissions in the YRB within a sustainability framework by using a spatial–temporal distribution model, providing relevant support for promoting innovation in low-carbon technologies and enhancing carbon emission management capabilities.

As the significance of carbon reduction becomes more prominent, carbon emissions have become one of the most important topics. Generally speaking, existing research has primarily focused on three main areas of discussion. The first one is to quantify carbon emissions within a given region. In terms of research methodology, previous scholars proposed measurement indicators from different perspectives [3], employing a “single factor” approach to measure them. As the research deepened, scholars integrated relevant factors, such as fixed capital, labor, and GDP, and embraced the data envelopment analysis measure that reflects a total factor philosophy as the mainstream to assess carbon emission efficiency [4,5]. In addition, in terms of research methodology, studies have been conducted to evaluate the scale and efficiency of carbon emissions, mainly through metrics such as the Thiel index [6] and the Dagum Gini coefficient, etc. In terms of data, advancements in geo-observation technology have enabled scholars to measure carbon emissions from a geospatial perspective. Considering the significant correlation between nighttime lighting and socioeconomic activities, Shi et al. used DMSP-OLS data to analyze China’s carbon situation of multiple scales [7]. Additionally, Meng et al. [8] and Cui et al. estimated carbon emissions in specific regions of China based on NPP-VIIRS nighttime lighting data [9], which provide a basis for developing models for carbon emissions.

The second area of discussion within existing research is the exploration of the spatiotemporal evolution characteristics of carbon emissions. This research measures carbon emissions using methods such as the Theil index [10]. On the basis of this, subsequent research analyzed the spillover effects of China’s carbon emissions through global and local spatial correlation [11]. For instance, Zhang et al. [12], Nie et al. [13], and Zhang et al. [14] employed spatial econometric models to calculate Moran’s I index and examined the spatial correlation of carbon emissions among different regions. They concluded that the eastern coastal areas are more capable of achieving the expected carbon emission reduction goals, while the less developed central and western regions have been less effective in this cause. The study conducted by Wang et al. [15] employs the Gini coefficient along with the method of dynamic distribution (MEDD) incorporating kernel density estimation to examine the dynamic evolution of carbon emissions in central, eastern, and western China. Nie and Lee investigated the convergence of regional differences [16]. Zhang et al. [17] found significant differences in carbon emissions across eight economic regions of China. Additionally, recent studies have employed methods such as the standard deviation ellipse of gravity to explore carbon emissions [18], providing methodological support for this paper.

Thirdly, the factors affecting carbon emissions have been explored from various perspectives by scholars. These studies examined economic and industrial aspects [19], energy intensity [20], land use [21,22], and urbanization [18] as important factors influencing carbon emissions. Vaninsky employed the generalized dyadic index decomposition (GMDI) and Monte Carlo models to analyze variations in factor contributions during the carbon peaking process [23]. Although many scholars have focused on exploring these factors, less attention has been paid to their combined effects, which presents an opportunity for further investigation in this paper.

Especially for the YRB, the current literature mainly focuses on the sustainable utilization of water resources [24]. The few studies available only provide a rough depiction of the spatial spillover effects [25,26], with a greater emphasis being placed on the analysis of energy-rich areas [27]. Furthermore, there is insufficient research on carbon emissions from the perspective of the watershed scale, and works analyzing their temporal and spatial evolution from a county or regional perspective within the YRB are even fewer. Moreover, the existing literature has only explored individual factors influencing carbon emissions [28], failing to consider that these factors not only have individual effects but also affect spatial distribution through superimposed effects. This aspect is evidently underemphasized in previous studies. Currently in China, it is necessary to systematically study the spatial distribution characteristics of carbon emissions and the superposition effect of influencing factors at the basin scale so as to clarify the effective approaches and policy systems for reducing carbon emissions in the YRB. This will enable the analysis of key bottlenecks and breakthrough points in promoting low-carbon transition in the YRB while also providing a sustainable pathway for the region’s economic growth. This paper aims to examine carbon emissions, covering a time span from 1997–2017, with two types of nighttime lighting data. We employ methods such as standard ellipse and three-stage nested decomposition models to analyze our sample. In addition, this paper explores the drivers of carbon emissions in brochures from five aspects: economic development, industrial agglomeration, population size, urbanization, and government support, providing an in-depth analysis of how these different factors interact with each other. Based on this, this paper proposes countermeasures in terms of overall basin management as well as local area interventions that take both environmental benefits and economic gains into consideration. These empirical evidence-based recommendations serve as theoretical references for formulating scientifically sound policies.

The potential innovation points of our study are as follows: (1) This paper examines the spatiotemporal patterns and evolutionary trends of the entire YRB from a county-level perspective. By applying methods such as kernel density estimation, it provides a detailed analysis of the entire basin and different sections of the river, systematically revealing the spatiotemporal patterns and dynamic evolution of carbon emissions in the YRB. This not only provides insights for understanding the growth and spatial expansion of carbon emissions from a county-level perspective but also complements the limitations of existing studies, which predominantly focus on provincial or municipal scales. (2) This paper adheres to the real-world situation and theoretical connotations of carbon emissions in the YRB. It employs kernel density estimation to examine the spatiotemporal dynamics of carbon emissions, identifying significant high carbon spillover effects and low carbon lock-in effects in the basin. This rigorous methodological foundation enhances the accuracy of carbon emission estimation and analysis for the YRB and provides intuitive, empirical evidence for comprehensively understanding and assessing the carbon emission characteristics of the region in the present era. (3) In terms of research content, this paper not only explores the drivers of carbon emissions in the entire basin and different sections of the river but also identifies the compound effects of economic development and other factors as key contributors to carbon emissions.

2. Research Data and Methods

2.1. Study Area

According to the administrative division of China, the YRB contains a total of 963 counties. To ensure accurate estimation results and avoid potential data gaps caused by administrative adjustments, this study excludes certain data, resulting in a final selection of 948 counties as the observation objects.

2.2. Carbon Emission Estimates and Data Sources

2.2.1. Estimation of Carbon Emissions

The carbon emissions in the YRB are quantified in this study using county-level data from the CEADs database (www.ceads.net, accessed on 20 June 2022) for the following two main reasons: Firstly, due to the top-down nature of China’s energy statistics system, there is a lack of carbon emissions data on the county level. Most of the existing literature analyzes the spatial and temporal patterns of carbon emissions by using energy consumption data to obtain provincial and urban data. But even for the same province (autonomous region and city), there are differences among counties under its jurisdiction, which leads to imprecision in estimating county-level emissions. Secondly, NOAA’s National Earth Data Center releases two types of nighttime lighting data that serve as important indicators of economic and social activities due to their consistency, relatively broad coverage and free accessibility. In fact, these nighttime lighting data have been used to estimate CO₂ emissions at different scales [29]. However, most of the carbon emissions data generated by nighttime lighting measures have been divided into two periods: 1992–2013 and post-2013. Although some scholars initially combined, re-projected, resampled, and cropped images of DMSP-OLS data and NPP-VIIRS data annually while performing outlier removal and time series correction [30], these two datasets were ultimately merged. However, some scholars have pointed out noticeable discrepancies between the two datasets, suggesting that using this method to estimate carbon emissions may result in disparities between the obtained results and actual values [31]. In contrast, the CEADs database provides more accurate estimations for carbon emissions for each county in China from 1997 to 2017. Consequently, it provides a basis for investigating county-level carbon emissions in the YRB in this paper.

2.2.2. Data Sources

The carbon emission data in this paper were obtained from the CEADs database. Data on economic development indicators are from the China County Statistical Yearbook.

2.3. Research Methodology

2.3.1. Standard Deviation Ellipse

A standardized ellipse model consisting of the center of gravity as a crucial indicator of spatial distribution is first developed. This paper takes Li et al. [5] as the reference and uses standard deviation ellipses to study the spatial evolution of carbon emissions in brochure districts and counties, as is shown in Equations (1)–(4).

$$\overline{x_i}={\displaystyle \frac{{\sum }_{i=1}^n{w}_i{x}_i}{{\sum }_{i=1}^n{w}_i}}\hspace{1em}\hspace{1em}\overline{y_i}={\displaystyle \frac{{\sum }_{i=1}^n{w}_i{y}_i}{{\sum }_{i=1}^n{w}_i}}$$

(1)

$${\rho }_x=\sqrt{{\displaystyle \frac{{\sum }_{i=1}^n{\left(w_i\overline{x_i}\cos \theta -w_i\overline{y_i}\sin \theta \right)}^2}{{\sum }_{i=1}^n{w_i}^2}}}$$

(2)

$${\rho }_y=\sqrt{{\displaystyle \frac{{\sum }_{i=1}^n{\left(w_i\overline{x_i}\sin \theta -w_i\overline{y_i}\cos \theta \right)}^2}{{\sum }_{i=1}^n{w_i}^2}}}$$

(3)

$$\tan \theta ={\displaystyle \frac{\left(\sum _{i=1}^n{w_i}^2{\overline{x_i}}^2-\sum _{i=1}^n{w_i}^2{\overline{y_i}}^2\right)+\sqrt{{\left(\sum _{i=1}^n{w_i}^2{\overline{x_i}}^2-\sum _{i=1}^n{w_i}^2{\overline{y_i}}^2\right)}^2+4{\sum }_{i=1}^n{w_i}^2{\overline{x_i}}^2{\overline{y_i}}^2}}{2{\sum }_{i=1}^n{w_i}^2\overline{x_i} \overline{y_i}}}$$

(4)

We obtain the centroid coordinates and azimuth angle through mathematical derivation. Among them, $\overline{x_i}$ and $\overline{y_i}$ represent the relative coordinates of ($x_i, y_i$) from the center of gravity of distribution, respectively.${\rho }_x$ and ${\rho }_y$ denote the standard deviations along the X and Y axes, correspondingly. $\theta$ is the azimuth of the standard deviation ellipse.

2.3.2. Theil Index

In this study, a three-stage nested Theil index decomposition method is employed, with counties serving as the fundamental unit of analysis to investigate both inter-county and intra-county variations in carbon emissions. This approach enables an accurate depiction of the spatial structural characteristics pertaining to discrepancies in carbon emissions. The model is constructed as follows.

$${\mathrm{T}\mathrm{h}\mathrm{e}\mathrm{i}\mathrm{l}}_k={\displaystyle \frac{1}{n}}{\sum }_i^n{\displaystyle \frac{Y_{ik}}{{\overline{Y}}_K}}ln{\displaystyle \frac{Y_{ik}}{{\overline{Y}}_K}}$$

(5)

$${\mathrm{T}\mathrm{h}\mathrm{e}\mathrm{i}\mathrm{l}}_k={\sum }_{i=1}^n({\displaystyle \frac{n_{ik}}{n}}{\displaystyle \frac{Y_{ik}}{\overline{Y}}})T_{ik}+{\sum }_{i=1}^n({\displaystyle \frac{n_{ik}}{n}}{\displaystyle \frac{Y_{ik}}{\overline{Y}}})\log ({\displaystyle \frac{Y_{ik}}{\overline{Y}}})=T_w+T_b$$

(6)

where ${\mathrm{T}\mathrm{h}\mathrm{e}\mathrm{i}\mathrm{l}}_k$ represents the Theil index, $T_w$ represents the within-group Theil index, and $T_b$ represents the between-group Theil index.

2.3.3. Kernel Density Estimates

In this paper, the methodology proposed by Guo and Fang [18] is used to examine the distribution pattern, peak value, and extensibility of carbon emissions with the aid of kernel density estimation, so as to obtain a comprehensive depiction of the distribution. The specific equation is shown in (7).

$$\mathrm{g}(\mathrm{x})={\displaystyle \frac{1}{nh}}{\sum }_{i=1}^{n}k({\displaystyle \frac{x-x_i}{h}})$$

(7)

2.3.4. Geographic Detectors

To detect the spatially stratified heterogeneity of carbon emissions and determine the extent to which these factors explain such variation, spatial heterogeneity is divided into spatial local heterogeneity (SLE) and spatial stratified heterogeneity (SSH), with the latter mainly referring to variations among multiple regions. This paper builds upon Song’s [31] theory and employs factor detection and interaction detection in a geodetector to analyze the determinants of carbon emissions at both county levels. Factor detection is quantified using q-values, $q=$1$-{\displaystyle \frac{{\sum }_{h=1}^L{W}_h{{\sigma }_n}^2}{w{\sigma }^2}}$, in which $w_h$ and $w$ represent the number of cells in layer h and the total area, respectively; L refers to variable stratification. Stratification here refers to dividing the study area into sub-regions, with spatial heterogeneity considering if the sum of sub-region variances is lower than that of the entire region. ${\sigma }_h$ and $\sigma$ represent carbon emission variances for stratum h and the overall region.

3. Results and Discussion

3.1. Results Analysis

Based on the CEADs database, we measured carbon emissions at the county level in the YRB.

Table 1. Average carbon emissions in the YRB.

	Entire Basin	Upper Reaches	Middle Reaches	Lower Reaches
1997	1.167	0.780	0.956	1.766
1998	0.846	0.667	0.916	0.954
1999	1.003	0.722	0.961	1.326
2000	1.025	0.759	1.021	1.294
2001	1.116	0.777	0.996	1.574
2002	1.162	0.832	1.091	1.564
2003	1.382	0.982	1.265	1.900
2004	1.585	1.113	1.403	2.239
2005	1.999	1.417	1.642	2.936
2006	2.244	1.610	1.858	3.263
2007	2.414	1.763	1.960	3.520
2008	2.615	1.967	2.101	3.777
2009	2.789	2.117	2.279	3.971
2010	3.059	2.401	2.512	4.265
2011	3.434	2.888	2.868	4.545
2012	3.496	2.933	2.938	4.619
2013	3.471	3.025	2.920	4.468
2014	3.549	3.082	2.959	4.604
2015	3.389	2.908	2.756	4.504
2016	3.492	2.972	2.829	4.674
2017	3.552	3.110	2.876	4.669

reports the average carbon emissions for the entire YRB, as well as the upstream and downstream regions, covering the period from 1997 to 2017. The carbon emissions in the middle and lower reaches of the basin are particularly concerning, with growth rates surpassing the overall average. A likely reason for this is that the middle and lower reaches are resource-dependent areas where economic growth has been driven by the intensive exploitation of natural resources, leading to persistently high carbon emissions. In terms of absolute values, there is a clear pattern of increasing carbon emissions from the upstream to the midstream and then to the downstream regions, highlighting a significant imbalance in carbon emissions across the basin. Cities in the downstream regions, located near the eastern coastal areas, are geographically positioned to absorb pollution transfers from the east, further contributing to elevated carbon emissions. In contrast, the midstream and upstream counties are primarily agricultural areas with abundant natural resources. In particular, the upstream region serves as an “ecological barrier zone”, with a key focus on ecological preservation.

3.2. Spatial and Temporal Distribution Characteristics

Figure 1 illustrates the spatial distribution of county-level carbon emissions in the YRB. It can be observed that over the 20-year period from 1997 to 2017, the total number of carbon emissions in the YRB showed an increasing trend, with high-emission areas expanding significantly in a contiguous manner. In 1997, carbon emissions were relevantly low, totaling 1.042 billion tons. By 2017, the carbon emissions in the YRB reached 1.149 billion tons, representing a nearly 1.1-fold increase. The Western Development Strategy was a major factor driving the gradual rise and imbalance in carbon emissions in the sample regions. Furthermore, carbon emissions in the Shandong and Shanxi provinces show a clear tendency to spread toward surrounding areas, indicating an outward spillover characteristic of emissions. This is especially evident at the borders of Shandong and Shaanxi, Gansu, and Inner Mongolia, where a pattern of layered expansion emerges, forming two primary core regions of carbon emissions. In 1997, the Zaozhuang, Jining, and Shizhong districts of Jinan; Shandong Province; and Qingxu County in Taiyuan, Shanxi Province, were identified as high-emission areas due to their resource-dependent economic development. Compared to other areas, the Ganzi Autonomous Prefecture in Sichuan has limited energy consumption and is classified as a low-carbon-emission region. This is due, on the one hand, to its reliance primarily on the primary sector, with its few industrial activities meaning carbon emissions remain low. On the other hand, the region benefits from abundant wind, photovoltaic, and hydropower resources on the Western Sichuan Plateau. Among them, hydropower stands out as the most reliable clean energy source in Sichuan, serving as a critical guarantee for achieving the “carbon neutrality” goal. By 2002, the high-emission counties had expanded, primarily concentrated in Zaozhuang, Jining, and Jinan, and also extended to Qingxu County in Taiyuan, as well as the cities of Jimo, Shouguang, and Dongying District in Shandong. In 2007, Ushen and Jungar Banner in Inner Mongolia, along with Pingdu City in Shandong, were classified as high-emission regions. The reason lies in Inner Mongolia’s abundant coal resources, with coal reserves and production accounting for a quarter of the national total. The reliance on coal resources to drive industrial development has resulted in persistently high carbon emissions in this region. By 2012, the high-emission areas further expanded, covering Qingxu County in Taiyuan; Ushen and Jungar Banner in Inner Mongolia; as well as Zhibu District, Jining, and Shizhong District in Jinan. In 2017, the carbon emissions of key high-emission counties showed steady growth, with Jungar Banner in Inner Mongolia topping the list with 25.679 million tons. In summary, throughout the observation period, Zaozhuang, Jining, and Jinan in Shandong Province; Qingxu County in Taiyuan; Shanxi Province; and Ushen and Jungar Banner in Inner Mongolia consistently ranked as high-emission areas and thus should be key targets for precise emission reduction efforts. This is because these regions are primarily resource-based cities with prominent high-carbon-emission characteristics. The low-carbon transition of related industries not only faces critical path dependency, but high-pollution firms also lack the motivation to transform due to the high costs associated with transition. Meanwhile, counties such as Longde in Ningxia and Jiuzhi in Qinghai, with limited resources and underdeveloped economies, consistently remained in the low-emission category, with no significant fluctuations being observed. The reason lies in the relatively limited resource consumption due to the underdeveloped industrial sector in these regions. Thus, it is essential for these areas to guard against the repeated construction of outdated capacities in the secondary industry while pursuing economic development, as this may lead to increased carbon emissions.

Figure 2 reports the changes in carbon emissions of the sample from 1997 to 2017. This paper classifies county-level carbon emissions into three types based on quartiles. It also defines counties that change from being higher to lower carbon emitters as an upward shift and the opposite as a downward shift, while counties indicating no significant change are identified as stable counties. It is observed that during the period from 1997 to 2017, a total of 146 counties, accounting for 15.401%, experienced upward shifts, with Shandong Province accounting for the highest number of such transitions (38 counties), indicating remarkable progress in their adjustment efforts. This is probably due to targeted energy saving and emission reduction strategies for industrial structure upgrading. This indicates the need to further promote innovation, application, and dissemination in this region, as well as encourage enterprises to increase investments in clean and renewable energy sources. The 134 counties experiencing a downward shift, which account for 14.135%, are primarily located in Inner Mongolia, Shaanxi, the junction area of Gansu, central Qinghai, and southeastern Shanxi. In the early stages of development, these regions ware excessively dependent on coal resources. The significant inertia makes short-term adjustments challenging, leading to high carbon emissions. Meanwhile, 668 counties maintain a stable carbon emission level, dispersed mainly across the middle reaches of China, accounting for 70.464% of the total. This suggests that most regional areas continue their original production patterns but require increased efforts towards energy conservation, emission reduction, and the elimination of backward production capacity [32].

3.3. Analysis of the Global and Local Characteristics of Carbon Emissions

Figure 3 illustrates Moran’s I index under different spatial weight matrices. Moran’s I index in the YRB remains above 0.15, indicating strong spatial autocorrelation. In terms of the performance, Moran’s I index exhibits a similar trend across various weight matrices. Before 2006, there was a growing trend. After 2006, the global Moran’s I index declined, with a slight recovery being observed in 2013, though to various degrees.

In general, the county-level carbon emissions in the YRB exhibit the following two characteristics: (1) Geographical distance serves as an important basis for the spatial spillover of carbon emissions, with a “club convergence” effect being present among counties in the YRB. The carbon emission development trends in different counties exhibit similar characteristics. When neighboring counties have high carbon emissions, the carbon emissions of the focal county tend to be elevated as well. Moreover, this study finds that even when high-carbon-emission counties are adjacent to low-carbon-emission ones, the probability of its transitioning to a lower carbon emission level remains relatively low. This may be due to the fact that the initial endowments of high-carbon-emission areas determine their industrial structure, which requires more energy resources for support. It is difficult for high-carbon-emission counties to transform their production methods and achieve low-carbon development in the short term. (2) Based on the weighted matrix constructed from economic development, the average Moran’s I indicates a higher degree of association between carbon emissions and economic factors. Counties with similar levels of economic development exhibit high carbon spillover effects and low carbon lock-in effects. Counties with similar economic development levels may experience spatial spillover effects of carbon emissions through industrial interconnection [33], resulting in a “Matthew effect” where high-carbon-emission areas remain high and low-carbon-emission areas remain low. In conclusion, the influence of geographical and economic factors is more pronounced.

3.4. Shift in the Center of Gravity of Carbon Emissions and Standard Deviation Ellipse Analysis

In this section, three distinct time points are selected, namely 1997, 2007, and 2017. The center of gravity coordinates are measured based on county-level carbon emission data to further depict the shifting trajectory (refer to Figure 4). In terms of the basin-wide situation, the center of gravity of its carbon emission distribution primarily lies within the range of 112.028° E–112.492° E and 36.190° N–36.652° N, shifting from Anze County to Huo City County. Throughout the observed period, the standard deviation ellipse of carbon emissions is mainly located in the middle and lower reaches, demonstrating a clustering distribution pattern. The area encompassed by this ellipse gradually increases over time, indicating a trend of spatial spillover in carbon emissions. The change in the turning angle $\theta$ mainly falls in a range of 48.308–54.364°, and its variation is not significant, indicating a relatively stable spillover direction. In terms of the axis length, the long half-axis length increased from 791.183 km in 1997 to 798.357 km in 2017, while the short half-axis length increased from 504.510 km to 525.091 km, indicating a concentration trend of carbon emissions in the middle and lower reaches of the counties and dispersion towards the upper counties.

As is shown in Figure 4, in the upstream region of the YRB, the center of gravity in 2017 was located in Ordos City. During the sample period, the angle $\theta$ displayed a counterclockwise trend, indicating a northwestward concentration of carbon emissions in upstream counties. The standard deviation of the long axis showed a decreasing trend, suggesting that the spillover effect of carbon emissions along the long axis was reduced. Meanwhile, the standard deviation of the short axis continuously shifted toward the northwest. In the midstream region, the center of gravity was located in Linfen City, and the angle $\theta$ remained relatively stable throughout the sampling period, reflecting a steady growth in carbon emissions in this area. The standard deviation of the long axis was 3.174, while that of the short axis was 1.566, indicating a “first agglomeration, then diffusion” pattern in both axes within the midstream counties. In the downstream region, the center of gravity was in Shandong Province, where the angle initially increased and then decreased.

3.5. Spatial Differences in Carbon Emissions and Their Decomposition

Figure 5 presents the results of the three-level nested decomposition of the Theil index from 1997 to 2017. It can be observed that carbon emissions in the YRB show a noticeable downward trend, decreasing from 0.425 in 1997 to 0.353 in 2017. This indicates that, against the backdrop of China’s active implementation of energy conservation and emission reduction policies, the issue of elevated carbon emissions in the YRB has been alleviated; however, the effects are not particularly significant. As for inter-municipal variation, which is the second most influential factor affecting the spatial pattern of carbon emissions, it demonstrates a gradual decline from 0.113 in 1997 to 0.088 in 2017. Inter-provincial differences exhibit a first decreasing and then increasing trend within the observed period, similarly representing one of the major sources contributing to the overall CO₂ variation. In terms of river basins, the intra-regional variation in the upper region is significantly higher than average, as is observed in Yushu, Qinghai, and Liangshan City, Sichuan. The majority of the Theil indexes fall in the range of (0.824, 1.792) and (0.582, 0.938). For the midstream provinces, Theil indices range from (0.020, 0.039) and (0.130, 0.159)], while the downstream municipalities exhibit negligible intra-municipal variation compared to the upstream areas. In summary, the variations within urban areas are the primary drivers of differences in carbon dioxide levels, with differences in the upstream areas being the most significant, followed by those in the downstream areas, and the differences in the midstream area being the least pronounced.

3.6. Dynamic Evolution of Spatial and Temporal Distribution

This paper further investigates the dynamic evolution of the temporal distribution of CO₂ in each county, focusing on its spatial location, distribution pattern, and wave number using the kernel density model. According to Figure 6, Figure 7, Figure 8 and Figure 9, regarding spatial location, the centers of CO₂ distribution curves for YRB as a whole progressively shift towards downstream areas, with little change being observed in upstream counties, thereby confirming the findings presented in Figure 2. In terms of distribution patterns, the kernel density function reveals a skewed distribution of CO₂ emissions among YRB counties, with a gradual increase in carbon emission disparity between them. Specifically, the peak values of carbon emissions in the counties of the YRB exhibit an alternating trend, indicating that the overall fluctuations in carbon emissions in the basin are relatively minor. However, the disparities among counties show a sign of widening. The gap in carbon emissions between the midstream and downstream regions of the YRB is expanding, which is consistent with the overall trend, but the degree of this widening is more pronounced compared to the overall situation. During the sampling period, the peak-to-valley comparison of the upstream counties in the YRB remained relatively stable, with minimal fluctuations. This stability can be primarily attributed to the significant differences in carbon emission levels among the counties within the basin. The upstream areas are predominantly water conservation zones and ecological barriers boasting a superior ecological environment. In contrast, the central regions serve as bases for heavy industries, while the downstream areas are important oil and gas production hubs, resulting in consistently high carbon emission levels. In terms of the number of peaks, carbon emissions in both the YRB and upstream counties display marked polarization phenomena, with middle and downstream-county peaks experiencing year-on-year increases amidst fluctuations. Although downstream-county peaks remain relatively stable, the polarization phenomenon tends to increase annually. Therefore, inter-county spillover effects associated with carbon emissions not only make neighboring counties face potential risks but also exacerbate the polarization phenomenon.

3.7. Analysis of Factors and Mechanisms Influencing Carbon Emissions

This paper utilizes the quantile method to construct a geoprobe model to measure the spatial difference factors affecting CO₂ emissions. The YRB has global and strategic significance in economic development. Moreover, numerous studies have pointed out that the current economic spatial pattern of the YRB shows an obvious spatial differentiation. The spatial heterogeneity of many factors such as industrial structure, population size, and urbanization is obvious [34]. Meanwhile, in the institutional context of China, local government’s policy support plays a pivotal role in shaping regional development strategies. Therefore, considering the availability of county data, the indicators selected in this paper are as follows: (1) For economic development, GDP per capita is used. (2) Industrial agglomeration (IA), calculated as IA$={\displaystyle \frac{I_n/{GDP}_n}{S/{GDP}_s}}$ where $I_n$ represents the total industrial output of each county, ${GDP}_n$ is the GDP of each county, $S$ is the total industrial output of the brochure, and ${GDP}_{s }$ is the GDP of the brochure. (3) Population size (HC), measured by the total resident population. (4) Urbanization (Urban), measured using the share of urban population in the total population. (5) Government support (Government), measured using fiscal expenditure as a share of GDP.

Table 2. Single factor detection results.

	Impact Factor	q
		Basin-Wide	Upstream	Midstream	Downstream
Economic development	GDP	0.359	0.469	0.394	0.370
Population	HC	0.108	0.109	0.200	0.280
Industrial agglomeration	IA	0.157	0.130	0.310	0.139
Urbanization	Urban	0.232	0.169	0.158	0.353
Government support	Government	0.022	0.022	0.064	0.207

reports the average contribution degree of each factor of carbon emissions from 1997 to 2017.

The results of the measurements suggest that the q-value is 0.359. This may be credited to China’s implementation of a series of policies in the early 20th century, which not only yielded substantial policy dividends to economic growth but also contributed to an increase in energy-intensive production at that time.

In recent years, although China puts forward the concept of green development, many inland counties in the region still rely on lenient environmental policies and abundant energy resources to drive economic growth. Moreover, it has a demonstration effect in the brochure, promoting the high carbon spillover effect. On the other hand, the YRB’s previous economic development model has a severe path dependence. Therefore, the traditional crude development model-driven economic growth is still the main source of carbon emissions in the YRB.

Industrial agglomeration is an important factor affecting carbon emissions, with a particularly pronounced impact being observed in the midstream sector. This paper contends that this phenomenon can be attributed to the dominance of coal consumption, accounting for over 70% in these regions. The persistent reliance on coal along with limited access to oil and gas resources has hindered the development of technologically advanced and high-value industries, resulting in an absence of concentrated industrial agglomeration. This agglomeration of energy-intensive enterprises will not only bring about huge energy consumption and wastes of resources, but will also lead to vicious competition among industrial enterprises, thereby exacerbating the intensity of carbon emissions.

In terms of population size, unlike the findings of Wang et al. [35], the results of this paper suggest that an increase in population increases the demand for electricity and transportation, which leads to an increase in carbon emissions. This is particularly evident in the lower reaches. For example, Shandong Province and Henan Province have large populations, yet the upper region boasts the obvious characteristic of sparsely populated areas. Therefore, it can be inferred that population size does not emerge as a predominant determinant influencing carbon emissions. The process of urbanization has led to a significant increase in the urban population, resulting in a proliferation of physical materials and infrastructure, as well as a shift towards high carbon production and consumption patterns. Moreover, the massive demolition and duplication of construction during the urbanization process have directly contributed to the increase of carbon emissions [36]. As a result, population movements and production activities associated with urbanization make certain counties and downstream regions high-carbon-emission areas.

In terms of government support, the regions mentioned in the brochure are generally characterized by limited economic development. The primary objective of government financial assistance is to enhance counties’ attractiveness and facilitate the absorption of industrial transfers from the coastal region. However, this concentration of outdated production capacity will undoubtedly raise regional CO₂ levels. This also provides a new explanation for the increased fiscal efforts observed in both the Shandong Peninsula and central provinces in recent years, which paradoxically contribute to elevated carbon emissions.

Based on the above analysis, it is very clear that the economy plays a paramount role in influencing carbon emissions. Therefore, this section delves deeper into the examination of the superimposed effect of economic development and other factors on carbon emissions in the YRB, with the findings presented in

Table 3. Influence factor superposition detection results.

Interaction Factor	Basin-Wide	Upstream	Midstream	Downstream
GDP × HC	0.5706	0.6915	0.7120	0.7458
GDP × IA	0.4595	0.5874	0.6098	0.6605
GDP × Urban	0.4231	0.6002	0.6244	0.4823
GDP × Government	0.3752	0.5359	0.4735	0.5681

. It can be found that both economic development and other factors exhibit a significant increase in their influence on CO₂ levels under the interaction. The overlapping effects of population size, industrial agglomeration, and economic development are the reason for the strongest correlation with carbon emissions throughout the region in the brochure, indicating that population dynamics, economic activities, and secondary industrial growth dominate as primary drivers for continued carbon emission escalation. In the upper and middle reaches of the brochure, population size and urbanization degree are important drivers of carbon emissions. In the lower reaches, the superposition of population size and industrial agglomeration has a stronger explanatory power. As for this paper, it can be attributed to two factors: For one thing, this particular region is a resource-based region characterized by a concentration on fossil energy with significant advantages in energy costs. Consequently, it tends to adopt an energy-driven economic development model, resulting in increased carbon emissions and displaying a trend of continuous expansion around the junction of Shandong and Shaanxi, Gansu, Ningxia, and Mongolia, with Shandong being the core. For another thing, with the industrial transfer from China’s coastal cities, the counties situated in the middle and lower reaches have attracted a large number of heavy industry sectors by virtue of their own resource endowments. Therefore, both its coal-based energy structure and the large number of industries have become important factors constraining the low-carbon development of the YRB, which is consistent with the results of existing studies [37].

3.8. Discussion

The research in this paper differs from the existing literature. First, unlike the work of Zhang, who used data envelopment analysis to measure carbon emissions [5], this paper examines comprehensively the spatial and temporal patterns as well as evolutionary trends of county areas in the YRB. It also provides a detailed analysis of different river sections, thereby offering insights into understanding the growth and spatial expansion patterns of carbon emissions from the county perspective. It also compensates the shortcomings of Wei et al. to a certain extent, which mainly focus on the provincial or municipal perspectives [38]. Second, we apply the kernel density estimation method to examine the spatial and temporal dynamics of carbon emissions. The results reveal a significant high carbon spillover effect and low carbon lock-in effect in the YRB. This demonstrates a trend of concentration in counties in the middle and lower reaches, while counties in the upper reaches show divergence. These findings not only provide direct empirical evidence for a comprehensive understanding and assessment of carbon emission characteristics in the YRB during the new era but also further highlight the importance of strengthening the joint prevention and control of carbon emissions in China at this stage. Third, unlike the studies by Song et al., which analyzed the factors affecting carbon emissions in the YRB from a single perspective [39], such as population or industrial structure, this paper not only explores the drivers of carbon emissions across the entire YRB and its different sections but also identifies the compounded effects of population size and industrial agglomeration as key determinants of carbon emissions [40]. The reasons are as follows: The YRB has become a major hub for energy-intensive industries relocated from China’s eastern coastal regions, whose production processes require substantial energy consumption and contribute to higher carbon emissions. This has resulted in an energy structure dominated by coal and a significant concentration of industrial activities, both of which pose challenges to low-carbon development in the basin. Population growth drives up electricity consumption, transportation demand, and ultimately carbon emissions, a phenomenon that is particularly pronounced in the downstream areas of the YRB. The findings of this paper offer a valuable complement to the literature on carbon emission determinants, helping to clarify the sources of spatial disparities in carbon emissions in the YRB and providing theoretical insights for formulating carbon peak action plans.

4. Conclusions and Policy Implications

4.1. Conclusions

The study finds the following: (1) At the grid scale, the spatiotemporal evolution of carbon emissions from the aspect of energy consumption in the YRB from 1997 to 2017 is significant. County-level carbon emissions exhibit a clear upward trend and an expansion pattern, with a contiguous spread characterized by a distribution pattern of higher emissions in the east and lower emissions in the west, as well as higher emissions in the south and lower emissions in the north, with the midstream and downstream areas being the center without reaching a peak. In particular, there is an expanding circle of carbon emissions at the junction of the Shandong, Shaanxi, and Gansu provinces, forming two core zones of high emission intensity. (2) Carbon emissions in these counties show a strong correlation with economic factors, demonstrating obvious high spillover effects but limited locking effects on carbon reduction efforts. And there is a gradual shift in the center of gravity towards the middle and lower reaches of this region’s distribution pattern, indicating a tendency for concentration within these areas while dispersion occurs in upstream counties. (3) Intra-regional differences play a crucial role in shaping the overall spatial pattern of carbon emissions in the YRB. The internal differences show a distribution pattern where the upper reaches have the highest emissions, followed by the lower reaches, and the middle reaches have comparatively lower emissions. This polarization phenomenon has been progressively intensifying over time. (4) Economic development consistently exerts the strongest influence on the spatial differentiation of carbon emissions, while population size and industrial agglomeration act as key driving forces for the annual growth of carbon emissions in the YRB. In contrast to the work of Guo et al. [41], who concluded that the relationship between industrial agglomeration and carbon emissions may be nonlinear depending on the type of industry, this paper’s findings demonstrate a positive association between industrial agglomeration and carbon emissions in the YRB. This can be attributed to the concentration of high-energy-consuming enterprises in this region, resulting in significant energy consumption and resource wastage. Consequently, industrial agglomeration exerts a positive influence on carbon emission intensity, which aligns with the conclusion drawn by Song et al. [28].

4.2. Policy Implications

First, a diversified approach should be tailored to local conditions and time-specific circumstances. The study’s findings indicate that carbon emissions in the YRB have been rising annually, with a tendency to concentrate in the middle and lower reaches while diffusing in the upper reaches. Therefore, on the one hand, the YRB should be treated as a community of life being regulated by integrated planning and regional coordination mechanisms, such as the river chief system. Fair negotiation, distribution, and compensation of interests among cities should be ensured, balancing the overall interests of the basin with those of individual cities and fostering synergy. On the other hand, carbon reduction policies should be tailored to the unique historical conditions, resource endowments, and industrial structures of the upper, middle, and lower reaches. In the middle and lower reaches, stricter environmental regulations should be implemented to achieve carbon peaking. In the ecologically sound upper reaches, resource advantages should be leveraged to develop clean energy and green industries while preventing the spillover effects from high-carbon regions.

Second, efforts should be made to establish a joint prevention and control mechanism for emergency water pollution incidents in the upstream and downstream inter-provincial regions of the YRB. The study shows that the boundaries between Shandong, Shaanxi, and Gansu have formed two core zones of carbon emissions in the YRB, where high carbon spillover effects and low carbon lock-in effects coexist. Therefore, for regions with significant spatial agglomeration of carbon emissions, it is crucial to promote cross-regional functional integration and coordinated governance. In addition to strictly controlling carbon emissions in spillover hotspots, efforts should be made to achieve the division of labor and optimal resource allocation within agglomerated areas. Special attention should be given to establishing joint mechanisms in high-carbon counties within Inner Mongolia, Shandong, Shaanxi, Ningxia, and Shanxi. On the one hand, it is essential to consider the roles and distribution characteristics of various regions within the carbon emission space correlation network. This will facilitate a reasonable allocation of emission reduction responsibilities across regions, moderately reinforcing the carbon emission responsibilities of central and western regions such as Inner Mongolia, Shandong, Shaanxi, Ningxia, and Shanxi. This approach aims to create a favorable institutional environment for green technological innovation and introduction, effectively mitigating the “upstream to downstream” tendency in carbon emission spatial correlation and the “carbon emission refuge” effect between regions. On the other hand, it is important to explore the establishment of a collaborative mechanism for low-carbon development across the upper, middle, and lower reaches of the YRB, featuring “unified planning, unified standards, unified environmental assessments, unified monitoring, unified law enforcement, and unified emergency responses”. This will enhance the law enforcement capabilities for ecological environmental protection throughout the basin and improve the joint law enforcement mechanism across regions and departments.

Third, carbon emission policies should balance economic and environmental benefits. The study shows that economic development and urbanization are the main drivers of carbon emissions in the YRB, with population scale and industrial agglomeration playing a key role through their combined effects. Therefore, population size and government support should gradually shift towards green industries to set an example and lead the way. Regions with high population density, strong economic agglomeration, and spillover effects should take the lead in the regionally coordinated strategy, achieving a low-carbon transition in the YRB through technological leadership and innovation. Other regions can actively participate in the low-carbon cooperation strategy, optimizing the allocation of low-carbon industries.