Multi-Scale Mapping of Energy Consumption Carbon Emission Spatiotemporal Characteristics: A Case Study of the Yangtze River Delta Region

Lv, Kangjuan; Wang, Qiming; Shi, Xunpeng; Huang, Li; Liu, Yatian

doi:10.3390/land14010095

Open AccessArticle

Multi-Scale Mapping of Energy Consumption Carbon Emission Spatiotemporal Characteristics: A Case Study of the Yangtze River Delta Region

by

Kangjuan Lv

¹

,

Qiming Wang

^2,*,

Xunpeng Shi

³

,

Li Huang

¹ and

Yatian Liu

⁴

¹

SHU-UTS SILC Business School, Shanghai University, Shanghai 201900, China

²

School of Economics, Shanghai University, Shanghai 201900, China

³

Australia-China Relations Institute, University of Technology Sydney, Sydney, NSW 2007, Australia

⁴

School of Economics and Management, Beijing Jiaotong University, Beijing 100080, China

^*

Author to whom correspondence should be addressed.

Land 2025, 14(1), 95; https://doi.org/10.3390/land14010095

Submission received: 2 December 2024 / Revised: 26 December 2024 / Accepted: 3 January 2025 / Published: 6 January 2025

(This article belongs to the Topic Advances in Multi-Scale Geographic Environmental Monitoring: Theory, Methodology and Applications)

Download

Browse Figures

Review Reports Versions Notes

Abstract

:

Climate issues significantly impact people’s lives, prompting governments worldwide to implement energy-saving and emission-reducing measures. However, many areas lack carbon emission data at the lower administrative divisions. Additionally, the inconsistency in the standards, scope, and accuracy of carbon dioxide emission statistics across different regions makes mapping carbon dioxide spatial patterns complex. Nighttime light (NTL) data combined with land use data enable the detailed spatial and temporal disaggregation of carbon emission data at a finer administrative level, facilitating scientifically informed policy formulation by the government. Differentiating carbon emission data by sector will help us further identify the carbon emission efficiency in different sectors and help environmental regulators implement the most cost-effective emission-reduction strategy. This study uses integrated remote-sensing data to estimate carbon emissions from fossil fuels (CEFs). Experimental results indicate (1) that the regional CEF can be calculated by combining NTL and Landuse data and has a good fit; (2) the high-intensity CEF area is mainly concentrated in Shanghai and its surrounding areas, showing a concentric circle structure; (3) there are obvious differences in the spatial distribution characteristics of carbon emissions among different departments; (4) hot spot analysis reveals a three-tiered distribution in the Yangtze River Delta, increasing from the west to the east with distinct spatial characteristics.

Keywords:

energy consumption; carbon emissions; nighttime light; spatiotemporal changes; Yangtze River Delta

1. Introduction

Carbon dioxide (CO₂) is one of the main factors contributing to global warming. Many countries have set long-term targets for carbon emissions and are attempting to introduce more clean energy to reduce reliance on carbon [1]. Effective regional governance of carbon reduction necessitates an understanding of the spatial distribution of carbon emissions, taking into account the differences in regional endowments. Therefore, accurately identifying the spatiotemporal distribution characteristics and trends of spatial carbon emissions will provide a practical basis for policy formulation [2]. China formally committed in 2020 to peak its carbon emissions by 2030 and achieve carbon neutrality by 2060 [3]. China is still classified as a developing country, and the government faces dual pressures of economic development and carbon reduction. To minimize the economic impact of decarbonization, significant regional disparities in development must be fully considered [4,5]. Estimating regional carbon emissions is highly significant in this context [6].

Understanding the spatial distribution of carbon dioxide emissions enables decision-makers to identify high-emission areas. Human activities are a major source of carbon dioxide emissions and the primary focus of the carbon emission estimation in this paper. County-level areas, as the primary spatial units for grassroots activities, are often used as the most basic units for economic activities [7]. In China, the administrative hierarchy features several levels, with the county serving as a crucial unit of local governance. Unlike U.S. counties, which mainly function as political subdivisions of a state, Chinese counties are directly involved in local administration. U.S. counties primarily handle local services and enforce state laws, while Chinese counties focus on grassroots governance and services. These differences in structure and function are important to understand when studying carbon emissions and economic development at the county level in China, as they reflect the unique governance system that influences local decision-making and policy implementation. We can achieve a more detailed understanding of emission patterns at the county level, enabling targeted analysis that surpasses broader provincial or municipal assessments. However, obtaining county-level carbon emission data is challenging. Due to the lack of statistical data, county-level statistics suffer from issues such as opaque statistical results, inadequate data reliability, and high levels of missing data [8].

The carbon emission estimates of counties provide a basis for the assessment of local governments, which is conducive to improving the awareness of local governments and residents about carbon emissions and promoting stakeholders to actively participate in environmental protection activities. Currently, carbon emission data for China’s regions are mostly reflected at the provincial level, and there are few carbon emission estimates for prefecture-level cities or counties [9]. Few studies on county carbon emissions currently exist, and they have limitations in methods, data refinement, and regional coverage [10]. In addition, considering that the proportion of carbon emissions in different industries is only available at the national or provincial level, the aggregated data obscure the differences in carbon emission characteristics between industries in different regions, which is more evident in the county. Therefore, it is necessary to distinguish carbon emissions from different sectors when estimating county carbon emissions. Refined carbon emission data enhance our ability to study how microeconomic activities influence carbon emissions, allowing for a more nuanced analysis of the temporal dynamics of emissions in relation to regional economic activities and land-use patterns. More detailed county carbon emission data will enhance policymakers’ understanding of regional emission characteristics, enabling better evaluation of past policies’ impact on various sectors. These data will help identify which governance strategies are most effective and differentiate carbon emission efficiency and reduction costs across industries. By considering emissions-reduction costs alongside welfare effects, policymakers can better balance environmental goals with economic growth, allowing for tailored policies in different regions that achieve maximum emission reductions at minimal cost. The compilation and analysis of carbon emission data at the county level facilitate a more precise identification of local emission sources and enable the formulation of effective strategies that align with national emission reduction targets. For enterprises, precise carbon emission data enable the identification of long-term economic impacts from investments in innovation and low-carbon technologies. By leveraging existing measures, companies can mitigate potential future costs associated with environmental regulations. A long-term analysis of sector-specific carbon emissions will allow businesses to accurately determine these costs, highlighting the practical significance of this article.

The advancement of satellite remote-sensing technology has enabled studies to estimate economic development levels [11,12], population density [13,14], and population distribution using satellite data [15]. More and more scholars are paying attention to how to use satellite data to detect carbon dioxide emissions [16]. Existing studies rarely consider the differences in the correlation between carbon emissions from different sectors and night light data when using night light data. Instead, they estimate the total regional carbon emissions directly through night light data. Obviously, due to the resolution and saturation problems of night lights, the simulation accuracy of carbon emissions still needs to be improved [17]. In this regard, by combining land use type data, not only can the carbon emissions of different sectors be estimated, but the heterogeneity of carbon emissions and night lights in different sectors is also taken into account. The spatiotemporal heterogeneity is also incorporated into the model through the Geographically and Temporally Weighted Regression model (GTWR) to achieve the purpose of a more accurate simulation of regional carbon emissions [18].

We propose a multi-scale carbon estimation strategy based on matched data from land use types and nighttime light (NTL-Landuse). The objectives of this study are as follows: (1) relying on the constructed NTL-Landuse, a multi-scale county CO₂ estimation strategy is developed to measure county CO₂ emissions under different sectors; (2) calculating the value of carbon dioxide emissions in the Yangtze River Delta; (3) introducing spatial statistical methods to examine the traits of various sectors and the overall spatiotemporal patterns of carbon emissions. (4) Evaluating carbon emission trends and formulating targeted emission reduction policies.

The main contributions of this study are as follows: (1) We combine nighttime light (NTL) data with land use information to simulate carbon emissions, improving accuracy over single-method approaches. This method addresses data limitations, offering more precise regional estimates. (2) In contrast to previous studies on total CO₂ emissions, we assess carbon emissions by industry sector using the NTL-Landuse framework. This multi-scale approach offers insights into sector-specific emission patterns, allowing for more targeted policy interventions. (3) This study tests the model in the Yangtze River Delta, examining carbon emissions trends and suggesting region-specific reduction policies.

2. Literature Review

Many scholars are now concentrating on carbon emissions accounting [19]. Due to different research objectives and purposes, the methods for accounting for carbon emissions also vary [20,21]. According to the approach of carbon accounting, it can generally be divided bottom-up and top-down [22]. The internationally accepted method is the Emission-Factor Approach proposed by the International Panel on Climate Change (IPCC) [23]. Although the IPCC method was initially developed for national-level carbon emission estimation, its theoretical framework is often used for provincial-level administrative divisions and urban [24]. The IPCC method has also been used for carbon emission estimation in various specific sectors such as the steel industry, timber industry, construction, and transportation [25,26,27]. These studies on carbon emission estimation provide strong practical evidence for regions to achieve policy goals of carbon neutrality and peak carbon emissions. However, the IPCC method relies on the energy balance sheet (EBT), and China only discloses these at the provincial level. Another standard bottom-up method is the carbon mass balance approach, which calculates emissions by tracking elemental inputs and outputs within an organization or system [28]. However, such methods’ limitations lie in the extensive manpower, material, and financial resources required for data collection, making it suitable only for estimating carbon emissions in a specific process or system rather than for regional carbon emissions [8].

There are currently two main methods for more refined spatial carbon emission estimation. The first method is based on existing energy balance sheets. For instance, Chen (2021) estimated industrial carbon emissions in Guangdong Province by compiling energy consumption and cement production data from various prefecture-level cities, depicting the spatiotemporal pattern of industrial carbon emissions in Guangdong Province from 2005 to 2015 [29]. Dong (2018) used input–output tables and IPCC methods to calculate carbon dioxide emissions in four direct-controlled municipalities, namely, Beijing, Tianjin, Shanghai, and Chongqing. Additionally, this study identified the urbanization rate as the primary driver of increased urban carbon dioxide emissions [30]. These studies rely on the collection of field survey data, which are not updated frequently. For example, China’s input–output tables are compiled every five years. The second issue is the inevitable presence of various noises and biases in survey data, affecting the research and decision-making based on these data [31]. The third issue is that the field survey method is too costly and lacks sustainability. The second method is to use remote-sensing data for estimation. With the advancement of remote-sensing technology, estimating economic activities through satellite data has emerged as a new method [32]. Among these, NTL, highly correlated with human activities and is often used by scholars as a proxy variable to investigate human activities [33]. Elvidge (1997) laid the groundwork for understanding the correlation between NTL and carbon emissions. Their work demonstrated the potential of using NTL as a proxy variable to measure carbon emissions [34]. Chen (2020) estimated carbon emissions data for Chinese countries from 2000 to 2017 using NTL [10]. Wang (2023) combined multi-source remote-sensing data to estimate carbon emissions at the grid scale in China from 2010 to 2018 and explored potential driving factors for carbon emissions, using Hunan Province as an example [35]. In more detailed research, Zheng (2024) estimated carbon emissions patterns at the “province-city-county-township” four-level scale in Fujian Province using NTL [36]. Zhang (2024) studied carbon emissions at the street level in Xi’an using NTL [37]. Wu (2025) estimated energy-related carbon emissions in the northeast by developing a model linking NTL to emissions. He also applied the Tapio decoupling model to examine the relationship between economic development and carbon emissions, concluding that both follow a three-stage decoupling pattern, with an overall state of decoupling marked by a growth linkage [38]. Lu (2024) utilized high-resolution NTL data obtained from the domestic satellite Luojia 1–01 to estimate electricity consumption in Shenzhen [39].

To further enhance the accuracy of the estimation results, existing studies have made substantial efforts. Some studies have considered incorporating additional data to construct more refined carbon emission estimation models. Meng (2017) further improved the estimation accuracy (R2 = 0.8796) by introducing data such as population density and combining it with NTL to estimate carbon emissions [40]. Wang (2023) combined NTL and XCO₂ concentration data to develop a carbon emission and energy consumption estimation model, achieving spatially refined measurements of energy consumption carbon emissions [35]. In terms of model selection, existing studies have identified a strong linear correlation between carbon emissions from human energy consumption, which is why linear regression analysis is often used with regional carbon emission statistics and nighttime light data [41]. Considering that carbon emissions from fossil fuels (CEFs) between cities are not isolated, with one region’s emissions being influenced by surrounding cities, a Spatial Dubin Model (SDM) is employed to address the spatial dependence issue in NTL based CEF estimation, and the use of a dynamic SDM model addresses endogeneity problems [42]. Considering the spatial heterogeneity, in addition to the application of the SDM model, some studies have incorporated the Geographically Weighted Regression (GWR) model into this estimation framework, aiming to improve the accuracy of the estimates [43].

Most related research primarily focuses on total carbon dioxide emissions, with little distinction between emissions from different sectors. Shi (2020) has started investigating the relationship between NTL and carbon emissions across various sectors. The study findings indicate that NTL can provide more accurate carbon emission assessments in urban areas with large populations and relatively developed social and economic conditions and that the precision of estimating urban carbon emissions through NTL is higher than that of estimating industrial carbon emissions [44]. Point of Interest (POI) data, as a form of multi-source geographic spatial big data, can be combined with NTL to obtain carbon emission estimates for specific sectors [45]. Wei (2024) effectively measured industrial carbon emissions in the Yellow River Basin by combining NTL and land use data, and further classified industrial carbon emissions using POI data. They categorized industrial carbon emissions into eight sectors and analyzed them individually [46]. Apart from POI, Landuse data are also commonly used to represent regional carbon emissions. Liu (2024) studied county-level Landuse carbon emissions (LUCEs) using changes in China’s land use data [47]. Since directly using NTL data results in carbon emission spatialization with high-value areas overly concentrated, making it difficult to discern the internal spatial heterogeneity, combining Landuse data helps to accurately depict carbon emissions [48]. Wei (2021) estimated carbon emissions for various provinces in China by differentiating NTL data under different land use types, further refining the categories into urban, rural, and industrial sectors [49].

In the field of using NTL to measure carbon emissions, existing studies have established a relatively comprehensive research framework. Although these studies have made significant progress, there are still certain limitations. First, current research mostly focuses on depicting the total carbon emissions, and NTL alone cannot differentiate between different categories or industries of carbon emissions. Secondly, aggregating NTL data to simulate carbon emissions masks the spatial heterogeneity of emissions, resulting in overly simplified carbon emission views based on average estimates and failing to capture more effective and refined differences.

The connection and distinction between this study and existing research lie in the fact that this study draws on existing mature approaches, such as the processing of nighttime light data and the calculation of carbon emissions. At the same time, this study attempts to address the issue of sector-specific carbon emission characterization by incorporating Landuse data and solving spatial heterogeneity through the application of the GTWR model. Building on existing research, we believe it is essential to consider both sectoral and spatial heterogeneity when estimating CEF using NTL. Therefore, we propose the research hypotheses of this study. Integrating nighttime light data with land use data will yield more accurate carbon emission estimates by accounting for spatial and sectoral differences.

3. Study Area and Research Methods

3.1. Study Area

The Yangtze River Delta lies at the intersection of China’s eastern coast and the Yangtze River Basin (see Figure 1). It is characterized by flat terrain, predominantly plains, with higher elevations and a combination of mountainous and hilly terrain in the southern region. The climate is hot and rainy in summer and mild and humid in winter, with an annual precipitation of over 800 mm. The area has diverse land use types, high urbanization rates, and a dense population.

According to data released by the Yangtze River Delta and Yangtze River Economic Belt Research Center, in 2023, the GDP of the three provinces and one city will exceed CNY 30 trillion, a year-on-year increase of 5.7%, accounting for 24.4% of the national total. It is one of China’s largest and most robust industrial bases, concentrating on highly competitive manufacturing industry chains and modern service sectors. It is the hub for modern services and advanced manufacturing industries in China, with core industries including electronics, automobiles, and modern finance. The Yangtze River Delta region is a crucial growth hub in China’s economic development and a key area for low-carbon transformation. Thus, choosing this region as the study area for this paper is both representative and significant.

3.2. Data Sources

We primarily utilized nocturnal light remote-sensing data, land use type data, and energy consumption data. The nocturnal light dataset was sourced from NOAA (https://earthdata.nasa.gov, accessed on 4 January 2025), predominantly employing Defense Meteorological Satellite Program (DMSP-OLS) and Visible infrared Imaging Radiometer (NPP-VIIRS) nocturnal light data. However, due to disparities in spectral resolution, spatial resolution, radiometric resolution, and product update frequency, the two sets of data are incompatible. Therefore, following Wu’s (2021) approach, we integrated and calibrated DMSP-OLS annual nocturnal light data and NPP-VIIRS monthly nocturnal light data using a convolutional neural network autoencoder model [50]. This integration yielded DMSP-OLS-like extended time series data, which underwent validation to ensure reasonable accuracy and spatiotemporal consistency. The land use type dataset was obtained from the Chinese Academy of Sciences Resource and Environmental Science Data Center (http://www.resdc.cn, accessed on 10 October 2024), primarily utilizing Landsat remote-sensing image data and employing manual visual interpretation to construct the dataset. We integrated land use types into five categories: agricultural land, urban land, rural land, built-up land, and undeveloped land. Energy consumption data were obtained from the China Energy Statistical Yearbook [51]. It was further categorized into agricultural consumption, transportation consumption, wholesale and retail trade consumption, town resident consumption, rural areas resident consumption, industry consumption, and building consumption.

3.3. Research Framework and Research Methods

3.3.1. Research Framework

In this study, we aimed to analyze the spatiotemporal variations of CEF in the Yangtze River Delta region. To achieve this goal, we can be broken down into four main phases: (1) Integration of Nighttime Light Data with Land Use Types. The first step involves integrating NTL with Landuse data for various regions within the Yangtze River Delta. This integration allowed us to correlate the intensity and distribution of lighting with specific land use categories, providing a spatially detailed basis for estimating CEF at a regional level. (2) Calculation of Total and Sectoral CEF Emissions. We calculated the total emissions of CEF across mainland China, as well as the sector-specific emissions, using official statistical data. These emissions were categorized into various sectors, such as industry, transportation, and residential, based on their respective energy consumption patterns and carbon intensity. This step provided a foundational dataset for the analysis of CEF patterns across different spatial units. (3) Selection of Relevant NTL-Landuse Data. The third phase involved selecting appropriate NTL-Landuse data by utilizing sector-specific CEF data as dependent variables. We performed a Pearson correlation analysis to identify the most relevant NTL-Landuse data that were strongly associated with CEF emissions. Once the suitable data were selected, we employed three different regression models—OLS, FE, and GTWR—to fit quantitative relationships between NTL-Landuse data and CEF emissions. These models were assessed using accuracy indicators, such as R-squared and Mean Squared Error (MSE), to determine the most appropriate model for estimation. (4) Spatiotemporal Analysis of CEF Dynamics. Finally, we integrated the estimation results obtained from the best-performing model and applied several analytical techniques to explore the spatiotemporal dynamics of CEF. This included trend analysis to identify temporal changes in CEF, hotspot analysis to detect areas with significant emission concentrations, and the use of standard deviation ellipses to examine the spatial distribution and shifts of emissions over time. These methods enabled us to visualize and interpret the regional variation and trends in CEF across the study period. The detailed process is illustrated in Figure 2.

3.3.2. Correction Processing of Nighttime Light Data

Nighttime light data have been widely applied in various studies. To obtain more valuable long-term time series data, existing research often uses a fusion of DMSP-OLS and NPP-VIIRS datasets. DMSP-OLS suffers from sensor aging issues, resulting in systematic biases and light saturation [52]. These problems are not present in NPP-VIIRS; however, NPP-VIIRS itself has issues with outliers, such as unfiltered transient light sources (e.g., lasers, ships, aircraft) [53]. These distortions affect the accuracy of carbon emission estimation, so it is necessary to calibrate both datasets before their fusion. Additionally, one of the datasets provides annual data, while the other provides monthly data, and their resolutions are inconsistent. Therefore, mutual calibration and fusion of the two datasets are also required to ensure data consistency and accuracy [54].

Following the common processing methods in the existing literature, we collated and integrated the two sets of data. Firstly, NOAA’s F162006 DMSP-OLS data were selected as reference data, using Hegang as the pseudo-invariant calibration region with a quadratic calibration model, following Elvidge’s approach (1997) [34]. Secondly, NPP-VIIRS data noise removal is essential due to the lack of filtering in the nighttime light data. Anomalous values in the original images with Digital Number (DN) values less than 0 were reassigned as 0. To account for high DN values likely representing noise (e.g., from airports or fires), Beijing, Shanghai, and Guangzhou were used as reference points, with peak DN values in other regions theoretically not exceeding these cities [55]. By optimizing the monthly data, we further obtain annual data and perform annual calibration based on the yearly data to ensure that there are no abnormal DN values. Thirdly, all spatial data were referenced to the WGS-84 datum, and the spatial resolution was resampled to 1 km [56]. Finally, logarithmic transformation of the NPP-VIIRS data reduced variance in radiance values, and, by selecting 2013 as the common coverage year, a sigmoid model was applied to convert the NPP-VIIRS data into a curve based on DMSP-OLS and the logarithmically transformed NPP-VIIRS data [57].

3.3.3. Method for Quantifying the Statistical CO₂ Emissions

This study adopts the carbon dioxide emission measurement model proposed by IPCC, which calculates carbon emissions by multiplying the combustion quantity (energy) of each energy source by its corresponding default emission factor and then summing them up to obtain the total carbon emissions [58]. The formula is as follows:

{C E F}_{i j} = C \times ({A D}_{i j} \times {N C V}_{i} \times {C C}_{i} \times O_{i j})

(1)

where i represents the type of fuel and j represents the sector. CEF denotes the carbon dioxide emissions produced from fossil fuel combustion. AD represents the physical consumption of fossil energy. NCV denotes the net calorific value, which refers to the heat released when a unit mass of sample burns in excess oxygen under constant volume conditions, with the combustion products composed of oxygen, nitrogen, CO₂, sulfur dioxide, gaseous water, and solid ash. CC denotes the carbon content per unit of heating value. O indicates the percentage of carbon in the energy source converted to CO₂. C represents the conversion factor, which is 44/12, converting the mass of carbon atoms to the mass of carbon dioxide molecules. According to the IPCC method, the emission factor (EF) can be expressed as (NCV × CC × O × 44/12). The emission factor represents the amount of CO₂ emitted per unit of activity or energy consumed.

The China Energy Statistical Yearbook provides energy balance sheets by province, listing 27 types of fossil energy consumption across various industrial sectors. Considering the discrepancies between the default net calorific values in the IPCC and actual investigation results in China, the average net calorific value coefficient used in this study is sourced from the “General Principles for Comprehensive Energy Consumption Calculation” [59].

Building upon this foundation, we simulated carbon emissions. Currently, abundant literature has evidenced a significant correlation between CEF and NTL [60]. Hence, we constructed the following linear regression model:

{C E F}_{i}^{’} = a {N T L}_{i} + b

(2)

where

{C E F}_{i}^{’}

represents the carbon emissions in the i year, a is the estimated parameter, and b is the constant term. Although there have been continuous improvements in estimating carbon emissions using NTL data, uncertainties still exist [61]. Firstly, the presence of light oversaturation and halo effects reduces the authenticity of NTL and weakens its correlation with economic activities [62]. Secondly, carbon emission estimates rely solely on a simple linear relationship with NTL [63], assuming that the correlation between NTL and carbon emissions is consistent across different spatial contexts without considering their spatial and temporal relationships, leading to estimation biases [64]. So, considering potential individual, time, and spatial effects, we further incorporated FE regression estimation and GWTR regression estimation. The optimal model was determined based on precision indicators such as MSE and R2.

In earlier related studies, most researchers adopted basic linear regression models such as OLS. While OLS has the advantage of being simple and straightforward, it is unable to account for individual-specific effects, leading to biases in carbon emissions modeling across individuals and time [65]. Panel estimation models have been widely applied to carbon emission estimation; however, the estimation strategies of panel fixed-effects regression also have certain limitations. For instance, these models typically assume that coefficients remain constant over space and time, which fails to account for localized effects [66]. To address this issue, we further introduce the GTWR model to overcome such limitations. First, GTWR allows the relationship between dependent and independent variables to vary across both space and time. This is particularly important in carbon emission estimation, as the complexity and diversity of carbon emissions result in differences across regions and over time. Second, GTWR permits local estimation of coefficients, which is especially useful in regional studies. Regional factors such as economic structures and policy environments vary across locations, and a single coefficient is insufficient to capture the nuanced spatiotemporal variations in carbon emissions. GTWR, by capturing local effects and spatiotemporal heterogeneity, typically achieves better model fit and higher explanatory power compared to panel fixed-effects models.

The specific regression model for GTWR is as follows:

{C E F}_{i k} = β_{0} (u_{i}, v_{i}, t_{i}) + β_{k} (u_{i}, v_{i}, t_{i}) {N T L}_{i k} + ε_{i}

(3)

where

{C E F}_{i k}

represents the value of the carbon emission for the k sector.

(u_{i}, v_{i})

represents the longitude and latitude coordinates.

t_{i}

represents time.

ε_{i}

is the model error term.

β_{k} (u_{i}, v_{i}, t_{i})

is the regression coefficient corresponding to the nighttime light data for the k sector at the i region.

3.3.4. Kernel Density Estimation

Kernel density estimation helps us generate continuous surface density maps, allowing for the intuitive identification of high-density regions of CEF. Kernel density estimation estimates a population distribution from a sample, generating a probability density function to analyze data properties such as clustering and dispersion. The formula is as follows:

f (x) = \frac{1}{n h} \sum_{i = 1}^{n} K [\frac{X_{i} - x}{h}]

(4)

\lim_{N \to \infty} h (n) = 0

(5)

\lim_{n \to \infty} n h (N) = N \to \infty

(6)

where n represents the sample size, K denotes the kernel density function, and h is the bandwidth. In urban and regional analysis, kernel functions compute values per unit area from point or line features, fitting each to a smooth, cone-shaped surface to visualize feature distribution. For point features, kernel density analysis calculates the density of each output raster cell around points, with the surface value peaking at each point and tapering to zero at the search radius. Each cell’s density is the sum of all kernel surfaces stacked at its center [67].

3.3.5. Directional Analysis

Directional analysis can quantitatively explain the overall characteristics of CEF in space, such as centrality, distribution, directionality, and spatial form. Common methods for spatial directional analysis include the mean center, standard deviation ellipse (SDE), and standard distance analysis. The SDE method can generate an elliptical space and describe the spatial distribution characteristics through the basic parameters such as the range, center, major axis, minor axis, and azimuth angle of the generated ellipse. Calculating the SDE first requires determining the mean center of the features.

\bar{X} = \frac{\sum_{i = 1}^{n} x_{i}}{n}, \bar{Y} = \frac{\sum_{i = 1}^{n} y_{i}}{n}

(7)

\bar{X_{w}} = \frac{\sum_{i = 1}^{n} w_{i} x_{i}}{\sum_{i = 1}^{n} w_{i}}, \bar{Y_{w}} = \frac{\sum_{i = 1}^{n} w_{i} y_{i}}{\sum_{i = 1}^{n} w_{i}}

(8)

Obtain the calculated average center as the center of the ellipse. The standard deviations of each element’s X and Y coordinates are calculated to define the ellipse’s major and minor axes and determine its direction. The formula is as follows:

{S D E}_{x} = \sqrt{\frac{\sum_{i = 1}^{n} {(x_{i} - \bar{X})}^{2}}{n}}, {S D E}_{y} = \sqrt{\frac{\sum_{i = 1}^{n} {(y_{i} - \bar{Y})}^{2}}{n}}

(9)

where

{S D E}_{x}

and

{S D E}_{y}

represent the coordinates of the center of the ellipse. To find the angle between the X-axis and true north (0 degrees), calculate accordingly.

A = (\sum_{i = 1}^{n} {\tilde{x}}_{i}^{2} - \sum_{i = 1}^{n} {\tilde{y}}_{i}^{2})

(10)

B = \sqrt{{(\sum_{i = 1}^{n} {\tilde{x}}_{i}^{2} - \sum_{i = 1}^{n} {\tilde{y}}_{i}^{2})}^{2} + 4 {(\sum_{i = 1}^{n} {\tilde{x}}_{i} {\tilde{y}}_{i})}^{2}}

(11)

C = 2 \sum_{i = 1}^{n} {\tilde{x}}_{i} {\tilde{y}}_{i}

(12)

t a n θ = \frac{A + B}{C}

(13)

{\tilde{x}}_{i}

and

{\tilde{y}}_{i}

represent the difference between the i-th element and the mean center. θ is the angle. Finally, calculate the lengths:

σ_{x} = \sqrt{2} \sqrt{\frac{\sum_{i = 1}^{n} {({\tilde{x}}_{i} c o s θ - {\tilde{y}}_{i} s i n θ)}^{2}}{n}}

(14)

σ_{y} = \sqrt{2} \sqrt{\frac{\sum_{i = 1}^{n} {({\tilde{x}}_{i} c o s θ + {\tilde{y}}_{i} s i n θ)}^{2}}{n}}

(15)

The mean center of the ellipse represents the centroid of the spatial distribution of CEF, while the azimuth angle indicates its primary direction. The major axis highlights areas with a higher concentration of CEF, while the minor axis represents more dispersed areas. If both axes are nearly equal in length, it suggests that the CEF distribution is uniform and does not exhibit clear directional characteristics.

4. Results and Analysis

4.1. Correlation Analysis of CEF and NTL-Landuse

The vast majority of nighttime light data recorded by satellites from space are generated by human activities and have been proven to be related to economic activities [68,69]. These human economic activities are significant sources of CEF, resulting in a high correlation between NTL and CEF [43]. Based on data availability, this study categorizes energy consumption-related carbon emissions into various sectors: agricultural consumption, transportation consumption, wholesale and retail trade consumption, urban residential consumption, rural residential consumption, industrial consumption, and building consumption. Considering the distinct characteristics of each sector, some sectors have minimal nighttime activities and, thus, cannot be directly captured through NTL data. These are classified as indirect sectors (e.g., agriculture, transportation, wholesale and retail trade). In contrast, sectors with substantial nighttime activities, which can be directly captured by NTL data, are classified as direct sectors (e.g., towns, rural areas, industry, and building).

To construct the estimation model more effectively, it is essential to match CEF from different industry sectors with NTL under various land use types. Apart from the insignificant correlation between NTL corresponding to undeveloped land and CEF emissions from various sectors, significant positive correlations are observed between NTL and CEF under other land use types. However, there are significant differences in the correlation coefficients between NTL and CEF across different land use types (see Table 1).

We need to compare the correlation coefficients between the NTL-Landuse and NTL-Total datasets and the carbon emissions across various sectors to select the appropriate independent variables. From a theoretical perspective, sectoral carbon emissions are closely related to land-use patterns (e.g., agricultural land tends to have more agricultural carbon emissions rather than industrial or construction-related emissions). The correlation between the matched NTL and CEF is shown in Table 2. By comparing with the correlation results of NTL-Total, it can be observed that, except for the agricultural sector, NTL-Landuse has higher correlation coefficients with CEF in all sectors. Although the difference in correlation coefficients between NTL-Landuse and NTL-Total for CEF is small, we still have reasons to believe that using NTL-Landuse can improve the accuracy of CEF estimation. Therefore, using NTL-Landuse as an independent variable is appropriate. Secondly, comparing the correlation coefficients of different sectors with NTL-Landuse data reveals that direct sectors exhibit coefficients ranging between 0.8 and 1, indicating a high linear correlation. This suggests that NTL-Landuse effectively captures the CEF of these sectors. In contrast, indirect sectors have correlation coefficients between 0.6 and 0.8, lower than those of direct sectors but still within a range of significant linear correlation. The existence of differences in correlation coefficients between NTL-Landuse and NTL-Total for CEF also indicates that the aggregated NTL neglects the sectoral heterogeneity when estimating carbon emissions. Therefore, incorporating NTL-Landuse will help identify this sectoral heterogeneity. This is especially important in regions where land use types have distinct structures, as the use of NTL-Landuse will be crucial for identifying CEF. Additionally, relying solely on NTL-Total cannot accurately assess the spatial distribution of sectoral CEF, as both are estimated based on the same nightlight distribution, which can introduce biases when mapping the spatial distribution of sectoral CEF. In such cases, even a slight improvement in accuracy makes the choice of NTL-Landuse for CEF estimation significant for enhancing the precision of identifying emission spatial distribution.

In addition, this study incorporates the Spatial Autoregressive Model (SAR) and Spatial Durbin Model (SDM) to examine whether these models can provide more accurate estimations. According to the results of the Akaike Information Criterion (AIC), the AIC value of the GTWR model is significantly smaller than those of the SAR and SDM models, indicating that the GTWR model is the more appropriate choice. Furthermore, the calculation of precision metrics also shows that the GTWR model achieves higher accuracy than both the SAR (Root Mean Squared Error (RMSE) = 0.6301; Mean Absolute Error (MAE) = 0.2602) and SDM (RMSE = 0.6505; MAE = 0.2825) models. This further demonstrates the rationality of using the GTWR model for carbon emission estimation.

4.2. CEF Calibration and Validation

We chose three models: OLS, FE, and GTWR. The regression parameters were all logarithmically transformed and considered time effects. In Table 3, by computing the RMSE, R2, and MAE for 30 provinces across the three models, using NTL-Total as the independent variable, the R2 values for OLS, FE, and GTWR were found to be 0.8408, 0.12575, and 0.93133. The corresponding RMSE values were 0.33204, 0.77811, and 0.21807. It can be concluded that the GTWR model outperforms OLS and FE in terms of fitting effect. Furthermore, by comparing the MAE, it is evident that the GTWR model has the smallest value at 0.17324, indicating that the GTWR model estimation results in smaller errors compared to OLS and FE. Similar results were observed when using NTL-Landuse as the independent variable. Additionally, through a comparison between the results obtained using NTL-Total and NTL-Landuse, it was found that both cases effectively reflected CEF. Therefore, it is deemed reasonable to adopt NTL-Landuse as the independent variable and GTWR as the estimation model.

4.3. Mapping CEF

Using NTL-Landuse as the independent variable input into the GTWR model yielded the estimation results of CEF. To observe the spatial heterogeneity of carbon emissions at the county level, we visualized the calculated CEF using ArcGIS (see Figure 3). The resulting map provides a spatial distribution of CEF at the county level across the study region. This map does not distinguish between emissions from specific industries within these counties but rather illustrates the overall carbon emissions at the county scale. The spatial distribution of CEF shows distinct patterns. Low CEF values are primarily observed in the northeast, northwest, and most rural areas of the central region. In contrast, high CEF values are predominantly concentrated in developed urban areas, where industrial activities, dense populations, and vigorous economic activities contribute to high energy consumption and CO₂ emissions. Notably, the Yangtze River Delta region stands out for its high CEF, which results from the concentration of energy-intensive industries, a large population, and a dynamic economy. This is also the reason why we chose the Yangtze River Delta as our research area. By comparing the spatial distribution of CEF in the Yangtze River Delta region in different periods, it can be found that carbon emissions show a clear upward trend. Specifically, CEF in Zhejiang Province increased by 281.19%, in Shanghai Municipality by 83.38%, in Jiangsu Province by 327.93%, and in Anhui Province by 228.85%.

4.4. Temporal Dynamics Characteristics of CEF

The variation of CEF during the period 2000–2020 is depicted in Figure 4. Overall, over the decade from 2005 to 2015, the CEF of the study area exhibited a rapid upward trend, followed by a subsequent decline in slope, with Shanghai being the first to show a downward trend. When examining the regions individually, apart from Shanghai, the other three provinces all demonstrated a continuous upward trend.

Jiangsu Province experienced the fastest growth in CEF, with a slope of 26.24. In 2007, industrial energy consumption was the primary source of its CEF, accounting for a remarkable 83.18%, the highest among all regions in the study area. By 2020, this proportion had decreased to 75.22%, yet it remained the highest level among the four provinces and municipalities examined. Although the growth of CEF in Jiangsu Province has slowed since the 13th Five-Year Plan period, a clear turning point in emissions has yet to be observed. This is largely due to surrounding areas transferring outdated, energy-intensive, and polluting industries to central and northern Jiangsu. At present, the northern Jiangsu region has formed a structural feature dominated by heavy industries such as chemical industry, coal, and machinery. The industrial sector consumes the most energy among the three major industries, which contributes to the ongoing increase in CEF emissions in the region.

Following Jiangsu Province, Zhejiang Province experienced the next highest growth rate, with a slope of 14.64. Due to the relative scarcity of oil, coal, and electricity resources in Zhejiang Province, it relies heavily on external energy consumption. With the acceleration of urbanization, total energy consumption in Zhejiang Province has continued to increase, with coal consumption reaching 12,758.22 million tons in 2020, accounting for 39% of total primary energy consumption. Although there was a significant decline in Zhejiang Province’s CEF in 2016, it resumed growth thereafter, with the growth rate showing an increasing trend year by year.

Anhui Province, another major industrial region, showed a slope of 11.12, with industrial CEF averaging 76.52%. Although there was a certain downward trend in industrial energy consumption carbon emissions during the study period, the decline was relatively weak. Energy consumption carbon emissions from transportation, urban residents, and rural residents have significantly increased, as the growth rate of energy consumption linked to economic activities in the region has outpaced that of industrial energy consumption. During the 13th Five-Year Plan period, the Anhui provincial government introduced a series of policies. However, this has not effectively curbed the overall growth of CEF, as evidenced by the results.

Shanghai Municipality, the most economically developed region in China, also experienced a rapid rise in CEF in the early stages. After 2014, industrial restructuring in Shanghai caused many industrial entities to relocate to nearby cities, resulting in a decrease in industrial CEF, which fell to 53.45% by 2020. This represented the peak of overall CEF since 2013, after which it began to decline. Meanwhile, as urbanization continued to expand, the proportion of CEF from urban residents and transportation increased. Conversely, the proportion of CEF from rural residents declined. Agriculture in Shanghai relies mainly on imports, with minimal local production, thus exerting a relatively minor impact on overall CEF, accounting for only 0.62% in 2020.

4.5. Spatial Change Characteristics of CEF

From 2000 to 2020, the CEF in the study area initially exhibited a bimodal distribution (See Appendix A). Apart from the low peak, a small high peak area formed in Shanghai and its surrounding areas. As the scale of CEF emissions continues to expand across regions, the bimodal distribution gradually shifts to a unimodal distribution. The kernel density curve shows a rightward shift, and, although it still exhibits a right-skewed distribution, the skewness value has decreased year by year, indicating a consistent increase in CEF. Meanwhile, the peak of the CEF kernel density has shown a declining trend, suggesting that the data distribution is becoming more dispersed. Specific to the performance in space, Figure 5 clearly shows that Shanghai is developing into a high CEF emission center, which gradually expands to surrounding areas, resulting in a higher proportion of high CEF regions.

In 2000, the overall CEF scale in the study area remained relatively low. However, Shanghai and its surrounding areas exhibited significantly higher CEF emissions levels than other regions during the same period, forming the embryonic stage of future carbon emission centers. As the level of economic development continued to rise, the regional industrialization level improved. By 2010, Shanghai had already demonstrated distinct features of high-density CEF centers, spreading to surrounding areas and displaying characteristics of a central pole. In 2015, CEF in Shanghai peaked and began to decline, with the regional diffusion effect during this phase being significantly greater than the agglomeration effect. The high-density CEF range gradually spread to counties and districts corresponding to Suzhou, Wuxi, Jiaxing, Hangzhou, and other areas. At the same time, scattered small, high-density CEF centers began to emerge in cities such as Nanjing, Hefei, Yangzhou, and Ningbo, forming the basic spatial structure of the carbon emission network. By 2020, the high-density carbon emission areas centered around Shanghai continued to increase and connect with secondary central regions. During this phase, the CEF in Shanghai and Suzhou decreased due to the optimization of their industrial structures and the relocation of inefficient capacity.

The spatial characteristics of CEF emissions across different sectors in the Yangtze River Delta Region generally exhibit similar distribution patterns. Still, there are subtle differences in high-density areas (see Figure 6). Industrial CEF emissions are primarily concentrated in the Minhang District, Jiading District, and Pudong New Area of Shanghai, as well as in surrounding areas such as Suzhou, Hangzhou, and Ningbo, showing distinct spatial agglomeration features. The spatial distribution of transportation CEF emissions is relatively dispersed, mainly forming small high-density areas around important transportation hubs such as Hefei, Nanjing, Yangzhou, and Wuxi. Wholesale and retail trade, construction, and urban resident CEF emissions exhibit similar distributions, mainly concentrated in several large cities and their surrounding areas within the region, without forming obvious spatial agglomerations. Rural resident CEF emissions form a high-density area from Wuxi to Shanghai to Hangzhou, with small high-density areas also present in Hefei and Lianyungang. Agricultural CEF emissions are primarily concentrated in southern Jiangsu Province, where the high proportion of plains and abundant water resources create contiguous high-density areas. In 2020, Jiangsu’s grain production reached 745.8 billion tons, 44.64% of the Yangtze River Delta’s total. Agricultural modernization is advanced, with crop cultivation mechanization at 93% and overall agricultural mechanization at 88%, leading the nation.

4.6. Spatial Correlation Analysis of CEF

Spatial correlation is primarily assessed using the Moran’s I index. When Moran’s I is greater than 0, it indicates positive spatial autocorrelation within the region; when Moran’s I is less than 0, it indicates negative spatial autocorrelation within the region; when Moran’s I equals 0, it suggests the data exhibit random spatial distribution characteristics within the region. The overall Moran’s I index in the study area can be calculated. In Table 4, the overall Moran’s I for CEF increased from 0.315 in 2000 to 0.38 in 2020, indicating a continuous strengthening of the spatial clustering of CEF.

Looking at different sectors, except for agriculture, which showed a significant decrease in global Moran’s I, the other sectors all exhibited a clear upward trend. This is mainly due to the increasing urbanization rate, where the population has transitioned from scattered distribution to concentrated distribution, leading to a concentration of economic activities. Agriculture, constrained by land resources, cannot change with population migration, resulting in a decrease in the clustering degree of agricultural CEF.

To further understand the local differences in CEF, hotspot analysis was conducted on the dataset. Hotspot analysis is a form of local autocorrelation analysis. It calculates a statistic for each feature in the dataset and provides an evaluation of whether its surrounding environment forms a hotspot or a cold spot. A hotspot signifies high values in both the area and its surroundings, while a cold spot indicates low values in both. In general, positive values of GiZScore indicate clusters of high values (hot spots), while negative values indicate clusters of low values (cold spots). CEF hotspots are primarily found in the central and eastern regions, whereas cold spots are mainly in the western areas, with an intermediate buffer zone in between (see Figure 7). After experiencing regional expansion from 2000 to 2015, hotspot areas showed signs of contraction from 2015 to 2020. The number of cold spot areas remained relatively stable at around 35 counties (districts).

In 2000, 23 counties (districts) were identified as hotspots at a 99% confidence level, covering most of Shanghai, Suzhou, Wuxi, and areas like Haiyan County in Jiaxing, and Yuyao and Jiangbei in Ningbo. Only Huangshan District in Huangshan City was identified as a cold spot at the same confidence level. By 2020, hotspots at 99% confidence rose to 26 counties, though, in Shanghai, only Qingpu, Fengxian, and Chongming remained. New hotspots included Chongxing County and Wuxing District in Huzhou and Rugao City in Nantong. Cold spots at this confidence level also grew to 12, encompassing areas like Anqing, Huangshan, Xuancheng, and Longquan in Zhejiang, showing a strengthened spatial clustering of CEF.

Further calculations of the standard deviation ellipse parameters for CEF in the study area showed a continuous increase in ellipse area from 2000 to 2020, indicating a gradual expansion of CEF coverage. The “Rotation” angle parameter suggests a distribution pattern of CEF in the “southeast–northwest” direction (see Table 5).

The CEF emissions center in the study area, initially located in Wuxi City, Jiangsu Province (31.2356° N, 119.9208° E), gradually shifted northwestward. By 2020, the CEF emissions center had relocated to Changzhou City, Jiangsu Province (31.36714° N, 119.4639° E). The flattening rate decreased from 42.38% to 41.63%, showing a weakening directional orientation in the southeast–northwest direction.

5. Discussion

5.1. Cross-Validation

While NTL data offer a feasible approach to estimating CEF at the county or sub-regional levels, some errors remain unavoidable, including NTL data, model estimation, and carbon measurement errors, causing deviations from true CEF values. This research does not aim to replace existing carbon estimation methods but to enhance them, focusing on capturing CEF trends and spatial patterns to identify key hotspot and coldspot areas for policymakers. To validate our results, we compare them with established carbon emission databases, specifically the CEADs China City Carbon Emission Database [70], derived from energy balance sheets and industrial energy consumption, and the China County Carbon Emission Database, based on remote-sensing data [10]. Since both databases provide total emissions estimates, this paper also aggregates sector carbon emissions to generate a total and examines the correlation among the various datasets. Ideally, since the errors of each method are independent, they do not have a significant impact on the correlation between the datasets. Therefore, a high correlation should be observed between different datasets.

Table 6 indicates that the correlation between CEF and county CE is 0.7416, city CE_CEADs is 0.7202, and province CE_CEADs is 0.8773. All these correlation coefficients are more significant than 0.7. Therefore, our calculated results are significantly linearly correlated and relatively reliable. Additionally, the CEF estimated for direct departments using NTL data shows a more significant correlation with other datasets than that for indirect departments. This discrepancy highlights the varying performance of NTL in estimating CEF across different sectors, primarily due to the limited nighttime activities of indirect departments, which introduces more noise into the estimates.

5.2. Result Analysis

This study uses nighttime light data to analyze county-level carbon emissions, focusing on the Yangtze River Delta. It measures the spatiotemporal distribution of carbon emissions across counties in the region.

Figure 8 shows the overall characteristics of the area, except for negative growth in CEF emissions in 2016, the CEF emissions in the study area showed positive growth throughout the study period. This conclusion is consistent with that of Wu (2021) [71]. This trend deviates from the “dual carbon” targets set by the central government, indicating significant challenges in achieving these goals. The Yangtze River Delta Region faces relative energy scarcity, relying heavily on external inputs for production activities [72]. The region has many industrial parks that contribute to high energy consumption, especially of fossil fuels, causing an annual rise in CEF emissions.

Observing the local characteristics of the region, we identified carbon emission hotspots in the Yangtze River Delta. In the early years, carbon emissions in the Yangtze River Delta were primarily concentrated in Shanghai, but, over time, they gradually expanded to surrounding areas. Meanwhile, other cities in the region, such as Nanjing, Hefei, Yangzhou, and Ningbo, also emerged as secondary growth poles for carbon emissions. In 2000, Shanghai’s CEF emissions accounted for 22.34%. By 2020, Shanghai’s share had decreased to 11.97%. Jiangsu Province has the highest proportion, with its CEF emissions accounting for 34.39% of the total in 2020 and rising to 43.02% by 2020. This is mainly due to Jiangsu Province’s rapid industrial development and its absorption of outdated production capacity transferred from Shanghai. This indicates that, over the past 20 years, carbon emissions in the Yangtze River Delta have shown spillover effects. In addition to the original growth pole, Shanghai, new secondary growth poles have emerged. Therefore, it is essential to focus on the carbon emissions from the CEF centers and implement targeted policy interventions. At the same time, due to the existence of spillover effects, achieving carbon reduction through independent local governance models has become increasingly difficult. Thus, coordinated carbon reduction efforts among urban agglomerations are urgently needed to promote high-quality development and effectively reduce regional CEF emissions.

According to the second law of geography, which states that spatial separation creates differences among geographic features, there is significant heterogeneity in sectoral carbon emissions across regions [73]. The distribution of carbon emissions also varies by sector. Industrial CEF emissions are primarily concentrated in Shanghai’s Minhang District, Jiading District, and Pudong New Area, as well as in surrounding areas such as Suzhou, Hangzhou, and Ningbo. Transportation CEF emissions are more spatially dispersed, forming small high-density clusters around key transportation hubs such as Hefei, Nanjing, Yangzhou, and Wuxi. The CEF emissions from wholesale and retail, construction, and urban residential sectors share similar spatial distributions, primarily concentrated in major cities within the region and their surrounding areas. Agricultural CEF emissions are mainly concentrated in southern Jiangsu. Given this sectoral heterogeneity, targeted emission reduction policies can be developed. For instance, regions where emissions are concentrated in the industrial sector may benefit from investments in clean technology and energy efficiency. Similarly, areas with high transportation emissions could focus on improving public transportation infrastructure or incentivizing the adoption of electric vehicles.

5.3. Strengths and Limitations

The combination of nighttime light data and land use type produces better regional carbon emissions simulations, achieving a goodness of fit of 0.931, which exceeds estimates from traditional econometric methods [56,65]. Estimates indicate that carbon emissions in the study area rose from 2000 to 2020. Initially showing a bimodal spatial distribution, emissions evolved into a right-skewed pattern centered around Shanghai, with concentric radiation and significant spatial autocorrelation. Hotspots and cold spots form a three-tiered west-to-east pattern, aligning with past findings [74]. Unlike previous models relying solely on nighttime light data for carbon emissions, our model includes land use data, allowing for sector-specific carbon estimates and multiscale analysis of regional emission patterns.

The carbon emission model proposed in this study applies to China and other countries or regions, given the difficulty in obtaining statistical data on carbon emissions. In such cases, the framework proposed in this study can be used for estimation, enabling finer monitoring of carbon emission spatial patterns and providing practical evidence for formulating regional carbon reduction policies.

Although our study provides more accurate multiscale estimates of carbon emissions, there are still limitations. Firstly, the model must demonstrate a better fit for the CEF of indirect sectors. This is primarily due to the weak nighttime activity in these sectors, making it difficult for NTL to capture their CEF directly. Instead, CEF must be estimated indirectly through the correlation between NTL and human activities. However, given that the CEF of indirect sectors constitutes a tiny portion of the total CEF, this does not significantly impact the main research conclusions. Secondly, we focus on using nighttime light data as a proxy for carbon emissions, which serves as a macro indicator of carbon emissions at the regional level. Additionally, considering the heterogeneity across industries and regions, we incorporate land use data and the GTWR model for further identification. However, due to the lack of more granular carbon emission data, we were unable to further refine the analysis at the industry level. Previous studies have used input–output tables to achieve more detailed sectoral carbon emission estimates; however, due to the large time span of input–output tables, the fit is not ideal [75]. In the future, with updated data, subsequent research can consider improving the carbon emission estimates for different sectors, which will provide more accurate cross-industry and cross-region carbon emission estimates. This will help capture the diversity and complexity of carbon emissions across regions and industries more accurately. Finally, the nighttime light data offer carbon emission grid data at a 1 km × 1 km resolution, failing to capture detailed urban emission patterns [76]. Future research should explore higher-resolution satellite data for more precise carbon emissions analysis.

6. Conclusions

This study estimates the CEF of various industries in China. It helps to supplement the existing database. The optimal fitting model is selected based on accuracy indicators to analyze the spatiotemporal patterns of CEF at different scales. The main conclusions are as follows:

(1): The dataset combining nighttime light data with land use type demonstrates a stronger correlation with CEF than nighttime light data alone. Model accuracy metrics show higher R-squared values and lower RMSE and MAE when NTL and land use data are used as independent variables, indicating that the NTL-Landuse approach is superior for CEF estimation. In model selection, we consider potential time, individual, and spatial effects, with results from OLS, FE, and GTWR models, where GTWR outperforms both OLS and FE.
(2): From 2000 to 2020, the total CEF in the Yangtze River Delta increased, with Shanghai experiencing a significant turning point while the other three provinces continued to rise. The primary driver was growth in industrial CEF, followed by urban population increases and expansions in transportation and wholesale/retail CEF due to urbanization. At the same time, agricultural and rural resident CEF saw smaller gains.
(3): In 2000, the Yangtze River Delta region had low carbon emissions and lacked distinct spatial distribution characteristics. By 2010, Shanghai emerged as a significant growth pole, influencing cities like Suzhou and Jiaxing. By 2015, the high-density CEF area radiating from Shanghai expanded to include Suzhou, Wuxi, Jiaxing, and Hangzhou, with other regions also beginning to show small high-density centers. In 2020, while the total CEF in Shanghai began to decline, its radiating area expanded further, connecting with secondary centers to form large contiguous regions. Although industrial CEF decreased across all provinces during the study period, only Shanghai saw a notable drop in its industrial CEF proportion, while other provinces experienced smaller decreases. Additionally, industrial CEF emissions clustered in adjacent areas of Jiangsu, Zhejiang, and Shanghai, whereas agricultural CEF displayed a dispersed trend. The transportation CEF concentrated around major transport hubs, while the wholesale and retail, construction, and urban resident CEF were predominantly found in larger cities.
(4): The global Moran’s I index results indicate that the spatial clustering of CEF in the study area region is continuously strengthening, except for the agricultural CEF, which shows a decline. The spatial pattern of hot and cold spots remains stable, exhibiting a three-level gradient increasing from west to east. The CEF center of gravity has gradually shifted northwest, weakening the characteristics in the southeast–northwest direction.

To achieve regional sustainable development, local governments must implement effective intervention policies. The carbon emission measurements for the Yangtze River Delta region from 2000 to 2020 reveal that carbon emissions exhibit significant spatial spillover, evolving from a single growth pole to a multi-center trend. Since carbon emissions are not confined by administrative boundaries, inter-regional cooperation is crucial. To meet the central government’s “dual carbon” goals, the Yangtze River Delta should leverage its resource advantages and promote coordinated governance for pollution and carbon reduction within a regional integration framework. For the existing carbon emission growth pole, Shanghai, it is crucial to focus on and monitor its emissions to prevent further diffusion while also closely observing emerging secondary growth poles such as Nanjing, Hefei, Yangzhou, and Ningbo. Efforts should be made to accelerate the construction of low-carbon cities in these areas to prevent an accelerated increase in carbon emission growth rates.

In addition to requiring close cooperation between regions, different regions should implement appropriate policies based on their unique characteristics to mitigate carbon emissions. Tailored emission reduction measures can be proposed based on the carbon emissions of different sectors. For industrial carbon emissions centers, such as Shanghai, Suzhou, Jiaxing, Hangzhou, Shaoxing, and Ningbo, the promotion of clean energy should be prioritized. Efforts should be made to create green industrial parks and encourage enterprises to transition to greener technologies through tax subsidies and pollution control measures. For regions with high transportation-related carbon emissions, such as Nanjing, Hefei, Suzhou, Hangzhou, Yangzhou, and Wuxi, public transportation infrastructure should be expanded, and the substitution of traditional fuel-powered vehicles with new energy vehicles should be accelerated. Additionally, efforts should be made to reduce reliance on private cars by improving public transportation facilities. In residential areas, subsidies for new energy vehicles should be increased to encourage residents to scrap older, high-emission vehicles. Incentives, such as discounts on public transportation fares, should also be provided to encourage residents to choose more environmentally friendly modes of transportation. For high carbon emission wholesale and retail sectors in regions such as Shanghai, Hangzhou, Wuxi, and Hefei, local governments should encourage the production and sale of goods locally, shorten supply chains, and establish energy-efficient factories and warehouses to reduce emissions along the supply chain. For regions with high carbon emissions in the construction industry, such as Shanghai, Hangzhou, Hefei, and Suzhou, governments should consider introducing green building certification systems. This would encourage the use of low-carbon materials and energy-efficient construction practices to minimize energy consumption and maximize the use of renewable energy, providing the most comfortable indoor environments with the least energy consumption. In southern Jiangsu, where there are significant agricultural carbon emissions, efforts should be made to reduce the proportion of land used for field embankments, increase the level of agricultural mechanization, and minimize unnecessary fertilizer use. This would enhance agricultural productivity and reduce the carbon emission intensity of the agricultural sector. For cities and rural areas with high carbon emissions, such as Shanghai, Hangzhou, Nanjing, Hefei, and Suzhou, sustainable urbanization should be planned carefully. Urban sprawl should be avoided, and efforts should be made to guide residents to concentrate in specific areas to create economies of scale. Green infrastructure should be laid in densely populated areas to reduce carbon emissions while improving residents’ quality of life.

Author Contributions

K.L.: conceptualization, project administration, funding acquisition. Q.W.: data curation, formal analysis, software, writing—original draft, methodology. X.S.: writing—review and editing, supervision. L.H.: resources, validation. Y.L.: software, visualization. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the Major Program of the National Fund of Philosophy and Social Science of China (grant number 23AZD056).

Data Availability Statement

The data that have been used are confidential.

Conflicts of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A

Table A1. Energy emission factors.

Energy Type	Emission Factor
Raw Coal	1.4875
Cleaned Coal	1.8743
Other Washed Coal	0.595
Briquettes	0.595
Gangue	0.595
Coke	2.023
Coking Oven Gas	1.19
Blast Furnace Gas	0.2677
Converter Gas	0.2677
Other Gas	0.5751
Crude Oil	3.0209
Gasoline	2.9287
Kerosene	3.0372
Diesel Oil	3.0998
Fuel Oil	3.1744
Naphtha	3.057
Lubricating	3.057
Paraffin	3.057
Solvent Oil	3.057
Bitumen Asphalt	3.057
Petroleum Coke	3.057
Liquefied Petroleum Gas	3.1052
Refinery Gas	3.0119
Natural Gas	1.9802
Liquefied Natural Gas	3.2153
Heat	0.095

Table A2. Electricity energy emission factors by region.

Province	Before 2016	2016 and After
Beijing	0.8292	0.6168
Tianjin	0.8733	0.8119
Hebei	0.9148	0.9029
Shanxi	0.8798	0.7399
Inner Mongolia	0.8503	0.7533
Liaoning	0.8357	0.7219
Jilin	0.6787	0.6147
Heilongjiang	0.8158	0.6634
Shanghai	0.7934	0.5641
Jiangsu	0.73566	0.6829
Zhejiang	0.6822	0.5246
Anhui	0.7913	0.7759
Fujian	0.5439	0.391
Jiangxi	0.7635	0.6339
Shandong	0.9236	0.8606
Henan	0.8444	0.7906
Hubei	0.3717	0.3574
Hunan	0.5523	0.4987
Guangdong	0.6379	0.4512
Guangxi	0.4821	0.3938
Hainan	0.6463	0.5147
Chongqing	0.6294	0.4405
Sichuan	0.2891	0.1031
Guizhou	0.6556	0.4275
Yunnan	0.415	0.0921
Shaanxi	0.8696	0.7673
Gansu	0.6124	0.4912
Qinghai	0.2263	0.2602
Ningxia	0.8184	0.6195
Xinjiang	0.7636	0.622

Figure A1. Kernel Density Plot.

References

Lin, B.; Li, Z. Towards world’s low carbon development: The role of clean energy. Appl. Energy 2022, 307, 118160. [Google Scholar] [CrossRef]
Guo, B.; Xie, T.; Zhang, W.; Wu, H.; Zhang, D.; Zhu, X.; Ma, X.; Wu, M.; Luo, P. Rasterizing CO₂ emissions and characterizing their trends via an enhanced population-light index at multiple scales in China during 2013–2019. Sci. Total. Environ. 2023, 905, 167309. [Google Scholar] [CrossRef]
Yang, W.-J.; Tan, M.-Z.; Chu, S.-H.; Chen, Z. Carbon emission and financial development under the “double carbon” goal: Considering the upgrade of industrial structure. Front. Environ. Sci. 2023, 10, 1091537. [Google Scholar] [CrossRef]
Yang, Z.; Shao, S.; Xu, L.; Yang, L. Can regional development plans promote economic growth? City-level evidence from China. Socio-Econ. Plan. Sci. 2022, 83, 101212. [Google Scholar] [CrossRef]
Sun, X.; Ma, X.; Shi, F.; Han, M.; Xie, H.; He, Y. Decomposition of China’s regional carbon emission paths: An analysis of environmental input and output considering regional development differences. Environ. Sci. Pollut. Res. 2022, 29, 62537–62559. [Google Scholar] [CrossRef] [PubMed]
Zhang, X.; Cai, Z.; Song, W.; Yang, D. Mapping the spatial-temporal changes in energy consumption-related carbon emissions in the Beijing-Tianjin-Hebei region via nighttime light data. Sustain. Cities Soc. 2023, 94, 104476. [Google Scholar] [CrossRef]
Sun, Y.; Zhao, T.; Xia, L. Spatial-temporal differentiation of carbon efficiency and coupling coordination degree of Chinese county territory and obstacles analysis. Sustain. Cities Soc. 2022, 76, 103429. [Google Scholar] [CrossRef]
Jing, Q.; Bai, H.; Luo, W.; Cai, B.; Xu, H. A top-bottom method for city-scale energy-related CO₂ emissions estimation: A case study of 41 Chinese cities. J. Clean. Prod. 2018, 202, 444–455. [Google Scholar] [CrossRef]
Sun, C.; Wang, B.; Miao, H. Spatiotemporal dynamics of CO₂ emissions: A case study of the “New Yangtze River Delta” in China. Environ. Sci. Pollut. Res. 2023, 30, 40961–40977. [Google Scholar] [CrossRef]
Chen, J.; Gao, M.; Cheng, S.; Hou, W.; Song, M.; Liu, X.; Liu, Y.; Shan, Y. County-level CO₂ emissions and sequestration in China during 1997–2017. Sci. Data 2020, 7, 391. [Google Scholar] [CrossRef]
Keola, S.; Andersson, M.; Hall, O. Monitoring economic development from space: Using nighttime light and land cover data to measure economic growth. World Dev. 2015, 66, 322–334. [Google Scholar] [CrossRef]
Watmough, G.R.; Marcinko, C.L.J.; Sullivan, C.; Tschirhart, K.; Mutuo, P.K.; Palm, C.A.; Svenning, J.-C. Socioecologically informed use of remote sensing data to predict rural household poverty. Proc. Natl. Acad. Sci. USA 2019, 116, 1213–1218. [Google Scholar] [CrossRef]
Doll, C.N.; Pachauri, S. Estimating rural populations without access to electricity in developing countries through night-time light satellite imagery. Energy Policy 2010, 38, 5661–5670. [Google Scholar] [CrossRef]
Bustos, M.F.A.; Hall, O.; Andersson, M. Nighttime lights and population changes in Europe 1992–2012. Ambio 2015, 44, 653–665. [Google Scholar] [CrossRef] [PubMed]
Mossoux, S.; Kervyn, M.; Soulé, H.; Canters, F. Mapping population distribution from high resolution remotely sensed imagery in a data poor setting. Remote Sens. 2018, 10, 1409. [Google Scholar] [CrossRef]
Yang, D.; Luan, W.; Qiao, L.; Pratama, M. Modeling and spatio-temporal analysis of city-level carbon emissions based on nighttime light satellite imagery. Appl. Energy 2020, 268, 114696. [Google Scholar] [CrossRef]
Han, J.; Meng, X.; Liang, H.; Cao, Z.; Dong, L.; Huang, C. An improved nightlight-based method for modeling urban CO₂ emissions. Environ. Model. Softw. 2018, 107, 307–320. [Google Scholar] [CrossRef]
Fotheringham, A.S.; Crespo, R.; Yao, J. Geographical and temporal weighted regression (GTWR). Geogr. Anal. 2015, 47, 431–452. [Google Scholar] [CrossRef]
Wang, P.; Li, Y.; Huang, G.; Wang, S.; Suo, C.; Ma, Y. A multi-scenario factorial analysis and multi-regional input-output model for analyzing CO₂ emission reduction path in Jing-Jin-Ji region. J. Clean. Prod. 2021, 300, 126782. [Google Scholar] [CrossRef]
Harris, S.; Weinzettel, J.; Bigano, A.; Källmén, A. Low carbon cities in 2050? GHG emissions of European cities using production-based and consumption-based emission accounting methods. J. Clean. Prod. 2020, 248, 119206. [Google Scholar] [CrossRef]
Souza Leao, E.B.; Nascimento, L.F.M.; Andrade, J.C.S.; Oliveira, J.A.P. Carbon accounting approaches and reporting gaps in urban emissions: An analysis of the Greenhouse Gas inventories and climate action plans in Brazilian cities. J. Clean. Prod. 2020, 245, 118930. [Google Scholar] [CrossRef]
Froemelt, A.; Geschke, A.; Wiedmann, T. Quantifying carbon flows in Switzerland: Top-down meets bottom-up modelling. Environ. Res. Lett. 2021, 16, 014018. [Google Scholar] [CrossRef]
Eggleston, H.S.; Buendia, L.; Miwa, K.; Ngara, T.; Tanabe, K. 2006 IPCC Guidelines for National Greenhouse Gas Inventories; IPCC: Geneva, Switzerland, 2006. [Google Scholar]
Long, Y.; Sharifi, A.; Huang, L.; Chen, J. Urban carbon accounting: An overview. Urban Clim. 2022, 44, 101195. [Google Scholar] [CrossRef]
Ahmetoğlu, S.; Tanık, A. Management of carbon footprint and determination of GHG emission sources in construction sector. Int. J. Environ. Geoinform. 2020, 7, 191–204. [Google Scholar] [CrossRef]
Trofimenko, Y.; Komkov, V.; Trofimenko, K. Forecast of energy consumption and greenhouse gas emissions by road transport in Russia up to 2050. Transp. Res. Procedia 2020, 50, 698–707. [Google Scholar] [CrossRef]
Petrov, S.P. Application of the IPCC Methodology to Assess the Spatial Distribution of CO₂ Emissions in Russian Ferrous Metallurgy. Stud. Russ. Econ. Dev. 2024, 35, 54–59. [Google Scholar] [CrossRef]
Jiang, M.; Medlyn, B.E.; Drake, J.E.; Duursma, R.A.; Anderson, I.C.; Barton, C.V.M.; Boer, M.M.; Carrillo, Y.; Castañeda-Gómez, L.; Collins, L.; et al. The fate of carbon in a mature forest under carbon dioxide enrichment. Nature 2020, 580, 227–231. [Google Scholar] [CrossRef]
Chen, L.; Xu, L.; Cai, Y.; Yang, Z. Spatiotemporal patterns of industrial carbon emissions at the city level. Resour. Conserv. Recycl. 2021, 169, 105499. [Google Scholar] [CrossRef]
Dong, F.; Yu, B.; Hadachin, T.; Dai, Y.; Wang, Y.; Zhang, S.; Long, R. Drivers of carbon emission intensity change in China. Resour. Conserv. Recycl. 2018, 129, 187–201. [Google Scholar] [CrossRef]
Burke, M.; Driscoll, A.; Lobell, D.B.; Ermon, S. Using satellite imagery to understand and promote sustainable development. Science 2021, 371, eabe8628. [Google Scholar] [CrossRef]
Bennett, M.M.; Smith, L.C. Advances in using multitemporal night-time lights satellite imagery to detect, estimate, and monitor socioeconomic dynamics. Remote Sens. Environ. 2017, 192, 176–197. [Google Scholar] [CrossRef]
Zheng, Q.; Seto, K.C.; Zhou, Y.; You, S.; Weng, Q. Nighttime light remote sensing for urban applications: Progress, challenges, and prospects. ISPRS J. Photogramm. Remote Sens. 2023, 202, 125–141. [Google Scholar] [CrossRef]
Elvidge, C.D.; Baugh, K.E.; Kihn, E.A.; Kroehl, H.W.; Davis, E.R.; Davis, C.W. Relation between satellite observed visible-near infrared emissions, population, economic activity and electric power consumption. Int. J. Remote Sens. 1997, 18, 1373–1379. [Google Scholar] [CrossRef]
Wang, M.; Wang, Y.; Teng, F.; Ji, Y. The spatiotemporal evolution and impact mechanism of energy consumption carbon emissions in China from 2010 to 2020 by integrating multisource remote sensing data. J. Environ. Manag. 2023, 346, 119054. [Google Scholar] [CrossRef] [PubMed]
Zheng, Y.; Fan, M.; Cai, Y.; Fu, M.; Yang, K.; Wei, C. Spatio-temporal pattern evolution of carbon emissions at the city-county-town scale in Fujian Province based on DMSP/OLS and NPP/VIIRS nighttime light data. J. Clean. Prod. 2024, 442, 140958. [Google Scholar] [CrossRef]
Zhang, Y.; Quan, J.; Kong, Y.; Wang, Q.; Zhang, Y.; Zhang, Y. Research on the fine-scale spatial-temporal evolution characteristics of carbon emissions based on nighttime light data: A case study of Xi’an city. Ecol. Inform. 2024, 79, 102454. [Google Scholar] [CrossRef]
Wu, R.; Wang, R.; Nian, Z.; Gu, J. Spatio-temporal variation and decoupling effects of energy carbon footprint based on nighttime light data: Evidence from counties in northeast China. Atmos. Pollut. Res. 2025, 16, 102366. [Google Scholar] [CrossRef]
Lu, S.; Xiao, Y.; Lu, Y.; Lin, J. Spatialization of electricity consumption by combining high-resolution nighttime light remote sensing and urban functional zoning information. Geo-Spat. Inf. Sci. 2024, 27, 1–14. [Google Scholar] [CrossRef]
Meng, X.; Han, J.; Huang, C. An improved vegetation adjusted nighttime light urban index and its application in quantifying spatiotemporal dynamics of carbon emissions in China. Remote Sens. 2017, 9, 829. [Google Scholar] [CrossRef]
Shi, K.; Chen, Y.; Yu, B.; Xu, T.; Chen, Z.; Liu, R.; Li, L.; Wu, J. Modeling spatiotemporal CO₂ (carbon dioxide) emission dynamics in China from DMSP-OLS nighttime stable light data using panel data analysis. Appl. Energy 2016, 168, 523–533. [Google Scholar] [CrossRef]
Wu, H.; Yang, Y.; Li, W. Dynamic spatiotemporal evolution and spatial effect of carbon emissions in urban agglomerations based on nighttime light data. Sustain. Cities Soc. 2024, 113, 105712. [Google Scholar] [CrossRef]
Wang, G.; Hu, Q.; He, L.; Guo, J.; Huang, J.; Zhong, L. The estimation of building carbon emission using nighttime light images: A comparative study at various spatial scales. Sustain. Cities Soc. 2024, 101, 105066. [Google Scholar] [CrossRef]
Shi, K.; Chen, Z.; Cui, Y.; Wu, J.; Yu, B. NPP-VIIRS nighttime light data have different correlated relationships with fossil fuel combustion carbon emissions from different sectors. IEEE Geosci. Remote Sens. Lett. 2020, 18, 2062–2066. [Google Scholar] [CrossRef]
Xia, B. Spatial Characteristics and Driving Mechanisms of Carbon Neutrality Progress in Tourism Attractions in the Qinghai–Tibet Plateau Based on Remote Sensing Methods. Remote Sens. 2024, 16, 4481. [Google Scholar] [CrossRef]
Wei, W.; Chen, D.; Zhang, X.; Ma, L.; Xie, B.; Zhou, J.; Wang, M.; Wei, X.; Yang, S.; Yan, P. Estimation of carbon emissions from different industrial categories integrated nighttime light and POI data—A case study in the Yellow River Basin. J. Environ. Manag. 2024, 370, 122418. [Google Scholar] [CrossRef]
Liu, C.; Hu, S.; Wu, S.; Song, J.; Li, H. County-level land use carbon emissions in China: Spatiotemporal patterns and impact factors. Sustain. Cities Soc. 2024, 105, 105304. [Google Scholar] [CrossRef]
Gao, F.; Wu, J.; Xiao, J.; Li, X.; Liao, S.; Chen, W. Spatially explicit carbon emissions by remote sensing and social sensing. Environ. Res. 2023, 221, 115257. [Google Scholar] [CrossRef] [PubMed]
Wei, W.; Zhang, X.; Cao, X.; Zhou, L.; Xie, B.; Zhou, J.; Li, C. Spatiotemporal dynamics of energy-related CO₂ emissions in China based on nighttime imagery and land use data. Ecol. Indic. 2021, 131, 108132. [Google Scholar] [CrossRef]
Wu, Y.; Shi, K.; Chen, Z.; Liu, S.; Chang, Z. Developing improved time-series DMSP-OLS-like data (1992–2019) in China by integrating DMSP-OLS and SNPP-VIIRS. IEEE Trans. Geosci. Remote Sens. 2021, 60, 4407714. [Google Scholar] [CrossRef]
National Bureau of Statistics of China. China Energy Statistical Yearbook. 2023. Available online: https://www.stats.gov.cn/sj/ndsj/2023/indexeh.htm (accessed on 10 October 2024).
Xia, S.; Shao, H.; Wang, H.; Xian, W.; Shao, Q.; Yin, Z.; Qi, J. Spatio-temporal dynamics and driving forces of multi-scale CO₂ emissions by integrating DMSP-OLS and NPP-VIIRS data: A case study in Beijing-Tianjin-Hebei, China. Remote Sens. 2022, 14, 4799. [Google Scholar] [CrossRef]
Ustaoglu, E.; Bovkır, R.; Aydınoglu, A.C. Spatial distribution of GDP based on integrated NPS-VIIRS nighttime light and MODIS EVI data: A case study of Turkey. Environ. Dev. Sustain. 2021, 23, 10309–10343. [Google Scholar] [CrossRef]
Li, S.; Cao, X.; Zhao, C.; Jie, N.; Liu, L.; Chen, X.; Cui, X. Developing a pixel-scale corrected nighttime light dataset (PCNL, 1992–2021) combining DMSP-OLS and NPP-VIIRS. Remote Sens. 2023, 15, 3925. [Google Scholar] [CrossRef]
Shi, K.; Yu, B.; Huang, Y.; Hu, Y.; Yin, B.; Chen, Z.; Chen, L.; Wu, J. Evaluating the ability of NPP-VIIRS nighttime light data to estimate the gross domestic product and the electric power consumption of China at multiple scales: A comparison with DMSP-OLS data. Remote Sens. 2014, 6, 1705–1724. [Google Scholar] [CrossRef]
Zhao, J.; Ji, G.; Yue, Y.; Lai, Z.; Chen, Y.; Yang, D.; Yang, X.; Wang, Z. Spatio-temporal dynamics of urban residential CO₂ emissions and their driving forces in China using the integrated two nighttime light datasets. Appl. Energy 2019, 235, 612–624. [Google Scholar] [CrossRef]
Li, X.; Zhou, Y.; Zhao, M.; Zhao, X. A harmonized global nighttime light dataset 1992–2018. Sci. Data 2020, 7, 168. [Google Scholar] [CrossRef]
Pachauri, R.K.; Allen, M.R.; Barros, V.R.; Broome, J.; Cramer, W.; Christ, R.; Church, J.A.; Clarke, L.; Dahe, Q.; Dasgupta, P.; et al. Climate Change 2014: Synthesis Report. In Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change; IPCC: Geneva, Switzerland, 2014; p. 151. [Google Scholar]
GB/T 2589-2020; National Standard of the People’s Republic of China. General Principles for Calculation of the Comprehensive Energy Consumption. Standardization Administration of China: Beijing, China, 2020.
Yang, X.; Liu, X. Accuracy analysis of inverting provincial-level carbon emissions from night-time light data in China: Comparison based on international carbon emission data. In IOP Conference Series: Earth and Environmental Science; IOP Publishing: Bristol, UK, 2020; Volume 601, p. 012046. [Google Scholar] [CrossRef]
Ou, J.; Liu, X.; Wang, S.; Xie, R.; Li, X. Investigating the differentiated impacts of socioeconomic factors and urban forms on CO₂ emissions: Empirical evidence from Chinese cities of different developmental levels. J. Clean. Prod. 2019, 226, 601–614. [Google Scholar] [CrossRef]
Letu, H.; Hara, M.; Tana, G.; Nishio, F. A saturated light correction method for DMSP/OLS nighttime satellite imagery. IEEE Trans. Geosci. Remote Sens. 2011, 50, 389–396. [Google Scholar] [CrossRef]
Cai, M.; Shi, Y.; Ren, C.; Yoshida, T.; Yamagata, Y.; Ding, C.; Zhou, N. The need for urban form data in spatial modeling of urban carbon emissions in China: A critical review. J. Clean. Prod. 2021, 319, 128792. [Google Scholar] [CrossRef]
Cui, Y.; Zhang, W.; Wang, C.; Streets, D.G.; Xu, Y.; Du, M.; Lin, J. Spatiotemporal dynamics of CO₂ emissions from central heating supply in the North China Plain over 2012–2016 due to natural gas usage. Appl. Energy 2019, 241, 245–256. [Google Scholar] [CrossRef]
Meng, L.; Graus, W.; Worrell, E.; Huang, B. Estimating CO₂ (carbon dioxide) emissions at urban scales by DMSP/OLS (Defense Meteorological Satellite Program’s Operational Linescan System) nighttime light imagery: Methodological challenges and a case study for China. Energy 2014, 71, 468–478. [Google Scholar] [CrossRef]
Zhang, W.; Cui, Y.; Wang, J.; Wang, C.; Streets, D.G. How does urbanization affect CO₂ emissions of central heating systems in China? An assessment of natural gas transition policy based on nighttime light data. J. Clean. Prod. 2020, 276, 123188. [Google Scholar] [CrossRef]
Lotwick, H.W.; Silverman, B.W. Methods for analysing spatial processes of several types of points. J. R. Stat. Soc. Ser. B Stat. Methodol. 1982, 44, 406–413. [Google Scholar] [CrossRef]
Henderson, J.; Storeygard, V.A.; Weil, D.N. Measuring economic growth from outer space. Am. Econ. Rev. 2012, 102, 994–1028. [Google Scholar] [CrossRef] [PubMed]
Pinkovskiy, M.; Sala-i-Martin, X. Lights, camera… income! Illuminating the national accounts-household surveys debate. Q. J. Econ. 2016, 131, 579–631. [Google Scholar] [CrossRef]
Shan, Y.; Guan, Y.; Hang, Y.; Zheng, H.; Li, Y.; Guan, D.; Li, J.; Zhou, Y.; Li, L.; Hubacek, K. City-level emission peak and drivers in China. Sci. Bull. 2022, 67, 1910–1920. [Google Scholar] [CrossRef] [PubMed]
Wu, W.; Zhang, T.; Xie, X.; Huang, Z. Regional low carbon development pathways for the Yangtze River Delta region in China. Energy Policy 2021, 151, 112172. [Google Scholar] [CrossRef]
Chen, X.; Di, Q.; Liang, C. The mechanism and path of pollution reduction and carbon reduction affecting high quality economic development—Taking the Yangtze River Delta urban agglomeration as an example. Appl. Energy 2024, 376, 124340. [Google Scholar] [CrossRef]
Cui, Y.; Khan, S.U.; Deng, Y.; Zhao, M. Spatiotemporal heterogeneity, convergence and its impact factors: Perspective of carbon emission intensity and carbon emission per capita considering carbon sink effect. Environ. Impact Assess. Rev. 2022, 92, 106699. [Google Scholar] [CrossRef]
Lv, T.; Hu, H.; Zhang, X.; Xie, H.; Fu, S.; Wang, L. Spatiotemporal pattern of regional carbon emissions and its influencing factors in the Yangtze River Delta urban agglomeration of China. Environ. Monit. Assess. 2022, 194, 515. [Google Scholar] [CrossRef]
Chen, G.; Chen, Z. Carbon emissions and resources use by Chinese economy 2007: A 135-sector inventory and input–output embodiment. Commun. Nonlinear Sci. Numer. Simul. 2010, 15, 3647–3732. [Google Scholar] [CrossRef]
Hu, K.; Liu, Z.; Shao, P.; Ma, K.; Xu, Y.; Wang, S.; Wang, Y.; Wang, H.; Di, L.; Xia, M.; et al. A review of satellite-based CO₂ data reconstruction studies: Methodologies, challenges, and advances. Remote Sens. 2024, 16, 3818. [Google Scholar] [CrossRef]

Figure 1. Geographical position of the Yangtze River Delta region.

Figure 2. Flowchart of the methodology.

Figure 3. Spatial mapping of CEF.

Figure 4. Changing trend of CEF in the Yangtze River Delta.

Figure 5. Spatial distribution of CEF in the Yangtze River Delta.

Figure 6. Spatial distribution of CEF in various departments in the Yangtze River Delta.

Figure 7. Distribution pattern of CEF hot spots in the Yangtze River Delta.

Figure 8. CEF growth rate of Yangtze River Delta.

Table 1. Pearson correlation coefficient between NTL-Landuse and CEF.

Type	Total	Indirect Section
Type	Total	Agriculture	Transportation	Wholesale and Retail Trade
Agricultural land	0.8720 *	0.6554 *	0.7088 *	0.7027 *
Urban land	0.8708 *	0.5840 *	0.7576 *	0.7037 *
Rural land	0.8648 *	0.5797 *	0.6290 *	0.6010 *
Construction land	0.7163 *	0.4485 *	0.7151 *	0.6810 *
Undeveloped land	0.0597	0.2671 *	−0.0158	0.0281
Type	Direct Section
Type	Towns	Rural areas	Industry	Building
Agricultural land	0.8271 *	0.8052 *	0.8425 *	0.7568 *
Urban land	0.8678 *	0.7453 *	0.8389 *	0.7547 *
Rural land	0.7881 *	0.7870 *	0.8559 *	0.6589 *
Construction land	0.7362 *	0.6814 *	0.6665 *	0.7770 *
Undeveloped land	−0.0576	−0.0606	0.0754	0.0137

Note: * denotes p-values were all <1%, passing the significance test of 1%.

Table 2. NTL-Landuse and NTL-Total correlation comparison.

	Sector	NTL-Landuse	NTL-Total
Indirect Section	Agriculture	0.6554 *	0.6622 *
	Transportation	0.7961 *	0.7300 *
	Wholesale and retail trade	0.7430 *	0.7143 *
Direct Section	Towns	0.8945 *	0.8479 *
	Rural areas	0.8295 *	0.8078 *
	Industry	0.8681 *	0.8661 *
	Building	0.8065 *	0.7708 *
	Total	0.8946 *	0.8946 *

Note: * denotes p-values were all <1%, passing the significance test of 1%.

Table 3. Performance comparison of three models for estimating CEF.

Independent Variables	Index	OLS	FE	GTWR
NTL-Total	RMSE	0.33204	0.77811	0.21807
	R2	0.8408	0.8751	0.93133
	MAE	0.25601	0.64211	0.17324
NTL-Landuse	RMSE	0.33171	0.44904	0.21858
	R2	0.84112	0.70884	0.93101
	MAE	0.25223	0.34431	0.17278

Table 4. Global Moran’s I values for CEF.

Year	Sector CEF	Moran’s I	p-Value
2000	Total	0.315	0.001
	Agriculture	0.415	0.001
	Industry	0.339	0.001
	Building	0.236	0.002
	Transportation	0.216	0.001
	Wholesale and retail trade	0.256	0.002
	Towns	0.262	0.001
	Rural areas	0.240	0.001
2010	Total	0.372	0.001
	Agriculture	0.404	0.001
	Industry	0.394	0.001
	Building	0.302	0.001
	Transportation	0.310	0.001
	Wholesale and retail trade	0.307	0.001
	Towns	0.309	0.001
	Rural areas	0.311	0.001
2015	Total	0.366	0.001
	Agriculture	0.402	0.001
	Industry	0.411	0.001
	Building	0.266	0.001
	Transportation	0.281	0.001
	Wholesale and retail trade	0.274	0.001
	Towns	0.286	0.001
	Rural areas	0.278	0.001
2020	Total	0.380	0.001
	Agriculture	0.354	0.001
	Industry	0.413	0.001
	Building	0.326	0.001
	Transportation	0.352	0.001
	Wholesale and retail trade	0.335	0.001
	Towns	0.358	0.001
	Rural areas	0.308	0.001

Table 5. CEF standard deviation ellipse of the Yangtze River Delta.

Year	Center X	Center Y	X Std Dist	Y Std Dist	Shape-Area	Rotation
2000	119.9208	31.2356	1.548122	2.686693	13.54527	142.8634
2010	119.7034	31.31735	1.601379	2.696235	13.72098	139.1775
2015	119.5993	31.25213	1.631479	2.820925	14.23804	139.3094
2020	119.4639	31.36714	1.675366	2.870402	14.52843	138.3116

Table 6. Cross validation results.

	County CE	City CE_CEADs	Province CE_CEADs
CEF	0.7416 *	0.7202 *	0.8773 *
Direct CEF	0.7641 *	0.7339 *	0.8720 *
Indirect CEF	0.6240 *	0.6405 *	0.8559 *

Note: * denotes p-values were all <1%, passing the significance test of 1%.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

Share and Cite

MDPI and ACS Style

Lv, K.; Wang, Q.; Shi, X.; Huang, L.; Liu, Y. Multi-Scale Mapping of Energy Consumption Carbon Emission Spatiotemporal Characteristics: A Case Study of the Yangtze River Delta Region. Land 2025, 14, 95. https://doi.org/10.3390/land14010095

AMA Style

Lv K, Wang Q, Shi X, Huang L, Liu Y. Multi-Scale Mapping of Energy Consumption Carbon Emission Spatiotemporal Characteristics: A Case Study of the Yangtze River Delta Region. Land. 2025; 14(1):95. https://doi.org/10.3390/land14010095

Chicago/Turabian Style

Lv, Kangjuan, Qiming Wang, Xunpeng Shi, Li Huang, and Yatian Liu. 2025. "Multi-Scale Mapping of Energy Consumption Carbon Emission Spatiotemporal Characteristics: A Case Study of the Yangtze River Delta Region" Land 14, no. 1: 95. https://doi.org/10.3390/land14010095

APA Style

Lv, K., Wang, Q., Shi, X., Huang, L., & Liu, Y. (2025). Multi-Scale Mapping of Energy Consumption Carbon Emission Spatiotemporal Characteristics: A Case Study of the Yangtze River Delta Region. Land, 14(1), 95. https://doi.org/10.3390/land14010095

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

Multi-Scale Mapping of Energy Consumption Carbon Emission Spatiotemporal Characteristics: A Case Study of the Yangtze River Delta Region

Abstract

1. Introduction

2. Literature Review