Introduction

In the process of the global economy to green, carbon emissions have intensified the pressure on countries to tackle the challenge of climate warming (Chaudhari et al., 2023; Ho Kugele and Sarkar, 2023). Therefore, curbing environmental deterioration and protecting the environment has become the responsibility and obligation of all countries in the world (Jiang et al., 2022). China has the largest total carbon emissions in the world (Dong et al., 2018; Kong et al., 2022; Peng et al., 2024). The International Energy Agency reported on carbon dioxide emissions in 2023, with China emissions reaching 12.6 billion tons, accounting for about 34 percent of the global total (Ge et al., 2023). This phenomenon shows that China is facing a challenge in carbon emission reduction. Ma et al. (2011) identifies energy consumption as the main reason for the increase in carbon emissions.

The Yangtze River Economic Belt (YEB) spans 11 provinces in China, including Shanghai, Hubei and Sichuan. In 2023, the region’s GDP reached 5,842.74 billion yuan, accounting for 46 percent of the country’s total (Guchhait and Sarkar, 2024; Lu et al., 2024; Xu et al., 2024). Therefore, the YEB has huge economic potential. At the same time, it faces the important task of maintaining ecological balance in the rapid development. In this context, the YEB has followed the guidance of the Action Plan for Energy Conservation and Carbon Reduction and other documents, and formulated an action plan for carbon emission reduction based on its own actual situation (Dey et al., 2023; You et al., 2023). For example, Wuhan’s carbon reduction, pollution reduction and green growth ten action plans. The implementation of these policies aims to promote the YEB to achieve a win-win situation for economic development and environmental protection (Gong and Zhao, 2024; Zhang et al., 2024a, 2024b, 2024c). However, the carbon emissions of the YEB are not optimistic, with significant emission differences between regions and the overall carbon emission level remaining at a high level (Liu and Hao, 2022; Wu and Zhang, 2021). With the rapid economic development, the task of carbon emissions control will only become more difficult (Sarkar et al., 2025a; 2025b). Therefore, it is beneficial to optimize and implement policies to evaluate the evolution trend and influencing factors of carbon emissions in Chinese cities.

Based on the above background, there have been studies on the relationship between nighttime light image data and urban carbon emissions (Guchhait et al. 2024; Sarkar et al., 2025a; 2025b). Liu et al. (2019) focus on the county social economy in Shandong province unit of space-time analysis. In the study, the trend analysis, stability analysis and correlation analysis method proved that the night light pattern and spatial pattern between consistency, and illustrates the population growth and economic development in light plays an important role in the process of change at night. Zhang et al. (2023) believes that the nighttime light remote sensing of the YEB can provide a solid theoretical and technical foundation for scientific management and city planning. Wang et al. (2024) explores the change of light intensity in the conflict between Ukraine and Russia, and indicates the nighttime light remote sensing in Ukraine was in an obvious downward trend in the war, and then believes the war has a serious impact on the social and economic development of Ukraine. Yu et al. (2023) uses nighttime light remote sensing to evaluate the traffic status of expressways, and points out that China’s expressways have regional heterogeneity, the eastern coastal areas are better than inland areas, and the performance of national roads at night is better than the day.

On the other hand, some literature mainly analyzes the influencing factors, carbon emission efficiency and differences of carbon emission reduction (Sarkar et al. 2024; Zhang et al. 2020). Among them, Li et al. (2024) measured the current situation of carbon emissions in China’s circuit industry from the two aspects of direct emissions and consumption of the power industry. Adams and Opoku (2020) measures the carbon emission intensity in Africa, and combines it with the consumption of fossil fuels and the import and export trade. Feng and Publishing (2020) believes that under the influence of industrial structure upgrading and emission reduction policies, carbon emission right trading will play a positive role in carbon emission reduction undertakings. Mahmood (2023) proves that NRR will cause the attenuation of gas emissions, gas combustion emissions and carbon dioxide emissions. Liu and Yan (2023) believe that population size, industrial structure and affluence have an important impact on carbon emissions. However, Zhang et al. (2021) uses the generalized Divisia index method to explore the influencing factors for carbon emission reduction of the logistics industry, and believe that energy consumption and population size have a great impact on carbon emissions. Gao et al. (2021) explores the carbon emission efficiency in the industrial field from the implied carbon emission and the direct carbon emission, and indicates that both the implied carbon emission and the direct carbon emission efficiency in China’s industrial field are at the rising stage. Li et al. (2022) believes that economic output, energy efficiency and energy structure have an important impact on carbon emission efficiency. Wang et al. (2022) proves that industries with low carbon emission efficiency are mainly concentrated in agriculture, services, light industry and construction, while industries with high carbon emission efficiency are concentrated in the power sector, fossil energy sector, transportation industry, chemical industry and heavy industry. Zhao et al. (2022) believes that the main reason for the differences in regional carbon emissions is the differences in resources and economic development between regions. In terms of spatial performance, the efficiency of carbon emission in the western region is significantly lower than that in the eastern region. In addition, Zhang et al. (2024a, 2024b, 2024c) uses remote sensing data of night light when exploring the spatiotemporal change and influence mechanism of China’s carbon emissions, which provides convenience for exploring the key attributes of China’s carbon emissions.

According to the existing research literature, although the research on carbon emissions has been increasing, there is still a lot of room for research in this field. Firstly, in the measurement of carbon emissions, current research relies on energy data. Therefore, the phenomenon can have two reasons: On the one hand, there may be two problems in this situation. On the one hand, due to the performance assessment of some local governments, the energy data of poor carbon emission control years may not be disclosed, and then affects the accuracy of carbon emissions measurement. On the other hand, the accuracy of the carbon emissions factor coefficient method is insufficient, and there are large errors in the measurement of some regions. Secondly, the research samples of existing literature are mostly on provinces, but few focus on cities. There is a relative lack of temporal-spatial evolution characteristics and influencing factors of urban carbon emissions in the YEB. Thirdly, the path research of carbon emission reduction is mostly analyzed by linear regression model, but little literature explores the impact path of carbon emission reduction from a spatiotemporal perspective.

For that reason, this paper is based on 108 cities in the YEB from 2005 to 2022 to analyse the disequilibrium and influencing factors of urban carbon emissions in the Yangtze River Economic Belt, can have the following three marginal contributions. First, this paper uses nighttime light remote sensing image data to reverse the carbon emission data of cities in the Yangtze River Economic Belt, which avoids the impact caused by the lag and imprecision of economic data statistics, so that the estimation of carbon emission of cities in the Yangtze River Economic Belt is more scientific and accurate. Second, this paper explores the spatial disequilibrium of urban carbon emissions from 108 prefecture-level cities in the Yangtze River Economic Belt, which is conducive to regional governments to accurately grasp the spatiotemporal evolution trend and distribution pattern of carbon emissions, and provides theoretical reference and reference for the government to formulate relevant policies according to local conditions. Third, this paper uses Geographically and Temporally Weighted Regression (GTWR) method to discuss the influencing factors of carbon emission reduction in cities along the Yangtze River Economic Belt from both time and space dimensions, which enriches the existing academic research system. Based on this, this research provides a theoretical of the large emission reduction and carbon reduction policy in the Yangtze River economy, and provides quantitative support for accelerating the pace of “dual-carbon”.

Research design

Problem description

Based on the above analysis, it can be found that the YEB, as an important part of China’s economy, has still been facing significant environmental pressures in recent years. Therefore, to fully understand the changing trends and influencing factors of carbon emissions in cities within the YEB across time and space, this paper intends to explore and analyze from the following two aspects: On the one hand, investigating the spatial imbalance of carbon emissions in cities within the YEB, specifically examining the temporal trends, spatial trends, spatial autocorrelation, and spatial migration characteristics of carbon emissions in these cities; On the other hand, exploring the influencing factors of carbon emissions in cities within the YEB, mainly analyzes the influence of time and dimension on the carbon emission of YEB cities. This leads to the construction of a logical framework as shown in Fig. 1.

Fig. 1: Logical framework of thinking.
figure 1

This figure illustrates the research idea and framework of this paper, and explains the research area, thinking and method of this paper from two aspects: spatial disequilibrium and influencing factors.

Full size image

Model design

Kernel density estimation

In the comparison of kernel density estimation (KDE) and other estimation methods, the former can effectively avoid the preset distribution assumptions when dealing with fuzzy or complex data, and then improve the flexibility in the data processing process (Xie et al., 2025). In exploratory research, this method can accurately catch the distribution properties of complex data, and promotes understanding of the data nature. In conclusion, the KDE is able to effectively respond to the diversity and uncertainty of the data, and ensures the flexibility and veracity of the result.

$${\hat{f}}_{h}(x)=\frac{1}{nh}\mathop{\sum }\limits_{i=1}^{n}K((x-{x}_{i})/h)$$
(1)

In Eq. (1), x represents the research observation point, xi indicates the research sample, n is the value range of the research sample. The equation K((x − xi)/h) represents the kernel function. The equation is meaningful only when the function is greater than or equal to 0 and the integral is 1. h represents the bandwidth.

Spatial autocorrelation

In the process of data processing, Moran’s I not only considers the spatial proximity characteristics, but also introduces the spatial weight matrix (SWM) into the model to refine the spatial proximity relationship (Meng et al., 2025; Peng and Li, 2025). Therefore, the scholar tends to use Moran’s I to analyze spatial autocorrelation. This paper combines the advantages of global Moran’s I and local Moran’s I, and explores the spatial agglomeration characteristics of urban carbon emissions in the YEB. The calculation equation is as follows:

$$Moran{\hbox{'}}s\,I=\frac{\mathop{\sum }\nolimits_{i=1}^{n}\mathop{\sum }\nolimits_{j=1}^{n}{W}_{ij}({Y}_{i}-\overline{Y})({Y}_{j}-\overline{Y})}{{S}^{2}\mathop{\sum }\nolimits_{i=1}^{n}\mathop{\sum }\nolimits_{j=1}^{n}{W}_{ij}}$$
(2)
$${I}_{i}=\frac{({Y}_{i}-\overline{Y})}{\mathop{\sum }\nolimits_{i=1}^{n}({Y}_{i}-\overline{Y})}\mathop{\sum }\limits_{j=1}^{n}{W}_{ij}({Y}_{j}-\overline{Y})$$
(3)

In Eqs. (2) and (3), n indicates the number of cities in the YEB, Yi and Yj represents the carbon emissions sample values of cities i and cities j. Wij indicates the SWM, \(\bar{Y}\) is the research sample mean, S2 is the research sample variance.

Spatial Markov chain

The Markov chain reveals the characteristics that the future variable states are mainly affected by the current state, but are independent of the historical state. This principle can effectively guarantee that the research decision is based on the objective state. The spatial Markov chain is an extension of the traditional Markov chain (Wang et al., 2025). This method can quantify the spatial state change by combining time and spatial dimension, and improve the analytical power of the evolution of variable state in complex environment. The model is calculated as follows:

$${P}_{ij}=\frac{{Z}_{ij}}{{Z}_{i}}$$
(4)
$${L}_{a}=\mathop{\sum }\limits_{b=1}^{n}{W}_{ab}{Y}_{b}$$
(5)

In Eq. (4), Pij is the probability that the values shift from i level to j level, Zij indicates the total number that transition from state i to j at time t, Zi is the total number of provincial units with carbon emissions in a state i. In Eq. (5), Wab indicates the spatial weight matrix, n is the number of sample cities.

GTWR

GTWR integrates spatialtemporal analysis dimensions, and shows the dynamic evolution characteristics of the data in different geographical coordinates and time nodes. In addition, GTWR enhances the ability to detect variation patterns in a spatiotemporal staggered background, and can clearly observe the temporal-spatial heterogeneity of the data (Xiang et al., 2025; Zhang et al., 2024a, 2024b, 2024c). The model is calculated as follows:

$${C}_{i}={\alpha }_{0}({\mu }_{i},{\nu }_{i},{t}_{i})+\mathop{\sum}\limits_{k}{\alpha }_{k}({\mu }_{i},{\nu }_{i},{t}_{i}){x}_{ik}+{\beta }_{i}$$
(6)

In Eq. (6), Ci is the carbon emissions of the city i, xik indicates the observed value of the variable k in the city i, \(\left({\mu }_{i},\,{v}_{i},\,{t}_{i}\right)\) represents the spatial coordinate position of the city i, \({\alpha }_{k}\left({\mu }_{i},\,{v}_{i},\,{t}_{i}\right)\) represents the regression coefficient of the variable k in the city i, βi refers to the residual.

$${({d}_{ij})}^{2}=\lambda [{({\mu }_{i}-{\mu }_{j})}^{2}+{({\nu }_{i}-{\nu }_{j})}^{2}]+\mu {({t}_{i}-{t}_{j})}^{2}$$
(7)
$${W}_{ij}=\exp ({(-{d}_{ij}/h)}^{2})$$
(8)

In Eq. (7), dij reflects the Mahalanobis distance between city i and city j, λ and μ are the scale factor sign of the spatial distance to the time distance. In Eq. (8), Wij is the calculated weight matrix, h is the bandwidth.

Research area and data sources

Research area

In the process of China’s low-carbon economic transformation, the YEB has an important strategic position. The map of the research area is shown in Fig. 2. From the perspective of geographical location, the YEB spans 11 provincial administrative units in China. It not only shows significant economic development potential and natural resource endowment, but also creates natural conditions for information reciprocity between regions. From the perspective of carbon emissions, the residential population of the YEB accounts for 42.9% of the national population. The region has a diversified industrial structure such as agriculture, industry and services. The high-carbon industry occupies an important position in the YEB. Therefore, the carbon emissions of the YEB may be affected by many factors.

Fig. 2: The research area.
figure 2

This map shows the administrative divisions of the Yangtze River Economic Belt. The shaded areas represent the research area, while the unshaded regions indicate areas with no data.

Full size image

Data source and processing

(1) The nighttime light remote sensing. This paper collects DMSP-OLS data from 2005–2013 and NPP-VIIRS data from 2014–2022 on the NOAA platform (https://www.noaa.gov/organization/ information-technology/noaa-geoplatform), and processes the data rigorously (Zhou et al., 2023). The steps are as follows: Firstly, the DMSP-OLS data of the YEB were corrected successively by the invariant target method (Ash and Mazur, 2020). Secondly, the NPP-VIIRS images of the YEB were corrected by log transformation and noise treatment (Bian et al., 2019). Thirdly, in this paper, we complete a unified time-series continuity correction between NPP-VIIRS data and DMSP-OLS data (Wu et al., 2018).

(2) The social and economic statistics. This paper comprehensively measures the urban carbon emissions, education, industrial and population factors in the YEB, and refers to the studies of Liu et al. (2023). This research selects industrial structure, government intervention, science and technology development, education development, economic development and human capital from 2005–2022, and explores the influencing factors of carbon emission reduction in the YEB. Data from China Statistical Yearbook (https://www.stats.gov.cn/sj/ndsj/). The selection table of carbon emissions impact factors in the YEB was constructed as shown in Table 1. Among them, this paper uses the proportion of the added value of the tertiary industry and the added value of the secondary industry to measure the industrial structure, through the general budget expenditure for the proportion of GDP to measure government intervention level. Using the proportion of science and technology expenditure in the general budget expenditure to measure the level of science and technology development. Using the proportion of education expenditure in the general budget expenditure to measure the education development level. The level of economic development is measured by the logarithm of the per capita GDP. The level of human capital uses the number of college students per 100 people.

Table 1 Selection of carbon emission impact factors in the YEB.
Full size table

Note: This study takes 108 cities in the Yangtze River Economic Belt as the research object and collects sample data of these cities from 2005 to 2022. The data are from China Statistical Yearbook (https://www.stats.gov.cn/sj/ndsj/).

Analysis of results

Spatial disequilibrium of urban carbon emissions in the YEB

Temporal agglomeration of urban carbon emissions in the YEB

This research describes the KDE map of urban carbon emissions in the YEB in Fig. 3. From the perspective of the changes in the center of the curve, the center of the KDE curve is consistently shifted to the right in 2006–2022. Among them, the center of curve shifts more slowly between 2006–2016 and faster between 2016–2022. This phenomenon points out that carbon emissions in the YEB are on the rise. From the perspective of the tail morphology of curve, the curve shows the tail-dragging phenomenon in the early stage of the sample, and indicates the imbalance of carbon emissions among cities. The curve tends to smooth in the later stages of the sample, and indicates the imbalance is improving. From the perspective of the peak of the curve, this phenomenon shows that the differences in urban carbon emissions in the YEB are gradually narrowing. Above all, this change did not show extreme high or low values. That is, there is no polarization of carbon emissions.

Fig. 3: The map of kernel density estimation for urban carbon emissions in the YEB.
figure 3

This figure shows the distribution characteristics of the kernel density estimation curve of urban carbon emissions in the Yangtze River Economic Belt. The color gradient represents the high and low values of kernel density, with red indicating higher positive values, and black representing lower values.

Full size image

Spatial distribution characteristics of urban carbon emissions in the YEB

This research uses ArcGIS10.2 software to draw the evolution map of the spatial pattern for carbon emissions in the YEB, as shown in Fig. 4.

Fig. 4: The map spatial distribution characteristic for urban carbon emissions in the YEB from 2005 to 2022.
figure 4

This figure illustrates the spatial distribution of carbon emissions in Yangtze River Economic Belt across different years. The maps depict the carbon emission levels in 2005, 2011, 2016, and 2022. The darker red areas indicate higher carbon emission levels, while lighter red areas represent lower emission levels. White areas indicate regions with no available data.

Full size image

According to the results of Fig. 3, the carbon emissions of the YEB show a significant trend of enhancement. Specifically, the number of cities whose carbon emissions reached or exceeded the median level between 2005 and 2022 was increasing from 4 to 16, and the number of cities with low carbon emissions is decreasing. Further observation shows that the development of carbon emissions in the YEB shows significant regional differences in the spatial distribution. In 2005, the carbon emissions of the YEB reached or exceeded the medium level were scattered in Chongqing, Shanghai, Suzhou and Ningbo. This phenomenon is the result of the joint effect of the industrialization process and economic development on industrial structure. After 2005, the carbon emissions of the YEB has a trend of increasing outward with the above cities as the center. Chongqing has increased from medium carbon emission level to high carbon emissions. The reasons for presenting the above phenomenon are as follows: As an important strategic fulcrum of western development, the accelerated development of the industrialization process leads to the increasing trend of carbon emissions. It is worth noting that the carbon emissions of Ningbo, Suzhou and Shanghai have been reduced in recent years, which has a great correlation with the national policy orientation and the upgrading of industrial structure.

Spatial autocorrelation of urban carbon emissions in the YEB

Based on the above analysis, this paper believes the urban carbon emissions of the YEB are correlated in the spatial dimension. Therefore, this research uses the global Moran’s I to measure the spatial autocorrelation of carbon emissions. The specific results are shown in Table 2. This result shows the global Moran index remained positive during the 2005–2022 period. The global Moran’s I is significant at the significance level of 1%. Therefore, the phenomenon reveals that the carbon emissions of the YEB show a strong positive spatial correlation. In addition, the carbon emissions in the YEB may have a significant aggregation effect on space. That is, high or low levels of carbon emissions tend to be geographically concentrated, and leads to a prominent positive spatial correlation.

Table 2 Global Moran’s index of urban carbon emissions in the YEB.
Full size table

In order to explore the spatial aggregation characteristics of carbon emissions in the YEB, this research uses the ArcGIS 10.2 software to draw the local spatial autocorrelation map of carbon emissions in each city, as shown in Fig. 5. Through the analysis of Fig. 5, it is found that the number of cities with high-high agglomeration in the YEB decreased significantly, while the number of cities with low-low agglomeration increased significantly. This phenomenon shows that the low-carbon economic development of the YEB has achieved remarkable results, and positive progress has been made in ecological environmental protection and governance. Specifically, the urban carbon emissions in the lower reaches of the YEB are in a state of high-high agglomeration, while the cities with low-low agglomeration state are mainly distributed in the middle and upper reaches of the YEB, such as Sichuan Province and Yunnan Province. This phenomenon not only conforms to the industrial structure characteristics of the YEB, but is also closely related to the population density and the urbanization process.

Fig. 5: LISA diagram of urban carbon emissions in the YEB from 2005 to 2022.
figure 5

This figure illustrates the LISA clustering results of carbon emissions in the Yangtze River Economic Belt over different years. The maps depict the spatial clustering patterns in 2005, 2011, 2016, and 2022. The light red areas indicate high-high clusters, the dark red areas indicate high-low clusters, the dark blue areas indicate low-high clusters, and the light blue areas indicate low-low clusters. The off-white areas represent regions that are not significant, while white areas indicate regions with no available data.

Full size image

Spatiotemporal migration characteristics of urban carbon emissions in the YEB

This research uses the spatial Markov chain to explore the spatiotemporal migration characteristics of urban carbon emission in the YEB. First, the Markov chain was used to explore the transition probability of carbon emission, and the research results are shown in Table 3.

Table 3 The Markov transition probability matrix of urban carbon emission in the YEB from 2005 to 2022.
Full size table

Specifically, in the two time intervals of 2005–2013 and 2014–2022, the value of the elements in the diagonal position in the transition probability matrix is larger than that in the other positions. Among them, the maximum and minimum values are 0.991 and 0.753. This phenomenon reveals the stability of carbon emissions from the YEB in the transfer process of different zones. In the Markov transition probability matrix, the probability of high carbon emissions levels maintaining the status quo is 99.1%, and the probability of low carbon emissions levels maintaining the status quo is 75.3%. From the number of elements in the diagonal, the probability maximum and minimum are 0.209 and 0.000, respectively. This data shows that the urban carbon emissions in the YEB are in an upward development trend, rather than a significant grade leap. In this regard, if the urban carbon emissions are at a low level, the probability of shifting to a high level in the future development process is 16.5%, and the probability of maintaining a low level is 83.5%, but the probability of leaping to a high level is 0. Therefore, if the city is in a low level state, the future development path may still be in a low level state. In addition, the lower carbon emissions are also likely to develop downward. According to the data, the probability of the lower carbon emissions is 1.8% to the low level, but the probability of the higher and high carbon level to the low carbon is 0.

In order to further explore the spatial migration characteristics of urban carbon emissions in the YEB, this paper adds spatial lag terms based on the traditional Markov chain, and calculates the spatial Markov transfer matrix of the YEB from 2005 to 2022. This analysis combines the results of Tables 3 and 4, and finds the spatial Markov transition probability matrix in Table 4 is significantly changed compared with Table 3. The urban carbon emissions in the YEB have obvious spatial linkage. That is, when the domain type is low carbon emissions, the number of regions with low carbon emissions during this period is 6. When the domain type is at the high carbon emissions, the number of regions with high carbon emissions during this period is as high as 151, ranking first among various states. This phenomenon confirms that the urban carbon emission in the YEB has a significant interactive and promoting effect in the spatial dimension.

Table 4 Spatial Markov transfer matrix of urban carbon emission in the YEB from 2005 to 2022.
Full size table

Analysis of the influencing factors of urban carbon emission in the YEB

Model testing

Before using the GTWR model, it is necessary to conduct a collinearity test on the variable data in this paper to ensure that the variables selected in this study and the variable data used will not affect the research conclusions due to the interrelationships between the variables. This paper uses the VIF (Variance Inflation Factor) to test for multicollinearity among variables, with the results of the multicollinearity test summarized in Table 5. From the multicollinearity test results in Table 5, all variable VIF values are greater than 1, and the maximum value is less than 3. This indicates that there is no multicollinearity among the variables in this study, allowing the use of the GTWR model for analysis.

Table 5 Collinearity test.
Full size table

On the basis of the above tests, GTWR is fitted to the data separately with GWR (Geographically Weighted Regression), TWR (Temporally Weighted Regression) and OLS (Ordinary Least Squares) models, and Table 6 shows the fitting results of the four models. As can be seen from the results in the table, by comparing the results, it can be found that the R2 value of the GTWR model is the highest at 0.837, and the AICc value is the smallest at 668.36. Therefore, by comparing the fitting results of the four models, it can be found that the GTWR model is better than the GWR, TWR and OLS models.

Table 6 Fitting results of regression model.
Full size table

Time trend of the influencing factors of urban carbon emissions in YEB

In order to further explore the influence factors of urban carbon emissions in the YEB, this paper uses the GTWR model to analyze the influence of variables on carbon emissions in the spatiotemporal dimension, and the results of the time dimension of the regression coefficient are shown in Fig. 6. On the whole, the influence factors in this paper will promote the urban carbon emissions in the YEB, but the effect characteristics of each factor are different. Among them, the average regression coefficient of educational development, economic development and human capital is always greater than 0, and indicates that the above variables will always improve the carbon emissions in the YEB during the study period. In terms of influence characteristics, the promoting effect of human capital on carbon emissions is weakening year by year. While the impact of education development and economic development fluctuates, it still shows an overall trend of enhancement. In terms of the impact degree of YEB on urban carbon emissions, the level of economic development has the most significant impact on carbon emissions, and its average coefficient is the largest. The factor with the smallest impact coefficient is the level of scientific and technological development. The level of economic development is directly linked to energy consumption and industrial structure. As the industrialization and urbanization of the Yangtze River Economic Belt accelerate, the expansion of high-carbon industries and the surge in energy demand have significantly increased carbon emission pressures. Although technology can improve energy efficiency and promote low-carbon technologies, its application is constrained by R&D cycles, costs, policy implementation, and the lag in industrial transformation. In the short term, the reduction effect is difficult to fully manifest, thus having a relatively smaller impact on carbon emissions.

Fig. 6: Time distribution of the regression coefficients for variables in the GTWR model.
figure 6

This figure illustrates the time distribution of regression coefficients for key variables in the GTWR model from 2005 to 2022. The boxes indicate the 25–75% interquartile range (IQR), the whiskers represent 1.5 times the IQR range, the solid horizontal lines inside the boxes indicate the median values, and the black triangles represent the mean values.

Full size image

From the perspective of social development, education has a long-term lag, and the modern average length of education is getting higher and higher. When there is a large gap between training and output, it will have a promoting effect on carbon emissions. From the perspective of economic development mode, China’s economic development still depends on high intensity of energy consumption, so economic development will increase urban carbon emissions. High human capital level may also be because of the crowding effect of education investment and in a certain period of time to promote the increase of carbon emissions, but with the rise of human capital level, clean technology research and development and application of the labor base, the inhibitory effect of carbon emissions gradually offset in the process of education investment, promote the human capital level of carbon emissions show a trend of weakening year by year. In addition, the level of industrial structure, government intervention level and science and technology development level will promote carbon emissions, but the industrial structure level in the early showed inhibition. This is due to the early development of carbon emissions base being small, through the adjustment of industrial structure and the level of science and technology, and the emissions and governance have a positive effect on carbon reduction.

Spatial trend of the influencing factors of urban carbon emissions in YEB

Based on the above analysis, this paper draws that the spatial distribution of the regression coefficient of each variable in the GTWR model, and the results are shown in Fig. 7. Specifically, the industrial structure of the upper Yangtze river cities and the lower Yangtze river cities on carbon emissions has an obvious promoting effect. and the middle reaches of the Yangtze river (in Hunan province and Hubei province) most urban industrial structure level, the phenomenon and the industrial structure characteristics of the YEB, technical level and energy efficiency, stage of economic development and policy guidance factors. The industrial structure of most cities in the middle reaches of the Yangtze River will restrain the carbon emissions. This phenomenon is related to science and technology, energy utilization efficiency and policy guidance of the cities along the YEB. At the same time, although the development of the tertiary industry in the upper reaches of the Yangtze River is dominated by low-carbon industries such as catering and tourism, it may not have formed enough scale to completely offset the emissions of other high-carbon industries.

Fig. 7: Spatial distribution of the regression coefficients for variables in the GTWR model.
figure 7

This figure illustrates the spatial distribution of regression coefficients for key variables in the GTWR model. The color gradient represents the magnitude of the regression coefficients, with darker red indicating higher positive values, lighter red representing lower values. White areas indicate regions with no available data.

Full size image

The level of government intervention in Sichuan and Chongqing and parts of Hubei, Hunan and Jiangsu provinces will promote the carbon emissions in the region, while the rest of the region will inhibit the carbon emissions. The primary reason for this phenomenon lies in the fact that Sichuan-Chongqing and Hubei-Xiangjiang-Su regions are industrialized areas. The government stimulates growth through infrastructure investment and support for heavy industry. However, the high energy consumption characteristics of traditional industries and their reliance on extensive development models lead to a path dependence. Short-term policy goals prioritize green transformation, resulting in an increase in carbon emissions. In contrast, developed regions like Southern Jiangsu have completed industrial upgrading earlier. Government intervention has shifted towards low-carbon technology subsidies and the construction of carbon trading markets, forming a synergistic effect between policy and market in reducing emissions.

The level of scientific and technological development in parts of Sichuan, Yunnan, Guizhou and Hunan will inhibit the level of carbon emissions, while the level of the rest of the regions will promote the development of local carbon emissions. Government intervention in regional differences in carbon emissions stems from the gradient nature of socio-economic development stages and the asynchronous transformation of industrial structures. Regions like Sichuan and Chongqing, relying on hydropower resources, have low emission reduction bases but are constrained by the west-to-east power transmission, limiting policy space; whereas industrial-dominated areas such as Jiangsu and Hunan, through measures like dual control of energy consumption and green manufacturing, directly promote the transformation of high-carbon industries like steel and chemicals, creating an “policy-driven upgrade” effect.

Only the educational development level of Chongqing and Tongren will inhibit the development of regional carbon emission level, while with the improvement of educational development level in the other regions, the carbon emission level will also increase accordingly. The inverse relationship between educational development and carbon emissions is essentially a spatiotemporal misalignment in the evolution of regional industrial structures and environmental awareness. In southwestern China, education is deeply integrated into ecological conservation practices; Chongqing vocational schools offer courses on carbon emissions, and preschool education in Tongren incorporates green concepts. In contrast, central and eastern regions focus on technological breakthroughs; Fujian strengthens its layout in hydrogen energy and other disciplines, while Jiangsu promotes industrial greening. The process of technological industrialization temporarily increases carbon emissions. This structural divergence leads to an “inhibitory-promotive” effect of education on carbon emissions.

The cities with the inhibitory effect of economic development level on carbon emissions are mainly concentrated in the middle reaches of the Yangtze River, while a small number of cities are also located in Yunnan Province. The economic development level of the other cities will promote the growth of regional carbon emissions. The mid-Yangtze region relies on water resources to develop low-carbon services, while Yunnan leverages its ecological advantages to layout green energy. Both regions are dominated by low-energy-consuming industries, with carbon emission intensity per unit of GDP significantly lower than that of industrially-led areas. Meanwhile, intensive management enhances agricultural efficiency, and the proportion of green energy exceeds the national average, reducing the base emissions from the energy sector. In contrast, other regions, due to high industrial dependence and lagging traditional energy transitions, have seen their economic growth and carbon emissions rise in tandem.

The cities where the human capital level will inhibit carbon emissions are mainly concentrated in the lower Yangtze River regions such as Jiangsu, Shanghai and Zhejiang provinces, while the human capital level in other regions promotes the growth of carbon emissions. The lower reaches of the Yangtze River usually have higher levels of human capital, which means more high-quality talents and stronger innovation ability. These regions can more effectively drive technological progress, especially in green technology, energy efficiency and environmental technologies. Technological advances can significantly improve energy efficiency, reduce waste, and thus reduce carbon emissions. In contrast, other regions may have had poor carbon emission control results due to the low levels of human capital and the slow pace of technological innovation and adoption of new technologies.

Summary

Research conclusion

This study focuses on 108 cities along the YEB. Firstly, nighttime light remote sensing is used to measure the urban carbon emissions in the YEB. Secondly, kernel density estimation, Moran’s index and spatial Markov chain are used to explore the spatiotemporal evolution characteristics of urban carbon emission in the YEB. Finally, the GTWR model is used to explore the spatial spillover effect of carbon emission reduction in the YEB. The results show that the urban carbon emissions in the YEB show significant spatial heterogeneity and complex dynamic changes. The specific conclusions are as follows:

First, there are significant regional differences in spatiotemporal of urban carbon emissions in the YEB in time and space. In terms of time trend, carbon emissions in cities along the YEB show a significant upward trend. In terms of space, the kernel density curve of YEB urban carbon emissions shows significant tail-dragging and peak phenomenon, indicating that urban carbon emissions have significant agglomeration characteristics, and the spatial difference is in an increasing trend. Meanwhile, there are significant spatial agglomeration characteristics of urban carbon emissions in the YEB region, and the regions with carbon emissions reaching or exceeding the medium level are mainly concentrated in the lower reaches of the Yangtze River and Chongqing. After 2005, the urban carbon emissions in the YEB have a trend of increasing outward with the above cities as the agglomeration center.

Second, the urban carbon emission in the YEB has obvious spatial migration characteristics. On the one hand, many cities in the lower reaches of the Yangtze River are in a high-high agglomeration state, while cities in low-low agglomeration are mainly distributed in Sichuan, Yunnan and the middle reaches of the Yangtze River, indicating that there is significant spatial autocorrelation of urban carbon emissions in the region, and there are significant spatial distribution differences. On the other hand, the urban carbon emissions in the YEB region showed a significant upward trend, but not a jump. This phenomenon is mainly manifested in the cities with low carbon emissions will maintain the current characteristics or upward migration trend. The cities with high carbon emissions will maintain the status quo, but there will be no downward trend.

Thirdly, the factors influencing urban carbon emissions in the YEB are diverse. Among them, education development, economic development and human capital will always improve the carbon emissions in the YEB during the research period. The human capital in promoting carbon emissions is weakening year by year. Although the impact of education development and economic development on carbon emissions has undulation, the overall trend still shows a strengthening trend. Industrial structure, government intervention and scientific and technological development will have a promoting effect on carbon emissions in the early stage, but its promoting effect gradually weakens in the later stage of the sample, and then leads to an inhibitory effect in the late sample.

Policy implication

Based on the above conclusions, this paper puts forward the following policy implications:

First, the YEB region implements a carbon reduction strategy of “dual-core linkage and stepwise diffusion.” It takes the Yangtze River Delta urban agglomeration and the Chongqing metropolitan area as the core engines for low-carbon development, establishing a cross-regional carbon emission trading market. Through economic measures, it guides high-carbon industries to gradually relocate to surrounding areas. On the one hand, it enforces dual control over energy intensity and total volume in core cities, promoting the low-carbon transformation of traditional industries; in expansion zones, it establishes a mechanism decoupling carbon emission intensity from economic growth, fostering green and low-carbon industrial clusters. On the other hand, it leverages the Golden Waterway of the Yangtze River to create a low-carbon logistics corridor, promoting new energy shipping and multimodal transport to reduce carbon emissions in transportation. This strategy not only enhances the emission reduction efficiency of existing agglomeration areas but also guides peripheral regions to orderly accept industrial transfers, forming a well-structured and highly efficient low-carbon development pattern along the YEB.

Second, the YEB region implements a differentiated emission reduction strategy of “tiered control and dynamic adjustment.” Based on the spatial autocorrelation of carbon emissions, construct a cross-regional governance mechanism with “high-low linkage”: in the highly concentrated urban agglomerations downstream of the Yangtze River, promote refined carbon quota management and establish an enterprise carbon account system to minimize emission reduction costs through market transactions; in the low-carbon regions of Sichuan, Yunnan, and the midstream areas, set up carbon sink growth poles and guide their expansion of forest carbon sinks through ecological compensation. At the same time, establish a dynamic classification control mechanism to implement preventive carbon budget management for low-carbon cities, setting carbon emission growth thresholds. Innovatively establish a carbon migration warning system to predict urban carbon emission migration trajectories using satellite monitoring and AI models, allowing for preemptive deployment of emission reduction facilities. This will form a sustainable emission reduction pattern where “high-carbon cities remain stable and low-carbon cities do not leap forward.”

Third, the YEB region implements a “factor-driven transformation” carbon reduction strategy. Given the continuous improvement effect of education and economic development on carbon emissions, increased investment in higher education should be made to focus on cultivating specialized talents in new energy, carbon management, and other fields. A collaborative innovation platform for industry, academia, and research should be established to promote the industrialization of low-carbon technologies. In response to the characteristics of industrial structure changes, the green transition of traditional manufacturing needs to be accelerated, with a carbon intensity threshold set for entry. The focus should be on developing low-carbon industries such as the digital economy and modern services. Regarding government intervention, a dynamic carbon tax adjustment mechanism should be established, adjusting rates according to the lifecycle of industries. Initially, there should be appropriate incentives for low-carbon technology R&D, while later efforts should focus on curbing the expansion of high-carbon industries. In the field of scientific and technological innovation, breakthroughs in CCUS technology bottlenecks are needed, along with the establishment of a carbon removal technology certification system. Government procurement should guide technological iteration. At the same time, a carbon emission monitoring platform integrating multi-source data should be built to assess the real-time contribution of various factors to carbon emissions, providing a scientific basis for policy adjustments.

Limitations and future prospects of the study

This paper uses sample data from 108 cities in the YEB region from 2005 to 2022 to analyze the spatiotemporal evolution of urban carbon emissions. Although there are certain marginal contributions, there are still the following limitations. First, in terms of data dimensions, limited by the consistency of statistical criteria, city-level carbon emission data mostly originates from energy consumption estimates and fails to directly obtain micro-emission data for specific sectors such as enterprises and transportation. This may understate the heterogeneity of industrial point sources and living source emissions. Second, at the method application level, the spatial Markov chain’s portrayal of spatial lag effects is still crude, failing to construct multi-scale nested models to analyze internal emission hot spots within cities. Moreover, the time weight function setting in geographically weighted regression is subjective. Third, in terms of theoretical framework, there is insufficient consideration of institutional factors, such as the quantification of policy variables like carbon trading markets and environmental regulations, which fails to reveal the spatial spillover effects of policy tools on carbon emissions.

Future research can expand in three directions: First, construct an emission inventory system that integrates multi-source data, combining satellite remote sensing inversion data, corporate pollution declaration data, and traffic GPS trajectory data to establish a high-precision spatial grid model for carbon emissions. Second, develop deep learning algorithms such as spatiotemporal neural networks to capture the nonlinear dynamic characteristics of carbon emissions, focusing on analyzing how industrial relocation and the evolution of transportation networks reshape emission patterns. Third, build ABM models that include policy scenario simulations to quantify the spatial differential effects of policy tools like carbon taxes and green subsidies, and simulate emission reduction pathways under different policy combinations. Additionally, the research can expand to include international comparisons, placing the YEB in the context of global major city cluster emission reduction processes to distill universally applicable low-carbon transition patterns.