Spatio-Temporal Diversification of per Capita Carbon Emissions in China: 2000–2020

Abstract

Exploring the low-carbon transition in China can offer profound guidance for governments to develop relevant environmental policies and regulations within the context of the 2060 carbon neutrality target. Previous studies have extensively explored the promotion of low-carbon development in China, yet no studies have completely explained the mechanisms of the low-carbon transition in China from the perspective of per capita carbon emissions (PCEs). Based on the statistics and carbon emissions data of 367 prefecture level cities in China from 2000 to 2020, this study employed markov chain, kernel density analysis, hotspots analysis, and spatial regression models to reveal the spatiotemporal distribution patterns, future trends, and driving factors of PCEs in China. The results showed that China’s PCEs in 2000, 2010, and 2020 were 0.72 ton/persons, 1.72 ton/persons, and 1.91 ton/persons, respectively, exhibiting a continuous upward trend, with evident regional heterogeneity. PCEs in northern China and the eastern coastal region were higher than those of southern China and the central and southwestern regions. The PCEs in China showed obvious spatial clustering, with hot spots mainly concentrated in Inner Mongolia and Xinjiang, while cold spots were mainly in some provinces in southern China. The transition of PCEs in China exhibited a strong stability and a ‘club convergence’ phenomenon. A regression analysis revealed that the urbanization level and latitude had negative effects on PCEs, while the regional economic development level, average elevation, average slope, and longitude showed positive effects on PCEs. These findings have important implications for the promotion of the low-carbon transition and the effective achievement of the “dual carbon” goal.

1. Introduction

The phenomenon of global warming, coupled with the melting of glaciers and a series of extreme weather events, has a significant impact on human survival and social activities, thereby posing ongoing challenges to social development [1,2]. The acceleration of CO₂ emissions has been unequivocally demonstrated as the dominant contributing factor to global warming [3,4]. Climate change has been induced by the explosive growth of greenhouse gases, primarily carbon emissions (CEs). How to tackle climate issues is the biggest environmental challenge in the world [5]. According to Climate Watch (2024) [6], China was the leading contributor to greenhouse gas emissions in 2020, accounting for 27%. In the ranking, the United States accounted for 11%, India accounted for 7%, while the European Union accounted for 6% of the emissions. The achievement of the Paris Agreement in 2015 has accelerated the global shift towards a low-carbon economy, thereby prompting significant transformations across various aspects of the world economy and society [7]. The Intergovernmental Panel on Climate Change reports underscore the importance of integrating renewable energy, enhancing energy efficiency, and fostering sustainable urban development. Initiatives such as the Global Carbon Project and the International Energy Agency’s sustainable development scenarios provide pathways and policy recommendations for achieving low-carbon transformation. As the most rapidly developing country in the world, China has become the largest emitter of CEs globally. In 2022, China’s CO₂ emissions reached 12.1 billion tons, accounting for 30% of the world, placing the country under severe pressure to reduce CEs. Although China ranks first in total CEs, the United States has the highest per capita carbon emissions (PCEs). Reducing CO₂ emissions has become a common goal for all countries [8]. In response, China is making efforts to carry forward low-carbon transition to reduce CO₂ emissions.

The evolution of CEs has been the subject of the existing research. Examining CEs from a per capita perspective is particularly important for understanding the unequal distribution. China exhibits significant differences across regions in terms of natural conditions, economic development, and population density. By approaching CEs from a per capita perspective, we can achieve a more precise measurement of regional contributions and responsibilities in CEs, reflecting the intensity and efficiency of CEs in each region. Hickel (2020) highlighted that China, despite its significant contribution to cumulative emissions, has four times the population of the United States [9]. If this difference is taken into account, the United States should be more responsible for CEs than China. Matthews (2016) quantified climate debts by the principle of atmospheric commons from a per capita perspective between 1960 (or 1990) and 2013 and determined whether countries’ PCEs are above or below the global average [10]. It was observed that if a country’s PCEs exceed the global average, it can be considered to be in debt. Conversely, if a country’s PCEs are below the global average, it can be considered to be in credit. Therefore, exploring CEs from a per capita perspective is based on fairness and equity. Thus, it is imperative to elaborate the spatio-temporal patterns and driving mechanism of PCEs in China, which would provide significantly insights for understanding the low-carbon transition and formulating effective action plans to achieve “dual carbon” targets.

The reduction in CEs has emerged as a key area of interest and investigation. The primary focuses of current research concerning CEs are on three aspects: CEs measurement [11,12,13], spatial distribution patterns and discrepancy [14,15], and driving mechanisms [16,17,18]. The degree of CEs output has been widely measured in many fields, such as the energy [19,20], manufacturing [21], service [22], tourism [23], and transportation industries [24]. The study of CEs has been progressively integrated with that of land use [25], thereby significantly enhancing the depth and breadth of research into CEs. In terms of scale, this research has included countries [26,27], urban agglomerations [28], river basins [29], provinces [30], and counties [31]. Moreover, there is a growing trend in researching spatial patterns and heterogeneity related to CEs. Scholars employed quantitative methods to quantify spatial autocorrelation and discrepancies, such as Moran’s I, the Theil index, and Dagum’s Gini index [32,33]. Moran’s I is extensively applied to quantitatively describe spatial autocorrelation [8,34]. For example, Chen et al. (2022) confirmed that building CEs had a positive spatial autocorrelation [35]. Wang et al. (2022) argued that the distribution of PCEs demonstrated self-reinforcing agglomeration in spatial distribution [36]. Furthermore, the spatio-temporal patterns and disparities were assessed using the Theil index and Dagum’s Gini coefficient. For example, Ma et al. (2022) employed Dagum Gini coefficient to assess the PCEs of commercial buildings in different regions, revealing that intra-regional disparities were smaller than inter-regional disparities [37].

The investigation of the driving factors contributing to CEs has been a topic of considerable interest and debate, drawing upon both theoretical deduction and empirical evidence. Existing academic studies have discussed the impacts of influencing factors, such as financial development [38], economic development [39,40], foreign direct investment [41], urbanization level [42], high-speed railways [43], and environmental policies [44]. The findings of previous studies indicated that factors such as economic development, population growth, and traffic congestion were positively associated with CEs. In addition, economic growth may also induce the unintended consequence of environmental pollution, thus resulting in higher CEs [45]. In the early stage, the conventional economic model resulted in a considerable amount of CEs. The process of urbanization has been identified as a significant driver of CEs [46]. As a consequence of urbanization, the rising demand for clothing, food, shelter, and transportation leads to different lifestyles, consequently increasing CEs. From this perspective, the process of urbanization inevitably increases CEs [47].

Meanwhile, when analyzing which factor exerts the greatest influence on CEs, a variety of analytical techniques are available for this purpose, including structural decomposition analysis (SDA) [48], Stochastic Impacts by Regression on Population, Affluence and Technology (STIRPAT), the logarithmic mean Divisia index (LMDI), and spatial econometric methods. For instance, Xu et al. (2021) took Guangdong province in China as the research object, and revealed that the factors of consumption structure, per capita consumption, and population caused the largest increase in CEs [48]. Yu et al. (2023) used the STIRPAT method to analyze the factors of household CO₂ emissions and identified that the total population, household size, unemployment rate, and urbanization level were the most significant factors [49]. Using a spatial econometric model, Liu et al. (2023) revealed that rapid economic growth and traditional industrial structure transform had promotion effects on CEs, while government expenditure, population clustering, and scientific innovation had inhibitory effects on CEs [50].

By the way of conclusion, previous research has provided a comprehensive examination of CEs, establishing a robust theoretical and empirical foundation. However, the previous research is not without limitations. Few studies have completely explained the mechanisms of the low-carbon transition in China from the perspective of PCEs. Furthermore, the study of how geographical factors influence PCEs has received only sporadic attention to date. To bridge the research gaps, this study aimed to analyze the spatiotemporal diversification of PCEs across different regions in China from 2000 to 2020, identify the trends and spatial autocorrelation in PCEs, assess the impact of economic and geographical factors on PCE levels in various regions, and provide insights into effective strategies for regional low-carbon transitions. So, we aimed to construct a framework to characterize the spatio-temporal patterns of PCEs, thereby expanding the depth and scope of research on PCEs. Secondly, we aimed to analyze the dynamic evolution of PCEs, recognizing that the CEs varied significantly across regions due to different influencing factors. Therefore, it is essential to explore this spatial heterogeneity, which can be visualized through spatial patterns. Finally, we attempted to explore the driving factors from both socio-economic and natural perspectives. While previous studies have identified various driving factors affecting CEs, there has been a tendency to focus on socio-economic factors, with relatively little attention paid to the role of natural and geographical factors. However, there is evidence that these factors exert a significant influence on CEs. Therefore, we integrated the effects of the urbanization rate, gross domestic product (GDP) density, average slope, average digital elevation model (DEM), latitude, and longitude factors to explore the effects on CEs. Revealing their driving mechanisms will help us to uncover the primary factor affecting China’s low-carbon transition, thereby contributing to achieving the goals of carbon peaking and neutrality.

2. Materials and Methods

2.1. Study Area

This study analysed 367 units at the prefecture level and above in mainland China (including a few county-level administrative districts; data for Hong Kong, Macao, and Taiwan were not yet available), including 4 municipalities directly under the central government, 293 prefectural-level cities, 7 districts, 30 autonomous prefectures, 3 leagues, and 30 provincial-level direct-administration units, for a total of 367 computational units (Figure 1). To scientifically reflect the socio-economic development of different regions in China, this study area was divided into four major economic subregions: Northeast, Central, Western, and Eastern China. The northeastern region mainly included the municipal units of Liaoning, Jilin, and Heilongjiang provinces; the eastern region includes the municipal units of Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan provinces; the central region includes the municipal units of Shanxi, Anhui, Jiangxi, Henan, Hubei, and Hunan provinces; and the rest is the western region.

2.2. Data Sources

This study selected 367 cities in China as the research sample, covering 30 provinces, municipalities, and autonomous regions. The CE data were derived from the Center for Global Environmental Research (https://db.cger.nies.go.jp/dataset/ODIAC/, accessed on 2 August 2024). The population data were mainly derived from the main data bulletins of the fifth, sixth, and seventh national censuses in 2000, 2010, and 2020, respectively. The land use data, DEM, and slope were derived from EPS data platform (https://www.epsnet.com.cn/index.html#/Index, accessed on 2 August 2024) and Data Center for Resources and Environmental Sciences (https://www.resdc.cn/, accessed on 2 August 2024). The GDP data were derived from the Statistical Yearbooks. In addition, for areas within the Xinjiang Uygur Autonomous Region where the seventh census data was unavailable, this study supplemented the missing values using linear regression.

2.3. Per Capita Carbon Emissions

In this context, PCEs represent the ratio of CEs to population [51,52,53]. The calculation equations were as follows:

$$PCE=CEs/Population$$

(1)

$$PCEC={\displaystyle \raisebox{1ex}{$CE{s}_{t2}$}\!\left/ \!\raisebox{-1ex}{$Populatio{n}_{t2}$}\right.}-{\displaystyle \raisebox{1ex}{$CE{s}_{t1}$}\!\left/ \!\raisebox{-1ex}{$Populatio{n}_{t1}$}\right.}$$

(2)

where PCEC represents the change in PCEs; CEs_t1 and CEs_t2 represent CEs of a unit at time t1 and t2, and Population_t1 and Population_t2 represent the population of a unit at time t1 and t2.

2.4. Markov Chain

The markov chain provides a mechanism for explaining how the probability distribution of transitions from one state to another eventuates [54]. Due to its steady-state analysis, the markov chain can be utilized to predict the long-term patterns and trends of geographic phenomena with a high degree of accuracy. In this study, we used markov chain to predict the trend of PCEs in each region, and the probabilistic transfer model constructed by the markov chain explaining the PCE conversion state in each region, reflecting the future interconversion of the values of different magnitudes. In general, the state type of PCEs at moment t is represented by a 1 × k state probability vector as Et = [E1,t, E2,t, …, Ek,t], and the transition process of PCE state types can be represented by a k × k Markov transition matrix M. Based on the similar counts for each type of county, PCEs at the county level were categorized into four types using quartiles (0.25/0.5/0.75), labeled as k = 1, 2, 3, and 4, respectively.

2.5. Kernel Density Analysis

As one kind of nonparametric estimation, kernel density estimation was characterized by relative smoothness and unbiasedness, which can accurately describe the distribution of random variables [55,56]. The calculation equations were as follows:

$$f(x)={\displaystyle \frac{1}{nh}}{\displaystyle \sum _{i=1}^{n}K\left(\frac{D_i-D}{h}\right)}$$

(3)

$$K(t)={\displaystyle \frac{1}{\sqrt{2\pi }}}{e}^{-{\displaystyle \frac{t^2}{2}}}$$

(4)

where K(x) is the kernel function; D_i is the independent and equally distributed observed values; D is mean value; n is the number of observed values; and h is the bandwidth. This study used Gaussian kernel function to effectively explore the changing law of PCEs in various regions so as to reflect the heterogeneity and the dynamic trend of PCEs in each region.

2.6. Hotspots Analysis

This study intended to use the Moran’s I index to reveal the spatial autocorrelation characteristics of PCEs. To further reflect the spatial agglomeration of PCEs and changes in PCEs, the Getis-Ord Gi* index was used to measure the statistically significant hot and cold spots of PCEs [35]. The Getis-Ord Gi* index can be used for measuring the spatial clustering, which can test whether variables exhibit the feature of high-value cluster or low-value cluster, thereby identifying the locations of hot and cold spots, and spatial outliers [57,58].

2.7. Spatial Regression

Spatial autocorrelation is a common phenomenon among geographical elements, and non-spatial models may overlook the spatial dependence between factors, which may lead to potential biases in identifying underlying influencing factors. To illustrate, an element is influenced not only by the level of the element in the study unit, but also by the level of the element in adjacent or distant units. Based on this, this study intended to use the least squares method (OLS), the spatial lag model (SLM), and spatial error model (SEM) to reveal the mechanism of China’s low-carbon transition [59]. The cross-sectional benchmark model was expressed as Equation (5):

$$PC{E}_i={\alpha }_0+{\alpha }_{1}LU{I}_i+{\alpha }_{2}GDP{D}_i+{\alpha }_{3}ASL{O}_i+{\alpha }_{4}ADE{M}_i+{\alpha }_{5}LA{T}_i+{\alpha }_{6}LO{N}_i+{\epsilon }_i$$

(5)

where i represents a city, and LUI is the development and utilization intensity of land [60,61], which represents the land urbanization level. The function, intensity, and efficiency of CEs are influenced by important driving forces such as land urbanization and associated human activities [62]. GDPD is GDP density, representing economic growth, which is an important factor. GDP promotes PCEs, while this effect weakens with the growth of GDP [63,64]. ASLO represents average slope, and ADEM represents average DEM. These two factors represent the topographic elements [20,51]. The ASLO and ADEM factors exert notable effects on CEs by influencing urban expansion [65,66]. In this study, we also added longitude (LON) and latitude (LAT) factors to reveal the rules of CEs in terms of latitude and longitude.

SLM was expressed as shown in Equation (6):

$$PC{E}_i=\rho {\displaystyle {\sum }_{i=1}^n{W}_{ij}\times PC{E}_i}+\beta X_i+{\epsilon }_i$$

(6)

where ρ is the spatial autoregressive coefficient, W_ij is the spatial weight matrix, β is a vector of coefficients X_i, and X_i represents the set of independent variables.

SEM was expressed as shown in Equation (7):

$$PC{E}_i=\beta X_i+{\epsilon }_i, {\epsilon }_i=\lambda {\displaystyle {\sum }_{j=1}^n{W}_{ij}{\epsilon }_i+{\phi }_i}$$

(7)

where λ is the spatial autocorrelation coefficient of the error term.

By combining multiple methods, this study aimed to evaluate the influencing factors to provide a reference for integrated regional green development (see Figure 2).

3. Results

3.1. PCEs in China

China’s PCEs in 2000, 2010, and 2020 were 0.72 ton/persons, 1.72 ton/persons, and 1.91 ton/persons, respectively. These figures demonstrated a continuous increasing trend, with notable regional variations in PCEs. The PCEs were the highest in the northeast and the lowest was in the central region with values of 0.53 ton/persons in 2000, 1.23 ton/persons in 2010, and 1.43 ton/persons in 2020. Additionally, the PCEs southeast of Hu-line were was lower than those in the northwest. Specifically, in 2000, 2010, and 2020 the PCEs southeast of Hu-line were 0.72 ton/persons, 1.63 ton/persons, and 1.81 ton/persons, respectively, while that northwest of Hu-line were 1.09 ton/person, 2.81 ton/persons, and 3.06 ton/persons, respectively (Figure 3). From the perspective of spatial distribution, the PCEs in southern China were obviously lower than those in northern China. Also, the eastern coastal areas had higher PCEs than central and southwest China. Seeing the change from 2010 to 2020, the PCEs in different regions exhibited an increasing tendency (Figure 4).

3.2. Spatiotemporal Dynamics of PCEs

The markov transfer probability matrix was employed to predict future PCEs with a time span of 10 years (

Table 1. Markov transfer probability matrix of PCEs.

Variables	Spatial Extent	Time Span	Type	L	ML	MH	H	Number
PCE	China	10	L	0.720	0.274	0.004	0	211
			ML	0.011	0.655	0.327	0.005	180
			MH	0	0.016	0.661	0.322	183
			H	0	0	0.025	0.975	160
	Eastern China	10	L	0.724	0.241	0.034	0	58
			ML	0.035	0.482	0.428	0.053	56
			MH	0	0.060	0.460	0.480	50
			H	0	0	0.100	0.900	40
	Central China	10	L	0.714	0.285	0	0	49
			ML	0	0.553	0.446	0	47
			MH	0	0.026	0.736	0.236	38
			H	0	0	0	1	40
	Western China	10	L	0.581	0.383	0.034	0	86
			ML	0	0.686	0.313	0	67
			MH	0	0.014	0.720	0.264	68
			H	0	0	0.046	0.953	65
	Northeastern China	10	L	0.545	0.409	0.045	0	22
			ML	0	0.473	0.526	0	19
			MH	0	0	0.350	0.650	20
			H	0	0	0	1	11

Notes: L stands for low-level type; ML stands for medium–low-level type; MH stands for medium–high-level type; and H stands for high-level type.

). In overall spatial terms, PCEs were expected to exhibit stable growth over the next 10 years, with PCEs showing a tendency to converge to higher values. Specifically, this tendency was particularly pronounced in the northeast, followed by the central region. It was also observed that, with the exception of the northeast and the east, the shift probabilities on the main diagonal were higher than those off the diagonal, indicating relatively stable PCE levels. The probability of shifting from high to higher values in the northeast and eastern regions suggested that higher growth rates in PCEs exist in these regions.

3.3. Kernel Density

The changes in the study elements over the course of the study can be encapsulated in the contour plots of the kernel density of the normalized PCEs. The contour lines should remain close to the 45° diagonal if no drastic changes have occurred. Figure 5a,b,d showed the kernel density of PCEs over a 10-year time span for northeast, eastern, and central China, respectively, and it was found that there were shifts to higher or lower values of PCEs in these regions, with a more pronounced shift towards higher values in eastern and central China. Figure 5c shows the trend of the shift in western China over a 10-year time span, revealing that the PCEs shifted to higher values, while in some regions with higher values, the PCEs shifted to lower values. The changes in PCEs in China, as shown in Figure 5e, indicated that some peaks of the kernel density contours were basically located near 45°, showing that there was some degree of stability in PCEs, as well as shifts from low to high and from high to low.

3.4. Change in PCEs in China

From 2000 to 2010, the number of cities with a reduction in PCEs accounted for 0.544%, and only a small proportion of prefecture-level cities achieved a decreases in PCEs during that period. Correspondingly, a large number of prefecture-level cities had an evident increase in PCEs. For example, the proportion of prefecture-level cities with PCEs increasing by more than 1 ton/persons was 40.871%. However, the number of cities with a decrease in PCEs from 2010 to 2020 was 10.354%, which was higher than that of the previous decade. It is noteworthy that the proportion of cities with an increase in PCEs greater than 1 ton/persons was 1.907%, which was evidently lower than that in the previous decade. Most cities (29.700%) had PCEs increases of less than 1 ton/persons, indicating that China has undergone a drastic transition in terms of PCEs, and there has been remarkable progress towards the low-carbon transition (Figure 6).

To gain further insight into the distribution of PCEs, we calculated the Moran’s I indexes of PCEs, which were 0.258, 0.316, and 0.316 in 2000, 2010, and 2020, respectively, and the p-value was significant at the level of 0.001, obviously showing a significant spatial agglomeration pattern in PCEs. The specific spatial distribution characteristics can be obtained from the hot spot distribution map. We found that the hot spots were mainly distributed in Inner Mongolia and Xinjiang, while the cold spots were primarily distributed in some provinces of southern China (e.g., Sichuan, Chongqing, Guizhou, Hubei, Hunan, Jiangxi, and Hainan). From 2000 to 2020, it is evident that the PCE hot spots gradually expanded eastward into Inner Mongolia and northeastern China (Figure 7).

To further reveal the hotspots of changes in PCEs, we calculated the spatial autocorrelation index of PCE changes during 2000–2010 and 2010–2020. The spatial autocorrelation indexes of PCE changes during 2000–2010 and 2010–2020 were 0.354 and 0.179, respectively, and the p-value was significant at the level of 0.001. The hot spot analysis results demonstrated that the hot spots of PCE changes during 2000–2010 were distributed in Inner Mongolia, Northern Xinjiang, and Northern Qinghai Province. The cold spots of PCE changes were distributed in southern China, showing a similar pattern to those of the PCEs. The cold spots of PCE changes in China during 2010–2020 were distributed in the west of Xinjiang and the Pearl River Delta, while the hot spots were distributed in northeast China (Figure 8).

3.5. Driving Mechanism of Low-Carbon Transition

To reveal the driving mechanism of the low-carbon transition in China, the OLS, SEM, and SLM were employed in this study. This study first carried out OLS model (

Table 2. Regression results of OLS.

Variables	2000	2010	2020
LUI	−3.239 ***	−8.440 ***	−7.027 ***
	(0.950)	(1.818)	(1.816)
GDPD	26.842 *	44.587 ***	4.659 *
	(13.565)	(11.510)	(2.115)
ASLO	−1.209 *	−2.707 *	−2.993 **
	(0.578)	(1.109)	(1.127)
ADEM	−2.076 ***	−3.205 **	−3.131 **
	(0.574)	(1.120)	(1.143)
LAT	0.068 ***	0.152 ***	0.156 ***
	(0.010)	(0.020)	(0.020)
LON	−0.043 ***	−0.066 ***	−0.056 ***
	(0.008)	(0.016)	(0.016)
Constant	5.883 ***	10.269 ***	8.721 ***
	(1.029)	(2.033)	(2.088)
Moran’s I (error)	3.225 **	4.291 ***	4.341 ***
LM (lag)	8.744 **	16.081 ***	16.967 ***
Robust LM (lag)	1.946	2.522	3.339
LM (error)	7.324 **	13.865 ***	14.408 ***
Robust LM (error)	0.526	0.305	0.779
LM (SARMA)	9.271 **	16.387 ***	17.747 ***
Measures of fit
Log likelihood	−607.080	−846.958	−853.786
AIC	1228.160	1707.920	1721.570
SC	1255.500	1735.250	1748.910
R-Squared	0.248	0.272	0.262
N	367	367	367

Notes: *** p ≤ 0.001, ** p ≤ 0.01, * p ≤ 0.05. Standard deviation is shown in parentheses.

) to diagnose. The coefficients of LUI in 2000, 2010, and 2020 were −3.239, −8.440, and −7.027, respectively, which indicated that the LUI had a significant negative impact on PCEs. The coefficients of GDPD in 2000, 2010, and 2020 were 26.842, 44.587, and 4.659, which indicated that GDPD had a significant positive impact on PCEs. Higher GDPD typically indicates more economic activity and a greater population density, thus leading to higher energy consumption and CEs. The coefficients of ASLO in 2000, 2010, and 2020 were −1.209, −2.707, and −2.993, respectively, indicating that the ASLO had a significant negative impact on PCEs. Steeper slopes may limit large-scale agricultural or construction activities, potentially leading to lower levels of energy consumption and CEs. The coefficients of ADEM in 2000, 2010, and 2020 were −2.076, −3.205, and −3.131, illustrating that the ADEM had a negative impact on PCEs. Regions with higher DEM, characterized by mountainous and hilly landscapes, were less suitable for extensive agricultural and industrial operations, leading to a reduction in energy consumption and CEs. The coefficients of LAT in 2000, 2010, and 2020 were 0.068, 0.152, and 0.156, respectively, indicating that the LAT had a significant positive impact on PCEs. Higher latitudes often have colder climates, which may lead to increased energy consumption for heating, thus increasing PCEs. The coefficients of LON in 2000, 2010, and 2020 were −0.043, −0.066, and −0.056, respectively, indicating that the LON had a negative impact on PCEs. This is largely because that the eastern regions were the most economically developed areas in China, boasting advanced technological levels that were conducive to reducing CEs. The results showed that the Moran’s I indexes of PCEs in 2000, 2010, and 2020 were 3.225, 4.291, and 4.341, respectively, and the p-value was significant at the level of 0.05. Therefore, the results obtained using the OLS regression model cannot fully and scientifically explain the relationship between them. When considering spatial lag and spatial error terms, the SLM and SEM performed better.

The coefficient of LUI was negative, indicating that the urbanization promotes low-carbon transition (

Table 3. Regression results of SLM and SEM.

Variables	SLM	SEM	SLM	SEM	SLM	SEM
	2000		2010		2020
LUI	−2.470 **	−2.468 *	−6.279 ***	−6.749 **	−5.061 **	−5.034 **
	(0.937)	(1.109)	(1.795)	(2.163)	(1.771)	(2.153)
GDPD	21.547	19.399	34.852 **	38.262 **	3.428	3.013
	(13.206)	(13.780)	(11.219)	(12.964)	(2.032)	(2.162)
ASLO	−0.988	−1.200	−2.105	−2.639 *	−2.299 *	−2.818 *
	(0.565)	(0.656)	(1.075)	(1.288)	(1.090)	(1.319)
ADEM	−1.858 **	−2.316 ***	−2.814 **	−3.76 **	−2.716 *	−3.641 *
	(0.574)	(0.685)	(1.098)	(1.387)	(1.116)	(1.434)
LAT	0.055 ***	0.072 ***	0.114 ***	0.162 ***	0.115 ***	0.167 ***
	(0.011)	(0.013)	(0.022)	(0.027)	(0.022)	(0.027)
LON	−0.038 ***	−0.051 ***	−0.058 ***	−0.085 ***	−0.050 **	−0.075 ***
	(0.009)	(0.010)	(0.016)	(0.021)	(0.016)	(0.022)
Constant	5.085 ***	6.259 ***	8.599 ***	11.161 ***	7.322 ***	9.366 ***
	(1.067)	(1.297)	(2.062)	(2.708)	(2.075)	(2.826)
Spatial lag term	0.238 ***		0.290 ***		0.304 ***
	(0.070)		(0.068)		(0.067)
Spatial error term		0.257 ***		0.310 ***		0.326 ***
		(0.071)		(0.069)		(0.068)
Measures of fit
Log likelihood	−602.400	−602.588	−839.089	−839.433	−845.333	−845.680
AIC	1220.800	1219.18	1694.180	1692.870	1706.670	1705.360
SC	1252.040	1246.510	1725.420	1720.210	1737.910	1732.700
R-Squared	0.275	0.276	0.315	0.315	0.309	0.309
N	367	367	367	367	367	367

Notes: *** p ≤ 0.001, ** p ≤ 0.01, * p ≤ 0.05. Standard deviation is shown in parentheses.

). The coefficient of the GDPD was positive, but it was significant only in the SEM and SLM models of 2010. The coefficient of the ASLO was negative, but it was significant only in the SEM model of 2010 and the SLM and SEM models of 2020. The coefficient of the ADEM was negative in all models, indicating that PCE decreases with increasing altitude. PCEs in China increased significantly with rising latitude, while PCEs decreased significantly with rising longitude. In addition, PCEs were influenced not only by their own unit factors but also by those of adjacent units, as indicated by the spatial lag terms being significant at the 0.001 level. An average 1% increase in PCEs in surrounding areas in 2000, 2010, and 2020 would result in increases of 0.238%, 0.290%, and 0.304% increases in PCEs in their own units. In addition, the spatial error terms were also statistically significant at the 0.001 level, indicating that PCE changes were not only influenced by the aforementioned factors, but also by other factors.

The OLS regression analysis results (

Table 4. Regression results of OLS for different regions of China.

Variables	Western China			Northeastern China			Central China			Eastern China
	2000	2010	2020	2000	2010	2020	2000	2010	2020	2000	2010	2020
LUI	−1.592	−10.968 *	−10.676 *	−0.667	−6.731	−5.800	−3.502 *	−10.198 **	−9.015 *	−2.423 ***	−5.624	−1.775
	(2.550)	(4.397)	(4.456)	(1.738)	(4.465)	(5.667)	(1.398)	(3.276)	(3.681)	(1.284)	(2.963)	(3.003)
GDPD	50.300	361.097 **	90.605 **	90.515	66.990	2.065	53.920	91.123 ***	15.105 *	17.800	24.656 ***	0.988
	(228.985)	(115.654)	(30.547)	(49.639)	(52.383)	(29.286)	(37.676)	(24.616)	(7.083)	(4.536)	(5.562)	(0.945)
ASLO	−1.234	−1.811	−2.133	−1.606	−6.625	−5.410	−1.699 *	−4.247 *	−3.269	−0.275 *	−2.146	−2.965
	(1.099)	(2.028)	(2.008)	(1.724)	(4.490)	(5.302)	(0.820)	(1.864)	(2.140)	(0.720)	(1.593)	(1.797)
ADEM	−1.954 *	−3.098	−3.235	2.107	0.443	−2.997	2.021	5.196	3.328	−3.360	−3.237	0.309
	(0.991)	(1.871)	(1.879)	(3.246)	(8.574)	(9.434)	(1.667)	(3.762)	(4.264)	(1.699)	(3.875)	(4.229)
LAT	−0.038	−0.017	0.002	0.036	0.135 *	0.140	0.113	0.123 *	0.134 *	−0.025	0.006	0.061
	(0.020)	(0.035)	(0.035)	(0.024)	(0.062)	(0.072)	(0.026)	(0.058)	(0.067)	(0.014)	(0.033)	(0.036)
LON	0.095 ***	0.213 ***	0.228 ***	0.037	0.002	−0.014	0.113 ***	0.221 ***	0.279 ***	0.037 ***	0.072 *	0.025
	(0.029)	(0.052)	(0.052)	(0.038)	(0.100)	(0.115)	(0.029)	(0.069)	(0.078)	(0.011)	(0.028)	(0.027)
Constant	3.568	3.787	1.644	−4.647	−10.126	−9.627	0.028	−13.254	−16.710*	4.350 **	2.614	−4.425
	(2.702)	(5.116)	(5.141)	(3.249)	(8.346)	(9.460)	(3.079)	(6.893)	(7.909)	(1.632)	(3.801)	(3.825)
Moran’s I (error)	1.915	1.676	1.711	−1.055	−1.000	−1.404	1.706	2.182 *		1.688	2.709 **	3.959 ***
LM (lag)	1.331	1.567	1.504	3.379	3.790 *	3.353	1.333	0.960	1.191	12.149 ***	5.503	8.524 ***
Robust LM (lag)	0.109	1.324	0.750	0.318	0.791	0.529	2.591	0.028	2.044	1.911	2.654	0.263
LM (error)	1.500	1.002	1.097	3.065	3.099	4.139 *	0.545	1.253	0.529	10.240 ***	3.525	9.558 ***
Robust LM (error)	0.278	0.759	0.343	0.004	0.100	1.315	1.803	0.321	1.381	0.001	0.675	1.298
LM (SARMA)	1.609	2.326	1.847	3.384	3.890	4.669	3.136	1.282	2.573	12.151 **	6.179 *	9.822 **
Measures of fit
Log likelihood	−296.557	−383.134	−382.904	−12.723	−46.081	−52.108	−46.992	−117.917	−129.776	−32.805	−115.618	−126.148
AIC	607.113	780.268	779.808	39.447	106.162	118.218	107.985	249.833	273.552	79.611	245.235	266.296
SC	627.853	801.008	800.548	50.531	117.247	129.302	125.246	267.095	290.814	97.986	263.61	284.671
R-Squared	0.257	0.319	0.326	0.294	0.315	0.264	0.508	0.504	0.439	0.235	0.349	0.250
N	143	143	143	36	36	36	87	87	87	102	102	102

Notes: *** p ≤ 0.001, ** p ≤ 0.01, * p ≤ 0.05. Standard deviation is shown in parentheses.

) indicated that the Moran’s I indexes of the 2000, 2010, and 2020 settlements in the eastern region were 4.070, 2.709, and 3.959, respectively, with a significant p-value at the 0.01 level, which indicated that the model’s settlements in the eastern region had a strong spatial autocorrelation, so the spatial lag and spatial error terms were considered, and the SLM and SEM were used to improve the goodness of fit (

Table 5. Regression results of SLM for different regions of China.

Variables	Western China			Northeastern China			Central China			Eastern China
	2000	2010	2020	2000	2010	2020	2000	2010	2020	2000	2010	2020
LUI	−0.874	−9.350 *	−9.289 *	−0.583	−6.453	−4.865	−3.156 *	−9.456 **	−8.175 *	−2.007	−4.197	−1.361
	(2.467)	(4.279)	(4.325)	(1.341)	(3.406)	(4.404)	(1.366)	(3.259)	(3.606)	(1.136)	(2.756)	(2.721)
GDPD	23.714	335.386 **	87.693 **	49.497	32.640	−12.276	52.691	87.954 ***	14.178 *	13.715 ***	18.992 ***	0.501
	(221.392)	(111.95)	(29.534)	(38.662)	(39.884)	(22.700)	(35.743)	(24.093)	(6.747)	(4.015)	(5.279)	(0.855)
ASLO	−1.171	−1.675	−1.966	−2.343	−7.222 *	−6.313	−1.424	−3.755 *	−2.729	−0.201	−1.484	−2.227
	(1.066)	(1.971)	(1.951)	(1.336)	(3.443)	(4.119)	(0.796)	(1.815)	(2.065)	(0.639)	(1.484)	(1.643)
ADEM	−1.790	−2.802	−2.885	1.702	−2.903	−4.449	1.561	4.349	2.405	−2.803	−2.815	0.440
	(0.977)	(1.833)	(1.840)	(2.518)	(6.543)	(7.315)	(1.596)	(3.620)	(4.071)	(1.499)	(3.585)	(3.827)
LAT	−0.037	−0.024	−0.005	0.061 ***	0.200 ***	0.218 ***	0.024	0.108	0.115	−0.019	−0.002	0.039
	(0.019)	(0.034)	(0.034)	(0.019)	(0.048)	(0.058)	(0.024)	(0.057)	(0.065)	(0.013)	(0.031)	(0.033)
LON	0.082 **	0.179 ***	0.191 ***	0.038	0.023	0.0204	0.096 **	0.192 **	0.241 **	0.027 **	0.054 *	0.015
	(0.030)	(0.056)	(0.057)	(0.031)	(0.082)	(0.094)	(0.032)	(0.073)	(0.083)	(0.010)	(0.027)	(0.025)
Constant	3.408	4.265	2.362	−6.997 **	−17.122 *	−19.265 *	−3.247	−11.318	−14.162	3.301 *	2.720	−2.592
	(2.625)	(4.949)	(4.970)	(2.572)	(6.741)	(7.809)	(2.976)	(6.820)	(7.832)	(1.452)	(3.514)	(3.490)
Spatial lag term	0.163	0.160	0.160	−0.743 ***	−0.763 ***	−0.733 ***	0.189	0.160	0.182	0.435 ***	0.316 **	0.387 ***
	(0.111)	(0.109)	(0.109)	(0.211)	(0.203)	(0.212)	(0.146)	(0.143)	(0.147)	(0.099)	(0.111)	(0.109)
Measures of fit
Log likelihood	−295.739	−382.252	−382.037	−9.469	−42.470	−48.971	−46.287	−117.397	−129.135	−26.245	−112.564	−121.633
AIC	607.478	780.503	780.075	34.938	100.942	113.942	108.575	250.794	274.27	68.491	241.128	259.266
SC	631.181	804.206	803.777	47.606	113.61	126.61	128.302	270.521	293.997	89.490	262.128	280.265
R-Squared	0.270	0.331	0.338	0.478	0.507	0.451	0.519	0.513	0.451	0.361	0.403	0.340
N	143	143	143	36	36	36	87	87	87	102	102	102

Notes: *** p ≤ 0.001, ** p ≤ 0.01, * p ≤ 0.05. Standard deviation is shown in parentheses.

and

Table 6. Regression results of SEM for different regions of China.

Variables	Western China			Northeastern China			Central China			Eastern China
	2000	2010	2020	2000	2010	2020	2000	2010	2020	2000	2010	2020
LUI	−0.275	−9.211 *	−8.918 *	−0.142	−5.562	−3.579	−3.511 *	−10.718 ***	−9.135 *	−2.045	−4.122	−2.234
	(2.653)	(4.527)	(4.544)	(1.191)	(3.115)	(3.765)	(1.403)	(3.360)	(3.692)	(1.258)	(2.999)	(2.962)
GDPD	−15.509	27.583 **	85.35 **	34.575	−3.413	−21.060	53.435	97.235 ***	14.347 *	11.673	18.752 ***	0.305
	(222.797)	(113.220)	(29.620)	(35.537)	(32.089)	(20.643)	(36.688)	(24.908)	(6.892)	(3.804)	(5.803)	(0.842)
ASLO	−1.410	−2.156	−2.372	−2.748 *	−7.250 *	−6.033	−1.587	−3.885	−3.069	−0.805	−2.083	−3.670
	(1.177)	(2.128)	(2.117)	(1.119)	(3.008)	(3.620)	(0.829)	(1.934)	(2.164)	(0.769)	(1.715)	(1.964)
ADEM	−2.078	−3.158	−3.221	2.559	−2.851	−3.830	1.745	4.237	2.683	−2.527	−2.621	0.790
	(1.110)	(2.033)	(2.045)	(1.960)	(5.208)	(5.436)	(1.672)	(3.863)	(4.286)	(1.535)	(3.686)	(3.882)
LAT	−0.046 *	−0.031 ***	−0.010	0.052 ***	0.150 ***	0.166 ***	0.027	0.122	0.133	−0.012	0.015	0.071
	(0.023)	(0.039)	(0.040)	(0.011)	(0.031)	(0.035)	(0.027)	(0.066)	(0.071)	(0.022)	(0.043)	(0.051)
LON	0.099 **	0.213	0.230 ***	0.001	−0.075	−0.067	0.117 ***	0.238 ***	0.289 ***	0.027	0.059	0.023
	(0.033)	(0.058)	(0.058)	(0.024)	(0.064)	(0.072)	(0.030)	(0.071)	(0.079)	(0.014)	(0.032)	(0.033)
Constant	3.776	4.445	2.082	−5.238 **	−8.728	−11.849 *	−3.917	−13.374	−16.717 *	2.941	1.047	−5.023
	(3.166)	(5.760)	(5.800)	(1.801)	(4.683)	(4.956)	(3.246)	(7.724)	(8.344)	(2.384)	(4.728)	(5.387)
Spatial error term	0.207	0.164	0.168	−0.923 ***	−0.810 ***	−0.957 ***	0.134	0.213	0.133	0.513 ***	0.336 **	0.449 ***
	(0.110)	(0.112)	(0.112)	(0.199)	(0.214)	(0.193)	(0.156)	(0.149)	(0.156)	(0.096)	(0.115)	(0.104)
Measures of fit
Log likelihood	−295.458	−382.430	−382.149	−7.776	−40.236	−47.226	−46.686	−117.175	−129.477	−25.921	−113.195	−120.578
AIC	604.917	778.861	778.299	29.553	94.472	108.452	107.374	248.351	272.954	65.843	240.391	255.156
SC	625.657	799.601	799.039	40.637	105.557	119.537	124.635	265.612	290.216	84.218	258.766	273.531
R-Squared	0.275	0.330	0.337	0.555	0.555	0.545	0.513	0.517	0.445	0.380	0.397	0.364
N	143	143	143	36	36	36	87	87	87	102	102	102

Notes: *** p ≤ 0.001, ** p ≤ 0.01, * p ≤ 0.05. Standard deviation is shown in parentheses.

). The SLM and SEM regression results analysis in Table 5 and Table 6 indicated that the main factors affecting PCEs varied across the four regions of China. In the western region, LUI and GDPD were the main factors significantly impact on PCEs. LUI exhibited a significant negative effect, while GDPD had a significant positive effect. The reason was that urbanization often led to a shift from high-carbon-emission industries and agriculture to low-energy-consuming service and high-tech industries, thereby resulting in reduced PCEs. However, economic growth often lead to increased energy consumption and production, which can result in higher CEs. In the northeastern region, LUI and ASLO were the main factors, which both had significant negative effects on PCEs. This can be attributed to the abundant forest resources in northeastern China. The steep areas with dense vegetation act as carbon sinks, absorbing CO₂ from the atmosphere and thereby reducing the overall carbon footprint. In the central region, LUI and GDPD were the main factors which had significant effect on PCEs. LUI had a significant negative effect, while GDPD had a significant positive effect. In the eastern region, GDPD and LON were the main factors, which both had significant positive effects on PCEs.

4. Discussion

4.1. Interpretation of Findings

Theoretical and empirical research on PCEs made a significant contribution to the promotion of China’s low-carbon transition. In this research, we analyzed the various spatiotemporal patterns and projected future state of PCEs. Our findings revealed an upward trend in China’s PCEs, with northern China’s PCEs being significantly higher than those of southern China, and the east coastal region’s PCEs exceeding that of the central and southwestern regions. With economic growth, an increasing number of individuals have migrated to urban areas, which has subsequently resulted in a rise in CEs [67]. Northern China is currently the largest energy base, and energy consumption and greenhouse gas emissions have been rising sharply due to the rapid development of energy-intensive and polluting industries. For instance, Xue et al. (2011) confirmed that fossil energy production regions produce significantly more CEs than other regions [68]. Wang et al. (2022) demonstrated that the PCE level was positively correlated with the geographical and scale distribution of cities [69]. Compared to other regions, the eastern region has higher PCEs and CEs, demonstrating a connection between the region’s economic level and PCE level.

In the markov transition probability matrix, the highest probability was found in the H-type, reaching 97.5%, with an average stability probability of 75.27%, indicating that China’s PCEs were stable and exhibited a “club convergence” phenomenon. At the same time, the low-level (0.720) and high-level (0.975) convergence probabilities were greater than the medium–low-level (0.655) and medium–high-level (0.661) convergence probabilities, indicating that low- and high-level regions tended to stabilize at their own levels, while medium–low and medium–high levels tended to transfer to low and high levels. From the perspective of probability values, the probabilities of downward transfer were 32.7% and 32.2%, while those of upward transfer were 1.10% and 1.60%, indicating that the types with lower and higher intensities showed a good trend of downward transfer. The result obtained was consistent with the findings of Cui et al. (2022), Liu et al. (2023), and Wang et al. (2019) [16,50,52]. A dynamical procedure underlying the transition matrix showed that a discernible “Matthew effect” was observed in PCEs, whereby cities with both low and high emissions exhibited tendencies to maintain their initial state throughout the transfer process. This illustrated that the region may become trapped in a development path locked into specific trajectories, making it challenging to swiftly reduce PCEs through technological means.

By conducting a comprehensive review on China’s CEs, we can clearly identify the differences between this study and previous research. These differences primarily focus on the following three main aspects. (1) Regional level. Previous studies may focus on the disparities in CEs across different regions. This study analyzed 367 municipal units at the prefecture level. There were presently many studies in the literature on different regions. Jin et al. (2024) [70] took the urban agglomeration of the Yangtze River Economic Belt’s as the study areas of CEs. Bei et al. (2024) [71] conducted the study of CEs at Wuhan. Yan et al. (2023) [72] considered 30 provincial units (provinces, cities, and autonomous regions) in mainland China as their basic research units. (2) Research methods. This study used markov chain, kernel density, and spatial economic regression analyses. Feng et al. (2024) [73] and Chen and Bi (2022) [74] used Geodetector models to empirically analyze the driving mechanisms of CEs. Li et al. (2024) [75] used the LMDI model to investigate each province’s CEs drivers. Chen et al. (2023) [76] used kernel density estimation, a spatial autocorrelation analysis, and the spatial-temporal LMDI model to explore the spatiotemporal patterns and driving mechanism of CEs. (3) Driving factors. This study utilized a spatial economic model to explore the driving factors of CEs. Wei et al. (2021) [77] verified that GDP was the dominant factor affecting CEs efficiency. Wang et al. (2023) [78] revealed that technology and population size played important roles in CEs reductions. Jiang et al. (2024) [79] deduced that the scale of urban construction presented different promoting or inhibiting effects in different regions. Zhang et al. (2024) [80] demonstrated that technological progress was the main factor influencing CEs.

4.2. Driving Mechanisms of Low-Carbon Transition

Based on the results of the OLS, SLM, and SEM, we revealed the driving mechanism of China’s low-carbon transition and explored how socioeconomic and geographical factors that have significantly influenced low-carbon transition. The LUI was proven to have a significantly negative impact on PCEs, which indicated that the increase in urbanization promoted the low-carbon transition, which was inconsistent with previous studies showing increasing CEs due to urbanization [81]. It is a widely accepted hypothesis that the LUI can increase PCEs. On the contrary, some studies have pointed out that urbanization can indeed lead to a reduction in CEs [82,83]. The reason for this was that urbanization promoted the transformation of industrial structures and thus contributes to CE efficiency. The coefficients of the LUI affecting PCEs were −3.239, −8.440, and −7.027 in 2000, 2010, and 2020, which showed that urbanization was more crucial to the reduction in PCEs in 2010 and 2020 than in 2000. In the early stage, due to the problem of blindly pursuing quantity rather than quality under the traditional urbanization and economic model, the reduction in CEs has been inhibited [84]. However, as urbanization levels increased and shifted towards equity, a green economy, and efficiency as core objectives have contributed to the reduction in CEs. Moreover, several studies have corroborated the assertion that urbanization can result in a reduction in CEs, albeit with notable spatial heterogeneity. For instance, Li et al. (2023) [85] revealed that a higher level of urbanization reduced the CEs intensity and explored the significant implications of urbanization on the low-carbon transition by promoting green technological innovation. GDPD was also proven to be positively correlated with CEs. This is because economic growth was usually accompanied by the development of industrialization and manufacturing, which often required more energy, especially fossil fuels, thereby releasing large amounts of CEs [86]. Many scholars have found that economic growth inevitably promoted CEs, leading to global warming and glacier melting, which posed serious challenges to sustainable economic development. For example, Guo and Fang (2023) [87] revealed that CEs was positively correlated with economic growth but present a fluctuating trend. Similarly, Li et al. (2022) [88] concluded that there was a positive effect on GDP per capita and total CO₂. The difference of this study was investigating the impact of GDPD on PCEs, with GDP being categorized by land area.

The regression results also demonstrated that the low-carbon transition was driven by the joint action of socio-economic systems and natural systems, rather than being solely influenced by either socio-economic factors or geographical factors. This finding underscores the comprehensiveness of the driving mechanism and emphasized the importance of formulating region-specific policies. The European Union emissions trading system (EU ETS) was launched in 2005, which has demonstrated significant success in reducing greenhouse gas emissions across 31 countries, accounting for over 40% of the EU’s total greenhouse gas emissions. The EU ETS has led to substantial emission reductions while increasing the regulated firms’ revenues and fixed assets, as verified by Dechezleprêtre et al. (2023) [89]. The Regional Greenhouse Gas Initiative (RGGI) in the United States was officially implemented in 2009 and established a cap-and-trade program for CEs in the power sector. The RGGI not only achieved significant emission reductions but also generated economic benefits by reinvesting auction proceeds into energy efficiency and renewable energy projects. However, the RGGI has caused significant emissions reductions within regulated states and emissions leakages in nearby unregulated states (Yan, 2021) [90]. In future research, we would like to explore this driving mechanism by taking China’s environmental policy into consideration.

4.3. Policy Implications of Low-Carbon Transition

4.3.1. Promoting the Low-Carbon Transition

A significant proportion of prefecture-level cities exhibited a noticeable increase in PCEs. China’s CEs have remained high level due to the decision that was made for its economic development mode. As rising CEs contributed to global warming, China, as the biggest developing country, is enacting a low-carbon transition and has made remarkable achievements in curbing the growth of CEs. New-type urbanization is the strategic fulcrum of future development, which is crucial for promoting the low-carbon transition. New urbanization has the potential to establish a favorable policy environment and offer significant scientific and technological support, which can facilitate the low-carbon transition. Throughout the progress of new urbanization [91,92], it is of paramount importance to underscore the significance of enhancing the quality of urbanization [93].

4.3.2. Focusing on the Spatial Agglomeration Effects of PCE

PCEs showed a relatively obvious distribution pattern of spatial agglomeration, and the degree of this continued to strengthen. The PCEs of each city were susceptible to influences from neighboring cities. Therefore, it is crucial for each city to not only prioritize its own PCEs but also to collaborate with neighboring cities. By establishing regional cooperation mechanisms, cities can share information and technology related energy and environmental protection, jointly implement emission reduction measures, as well as collectively assume the responsibility of promoting coordinated economic development and environmental protection. Considering the differences across regions, for provinces such as Chongqing, Sichuan, Guizhou, Anhui, and Jiangxi, which are experiencing rapid economic growth, it is recommended to optimize their industrial structure, support low-energy consuming and clean industries, and promote the shift from high-speed to high-quality economic development. For resource-rich provinces such as Inner Mongolia and Xinjiang, which have resource-endowed advantages, it is suggested that these governments appropriately increase their carbon allowances and set higher thresholds for their high-energy-consuming enterprises, thereby promoting the introduction of high-energy-efficiency technologies from the east to the west.

4.3.3. Coordinating Economic Growth and CEs

In terms of driving factors, economic growth acts as a “double-edged” sword, not only increasing CEs but also providing momentum for the low-carbon transition. The contradiction between the increase in GDP and CEs still exists. However, with the improvement of quality, economic progress can stimulate technological innovation and accelerate emission reductions. Enhancing the quality of economic growth quality would inevitably result in a reduction in CEs and the promotion of the low-carbon transition. Economic growth should leverage the advantages of agglomeration effects and scale effects.

4.3.4. Formulating Differential Strategy

Based on the different regions of CEs and their spatial-temporal characteristics, we proposed a targeted strategy to reduce PCEs. At the national level, the focus should be on improving the efficiency of fossil energy use and replacing fossil energy with non-fossil energy sources, such as developing hydropower, wind power, and solar power. Additionally, carbon sinks can be increased by using trees or other plants to absorb the carbon dioxide emitted into the atmosphere. At the regional level, for the eastern and central regions, the priority should be to strengthen green technology upgrades, leverage their advantages in low-carbon technology research and development, and actively explore innovative systems and mechanisms for low-carbon economic development. For the northeastern regions, it is essential to promote the transformation of resource-depleted cities and the adjustment of old industrial bases in conjunction with the development of a low-carbon economy. Considering the reality that the northeastern region has more per capita arable land and forestry resources, agriculture and forestry should play a significant role in the low-carbon economic development in this region. For the western region, the focus should be on reducing CEs in resource development, distinctive industry development, and infrastructure construction. In terms of increasing carbon sinks, efforts should be made to strengthen carbon sink construction in ecological functional zones, integrating the construction of carbon sinks in the western region with the establishment of a national ecological security barrier.

5. Conclusions

This study aimed to investigate the distribution patterns, future trends, and driving factors of PCEs in 367 cities in China. The results revealed that the PCEs presented a continuous upward trend, with obvious differences across regions. Therefore, CEs reduction policies should be formulated based on a per capita perspective and regional differences. The PCEs also showed obvious spatial clustering, with hot spots mainly concentrated in Inner Mongolia and Xinjiang, while cold spots were mainly distributed in some provinces in southern China. Meanwhile, China’s low-carbon transition has achieved noteworthy outcomes, with the majority of prefecture-level cities (29.700%) having PCEs of less than 1 ton/persons. The overall PCE level in China remained stable despite the dynamic changes observed over the ten-year period. In particular, the shift of the PCEs of eastern and central China to higher values was obvious. The regression analysis indicated that China’s PCEs undergone changes due to a confluence of factors. The LUI and latitude had negative effects on PCEs in prefecture-level cities, and regional economic development, elevation, slope, and longitude had positive effects. The results can serve as a basis for developing CEs policies, which are of strategic importance for promoting sustainable development and formulating rational CE reduction policies.

However, this study still has several limitations. First, we acknowledge that it is insufficient to use data from 2000, 2010, and 2020 to explore the distribution patterns, future trends, and driving factors of China’s PCEs. This is because the PCE calculations were based on data extracted from the bulletins of the fifth, sixth, and seventh national censuses, which covered a ten-year period. In future studies, we will use annual data to supplement or verify the findings of the studied ten-year period. Second, although this study verified the importance of natural geographical factors and socioeconomic factors in influencing the low-carbon transition, due to data availability limitations, other driving forces, such as the digital economy and trading policies, were not taking into account. In the future, we will consider more driving factors to analyze the mechanism.