Temporo-Spatial Relationship Between Energy Consumption, Air Pollution and Carbon Emissions in the Guangdong–Hong Kong–Macao Greater Bay Area, China

Abstract

The Guangdong–Hong Kong–Macao Greater Bay Area (GBA) is a key economic region in China facing increasing pressure to balance socioeconomic development with environmental protection and energy conservation. This study examines the interrelationships among energy consumption, air pollutants (PM_2.5, NO₂, and SO₂), and carbon dioxide (CO₂) emissions in the GBA from 2000 to 2020. Using spatial correlation matrices and temporo-spatial decoupling analysis, we assess spatial patterns, temporal dynamics, and interactions among these factors. Results show that the GBA has made significant progress in reducing air pollution and carbon emissions. Notably, since 2013, concentrations of PM_2.5, NO₂, and SO₂ have decoupled markedly from energy consumption, reflecting effective pollution control measures. Although CO₂ emissions have decreased more gradually, the trend remains positive, indicating steady advances in carbon management. These findings underscore the need for continued optimization of the energy structure to achieve coordinated control of energy use, air quality, and carbon emissions—essential for promoting sustainable, high-quality development in the region.

1. Introduction

The intersection between energy and environment constitutes one of the most pressing challenges in global sustainable development; however, while energy serves as the foundation and driving force for socioeconomic development, its consumption patterns invariably generate significant environmental impacts, including atmospheric pollution and carbon emissions, which exert enormous influence on human health and environmental sustainability [1,2,3]. The acceleration of climate change has amplified the environmental footprint of energy consumption patterns, in developing countries, particularly in those undergoing rapid industrialization. Among these patterns, China’s energy structure presents a critical case study: fossil fuels, notably coal and oil, dominate the national energy portfolio, accounting for approximately 80% of total consumption [4]. These conventional energy sources are responsible for disproportionate emissions of greenhouse gases (e.g., CO₂, and CH₄) and atmospheric pollutants (e.g., SO₂, NO₂, PM_2.5, and PM₁₀), collectively contributing to persistent smog episodes and acid rain formation [5,6]. Implementing targeted pollution control measures, optimizing energy systems and enhancing air quality are therefore crucial in China with both immediate practical benefits and long-term strategic value. The effective execution of these strategies depends on a thorough understanding of the relationship between energy consumption, air pollution and carbon emissions.

Numerous studies have explored the energy-environment relationship at various scales. The application of methods such as index decomposition analysis has helped clarify the structural, intensity, and other effects within changes in energy consumption and emissions [7]. Further, causality studies based on dynamic panel data have revealed the complex interconnections between energy consumption, economic growth and CO₂ emissions in many countries and economies [8]. Dynamic causality analysis targeting countries like Pakistan indicates tight linkages among its energy consumption, carbon emissions, and economic growth [9]. At the regional level, research on key Chinese regions like the Beijing-Tianjin-Hebei area and the Yangtze River Delta has revealed strong spatial coupling between energy-intensive areas and PM_2.5 concentrations, often attributed to industrial agglomeration and coal consumption [10,11]. The study by Xu et al. [12] specifically highlighted the close temporal correlation between tropospheric NO₂ column concentrations and energy consumption trends in China from 2005 to 2013. Recently, Lei et al. [13] investigated the spatiotemporal trajectory of energy efficiency in the (hereinafter referred to as the GBA) and its implications for economic transformation pathways.

Despite these contributions, significant research gaps persist. Although previous studies have examined energy efficiency and specific pollutants (e.g., land use-related nitrogen emissions) in the GBA [14], a comprehensive long-term, large-scale spatiotemporal analysis simultaneously encompassing energy consumption, multiple key air pollutants (PM_2.5, SO₂, NO₂), and carbon emissions is notably absent. Most existing research focuses on single pollutants or a single dimension (spatial or temporal), lacking a systematic revelation of the synergies and trade-offs within the complex “energy-air pollution-carbon emissions” system over long time series. The GBA, as a coastal, energy-importing hub experiencing extremely rapid development, coupled with the implementation of the unique national strategy outlined in the “Outline Development Plan for the Guangdong–Hong Kong–Macao Greater Bay Area,” makes it a critical and underexplored case study [15,16,17]. The Plan mandates a green development model prioritizing energy conservation, resource recycling, and low-carbon transitions. Therefore, systematically examining its spatiotemporal effectiveness and extracting transferable governance insights is paramount.

Therefore, to address these literature gaps and provide a robust analytical framework for the GBA, this study establishes the following specific scientific objectives: (1) to quantify the spatial correlation and identify the evolving spatial patterns between energy consumption and the concentrations of PM_2.5, NO₂, SO₂, and CO₂ across the GBA from 2000 to 2020; (2) to evaluate the temporal decoupling states between energy consumption and each atmospheric environmental indicator (PM_2.5, NO₂, SO₂, CO₂) over distinct phases within the study period, identifying key transition points and trends; (3) to synthesize the spatiotemporal findings into actionable insights for regional energy-environment policy optimization and to discuss their potential applicability for other urban agglomerations globally. By integrating spatial correlation matrices and temporo-spatial decoupling analysis, this research aims to systematically unravel the complex interactions within the GBA’s energy-environment system, thereby contributing scientific evidence for achieving coordinated control over energy consumption, air quality, and carbon emissions.

2. Study Area and Dataset

2.1. Study Area

Figure 1 illustrates the GBA urban agglomeration that encompasses nine cities in Guangdong Province (Guangzhou, Shenzhen, Zhuhai, Foshan, Zhongshan, Dongguan, Zhaoqing, Jiangmen and Huizhou) and two Special Administrative Regions (Hong Kong and Macao). Established as a national strategic initiative to build a globally competitive urban agglomeration, the GBA has emerged as one of the world’s four leading bay areas, alongside the New York Bay Area and the San Francisco Bay Area in the USA and the Tokyo Bay Area in Japan.

Spanning 56,000 km² with a population exceeding 86 million, the GBA has emerged as one of China’s most economically vibrant and globally integrated regions, boasting a GDP of over RMB 13 trillion in 2022. Over the past two decades, the GBA has witnessed extraordinary economic expansion, with its GDP growing from 2.34 trillion RMB in 2000 to its current level, reflecting its rapid development trajectory [18]. This economic boom has driven a significant increase in energy consumption, a trend likely to continue as the region further develops. Historically, the GBA’s energy consumption was dominated by conventional, fossil fuels, in particular coal and petroleum, which have been the primary sources of air pollutants and CO₂ emissions. In recent years, however, the region has been actively diversifying its energy portfolio through renewable energy technologies, signaling a shift toward more sustainable development pathways [17].

2.2. Dataset

The dataset utilized in this study include energy consumption, air pollution and carbon emission data of the GBA. City-level energy consumption data in China are notably incomplete, with available records covering only 53.5% of cities in Guangdong Province from 2000 to 2020. To enable analysis at this scale, a random forest–based modeling framework was implemented to estimate annual energy consumption for all cities within the GBA. The dataset of annual energy consumption used in this study was converted to million tons (i.e., 10⁶ Mt) of standard coal at a resolution of 500 m. The model leverages spatially explicit variables: nighttime light data [19], population data [20], and urban impervious surface data [21]. Full methodological details are available in a dedicated publication [22], and subsequent validation exercises demonstrate that the estimated dataset meets acceptable accuracy standards for research applications.

The atmospheric pollution data, including PM_2.5, SO₂ and NO₂ concentrations, were obtained from the high-resolution and high-quality near-surface air pollutant database High Air Pollutants (CHAP) dataset [23,24]. The following types of air pollution data are included:

• Annual mean concentrations of PM_2.5 during 2000–2020 at 1 km spatial resolution;
• Annual mean concentrations of SO₂ during 2013–2020 at 10 km spatial resolution;
• Annual mean concentrations of NO₂ during 2008–2020 at 10 km spatial resolution.

The carbon emission data were gathered from the Emissions Database for Global Atmospheric Research (EDGAR) [25], with a spatial resolution of 0.1° and a temporal coverage from 2000 to 2020. The original data, provided in kg/m²/s, were converted into total emission quantities measured in 10⁶ Mt for consistency.

Table 1. Description of the data used in this study.

Data	Unit	Spatial Resolution	Year	Source
Annual energy consumption	106 Mt of standard coal	500 m	2000–2020	Annual energy consumption at the city level retrieved from nighttime light data [19], population data [20], and urban impervious surface data [21] using a random forest model [22]
PM2.5 Concentrations	µg/m3	1 km	2000–2020	The high-resolution and high-quality near-surface air pollutant database (CHAP) (https://data.tpdc.ac.cn (accessed on 25 September 2023)
SO2 Concentrations	µg/m3	10 km	2013–2020	The high-resolution and high-quality near-surface air pollutant database (CHAP) (https://data.tpdc.ac.cn (accessed on 15 October 2023)
NO2 Concentrations	µg/m3	10 km	2008–2020	The high-resolution and high-quality near-surface air pollutant database (CHAP) (https://data.tpdc.ac.cn (accessed on 22 October 2023)
CO2 Concentrations	kg/m2/s	0.1°	2000–2020	Emissions Database for Global Atmospheric Research (EDGAR) (https://edgar.jrc.ec.europa.eu (accessed on 26 October 2023)
Annual energy consumption structure of Zhuhai City	106 Mt of standard coal	-	2010–2020	Zhuhai Statistical Yearbooks

contains a brief description of the relevant data.

3. Methods

3.1. Data Preprocessing for Spatial Analysis

Prior to performing spatial correlation analysis, a critical data preprocessing step was undertaken to ensure spatial comparability and consistency across all raster datasets. The original data for energy consumption, PM_2.5, SO₂, NO₂, and CO₂ emissions were obtained from diverse sources with varying spatial resolutions and coordinate systems. To enable pixel-to-pixel correlation analysis, all raster datasets were harmonized to a common spatial framework.

Specifically, a uniform spatial resolution of 10 km × 10 km was established for the analysis. Datasets with finer resolutions were aggregated to this target resolution, which preserves the representativeness of the original data. Concurrently, all datasets were projected into a consistent geographic coordinate system (WGS 84) and clipped to the exact administrative boundary of the GBA using a standard mask.

This resampling and harmonization process is essential for minimizing artifacts caused by scale mismatch and ensuring that the calculated spatial correlation coefficients genuinely reflect the substantive relationships between energy consumption and atmospheric variables, rather than technical discrepancies in data preparation.

3.2. Spatial Correlation Matrices

In geospatial studies, analyzing relationships between spatially aligned raster datasets constitutes a core analytical task. The methodological framework of spatial correlation matrices provides a rigorous approach to quantify these interdependencies [26]. The resulting correlation coefficients hold particular interpretive significance as they systematically capture spatial patterns of association between variables. These coefficients serve as fundamental metrics for identifying co-occurrence trends, validating spatial hypotheses and informing subsequent geospatial modeling procedures. The application of this methodology enables researchers to discern structured relationships within complex spatial systems while accounting for geographic context and scale effects. The correlation coefficient derived from these matrices carries specific interpretive significance:

• A coefficient approaching +1 indicates a strong positive correlation, reflecting similar spatial distribution patterns between variables;
• A coefficient approaching −1 indicates a strong negative correlation, manifesting as divergent spatial distribution patterns between variables.

The calculation formula is defined as follows:

$${Corr}_{ij}={\displaystyle \frac{{\sum }_{k=1}^n\left(v_{ik}-{\mu }_i\right)\left(v_{jk}-{\mu }_j\right)}{(n-1){\sigma }_i{\sigma }_j}}$$

(1)

In this context, ${Corr}_{ij}$ represents the spatial correlation coefficient between the energy consumption distribution and the spatial distributions of PM_2.5, SO₂, NO₂ and CO₂ concentrations, i denotes the spatial distribution dataset of energy consumption, j denotes the spatial distribution dataset of PM_2.5, SO₂, NO₂ and CO₂, μ indicates the mean pixel value of the corresponding raster layer, n represents the number of pixels, $v_{ik}$ signifies the value of the k-th pixel in layer i, $v_{jk}$ signifies the value of the k-th pixel in layer j, σ denotes the standard deviation of the raster layer.

3.3. Decoupling Analysis

Decoupling analysis represents a quantitative framework for assessing the dissociation between economic expansion and resource utilization patterns, with the primary objective of identifying scenarios where economic development occurs independently of proportional increases in resource demand. This analytical approach holds particular significance in sustainability studies, where it is systematically applied to investigate the dynamic interplay between Gross Domestic Product (GDP) growth and critical environmental indicators—notably energy consumption and carbon emissions. By quantifying the elasticity of resource use relative to economic output, decoupling analysis provides empirical evidence for evaluating whether economic growth aligns with environmental sustainability.

Among the various decoupling models, the Tapio decoupling index [27] has emerged as a predominant analytical framework for assessing the dissociation between economic growth and environmental pressures since its introduction. This methodology classifies decoupling states into eight distinct categories, ranging from strong decoupling (absolute decline in environmental pressure amid economic growth) to strong negative decoupling (accelerated environmental degradation with economic expansion), to provide a nuanced typology for policy evaluation. In this study, we employ the Tapio index to systematically examine the spatiotemporal dynamics between energy consumption and atmospheric environmental indicators (e.g., CO₂ emissions and atmospheric pollution). The decoupling index is calculated as

$$I_t={\displaystyle \frac{∆{AE}_t}{∆E_t}}={\displaystyle \frac{({AE}_t-{AE}_{t-1})/{AE}_{t-1}}{(E_t-E_{t-1})/E_{t-1}}}$$

(2)

where

• $t-1$ and $t$ represent the base period and end period, respectively;
• $I_t$ denotes the decoupling index between atmospheric environmental indicators and energy consumption at the end period relative to the base period;
• $∆{AE}_t$ is the change in atmospheric environmental indicators;
• $∆E_t$ is the change in energy consumption;
• $AE$ represents the values of atmospheric environmental indicators;
• $E$ denotes energy consumption values.

Following the methodology of Feng et al. [28], we categorize decoupling states into six distinct types (see

Table 2. The decoupling status and decoupling index between energy consumption and atmospheric environment.

Decoupling Status	Decoupling Types	$∆{\mathit{A}\mathit{E}}_{\mathit{t}}$	$∆{\mathit{E}}_{\mathit{t}}$	${\mathit{I}}_{\mathit{t}}$
Decoupling	Strong Decoupling	≤0	>0	$I_t$ ≤ 0
	Weak Decoupling	>0	>0	$0 < I_t$ < 1
	Recessive Decoupling	<0	<0	$I_t$ ≥ 1
Negative Decoupling	Strong Negative Decoupling	≥0	<0	$I_t$ ≤ 0
	Weak Negative Decoupling	<0	<0	$0 < I_t$ < 1
	Expansive Negative Decoupling	>0	>0	$I_t$ ≥ 1

):

• Strong Decoupling: Represents an optimal scenario where increased energy consumption coexists with reduced atmospheric pollution/carbon emissions;
• Weak Decoupling: Denotes partial decoupling, characterized by energy consumption growth outpacing the growth rate of atmospheric pollution/carbon emissions;
• Recessive Decoupling: Occurs when reductions in atmospheric pollution/carbon emissions exceed the rate of energy consumption decline;
• Strong Negative Decoupling: An undesirable state where energy consumption decreases but atmospheric pollution/carbon emissions increase;
• Weak Negative Decoupling: Reflects concurrent reductions in both energy consumption and atmospheric pollution/carbon emissions, though with smaller environmental improvement relative to energy reduction;
• Expansive Negative Decoupling: Represents both energy consumption and atmospheric pollution/carbon emissions increase, though with environmental degradation outpacing energy demand growth.

All spatially explicit data (energy consumption, atmospheric pollution and carbon emissions) were aggregated to a consistent city-level administrative unit. This involved summing or averaging pixel values within each city’s boundary, transforming the high-resolution raster data into a panel dataset of annual city-level totals or mean concentrations. This ensures that the decoupling index, which compares percentage changes in aggregate city-level variables, is computed on internally consistent spatial units. Data preprocessing, spatial analysis and visualization of results in this study were carried out using the ArcGIS 10.6 software package.

4. Results

4.1. Spatial Correlation Analysis

Figure 2 displays the results of spatial analysis of energy consumption and atmospheric pollutant distributions (PM_2.5, NO₂, SO₂ and CO₂) across the GBA. The visualization reveals distinct spatial coupling patterns: The CO₂ concentrations demonstrated the strongest alignment with energy consumption, with high-value zones exhibiting nearly 80% spatial overlap and parallel development trajectories across urban cores. Similarly, the PM_2.5 concentrations displayed consistency with energy consumption, particularly in the Guangzhou and Foshan where the energy consumption correlates with PM_2.5 hotspots; However, divergence occurred in Shenzhen, Dongguan and Hong Kong where advanced emission control technologies decouple energy use from particulate pollution. In stark contrast, the SO₂ manifests the weakest spatial relationship, with its primary concentration zones in Foshan and Zhaoqing exhibiting northwestward migration patterns contrary to energy consumption gradients. Among all types of pollutants, the NO₂ concentrations emerged as the most spatially congruent, maintaining over 60% co-location with energy-intensive zones across all urban agglomeration. These findings collectively unravel that the relationships between energy consumption, air pollution and carbon emissions in megacity regions were governed by complex interactions between industrial structure, technological adoption, and regional atmospheric circulation patterns, instead of simplistic linear correlations. The observed spatial heterogeneities imply the necessity for differentiated environmental policies tailored to specific urban sub-regions rather than uniform mitigation strategies.

Table 3. The spatial correlation coefficient between total energy consumption and atmospheric environmental factors of PM_2.5 (2000–2020), SO₂ (2013–2020), NO₂ (2008–2020), and CO₂ (2000–2020) in the GBA.

Year	Spatial Correlation Coefficient
	PM2.5	SO2	NO2	CO2
2000	0.2515	\	\	0.4729
2001	0.3156	\	\	0.5829
2002	0.2521	\	\	0.5676
2003	0.3318	\	\	0.5588
2004	0.3485	\	\	0.5690
2005	0.3251	\	\	0.5267
2006	0.3492	\	\	0.4970
2007	0.2855	\	\	0.4436
2008	0.3578	\	0.6975	0.4618
2009	0.2387	\	0.7023	0.4646
2010	0.2982	\	0.6500	0.4186
2011	0.3393	\	0.7053	0.4218
2012	0.3000	\	0.6465	0.4075
2013	0.3512	−0.1621	0.7568	0.4063
2014	0.2035	−0.2574	0.7373	0.4036
2015	0.2920	−0.2428	0.7661	0.4084
2016	0.2892	−0.1341	0.7639	0.4028
2017	0.2112	−0.0943	0.7157	0.3800
2018	0.2846	−0.2547	0.7219	0.3657
2019	0.3434	−0.3815	0.7617	0.3734
2020	0.2116	−0.3814	0.7571	0.3808
Average level	0.2943	−0.2385	0.7217	0.4388

displays the results of spatial correlation analysis that quantifies the spatial relationships between energy consumption and atmospheric pollutants, with distinct temporal coverage for each pollutant. For the concentrations of PM_2.5 and CO₂, the analysis utilized data from 2000 to 2020, revealing considerable spatial alignment with energy consumption patterns in industrial zones like Guangzhou and Foshan, while showing notable deviations in technology-driven regions such as Shenzhen, Hong Kong and Dongguan. The NO₂ concentrations during 2008–2020 demonstrated the strongest correlation with energy consumption among all pollutants, particularly in urban centers with high traffic and industrial activity. On the contrary, the SO₂ concentrations during 2013–2020 exhibited the weakest spatial relationship with energy consumption, with their hotspots showing a northwestward shift that diverged from energy consumption patterns, likely reflecting changes in industrial fuel composition and regional pollution control measures. These findings underscore the complex interplay between energy use and atmospheric pollutants and carbon dioxide, where spatial correlations vary significantly depending on pollutant characteristics, industrial structure and environmental policies. Despite the complexity, the temporal differences in correlations between multiple factors reflect both the evolution of energy efficiency and shifting policy priorities in the region.

The analysis of spatial correlations between energy consumption and atmospheric pollutants reveals distinct patterns across different pollutants in the GBA. The PM_2.5 concentrations demonstrated a moderate relationship with energy consumption, showing only localized overlaps in high-value areas but an overall weakening trend. In the same way, the SO₂ concentrations exhibited no meaningful spatial correlation with energy use, suggesting its distribution was not influenced by non-energy factors like industrial processes or atmospheric chemistry. By comparison, the NO₂ concentrations maintained a strong, stable positive correlation with energy consumption, reflecting its direct linkage to combustion processes in transportation and power generation. In terms of CO₂-energy consumption relationship, the CO₂ concentrations showed a progressive decoupling since 2006, indicative of successful energy efficiency measures and cleaner energy transitions.

4.2. Time Series Analysis

The spatial correlation analysis provides only a partial perspective on the trade-offs in energy efficiency and environmental impact. For a more comprehensive examination of the evolving relationships between energy consumption and emissions of PM_2.5, NO₂, SO₂ and CO₂ in the GBA, decoupling analysis is subsequently applied to quantify the dynamic interactions between the multiple variables.

Figure 3 illustrates the time-varying energy consumption of various sectors and concentrations of PM_2.5, NO₂ and SO₂ in the GBA between 2000 and 2020. While energy consumption (Figure 3a) displayed a consistently upward trend with periodic fluctuations throughout the two decades, the PM_2.5 concentrations (Figure 3b) demonstrated a clear inverted U-shaped pattern, peaking in around 2005 before entering a sustained decline. This divergence in trends reflects a fundamental shift in the region’s development paradigm: the initial phase saw PM_2.5 levels escalated simultaneously with rapid industrialization and surging energy demands, whereas subsequent years of 2005–2013 witnessed a pronounced decoupling between economic growth and air pollution. The period of 2013–2020 particularly exemplified this transformation, as energy consumption maintained its upward trajectory while PM_2.5 concentrations exhibited a significant downward trend. This inverse relationship stresses the cumulative impact of multifaceted interventions including stringent environmental regulations, technological innovations in emission control and systematic improvements in energy efficiency and structural optimization. The observed decoupling phenomenon highlighted the effectiveness of integrated policy measures in mitigating environmental degradation without compromising economic vitality.

In contrast, the NO₂ levels (Figure 3c) followed a messier pattern from 2008 to 2020, during which the energy consumption of the GBA exhibited consistent growth. The first phase of 2008–2013 saw the NO₂ concentrations rose steadily across the region, mirroring the expanding energy needs of rapid industrialization. However, a significant shift occurred after 2013, when NO₂ levels began to decline despite ongoing economic growth and increasing energy consumption. This decoupling trend identifies the effectiveness of emission control measures, including vehicle emission standards and industrial pollution controls, which helped mitigate NO₂ pollution even as energy demand continued to climb. The subsequent period of 2013–2020 was characterized by fluctuating but generally decreasing NO₂ concentrations, indicating that while economic development remained energy-intensive, targeted environmental policies successfully reduced NO₂ emissions.

Similarly, SO₂ (Figure 3d) concentrations demonstrated a clear and continuous decline from 2013 to 2020, during which the energy consumption of the GBA remained stable. This decoupling pattern suggests that despite maintaining consistent energy use density, the GBA successfully implemented effective measures to reduce SO₂ emissions. The sustained decrease in SO₂ levels likely resulted from comprehensive air quality management strategies, including the widespread adoption of cleaner energy sources, stricter industrial emission controls and the phase-out of high-sulfur fuels. Hence, these findings suggest the region’s progress in achieving economic growth while simultaneously improving environmental quality through targeted pollution reduction initiatives.

The rest of Figure 3 presents the temporal dynamics of CO₂ emissions (Figure 3e) in the GBA from 2000 to 2020, revealing a sustained upward trajectory that closely mirrors the region’s growing energy demand. This parallel development indicates the intrinsic linkage between energy consumption and CO₂ emissions, as the expansion of energy-intensive activities directly drives atmospheric carbon output. The consistent positive correlation between these two variables highlights the fundamental challenge of balancing economic development with emission reduction in rapidly urbanizing regions. Notably, while the GBA achieved significant economic growth during this period, the persistent rise in CO₂ emissions advises that current mitigation strategies have yet to fully decouple energy use from carbon output, pointing to the need for more aggressive decarbonization measures in the region’s energy sector.

To elucidate the temporal correlation between energy consumption and emissions of PM_2.5, SO₂, NO₂ and CO₂, the decoupling theory is then applied to quantitatively evaluate the coupling dynamics. The Tapio decoupling index was used to calculate decoupling indices between energy consumption and emissions of each air pollutant and CO₂ across the full study period (2000–2020) and the three sub-periods (2000–2008, 2008–2013 and 2013–2020), as visualized in Figure 4. The decoupling results demonstrate that the GBA achieved significant progress in balancing the relationships between energy consumption, air pollution and carbon emissions. During 2000–2020, PM_2.5 concentrations were effectively reduced across all cities in the GBA despite the growing energy use, resulting in strong decoupling between energy consumption and PM_2.5. Similarly, strong decoupling was observed between energy consumption and NO₂ during 2008–2020 and between energy consumption and SO₂ during 2013–2020. The decoupling relationship between energy consumption and air pollution clearly demonstrated substantial achievements in controlling atmospheric pollutant emissions across the GBA cities. However, the decoupling status of CO₂ emissions and energy consumption is far more complex. During 2000–2020, six cities in the GBA, namely Macao, Dongguan, Guangzhou, Shenzhen, Zhuhai and Zhaoqing, exhibited weak decoupling between CO₂ emissions and energy consumption. Although energy consumption increased in these areas, the growth rate of carbon emissions was relatively lower, showing some degree of decoupling trend. At the same time, five cities in the GBA, namely Foshan, Huizhou, Hong Kong, Zhongshan and Jiangmen exhibited expansive negative decoupling between CO₂ emissions and energy consumption. While both energy consumption and carbon emissions increased in these cities, carbon emissions possessed a higher growth rate than that in energy consumption during 2000–2020.

5. Discussion

5.1. Spatiotemporal Decoupling Characteristics and Evolution of Energy Consumption and Atmospheric Environment

The most crucial finding from the decoupling analysis is that, between 2000 and 2020, the GBA has made significant yet uneven progress in coordinating the relationships among atmospheric pollution, carbon emissions, and energy consumption, demonstrating substantive advancements in environmental governance. This conclusion aligns with existing studies on environmental policy effectiveness in the GBA [29,30,31]. Research indicates that under the context of sustained growth in energy consumption, the concentrations of major atmospheric pollutants have been effectively controlled, and the linkage between energy consumption and atmospheric pollution has achieved significant decoupling. However, the decoupling process between carbon emissions and energy consumption remains relatively lagging and exhibits regional disparities. The decoupling status analysis reveals three distinct phases:

(1)

2000–2008: During this initial phase, the GBA cities initiated their air pollution control measures in response to growing environmental concerns. PM_2.5 exhibited predominantly weak decoupling from energy consumption, reflecting early-stage mitigation efforts. However, the situation regarding carbon emission control remains suboptimal, as the relationship between CO₂ emissions and energy consumption in most cities exhibits an expansive negative decoupling, indicating that the GBA still requires strengthened efforts in carbon emission management.

(2)

2008–2013: Marked by intensified environmental policies, this period saw improved air quality outcomes. While PM_2.5 maintained weak decoupling patterns, some cities achieved strong decoupling as concentrations began declining. NO₂ demonstrated emerging weak decoupling from energy consumption, signaling progress in vehicular emission controls. However, although the GBA also shows a positive trend in carbon emission control, the progress is relatively slow and further efforts are still needed.

(3)

2013–2020: Representing a watershed period, the GBA achieved strong decoupling for all three air pollutants relative to energy consumption, reflecting comprehensive air quality improvements. Notably, carbon emissions entered a declining phase, with the relationship between carbon emissions and energy consumption transitioning from a phase of expansive negative decoupling to one characterized predominantly by weak decoupling in most cities, demonstrating the effectiveness of integrated climate policies.

The decoupling analysis results demonstrate the effectiveness of emission control measures in mitigating air pollution despite rising energy use. The turning point in 2013 underscored the efficiency of air quality management strategies which rely on coordinated allocation of energy consumption and air pollution control within the GBA’s sustainable development framework. The GBA’s success proves that energy consumption growth can be decoupled from emissions, even in rapidly developing regions. While industrialization has been shown to inhibit the decoupling of economic growth from carbon emissions [32], these findings reveal complex and interrelated dynamics between energy consumption and atmospheric environmental quality. Although energy consumption inherently causes air pollution and carbon emissions, measures such as optimizing energy structures and implementing emission reduction policies can significantly mitigate environmental impacts. The observed differences in decoupling relationships between energy consumption and PM_2.5, SO₂, NO₂ and CO₂ may stem from multiple contributing factors including energy structure, emission control technologies, meteorological conditions and environmental policy implementation.

5.2. Energy Structure–Atmospheric Environment Relationship Based on a Representative City (Zhuhai)

To carry out such a case study at city-level presents significant challenges, especially for specific regions where energy consumption structure data are often incomplete or unavailable. Since Zhuhai is the only city within the GBA that has published a complete dataset of its energy consumption structure (2010–2020), this study used Zhuhai city as representative in the GBA to analyze the relationship between energy consumption structure and atmospheric environment. As illustrated in Figure 5, Zhuhai city accelerated the transformation to a clean and low-carbon energy consumption structure during 2010–2020 (Figure 5a). Despite the rapid growth of total energy consumption over this period, the proportion that coal and oil account for remained at a moderate level. Meanwhile, the decade witnessed a significant increase in clean energy consumption, which notably elevated the share of natural gas (from 1.5% to 6.2%) and thermal power (from 0.4% to 3.1%) in the overall energy composition. Over this period, consumption changes in different energy types showed diverse trends. A comparison of energy consumption and atmospheric pollution trends in Zhuhai from 2010 to 2020 reveals that despite stable fossil fuel consumption, significant reductions in air pollution have been achieved (Figure 5b). It is particularly noteworthy that since 2013, concentrations of PM_2.5, NO₂ and SO₂ have exhibited a pronounced downward trend. The case of Zhuhai aligns with the synergistic integration of carbon capture and renewable energy, a pathway frequently proposed to achieve pollution control and carbon reduction [33,34].

5.3. Implications, Research Limitations and Future Directions

This study systematically examines the spatiotemporal relationship among energy consumption, air pollution, and carbon emissions in the GBA, with a particular focus on understanding the impacts of low-carbon energy transition on air quality. By employing the Tapio decoupling model, we analyze the dynamic interactions between these critical environmental factors. Five distinct decoupling scenarios: strong decoupling, weak decoupling, expansive negative decoupling, weak negative decoupling and strong negative decoupling are identified in the decoupling status of energy consumption and PM_2.5, SO₂, NO₂ and CO₂ during 2000–2020. Through a case study of Zhuhai City, we investigate how lean energy adoption and fossil fuel purification technologies contribute to decoupling economic growth from environmental degradation. These findings reveal actionable insights for optimizing energy-environment governance, offering transferable strategies for sustainable development at both regional and global scales. This research holds significant reference value for countries and cities undergoing rapid urbanization and industrialization, providing a methodological framework for balancing economic growth with environmental protection.

However, as with any research, this study is subject to limitations, which provide avenues for future work. For instance, the energy consumption data estimated from nighttime lighting have not been divided into more specific types and sources. Due to the absence of comprehensive energy structure data for 10 out of the totally 11 major cities within the GBA, the energy types, sources, and responsible activities for the cities cannot be accurately identified. This data gap significantly hinders the assessment of the energy consumption dynamics at a city level. To address this limitation, future research will prioritize integrating satellite-based remote sensing data with authoritative statistical datasets such as official yearbooks. In addition, our discussion on how atmospheric environmental conditions can benefit from clean energy deployment and technological advances in traditional fuels has only been based on the data of Zhuhai city. Given that cities in the GBA vary considerably in their economic profiles, industrial structures, and energy resource allocations, the formulation of differentiated emission reduction strategies is essential for each individual city in the region [35]. We hope to thoroughly investigate the temporo-spatial relationship between industry sectoral energy consumption, air pollution and carbon emissions across the GBA urban agglomeration in the future work. It will be helpful for implementing more targeted emission reduction strategies for each city.

6. Conclusions

With spatial correlation and decoupling index analysis, this study thoroughly investigates the complex relationships between energy consumption, atmospheric pollution and carbon emissions in the GBA. All cities across the region have demonstrated significant advancements in tackling atmospheric environmental governance challenges, marking a critical milestone in regional low-carbon transition and collaborative pollution control. The period of 2013–2020 witnessed significant decoupling between energy consumption and major air pollutants (PM_2.5, NO₂ and SO₂), demonstrating substantial progress in air pollution control. In contrast, progress in carbon emission control has been relatively slower, with some cities remaining in expansive negative decoupling states, though showing positive trends, as evidenced by significant improvements in carbon emission control since 2013. To achieve high-quality sustainable development, the GBA shall implement a dual-track governance framework, prioritizing energy structure optimization through accelerated clean energy deployment, while concurrently enforcing stringent carbon emission regulations and fostering innovation in decarbonization technologies. Drawing on this insight, we aim to deliver adaptable policy frameworks for energy optimization, pollution reduction, and carbon regulation, converting the GBA’s experience into actionable blueprints for sustainable development worldwide.