Study on Spatial and Temporal Characteristics and Influencing Factors of Carbon Emissions in the Urban Agglomeration of the Middle Reaches of the Yangtze River

Abstract

The industrial transfer of heavy industries such as non-metallic mineral manufacturing, metal smelting and manufacturing from the eastern coast of China to the central region is beneficial to the economic development of the central region on the one hand, but increases carbon emissions in the central region on the other hand. In February 2022, the National Development and Reform Commission approved the “14th Five-Year Plan for the Development of the Urban Agglomeration in the Middle Reaches of the Yangtze River”. This indicates that the urban agglomeration of the middle reaches of the Yangtze River is an important region for implementing green development in the central area. The spatial and temporal evolution of carbon emissions and influencing factors in this region are the foundation for achieving carbon peaking and the carbon neutrality goal. This paper calculates the total carbon emissions of the cities in the urban agglomeration of the middle reaches of the Yangtze River and uses models such as spatial autocorrelation, geographically weighted regression, and Geodetector to explore the spatial–temporal pattern of carbon emissions. The results show the following: (1) The total carbon emissions of the middle reaches of the Yangtze River urban agglomeration showed fluctuations during 2010–2020, and the carbon emission reduction effect is unstable. Additionally, the carbon emissions of the middle reaches of the Yangtze River city cluster show obvious spatial variability, but the high carbon emission area is always concentrated in Wuhan, and this remains unchanged. (2) In 2010, 2014 and 2017, population size was the most important factor affecting carbon emission divergence, and in terms of interaction, the interaction between energy intensity and GDP and urbanization is the reason for the increasing carbon emissions. (3) The influence of population size on carbon emissions decreases from north to south, the influence of energy intensity on carbon emissions shows a spread from the most influential region in the northwest to the centre and then to the northeast, and the GDP per capita has little influence on the difference of carbon emissions spatial distribution.

1. Introduction

Since 2010, when the General Office of the State Council first issued the “Guidance Opinions of the State Council on Undertaking Industrial Transfer in Central and Western Regions”, the industrial transfer of heavy industries such as non-metallic mineral manufacturing, metal smelting, and manufacturing from the eastern coastal regions of China to the central region has benefited the economic development of the central region on the one hand, but has increased carbon emissions the central region on the other hand [1]. In 2015, the Chinese government proposed promoting the development of the urban agglomeration in the middle reaches of the Yangtze River, which was of great significance for accelerating the rise of central China and developing the Yangtze River Economic Belt [2]. In February 2022, the National Development and Reform Commission approved the “14th Five-Year Plan for the Development of the Urban Agglomeration in the Middle Reaches of the Yangtze River” [3]. However, most of the current research on urban agglomerations focuses on the most economically developed regions, such as the Yangtze River Delta and the Beijing–Tianjin–Hebei region [4,5,6], or regions with concentrated coal resources, such as Shanxi, Shaanxi, and Inner Mongolia [7,8]. There is a lack of research on urban agglomerations in the central region. Therefore, studying the carbon emissions of the urban agglomeration in the middle reaches of the Yangtze River is of great importance for promoting high-quality development and carbon reduction in the central region. Overall, the main contributions of this paper are as follows:

(1) It fills a gap in the study of carbon emissions in the middle reaches of the Yangtze River urban agglomeration. (2) The interaction between different factors influencing carbon emissions is considered, rather than treating the influencing factors in isolation. (3) For the first time, two spatially related methods are used to study the carbon emission influencing factors in the middle reaches of the Yangtze River urban agglomeration, taking into account the spatial heterogeneity of the region in the study to a greater extent.

The study of carbon emissions has been a popular research area. Carbon emission studies are mainly on the national [9], provincial [10,11], urban cluster [12,13], and occasionally county [14,15] scales. For the study of carbon emission influencing factors, most of them adopt the Kaya identity [16,17], the LMDI decomposition method [18,19,20,21], the IPAT model [22], the STIRPAT model [23,24,25], the spatial econometric model [26,27,28], the geographically weighted regression model [29,30,31], Geodetector [32,33], and so on. Huong [16] used Kaya identity and LMDI decomposition analysis to analyse carbon emission patterns in Vietnam and identify their critical drivers for the period 1990–2016. Behera and Dash [25] calculated the effects of urbanisation, energy consumption, and foreign direct investment on CO₂ emissions in the South and Southeast Asian countries during the period 1980–2012. Usama Al-mulali [34] used econometric methodology to investigate the major factors that influence the CO₂ emission in 12 Middle Eastern countries during the period 1990–2009. Zeng et al. [35] selected 30 provincial administrative units as spatial units and used the exploratory spatial data analysis (ESDA) method to study the spatial and temporal distribution patterns of transportation carbon emissions, while taking into account the differences in spatial units and constructed a GWR model to analyse the spatial and temporal heterogeneity of factors influencing transportation carbon emissions. However, current research on carbon emissions has only focused on the use of individual methods [36,37,38], each with different conditions of use and different emphases. The STIRPAT model, for example, does not take into account differences in the spatial distribution of data, and the spatial Durbin model generally requires spatial autocorrelation within the data, etc. There are also limitations regarding the research conclusions that can be drawn from individual research methods. A Geodetector model is a model used to explore non-linear relationships in spatial data, which can detect interactions and non-linear effects between spatial variables [39]. The geographically weighted regression model, on the other hand, is a local regression model that takes into account the spatial heterogeneity of the data. The main difference between the two models is that the Geodetector can explore the explanatory power of a single variable and the interaction between two variables, and is insensitive to multicollinearity, whereas the geographically weighted regression model can visualise the magnitude of the effect of a single variable on different geographical areas in space. The use of the two different models enables a better exploration of the strength of influencing factors in different spatial locations, as well as the interactions between the influences, leading to more comprehensive conclusions.

As for the research on the factors influencing carbon emissions, the current domestic and international studies generally agree that population factors [40,41], economic development level [42,43], industrial structure [44,45], energy intensity [46], urbanisation rate [47], etc., are important factors influencing carbon emissions, and foreign direct investment [48,49] and international trade [50,51] are also considered to have an impact on carbon emissions. For example, Birdasall [40] and Knapp et al. [41] studied the relationship between carbon emissions and total population, and the results showed that the growth in a global population is an important reason for the increase in carbon emissions. Douglas [42] studied the relationship between GDP and carbon emissions, and the results showed that as the economy grows, marginal carbon emissions show a decreasing trend. In a study on economic growth and carbon emissions in China, Tian et al. [43] found that the contribution of economic growth to per capita carbon emissions in China was exponential, with economic growth factors contributing 42.9% of the change in carbon emissions. In their study, Ang [44] and Sahu [45] found that the shift in industrial structure from heavy industries with high energy consumption to light industries with low energy consumption and the internal restructuring of industries would reduce the energy consumption intensity of the whole economy. Zhu et al. [46] categorised the change in CO₂ emissions into five main influencing factors, namely the emission factor of fossil fuels, energy consumption structure, energy intensity, GDP per capita, and total population. The study shows that energy intensity is the main driver. Through their research, Bi [47] found that urbanisation affects carbon emissions in terms of three aspects: urban industrial structure, residents’ lifestyles, and urban spatial layout. However, in most studies, demographic factors, economic development level, industrial structure, energy intensity, and urbanisation rate are important factors affecting carbon emissions. Therefore, these five indicators were selected as the influencing factors of carbon emissions in this study [52,53,54].

In this paper, 27 cities in the middle reaches of the Yangtze River were selected as the research objects, and the carbon emissions of the middle reaches of the Yangtze River urban agglomeration were counted using the measurement method of IPCC. Through the comprehensive use of the spatial autocorrelation model, Geotector and geographically weighted regression model, the influencing factors of carbon emissions in the middle reaches of the Yangtze River urban agglomeration were studied in depth, providing a reference basis for the formulation of the regional differentiated carbon emission reduction policies.

2. Materials and Methods

2.1. Study Area Overview and Data Processing

The urban agglomeration in the middle reaches of the Yangtze River is located in the core area connecting the east and the west, running through the north and the south. It is one of the most important city clusters in the Yangtze River Economic Belt, and a key area for promoting the “Rise of Central China” strategy, comprehensive high-quality development and new urbanisation. It is a mega-city cluster consisting of three city clusters: the Poyang Lake city cluster, the Changzhutan city cluster, and the Wuhan city circle. It has a land area of about 317,000 km² and includes 31 cities with a total population of 130 million in 2020.

The energy situation in the middle reaches of the Yangtze River urban agglomeration is that the three provinces are short of coal, oil, and natural gas, with insufficient energy self-sufficiency and high external dependence. Jiangxi province is rich in metal and mineral resources and its economy is highly dependent on high carbon emission industries. Hubei province is rich in hydropower resources due to the Three Gorges Dam, but relies almost entirely on external supplies of thermal power, while wind energy has large theoretical potential there but is under-exploited. Hunan province has large hydro energy reserves but is highly seasonal, and biomass has great potential there but is underutilised.

Considering the availability of data, the following 27 cities were selected to form the study area. The details are shown in

Table 1. Twenty-seven research cities in the middle reaches of the Yangtze River.

Province	City
Hubei Province	Wuhan, Jingmen, Huanggang, Xianning, Yichang, Xiangyang, Ezhou, Huangshi, Xiaogan, Jingzhou
Hunan Province	Changsha, Yueyang, Changde, Zhuzhou, Xiangtan, Hengyang, Yiyang
Jiangxi Province	Nanchang, Jiujiang, Yichun, Fuzhou, Shangrao, Jingdezhen, Ji’an, Pingxiang, Yingtan

.

The data sources used include the “Statistical Yearbook of Hubei Province”, “Statistical Yearbook of Jiangxi Province”, “Statistical Yearbook of Hunan Province”, “China Urban Statistical Yearbook” and statistical yearbooks of various cities such as “Statistical Yearbook of Wuhan” and “Statistical Yearbook of Changsha”, covering data on population, GDP per capita, energy consumption, and the proportion of secondary industry structure for each city from 2010 to 2020. Administrative boundary data were obtained from the Resource and Environment Science and Data Centre of the Chinese Academy of Sciences. Based on previous research [52,53,54], the five indicators of population size (pop), GDP per capita (p_gdp), the proportion of output value of secondary industry (gdp2), energy intensity (c_gdp), and urbanisation rate (ur) were selected as the independent variables, with carbon emissions as the dependent variable, in terms of population size, economic development level, industrial structure, energy intensity, and urban development level.

China currently lacks long-term real-time monitoring data on carbon emissions at the city level and a unified calculation standard for carbon emissions. Therefore, based on the data available at the city level, this study used the carbon emissions calculation formula proposed by the IPCC and calculated the carbon emissions of 27 cities in the urban agglomeration in the middle reaches of the Yangtze River. We considered eight types of energy sources including raw coal, coke, crude oil, gasoline, kerosene, diesel, fuel oil, and natural gas. The conversion factors for different energy sources and the carbon emission factors were obtained from the “China Energy Statistical Yearbook (2007–2016)” and the “National Greenhouse Gas Emissions Inventory Guidelines”. The calculation method is shown in Equation (1):

$$\mathrm{C}{\mathrm{O}}_2=E_i\times \mathrm{S}\mathrm{C}\times \mathrm{C}\mathrm{F}\times \frac{44}{12}$$

(1)

where CO₂ is total carbon emissions, E is energy consumption, i is energy type, SC is the standard discount factor, and CF is the carbon emission factor.

2.2. Research Methods

The first measure of the interaction of carbon emissions within a region was through the use of global spatial autocorrelation analysis. If the spatial autocorrelation was weak, then intra-regional variation analysis using a Geodetector and geographically weighted regression model was appropriate.

2.2.1. Global Spatial Autocorrelation Analysis

Spatial autocorrelation analysis is usually used to reflect a correlation within a space. In this paper, Moran’s I index was used to measure the degree of spatial clustering of carbon emissions in adjacent areas. The formula for global autocorrelation is as follows:

$$\mathrm{I}=\frac{n}{S_0}=\frac{\sum _{i=1}^n\sum _{j=1}^n{W}_{ij}(y_i-\stackrel{-}{y})(y_j-\stackrel{-}{y})}{\sum _{i=1}^n{(y_i-\stackrel{-}{y})}^2}$$

(2)

where $S_0=\sum _{i=1}^n\sum _{j=1}^n{W}_{ij},$ n is the total number of spatial units, y_i and y_j are the i-th and j-th spatial attributes, and W_ij is the spatial weight value.

Using the global Moran’s I index, the clustering situation of the data in the region could be judged, that is, whether there was a spatial correlation. When the index is greater than or equal to 0, it indicates that the data in the region have positive autocorrelation, while when the index is less than or equal to 0, it indicates that the data in the region have negative autocorrelation. In addition, the numerical value of the global Moran’s I index could be used to measure the concentration of the data, and the Moran’s I index was ∈[−1, 1].

The spatial weight matrix is constructed using the spatial adjacency weight matrix, and its expression is:

$$w_{ij}\left\{\begin{array}{c}1,when region i and j have a common border\\ 0,when region i and j have no common border\end{array}\right.$$

(3)

where i,j is the spatial region block number, i,j∈[1, n], n is the number of this spatial region. This paper sets up to use the rook adjacency matrix, i.e., only the spatial weight matrix of regions with common boundaries was assigned to 1, without considering the connection of regions with only common points.

For the results of global Moran’s I, a statistical test for z-significance was performed, calculated as

$$\mathrm{Z}\left(\mathrm{I}\right)=\frac{1-E\left(I\right)}{\sqrt{Var\left(I\right)}}~N(\mathrm{0,1})$$

(4)

where Var(I) is the theoretical variance in Moran’s I index, and E(I) is the theoretical expectation. If the statistical test for the significance of the Z-value is passed, it indicates that the observations have been autocorrelated in terms of spatial distribution rather than randomly distributed.

2.2.2. Geodetector

Through the use of a Geodetector, not only the explanatory power of a single factor on the dependent variable can be explored, but the interaction between factors can also be examined, showing linear or non-linear enhancement features. The core idea of Geodetector is based on the assumption that if an explanatory variable has a significant impact on another dependent variable, then their spatial distributions should also be similar.

The advantages of Geodetector are as follows:

(1)

It can explore both quantitative and qualitative data.

(2)

It can detect the interaction between two factors on the dependent variable [39].

The degree of explanatory power of factor X on the spatial differentiation of the dependent variable Y is measured using the q-value.

$$\mathrm{q}=1-\frac{\sum _{h=1}^L{N}_h{\sigma }_{\mathrm{h}}^2}{N{\sigma }^2}$$

(5)

where h = 1,2,…K represents the layers of the carbon emission influencing factor X, N represents the number of cities in the study area, N_h represents the number of cities included in the m layer of the influencing factor X, and ${\sigma }_{\mathrm{h}}^2$ and ${\mathsf{\sigma }}^2$ represent the variance in layer h and the variance in the entire carbon emission value of the region, respectively. Regarding q∈[0, 1], when q = 0, carbon emissions are randomly distributed. The larger the q-value, the greater the explanatory power of the influencing factor X on the carbon emissions, and the q-value indicates that X explains q × 100% of Y.

2.2.3. Geographically Weighted Regression Model

The geographically weighted regression model (GWR) has been a commonly used spatial analysis model in recent years for exploring the influencing factors of carbon emissions. The following gives a comparison of the advantages of global regression models:

• In terms of analysis results, the global regression model can only obtain the overall status of the entire study area and may ignore the relationships between variables. The GWR model introduces the spatial geographic location of the data in the model, which can reflect the specific situation of the distribution of the relationships between variables in different geographic spaces.
• In terms of visualisation, GWR can use ArcGIS to present the numerical values of various parameters on the map, making it easier to study the changes in parameters as the spatial geographic relationship changes.

The GWR model is expressed by the following function:

$$Z_i={\beta }_0(x_i,y_i)+\sum _{k=1}^p{\beta }_k(x_i,y_i){\theta }_{ik}+{\sigma }_i$$

(6)

Y is the dependent variable, β₀ is the intercept, (x_i,y_i) are the coordinates of the sampling point, ${\beta }_k(x_i,y_i)$ is the k-th fitting parameter on the sampling point i, which is a function related to the geographic location, ${\theta }_{ik}$ is the value of the k-th explanatory variable at the i-th sample point, and ${\sigma }_i$ is the random error.

3. Results

3.1. Spatiotemporal Characteristics of Carbon Emissions

The total carbon emissions of the Yangtze River midstream urban agglomeration have gradually increased from 49,141.30 million tonnes in 2010 to 51,162.98 million tonnes in 2020, but there have been fluctuations in the total carbon emissions between different years during the period of 2010–2020. From 2013 to 2017, the carbon emissions decreased from 54,121.37 tonnes to 49,525.63 tonnes, indicating a significant reduction in carbon emissions during this period. Similar time-series features were also observed in a study by Guangyao Deng [55] on the carbon emissions of the Lanxi urban agglomeration in Gansu province. The carbon emissions of the Yangtze River midstream urban agglomeration sharply declined from 2019 to 2020, possibly due to the outbreak of the COVID-19 pandemic in Wuhan at the end of 2019, leading to lockdowns in Wuhan and the surrounding cities, followed by a nationwide lockdown, resulting in reduced energy consumption and a decrease in carbon emissions (Figure 1).

From 2010 to 2020, the carbon emissions of each city in the Yangtze River midstream urban agglomeration showed a trend of initially increasing and then decreasing. Specifically, the high carbon emission area remained concentrated in Wuhan and did not change significantly, while the secondary high-emission area showed slight changes but remained stable overall. In 2010, the secondary high-emission areas were located in Xiangyang, Jingmen, Yueyang, Jiujiang, and Hengyang. The later secondary high-emission areas were mainly distributed in Yueyang, Jingmen, and Jiujiang. In 2020, the carbon emissions of Jiujiang surged to the same level as in Wuhan (Figure 2).

3.2. Global Spatial Autocorrelation

Global spatial autocorrelation analysis was conducted on the data from 2010 to 2020 to calculate Moran’s I index and perform significance tests (

Table 2. Global Moran’s I of total carbon emissions of Yangtze River midstream urban agglomerations from 2010 to 2020.

Variable	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020
Moran’I	−0.19	−0.20	−0.22	−0.20	−0.17	−0.17	−0.15	−0.15	−0.11	−0.07	−0.09
Z-value	−1.63	−1.70	−1.82	−1.73	−1.55	−1.39	−1.14	−1.17	−0.71	−0.33	−0.52
p-value	0.10	0.09	0.07	0.084	0.12	0.16	0.26	0.242	0.48	0.74	0.61

).

The results show that only 2011, 2012, and 2013 passed the significance test with p < 0.1, but not p < 0.05. This indicates that the carbon emissions in the Yangtze River midstream city cluster had relatively strong spatial autocorrelation in these three years, and Moran’s I index was negative, indicating negative spatial autocorrelation. For the other years, there was no significant spatial autocorrelation, and Moran’s I index decreased, indicating that the spatial correlation between carbon emission intensity among cities is becoming weaker and learning towards a random distribution.

3.3. The Result of Geodetector

(1)

Single-factor detection results (

Table 3. Single factor detection results for 2010, 2014, 2017, and 2020.

Impact Factor	q-Value
	2010	2014	2017	2020
Population size (pop)	0.7012	0.7012	0.7013	0.4666
GDP per capita (p_gdp)	0.3777	0.3774	0.3610	0.7225
Energy intensity (c_gdp)	0.4019	0.4031	0.3112	0.2862
Share of secondary sector (gdp2)	0.2432	0.2431	0.3246	0.1727
Urbanisation rate (ur)	0.2430	0.2693	0.7408	0.7064

).

The main factors affecting carbon emissions in the middle reaches of the Yangtze River urban agglomeration have changed significantly during the sample year period. In 2010, 2014, and 2017, the influence of population size on carbon emissions was the strongest, with q-values of 0.7012, 0.7315, and 0.7012, respectively; population size had strong explanatory power in terms of carbon emissions, but by 2020, its explanatory power had gradually decreased, with q value decreases to 0.4666.

The q-values of the urbanisation rate in 2010, 2014, 2017, and 2020 are 0.2430, 0.2693, 0.7408, and 0.7064, respectively, indicating that the explanatory power of urbanisation rate on the impact of carbon emissions has gradually become higher over time, and even dominated in 2017. This may be due to the fact that since the implementation of the Yangtze River Central River Urban Agglomeration Development Plan in 2015, the integrated network pattern of the middle reaches of the Yangtze River urban agglomeration has gradually strengthened, accelerating the urbanisation process, while the massive consumption of energy in the urbanisation process has increased the growth in carbon emissions.

The explanatory power of energy intensity and the share of secondary industry on carbon emissions is also weakening, with the q-value of the impact of energy intensity on carbon emissions gradually decreasing from 0.4019 in 2010 to 0.286, and the q-value of the impact of the share of secondary industry on carbon emissions rising from 0.2432 and then decreasing to 0.1727.

(2)

Interaction factor detection results (Figure 3).

The q-values obtained from the impact factor interactions for 2010, 2014, 2017, and 2020 were ranked separately and the top five factor combinations were collated with the results shown in

Table 4. Top five combinations for interaction factor detection in 2010, 2014, 2017, and 2020.

2010		2014		2017		2020
Interaction Factor	q	Interaction Factor	q	Interaction Factor	q	Interaction Factor	q
ur∩c_gdp	0.7452	pop∩c_gdp	0.7087	pop∩c_gdp	0.7799	ur∩c_gdp	0.8025
c_gdp∩gdp2	0.7400	c_gdp∩p_gdp	0.7024	c_gdp∩p_gdp	0.6490	ur∩p_gdp	0.7947
pop∩c_gdp	0.7215	ur∩c_gdp	0.7520	p_gdp∩gdp2	0.6063	pop∩c_gdp	0.7688
c_gdp∩p_gdp	0.6737	p_gdp∩gdp2	0.6624	ur∩c_gdp	0.5401	c_gdp∩p_gdp	0.6868
p_gdp∩gdp2	0.6624	ur∩c_gdp	0.6121	ur∩p_gdp	0.5244	p_gdp∩gdp2	0.5666

.

The results show that the explanatory power of the two-by-two interactions of the impact factors has eithers enhanced or weakened compared to the individual impact factors. The specific linear or non-linear enhancement can be found in Figure 4. The impact factors after the two-by-two interaction, with q-values > 0.7 for 10 pairs, were namely energy intensity (c_gdp) and urbanisation rate (ur), energy intensity (c_gdp2) and share of output in the secondary sector (gdp2) in 2010, energy intensity (c_gdp) and population size (pop) in 2014, energy intensity (c_gdp) and GDP per capita (p_gdp), energy intensity (c_gdp) and urbanisation rate (ur), energy intensity (c_gdp) and population size (pop) in 2017, energy intensity (c_gdp) and urbanisation rate (ur) in 2020, energy intensity (c_gdp) and population size (pop) and energy intensity (c_gdp) and urbanisation rate (ur), energy intensity (c_gdp) and population size (pop) and energy intensity (c_ gdp) and urbanisation rate (ur). It can be seen that the urbanisation rate had a limited impact on carbon emissions in 2010 and 2014, but interacted with the energy intensity factor to leapfrog the explanatory power of carbon emissions, and the same for industrial structure (Table 4).

The interaction factors with the largest q-values have shown similarities across time, with the largest q-values in 2014 and 2017 being for energy intensity (c_gdp) and urbanisation rate (ur), and the largest q-values in 2010 and 2020 being for energy intensity (c_gdp) and population size (pop).The explanatory power of the interaction factor between energy intensity (c_gdp) and GDP per capita (p_gdp) rises and then falls, showing the same trend of change. The explanatory power of the interaction factor for the share of the secondary sector in total output (gdp2) and GDP per capita (p_gdp) also rises and then falls.

3.4. Geographically Weighted Regression Model

As the use of geographically weighted regression methods requires the exclusion of multicollinearity in the selected variables, a multicollinearity diagnosis was first performed on the selected impact factors. The magnitude of the variance inflation factor (VIF) is generally used to determine whether there is multicollinearity between factors. In Arcgis, a VIF > 7.5 is generally considered to be a problem of multicollinearity.

In this paper, with the help of SPSS, the independent variables were tested for covariance in 2010, 2014, 2017, and 2020, respectively. The variance inflation factor (VIF) values of all five independent variables were less than 7.5, and the share of output value accounted for by secondary industry (gdp2) and urbanisation rate (UR) failed the significance test in 2010, 2014, 2017, and 2020. Summing up the above results, population size (pop), GDP per capita (p_gdp), and energy intensity (c_gdp) were finally selected as the independent variables.

The following results were obtained via geographically weighted regression of Arcgis data for 2010, 2014, 2017, and 2020.

3.4.1. The Impact of Population on Carbon Emissions

According to Figure 5, it can be seen that, in 2010, 2014, 2017, and 2020, the regression coefficient of population size showed an increasing trend in spatial distribution from south to north. Moreover, there is a large intra-regional spatial distribution difference for 2010, 2014, and 2017. Comparing the differences in values for 2010, 2014, 2017, and 2020, the regression coefficient of population size generally shows a decreasing trend from 2010 to 2017 and then an increase in 2020 in terms of the absolute value of the data.

From 2010 to 2020, the spatial differences in the population size coefficients showed a trend of increasing and then decreasing, and by 2020, they narrowed to a smaller value, indicating that the contribution of population changes to carbon emissions basically converged, and the overall regression coefficient of population size has decreased. On the one hand, since the Yangtze River Central River Urban Agglomeration Development Plan came into effect in 2015, the coordinated development of cities has been vigorously promoted, gradually forming the pattern of the central river urban agglomeration, which has also led to the balanced development of the population. The impact of population size on carbon emissions is also gradually decreasing.

3.4.2. The Impact of GDP per Capita on Carbon Emissions

The spatial distribution pattern of GDP per capita regression coefficients from 2010 to 2020 shows little change, with a slight trend of high values shifting from the northeast to the northwest (Figure 6). The regression coefficient of GDP per capita shows a significant decrease in its impact on carbon emissions.

The GDP per capita regression coefficient has continued to fall, with a healthier and more sustainable increase in GDP, either due to there being more services and high-tech industries, or due to technological advances that have also reduced energy consumption in terms of output, and with all examples showing a trend of higher in the north and lower in the south.

3.4.3. The Impact of Energy Intensity on Carbon Emissions

From 2010 to 2020, the regression coefficient of energy intensity in the urban agglomeration of the middle reaches of the Yangtze River showed a trend of high values spreading from the northwest towards the centre and then towards the northeast, and the spatial differences showed an increase followed by a decrease (Figure 6). This is mainly due to changes in the spatial distribution of the technological innovation capabilities within the urban agglomeration of the middle reaches of the Yangtze River. On the one hand, with technological progress, energy intensity is gradually decreasing, and its impact on carbon emissions is also decreasing. On the other hand, under the guidance of national policies such as the “Planning of the Urban Agglomeration of the Middle Reaches of the Yangtze River”, the development in technological innovation capabilities in the region is becoming more integrated. Although there is still a gap in the technological innovation capabilities within the urban agglomeration of the middle reaches of the Yangtze River, this gap is gradually narrowing as similar development plans are being implemented in the region.

4. Conclusions

(1)

In terms of total carbon emissions, the total carbon emissions of the middle reaches of the Yangtze River urban agglomeration show fluctuations between 2010 and 2020, the carbon emission reduction effect is unstable, and there is still room for more carbon emission reduction.

(2)

In terms of the spatial and temporal distribution of carbon emissions, those in the middle reaches of the Yangtze River urban agglomeration show obvious spatial variability, but the high carbon emission area is always concentrated in Wuhan and this remains unchanged.

(3)

In terms of the spatial distribution of factors influencing carbon emissions in the middle reaches of the Yangtze River urban agglomeration, in 2010, 2014, and 2017, population size was the most important factor influencing carbon emission divergence, and in terms of interaction, the interaction between energy intensity, GDP, and urbanisation was the cause of increasing carbon emissions. The impact of population size on carbon emissions has decreased from north to south, and the impact of energy intensity on carbon emissions has shown a spread from the most influential region from in the northwest to the centre and then to the northeast, while the GDP per capita has had little impact on the differences in the spatial distribution of carbon emissions.

(4)

In terms of the explanatory power of the factors influencing carbon emissions in the middle reaches of the Yangtze River urban agglomeration, given that the impact of population size on carbon emissions is weakening and will no longer be an important factor related to carbon emissions by 2020, and that the interaction between energy intensity, population size, urbanisation, and GDP capita gdp is the cause of increasing carbon emissions, carbon reduction strategies should focus on controlling energy intensity, and policies should be tailored to local conditions based on spatial differences in the degree of influence of different factors on carbon emissions.

5. Recommendations and Prospects

(1)

Strengthen technological innovation and improve energy use efficiency. The interaction between energy intensity and the level of economic development and urbanisation is the main driver of carbon emissions in the middle reaches of the Yangtze River urban agglomeration. The key factor influencing energy intensity is the development in the level of technology. The impact of energy intensity in the middle reaches of the Yangtze River urban agglomeration shows a distribution of high in the north and low in the south, and the high values are slowly shifting from the northwest to the northeast. this indicates that in the carbon emission control of the middle reaches of the Yangtze River urban agglomeration, the adjustment for Hubei cities should be prioritised more in the region, i.e., strengthening the technological innovation of Hubei cities and improving energy use efficiency, while allowing for more high-precision industries from the east coast to the central part, especially regarding the cities in the middle reaches of the Yangtze River cities in the northern part of the Yangtze River urban agglomeration.

(2)

Promote new urbanisation and build a green development city. This study also found that the urbanisation rate has been the main driver of carbon emissions in the middle reaches of the Yangtze River urban agglomeration since 2017 and 2020. Therefore, how to promote the intensive development of urban agglomeration and promote urban-rural integration and realise the promotion and popularisation of new green urbanisation in the middle reaches of the Yangtze River urban agglomeration in the process of urbanisation, are priority issues.

(3)

Based on resource endowment, optimise energy structure. The six high-energy-consuming industries in the middle reaches of the Yangtze River urban agglomeration account for a high proportion of industry and have still been on the rise in recent years. The consumption of fossil energy still accounts for a large proportion of the energy structure, and to achieve the goal of carbon peaking and carbon neutrality, it is necessary to adopt a variety of clean energy sources such as solar, wind, and geothermal energy according to local conditions. For example, Hubei province is not in a good position for photovoltaics; thus, the advantages of hydropower should be fully exploited, and wind power arrangements should be strengthened to adjust its energy structure. The economy of the Jiangxi province relies heavily on high carbon emission industries and it is difficult to adjust the industrial structure in the short term; thus, it should promote the conversion of coal to gas and coal to electricity and develop clean coal power where appropriate.