Introduction

Considering the dire circumstances surrounding global climate change, it has become imperative for all nations worldwide to assume the inescapable duty of environmental conservation. Carbon emissions are the most crucial factor influencing the global ecological environment and sustainable growth of humanity. Consequently, reducing the intensity of carbon emissions has been the primary focus of all nations worldwide (Jiang et al., 2022; Williams et al., 2017). China introduced a “dual carbon” policy in September 2020 as a means to efficiently manage the level of carbon emissions (Gao et al., 2021; Xiao et al., 2024). Studying the changes in carbon emissions across space and time in the Yellow River Basin, a crucial ecological and economic region in China, is necessary for understanding regional development (Ren et al., 2023). The objective of this research is to do a comprehensive examination of carbon emissions in the Yellow River Basin, with a focus on their spatial and temporal characteristics. This will be achieved by utilizing nighttime light data to uncover the patterns of evolution and analyze the primary elements that impact carbon emissions.

The Yellow River Basin, as a prominent center of Chinese culture, has brought attention to the dilemma of reconciling economic development with environmental preservation. The swift progress of industrialization and urbanization has led to a significant rise in energy consumption and carbon emissions. This has created immense ecological pressure on the Yellow River Basin, which has experienced an average annual carbon emission of 2.361 billion tons from 1997 to 2019 (Liu et al., 2022; Song et al., 2022). Over this period, carbon emissions in the Yellow River Basin experienced three distinct phases: no growth, rapid growth, and gradual increase. Since 2012, the rate of rise has declined, but carbon peaking has yet to be reached (Wang et al., 2023). Hence, it is imperative to precisely comprehend the spatial and temporal distribution of carbon emissions in the Yellow River Basin. This understanding is crucial for devising efficient carbon reduction programs and advancing sustainable development in the region. However, nighttime light data, being a novel form of data source, offers distinct advantages in the study of carbon emissions (Lu & Li, 2021).

Remote sensing data from night-light sources offers a more intuitive representation of the intensity and spatial distribution of human activities compared to conventional statistics data. This provides novel concepts and methodologies for examining the spatial and temporal dimensions of carbon emissions (Liu et al., 2021; Wang et al., 2024). By using extensive nighttime light data over a significant period, we can depict the evolutionary pattern of carbon emissions in the Yellow River Basin with greater precision and delve into the underlying processes that contribute to it (Du et al., 2018; Wang et al., 2023).

Researchers, both domestically and internationally, have extensively examined many aspects of carbon emissions and have generated substantial findings. Chen et al. (2023) distinguish between “unavoidable carbon emissions” and “excessive carbon emissions” in their investigation and address the optimization of carbon emissions. Wang et al. (2024) reported on the research conducted by Chinese scholars regarding different aspects of carbon emission analysis. These aspects include the computation and projection of carbon emissions, factors that influence carbon emissions, carbon footprint assessment, carbon emission efficiency, and examination of disparities in carbon emissions. The results were acquired with the CiteSpace tool. Ge et al. (2024) provided the concept of a “meta-block” that consists of static and variable carbon blocks. They then measured sectoral carbon emissions based on this concept. Wang et al. (2023) conducted a study explicitly targeting the iron and steel business. They devised a corporate carbon accounting model to quantify carbon emissions within this industry accurately. Wu et al. (2023) contended that the development of a scientific carbon emission measurement system provides enterprises with the ability to define strategies for reducing carbon emissions. Zhang et al. (2023) constructed a carbon emission evaluation model for assembled buildings in the coordination stage using the carbon emission factor method. Their study examined the influence of the overall load rate on carbon emissions and determined that the total load rate can be efficiently decreased during transportation.

In addition to the research mentioned above on carbon emissions measurement, other scholars have also investigated the variables that influence carbon emissions. Zeng et al. (2023) developed a carbon emission impact factor evaluation system using a system dynamics approach. This system considers both the origin and the destination of carbon emissions. The study concluded that policy support and adjustments in energy structure have beneficial effects in decreasing the intensity of carbon emissions. Mahmood (2023) examined the relationship between carbon emission intensity and economic growth, natural resource rent (NRR), trade openness, and foreign direct investment (FDI) in Saudi Arabia from 1968 to 2021. The study found that these indicators have both linear and non-linear impacts on carbon emissions, exhibiting phase features. Feng et al. (2024) contend that carbon trading regimes can successfully mitigate carbon emissions, and this policy has gained widespread acceptance worldwide. In their study, Zhang et al. (2023) examined the urban-rural disparities in China while investigating the problem of carbon emissions. They highlighted that raising people’s income would effectively mitigate the underlying factors driving carbon emissions.

Regarding driving factors, researchers have identified four distinct patterns to describe China’s carbon emission trends: aggregated, discrete, overlapping, and idiosyncratic (Li et al., 2024). Wu et al. (2023) contend that financial agglomeration can contribute to the reduction of carbon emissions. Chen et al. (2023) investigated the impact of urban polycentricity and decentralization on carbon emissions in 95 prefecture-level communities in southeastern China by utilizing nighttime light data. The study focused on various sectors. The study highlighted that nighttime light data may be utilized to assess the spatial organization of the town. The study also found that a city with a polycentric and dispersed spatial form is linked to a reduction in both the total amount of carbon emissions and carbon emissions, specifically from transportation. The scholar’s research concept serves as the guiding principle for this study. In addition, some individual scholars utilize nighttime light data to analyze different land use types, which serves as a relevant point of reference for this paper (Zheng et al., 2020).

Furthermore, particular academics have done comprehensive investigations utilizing spatial measures derived from the attributes of spatial dispersion. Du et al. (2022) showed that carbon emissions from China’s construction industry had a spatial spillover impact. Additionally, it was shown that the constructing industry’s carbon emissions exhibit spatial agglomeration traits, with notable variations in regional development. This is evident from the elevated levels in the eastern regions and the lower levels in the western regions. Wang et al. (2024) utilized a spatial inversion approach to analyze the characteristics of carbon emissions’ geographical distribution in the Beijing-Tianjin-Hebei region. They employed nighttime light data and land use data to reveal the geographical distribution characteristics of carbon emissions over an identical region. Zhang et al. (2022) used energy consumption data to investigate the spatial distribution patterns and carbon emission levels of different land use categories in the Yellow River Delta region. Some scholars have also employed panel data to investigate the geographical dispersion of China’s carbon emissions and the factors that influence it (Li et al., 2021; Pan et al., 2022).

Although research on carbon emissions has steadily increased in recent years, there remains significant scope for further exploration in this field. Regarding the estimation of carbon emissions, current studies primarily rely on statistical data on energy. However, due to factors such as local government performance evaluations, energy consumption data may be withheld or misreported. This can lead to inaccuracies in the results derived from multiplying fossil fuel consumption by carbon emission factors, with some regions exhibiting significant deviations from actual levels. Moreover, existing studies often focus on provincial-level units, with relatively few examining cities or conducting regional analyses based on city-level data. Furthermore, research on the factors influencing carbon reduction frequently employs linear regression or spatial econometric models, which are limited to analyzing overall effects and cannot capture the temporal and spatial variations in influencing factors for individual provinces across different years. To address these gaps, this study uses city-level data to investigate the spatiotemporal evolution and influencing factors of carbon emissions in the Yellow River Basin. Nighttime light remote sensing data is utilized to estimate the carbon emissions of 94 prefecture-level cities in the region, forming the primary dataset for this research. Spatial autocorrelation models and Gini coefficients are applied to explore the spatiotemporal evolution of carbon emissions. In contrast, the GTWR model is used to analyze the temporal and spatial variations in the impact coefficients of carbon reduction factors across the region. These analyses provide quantitative support for developing national and regional emission reduction and carbon neutrality policies in the Yellow River Basin, contributing to the acceleration of carbon peaking and carbon neutrality efforts.

Regional overview and data sources

Overview of the study area

This paper focuses on nine provinces in the Yellow River Basin: Qinghai, Sichuan, Gansu, Ningxia, Inner Mongolia, Shanxi, Shaanxi, Henan, and Shandong. For data validity, 94 prefectural-level municipalities are selected as samples. As shown in Fig. 1, The data primarily encompass night-time lighting impact, energy consumption, and socio-economic statistics.

Fig. 1: Map of administrative divisions of the Yellow River Basin.
figure 1

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

Full size image

Data sources

  1. (1)

    Nighttime Light Data. This study predominantly gathers data using the NOAA platform, utilizing DMSP-OLS data for the time frame between 2005 and 2013 and NPP-VIIRS data for the period from 2012 to 2022. Due to the differences in format between these datasets, it is necessary to perform an initial data normalization operation. Accordingly, this study employed calibrated nighttime light remote sensing data, building upon the findings of various scholars (Wang et al., 2023; Wang et al., 2022). Upon the completion of unified time-series continuity corrections for NPP-VIIRS and DMSP-OLS data, a long-term series of Yellow River Basin nighttime light data from 2010 to 2022 is collected (Wang et al., 2023).

  2. (2)

    Energy consumption data. This analysis utilized energy consumption data extracted from the ‘China Energy Statistical Yearbook’ spanning the years 2010 to 2022. Subsequent estimations of carbon emissions were then derived from this data.

  3. (3)

    This study conducted an extensive analysis of both national and international literature (Sun et al., 2022; Zhang & Tan, 2016) to discover the factors that influence the levels of carbon emissions in different regions. After assessing the practical aspects and considering the availability of data and its practical significance, the study selected economic growth, industrial structure, scientific advancement, educational attainment, financial development, and infrastructure as the five determinants. ①Economic Development: Based on the Environmental Kuznets Curve (EKC) theory, economic development, and carbon emissions exhibit an inverted U-shaped relationship, where economic growth initially increases emissions but later reduces them. This study uses per capita GDP as a measure of economic development. ②Industrial Structure: A shift in industrial structure towards low-carbon industries, such as services and high-tech sectors, reduces carbon emissions. High-tech industries not only emit less but also drive emission reductions in other sectors through technological innovation. The advanced industrial structure index is used to measure this, calculated as: value added of the primary industry * 1 + proportion of the secondary industry * 2 + proportion of the tertiary industry * 3. ③Scientific Development: Advanced industrial technologies enhance energy efficiency by reducing energy consumption and emissions per output unit, fostering sustainable economic and environmental development. This study measures scientific development using the proportion of fiscal expenditure on science and technology to total fiscal expenditure. ④Education Level: High-quality education fosters talent that drives green technology research and industrial transitions towards low-carbon pathways, improving energy efficiency and reducing emissions. Education levels are measured by the proportion of fiscal expenditure on education to total fiscal expenditure. ⑤Financial Development: Financial development supports carbon finance mechanisms, such as carbon trading markets, which incentivize emission reductions and channel investment into low-carbon fields like clean energy. This study measures financial development using the ratio of total loans and deposits of financial institutions to GDP. ⑥Infrastructure: The growth of postal and telecommunications networks enables remote work and online transactions, reducing travel demands and transportation-related emissions. Infrastructure is measured using the postal and telecommunications business volume ratio to GDP. The carbon emission effect factors in Table 1 were obtained for the Yellow River Basin. Economic and social statistical data for cities in the region from 2010 to 2022 were obtained from the China Statistical Yearbook and China City Statistical Yearbook.

    Table 1 Selection of carbon emission impact factors in the Yellow River Basin.
    Full size table

Methodology

Estimation of energy carbon emissions

This study utilizes energy consumption data collected from 94 prefecture-level cities in the Yellow River Basin spanning from 2010 to 2022. The energy conversion coefficients provided by the IPCC (2006) are employed to ascertain and compute carbon dioxide emissions from nine primary energy sources, Table 2 presents the carbon emission factors for various energy types. The formula used to calculate carbon emissions is shown below:

$${{CO}}_{2}=\frac{44}{12}\times \mathop{\sum }\limits_{i=1}^{9}{K}_{i}{E}_{i}$$
(1)
Table 2 Carbon emission factors for various energy sources.
Full size table

In Eq. (1): \(i\) denotes the energy category; \({K}_{i}\) denotes the carbon emission factor for \(i\) energy source (as shown in Table 1); \({E}_{i}\) denotes the amount of \(i\) energy used.

Dagum Gini coefficient

In 1912, Italian economist Corrado Gini established the Gini coefficient as a means of precisely measuring the income and distribution differences among residents in a country or region. Subsequently, American economist Albert O. Hirschman further developed and improved this notion, which has now become the universally accepted approach worldwide for evaluating inequality in the distribution of income or wealth (Gao et al., 2023). The Gini coefficient is a crucial economic indicator in statistics that quantifies the extent of income inequality within a population by calculating the ratio of unequal income to total income. The utilization of it is described in Eqs. (2) to (10):

$$G=\frac{1}{2\bar{y}{n}^{2}}\mathop{\sum }\limits_{j=1}^{k}\mathop{\sum }\limits_{h=1}^{k}\mathop{\sum }\limits_{i=1}^{{n}_{j}}\mathop{\sum }\limits_{i=1}^{{n}_{h}}|{y}_{{ji}}-{y}_{{hr}}|$$
(2)
$${G}_{{jj}}=\frac{1}{2{\bar{y}}_{j}{{n}^{2}}_{j}}\mathop{\sum }\limits_{i=1}^{{n}_{j}}\mathop{\sum }\limits_{r=1}^{{n}_{j}}|{y}_{{ji}}-{y}_{{jr}}|$$
(3)
$${G}_{{jh}}=\frac{1}{{n}_{j}{n}_{h}({\bar{y}}_{j}+{\bar{y}}_{h})}\mathop{\sum }\limits_{i=1}^{{n}_{j}}\mathop{\sum }\limits_{r=1}^{{n}_{h}}|{y}_{{ji}}-{y}_{{hr}}|$$
(4)
$${G}_{w}=\mathop{\sum }\limits_{j-1}^{k}{G}_{{jj}}{P}_{j}{S}_{j}$$
(5)
$${G}_{{nb}}=\mathop{\sum }\limits_{j=2}^{k}\mathop{\sum }\limits_{h=1}^{j-1}{G}_{{jj}}({P}_{j}{S}_{h}+{P}_{h}{S}_{j}){D}_{{jh}}$$
(6)
$${G}_{t}=\mathop{\sum }\limits_{j=2}^{k}\mathop{\sum }\limits_{h=1}^{j-1}{G}_{{jj}}({P}_{j}{S}_{h}+{P}_{h}{S}_{j})(1-{D}_{{jh}})$$
(7)
$${D}_{{jh}}=\frac{({d}_{{jh}}-{P}_{{jh}}{S}_{j})}{({d}_{{jh}}+{P}_{{jh}}{S}_{j})}$$
(8)
$${d}_{jh}={\int }_{0}^{\infty }d{F}_{j}(y){\int }_{0}^{\infty }(y-x)d{F}_{h}(x)$$
(9)
$${P}_{jh}={\int }_{0}^{\infty }d{F}_{j}(y){\int }_{0}^{\infty }(y-x)d{F}_{j}(x)$$
(10)

The symbols \(k\) and \(n\) in the above equation reflect the number of regions and the number of samples chosen from these regions, appropriately. \({n}_{j}\) and \({n}_{h}\) represent the sample sizes for areas \(j\) and \(h\), respectively. \({y}_{{ji}}\) and \({y}_{{hr}}\) indicates the level of carbon performance in various places. \(i\) indicates the mean carbon emissions level of every prefecture-level city in the Yellow River Basin.

Spatial-temporal geographically weighted regression models

The Geographically and Temporally Weighted Regression (GTWR) model is a frequently used data analysis approach that is highly skilled at analyzing data that exhibits both spatial and temporal dynamics. This model enhances the traditional Geographically Weighted Regression (GWR) model by including non-stationary fluctuations in both geographical and temporal dimensions. As a result, it overcomes the constraints of the GWR model when dealing with data that show spatiotemporal variability (Tan et al., 2021). When selecting the GTWR model, it is crucial to consider the geographical correlation between the dependent and independent variables, as well as the spatial and temporal features of the data. Bandwidth selection is a crucial stage in the GTWR model as it defines the extent of local variations across various locations and periods. A narrower bandwidth can lead to more rapid geographical variations in the results, whereas a broader bandwidth brings the conclusions closer to those of a global model. The model is calculated as follows:

$${C}_{i}={\alpha }_{0}\left({u}_{i},{v}_{i},{t}_{i}\right)+\mathop{\sum}\limits _{k}{\alpha }_{k}({u}_{i},{v}_{i},{t}_{i}){x}_{{ik}}+{\beta }_{i}$$
(11)

In Eq. (11), \({C}_{i}\) represents the carbon emission level of region \(i\); \({x}_{{ik}}\) is the observed value of indicator \(k\) in city \(i\); \(\left({u}_{i},{v}_{i},{t}_{i}\right)\) denotes the spatial coordinates of city \(i\); \({\alpha }_{0}({u}_{i},{v}_{i},{t}_{i})\) is the spatiotemporal intercept for city \(i\); \({\alpha }_{k}({u}_{i},{v}_{i},{t}_{i})\) is the regression coefficient of indicator \(k\) in region \(i\); \({\beta }_{i}\) indicates the residuals.

$${\left({d}_{{ij}}\right)}^{2}=\lambda \left[{\left({u}_{i}-{u}_{j}\right)}^{2}+{\left({v}_{i}-{v}_{j}\right)}^{2}\right]+\mu {\left({t}_{i}-{t}_{j}\right)}^{2}$$
(12)
$${W}_{{ij}}=\exp \left[-{\left(\frac{{d}_{{ij}}}{h}\right)}^{2}\right]$$
(13)

In Eq. (12), \({d}_{ij}\) represents the distance between city \(i\) and \(j\); \(\lambda\) and \(\mu\) are scaling factors for spatial and temporal distances respectively, both defined as non-negative number. In Eq. (13), \({W}_{{ij}}\) is the calculated weight matrix, and \(h\) denotes the bandwidth.

Spatial autocorrelation analysis

Spatial autocorrelation analysis ensures data values are correlated in geographic space before conducting spatial analyses of indicators. Using the Moran’s I, this analysis determines the degree of correlation between spatial locations and their neighbors. Positive values signify positive spatial autocorrelation, negative values indicate negative spatial autocorrelation, and the absence of spatial dependence denotes spatial randomness. To accurately capture the spatial agglomeration characteristics of carbon emissions in the Yellow River Basin, this paper employs both global and local Moran’s I.

The global Moran’s I provide insight into the overall spatial pattern of data, indicating if there is a tendency towards clustering or dispersion. On the other hand, the local Moran’s I identify the specific locations where clustering or dispersion is occurring. The equation for the above method is as follows:

$${Mora}{n}^{{\prime} }s\,I=\left[\mathop{\sum }\limits_{i=1}^{n}\mathop{\sum }\limits_{j=1}^{n}{W}_{{ij}}({Y}_{i}-\bar{Y})({Y}_{j}-\bar{Y})\right]\div\left({S}^{2}\mathop{\sum }\limits_{i=1}^{n}\mathop{\sum }\limits_{j=1}^{n}{W}_{{ij}}\right)$$
(14)
$${I}_{i}=({Y}_{i}-\bar{Y})\div\mathop{\sum }\limits_{i=1}^{n}{({Y}_{i}-\bar{Y})}^{2}\times \mathop{\sum }\limits_{j=1}^{n}{W}_{{ij}}({Y}_{j}-\bar{Y})$$
(15)

Equations (14) and (15) describe the calculation methods for the global Moran’s I, and the local Moran’s I, respectively. In Eq. (14), \(n\) represents the number of cities, with this study including 94 prefecture-level cities, hence n = 94; \({Y}_{i}\) and \({Y}_{j}\) denote the carbon emission levels of cities \(i\) and \(j\) respectively; \({W}_{{ij}}\) is the spatial weights matrix; \(\bar{Y}\) is the sample mean; \({S}^{2}\) is the sample variance; In Eq. (15), \({I}_{i}\) represents the local autocorrelation value, with other symbols retaining the same meanings as in Eq. (14).

Analysis of results

Measurement of carbon emissions from energy consumption in the Yellow River Basin

This study estimates the carbon emissions of 94 prefecture-level cities in the Yellow River Basin using nighttime light remote sensing data and calculates the provincial averages. The results, shown in Table 3, summarize the temporal trends of carbon emissions across regional provinces. As indicated in Table 3, all provinces exhibited an upward trend in carbon emissions during the sample period. Sichuan experienced the largest increase, with emissions rising from 66.57 Mt CO₂ in 2010 to 76.17 Mt CO₂ in 2022, an increase of 9.60 Mt CO₂ and an average annual growth rate of approximately 1.202%. Conversely, Shandong showed the smallest increase, with emissions rising by only 0.63 Mt CO₂ during the same period, indicating minimal change. Overall, the average carbon emissions of cities in the Yellow River Basin rose from 76.32 Mt CO₂ in 2010 to 80.59 Mt CO₂ in 2022, an increase of 4.27 Mt CO₂, with an average annual growth rate of approximately 0.466%. Notably, the growth rates in Inner Mongolia, Sichuan, Gansu, and Ningxia exceeded the regional average, while those in other provinces were below average. Nonetheless, the overall trend indicates a significant increase in carbon emissions across the Yellow River Basin.

Table 3 Average carbon emissions growth in provinces of the Yellow River Basin.
Full size table

To further explore the development trends and characteristics of carbon emissions in the Yellow River Basin. The study used ArcGIS software to generate geographic distribution maps of carbon emissions resulting from energy consumption in the Yellow River Basin. These maps are depicted in Fig. 2. Four specific time points were chosen: 2010, 2014, 2018, and 2022. The carbon emission levels were classified into five stages: high, relatively high, medium, low, and comparatively low.

Fig. 2: Spatial distribution of carbon emissions from energy consumption in the Yellow River Basin.
figure 2

This figure illustrates the spatial distribution of carbon emissions in the Yellow River Basin across different years. The maps depict the carbon emission levels in 2010, 2014, 2018, 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

Significant trends are revealed by analyzing the regional distribution maps of carbon emissions in the Yellow River Basin for 2010, 2014, 2018, and 2022. Concentrations of high and relatively high-emission cities were found in the E-Yu urban agglomeration, Xi’an, Chengdu, and the Shandong Peninsula in 2010, whereas 29 cities had low emissions and 54 had comparatively low emissions. Three cities had high emissions, nine had relatively high, two had medium, and 54 had relatively low. A few towns in the area have already reached high emission levels by 2014, including Yulin, Xi’an, Yan’an, Zhengzhou, Linyi, and Jinan. Overall emissions increased in 2018, with the E-Yu urban agglomeration, Xi’an, and portions of the Shandong Peninsula shifting to higher levels. Mid- and high-emission cities increased significantly between 2010 and 2022, whereas low- and medium-emission cities decreased.

Half of the locations in the Yellow River Basin had medium carbon emissions by 2022, down from 52 in 2010. The number of cities with low or lower emissions fell to 42. Cities in the E-Yu region, Xi’an, Chengdu, and Shandong Peninsula remained hotspots for pollution. The carbon emissions in the Yellow River Basin experienced a rise from 2010 to 2022, primarily concentrated in urban agglomerations where both high- and low-emission cities are located. Rising overall energy usage, particularly of coal, had a major effect on emissions of carbon dioxide. Urbanization and population growth pushed up energy use and demand, which in turn pushed up carbon emissions.

Characterization of the spatial distribution of carbon emissions in the Yellow River Basin

Spatial autocorrelation analysis of carbon emissions in the Yellow River Basin based on the Moran’s I

To further investigate the spatiotemporal characteristics of urban carbon emissions in the Yellow River Basin, this study employs three spatial weight matrices: the spatial critical distance weight matrix, the first-order inverse distance spatial weight matrix, and the spatial adjacency weight matrix. Using carbon emission data from 2010 to 2022, we calculated the global Moran’s I for the sample period, with the results summarized in Table 4. The results indicate that the global Moran’s I calculated under all three spatial weight matrices consistently produced positive values, with p-values remaining below 0.001. This confirms the presence of significant spatial autocorrelation in urban carbon emissions within the Yellow River Basin, indicating a strong spatial clustering pattern. However, when examining the temporal evolution trend, it is observed that the Moran’s I did not exhibit significant fluctuations over the study period. This suggests that the spatial agglomeration arrangement of carbon emissions in this area remains relatively stable. Meanwhile, from the results of different spatial weight matrices, it can be found that although there are variations in the global Moran’s I values, the overall trend, direction, and significance remain unchanged. This verifies the robustness of the above spatial autocorrelation.

Table 4 Moran’s I for the Yellow River Basin, 2010–2022.
Full size table

This paper conducts a local Moran’s I test on carbon emissions in the Yellow River Basin at four distinct time intervals: 2010, 2014, 2018, and 2022. The purpose is to investigate further the spatial patterns of carbon emissions in the area. The test aims to identify the internal clustering characteristics of carbon emissions within the region. Figure 3 displays the results. In general, the annual count of cities in China’s Yellow River Basin that exhibit high carbon emissions and high agglomeration is falling annually, whereas the quantity of cities with low carbon emissions and low agglomeration is initially increasing and subsequently declining.

Fig. 3: LISA of carbon emissions in the Yellow River Basin, 2010–2022.
figure 3

This figure illustrates the LISA clustering results of carbon emissions in the Yellow River Basin over different years. The maps depict the spatial clustering patterns in 2010, 2014, 2018, 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

In 2010, the cities in Shanxi Province, Shandong Province, and Henan Province had high levels of agglomeration. On the other hand, cities in Gansu Province, such as Wuwei City, Xining City, Dingxi City, and Tianshui City, as well as most cities in Sichuan Province, had low levels of agglomeration. The phenomenon can be ascribed to the higher demand for resources within the areas of energy consumption, population growth, and openness to the outside world in Shanxi, Henan, and Shandong provinces. As a result, these provinces exhibit high levels of agglomeration in resource utilization. On the other hand, Northwest China and Sichuan have a slower pace of industrialization and enjoy an advantage in terms of total carbon emissions compared to heavily populated provinces like Shandong and Henan. Consequently, these regions demonstrate low levels of agglomeration. Industrialization in the Northwest and Sichuan areas is relatively slower than in the populous provinces of Shandong and Henan. Nevertheless, the Northwest and Sichuan regions experience a low concentration of emissions because of their lower total carbon emissions.

In 2014, there was a slight rise in the number of communities in the Yellow River Basin that exhibited a significant amount of concentration of carbon emissions. Jinzhong City, Xinzhou City, Shuozhou City, and Yulin City have formed high-high agglomerations in terms of carbon emissions, while Xuchang City, Pingdingshan City, and Zhoukou City no longer hold the status of high-high agglomeration. At the same time, there is a minor increase in the number of cities exhibiting low-low concentration, with Baoji and Ankang emerging as new low-low-concentration cities. In 2018, there was a notable decline in the number of towns situated inside the high-high agglomeration area, particularly in the Shandong Peninsula and Lvliang. Additionally, there was a slight decrease in the number of cities located in the low-low agglomeration region, with Jiuquan transitioning into a low-low agglomeration state. In 2022, there will be a notable reduction in the number of cities in the Yellow River Basin in China that have both elevated and diminished levels of carbon emissions. The cities exhibiting significant carbon emissions are still primarily found in Shandong Province and Lvliang City, while the cities with low concentrations are spread across Gansu Province, Shaanxi Province, and Sichuan Province.

Characterization of regional differences in carbon emissions in the Yellow River Basin based on the Dagum Gini coefficient

This research incorporates the pertinent data of the Yellow River Basin into Eqs. (2) to (10) and utilizes the Gini coefficient to assess the decomposition outcomes of the regional inequalities in carbon emissions within the Yellow River Basin over the specified time. The results are presented in Table 5. This research will examine the attributes underlying regional disparities in carbon emissions within the Yellow River Basin, focusing on three aspects: overall disparities, disparities within regions, and disparities between regions, as shown in Table 5.

Table 5 Dagum Gini coefficient for emissions in the Yellow River Basin.
Full size table

Firstly, by measuring the Gini coefficient of carbon emissions from 2010 to 2022, it is evident that there is no clear pattern in the change of this coefficient. It fluctuates throughout this period, with the highest value occurring in 2016 and the lowest value in 2022. This suggests that the overall disparity in carbon emissions in the Yellow River Basin was most significant in 2016 and most minor in 2022. The overall difference coefficient exhibits a declining pattern, with the total difference Gini coefficient decreasing from 0.401 in 2010 to 0.317 in 2022. This represents a reduction of 20.9 percent, demonstrating that the overall disparity in carbon emissions within the Yellow River Basin is gradually diminishing over time despite irregular fluctuations.

Secondly, when examining the Gini coefficient of intra-regional disparities, it becomes evident that the average Gini coefficient is highest in the upstream region of the Yellow River, afterward in the middle reaches, and lowest in the downstream region during the period of 2010–2022. This indicates a pattern of increasing intra-regional disparities from the downstream to the middle reaches and further to the upstream of the Yellow River.

Thirdly, the Gini coefficient of inter-regional disparities reveals that the carbon emission disparity between the upstream and middle reaches of the Yellow River is more pronounced, with mean values of 0.407 and 0.408, respectively. In contrast, the mean value of the middle and downstream regions is 0.295, indicating a lower inter-regional difference in carbon emission compared to other areas. This indicates that the disparity in carbon emissions between the middle and lower sections of the Yellow River is smaller than the disparity between the upper and middle sections and the lower sections of the Yellow River. The phenomenon can be ascribed to the slower economic ascribed in the cities located in the upstream regions of the Yellow River. These cities have an industrial structure that is primarily focused on agriculture and light industry, with fewer heavy industries and energy-consuming sectors. Consequently, their carbon emissions are comparatively low. The cities situated in the middle region of the Yellow River have a greater degree of economic advancement, and their industrial structure is focused on heavy industries, particularly coal, electric power, and iron and steel. These industries have a high carbon emission intensity, which leads to higher overall carbon emissions in the middle reaches of the region. On the other hand, the cities in the lower reaches of the Yellow River have a more considerable proportion of heavy industries, along with the demand for transportation specific to coastal regions, such as maritime transport. This combination also results in relatively high carbon emissions.

Finally, upon careful examination, it is found that the mean value of the differential contribution rate between intra-region and inter-region is 31.190 and 37.097, respectively. Additionally, the mean value of hypervariable density is 31.713, which is higher than the mean value within the intra-region. The data above indicate that the variation in carbon emissions in the Yellow River Basin is impacted not only by the disparity in carbon emissions between different regions and within regions but also by the interactions among individuals or entities during population movement, trade, and the exchange of resources between different regions.

Analysis of carbon emission influencing factors in the Yellow River Basin

Prior to employing the GTWR model, it is imperative to conduct a covariance test on the variable data in this study to ascertain that the selected variables and their data do not influence the conclusions because of the interconnections between the variables. The covariance test is frequently employed using the Variance Inflation Factor (VIF) method. If the VIF value exceeds 10, it signifies the presence of an interaction between the variables and the data utilized in the study, which could impact the study’s conclusions. The statistical findings of the covariance test are provided in Table 6. The variables in this study have VIF values below 3, demonstrating no issue of multiple covariance. Therefore, the GTWR model can be used to explore these variables.

Table 6 Tests for covariance.
Full size table

After conducting collinearity tests, this study compared the fitting results of the GTWR, GWR, TWR, and OLS models, as presented in Table 7. The results show that the GTWR model achieved an AICc value of 1723.02 and an R² of 0.502. Compared with the TWR, GWR, and OLS models, the GTWR model exhibits the lowest AICc value and the highest R², indicating superior model performance. Thus, the GTWR model was selected for constructing the regression model to investigate the influencing factors of urban carbon emissions in the Yellow River Basin, as it provides better explanatory power and reliability than the GWR, TWR, and OLS models.

Table 7 Regression model fitting results.
Full size table

This study conducted a cross-validation analysis to further validate the model selection’s appropriateness. The Yellow River Basin was divided into three sections based on geographic location: the upper, middle, and lower reaches. Regression fitting was performed separately for each region; the results are shown in Table 8. The GTWR model achieved R² values of 0.664, 0.488, and 0.655 for the upper, middle, and lower reaches, respectively, and corresponding AICc values of 572.39, 481.33, and 33.69. Compared with other models, the GTWR model consistently exhibited the highest R² values and the lowest AICc values across all regions. These results further prove that the GTWR model is the most suitable choice for constructing the regression model to explore the influencing factors of urban carbon emissions in the Yellow River Basin.

Table 8 Regression fitting results of regional models in the Yellow River Basin.
Full size table

The regression findings indicate that the GTWR model consistently demonstrates that economic development, industrial structure, scientific advancement, and education exert a beneficial influence on carbon emissions in the Yellow River Basin. Conversely, the progress of financial development consistently restrains carbon emissions. Infrastructure first contributes to carbon emissions but eventually reduces them. The findings are depicted in Fig. 4.

Fig. 4: Temporal distribution of regression coefficients of variables in the GTWR model.
figure 4

This figure illustrates the temporal distribution of regression coefficients for key variables in the GTWR model from 2010 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

Economic development’s average regression coefficients reveal a pattern of modest expansion from 2010–2017 and a slight decrease from 2017–2022. The initial positive effect of economic growth in the Yellow River Basin on carbon emissions diminishes over time as energy-intensive industries are replaced with more environmentally friendly, low-carbon alternatives. There is a clear pattern of declining average regression coefficient for industrial structure, with the steepest drop happening between 2019 and 2022. This suggests that the enabling influence on carbon emissions is gradually reducing.

Since plenty of funds and assets are needed to implement new technology and phase out old ones, carbon emissions could spike for a while during the first stage of industrial upgrading. On the other hand, renewable energy and energy efficiency are two examples of new technologies and procedures that will mitigate the influence of industrial upgrading on carbon emissions. Scientific progress also shows a constant declining trend in its regression coefficients, suggesting that its positive influence on carbon emissions is gradually waning. Consistently promoting carbon emissions, the average regression coefficient for education level changes but stays over zero. One explanation is that governmental funding for education is taking resources away from environmental stewardship and the advancement of low-carbon technologies. Yet, the average regression coefficient for financial development’s impact on carbon emissions is always negative, indicating a constant inhibitory influence. To facilitate the shift to a low-carbon manufacturing model, economic development initiatives, including green industry preferential loans and other forms of technical and industrial support, are put in place.

Infrastructure’s average regression coefficient is positive from 2010 to 2016 and negative from 2017 to 2022, suggesting that it initially induces carbon emissions in the Yellow River Basin but subsequently diminishes them. Phases of infrastructure construction and utilization are associated with this pattern. Early stages necessitate substantial amounts of energy and basic materials, including steel and cement, which cause the rise in carbon emissions. Construction and transportation equipment also contribute to elevated emissions. During the later stages, energy efficiency is enhanced by the implementation of more stringent environmental standards and the development of more efficient technologies, equipment, and management strategies, thereby reducing carbon emissions.

In the previous analysis, we examined the temporal distribution of the regression coefficients in the GTWR model. Based on these findings, we have generated a spatial distribution plot of the regression coefficients for the variables, as shown in Fig. 5. The cities that exhibit a robust positive correlation between economic growth and carbon emissions are primarily situated in three districts along the Yellow River: the upstream region (including Lanzhou, Tianshui, Qingyang, Guyuan, Xianyang, Erdos, Chengdu, Mianyang, and Dazhou), the middle reaches region (including Taiyuan and Jiaozuo), and the lower reaches region (including Jinan, Weifang, Jining, and Linyi). The cities exhibiting a tenuous positive link are primarily situated in Zhangye, Baotou, Datong, Xi’an, Tai’an, and Anyang. Regional variations in the link between economic growth and carbon emissions are substantial, and its average regression coefficient exhibits the characteristics of the upper reaches of the Yellow River > the lower reaches of the Yellow River > the middle reaches of the Yellow River.

Fig. 5: Spatial distribution of regression coefficients of variables in the GTWR model.
figure 5

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 red indicating higher positive values, yellow representing lower values.

Full size image

The relationship between the composition of industries and carbon emissions is notably strong in the cities located in the hinterland of Gansu, southern Sichuan, northern Inner Mongolia, south Henan, and western Shandong. However, this correlation is comparatively weaker in the other prefecture-level cities. The locations above likely possess robust urban centers with significant industrial growth in each province, fostering the advancement of industries from primary to secondary and tertiary sectors that demand increasing levels of energy consumption. The association between scientific progress and carbon emissions in the cities of Zhangye, Tianshui, Xi’an, Baotou, Datong, Chengdu, Zhengzhou, and Tai’an is quite evident, with a more significant average regression coefficient compared to other factors. This could be attributed to the fact that the regions above are either provincial capitals or near provincial capitals. These places experience a greater demand for and utilization of resources along the course of technological advancement, resulting in a more pronounced positive link with carbon emissions. The spatial distribution map of education level reveals that cities such as Lanzhou City, Guyuan City, Xianyang City, E-Yu urban agglomeration, Taiyuan City, and part of Shandong Peninsula exhibit a strong positive correlation with carbon emission. However, the association between education level and carbon emission in Xi’an City is inhibitory. The tendency can be ascribed to the fact that the cities above are advanced in terms of education and receive more subsidies for education than for carbon emission reduction initiatives. This ultimately results in a rise in carbon emissions. However, Xi’an has abundant educational resources, and the reduction in carbon emissions arising from the export of talented individuals outweighs the indirect increase in carbon emissions caused by high financial subsidies. As a result, Xi’an exhibits a suppressive effect.

Moreover, the average regression coefficients of financial development and infrastructure are both positive and negative, suggesting that the variables above have varying effects on carbon emissions across different locations. To be more precise, the correlation between financial development and carbon emissions is positive in Qingyang, Tianshui, Xianyang, Guyuan, Chengdu, Mianyang, Bazhong, and certain cities in Shandong and Shanxi Provinces. Conversely, it is harmful in Jiuquan, Zhangye, and other cities in northern Gansu Province; Ulanqab, and other towns in the north of Inner Mongolia; Yulin, Baoji, and other cities in Shaanxi Province; Leshan, Deyang, and other cities in Sichuan Province, together with Liaocheng, Heze, and other cities in Shandong Province. This might be because the cities above are still in the initial stages of their financial development, where there is not a great demand for resources. Out of all the prefecture-level cities, the ones in Gansu’s northern regions, Shaanxi’s central cities, Sichuan’s southern cities, Shanxi’s eastern cities, Inner Mongolia’s northern cities, and Shandong’s western cities are the only ones where infrastructure and carbon emissions exhibit a negative correlation. The Yellow River Basin is seeing uneven infrastructure development, which is having an uneven impact on carbon emissions. The cities that are displaying a negative connection tend to be less developed ones, as they use less energy during infrastructure construction.

Summary

Conclusions

This study focuses on 94 prefecture-level cities in the Yellow River Basin from 2010 to 2022. First, the carbon emissions of these cities were estimated based on nighttime light remote sensing data. Second, the regional disparities and spatial evolution characteristics of urban carbon emissions were examined using spatial autocorrelation models and the Dagum Gini coefficient. Finally, the GTWR model was employed to analyze the factors influencing urban carbon reduction in the region.

First, the results indicate that urban carbon emissions in the Yellow River Basin showed an overall upward trend from 2010 to 2022, with the average emissions increasing from 76.32 Mt CO₂ in 2010 to 80.59 Mt CO₂ in 2022. The spatial distribution characteristics reveal that urban carbon emissions in the Yellow River Basin exhibit a significant spatial clustering pattern. Cities with high and relatively high emission levels are primarily concentrated in the Eyu urban agglomeration, Xi’an, Chengdu, and the Shandong Peninsula urban agglomeration. In contrast, cities with medium and lower emission levels are distributed around these urban clusters. Carbon emissions in the Yellow River Basin exhibit significant spatial clustering characteristics. As of 2022, the number of cities with high-high and low-low spatial clusters in the region has significantly decreased. High-high clusters are primarily concentrated in Shandong Province and Lvliang City, while low-low clusters are mainly scattered across Gansu, Shaanxi, and Sichuan Provinces.

Second, significant spatial disparities in urban carbon emissions were observed. Within-region differences followed the trend: Upper Yellow River > Middle Yellow River > Lower Yellow River. Inter-region disparities showed that the differences between the middle and lower reaches were smaller than between the upper and middle reaches or between the upper and lower reaches. These disparities were influenced not only by inter- and intra-regional differences but also by interactions among regions, such as population movement, trade, and factor flows, which played a critical role in shaping the observed patterns.

Third, the factors influencing urban carbon emissions in the Yellow River Basin are complex and multifaceted. The GTWR model results show that economic development, industrial structure, scientific progress, and education levels consistently promoted carbon emissions. However, the advancing industrial structure and scientific development levels gradually reduced their promoting effects, suggesting that optimizing industrial structures and fostering scientific development can effectively lower carbon emissions. Financial development exhibited a sustained inhibitory effect on carbon emissions, while infrastructure showed a pattern of initially promoting and later suppressing emissions. These findings provide valuable insights for developing strategies to reduce carbon emissions in the Yellow River Basin.

Discussion

This study investigates the spatiotemporal evolution and influencing factors of urban carbon emissions in the Yellow River Basin from 2010 to 2022. Based on provincial carbon emissions data, carbon emissions for 94 prefecture-level cities in the Yellow River Basin were estimated. Using Moran’s I, the Dagum Gini coefficient, and the GTWR model, this study explored the region’s spatiotemporal evolution and influencing factors of carbon emissions. This study reached three main conclusions:

First, carbon emissions in the Yellow River Basin show a significant upward trend. Wang and Qu (2024) utilized industrial development and energy consumption data from 2000 to 2019 to estimate industrial carbon emissions in the Yellow River Basin using IPCC methods. They found that industrial carbon emissions in the region exhibited a fluctuating upward trend, with a spatial distribution pattern of “low emissions in the upper reaches and high emissions in the middle and lower reaches,” consistent with this study’s findings. Similarly, Song et al. (2022) noted that carbon emission efficiency in the Yellow River Basin has significantly improved but remains unevenly distributed, generally decreasing from east to west. This distribution pattern aligns with the “low emissions in the upper reaches and high emissions in the middle and lower reaches” trend identified in this study, further validating its conclusions.

Second, significant spatial disparities and clustering characteristics in carbon emissions were identified. Tian et al. (2022) observed strong spatial associations in land-use carbon emissions in the Yellow River Basin, with urban spatial clustering levels initially rising and then declining. They reported that “Low-Low” clusters remained concentrated in the upper reaches, while “High-High” clusters in the middle and lower reaches exhibited an eastward migration trend. This study also found that “High-High” clustering cities are mainly located in Shandong Province. In contrast, “Low-Low” clustering cities are primarily scattered across Gansu, Shaanxi, and Sichuan Provinces, which aligns with existing findings (Zhang et al., 2023).

Third, the factors driving the growth of urban carbon emissions in the Yellow River Basin are complex and varied. Gong et al. (2023) highlighted the significant positive direct effects of economic development levels and industrial structure upgrades, consistent with this study’s findings. Liu et al. (2023) pointed out that digital economies and foreign investments increased carbon emissions in some cities, similar to this study’s observation that economic development promotes higher emissions. However, this study also identified a weakening effect of industrial structure development on emissions, offering a novel perspective that complements existing research.

While this study provides insights into the spatiotemporal evolution and influencing factors of urban carbon emissions in the Yellow River Basin, it also has three main limitations and future research directions.

First, there are limitations in the research sample. This study focuses on prefecture-level cities, which is a step forward from provincial-level data but still tends to be analyzed from a macro perspective. This approach does not fully capture the nuanced carbon emission trends at a micro level. Future studies could refine the sample by analyzing county-level cities to explore better the spatiotemporal characteristics of urban carbon emissions in the Yellow River Basin.

Second, this study uses the GTWR model to analyze influencing factors. However, from a geographical perspective, the impact of factors on carbon emissions may involve spatial spillover effects beyond simple direct effects. Characteristics such as similar economic development levels or geographic proximity may lead to spillover influences on urban carbon emissions. Future research could employ spatial Durbin models to explore deeper pathways for carbon reduction in the region.

Third, this study relies primarily on regression methods to estimate carbon emissions and uses Moran’s I and the Dagum Gini coefficient to analyze spatiotemporal differences. These methods may limit the depth of analysis on the spatiotemporal characteristics of urban carbon emissions. Future research could adopt related geographical methods, such as spatial association network features or spatial Markov models, to further explore the spatiotemporal evolution of the Yellow River Basin. Such approaches would provide more in-depth insights into the current state of urban carbon emissions and offer practical recommendations for sustainable development.