Spatial-Temporal Evolution and Environmental Regulation Effects of Carbon Emissions in Shrinking and Growing Cities: Empirical Evidence from 272 Cities in China

Abstract

Shrinking and growing cities are categories of cities characterized by population loss or add, and the issue of carbon emissions in these cities is often neglected. Environmental regulation, as an important influence on carbon emissions, plays an important role in promoting the low-carbon transition in Chinese cities. This study focused on the carbon emissions of 272 cities in China from 2012–2021, constructed a comprehensive indicator to classify four city types, and calculated carbon emissions. Spatial-temporal characteristics and evolution of carbon emissions and impacts of environmental regulation were investigated. Carbon emissions of rapidly growing cities showed a downward trend, whereas those of slightly growing, rapidly shrinking, and slightly shrinking cities showed upward trends. The more rapidly a city grew or shrunk, the higher its average carbon emissions. Growing cities’ center of gravity of their carbon emissions migrated northwest. Carbon emissions of rapidly and slightly shrinking cities were high in the northeast, and their carbon emission centers migrated northeast and southwest, respectively, with obvious spatial autocorrelation of city types. Strengthening environmental regulations significantly positively affected carbon emission reduction. The impact of environmental regulation on carbon emissions reduction was temporally and spatially heterogeneous and more significant in non-resource cities.

1. Introduction

China, being the largest developing country globally, has produced substantial carbon emissions throughout its rapid growth. With the goal of addressing climate change and promoting sustainable development, the Chinese government has set a target to reach peak carbon emissions by 2030 and achieve carbon neutrality by 2060 [1]. Existing studies show that in China, more than 80% of carbon emissions come from cities [2], so controlling urban carbon emissions is the key to achieving carbon reduction targets. As China faces uneven regional economic development and urbanization, urban growth and contraction occur in certain regions [3,4], the issue of carbon emissions in these cities is often neglected. So it is necessary to explore the carbon emission characteristics and carbon reduction paths of shrinking and growing cities to realize the “dual carbon” goal. Environmental regulation, as an important influence on carbon emissions, plays an important role in promoting low-carbon development in Chinese cities, and the environmental regulation system has become complex and multilayered after almost 50 years of development [5]. Strengthening environmental regulation has become one of the most important ways to promote low-carbon development in Chinese cities. Therefore, the primary goal of this study was to clarify the spatial-temporal evolution characteristics and environmental regulation effects of carbon emissions in shrinking and growing cities.

Under the dual-carbon background, studies on urban carbon emission measurement methods, spatial–temporal evolution characteristics, and related influencing factors have received extensive attention from the academic community. Urban carbon emissions have been measured using various methods, including the STIRPAT model, Kaya identity [6], building carbon model, LMDI method [7], four-part methodology, CGE model, hybrid analysis method, FFDAS model, GRID model, four-part methodology, building carbon model, and EC carbon flux observation. A widely used method is the inventory method, which uses energy consumption inventories to calculate carbon emissions [8]. For example, Fan et al. [9] used energy consumption to calculate carbon emissions data for 30 provinces in China from 2003–2019. Shan et al. [10] measured carbon emissions for 287 cities in China from 2001–2019 using an inventory of energy-related emissions that included 17 fossil fuels. However, inventories of urban energy consumption often suffer from different calibers and lack of statistical data. Existing studies have shown that a strong correlation between the nighttime lighting index and carbon emissions [11]. The Nighttime Lighting Index has become an effective method of accounting for urban carbon emissions.

For example, Chen et al. [12] used NPP/VIIRS satellite imagery to estimate the carbon emissions of 2735 counties in China from 1997 to 2017. Yang et al. [13] used a combination of nighttime lighting index and energy consumption data to calculate urban carbon emissions.

Some Scholars have also analyzed the spatial and temporal evolution characteristics of carbon emissions. For example, Xu et al. [14] analyzed the spatial and temporal characteristics of carbon emissions from international shipping in 19 coastal countries of the European Union, and found positive spatial correlation and spatial clustering of their carbon emissions. Xia et al. [15] examined rural China’s carbon emissions intensity, regional differences, and dynamic evolution trends using Gini coefficient decomposition and kernel density estimation methods. A Tapio decoupling index was used by Kang et al. [16] to examine the decoupling relationship between urban carbon emissions and economic growth and analyze the spatial and temporal evolution characteristics of carbon emissions and the decoupling index systematically. These studies demonstrate a spatial and temporal heterogeneity in carbon emissions.

The influencing factors of carbon emissions are complex and diverse, and many scholars have carried out extensive research. Pan et al. [17], According to the extended STIRPAT model examined a number of factors affecting carbon emissions in the United States, including the per capita gross domestic product (GDP), the urban population, and the merchandise trade in terms of GDP, all of these factors were found to have a significant effect on carbon emissions. Wu et al. [18] found that the factors affecting carbon emissions in China include energy structure, energy emission intensity, energy efficiency, economic development, technological progress, industrial structure, and environmental regulations [19]. Guo et al. [20] studied the driving factors of carbon emissions in China’s resource-based cities and found that factors such as population size, level of economic development, carbon abatement technology, and the proportion of resource-based industries all have an impact on carbon emissions. It is clear that the factors affecting carbon emissions are complex and varied. As one of the factors influencing carbon emissions, the role of environmental regulation in urban carbon reduction is becoming increasingly important. Pei et al. [21] found that environmental regulation has the potential to reduce carbon emissions either directly or indirectly through technological efficiency of energy-intensive industries. Zhang et al. [22] found that strict environmental regulation in China is positively correlated with CO₂ emission efficiency and can promote development of a low-carbon economy. Wu et al. [18] noted that countries can reduce carbon emissions by strengthening their environmental protection. It is thus clear that environmental regulation plays a crucial role in controlling carbon emissions and promoting low-carbon development.

As a result of uneven economic and social development, cities have undergone differentiated development, with urban shrinkage or growth occurring worldwide. A distinctive feature of shrinking cities is sustained population decline [23]. For example, Wolff et al. [24] showed that 20% of shrinking cities in Europe underwent shrinkage from 1990–2010. Ribant et al. [25] classified 367 shrinking central cities in the United States to provide a new perspective for addressing population loss. The study of urban growth has also shown a new direction; for example, Yang et al. [26] analyzed the spatial and temporal changes in global urban growth from 1995 to 2015 using remote sensing imagery data and found that urban growth remains in the typical stage of marginal growth. Many scholars have taken different approaches to identify shrinking and growing cities, with the majority using population growth and decline data. Li et al. [27] used demographic and economic indicators to identify shrinking cities. Other methods have also been used to identify shrinking cities; for example, Tan et al. [28] used the nighttime lighting index to identify 36 shrinking prefectural-level cities and 644 county-level cities in the Yangtze River Economic Belt. Tong et al. [29] used a composite index that incorporates demographic, economic, and social development to identify shrinking cities in China. Overall, population is the most critical indicator in identifying growing and shrinking cities.

Research already conducted has shown that urban growth and shrinkage have a significant impact on carbon emissions [30,31]. Hwang et al. [32] examined climate change in growing and shrinking cities in the post-industrial Rust Belt region of the United States and found that both types of city temperatures are generally increasing but that shrinking cities are growing slightly less. Carter et al. [33] study focused on Greater Manchester in North West England and found that shrinking cities face a different set of issues, including climate change, than other cities. In China, the impact of changes in urban form on carbon emissions is even more pronounced. For example, Huang et al. [34] studied the mechanisms by which socio-economic factors and urban form affect carbon emissions in growing and shrinking cities, using 285 prefecture-level cities in China as examples. The results show that shrinking cities have a continuing upward trend in emissions due to population loss and spatial expansion, while growing cities with an irrational urban structure can also exacerbate carbon emissions. Liu et al. [35] analyzed the relationship between urban form and residential carbon emissions at different city levels in China. It turns out that shrinking cities tend to be less energy efficient than growing cities, meaning that these cities not only face shrinking populations and economies but also generate higher carbon emissions and serious environmental problems. These studies show that urban growth and shrinkage have a considerable impact on carbon emissions, even if each has a different focus.

Previous studies have extensively discussed carbon emissions from shrinking and growing cities; however, some shortcomings remain. First, population indicators are mostly used to identify growing and shrinking cities; whereas, energy consumption indicators are mostly used to measure carbon emissions, which are vulnerable to the influence of other factors and lead to inaccuracies. Second, when analyzing the spatial and temporal distribution of carbon emissions, researchers are less likely to categorize cities, ignoring the potential differences between different city types, and there has been insufficient research on the trend of the shift in the center of gravity of carbon emissions. Finally, many scholars have used multiple influencing factors when studying carbon emission drivers; however, the study of individual drivers has not been in-depth.

The innovation of this study is reflected threefold: first, an urban development index was used to identify growing and shrinking cities. A composite indicator is used to measure carbon emissions by combining energy consumption with the NPP/VIIRS nighttime lighting index, thus accurately explaining urban carbon emissions and compensating for the different calibers of urban energy consumption and the lack of statistical data. Second, a detailed distinction was made between growing and shrinking cities, and the spatial and temporal distribution characteristics of carbon emissions as well as the trend of the center of gravity shift of the four city types and their differences were analyzed. Finally, the impacts of environmental regulation on carbon emissions of the four city types were analyzed in depth, and their temporal heterogeneity was analyzed to provide targeted references for formulating carbon emission reduction policies. This study enriches the research on a single driver of carbon emissions with greater targeting.

2. Materials and Methods

2.1. Study Scope and Duration

In this study, 272 prefecture-level cities in China were chosen as the study area. Measurements using the urban development degree index (UDD) showed that 72 prefecture-level cities in China contracted during the period from 2012–2021, including 27 RSCs and 43 SSCs. In addition, 204 prefectural-level cities in China grew from 2012 to 2021, including 63 RGCs and 139 SGCs (Figure 1).

2.2. Research Methods

The research framework of this study is shown in Figure 2.

2.2.1. Identification of Shrinking and Growing Cities

There are two dominant academic approaches for defining growing and shrinking cities. One is to look at the population change over a certain time period, whereby a city with a population increase over a certain period is identified as a growing city and that with a population decrease is identified as a shrinking city. Another approach is to consider temporal changes in the nighttime lighting index. When nighttime lights become brighter, the city is considered to be growing, and vice versa. However, both methods have shortcomings. For example, population classification can only reflect population changes within a certain time period. Although the nighttime lighting index can reflect population and economic changes, the data lacks a temporal connection; Therefore, many scholars have constructed comprehensive indices to identify shrinking and growing cities. Tong et al. selected nine indicators from three aspects (i.e., population, economy, and society) to construct an UDD to identify growing and shrinking cities [29]. The present study used the UDD to identify shrinking and growing cities.

Calculating the UDD requires demographic, economic, and social indicators, as detailed in

Table 1. Urban development degree indicators selection.

Dimensions	Indicators	Unit
	Growth rate of the natural population	‰
Population	total population	104 Person
	Population density	person/km2
	GDP per capita	Yuan
Economic	Per capita fiscal revenue	person/yuan
	GDP growth rate	%
	Total sales of consumer items at retail	104 yuan
Social and Land use	Fiscal expenditures per capita	person/yuan
	Built-up area	km2

Note: GDP—gross domestic product.

.

Equations (1)–(4) show the calculation process for UDD; when ${\mathrm{X}}_{\mathrm{i}\mathrm{j}}$ is positive:

$${\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }={\displaystyle \frac{{\mathrm{X}}_{\mathrm{i}\mathrm{j}}-\mathrm{m}\mathrm{i}\mathrm{n}{\mathrm{X}}_{\mathrm{j}}}{\mathrm{m}\mathrm{a}\mathrm{x}{\mathrm{X}}_{\mathrm{j}}-\mathrm{m}\mathrm{i}\mathrm{n}{\mathrm{X}}_{\mathrm{j}}}},$$

(1)

and when ${\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }$ is negative,

$${\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }={\displaystyle \frac{\mathrm{m}\mathrm{a}\mathrm{x}{\mathrm{X}}_{\mathrm{j}}-{\mathrm{X}}_{\mathrm{i}\mathrm{j}}}{\mathrm{m}\mathrm{a}\mathrm{x}{\mathrm{X}}_{\mathrm{j}}-\mathrm{m}\mathrm{i}\mathrm{n}{\mathrm{X}}_{\mathrm{j}}}},$$

(2)

$$\begin{gathered} {\mathrm{Y}}_{\mathrm{i}\mathrm{j}}={\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }/{\sum }_{\mathrm{i}=1}^{\mathrm{m}}\hspace{0.17em}{\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }, \\ {\mathrm{e}}_{\mathrm{j}}=-\mathrm{k}{\sum }_{\mathrm{i}=1}^{\mathrm{m}}\hspace{0.17em}({\mathrm{Y}}_{\mathrm{i}\mathrm{j}}\times \mathrm{l}\mathrm{n}{\mathrm{Y}}_{\mathrm{i}\mathrm{j}}),\mathrm{k}={\displaystyle \frac{1}{\mathrm{l}\mathrm{n}(\mathrm{m})}}, \\ {\mathrm{d}}_{\mathrm{j}}=1-{\mathrm{e}}_{\mathrm{j}}, \\ {\mathrm{W}}_{\mathrm{j}}={\mathrm{d}}_{\mathrm{j}}/{\sum }_{\mathrm{j}=1}^{\mathrm{n}}\hspace{0.17em}{\mathrm{d}}_{\mathrm{j}}, \end{gathered}$$

(3)

$$\begin{gathered} {\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}}_{\mathrm{i}\mathrm{t}}={\sum }_{\mathrm{j}=1}^{\mathrm{n}}\hspace{0.17em}({\mathrm{W}}_{\mathrm{j}}\times {\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }), \mathrm{a}\mathrm{n}\mathrm{d} \\ {\mathrm{U}\mathrm{D}\mathrm{D}}_{(\mathrm{i}{\mathrm{t}}_0,\mathrm{i}{\mathrm{t}}_1)}={\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}}_{\mathrm{i}{\mathrm{t}}_1}-{\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}}_{\mathrm{i}{\mathrm{t}}_0}, \end{gathered}$$

(4)

where i represents the city, j represents the indicator, t represents the year, n represents the total number of indicators, and m represents the total number of cities. The original data is normalized using the procedures outlined in Equations (1) and (2). ${\mathrm{X}}_{\mathrm{i}\mathrm{j}}$ and ${\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }$ represent the original and standardized data of the j index of city i, respectively. ${\mathrm{W}}_{\mathrm{j}}$ is the weight of the index, and Equation (3) shows the process of calculating ${\mathrm{W}}_{\mathrm{j}}$. Equation (4) provides the calculation process for the UDD; index represents the urban development degree index, and ${\mathrm{U}\mathrm{D}\mathrm{D}}_{(\mathrm{i}{\mathrm{t}}_0,\mathrm{i}{\mathrm{t}}_1)}$ represents the UDD from ${\mathrm{t}}_0$ to ${\mathrm{t}}_1$.

Based on research, we divided cities into the following four groups: (1) rapidly growing cities (RGCs), UDD ≥ 0.05; (2) slightly growing cities (SGCs), 0 ≤ UDD < 0.05; (3) rapidly shrinking cities (RSCs), UDD < −0.02; (4) slightly shrinking cities (SSCs), −0.02 ≤ UDD < 0.

2.2.2. Urban Carbon Emission Calculations

Currently, two methods are commonly used to measure urban carbon emissions: one is based on energy consumption data and the other on the nighttime lighting index. Existing studies have shown that the nighttime lighting index can accurately explain urban carbon emissions and compensate for the problems of different calibers of urban energy consumption and the lack of statistical data. The NPP/VIIRS nighttime lighting index has the advantages of high precision and stability. Therefore, this study combined these two methods to measure carbon emissions.

This study referred to the existing method of calculating sample city carbon emissions presented by Yang et al. [13]. First, nighttime lighting remote-sensing images of the study area were cropped, corrected, and reprojected using ArcGIS10.7 software to extract the gray values of nighttime lighting from 2012–2021 (DN). Then, the mathematical relationship between provincial energy consumption carbon emissions and provincial DN values in China was constructed, as shown in Equation (5). Finally, the carbon emission data of 63 RGCs, 139 SGCs, 27 RSCs, and 43 SSCs in China from 2012–2021 were calculated using the DN values of the cities, as shown in Equation (6). Carbon emission factors of various energy sources are shown in

Table 2. Conversion factor of standard coal and carbon emission factor for different energy types.

Fuel	Discount Factor for Standard Coal (t Standard Coal/t)	Carbon Emission Coefficient (104 t Carbon/104 t Standard Coal)
raw coal	0.7143	0.7559
coke	0.9714	0.855
crude oil	1.4286	0.5857
gasoline	1.4714	0.5538
kerosene	1.4714	0.5714
diesel	1.4571	0.5921
fuel oil	1.4286	0.6185
natural gas	1.33	0.4483
heat	34.12	0.67
electricity	/	0.272

Note: Based on the Intergovernmental Panel on Climate Change’s data (http://www.ipcc.ch (accessed on 10 March 2024)).

. The provincial carbon emissions from energy consumption were calculated using Equation (7):

$$\mathrm{C}\mathrm{E}=\mathrm{n}\times \mathrm{D}\mathrm{N},$$

(5)

where CE represents provincial carbon emissions, n is the mathematical relationship between the provincial carbon emissions and DN values, and DN represents Provincial nighttime light grey scale values.

$${\mathrm{C}\mathrm{E}}_{\mathrm{i}\mathrm{t}}=\mathrm{n}\times {\mathrm{D}\mathrm{N}}_{\mathrm{i}\mathrm{t}},$$

(6)

where ${\mathrm{C}\mathrm{E}}_{\mathrm{i}\mathrm{t}}$ represents city carbon emissions, n is the mathematical relationship between the provincial carbon emissions and Provincial DN values, and ${\mathrm{D}\mathrm{N}}_{\mathrm{i}\mathrm{t}}$ represents city nighttime light grey scale values, i represents the city, t represents the year.

$$\mathrm{C}\mathrm{E}={\displaystyle \frac{44}{12}}\times {\sum }_{\mathrm{i}=1}^{10}\hspace{0.17em}{\mathrm{K}}_{\mathrm{i}}{\mathrm{E}}_{\mathrm{i}}$$

(7)

Here, CE denotes carbon emissions, K denotes the energy carbon emission factor, E denotes energy consumption, and i denotes the energy type.

2.2.3. Spatial Analysis of Carbon Emissions

• Moran’s index

Detecting global spatial autocorrelation characteristics of carbon emissions can be achieved with Moran’s index (Moran’s I); therefore, it was chosen to test the global spatial autocorrelation characteristics of carbon emissions in China’s growing and shrinking cities. The calculation of Moran’s I and the method of significance testing are shown in Equation (8):

$$\begin{gathered} \mathrm{M}\mathrm{o}\mathrm{r}\mathrm{a}{\mathrm{n}}^{\prime }\mathrm{s} \mathrm{I}={\sum }_{\mathrm{i}=1}^{\mathrm{n}}\hspace{0.17em}{\sum }_{\mathrm{j}=1}^{\mathrm{n}}\hspace{0.17em}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}\left({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}}\right)\left({\mathrm{x}}_{\mathrm{j}}-\overline{\mathrm{x}}\right)/{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\hspace{0.17em}{\sum }_{\mathrm{j}=1}^{\mathrm{n}}\hspace{0.17em}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\hspace{0.17em}({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}}{)}^2,\mathrm{a}\mathrm{n}\mathrm{d} \\ \mathrm{P}={\displaystyle \frac{\mathrm{I}-\mathrm{E}\left(\mathrm{I}\right)}{\sqrt{\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{I}\right)}}}\stackrel{\mathrm{d}}{\to }\mathrm{N}(0,1) \end{gathered}$$

(8)

where n is the number of cities; ${\mathrm{x}}_{\mathrm{i}}$ and ${\mathrm{x}}_{\mathrm{j}}$ are the carbon emissions of cities i and j, respectively; $\overline{\mathrm{x}}$ denotes the mean value, and ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ denotes the weight. The P value denotes the level of significance; and E, Var, and N are the mathematical expectation, normal distribution, and variance, respectively.

• Hotspot analysis

The “Getis-Ord Index (G)” can distinguish the “hot” and “cold” areas of carbon emissions; therefore, the G index was chosen to test the local spatial autocorrelation characteristics of carbon emissions in the growing and shrinking cities in China and was calculated using Equation (9):

$$\mathrm{G}={\sum }_{\mathrm{j}\ne \mathrm{i}}^{\mathrm{n}}\hspace{0.17em}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}(\mathrm{d}){\mathrm{x}}_{\mathrm{j}}/{\sum }_{\mathrm{j}\ne \mathrm{i}}^{\mathrm{n}}\hspace{0.17em}{\mathrm{x}}_{\mathrm{j}},$$

(9)

where G is the hotspot; d denotes the designated radius; n is the number of growing and shrinking cities; ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ is the spatial connectivity matrix; i and j denote growing and shrinking cities, respectively, and x are carbon emissions.

• Standard deviation ellipse

Standard deviational ellipse (SDE) analysis is a spatial statistical method that can accurately reveal the spatial distribution of geographic elements in terms of center, dispersion, and directional trends. The calculation method is shown in Equation (10):

$$\begin{gathered} {\stackrel{-}{\mathrm{X}}}_{\mathrm{w}}={\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}}{\mathrm{x}}_{\mathrm{i}}/{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}}, \\ {\stackrel{-}{\mathrm{Y}}}_{\mathrm{w}}={\sum }_{\mathrm{i}=1}^{\mathrm{n}}\hspace{0.17em}{\mathrm{w}}_{\mathrm{i}}{\mathrm{y}}_{\mathrm{i}}/{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\hspace{0.17em}{\mathrm{w}}_{\mathrm{i}}, \\ {\mathsf{\sigma }}_{\mathrm{x}}=\sqrt{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\hspace{0.17em}({\mathrm{w}}_{\mathrm{i}}{\stackrel{-}{\mathrm{x}}}_{\mathrm{i}}\cos \mathsf{\theta }-{\mathrm{w}}_{\mathrm{i}}{\stackrel{-}{\mathrm{y}}}_{\mathrm{i}}\sin \mathsf{\theta })}/{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\hspace{0.17em}{\mathrm{w}}_{\mathrm{i}}^2, \mathrm{a}\mathrm{n}\mathrm{d} \\ {\mathsf{\sigma }}_{\mathrm{y}}=\sqrt{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\hspace{0.17em}({\mathrm{w}}_{\mathrm{i}}{\stackrel{-}{\mathrm{x}}}_{\mathrm{i}}\sin \mathsf{\theta }-{\mathrm{w}}_{\mathrm{i}}{\stackrel{-}{\mathrm{y}}}_{\mathrm{i}}\cos \mathsf{\theta })}/{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\hspace{0.17em}{\mathrm{w}}_{\mathrm{i}}^2, \end{gathered}$$

(10)

where ${\stackrel{-}{\mathrm{X}}}_{\mathrm{w}}$ and ${\stackrel{-}{\mathrm{Y}}}_{\mathrm{w}}$ denote the ellipse center; ${\mathsf{\sigma }}_{\mathrm{x}}$ and ${\mathsf{\sigma }}_{\mathrm{y}}$ represent the standard deviations of the two axes; ${\mathrm{x}}_{\mathrm{i}}$ and ${\mathrm{y}}_{\mathrm{i}}$ are the spatial coordinates of each element; ${\mathrm{w}}_{\mathrm{i}}$ is the weight; x and y represent the relative coordinates of each point from the center of the ellipse; θ means to the angle between the long axis and the true north of the ellipse; and n is the gross number of pixels.

2.2.4. Panel Data and Regression Model

Based on the influence of technological, economic, and demographic factors on the efficiency of carbon emissions, the fundamental formulation of the STIRPAT model was derived as shown in Equation (11):

$$\mathrm{I}=\mathsf{\mu }{\mathrm{T}}_{\mathrm{a}}{\mathrm{A}}_{\mathrm{b}}{\mathrm{P}}_{\mathrm{c}}\mathsf{\epsilon },$$

(11)

where I denotes the environment; T, A, and P are the technological advancement, affluence, and population size, respectively, and ε and μ are random error terms and constant, respectively. To eliminate the effect of heteroscedasticity as much as possible, the formula was converted to a logarithmic form, as shown in Equation (12).

$$\ln \mathrm{I}=\mathrm{u}+\mathrm{aln}\mathrm{T}+\mathrm{bln}\mathrm{A}+\mathrm{cln}\mathrm{P}+\mathsf{\epsilon }$$

(12)

Considering the influence of environmental regulation, industrial structure, foreign investment intensity, and other factors, the STIRPAT model was constructed as shown in Equation (13):

$$\begin{gathered} \mathrm{C}{\mathrm{E}}_{\mathrm{m}\mathrm{n}}={\mathsf{\mu }}_0+{\mathsf{\mu }}_1\ln \mathrm{E}{\mathrm{R}}_{\mathrm{m}\mathrm{n}}+{\mathsf{\mu }}_2\ln \mathrm{T}{\mathrm{A}}_{\mathrm{m}\mathrm{n}}+{\mathsf{\mu }}_3\ln \mathrm{E}{\mathrm{D}}_{\mathrm{m}\mathrm{n}}+{\mathsf{\mu }}_4\ln \mathrm{I}{\mathrm{S}}_{\mathrm{m}\mathrm{n}} \\ +{\mathsf{\mu }}_5\ln \mathrm{P}{\mathrm{Z}}_{\mathrm{m}\mathrm{n}}+{\mathsf{\mu }}_6\ln \mathrm{F}{\mathrm{D}}_{\mathrm{m}\mathrm{n}}+{\mathsf{\epsilon }}_{\mathrm{m}\mathrm{n}}, \end{gathered}$$

(13)

where CE, ER, TA, ED, IS, PZ, and FDI denote carbon emissions, environmental regulation, technological advancement, population size, economic development level, industrial structure, and foreign investment intensity, respectively; μ₁, μ₂, μ₃, μ₄, μ₅, and μ₆ are the corresponding elasticity coefficients, which denote the rate of change of carbon emissions caused by a 1% change in ER, TA, ED, IS, PZ, and FDI, respectively; m denotes the city, and n denotes the year.

This study selected carbon emissions as the explanatory variable, determined environmental regulation as the explanatory variable, and added technological advancement, economic development level, industrial structure, population size, and foreign investment intensity as control variables to comprehensively explain urban carbon emissions. Notably, the environmental regulation is an explanatory variable, using industrial wastewater, SO₂, and smoke (dust) emissions per unit of output value of three pollutant types as a basic variable. The entropy method is used to obtain a comprehensive indicator to characterize the intensity of environmental regulation, whereby the smaller the value, the greater the intensity of environmental regulation. These variables are presented in

Table 3. Calculations of variables required for the STIRPAT model.

Indicator Properties	Indicator (Variable Name)	Interpretation of Indicators
Explanatory variable	Carbon emissions (CE)	Energy consumption data combined with nighttime light index measurements
Explanatory variable	Environmental regulation (ER)	Industrial wastewater, SO2, and smoke (dust) per unit of output measured by the entropy weight method
Control variable	Technological advancement (TA)	Patent applications for inventions
	Economic development level (ED)	GDP per capita
	Industrial structure (IS)	Ratio of secondary sector output to GDP
	Population size (PZ)	Total population at end of the year
	Foreign investment intensity (FDI)	Foreign investment as a percentage of GDP for the year

.

2.3. Data Sources

In this study, NPP/VIIRS nighttime light raw images were obtained from the National Oceanic and Atmospheric Administration National Geophysical Data Center website (http://www.ngdc.noaa.gov/ (accessed on 15 March 2024)). The administrative vector boundaries at the Chinese and prefecture levels were based on the standard map GS(2020)4630 on the Ministry of Natural Resources website, with no modifications to the base map. ArcGIS 10.7 was used to extract the nighttime light DN values and construct geographic maps. Data on provincial energy consumption were gathered from the China Energy Statistical Yearbook 2012–2021.

The demographic, economic, and social data used to construct the UDDs as well as the data used to compute the STIRPAT model were obtained from the China Urban Statistical Yearbook. The STIRPAT model data were analyzed using Stata17 software. The number of invention patent applications was obtained from the Patent Search and Analysis System of the State Intellectual Property Office of China (http://pss-system.cnipa.gov.cn/ (accessed on 20 March 2024)). Missing data were supplemented using linear interpolation.

3. Results

3.1. Temporal Trends in Carbon Emissions in Growing and Shrinking Cities

The temporal trend of gross carbon emissions in growing and shrinking cities in China from 2012–2021 is shown in Figure 3. As shown in Figure 3a, the total carbon emissions of 63 RGCs displayed a downward trend; and 139 SGCs, 27 RSCs, and 43 SSCs in China all showed an upward trend in carbon emissions. Among them, the total carbon emissions of SGCs and SSCs showed a fluctuating upward trend.

As shown in Figure 3b, judging from the time trend of the average carbon emissions from 2012–2021, the average carbon emission distribution characteristics of the four city types decreased as follows: RGCs > RSCs > SGCs > SSCs. This showed that the more rapidly a city grew or shrunk, the higher its average carbon emissions, and at the same rate of urban change, the average carbon emissions of growing cities will be higher than those of shrinking cities.

3.2. Spatial Distribution Characteristics of Carbon Emissions in Growing and Shrinking Cities

Figure 4 illustrates spatial patterns of carbon emissions in growing and shrinking cities. Here, carbon emissions were graded as low, medium, or high in four city groups using the ArcGIS software’s natural bond breakpoint grading system. Figure 4 shows that among the growing cities, high-carbon emission areas in RGCs were mainly concentrated in the eastern coastal areas, whereas high-carbon emission areas in SGCs are mostly concentrated in North China, such as Beijing, Tianjin, and some cities south of Shenyang. From this, we found that high carbon emission areas in growing cities were mainly concentrated in more developed cities on the eastern coast and in North China. Among shrinking cities, carbon emissions in both RSCs and SSCs were high in the northeast and low in the central and southern regions. It can be seen that the carbon emissions of the four types of cities in China had strong regional heterogeneity.

Table 4. Moran index of carbon emissions from 2012–2021.

	RGCs		SGCs		RSCs		SSCs
Year	Moran I	Z	Moran I	Z	Moran I	Z	Moran I	Z
2012	0.04 *	1.41	0.15 ***	9.37	0.11 *	1.16	0.26 ***	3.51
2015	0.10 ***	2.86	0.11 ***	7.37	0.14 *	1.38	0.19 ***	2.75
2018	0.09 ***	2.67	0.09 ***	6.18	0.15 *	1.47	0.18 ***	2.69
2021	0.08 ***	2.52	0.06 ***	4.50	0.10 *	1.08	0.14 ***	2.20

Note: RGCs–rapidly growing cities, SGCs, slightly growing cities, RSCs–rapidly shrinking cities, SSCs–slightly shrinking cities. *** p < 0.01, * p < 0.10.

shows Moran’s I values of carbon emissions in RGCs, SGCs, RSCs, and SSCs in China from 2012–2021. The Moran’s I values of carbon emissions in the four groups of cities in 2012, 2015, 2018, and 2021 were all significantly positive, indicating that carbon emissions had a significant positive spatial correlation. The Moran’s I values of the four groups of cities showed a decreasing trend, indicating that the concentration trend of carbon emissions gradually decreased over time.

The hot and cold spot distribution features of carbon emissions in RGCs, SGCs, RSCs, and SSCs are displayed in Figure 5. Carbon emission hotspot areas of rapidly shrinking and growing cities were concentrated in the eastern coastal areas (i.e., mainly in Shandong and Jiangsu provinces). Carbon emission hotspots of SSCs were concentrated in North China (i.e., mainly in Hebei, Shandong, Shanxi, and other provinces), and cold spots were located in South China (i.e., concentrated in Hunan, Jiangxi, Fujian, and other provinces). The carbon emission hotspots of RSCs were located in Northeast China, whereas those of SSCs were mostly located in Northeast China and some were located in North China. The carbon emissions of RGCs, SGCs, RSCs, and SSCs in China exhibited local spatial autocorrelation characteristics.

3.3. SDEs and Center of Gravity Migration Paths of Carbon Emissions for Growing and Shrinking Cities

The SDEs and center of gravity migration paths of carbon emissions for the four city types are shown in Figure 6. The direction of the ellipse tilt showed that the standard deviation ellipses of carbon emissions in RGCs, RSCs, and SSCs all showed a northeast–southwest direction, and those in SGCs showed a southeast–northwest direction.

The center of gravity of RGCs showed a migration trend of southeast–northeast–northwest, and generally moved to the northwest; the center of gravity of SGCs showed a migration trend of northwest–southwest–northwest and generally moved to the northwest; the center of gravity of a RSCs showed a migration trend of northwest–northeast–northeast and generally moved to the northeast; and the center of gravity of SSCs showed a migration trend of northeast–northeast–southwest–southwest and generally moved to the southwest. The carbon emission center of gravity of growing cities (i.e., RGCs and SGCs) had obviously shifted to the northwest, the carbon emission center of RSCs obviously moved to the northeast, and the carbon emission center of SGCs obviously shifted to the southwest. This indicated that the centers of gravity of the carbon emissions of the four city types were spatially unstable and constantly changing.

3.4. Effect of Environmental Regulation on Carbon Emissions in Growing and Shrinking Cities in China

3.4.1. F-Test and Hausman Test

Based on the F-test and Hausman test results, the RGCS, SGCs, RSCs, and SSCS were regressed using a fixed-effects model (

Table 5. F-test and Hausman test of four city groups.

City Groups	F-Test		Hausman Test
	F-Value	Prob	Shi-Sq. Statistic	Prob
RGCS	F = 5.804	0.000	44.11	0.000
RGCS	F = 8.638	0.000	141.536	0.000
RSCS	F = 1.637	0.021	21.339	0.003
SSCS	F = 3.810	0.000	56.336	0.000

). Considering that the data used changed temporally and individually, the regression results were analyzed using a two-way fixed-effects model (Fe-tw).

3.4.2. Linear Regression Results on the Impact of Environmental Regulations on Four City Types

As shown in

Table 6. Regression results of the impact of environmental regulations on four types of cities.

Explanatory Variables	RGCS	SGCS	RSCS	SSCS
ER	0.018 **	0.011 **	0.029 **	0.016 **
	(2.103)	(1.980)	(2.528)	(1.884)
TA	0.017	0.035 ***	0.021	0.030
	(1.007)	(3.134)	(0.730)	(1.618)
ED	0.182 ***	0.079 **	−0.058	−0.042
	(3.603)	(2.430)	(−1.014)	(−0.808)
IS	0.026	0.188 ***	0.046	0.040
	(0.332)	(4.052)	(0.528)	(0.507)
PZ	−0.389 ***	0.004	0.270	0.121
	(−3.412)	(0.049)	(0.964)	(0.849)
FDI	0.011	0.008	0.018 *	0.019 **
	(1.209)	(1.609)	(1.905)	(2.120)
_cons	11.071 ***	8.611 ***	7.141 ***	8.475 ***
	(12.974)	(12.728)	(3.625)	(8.434)
N	630	1390	270	430
R2	0.976	0.968	0.975	0.971
F	289.946	245.268	214.631	218.604

ER–environmental regulation, TA–technological advancement, ED–economic development level, IS–industrial structure, PZ–population size, FDI–foreign investment intensity. *** p < 0.01, ** p < 0.05, * p < 0.10.

, the regression results indicated that the coefficients of the influence of industrial wastewater, SO₂, and smoke (dust) emissions per unit of output value measured by the entropy weighting method on carbon emissions in the four city types ranged from 0.011–0.029 and were all significant (p < 0.05). This suggested that the increase in environmental regulation significantly reduced carbon emissions from RGCS, SGCs, RSCs, and SSCS. In addition to environmental regulations, other factors also affected carbon emissions. In RGCs, GDP growth led to an increase in carbon emissions, and population growth at the end of the year led to a decrease in carbon emissions. In SGCs, growth in the output value of the secondary industry, invention patents, and GDP all led to increased carbon emissions. In RSCs and SSCs, in addition to the impact of environmental regulations, an increase in the level of foreign investment also led to increased carbon emissions.

3.4.3. Heterogeneity Regression Results of the Impact of Environmental Regulation on Four City Types

To explore whether environmental regulations’ effects on carbon emissions changed after the State Council issued the “Thirteenth Five-Year Plan for Ecological Environmental Protection” in 2016, the study took this as the time cut-off point, divided the sample into two time periods, and selected the regression results of the Fe-tw model based on the Hausman test to be analyzed (

Table 7. Temporal regression results of the impact of environmental regulations on carbon emissions in four city types.

Variables	2012–2016				2017–2021
	RGCS	SGCS	RSCS	SSCS	RGCS	SGCS	RSCS	SSCS
ER	−0.016	0.010	−0.005	0.011	0.042 ***	0.014 ***	0.040 ***	0.041 ***
	(−1.015)	(1.036)	(−0.245)	(0.723)	(3.743)	(2.026)	(2.636)	(3.741)
Control Variables	YES	YES	YES	YES	YES	YES	YES	YES
cons	YES	YES	YES	YES	YES	YES	YES	YES
City	YES	YES	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES	YES	YES
N	315	690	135	215	315	690	135	215
R2	0.982	0.973	0.979	0.977	0.992	0.989	0.991	0.988
F	181.834	133.047	128.099	132.047	394.583	336.699	283.976	262.046

Note: *** p < 0.01.

). The effect of environmental regulation on carbon emissions in the four city types in 2012–2016 was not significant, whereas the coefficient of the effect of environmental regulation on carbon emissions between 2017–2021 ranged from 0.014–0.042 and was significant (p < 0.01). This indicated that the inhibitory effect of environmental regulation on carbon emissions was gradually highlighted, and the strength of environmental regulation gradually increased.

To elucidate the impact of different regional environmental regulation levels on urban carbon emissions, the Fe-tw model was chosen based on the Hausman test, and a regression was performed on the panel data of the four city types in the eastern, central, and western regions (

Table 8. Subregional regression results of the impact of environmental regulations on carbon emissions in four city types.

Variables	Eastern Region				Central Region				Western Region
	RGCS	SGCS	RSCS	SSCS	RGCS	SGCS	RSCS	SSCS	RGCS	SGCS	RSCS	SSCS
ER	0.023 **	0.023 **	−0.009	−0.023	0.026	0.003	−0.027	0.003	−0.077 **	−0.000	−0.016	0.031 *
	(2.488)	(2.340)	(−0.815)	(−0.797)	(1.379)	(0.368)	(−1.121)	(0.316)	(−2.228)	(−0.027)	(−0.388)	(1.867)
Control Variables	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
cons	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
City	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
N	368	420	110	80	180	500	90	200	80	460	70	150
R2	0.983	0.974	0.986	0.978	0.967	0.956	0.962	0.966	0.986	0.972	0.987	0.979
F	359.461	239.536	244.896	117.102	134.301	148.733	72.070	139.967	176.400	228.545	179.355	195.128

Note: ** p < 0.05, * p < 0.10.

). In the eastern region, environmental regulation was positively correlated with carbon emissions in RGCs and SGCs, and strengthening environmental regulation had an inhibiting influence on carbon emissions, whereas the impact on RSCs and SSCs was not obvious. In the central region, the influence of environmental regulation on the carbon emissions of the four city types was not significant. In the western region, environmental regulation was negatively correlated with carbon emissions in RGCs and positively correlated with carbon emissions in SSCs. In the western region, environmental regulations were negatively correlated to carbon emissions in RGCs and positively correlated to carbon emissions in SSCs, whereas the effects on SGCS and RSCs were not significant.

The four city types were classified as resource-type and non-resource-type cities, and the Fe-tw model was selected according to the Hausman test to investigate the influence of environmental regulatory capacity on carbon emissions in cities with different characteristics (

Table 9. Regression results of the influence of environmental regulations on carbon emissions in four types of cities by city type.

Variables	Resource-Based City				Non-Resource-Based City
	RGCS	SGCS	RSCS	SSCS	RGCS	SGCS	RSCS	SSCS
ER	0.000	0.005	0.004	0.035 ***	0.025 **	0.019 ***	0.042 ***	0.034 **
	(0.002)	(0.633)	(0.161)	(3.386)	(2.529)	(2.744)	(3.030)	(−2.393)
Control Variables	YES	YES	YES	YES	YES	YES	YES	YES
cons	YES	YES	YES	YES	YES	YES	YES	YES
City	YES	YES	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES	YES	YES
N	110	570	150	260	518	810	120	170
R2	0.974	0.969	0.971	0.962	0.974	0.970	0.983	0.984
F	127.387	222.595	138.393	139.292	255.351	243.157	212.548	275.669

Note: *** p < 0.01, ** p < 0.05.

). In resource-based cities, the effects of environmental regulations on carbon emissions were not significant in RGCs, SGCs, and RSCs, whereas environmental regulations in SSCs significantly positively affected carbon emissions. Among non-resource cities, environmental regulations in RGCs, SGCs, and RSCs had significant positive impacts on urban carbon emissions, whereas environmental regulations in SSCs were negatively correlated with urban carbon emissions.

3.4.4. Robustness Test

Consider the possibility that there may be omitted variables in the model construction or the two-way causality that usually exists between variables. This could lead to an endogeneity problem, which could, in turn, result in instability in the model regression results [36]. To guarantee the dependability of the findings and to tackle the issue of endogeneity, this study employs environmental regulation variables lagged by one period as instrumental variables and regresses the four categories of cities using the two-stage least squares (2SLS) method, as illustrated in

Table 10. Robustness test results.

Explanatory Variables	RGCS	SGCS	RSCS	SSCS
ER	0.113 ***	0.1626 ***	0.1039 ***	0.2241 ***
	(0.015)	(0.014)	(0.027)	(0.018)
TA	−0.207	0.0919 ***	0.0296	0.0587 *
	(0.025)	(0.020)	(0.048)	(0.030)
ED	0.675 ***	0.7844 ***	0.6976	0.4219
	(0.063)	(0.056)	(0.105)	(0.091)
IS	0.144	1.1457 ***	0.2586 *	0.3430
	(0.102)	(0.077)	(0.152)	(0.111)
PZ	−0.763 ***	0.5204 *	0.3996	0.5157
	(0.045)	(0.041)	(0.093)	(0.055)
FDI	0.103 *	−0.0291	0.0285 ***	0.0483 **
	(0.019)	(0.012)	(0.026)	(0.020)
_cons	3.5271 ***	9.602 ***	3.7306 ***	2.6267 ***
	(0.408)	(13.565)	(0.720)	(0.603)
N	565	1242	243	387
R2	0.6436	0.479	0.524	0.578
F	309.824	248.286	272.68	552.05

Note: *** p < 0.01, ** p < 0.05, * p < 0.10.

. From the results, the nature of the effect of environmental regulation and the significance level are basically consistent with the original regression results, and the robustness test results are reliable.

4. Discussion

4.1. Characteristics of Spatial and Temporal Eevolution of Carbon Emissions in Growing and Shrinking Cities

China’s growing and shrinking cities have varying carbon emission characteristics. Temporally, total carbon emissions of RGCs showed a decreasing trend, and those of SGCs, RSCs, and SSCs showed an increasing trend. This finding is consistent with that reported by Yang et al. [13]. The study by Liu et al. [35] corroborates this assertion and elucidates the underlying causes of this phenomenon. Shrinking cities are less efficient in their energy use than other cities and not only face shrinking populations and economies but also environmental problems such as increasing carbon emissions. The study by Xiao et al. [31] revealed that RGCs undergo a more rapid transformation of industries and exhibit a higher energy efficiency level, which results in lower carbon emissions. This view is consistent with that of the present study. Zhu et al. [37] highlighted that urban form and structure directly impact carbon emissions, which can be effectively reduced by optimizing the urban form and improving energy efficiency. SGCs showed an upward trend in carbon emissions in the present study, which may have been related to challenges in industrial restructuring, population loss, and spatial expansion. For example, shrinking cities continue to have increased CO₂ emissions owing to secondary industry drivers, coexisting population loss, and spatial expansion [34]. Notably, the present study found that the more rapid the rate of urban growth or shrinkage, the higher the average carbon emissions. This may have been because rapidly changing urban environments lead to uncertainty in planning and management, thereby increasing the instability of carbon emissions. Conversely, slowly changing urban environments provide more time for adaptation and adjustment, contributing to more stable low-carbon development pathways [38].

Spatially, the high carbon emission zones of RGCs and SGCs were mainly concentrated in some of the more developed cities along the eastern seaboard and in northern China, corresponding to the results of Tong et al. [29]. This phenomenon was also verified and explained in the study by Fu et al. [39]. It may be related to the higher level of economic development in these regions, as increased economic activity is often accompanied by increased energy consumption and carbon emissions. In addition, land-use changes on the east coast and in North China may also have led to increased carbon emissions owing to their high urbanization and industrialization levels [40]. Both RSCs and SSCS showed high carbon emissions in the northeast and low emissions in the central and southern regions, which may have been related to economic restructuring, industrial upgrading, and improvements in energy use efficiency [41]. The study by Wu and Zhang et al. [42] also points out that Northeast China, as China’s old industrial base, has a particularly serious carbon emission problem in shrinking cities. These cities, as old industrial cities, tend to encounter problems such as population loss and resource depletion during the transition period, and thus, the problem of carbon emissions will be more serious. Carbon emissions of the four city types exhibited clear spatial autocorrelation characteristics, and the center of gravity was spatially unstable and constantly changing. This suggests that the geographical distribution of urban carbon emissions and the migration of the center of gravity is a dynamic process influenced by various factors, including economic development, industrial structure, energy consumption patterns, and policy orientation [37,43].

4.2. Impact of Environmental Regulation on Carbon Emissions in Growing and Shrinking Cities

Strengthening environmental regulation significantly positively affected carbon emission reduction in RGCs, SGCs, RSCs, and SSCs. This perspective was corroborated by the findings of the study conducted by L. Zhang et al. [44]. Furthermore, the study posited that environmental regulation exerts an influence on carbon emissions not only through direct means but also indirectly via pathways such as industrial structure, economic development, and technological innovations. The foreign investment level had an obvious driving action on the carbon emissions of RGCs and SSCs. Previous studies indicate foreign direct investment as an influential factor affecting the carbon emissions of cities [45]. This is because foreign investments aim to increase production, which increases energy consumption in shrinking cities and increases carbon emissions from natural gas production [46]. In their study, Zhang et al. [44] put forth the argument that GDP growth resulted in increased carbon emissions in RGCs and SGCs. Conversely, they found that end-of-year population growth led to a reduction in carbon emissions in RGCs. Population migration and economic development are important causes of changes in carbon emissions. Li et al. [47] argued that economic growth and increased energy intensity contribute to increased carbon emissions. In addition, an increase in secondary industry output value increased carbon emissions in SGCs. Carbon emissions were driven by the industrial structure of cities, similar to the findings of Tong et al. [29]. Therefore, it is necessary to upgrade industrial structures to reduce urban carbon emissions.

After the State Council published the “13th Five-Year Plan for Ecological and Environmental Protection” in 2016, the inhibitory influence of environmental regulation on carbon emissions was gradually highlighted and the strength of environmental regulation was gradually increased. This indicated strong temporal heterogeneity of environmental regulation on carbon emissions of RGCS, SGCs, RSCs, and SSCs. In the eastern region, the strengthening of environmental regulation dampened carbon emissions in RGCs and SGCs, whereas the effect on RSCs and SSCs was not notable. In the central region, the impact of environmental regulation on carbon emissions of the four city types was not significant. SGCs and RSCs were not significantly impacted by environmental regulation in the western region; environmental regulation negatively correlated with carbon emissions in RGCs and positively correlated with carbon emissions in SSCs. This suggested that there were differences in the effect of environmental regulation on carbon emissions in distinct regions owing to factors such as geographic location, economic development, policy support, and talent structure [48,49]. Moreover, improving environmental regulation in growing cities in the eastern region with stronger economic development significantly reduced carbon emissions, showing that economic development was closely related to the carbon emission reduction level of environmental regulation. The study by Song et al. [50] verifies this view; in addition, he points out that environmental regulation has a significant inhibiting effect on local carbon emissions but affects carbon emissions in neighboring areas.

Different city types have different environmental regulatory capacities and sustainable economic and social development levels; hence, there are also differences in the challenges and focuses that must be addressed in the process of low-carbon transition. In resource cities, the positive impacts of environmental regulation on carbon emissions in RGCS, SGCs, and RSCs were not significant, whereas environmental regulation in SSCs significantly positively impacted carbon emissions. In non-resource cities, environmental regulation in RGCS, SGCs, and RSCs significantly positively affected urban carbon emissions, whereas environmental regulation in SSCs negatively affected urban carbon emissions. Strengthening environmental governance is conducive to synergistic promotion of pollution and carbon reductions, indicating that the impacts of environmental regulation on carbon emissions differ significantly in different city types. This view is confirmed by a study by Wang et al. [51], which pointed out that both narrow and broad environmental regulations in the eastern and central regions showed significant “abatement and mitigation” effects, while in the western region, only narrow environmental regulations had a “reverse abatement” effect. The carbon emission decreases, and the effect of strengthening environmental governance in non-resource cities is more significant than in resource cities. This may have been because cities that are not resource-rich have more capacity to adjust their energy and industrial structures. Compared to resource cities, non-resource cities are more likely to reduce carbon emissions through technological innovation and industrial upgrades [52]. Finally, we can also consider a fresh perspective on technology in addressing urban carbon emissions. For example, revolutionizing sustainable energy production with quantum AI is a revolutionary technological approach that strongly dampens urban carbon emissions. Ajagekar et al. [53] similarly noted in their study that advances in quantum hardware and algorithms, quality control, and quantum artificial intelligence become promising tools for dealing with renewable and sustainable energy systems. Moreover, urban carbon emissions in coastal areas may be influenced by marine agricultural patterns. Therefore, the implementation of sustainable marine agriculture strategies will provide a new development model for coastal cities with high carbon emissions. This view is confirmed by the study of Del Valle et al. [54], in which it was noted that sustainable intensive mariculture can significantly improve coastal ecosystems.

4.3. Limitations and Directions for Improvement

There are some limitations to this study. First, although many counties and townships in China have experienced growth and shrinkage, current research is limited to the city level due to the availability of data. Second, when calculating carbon emissions, only carbon emissions from energy consumption combined with the nighttime lighting index are considered, ignoring carbon emissions from urban building activities. Example activities include cement production, which is also a significant source of urban carbon emissions. Finally, the impact of different types of environmental regulatory instruments on carbon emissions may be different. The impact of different types of environmental regulatory instruments on carbon emissions needs to be further analyzed due to the availability of data on environmental regulatory instruments and the study’s bias towards the heterogeneous impact of environmental regulation on carbon emissions in shrinking and growing cities.

To address these limitations, it is recommended that future research consider the following aspects. First, Counties and towns may also have larger carbon emission problems, and future research should focus on small towns. Second, when calculating carbon emissions, the carbon emissions generated by urban construction, as well as other activities, should be fully considered to improve the accuracy of the research data. Finally, when examining the impact of environmental regulation on carbon emissions, environmental regulation can be analyzed in categories, such as command and control and market-based instruments, to produce more nuanced results.

5. Conclusions and Policy Implications

To understand the carbon emissions of growing and shrinking cities, this study analyzed the carbon emissions of 63 RGCs, 139 SGCs, 27 RSCs, and 43 SSCs in China from v 2012–2021 by combining them with the NPP/VIIRS nighttime lighting index. Moran’s I, hotspot analysis, and SDE were used to explore the characteristics and evolution of spatio-temporal differentiation, and the influence of environmental regulation on carbon emissions was explored using the STIRPAT model. This study concludes with the following main conclusions:

In the aspect of the spatial and temporal distribution characteristics of carbon emissions, from 2012 to 2021, gross carbon emissions of RGCs in China displayed a descending trend, and those of SGCs, RSCs, and SSCs showed upward trends. The distribution of the average carbon emissions of the four city types decreased as follows: RGCs > RSCs > SGCs > SSCs. The more rapidly a city grew or shrunk, the higher the average carbon emissions. Spatially, High carbon emission areas of RGCs and SGCs were major focuses on the eastern coast and in some of the more developed cities in North China, and the center of gravity of their carbon emissions migrated to the northwest. Carbon emissions of RSCs and SSCs were high in the northeast and low in central and southern China. The center of gravity of carbon emissions of RSCs migrated to the northeast, and that of SGCs migrated to the southwest. Carbon emissions of the four city types exhibited clear spatial autocorrelation characteristics, and the center of gravity was spatially unstable and constantly changing.

In the aspect of the impact of environmental regulation on carbon emissions, strengthening environmental regulation had significantly positively affected carbon emission reduction in all four city types, and factors other than environmental regulation also affected carbon emissions. In RGCs, GDP growth led to higher carbon emissions, and year-end population growth led to lower carbon emissions. In SGCs, growth in the secondary sector’s output value, invention patents, and GDP all led to higher carbon emissions. In RSCs and SSCs, in addition to the influence of environmental regulation, an increase in foreign investment also led to increased carbon emissions. The inhibitory effects of environmental regulation on carbon emissions gradually increased after the State Council issued the “13th Five-Year Plan for Ecological and Environmental Protection.” The influence of environmental regulation on carbon emission reductions in the three major regions and different city types was heterogeneous.

The policy implications of this study are as follows. RGCS and SGCS, due to their sufficient human resources, can vigorously encourage the development of tertiary industry, reduce the proportion of secondary industry, and develop high-value-added industries. At the same time, RSCS and SSCS should focus on adjusting the capital structure of their investments, Regulating the behavior of foreign investment, and increasing subsidies for high-quality personnel to provide sufficient human capital for technological upgrading. In addition, the governments of all four types of cities should strengthen their environmental regulation efforts to curb carbon emissions through policy instruments supplemented by other instruments in a focused manner. For example, non-resource cities should focus on adopting policy-based environmental regulations to reduce carbon emissions, while non-resource cities should adopt measures such as the introduction of talent and industrial restructuring. Finally, each city should take appropriate measures to proactively regulate its rate of growth or contraction to respond to the call for high-quality development and to prevent carbon emissions from the outputs of rapid urban change.