A study of dominant factors in the coupled coordination of population–ecology–energy–digital economy in China based on random forest algorithm

Abstract

This study aims to examine the spatiotemporal patterns and dominant influencing factors of the coupling coordinated development (CCD) among population, ecology, energy, and digital economy (PEED) systems in China, contributing to the broader goal of sustainable regional development. Using panel data from 31 Chinese provinces over the period 2011–2020, we construct a PEED coordination index and analyze its evolution through coupling coordination models, spatial autocorrelation (Moran’s I), the Geodetector model, and a Random Forest algorithm with SHAP analysis. Results show a steady improvement in the overall CCD across provinces, although significant regional disparities persist—eastern provinces such as Guangdong and Beijing lead in coordination, while western and northeastern regions lag behind. Among the four subsystems, the ecological subsystem shows the greatest spatial variation, while the digital economy subsystem is more homogeneous. The Nighttime Light Index, Urbanization Rate, and Green Coverage Rate are identified as the most important drivers, with the Nighttime Light Index consistently exhibiting the strongest influence on CCD. SHAP analysis reveals nonlinear effects of all drivers, highlighting the complexity of subsystem interactions. The findings provide policy-relevant insights for promoting balanced and sustainable development. Policymakers should focus on enhancing urban planning, ecological protection, renewable energy adoption, and digital infrastructure investment, especially in less-developed regions, to further strengthen PEED coordination and support the achievement of Sustainable Development Goals (SDGs).

Introduction

Align with the Sustainable Development Goals (SDGs) outlined by the United Nations, grasping the intricate dynamics among the subsystems of population, ecology, energy, and the digital economy (PEED) plays a pivotal role in the successful realization of sustainable development. The SDGs underscore the imperative to balancing economic growth with environmental sustainability and social inclusiveness. It is now universally acknowledged that realizing sustainable development stands as the most effective approach to advancing economic growth and safeguarding the environment on a global scale¹. China, with its large population and significant ecological challenges, is particularly relevant in this context. Although energy and the digital economy are pivotal in driving economic progress, their expansion in China is often accompanied by severe ecological impacts and environmental pollution^2,3. This issue is particularly pronounced in areas with high population densities. Historically, industries with substantial energy consumption have been the driving forces behind China’s economic expansion. In return, due to its rapid economic growth, China has become a significant global energy consumer and carbon emitter^4,5. This scenario has led to an escalated demand for energy, enduring environmental degradation, and an incongruity between the supply of energy and the tenets of ecological sustainability. This scenario underscores the need for a balanced approach in PEED dynamics to ensure that China’s path toward economic development aligns with ecological conservation and the broader objectives of the SDGs.

The interrelated factors of PEED in China form a complex system that is vital for the country’s sustainable development. Statistics from China’s National Bureau reveal that between 2011 and 2020, the total energy consumption in China grew at an annual rate of approximately 3.6%. Concurrently, there was an average yearly growth rate of 4.6% in carbon emissions. By the year 2023, it is anticipated that China’s total energy consumption will escalate to around 4.7 billion tons of standard coal, predominantly sourced from the consumption of fossil fuels. This trend indicates an increasing demand for energy self-sufficiency⁶. Simultaneously, by 2022, China’s population had grown to 1.418 billion, and the large population base implies that any minor changes in population structure and size can have significant socio-economic impacts^7,8. In parallel, China grapples with increasing environmental concerns. The intensification of industrial activities has not only amplified energy consumption but also escalated environmental pollution, thereby exerting substantial pressure on the country’s ecological well-being. Due to industrialization, China’s energy consumption and environmental pollution have become increasingly serious, posing significant challenges to China’s environment. The intensity of energy usage has demonstrated a positive and statistically significant influence on environmental degradation, particularly in highly developed provinces⁹. Conversely, the swift advancement of the digital economy has injected fresh impetus into China’s economic growth. It has reached an impressive sum of 113.1 trillion RMB, contributing to 38.6% of the GDP. This clearly underscores the critical role of the digital economy¹⁰. Moreover, the “Digital China” strategy has been proven to enhance urban innovation by upgrading industrial structures, and the digital economy has a more substantial effect on substantive innovation in terms of the types of innovation¹¹. Safeguarding the robust and accelerated progression of the digital economy while harmonizing it with demographic trends, environmental conservation, and energy management presents a multifaceted challenge. Consequently, the challenge lies in determining how to foster the swift and beneficial growth of the digital economy while simultaneously aligning it with the demands of the population, environment, and energy sectors. These are complex issues that necessitate comprehensive investigations and analyses. Therefore, deciphering and comprehending the dynamic relationships among PEED subsystems and identifying a pathway toward coordinated and balanced development is crucial for sustainable development.

While existing studies have explored interactions within ecology, economy, and society systems, there remains a critical gap in understanding how the four subsystems of PEED dynamically coordinate at the provincial scale, especially concerning the spatiotemporal heterogeneity and driving factors within the specific context of China’s regional disparities. To address this, the present study focuses on three main questions: (1) How has the coupling coordination among PEED subsystems evolved across provinces from 2011 to 2020? (2) What regional disparities and spatial patterns can be identified in the CCD, and how are they distributed geographically? (3) What are the dominant driving forces of PEED coordination, and how do these factors interact in potentially nonlinear ways? The central aim is to systematically assess the dynamic coupling coordination among population, ecology, energy, and digital economy subsystems (PEED) at the provincial level, and to reveal their spatiotemporal patterns, regional disparities, and driving mechanisms. To achieve this aim, drawing on the foundational concepts of coupling coordination theory, this research targets 31 provincial-level divisions (encompassing municipalities and autonomous regions) within China, covering the decade from 2011 to 2020, constructing a four-dimensional dynamic analysis model from the perspective of PEED. This study utilizes the Technique for Order of Preference by Similarity to the Ideal Solution (TOPSIS) method and the Coupling Coordination Model (CCM) as tools to scrutinize the level of coupling coordination development. Then, Spatial patterns and regional clustering are further explored through Moran’s I index and LISA clustering, while the Geodetector model and Random Forest with SHAP analysis are used to investigate the driving mechanisms and nonlinear influences. This analysis not only provides a comprehensive understanding of the development of different provinces but also offers policy recommendations to promote China’s sustainable development. The data sources include authoritative statistics from national and provincial yearbooks, Nighttime Light Index data, and CO₂ intensity estimates, ensuring empirical reliability and robustness. Compared with previous studies, this paper not only brings the digital economy into the coupling coordination framework as an independent subsystem, but also integrates spatial statistics and machine learning methods to offer a more robust and scalable analytical path for assessing coordinated regional development.

This research makes three key contributions to the literature. First, it innovatively brings together four subsystems into a unified coupling framework, offering a broader perspective beyond traditional tripartite sustainability models. Second, it reveals the spatial heterogeneity and dynamic evolution of PEED coordination in China, emphasizing the asymmetry between coastal and inland regions and identifying the dominant roles of urbanization and environmental quality. Third, the findings provide timely policy implications to inform regional development strategies, such as optimizing resource allocation, enhancing ecological regulation, and balancing digital expansion with sustainable growth.

The rest of the paper is structured as follows: the section "Literature review" reviews related literature; the section "Methods" describes the methodology and data; the section "Results" presents empirical results; the section "Discussion" discusses key insights; and the section "Conclusion" concludes with summary findings and policy recommendations.

Literature review

The relationship between PEED

Recently, the investigation of the interconnections among PEED subsystems has emerged as a crucial area of scholarly interest. The cross-impacts of these areas reveal complex and dynamic system relationships. Early studies predominantly focused on the interactions between population growth and environmental pressures, Orenstein¹², Liddle¹³, and Scovronick et al.¹⁴ studied the internal interactions between population and the environment, finding that population growth may exert pressure on the environment, as increased population and wealth may lead to more energy consumption and environmental destruction. Building upon this foundation, further investigations expanded to the economy-society-environment triad, where scholars such as Lehtonen¹⁵, Strezov et al.¹⁶, and Awais et al.¹⁷ emphasized the integrated assessment of sustainability. They devised an all-encompassing evaluation index to gauge the extent of sustainability in urban development. Meanwhile, some studies began to incorporate energy utilization as an important dimension, recognizing its significant role in shaping environmental outcomes. Agabo et al.¹⁸, Rehman and Rehman¹⁹, and Yousaf et al.²⁰ explored the dynamic causal relationships linking population growth, energy consumption, and environmental preservation, suggesting that increasing population and energy demand jointly contribute to sustained environmental degradation. In a related vein, He et al.²¹ and Ali et al.²² probed into the long-term causal correlations between energy usage, economic expansion, and environmental degradation. Their findings indicated that economic growth spurs both energy consumption and environmental pollution, with energy consumption directly contributing to environmental degradation.

More recently, research attention has extended to the emerging influence of the digital economy. Bai et al.²³, Li and Wang²⁴, and Zhang et al.²⁵ found that the digital economy significantly influences environmental change, and it can alleviate pollution emissions by improving industrial structures, promoting green innovation, and developing finance. Meanwhile, other studies have turned attention to the structural energy-environment nexus, emphasizing marginal, source-based, and sector-level effects. For example, a recent study by Kartal et al.²⁶ applied the KRLS model on high-frequency daily data to uncover that different sources of electricity generation exhibit varying marginal impacts on sectoral CO₂ emissions, implying that the environmental cost of energy consumption in China is highly nonlinear and sector-specific. Similarly, Magazzino²⁷ examined the relationship between electricity use, GDP, and ecological footprint in China, revealing that while electricity consumption and GDP increase environmental pressure, urbanization plays a dual role by promoting both growth and environmental quality under certain conditions. From a green innovation perspective, Solangi et al.²⁸ identified key drivers—economic, environmental, and social—that influence the adoption of renewable energy technologies, underscoring the role of government and industry-led initiatives in aligning growth with sustainability. These studies collectively highlight the complex and dynamic nature of the relationships within the PEED framework.

From the literature review, PEED subsystems interact dynamically, exchanging and reflecting material energy within a spatiotemporal content, thus forming an integrated PEED system. The population influences productivity and consumption demand. The ecological environment’s transformations embody the repercussions of socio-economic endeavors on the natural surroundings, as well as the society’s commitment to safeguarding the environment. Fundamentally intertwined with economic expansion, the availability and efficiency of energy utilization have profound connections to economic growth and pose immediate influences on the ecological environment. A region’s innovation prowess and growth potential are mirrored in the digital economy, an emerging economic development model. Consequently, the interconnected and harmonious evolution of these four subsystems is fundamentally illustrative of humankind’s ongoing efforts and applications in achieving a synchronized development of economy, society, and environment.

Technique for order of preferences by similarity to ideal solution (TOPSIS) research

TOPSIS is a well-known classical method originally proposed for addressing multi-criteria decision-making (MCDM) problems²⁹. The underlying principle of TOPSIS is that the optimal solution is the one closest to the positive ideal solution and farthest from the negative ideal solution, and alternatives are prioritized based on their distances to the ideal solutions³⁰. Owing to its sound mathematical foundation, simplicity, and practical applicability³¹, TOPSIS and its variations have garnered substantial attention from researchers and practitioners across diverse domains³².

Wang et al.³³ constructed an index system that includes criteria for social development, economic benefit, ecological response, and environmental quality. They utilized the variation coefficient TOPSIS method to calculate the evaluated value of each criterion to support the study of the socio-economic and eco-environmental benefits of urban land intensive use. Solangi et al.³⁴ proposed an integrated framework combining the fuzzy analytical hierarchy process and fuzzy technique for order performance by similarity to an ideal solution (F-TOPSIS) to prioritize key strategies for effective sustainable energy technologies adoption in the construction sector. Foroozesh et al.³⁵ proposed a practical framework to assess the development sustainability of Karaj. Based on seven criteria and 24 sub-criteria of influencing factors on sustainable urban development, they developed an integrated model to identify and evaluate sustainable areas, using TOPSIS to rank their priority. Wątróbski et al.³⁶ proposed an innovative sustainability assessment method called the Data vARIability Assessment Technique for Order of Preference by Similarity to Ideal Solution (DARIA-TOPSIS) to evaluate sustainable cities and communities in 26 European countries based on ten criteria encompassing economic, social, and environmental dimensions. Wei et al.³⁷ established an assessment system consisting of 20 indicators for the water resources, economy, society, and ecology subsystems. They utilized the entropy weighted TOPSIS method to assign weights to these indicators and assess the water resource carrying capacity of Henan Province from 2005 to 2021.

In summary, the applications of TOPSIS in sustainability-related fields demonstrate its versatility and effectiveness as a decision-making tool. Once the evaluation system is established, the use of the TOPSIS algorithm allows for the treatment of indicators representing different dimensions of sustainability and enables the computation of the weights of these indices, thus providing valuable insights that help achieve sustainable development goals.

Coupling coordination research

Modern cities can be viewed as complex systems comprised of multiple elements, including population, environment, energy, and economy, which interact and influence each other. CCM emerges as an essential analytical tool for exploring and assessing the degree of coordination among different elements in urban development. By constructing CCM, researchers can quantitatively investigate and analyze the coupling relationships among different elements within urban systems, thereby predicting and assessing crucial factors, bottlenecks, and feasibility in urban development. Therefore, coupling effects and coupling theory have gained attention as fundamental approaches in examining the coordinated relationships among multiple systems.

Existing research has evaluated the coupling coordination degree (CCD) within urban systems from multiple elements, including population, environment, energy, economy, infrastructure, urbanization, agriculture, food, and transportation. Sun and Cui³⁸ proposed a metric system to evaluate the socio-economic, energy, and environmental advantages of public transport infrastructure. They further investigated the coordination level among China’s four primary municipalities from 2006 to 2015, offering a scientific benchmark for urban revitalization. Xing et al.³⁹ employed a system dynamics model to establish a system model integrating economic, resource, and environmental subsystem interactions, revealing the coupling relationships among these elements. Li et al.⁴⁰ analyzed the coordinated development of the social security and economic development systems nationwide from 2002 to 2018, highlighting disparities in regional development. Xiao et al.⁴¹ employed a model that measures the CCD to quantitatively analyze the efficiency of urbanization and land use in China’s Loess Plateau. This insightful approach sheds light on the overall development and coordination within this region. Sun et al.⁴² developed a carbon efficiency evaluation system, considering production, lifestyle, and ecology perspectives, and employed grey relational analysis to calculate the carbon efficiency index and CCD in four representative counties of China. By focusing on the Yellow River Basin as a case study, Feng et al.⁴³ delved into strategies to harmonize the use of water resources, the growth of industries, and the conservation of ecology. Zhong et al.⁴⁴ examined the effects of urban policies on the coordinated development of interconnected systems at the urban level.

In summary, CCM has proven to be valuable tools for quantifying the degree of coordination between different components, providing insights into their relationships within urban systems. Recent methodological work by Solangi and Magazzino⁴⁵ has evaluated the policy and financial implications of renewable energy development in China, proposing a hybrid fuzzy AHP—VIKOR framework to assess multi-dimensional criteria such as risk, economic viability, and subsidy mechanisms—an approach particularly relevant for understanding system-level coordination in sustainable energy planning. However, previous research employing these methodologies has often presented certain limitations. A significant portion of this work has concentrated on the coupling coordination of ecology, economy, and society systems, with comparatively less attention devoted to the integrated analysis of population, ecology, energy, and digital economy (PEED) subsystems. Moreover, existing research mainly focuses on larger and medium scales, such as national, provincial, and specific regional scales, employing a diverse research method, including indicator system methods, CCM, correlation analysis, data envelopment analysis, and system dynamics. These studies lay a foundational groundwork for our research on the interactive development of the PEED system in 31 provincial regions of China. Therefore, conducting an in-depth analysis of the coordinated development of the PEED system in China from a new, broader perspective holds significant value.

Methods

Study area

This research included 31 provinces of China, covering different stages of development (Fig. 1). The selection of such a diverse cross-regional sample offers a distinctive viewpoint for examining the complex interactions among population structure, ecological balance, energy use, and the digital economy. By conducting comparative analyses between economically thriving metropolises such as Beijing and Shanghai and the resource-abundant yet less developed western provinces like Qinghai and Yunnan, this study aims to achieve a profound understanding of the variances and potential linkages among provinces concerning population dynamics, ecological conservation, energy strategies, and digital evolution.

Fig. 1.

Study Area. The map was generated by QGIS 3.16 (https://qgis.org/).

In summary, this study intends to build an all-encompassing structure to fully grasp the interactions among population structure, environmental ecology, energy utilization, and the advancement of the digital economy across diverse provinces in China. It not only contributes to identifying unique and target strategies for addressing environmental challenges, promoting energy transition, and accelerating digitization but also provides policy guidelines for achieving coordinated progress among population growth, ecological protection, energy security, and the digital economy. The promotion of sustainable development, both within the borders of China and on a global scale, carries substantial practical significance.

Methods

Research framework

Figure 2 illustrates the primary stages of the analysis process for evaluating the CCD of the PEED system, as well as for identifying its influencing factors. The framework reflects not only the technical sequence of the analysis but also the conceptual structure linking the four subsystems—population, ecology, energy, and digital economy—which interact in both direct and indirect ways within regional development processes.

Fig. 2.

Research design framework of this research.

In the first stage, this research establishes an indicator system based on relevant literature and standardizes the raw data accordingly. In the second stage, the entropy weight TOPSIS model is used to calculate the weights of the indicators and evaluate the development level of each subsystem. The third stage involves the application of the Coupling Coordination Degree Model (CCDM) to assess the degree of interaction and coordination between the subsystems. Spatial autocorrelation methods are employed to analyze the geographic characteristics and clustering patterns of coordination across provinces.

In the final stage, a set of analytical tools is applied to explore the driving mechanisms of CCD. The Geodetector model is used to identify the explanatory power of different influencing factors. Correlation analysis further clarifies the relationships between those factors and the degree of coordination. In addition, a Random Forest model combined with SHAP analysis is introduced to capture nonlinear relationships and assess the relative importance of variables. Through this comprehensive framework, the study aims to reveal the internal structure and external drivers of regional PEED coordination, offering a basis for formulating targeted strategies that support sustainable development.

Entropy weight TOPSIS model

This study initially standardizes each indicator’s data to examine and contrast their impact on coupling coordination. The entropy weight method is subsequently employed to ascertain the weight of each indicator. Finally, the PEED subsystem’s values are computed using the TOPSIS model. The specific calculation formula is as follow:

1. Standardizing the Data. Given a data matrix , here is the corresponding formula:

   1

   2

In the formula, x_ij represents the value of the jth indicator for the ith sample under evaluation; m and n refer to the number of samples and indicators, respectively; is the minimum value of the jth indicator and is the maximum value of the jth indicator.

• 2)
  Utilizing the entropy weight method, the weights of each indicator are computed using the following formula:

  3

  4

  5

  6

In the formula, for the ith evaluation sample, the weight value of the jth indicator is symbolized by Y_ij; The jth indicator’s information entropy value is represented by e_j; g_j corresponds to the redundancy of information entropy of the jth indicator; The weight coefficient associated with the jth indicator is represented by w_j.

• 3)
  Weighted synthesis calculation, the formula is:

  7

In the formula, S_ij stands as the comprehensive weighted value, W_j acts as the weight corresponding to the jth indicator, and Z_ij represents standardized data.

• 4)
  The TOPSIS model incorporates the calculation of both the positive ideal solution, denoted as , and the negative ideal solution, denoted as . The formulas used to determine these are as follows:

  8

  9

In the formula, represents the collection of positive indicators, while denotes the collection of negative indicators.

• 5)
  Calculate the distance () between each decision-making unit and the positive ideal solution, as well as the distance () between it and the negative ideal solution. The formula is:

  10

  11
• 6)
  To garner a comprehensive assessment of each decision-making unit, it’s crucial to calculate their relative closeness (C_j) to the ideal solution. The formula is:

  12

The value of C_j lies within a spectrum of 0 to 1. As Cj approaches 1, it implies a superior overall assessment for each respective decision-making unit.

Coupling coordination model

This study calculates the CCD through the CCM, which has been widely adopted in interdisciplinary sustainability research. This model provides a quantitative framework to measure both the interaction strength and the coordination status among multiple subsystems. The method is particularly suitable for assessing the degree to which subsystems such as population, ecology, energy, and digital economy evolve in harmony over time and space. Compared with simple correlation-based assessments, the CCM accounts for both structural interdependence and developmental consistency, making it appropriate for analyzing complex socio-economic systems^42,43,46. The specific calculation formula is as follows:

13

14

15

In formulas (13)–(15), D represents the CCD (0 ≤ D ≤ 1); The degree of coupling among the PEED systems is symbolized by C, within a range of 0 to 1. A greater value of C implies a more robust correlation among the subsystems, resulting in more cohesive development. Prior research has categorized the degree of coupling into five different phases, as shown in Table 1. E₁ means the population system’s value; E₂ means the ecology system’s value; E₃ means the energy system’s value; E₄ means the economy system’s value. T means the coordination value, reflecting the overall synergistic effect of subsystems. The parameters a, b, c, and d correspond to the contribution rates of the population, ecology, energy, and digital economy systems, respectively, each set at a value of 1/4. The categorization of the CCD, as outlined in Table 2, draws from the distribution function suggested in earlier research⁴⁷.

Table 1.

Coupling stage classification.

Coupling degree(C)	Stage
(0,0.3]	Low level coupling
(0.3,0.5]	Antagonistic Stage
(0.5,0.8]	Running-in phase
(0.8,1.0]	High level coupling
1	Benign resonant coupling

Table 2.

Coupling coordination classification.

D	Category	Level	D	Category	Level
0–0.09	Extreme disorder	1	0.5–0.59	Barely coordination	6
0.1–0.19	Serve disorder	2	0.6–0.69	Primary coordination	7
0.2–0.29	Moderate disorder	3	0.7–0.79	Intermediate coordination	8
0.3–0.39	Mild disorder	4	0.8–0.89	Good coordination	9
0.4–0.49	Verge disorder	5	0.9–1.0	Excellent coordination	10

Exploratory spatial data analysis

To explore the spatial characteristics of PEED coordination across provinces, this study employs two classical spatial statistical tools: Global Moran’s I and Local Moran’s I. These indicators are chosen for their ability to detect overall spatial autocorrelation as well as identify local spatial clusters and outliers. This dual approach allows us to reveal both macro-level patterns and micro-level heterogeneities in regional coupling coordination. It is particularly useful in understanding whether coordinated development is spatially clustered or randomly distributed, which provides a spatial basis for further policy differentiation⁴⁸. The specific calculation formula is as follows:

The Global Moran’s I index. In the formula (16), is the spatial weight between spatial units i and j; is the sum of all spatial weights:; n denotes the aggregate number of spatial units, the attribute values of spatial units i and j are represented by and , respectively. The mean attribute value across all spatial units is represented by , and the value of the spatial weight is denoted by . The formula is:

16

The Local Moran’s I index. In the formula (17), ,, is the variance of the variable y across all spatial units, defined as , The local Moran’s I index of the ith region is symbolized by I_i. The value known as the spatial weight, w_ij, is also of importance. The corresponding formula is presented below:

17

Geodetector model

The Geodetector model is employed to identify the explanatory power of various potential influencing factors on regional coupling coordination. This method is particularly well-suited to this study: first, it is non-parametric and does not require linear assumptions, which aligns with the complex and potentially nonlinear relationships within the PEED system; second, it effectively quantifies how much of the spatial variation in CCD can be attributed to individual factors or their combinations. The model’s ability to reveal hidden spatial drivers and assess their relative importance makes it a valuable tool in policy-relevant spatial analysis^49–51. As this method does not rely on regression assumptions, diagnostic tests such as residual normality or multicollinearity are not applicable. The specific calculation formula is as follows:

18

In the formula: q stands for the impact of dynamic factors. The stratification of these dynamic factors is denoted by h. The count of units in both layer h and the entire area are symbolized by N_h and N, respectively. The variance of y values in the h layer and the complete area are also important, each represented by and respectively. q can hold any value within the range of [0, 1]. A higher value of q signifies a more potent explanatory power of dynamic factors on the CCD.

Random forest model and SHAP analysis

To further investigate the driving mechanisms of PEED coordination and validate the influence of key variables, this study applies a Random Forest (RF) model combined with SHAP (SHapley Additive exPlanations) analysis. The RF model is chosen for its robustness in handling high-dimensional data and nonlinear interactions among variables, making it well-suited to explore the complex dependencies within the PEED system. Unlike traditional regression models, RF can capture intricate, nonlinear relationships without overfitting. To enhance interpretability, SHAP is used to decompose the model predictions and assess the marginal contribution of each factor to the CCD outcome. This combined approach not only ensures prediction accuracy but also provides transparent and explainable results, which is crucial for policy insights and decision-making. Since the model is non-parametric, no conventional diagnostic tests are required. Meanwhile, in this study, the Random Forest model used default parameter settings as recommended by standard practice, as the model already achieved high explanatory performance without hyperparameter tuning.

Indicators and data sources

Indicator of PEED systems

This research establishes an extensive indicator system that facilitates the dynamic exploration of the coupling coordinated development of PEED. Based on coupling coordination theory and prior studies, this research meticulously chose a set of representative indicators to capture the key characteristics of each subsystem and grasp the context of their mutual interactions. The selection of each subsystem’s indicators is based on their ability to fully reveal the characteristics and dynamic changes of each system.

Indicator System for the coupling coordinated development of PEED is shown in Table 3. Specifically, in the population subsystem, indicators such as year-end permanent population, birth rate, and employed persons in urban unit reflect demographic scale, labor force availability, and social capacity—factors widely recognized as influencing resource demand and economic momentum^46,52,53. These indicators have been frequently included in spatially disaggregated CCD studies due to their direct relevance to regional development disparities.

Table 3.

Indicator System for the coupling coordinated development of PEED.

Subsystem	Indicator	Units of Measurement	Property
Population	Year-end Permanent Population	Ten thousand people	–
	Birth Rate	%	+
	Natural Population Growth Rate	A thousandth of a percent	+
	Employed Persons in Urban Units	Ten thousand people	+
	Urban Registered Unemployment Rate	%	–
Ecology	Sulfur Dioxide Emissions	Ten thousand tons	–
	Particulate Matter, Smoke (Dust) Emissions	Ten thousand tons	–
	Sewage Treatment Rate	%	+
	Solid Waste Comprehensive Utilization Rate	%	+
	Forest Coverage Rate	%	+
	Per Capita Park Green Space Area	Square meter	+
Energy	Electricity Generation	Billion kilowatt-hours	+
	Per Capita Total Supply of Oil and Gas	Cubic meters per person	+
	Per Capita Natural Gas Supply	Cubic meters per person	+
	Electricity Consumption	Billion kilowatt-hours	–
	Gas Popularization Rate	Tons of standard coal per ten thousand yuan	+
Digital Economy	Mobile Phone Penetration Rate	%	+
	Number of Internet Broadband Access Ports	Ten thousand pieces	+
	Online Mobile Payment Level	/	+
	Digital Finance Coverage Breadth	%	+
	Degree of Digitalization in Digital Finance	%	+

For the ecological subsystem, sulfur dioxide emissions and particulate matter (smoke and dust) emissions are chosen to represent the major sources of atmospheric pollution, which directly impact environmental health. On the positive side, indicators such as sewage treatment rate, solid waste comprehensive utilization rate, forest coverage rate, and per capita park green space area are included, as they collectively portray environmental management capacity, green space development, and ecological restoration efforts. These indicators are standard in regional sustainability assessments and have demonstrated explanatory power in coordination studies involving environmental constraints^{39,41,42,44,53}.

In the energy subsystem, electricity generation, per capita total supply of oil and gas, per capita natural gas supply, and gas popularization rate are employed to reflect the production capacity and accessibility of major energy sources, electricity consumption is incorporated to assess energy demand and consumption pressure. These variables are not only core to energy-environment-economic analyses^54–56, but also well-suited for integration into coupling models such as CCM and Geodetector due to their spatial and temporal variability.

Regarding the digital economy system, the mobile phone penetration rate and the number of internet broadband access ports are selected to measure digital infrastructure development. Online mobile payment level, digital finance coverage breadth, and the degree of digitalization in digital finance further capture the extent of digital financial inclusion and the deepening impact of the digital economy on socioeconomic activities^25,57,58. Their inclusion in the model allows the study to account for emerging forms of development that were often absent in earlier CCD studies. Additionally, the “Property” column in the table indicates whether the indicator has a positive or negative impact on the coupling coordinated development.

Overall, the indicator system is designed not only to reflect subsystem performance but also to enable multi-dimensional evaluation through entropy-weighted TOPSIS, spatial modeling, and machine learning-based estimation. The selected indicators have been validated in prior literature both individually and in integrated models of regional coordination, ensuring the robustness and interpretability of subsequent analytical results.

Indicators of driving factors for CCD

To comprehensively assess the external factors influencing coupling coordinated development, this study selected six indicators drawn from previous literature (Table 4): Gross Domestic Product (GDP), Nighttime Light Index, Urbanization Rate, Green Coverage Rate, Energy Consumption Intensity, and CO₂ Emission Intensity. Specifically:

Table 4.

Indicators for Analyzing the Driving Factors of CCD.

Type	Name	Unit of measurement	Code
Explained Variable	Coupling Coordination Degree	/	D
Explanatory variable	GDP	billions of yuan	X1
	Nighttime Light Index	/	X2
	Urbanization Rate	%	X3
	Green Coverage Rate	%	X4
	Energy Consumption Intensity	Tons of standard coal per ten thousand yuan	X5
	CO2 Emission Intensity	Tons per ten thousand yuan	X6

GDP and nighttime light index. These two economic activity indicators reflect the scale of China’s economy and provide real-time economic activity information, providing the critical means to measure the economic development level ⁵⁹. The specific calculation formula is as follows: In the formula, ANLI represents the Average Nighttime Light Index; DN_i represents the brightness value of the ith pixel in the region; n represents the total number of pixels in the region.

Urbanization rate. This indicator not only reveals the changes in the population structure and distribution, but also closely relates to economic development, ecological environment, and resource consumption. It unveils the interplay between economic growth and dynamic population changes^41,46.

Green coverage rate. As a measure of ecological quality, this indicator reflects the health conditions of the ecosystems and the expansion of green spaces, which are crucial for evaluating the functions of the ecosystem and its contribution to human well-being⁶⁰.

Energy consumption intensity and CO₂ emission intensity. These two indicators are related to energy utilization efficiency and environmental pressure. They hold substantial importance in the assessment of energy policy results and the effectiveness of environmental conservation, thus providing a crucial outlook on the capacity for sustainable growth in a region^6,21.

Data source

This study focuses on 31 provincial-level administrative units in China, including municipalities and autonomous regions, covering the period from 2011 to 2020. The administrative boundary data are obtained from the National Basic Geographic Information System (http://ngcc.sbsm.gov.cn). All indicators for the PEED subsystems and external driving factors are derived from official statistical publications, including the China Statistical Yearbook, China City Statistical Yearbook, China Energy Statistical Yearbook, and the Statistical Bulletin of National Economic and Social Development. When possible, data were accessed through the National Bureau of Statistics website (https://data.stats.gov.cn), ensuring source consistency and reliability. For each year, we used the corresponding edition of the yearbooks to maintain temporal accuracy. Measurement units for all indicators are provided in Tables 3 and 4, and data were used in their original form unless normalization or entropy-weighting was required as part of the evaluation process.

To prepare the data for analysis, we standardized all variables where necessary, applying min -max normalization prior to the TOPSIS evaluation to allow comparability across dimensions. Missing data, which accounted for a very small proportion of the dataset, were treated using linear interpolation based on adjacent-year values within each province. No cross-provincial estimation or extrapolation was conducted. These procedures were implemented using SPSS 22.0, following established practices for dealing with incomplete panel data in applied regional studies⁶¹. Since all data are obtained from publicly available statistical yearbooks and national databases, and the full indicator system and preprocessing steps have been described, the analysis is fully reproducible using standard tools and documented procedures. In addition, all indicators were visually and statistically screened for outliers, and no extreme values were found that required removal or adjustment. Finally, in this study, spatial analysis was conducted using GeoDa 1.14 and ArcGIS 10.4; entropy-TOPSIS and CCD computations were performed in Excel; Random Forest and SHAP analyses were implemented in Python 3.8.

Results

Temporal and spatial characteristics of general coupling coordination

Temporal characteristics of general coupling coordination

The time series analysis of this study reveals that from 2011 to 2020, the coupling degree and coordination level of provincial systems in China have been continuously and steadily rising (Table 5 and Fig. 3). Table 5 presents the average coupling degree (C) and coordination degree (D) of the PEED system from 2011 to 2020. Each year’s coupling degree and coordination degree are listed alongside their corresponding development stages and types. In 2011, the system was in the running-in stage, with a near-disorder coordination degree. Then, both coupling and coordination degrees improved, moving the system to the high-level coupling stage. In 2014, the system reached primary coordination. Until 2020, despite a slight decrease in the coupling degree, the coordination degree continues to improve, suggesting overall progress in coordinated development. This trend reflects China’s significant progress in various fields, such as economic growth, technological innovation, population policies, environmental protection implementation, and digital transformation. Particularly after 2012, provinces transitioned into a high-level coupling phase, and from 2014 to 2020, they all maintained a primary coordination state, validating the effectiveness of the national strategy of coordinated development.

Table 5.

Average Coupling Degree and Coordination Degree of PEED System from 2011 to 2020.

Year	C	D	Stage	Type
2011	0.774	0.498	Running-in Stage	Near disorder
2012	0.845	0.551	High-level Coupling	Barely coordination
2013	0.874	0.585	High-level Coupling
2014	0.901	0.607	High-level Coupling	Primary coordination
2015	0.920	0.621	High-level Coupling
2017	2016	0.925	0.657
2018	0.937	0.674	High-level Coupling
2019	0.943	0.686	High-level Coupling
2020	0.930	0.693	High-level Coupling

Fig. 3.

Comprehensive and Detailed Results of China’s CCD.

Specifically, due to their economic advantages and policy foresight, Guangdong and Beijing lead in coupling coordination degrees. Conversely, Heilongjiang and Tibet show relatively lower performance, likely due to their unique geographical, economic, and policy implementation conditions.

An intriguing observation is a “fluctuating” growth over the years in various provinces showcased by the CCD. From 2011 to 2013, there was general growth in CCD driven by economic and technological policies. From 2014 to 2017, there was a significant "W-shaped" fluctuation, implying periodic challenges in policy implementation and economic structural adjustments. From 2018 to 2020, nearly half of the provinces experienced a decline in coupling coordination degrees, possibly related to adjustments in environmental and energy policies and changes in the global economic situation.

Spatial characteristics of general coupling coordination

Spatiotemporal evolution analysis reveals significant spatial differences in the CCD of 31 provinces in China (Figs. 4 and 5). The majority of the provinces display a lack of coordination, while the eastern provinces exhibit a more elevated CCD in contrast with central and western areas. In 2011, only Guangdong was in good coordination, which changed to Xinjiang, Ningxia Hui Autonomous Region, Shaanxi, Sichuan, Jiangsu, Anhui, Hunan, Jiangxi, Fujian, and Guangxi by 2020. Over the previous ten years, the Chinese government has vigorously promoted new urbanization, green development, and digital transformation, which has somewhat driven the improvement of the CCD of the provinces^46,58. Compared to 2010, the number of provinces with good coordination and above in 2020 changed significantly, especially the high-quality coordinated provinces, which increased from 0 to 4, namely Beijing, Shandong, Zhejiang, and Guangdong.

Fig. 4.

Spatial Characteristics of China’s CCD.

Fig. 5.

Spatial Characteristics of China’s Coupling Coordination Level.

Temporal and spatial characteristics of coupling coordination in four PEED subsystems

This study conducts a thorough assessment of the development status of the PEED subsystems across China’s 31 provinces from 2011 to 2020 (Fig. 6). By constructing an evaluation indicator system and conducting visual analysis, this study analyzes the spatiotemporal evolution characteristics and regional disparities of each subsystem. This analysis also emphasizes the influence of policy changes on population structure changes, ecological environment improvement, energy efficiency enhancement, and digital economy development.

Fig. 6.

Time series results of subsystem index. POI Population Subsystem Index, ECI Ecology Subsystem Index, ENI Energy Subsystem Index, and DII Digital Economy Subsystem Index.

Temporal characteristics of coupling coordination in four subsystems

Population subsystem. The evaluation result indicates an overall upward trend, especially between 2014 and 2017. This trend correlates strongly with the adjustment of national population policy (National Population Development Plan (2016–2030)). However, the population subsystem has been low for a long time in some economically underdeveloped provinces with severe population outflow (such as Heilongjiang, Jilin, and Liaoning), indicating a need for targeted demographic and economic policies by the government to promote development in these provinces.

Ecology subsystem. Significant improvement has been observed since 2015, reflecting China’s efforts on bolstering ecological civilization and the efficacy of legal reforms. This improvement is particularly prominent in provinces like Guangdong, Chongqing, Zhejiang, and Beijing, indicating these areas’ emphasis on environmental protection and sustainable growth.

Energy subsystem. The generally low-performance level indicates the overall inefficiency of China’s energy framework and its substantial reliance on fossil fuels. However, provinces such as Xinjiang, Sichuan, and Beijing have shown significant improvements, which are related to their efforts in diversified energy structure and renewable energy utilization.

Digital economy subsystem. While generally low in performance level, substantial growth has been witnessed in provinces like Guangdong, Zhejiang, Jiangsu, and Beijing since 2015. This growth may be related to the implementation of the "Internet + " planning, illustrating the potential for comprehensive economic and social innovation driven by the Internet.

Moreover, the study revealed that the ecology subsystem attained the highest performance level, whereas the digital economy subsystem registered the lowest. The disparity in the performance of the ecology subsystem was the most pronounced across different provinces, in contrast to the digital economy subsystem, which demonstrated the least variation.

Spatial characteristics of coupling coordination in four subsystems

Population subsystem. The study found a conspicuous “concave” spatial pattern, i.e., characterized by lower values in the central regions and higher values in the surrounding areas, as shown in Fig. 7. Over the past decade, there has been an increase in the number of provinces with a high population value. This trend suggests positive developments in China’s population subsystem, which may be influenced by government policies such as promoting high-quality urbanization, enhancing public services, and optimizing population distribution^62,63.

Fig. 7.

Spatial characteristics of coupling coordination in population subsystem. The map was generated by QGIS 3.16 (https://qgis.org/).

Ecology subsystem. The results reveal regional disparities, with the northwest exhibiting poorer environmental quality compared to the southeast (Fig. 8). Between the years of 2011 and 2020, there was a decline in provinces with high ecology subsystem values, decreasing from 14 to 13. Similarly, provinces with low values also experienced a dip, falling from 11 to 8. However, the study found that the overall ecology subsystem has improved in recent years, which may be attributed to the Chinese government’s "Thirteenth Five-Year" planning policy⁶⁴.

Fig. 8.

Spatial characteristics of coupling coordination in ecology subsystem. The map was generated by QGIS 3.16 (https://qgis.org/).

Energy subsystem. There is a trend of value diffusion from the central to the north (Fig. 9). Between the years of 2011 and 2020, there was a decline in provinces with high energy subsystem values, decreasing from 7 to 10. This may result from the intensity of industrial activities in the northern regions and the reliance on coal and other fossil fuels⁶⁵.

Fig. 9.

Spatial characteristics of coupling coordination in energy subsystem. The map was generated by QGIS 3.16 (https://qgis.org/).

Digital economy subsystem. The findings suggest that coastal regions showcase a more advanced digital economy than the border regions (Fig. 10). This may be due to the more developed infrastructure construction in coastal areas compared to border areas. However, it is noteworthy that the digital economy in border areas has also shown an increase over the past decade.

Fig. 10.

Spatial characteristics of coupling coordination in digital economy subsystem. The map was generated by QGIS 3.16 (https://qgis.org/).

Spatio-temporal patterns of coupling coordination using spatial autocorrelation analysis

According to Moran’s I index by spatial autocorrelation analysis (Fig. 11), there is an improvement from 0.294 in 2011 to 0.307 in 2020, proving the enhanced coordination between these four systems and showing the clustering trend of similar provinces.

Fig. 11.

Moran’s I of China’s CCD from 2011 to 2020.

Based on the LISA cluster maps depicted in Fig. 12, there was an increase in the number of provinces falling into the high-high (HH) cluster, rising from a single province in 2011 to six by 2020. In 2011, only Fujian Province was in the high-high (HH) cluster, indicating its excellent performance in the coordinated development of the four subsystems. This is attributed to Fujian’s consistent dedication to sustainable development and the promotion of ecological civilization⁶⁶. Since then, Fujian Province has consistently remained in the high-high (HH) cluster throughout the decade, demonstrating its continuous advantage in the coordinated development of the four subsystems. By 2020, the high-high (HH) cluster had expanded to six provinces, including Jiangsu, Anhui, Zhejiang, Jiangxi, Fujian, and Guangdong. This is likely a result of these provinces’ robust economic development, abundant resources, clear policy orientation, and notable achievements in digitalization, environmental sustainability, and new energy development^67–69.

Fig. 12.

LISA cluster results of CCD from 2011 to 2020. The map was generated by QGIS 3.16 (https://qgis.org/).

On the contrary, the low-low (LL) cluster predominantly occupies the northern and western areas of China, encompassing the trio of Sichuan, Yunnan, and Inner Mongolia provinces. These regions may face challenges in the coordinated development of the four subsystems due to factors such as their remote geographical location, lack of resources, lower level of economic development, and unclear policies⁵⁴.

Driving factors of coupling coordination

It was found that the effect of driving factors on CCD varies over different periods (Table 6). The Nighttime Light Index consistently exhibits substantial impacts, with an average value of 0.432, highlighting its importance in coupling coordination across the decade. The Urbanization Rate and CO₂ Emission Intensity showed significant impacts in the earlier years, but their influence on CCD has weakened and is no longer significant in recent years. GDP also showed some significance in 2013 and 2016, but its impact is weaker compared to the other significant factors.

Table 6.

Yearly Impact Results of Driving Factors.

	X1	X2	X3	X4	X5	X6
2011	0.383	0.492*	0.637**	0.256	0.271	0.528*
2012	0.337	0.474*	0.562*	0.293	0.304	0.639**
2013	0.473*	0.566**	0.525**	0.4	0.193	0.531*
2014	0.377	0.497**	0.435	0.4	0.296	0.545
2015	0.446	0.468*	0.264*	0.446*	0.192	0.367
2016	0.497*	0.319	0.247	0.596*	0.321	0.319
2017	0.514	0.373	0.247	0.457	0.266	0.278
2018	0.449	0.426*	0.226	0.496	0.185	0.39
2019	0.435	0.358*	0.366	0.487	0.216	0.455
2020	0.487	0.346	0.264	0.495*	0.393	0.349
Average	0.440	0.432	0.377	0.433	0.264	0.440

X1 GDP, X2 Nighttime Light Index, X3 Urbanization Rate, X4 Green Coverage Rate, X5 Energy Consumption Intensity; X6 CO₂ Emission Intensity.

*P < 0.05; **P < 0.01.

Based on the number of years it is significant, the Nighttime Light Index (NLI) ranks first among these factors, followed by the Urbanization Rate. While the Nighttime Light Index (NLI) can be a useful indicator for observing CCD development trends, it is important to recognize that it is an outcome of various underlying factors. According to Xiao et al.⁵⁹, nighttime light growth is affected by technological innovation, industrial structure, population growth rate, openness to the international world, and resource consumption. Moreover, the study by Chen et al.⁷⁰ shows that nighttime light is a new alternative for digital economy evaluation. Therefore, policymakers should not focus solely on increasing the NLI as an end in itself; efforts should be directed toward attracting talent, improving the environment, developing the digital economy, etc. As for the Urbanization Rate, efforts need to be made in improving the population structure^71–73.

Not only does the Nighttime Light Index significantly influence as a standalone factor, but it also holds a substantial impact when amalgamated with other factors (Fig. 13). The innovation of this study lies in revealing the interactive relationships between driving factors, finding that the interaction effects of these factors are intensifying. For instance, the first four main interactive factors are X3 ∩ X4, X2 ∩ X3, X2 ∩ X6, X1 ∩ X2. This offers a fresh viewpoint for comprehending and forecasting the CCD for distinct subsystems.

Fig. 13.

Interaction Results. X1 GDP; X2 Nighttime Light Index; X3 Urbanization Rate; X4 Green Coverage Rate; X5 means Energy Consumption Intensity; X6 CO₂ Emission Intensity.

Correlation analysis

Using Origin software, the study conducted a correlation analysis between the system’s CCD and various driving factors, as shown in Fig. 14. The results show that the CCD of the four systems is significantly positively correlated with GDP, Nighttime Light Index, Urbanization Rate, and Green Coverage Rate, and significantly negatively correlated with CO₂ Emission Intensity. Among them, the Nighttime Light Index significantly influences the coupling coordination, demonstrating the highest correlation with an average correlation coefficient of 0.54; the second is Urbanization Rate, with an average correlation coefficient of 0.52; the Nighttime Light Index and CO₂ Emission Intensity have opposite correlations with the coupling coordination, with similar absolute values of the correlation coefficient, both at 0.41; GDP has the weakest correlation with the coupling coordination, with an average correlation coefficient of 0.32.

Fig. 14.

Correlation Analysis Results. X1 GDP; X2 Nighttime Light Index; X3 Urbanization Rate; X4 Green Coverage Rate; X5 Energy Consumption Intensity; X6 CO₂ Emission Intensity.

Through the Geodetector model and correlation analysis, it can be determined that the Nighttime Light Index, Urbanization Rate, CO₂ Emission Intensity, and Green Coverage Rate are actually the most significant influencing factors.

Random forest and SHAP analysis

In this section, we select the driving factors X1, X2, X3, X4, and X6, which were proven significant in the previous part, and employ a random forest model to reveal the mechanisms influencing the evolution of coupling coordination across the decade. The final result of R² in the random forest model is 0.954, indicating strong model interpretability and a good fit with the data.

Significance and relative importance of feature variables

Figures 15 and 16 illustrate the significance, relative importance, and density distribution of influencing variables on CCD. It can be observed (see Table 7) that five influencing variables significantly affect CCD. Among them, X3, contributing 33.60%, ranks first. This highest rank indicates that X3 has the most significant impact on coupling coordination changes. Furthermore, the distribution of X3 indicates that lower values have a significantly greater negative impact, while higher values contribute more positively, with a broader range of negative values than positive values. Secondly, X4 (30.95%) and X1 (28.74%), representing green coverage rate and GDP, respectively, have the second and third highest relative importance. Both generally show a tendency where higher values contribute positively, while lower values have a negative impact. Finally, X2 (28.59%) and X6 (28.56%) have relatively moderate explanatory power with similar magnitudes. X2 shows a rather chaotic pattern around the SHAP value of 0, with low and high values appearing in a concentrated range. However, X6 mainly exhibits a negative correlation with CCD changes, with low values leading to positive SHAP values and high values leading to negative SHAP values.

Fig. 15.

The result of the RF model (*** means P < 0.001). X1 GDP, X2 Nighttime Light Index, X3 Urbanization Rate, X4 Green Coverage Rate, X6 CO₂ Emission Intensity.

Fig. 16.

The feature effects of the SHAP model. X1 GDP; X2 Nighttime Light Index; X3 Urbanization Rate; X4 Green Coverage Rate; X6 CO₂ Emission Intensity.

Table 7.

Relative Importance of Factors with Significant Effects

Variable	X3	X4	X1	X2	X6
%IncMSE	33.60%	30.95%	28.74%	28.59%	28.56%

X1 GDP, X2 Nighttime Light Index, X3 Urbanization Rate, X4 Green Coverage Rate, X6 CO₂ Emission Intensity.

Nonlinear relationships of influence forces of CCD changes

Figure 17 illustrates the individual dependence of the five driving factors in the model. These dependence plots allow for detailed observation of each factor’s contribution to the model output at different values, providing a better understanding and interpretation of the relationship between independent variables and the dependent variable. The X-axis displays the attribute values of these factors, while the Y-axis shows the SHAP values corresponding to these attribute values. The findings reveal a clear nonlinear relationship between the selected influencing factors and the changes in CCD.

Fig. 17.

Partial Dependence Plots of Key Factors. X1 GDP; X2 Nighttime Light Index; X3 Urbanization Rate; X4 Green Coverage Rate; X6 CO₂ Emission Intensity.

As shown in Fig. 17, at lower values of X1 (0–15,000), there is a negative correlation between X1 and CCD. As the value of X1 increases, the SHAP values generally rise, indicating that higher values of X1 have a more positive impact on the model output. X4 and X6 also exhibit similar patterns, both displaying a V-shaped trend. At lower X4 values, SHAP values are negative. As X4 values increase, a noticeable fluctuation in the middle range (approximately 35–45) indicates variability in impact within this range. The SHAP values then continue increasing, showing a positive impact on the model output with higher attribute values. For X6 values from 0 to 2, SHAP values decrease rapidly. This indicates that in this range, X6 significantly negatively impacts the model output. Similarly, the SHAP values rise again afterward. The relationship between X3 and CCD first shows a clear monotonic increase and then stabilizes in the latter range (80 to 90), indicating that higher values of X3 have a more positive impact on the model output. For X2 values from 0 to 5, SHAP values increase rapidly and peak, indicating that in this range, X2 has a notable impact on the model output. However, as X2 values increase, this impact diminishes, and SHAP values start to decline and level off.

Discussion

Overview and interpretation of key findings

In this study, the development trend and spatial differences of the systemic CCD across various provinces in China have been explored. It was found that with the Chinese government’s policies aimed at comprehensive coordinated development, sustainable development, and innovation-driven development, the CCD has shown a steady upward trend. Especially from 2012 onward, the coupling degree transitioned into a high-level stage, and the coordination degree shifted to primary coordination after 2013. This trend likely correlates with the government’s introduction of a renewed cycle of scientific and technological innovation policies, in conjunction with rural revitalization strategies^54,64,65. However, as per this study, it’s also crucial to note that provinces in the eastern region tend to surpass their northern and western counterparts when it comes to coupling coordination. This phenomenon is also observed in the studies by Shi et al.⁷⁴ and Song et al.⁷⁵, with the former focusing on the coordination between economic development and the ecological environment, and the latter examining the coordination between low-carbon development and urbanization. This discrepancy can be attributed to variations in regional development, resource allocation, and policy directions. For instance, the eastern region was quicker to adopt and enact a series of liberalization policies and reformative actions. This strategic decision led to a considerable migration of population and capital, thus fostering speedy economic growth^67,69. These regional imbalances suggest that while national-level strategies are effective in elevating coordination overall, their impacts differ substantially across provinces depending on local conditions, governance capacities, and development baselines. Furthermore, the performance of the four subsystems is not uniform-population and ecology show more balanced improvements, whereas digital economy and energy subsystems exhibit stronger regional divergence, indicating differentiated policy effectiveness.

The variable importance results offer more than a statistical ranking-they reflect structural dynamics behind coordinated development. The Urbanization Rate, as the most influential factor, points to the centrality of spatial form and service accessibility in shaping subsystem linkages. Yet, its positive impact depends on planning quality rather than urban expansion alone. The Green Coverage Rate reflects long-term ecological investment and institutional commitment, showing that environmental capital supports broader systemic balance. Interestingly, GDP plays a relatively weaker role, suggesting that economic scale without structural depth or governance efficiency may not translate into better coordination. The Nighttime Light Index, as a proxy for digital and infrastructural vitality, shows moderate but policy-relevant influence-its effects depend on whether digital growth is accompanied by ecological safeguards. Meanwhile, the negative role of CO₂ Emission Intensity confirms that high-emission regions face barriers to integration, reinforcing the need for cleaner energy transitions to support systemic coordination.

This study also found that the CCD of PEED is influenced by policy. The improvement in the population subsystem may be attributed to adjustments in population policy and new urbanization plans^8,53; the advancements in the ecology subsystem correspond with the development of ecological civilization alongside policy modifications^41,58; the status of the energy subsystem reveals issues with the efficiency of China’s energy structure, yet some provinces have made improvements through diversification of energy structure and adoption of renewable energies⁵⁵; and advances in the digital economy subsystem are in line with the goals of the government’s Internet + action plan²⁵.

Comparison with existing studies and international patterns

To better situate these findings within a broader international context, it is useful to compare China’s PEED coordination with patterns observed in other economies. In OECD countries, particularly in Western Europe, the integration of environmental sustainability and digital transformation has developed over several decades, supported by institutional continuity and cross-sectoral policy frameworks⁷⁶. These countries often demonstrate more balanced subsystem coordination than emerging economies, where structural disparities remain pronounced. In India and Brazil, for example, recent studies highlight persistent challenges such as unequal energy access, regional ecological pressure, and fragmented digital development, which resemble those observed in China’s central and western regions⁷⁷. Spatially, the east–west gap in China’s coordination level is comparable to patterns found in the United States, where coastal states interior states due to early policy adoption and stronger innovation capacity. Furthermore, the identification of the Nighttime Light Index and Urbanization Rate as key driving factors aligns with global empirical research. Nighttime light data, for instance, have become a standard proxy for economic and infrastructural development in regions with limited official statistics, especially in Sub-Saharan Africa and Southeast Asia⁷⁸. Similarly, studies in South Korea and Indonesia have shown that rapid urbanization often precedes ecological coordination, creating temporary imbalances between growth and sustainability goals⁷⁹. These international parallels validate the theoretical assumptions of this study and further demonstrate that China’s experience reflects broader structural tensions commonly found in rapidly developing economies undergoing multidimensional transitions.

Moreover, Empirical results reveal that the Nighttime Light Index, Urbanization Rate, CO2 Emission Intensity, and Green Coverage Rate, identified as the primary driving factors influencing CCD, exhibit varying impacts on CCD across different periods. These findings are consistent with existing literature^80–85, but this study further emphasizes the importance of the Urbanization Rate in coupling coordination.

Policy implications

Based on the analysis of the driving factors, policymakers can formulate decisions to regulate CCD from the following aspects. The nighttime light index, as a new alternative for evaluating the digital economy, should be increased by concentrating on technological innovation, industrial structure, population growth rate, openness to the international world, and resource consumption. As for the Urbanization Rate, efforts need to be made in improving the population structure. Regarding the Urbanization Rate, efforts need to be made to improve the population structure.

In particular, regions with persistently low performance in energy or ecology subsystems should be supported through differentiated policy tools such as green finance mechanisms, regional ecological compensation, and adaptive energy subsidies. The eastern provinces, while generally advanced, must guard against digital-resource imbalance and overconcentration of development, which can widen interprovincial gaps. It is also important to promote inter-regional collaboration—for example, digital economy spillovers from advanced provinces can be enhanced through infrastructure sharing, digital training initiatives, and open data policies. Finally, coordination mechanisms between departments (e.g., natural resources, technology, urban planning) should be institutionalized to ensure integrated policy delivery across PEED dimensions.

Limitations and future research

This study has several limitations that should be acknowledged. Although the indicator system covers key dimensions of population, ecology, energy, and the digital economy, it does not fully capture all aspects of regional development. Institutional, cultural, and behavioral factors may significantly influence subsystem coordination but were excluded due to data availability. Additionally, while the entropy-TOPSIS and coupling coordination models provide a structured and interpretable framework, they remain static in nature and do not account for feedback loops or delayed effects between subsystems. The explanatory variables are also limited to six core drivers; other potentially relevant factors such as fiscal capacity, educational attainment, or interregional migration patterns could be considered in future studies. These limitations may affect the granularity and generalizability of the findings, particularly in highly localized contexts or under dynamic policy shifts.

Building on the current framework, future research can expand in several directions. First, the dataset could be updated beyond 2020 to include more recent developments, especially those related to the COVID-19 recovery period and China’s carbon neutrality commitments. Second, the analytical model could incorporate dynamic methods—such as panel vector autoregression (PVAR), system dynamics modeling, or agent-based simulation—to capture nonlinear interactions and temporal feedback among subsystems. Finally, the scale of analysis can be refined to include municipal or county-level data, which would provide more precise insights into spatial coordination. Comparative studies with other emerging economies—particularly those with similar regional disparities such as India or Indonesia—could further enhance the international relevance of the PEED framework.

Conclusion

This study explores the spatiotemporal evolution and driving mechanisms of the coupling coordinated development (CCD) among population, ecology, energy, and digital economy (PEED) subsystems across 31 provinces in China between 2011 and 2020. The findings indicate that although the overall level of coordination has steadily improved, significant regional disparities remain. Eastern provinces exhibit stronger systemic integration, while central and western provinces face persistent challenges, particularly in energy structure and digital economy development. Among the four subsystems, the ecological and population dimensions show more balanced progress, whereas digital and energy subsystems reflect greater spatial inequality.

The empirical analysis identifies the Nighttime Light Index, Urbanization Rate, Green Coverage Rate, and CO₂ Emission Intensity as major drivers of PEED coordination, with the Nighttime Light Index emerging as the most influential across both time and space. These findings suggest that economic activity intensity, urban form, and environmental quality are more decisive for systemic coordination than traditional macroeconomic indicators such as GDP. Based on these insights, this study offers targeted policy suggestions, emphasizing differentiated strategies across provinces, with a focus on improving spatial equity, infrastructure accessibility, and digital inclusion.

This research contributes to the scientific community by integrating digital economy into the traditional population-environment-energy framework and applying a combination of spatial statistics, Geodetector, and machine learning methods to reveal dynamic drivers of coordination. It enriches existing literature by offering a multidimensional, data-driven approach to regional sustainable development, and provides a replicable framework for other emerging economies facing similar structural imbalances. Despite certain limitations in data scope and model dynamics, the study lays a foundation for further research on coordinated development pathways under differentiated spatial contexts.

Abbreviations

CCD

Coupling coordinated development

PEED

Population, ecology, energy, and digital economy

SDGs

Sustainable Development Goals

CCM

Coupling Coordination Model

Author contributions

Conceptualisation: J.Z. and Q.Z.; methodology: Q.Z. and J.Z.; formal analysis: Q.Z. and J.Z.; resources: J.Z. and Q.Z.; writing—original draft preparation: Q.Z., J.Z., and Z.L.; writing—review and editing: Q.Z., J.Z., and Z.L.; visualisation: Q.Z. and J.Z.; supervision: J.Z. All authors have read and approved the final manuscript.

Funding

National Natural Science Foundation of China, 42301537.

Data availability

The data that support the findings of this study are available from the corresponding author, [Jiawei Zhu], upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval and consent to participate

Not applicable.

Consent for publication

Not applicable.