AHP–Entropy Method for Sustainable Development Potential Evaluation and Rural Revitalization: Evidence from 80 Traditional Villages in Cantonese Cultural Region, China

Abstract

Scientific assessment of sustainable development potential (SDP) and analysis of spatial heterogeneity mechanisms of traditional villages are crucial for promoting the synergy between cultural heritage conservation and rural revitalization strategies. With an emphasis on traditional villages in the Cantonese region, this study develops a thorough evaluation methodology that combines spatial analysis and multi-criteria decision-making. It aims to (1) systematically reveal the spatial differentiation characteristics of sustainable development potential; (2) develop and validate a combined weighting method that effectively integrates both subjective and objective weights; and (3) identify key driving factors and their interaction mechanisms influencing the formation of this potential. To achieve these objectives, the research sequentially conducted the following steps: First, an evaluation indicator system encompassing socioeconomic, cultural, ecological, and infrastructural dimensions was developed. Second, the Analytic Hierarchy Process and the Entropy Weight Method were employed to calculate subjective and objective weights, respectively, followed by integration of these weights using a combined weighting model. Subsequently, the potential assessment results were incorporated into a Geographic Information System, and spatial autocorrelation analysis was applied to identify agglomeration patterns. Finally, the Geographical Detector model was utilized to quantitatively analyze the explanatory power of various influencing factors and their interactions on the spatial heterogeneity of potential. The main findings are as follows: First, the sustainable development potential of traditional Cantonese villages exhibits a significant “core–periphery” spatial structure, forming a high-potential corridor in the Zhongshan–Jiangmen–Foshan border area, while peripheral areas generally display “low–low” agglomeration characteristics. Second, the combined weighting model effectively reconciled 81.0% of case discrepancies, significantly improving assessment consistency (Kappa coefficient above 0.85). Third, we identified economic income (q = 0.661) and ecological baseline (q = 0.616) were identified as key driving factors. Interaction detection revealed that the interaction between economic income and transportation accessibility had the strongest explanatory power (q = 0.742), followed by the synergistic effect between ecological baseline and architectural heritage (q = 0.716), highlighting the characteristic of multi-factor synergistic driving. The quantitative and spatially explicit evaluation framework established in this study not only provides methodological innovation for research on the sustainable development of traditional villages but also offers a scientific basis for formulating regionally differentiated revitalization strategies. The research findings hold significant theoretical and practical importance for achieving a positive interaction between the conservation and development of traditional villages.

1. Introduction

Traditional villages represent a vital spatial heritage of Chinese civilization, embodying millennia of regional wisdom in culture, architecture, and ecology [1,2]. Their revitalization is crucial for preserving historical continuity and cultural identity. However, these villages face multifaceted pressures from urbanization, including depopulation, cultural erosion, environmental degradation, and economic stagnation. These challenges are particularly acute and complex within the Cantonese Cultural Region (CCR). Population loss exhibits a dual outflow pattern toward both urban centers in the Greater Bay Area and overseas communities, reflecting the region’s historical role as an emigration hub. Cultural disruption threatens core societal foundations—the Cantonese language and clan systems. Ecological degradation is most evident in the pollution of the Lingnan water town river networks and the collapse of traditional circular agricultural models. Furthermore, industrial stagnation persists despite the region’s advanced economy, indicating reliance on low-end development paths [3,4]. Therefore, a scientific assessment of their SDP and the formulation of differentiated revitalization strategies are essential. This approach is necessary to prevent homogeneous redevelopment, safeguard this unique cultural specimen-defined by its dialect, clan structures, and pro-water ecology and activate endogenous village dynamics, constituting a critical “Guangfu Solution” for balancing conservation with development.

Evaluation methodologies for traditional villages have evolved from unidimensional material heritage conservation to multidisciplinary, integrated assessment systems (As shown in Figure 1). Early studies primarily focused on static indicators such as architectural morphology and historical value. With the advancement of sustainable development concepts, frameworks increasingly incorporate dynamic elements, such as ecological resilience, social networks, and economic vitality [5,6]. The integration of Geographic Information Systems (GIS) and spatial analysis tools has become a research focus. For example, Xu et al. (2018) systematically assessed flood risk spatial patterns using GIS-AHP integration, identifying topographic relief and drainage density as key drivers [7]. Zyoud et al. (2017) highlighted through bibliometric analysis that synergistic AHP-GIS applications are a growing trend in regional planning [8]. Quantitative tools like information entropy have also been introduced. Li et al. (2022) established a shallow geothermal energy evaluation system using the entropy weight method, demonstrating advantages in measuring resource synergy efficiency [9]. However, existing methods often rely heavily on either expert-driven subjective weighting or purely objective data-driven approaches, leading to a significant imbalance between subjective and objective weighting and difficulty in capturing the complexity and dynamics of cultural–ecological–economic–social systems in traditional villages [10].

The Analytic Hierarchy Process (AHP) and Entropy Weight Method (EWM) offer distinct advantages for cultural heritage assessment through integrated weighting. Compared to common hybrid methods (e.g., AHP-TOPSIS or entropy-TOPSIS), the AHP-EWM combination better addresses the multidimensional complexity of cultural heritage by reconciling subjective expertise with objective data dispersion. AHP effectively incorporates expert knowledge and qualitative indicators, such as cultural value, via hierarchical structures. Dos Santos et al. emphasized that AHP can quantify implicit factors such as “social capital” and “cultural identity,” although its subjective weights require consistency checks to mitigate bias [11]. Conversely, EWM objectively assigns weights based on data dispersion, making it suitable for quantifying explicit metrics like economic and infrastructure indicators. The synergy of AHP-EWM uniquely balances qualitative–cultural and quantitative–economic dimensions, avoiding the oversimplification risks associated with TOPSIS-based approaches. Chen (2020) dynamically evaluated eco-economic synergy in building material supply chains using an entropy-AHP-TOPSIS model, identifying transport costs and carbon emissions as highly weight-sensitive factors [12]. Recent studies have begun combining these methods to enhance precision. Wu et al. (2022) [10] applied an AHP–entropy model in Poyang Lake Basin flood risk assessment, demonstrating an 18.3% error reduction compared to single-methods approaches [10].

For CCR villages’ genetic features—such as overseas Chinese economic networks—AHP-EWM provides a systematic framework to integrate clan-based social capital (weighted by AHP) and remittance-driven economic data (weighted by EWM), In contrast, TOPSIS methods lack tailored mechanisms to address such nonlinear interactions. Particularly within complex “ecology–culture–economy” systems [10]. Nevertheless, existing research primarily focuses on current-state diagnosis, with insufficient validation for potential predictive scenarios and a lack of systematic frameworks specifically designed for the unique genetic features of CCR villages [6,13].

Addressing these gaps, this study focuses on 80 traditional villages within the CCR. As illustrated in the research workflow (Figure 2), we first conducted a bibliometric analysis to identify key evaluation indicators, which were subsequently refined through Delphi expert consultation. Building on this foundation, we developed an integrated four-dimensional evaluation system for sustainability potential—encompassing the geographical environment, socio-cultural context, historical heritage, and industrial economy—using a combined AHP and EWM.

To enhance the predictive applicability of the framework, the multidimensional indicator system serves as a proxy for future resilience. For example, the number of external routes and the degree of industrial diversification directly reflect adaptive capacity to socioeconomic changes, while ICH continuity quantifies key cultural capital essential for long-term sustainability. Furthermore, spatial autocorrelation analysis, including Global Moran’s I and Local LISA methods, was employed to uncover spatial differentiation patterns and clustering characteristics of sustainability potential, as well as to identify its core driving factors. We hypothesize that significant spatial agglomeration of sustainable development potential exists among Cantonese villages, exhibiting a core–periphery structure, with primary drivers closely tied to socio-cultural attributes and infrastructure conditions. The expected outcomes not only offer a robust and replicable quantitative tool for assessing the sustainability of traditional village but also facilitate precise decision-making for regional cultural heritage conservation and rural revitalization strategies through the identification of spatial patterns.

2. Materials and Methods

2.1. Study Area

2.1.1. Geographical and Cultural Delineation of the CCR

The study area is located within the continental portion of the Canton Cultural Region (CCR), which is widely recognized as the core area of Lingnan civilization. The broader CCR encompasses the Pearl River Delta and its adjacent hill lands, a region that culturally and historically includes the Special Administrative Regions of Hong Kong and Macau. However, this study specifically focuses on the following eight prefecture-level cities within Guangdong Province: Guangzhou, Foshan, Zhaoqing, Jiangmen, Zhongshan, Zhuhai, Shenzhen, and Dongguan (Figure 3). The delineated study area lies approximately between latitudes 21°27′~24°24′ N and longitudes 112°25′~114°38′ E.

Bordered by the Nanling Mountains to the north and the South China Sea to the south, this region features a convergence zone where the Xijiang, Beijiang, and Dongjiang rivers form an intricate network of waterways and deltaic plains. These hydrological conditions support unique agroecosystems, notably the “Sangji Fishponds”, aquaculture with mulberry cultivation, and the “Mountains-Waters-Fields-Settlements” complex, demonstrating adaptive village planning.

Characterized by a subtropical monsoon climate with a mean annual temperature of 21–23 °C and distinct wet and dry seasons, the CCR frequently experiences frequent typhoons that have influenced vernacular architectural innovations. These include wok-ear roof structures designed for storm resistance, oyster-shell mortar walls for humidity control, and elevated arcades for flood mitigation. As the historical hub of the Maritime Silk Road, the region exemplifies cultural hybridization in its built environments-evident in ancestral hall complexes featuring “three chambers and two corridors” layouts, watchtowers that combine defensive and residential functions, and educational institutions that merge Chinese and Western decorative motifs. Intangible cultural practices flourish through Cantonese opera’s distinctive vocal system, lion dance traditions rooted in martial arts, and intricate embroidery techniques such as Cantonese and Guangfu embroidery. This dual system of natural and human adaptation positions the CCR as a critical case for analyzing the civilizational evolution of Lingnan.

2.1.2. Village Selection and Spatial Distribution

A total of 80 traditional villages were selected for this study from the eight aforementioned cities, using a stratified and purposive sampling strategy to ensure representativeness across cultural, socioeconomic, and geographical dimensions (Figure 4). All selected villages are officially designated and listed in the List of Chinese Traditional Villages (from the first to the sixth batch) by the Chinese government, recognizing their exceptional historical, cultural, and architectural value.

The selection was designed to comprehensively capture the region’s diversity. The eight cities were chosen to represent the full socioeconomic spectrum of the continental CCR, ranging from global megacities (e.g., Guangzhou, Shenzhen) and advanced urban agglomerations to underdeveloped rural and mountainous areas (e.g., in Zhaoqing), thereby reflecting diverse developmental pressures and conservation contexts.

Furthermore, the village selection intentionally included the three predominant cultural subtypes within the region:

(1)

Core Guangfu villages, which represent the heartland of Lingnan culture and are known for their distinctive architectural styles.

(2)

Hakka villages, often located in peripheral hilly areas, are known for their collective fortified structures.

(3)

Overseas Chinese hometowns (Qiaoxiang), especially common in Jiangmen and Foshan, exhibit diasporic influences through architectural hybridity-such as watchtowers-and cultural practices.

2.2. Research Framework and Data Sources

The research framework (Figure 5) systematically integrates four methodological phases. First, evaluation indicators encompassing the geographical environment, socio-cultural context, historical heritage, and industrial economy dimensions were identified through a dual process of meta-literature analysis and Delphi expert consultations. Subsequently, an AHP–entropy integrated weighting model was applied to calculate comprehensive indicator weights, harmonizing subjective expertise with objective data dispersion. Subsequently, village-level datasets were analyzed using spatial autocorrelation and geographic detector methods to quantify regional development disparities and identify core driving mechanisms. Finally, evidence-based revitalization strategies—including adaptive zoning protection and cultural-tourism symbiosis pathways—were developed through policy benchmarking and multi-criteria decision analysis, establishing a replicable framework for sustainable village development.

2.2.1. Establish a System of Indicators

To systematically review existing research findings and extract evaluation indicators, this study employed a systematic review methodology, emphasizing process reproducibility and transparency. The literature search utilized the China National Knowledge Infrastructure (CNKI) and the Web of Science (WOS) Core Collection as data sources. The search period spanned from 1 January 2019, to 31 March 2025, with the final data extraction conducted on 10 April 2025.

In CNKI, the following search string was constructed using the professional search mode:

(SU = ‘村落’ OR SU = ‘乡村’ OR SU = ‘村庄’) AND (SU = ‘发展潜力’ OR SU = ‘可持续发展’ OR SU = ‘可持续性路径’ OR SU = ‘活化利用’ OR SU = ‘乡村振兴’ OR SU = ‘发展’ OR SU = ‘潜力’ OR SU = ‘价值’ OR SU = ‘驱动因素’) AND (SU = ‘评估’ OR SU = ‘评价’ OR SU = ‘指标体系’ OR SU = ‘保护策略’)

This preliminary search retrieved 1587 academic publications, including journal articles, conference reports, and dissertations.

In WOS, two distinct search strategies were employed:

Search Strategy 1: TS = ((rural OR countryside OR “traditional village*”) AND (“sustainable development” OR sustainability) AND (assessment OR evaluation) AND (indicator* OR index OR “index system”)) NOT WC = (“Medicine” OR “Biochemistry & Molecular Biology” OR “Chemistry” OR “Materials Science” OR “Physics” OR “Neurosciences” OR “Psychology” OR “Nursing” OR “Veterinary Sciences”) AND DT = (“Article” OR “Proceedings Paper” OR “Review”)

Search Strategy 2: TS = ((rural OR countryside OR “traditional village*”) AND (“sustainable development” OR sustainability) AND (multifunction* OR value) AND (assessment OR evaluation) AND (indicator* OR index OR “index system”)) NOT WC = (“Medicine” OR “Biochemistry & Molecular Biology” OR “Chemistry” OR “Materials Science” OR “Physics” OR “Neurosciences” OR “Psychology” OR “Nursing” OR “Veterinary Sciences” OR “Dentistry” OR “Pharmacology & Pharmacy”) AND DT = (“Article” OR “Proceedings Paper” OR “Review” OR “Book Chapter” OR “Case Report” OR “Book”)

The search was further refined by subject categories (encompassing Geography, Ecology, Urban Studies, Economics, Sociology, etc.) and publication date range to comprehensively capture research related to rural sustainable development evaluation.

Literature screening strictly adhered to the following three exclusion criteria, and a PRISMA flow diagram (Figure 6) was drawn to enhance transparency:

Exclusion of studies that focus exclusively on a single dimension, such as settlement spatial restructuring or land use optimization.

Exclusion of sector-specific analyses, including non-comprehensive assessments like power infrastructure or homestay development.

Screening out non-systematic research, that is, literature lacking a multi-element (e.g., industry, culture, population) collaborative evaluation framework.

Finally, 69 valid publications in Chinese and English were included, forming the corpus for subsequent analysis.

To objectively extract evaluation factors and reduce lexical redundancy and subjective bias, the Term Frequency–Inverse Document Frequency (TF-IDF) weighting algorithm was employed to quantify term importance [14]. The TF-IDF value is a statistical measure used to assess the significance of a specific word within a particular document or a collection of documents. The underlying principle is that the higher the frequency of a word appearing in a single document (TF) and the lower its frequency across the entire document collection (DF), the greater its discriminative power and importance of that word to the current document. The specific process was as follows: First, the full texts of the 69 publications were preprocessed and segmented to build an initial term library. Subsequently, the weight of each term was calculated using the TF-IDF algorithm according to the formula:

$$\mathrm{TF}\text{-}\mathrm{IDF}(t,d,D)=TF(t,d)\times \log \left({\displaystyle \frac{N}{DF(t)}}\right)$$

(1)

where t denotes a term, d a document, D the corpus, N the total number of documents, and DF(t) the number of documents containing the term t. A weight threshold (TF-IDF value ≥ 3.0 and term frequency ≥ 11) was set to identify 23 high-frequency core terms (

Table 1. Evaluation Factor Corpus.

Word	TF-IDF	Word	TF-IDF
distance	0.040	villager	0.012
facility	0.027	proportion	0.012
area	0.022	density	0.012
scale	0.020	town	0.011
travel	0.020	history	0.011
using land	0.019	environment	0.011
develop	0.018	population	0.010
resource	0.016	slope	0.010
coverage	0.015	Infrastructure	0.010
revenue	0.015	disposable	0.010
village	0.014	scenic area	0.009
Cultivated land area	0.013	Population density	0.009
building	0.013	scene	0.009
transportation	0.012	feature	0.009

), constituting a preliminary set of evaluation factors.

Addressing the limitations in multidimensional integration identified in previous research, this study further conducted theory-guided semantic clustering of the aforementioned high-frequency terms based on the man-land areal system theory and the triple bottom line principles of sustainable development. This process resulted in the formation of four core dimensions: geographical environment, socio-cultural, historical culture, and industrial economy. Building upon this, the study systematically referenced the national standard “Traditional Village Evaluation Indicator System” and specific targets closely related to rural development within the United Nations Sustainable Development Goals (SDGs) (e.g., SDG 11: Sustainable Cities and Communities, SDG 8: Decent Work and Economic Growth) to standardize, integrate, and semantically extend the initially extracted terms, resulting in a preliminary evaluation framework comprising 10 criterion layers.

To further enhance the scientific rigor and consensus degree of the indicator system, the Delphi Method was employed for multi-round screening and optimization of the preliminary indicators. A Delphi expert panel was established, comprising 12 experts from fields such as urban and rural sociology, This panel included eight scholars from universities and research institutions, two experts from provincial-level or higher departments of agriculture, rural affairs, and housing and urban-rural development, and two senior planners with extensive experience in village preservation and development. Additionally, five resident representatives from different traditional villages participated. Through two rounds of anonymous consultation, indicators were scored and reviewed based on criteria such as importance, operability, representativeness, and relevance to the SDGs. Indicators with an agreement level below 80% or significant divergence of opinion were modified or eliminated. The second round of consultation focused on converging opinions on the revised indicator set, ultimately finalizing a clearly structured, consensus-based set of 34 indicators (see

Table 2. SDP Evaluation Table and Its Hierarchical Structure.

Dimension	Theme	Metric	SDGs
A1 Geographical Environment	B1 Transportation Accessibility	C1 Distance to Administrative Center (−)	SDG 11
		C2 Distance to Transportation Node (−)	SDG 9
		C3 Distance to Scenic Area Linkage (−)	SDG 11
		C4 Number of External Routes (+)	SDG 9
	B2 Ecological Base	C5 Topographic Position Index (−)	SDG 15
		C6 Ecological Sensitivity Level (−)	SDG 15
		C7 Water Coverage Rate (+)	SDG 6
		C8 Geohazard Risk (−)	SDG11
A2 Socio-cultural Context	B3 Policy Empowerment	C9 Rural Honorary Titles (+)	SDG 1
		C10 Subsidy Amount (5-year) (+)	SDG 1
	B4 Population Structure	C11 Population Scale (+)	SDG 3
		C12 Population Aging Rate (+)	SDG 3
	B5 Public Facilities	C13 Basic Education Facilities (+)	SDG 4
		C14 Medical Coverage (+)	SDG 3
		C15 Cultural-Sports Facilities (+)	SDG 11
A3 Historical Heritage	B6 Settlement History	C16 Settlement Antiquity (+)	SDG 11
		C17 Cultural Richness (+)	SDG 11
		C18 Settlement Pattern Integrity (+)	SDG 11
		C19 Natural Integration Degree (+)	SDG 11
	B7 Architectural Heritage	C20 Building Antiquity (+)	SDG 11
		C21 Building Rarity (+)	SDG 11
		C22 Regional Architectural Features (+)	SDG 11
	B8 ICH Inheritance	C23 ICH Rarity (+)	SDG 11
		C24 ICH Diversity (+)	SDG 11
		C25 ICH Continuity (+)	SDG 11
		C26 ICH Inheritors (+)	SDG 11
A4 Industrial Economy	B9 Industrial Foundation	C27 Arable Land Resources (+)	SDG 2
		C28 Forest Resources (+)	SDG 15
		C29 Agricultural Modernization (+)	SDG 2
		C30 Industrial Diversification (+)	SDG 9
	B10 Economic Income	C31 Gross Industrial Output (+)	SDG 9
		C32 Agricultural Output (+)	SDG 8
		C33 Disposable Income (+)	SDG 1
		C34 Per Capita GDP (+)	SDG 8

). Each indicator is accompanied by a clear definition and its source in literature or standards, ensuring its operability and measurability (See Appendix A). To assess the potential multicollinearity among the 34 finalized indicators, a linear regression analysis was conducted. The Variance Inflation Factor (VIF) for each indicator was calculated, with all values were found to be below 5. Additionally, tolerance levels were consistently above 0.2. These results confirm the absence of significant multicollinearity, indicating that the selected indicators are statistically independent and suitable for subsequent comprehensive evaluation.

2.2.2. Data Sources

This study utilizes multi-source data, primarily consisting of geospatial data and information on Guangfu traditional villages. All data sources are described in detail, including their origins, resolutions, processing methods, and access permissions, to ensure the reproducibility and scientific rigor of the research.

(1)

Geospatial Data

NDVI Data: The vegetation index data were obtained from the National Aeronautics and Space Administration (NASA) Earth Data platform https://earthdata.nasa.gov (accessed on 25 March 2025). The core dataset is the MOD13Q1 Version 6.1 product, acquired by the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the Terra and Aqua satellites. This dataset has a spatial resolution of 250 m and a temporal resolution of 16 days. A monthly maximum value composite (MVC) method was employed to further minimize noise from clouds and atmospheric effects, generating a monthly NDVI dataset covering the period from January 2024 to May 2025. The MOD13Q1 product incorporates internal atmospheric corrections (e.g., for Rayleigh scattering and aerosols) and cloud masking algorithms. Data access complies with the NASA Earth Data usage policies, and the dataset can be cited via its platform [DOI:10.5067/MODIS/MOD13Q1.061].

Digital Elevation Model (DEM) Data: The topographic data were obtained from the Shuttle Radar Topography Mission (SRTM) SRTM DEMUTM 90 m resolution digital elevation data product. The original data were jointly measured by NASA and the National Geospatial-Intelligence Agency (NGA), and were accessed for this study through the Geospatial Data Cloud platform http://www.gscloud.cn (accessed on 26 March 2025). This dataset has a spatial resolution of 90 m and was used to extract topographic factors such as elevation and slope for the study area.

Other geospatial data: including vector base map data such as administrative boundaries of Guangdong Province, major rivers, and railways, were sourced from the Resource and Environment Science and Data Center of the Chinese Academy of Sciences http://www.resdc.cn (accessed on 25 March 2025). Land use data were obtained from the Geographic Monitoring Cloud Platform http://www.dsac.cn (accessed on 25 March 2025).

(2)

Guangfu Traditional Village Data

The research subjects are traditional villages in the Guangfu region, selected from the List of Chinese Traditional Villages jointly published by the Ministry of Housing and Urban-Rural Development of China and other relevant departments. A total of 80 villages across eight prefecture-level cities in Guangdong Province were ultimately identified as the study sample. Multidimensional attribute information for these villages (including historical evolution, architectural heritage, folk culture, industrial economy, population structure, and policy implementation) was acquired through the following channels:

A systematic review was conducted of local statistical yearbooks, academic publications, government planning documents, and official archives from provincial, municipal, and district-level archives and planning bureaus.

Field Investigations: Field surveys and key informant interviews were carried out in a subset of the sample villages. This primary data was used to cross-validate information from literature sources, mitigating potential issues inherent in public data such as delays in updates, fragmentation, or inaccuracies, thereby ensuring data reliability.

The data sources used in this study are summarized in

Table 3. Summary of data sources.

Data Category	Data Content	Primary Source	Spatial/Temporal Resolution
Core Spatial Data	NDVI	NASA Earth Data (MODIS MOD13Q1 V6.1)	250 m/16-day
Digital Elevation Model (DEM)	Geospatial Data Cloud (Source: SRTM)	90 m/Static	SRTM DEMUTM Product
Auxiliary Spatial Data	Admin. Divisions, Rivers, Railways	RESDC, CAS	Vector/Static
Land Use	Geographic Monitoring Cloud Platform	Unspecified/Multi-Temporal	Used for background analysis
Research Object Data	Guangfu Village Information	List of Chinese Traditional Villages, Statistical Yearbooks, Academic Literature, Planning Docs, Fieldwork	Village-scale/Multi-Temporal

.

2.3. Methods

2.3.1. Standardization

All factors are first divided into positive and negative factors. Then they were standardized in the range between 0 and 1 to eliminate the dimensional influence between different factors [15]. For positive factors, the higher value, the higher SDP, and for negative factors, the higher value, the lower SDP. The positive and negative factors are normalized by Equation (2) and Equation (3), respectively.

$$Y_{ij}={\displaystyle \frac{x_i-x_{\min }}{x_{\max }-x_{\min }}}$$

(2)

$$Y_{ij}={\displaystyle \frac{x_{\mathit{max}}-x_i}{x_{\max }-x_{\min }}}$$

(3)

where $Y_{ij}$ is the normalized value and it is range from 0 to 1, $x_i$ is the original value of the $j_{th}$ factor, $x_{max}$ is the maximum of the $j_{th}$ factor, $x_{min}$ is the minimum values of the $j_{th}$ factor.

Data Translation Operation: Since the normalized data may contain zero values, directly computing logarithms would cause mathematical issues. Therefore, a translation operation is applied to the normalized data $y_{ij}$, defined as $z_{ij}$ = $y_{ij}+ϵ$, where ϵ is a minimal positive constant (e.g., 0.0001 or 1 × 10⁻¹⁰). This operation ensures all values are strictly greater than zero.

2.3.2. AHP–Entropy Hybrid Weighting

The AHP method was first created in 1971 by Professor Thomas L. Saaty of the University of Pittsburgh. The AHP method reduces the complex problem system to a clear element system. Entropy is a measure of the information disorder degree of a specific system and could fully exploit the information contained in original data [10] Entropy weight can indicate useful information provided by the index although it excessively relies on the objective data that cannot reflect the knowledge and practical experience of experts, such that the results are occasionally inconsistent with reality and individual understanding. Therefore, a method integrating the AHP and the Entropy Weight Method was proposed, in which matrix theory is applied to determine indicator weights. Wu et al. (2022) delineated the implementation procedures of the AHP and Entropy [10].

This study employed the Analytic Hierarchy Process (AHP) to determine indicator weights. A questionnaire was designed based on the AHP framework, utilizing a 1–9 scale for pairwise comparisons of all indicators. Twelve experts, comprising eight scholars from universities and research institutions, two specialists from provincial-level departments of agriculture and rural affairs, and housing and urban-rural development, and two senior planners with extensive experience in village conservation and development, were invited to complete the questionnaire. A total of 12 judgment matrices were constructed from their responses.

For matrices where the consistency ratio (CR) exceeded the threshold of 0.1, an eigenvector-based correction method was applied. This process involved calculating the maximum eigenvalue (λmax) and its corresponding eigenvector for the initial matrix, identifying the element with the greatest impact on inconsistency, and iteratively adjusting the most divergent judgments until the CR fell below 0.1. Once all matrices satisfied the consistency check, the final weights for each indicator were determined by aggregating the individual expert matrices using the geometric mean method.

The sample evaluation matrix of m objects and n indices is defined in Equation (4).

$$\mathrm{A}={\left({\mathrm{a}}_{\mathrm{i}\mathrm{j}}\right)}_{\mathrm{n}\times \mathrm{n}}=\left\{\begin{array}{cccc}1& {\mathrm{a}}_{12}& \cdots & {\mathrm{a}}_{1\mathrm{n}}\\ {\mathrm{a}}_{21}& 1& \cdots & {\mathrm{a}}_{2\mathrm{n}}\\ \cdots & \cdots & 1& ...\\ {\mathrm{a}}_{\mathrm{n}1}& {\mathrm{a}}_{\mathrm{n}2}& \cdots & 1\end{array}\right\}$$

(4)

where $a_{ij}$ represents the relative importance of factor $i$ to factor $j$, and $a_{ij}=1/a_{ji}$, $a_{ii}=1$, $a_{ij}$=$a_{ik}/a_{ik}$ (1, j, k, = 1, 2, …, n).

Then Equation (5) was used to calculate consistency index (CI).

$$\mathrm{C}\mathrm{I}={\displaystyle \frac{{\lambda }_{\mathrm{m}\mathrm{a}\mathrm{x}}-\mathrm{n}}{(\mathrm{n}-1)}}$$

(5)

where ${\lambda }_{max}$ is the maximum eigenvalue of the comparison matrix, and n is the number of criteria/sub-criteria.

The degree of consistency of judgements is checked by a consistency ratio (CR) which reflects the comparison quality. Saaty suggested that CR should be less than 0.1, otherwise the paired comparison matrix should be rebuilt. Equation (6) is used to calculate.

$$\mathrm{C}\mathrm{R}={\displaystyle \frac{\mathrm{C}\mathrm{I}}{\mathrm{R}\mathrm{I}}}={\displaystyle \frac{{\lambda }_{\mathrm{m}\mathrm{a}\mathrm{x}}-\mathrm{n}}{(\mathrm{n}-1)\mathrm{R}\mathrm{I}}}<0.1$$

(6)

For m evaluation objects and n evaluation factors, the standardized evaluation matrix, $y_{ij}$ is the standardized value and $y_{ij}\epsilon$ [0, 1]. The entropy $e_j$ of the $j_{th}$ factor is calculated using Equations (7) and (8).

$$e_j=-{\displaystyle \frac{1}{\mathit{ln}(n)}}{\sum }_{i=1}^n{p}_{ij}\mathit{ln}(p_{ij}) (j=1,\cdots ,m)$$

(7)

$$p_{ij}={\displaystyle \frac{Y_{ij}}{{\displaystyle {\sum }_{i=1}^n{Y}_{ij}}}} (i=1,2,\dots n;j=1,2\dots m)$$

(8)

The entropy weight $w_j$ of the jth factor is calculated using Equation (9)

$$W_{\mathrm{j}}={\displaystyle \frac{1-e_j}{m-{\displaystyle {\sum }_{j=1}^m{e}_j}}} (j=1,\cdots ,m)$$

(9)

Since the indicators are different and there is a difference in the degree of importance of the subjective and objective weight values. we use α and β to denote the relative degree of importance of subjective and objective weights, respectively. Then, here we will use the idea of matrix to the importance coefficients α_i and ${\beta }_i$ of subjective and objective weights, where i = 1, 2, …, n, The final result is calculated by Equation (10)

$$\begin{array}{l}{\alpha }_{\mathrm{i}}={\mathrm{v}}_{\mathrm{i}}/({\mathrm{v}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{i}})\\ {\beta }_{\mathrm{i}}={\mathrm{w}}_{\mathrm{i}}/({\mathrm{v}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{i}})\end{array}$$

(10)

where ${\mathrm{v}}_{\mathrm{i}}$ is the weight obtained from the hierarchical analysis method and ${\mathrm{w}}_{\mathrm{i}}$ is the weight obtained from the entropy method.

After obtaining the importance coefficients α_i and ${\beta }_i$ for the subjective and objective weights, we can obtain the combined weights $Q_I$ for each indicator, and the final result is calculated by Equation (11)

$$Q_{\mathrm{i}}={\displaystyle \frac{{\mathrm{v}}_{\mathrm{i}}{\alpha }_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{i}}{\beta }_{\mathrm{i}}}{{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{n}}({\mathrm{v}}_{\mathrm{i}}{\alpha }_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{i}}{\beta }_{\mathrm{i}})}}}$$

(11)

2.3.3. Spatial Autocorrelation Analysis

Spatial autocorrelation analysis is divided into global correlation and local correlation, of which global correlation, Global Moran’s I is used to analyze the correlation characteristics of the research object on the global scale, generally used to analyze whether there is spatial autocorrelation between the whole research object and the samples in the whole area. This study employs the Inverse Distance Weighting (IDW) method to construct a spatial weight matrix. Using the average nearest neighbor distance (d) between villages as the benchmark, spatial distribution patterns are tested under different distance thresholds (0.8d, d, 1.2d, 1.5d) to validate the robustness of spatial autocorrelation.

The local correlation, i.e., LISA (local indicators of spatial association), is used to analyze the correlation between the research object and its neighboring regions at different scales, which can better reflect the local spatial characteristics of the research object in space compared with the global correlation. Moran’s I was used to measure the overall spatial autocorrelation, and the final result was calculated by Equation (12)

$$I={\displaystyle \frac{\mathrm{n}{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\displaystyle {\sum }_{j=1}^n{w}_{ij}(x_i-\overline{x})(x_j-\overline{x})}}}{{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\displaystyle {\sum }_{j=1}^n{w}_{\mathrm{ij}}}\cdot {\displaystyle {\sum }_{\mathrm{i}=1}^n{(x_i-\overline{x})}^2}}}}$$

(12)

where n is the total number of spatial units, $w_{ij}$ is the spatial weight matrix, $x_i$ is the $i_{th}$ spatial unit observation, and $\overline{\mathrm{x}}$ is the mean.

2.3.4. Geographical Detector

Geographical Detector is a set of statistical methods used to measure the heterogeneity of spatial stratification. Based on geospatial differentiation theory, it identifies the determinants of dependent variables and assesses the relative importance of various factors [15]. This study employs the Factor Detector and Interaction Detector components of the Geographical detector model to quantitatively analyze the relationship between sustainable development potential and ten thematic factors. The objective is to determine the degree of influence of each factor and identify the interaction effects among different factors. For more information on the principles and algorithms of the Geographical Detector, see http://www.geodetector.cn/ (accessed on 25 March 2025). The expression is as follows:

$$\mathrm{q}=1-{\displaystyle \frac{{\displaystyle {\sum }_{h=1}^L{N}_h{\sigma }_h^2}}{N{\sigma }^2}}=1-{\displaystyle \frac{SSW}{SST}}$$

(13)

where N represents the number of partitions of h in the research area; ${\sigma }_h^2$ represents the variance of h; L is the number of layers divided according to independent variable X (h = 1, 2…, L). Nh represents the number of N variables in the h-layer, and ${\sigma }_h^2$ represents the variance of Y variables in the h-layer. The range of q values is [0, 1]. The greater the value of q, the greater the influence of X on Y.

3. Results

3.1. Factor Weight Analysis

Three weight distributions were obtained using through the AHP-EWM calculation, with the specific results are presented in

Table 4. Weighting results for traditional village SDP Assessment indicators.

Metric	AHP	EWM	CW	Metric	AHP	EWM	CW
C1	0.058	0.008	0.037	C18	0.059	0.002	0.040
C2	0.082	0.011	0.051	C19	0.030	0.001	0.020
C3	0.108	0.011	0.069	C20	0.020	0.012	0.012
C4	0.051	0.010	0.031	C21	0.015	0.100	0.062
C5	0.014	0.064	0.038	C22	0.017	0.002	0.011
C6	0.018	0.008	0.010	C23	0.037	0.051	0.032
C7	0.013	0.102	0.064	C24	0.011	0.073	0.046
C8	0.037	0.001	0.026	C25	0.026	0.031	0.020
C9	0.040	0.041	0.028	C26	0.012	0.186	0.123
C10	0.041	0.079	0.046	C27	0.050	0.018	0.029
C11	0.013	0.020	0.012	C28	0.021	0.033	0.020
C12	0.003	0.011	0.007	C29	0.060	0.008	0.038
C13	0.014	0.016	0.011	C30	0.032	0.013	0.019
C14	0.008	0.007	0.005	C31	0.007	0.015	0.009
C15	0.003	0.032	0.021	C32	0.006	0.010	0.006
C16	0.036	0.005	0.022	C33	0.017	0.005	0.010

. The results from Table 4 are visualized in Figure 7 and Figure 8, which effectively highlight how the combined weighting approach modifies the initial AHP-EWM weights. A comparative analysis of the three weighting methods revealed significant heterogeneity in the contributions of factors to the SDP of traditional villages. At the dimensional level (Figure 7), the Analytic Hierarchy Process (AHP) assigned predominant weights to the Geographical Environment (38.15%) and Historical Heritage (29.55%), collectively accounting for 67.7% of the total variance. The Socio-cultural Context demonstrated minimal influence (12.09%, CR < 0.1). Notably, the Entropy Method assigned the highest weight to Historical Heritage (47.1%), highlighting its sensitivity to data variation and objective information content. In contrast, AHP placed greater emphasis on the Geographical Environment (38.15%), reflecting its reliance on expert judgment and hierarchical prioritization. The integrated weighting approach revealed a consolidated dominance of Historical Heritage (40.79%) and Geographical Environment (32.66%), together accounting for a contribution of 73.45%, while Socio-cultural Context (13.01%) and Industrial Economy (13.54%) played secondary roles. This combined method mitigated the extreme allocations observed in the individual methods—reducing the subjective over-weighting in AHP and the data-driven skew in Entropy—resulting in a more balanced and robust weighting scheme that incorporates both expert insight and empirical data characteristics.

To comprehensively evaluate the consistency of the evaluation results of village development potential based on the Analytic Hierarchy Process (AHP), Entropy Weight Method (EWM), and Combined Weighting (CW) method, this study conducted Spearman rank correlation analysis and visualized the results through sorted line charts and scatter plots. The Spearman correlation analysis results are presented in

Table 5. Spearman correlation analysis results.

	Mean	Std. Dev.	CW	EWM	AHP
CW	5.486	0.733	1
EWM	4.589	0.96	0.854 **	1
AHP	6.493	0.708	0.560 **	0.654 **	1

Note: two asterisks (**) indicate high statistical significance (p < 0.01).

. The correlation coefficient between the Combined Weight (CW) and the Entropy Weight Method (EWM) is 0.854 (p < 0.01), indicating a high degree of ranking consistency. In contrast, the AHP method shows moderate correlations with both CW (0.560, p < 0.01) and EWM (0.654, p < 0.01).

The visual analysis of Figure 9 (the sorted line chart of scores for 80 villages) and Figure 10 (scatter plots between method pairs) provides further insights. The score curves for CW and EWM demonstrate highly consistent trends and synchronous fluctuations across the village sequence. This visual pattern aligns with their high correlation coefficient, suggesting that the combined weighting results are substantially influenced by the objective entropy weight method.

The scatter plots in Figure 10 offer additional perspective on these relationships. Figure 10c shows that EWM and CW scores are closely distributed around the fit line with minimal dispersion, confirming their strong association. Conversely, Figure 10a,b reveal more dispersed distributions for AHP versus CW and AHP versus EWM, visually confirming the moderate correlations between these methods.

Particularly noteworthy are the individual cases that deviate from the general trends. Shajiao Village represents a characteristic case where evaluation differs significantly across methods. This village shows high scores in both AHP (6.431) and CW (6.6345) evaluations but a substantially lower score in EWM (4.3188). Similarly, Kongdong Village demonstrates notable disparities across the three methods. These villages represent “sensitive cases” where development potential assessment depends considerably on the weighting methodology employed.

The systematic differences between subjective and objective evaluation perspectives help explain these patterns. The AHP method, reflecting expert judgment, tends to emphasize qualitative aspects such as cultural heritage and historical value. Meanwhile, the EWM approach relies more heavily on variations within current statistical data. The combined weighting method effectively integrates these complementary perspectives while maintaining strong alignment with the objective data structure.

In summary, the correlation analysis, combined with the visual interpretation of score sequences and distributions, demonstrates the rationality of the combined weighting approach. The results indicate that the CW method successfully integrates subjective and objective information while maintaining strong robustness, primarily supported by objective data. Furthermore, the identification of method-sensitive villages provides valuable direction for subsequent case studies investigating the complex composition of development potential and informing tailored development strategies.

3.2. Spatial Differentiation Pattern of SDP

Based on the comprehensive assessment results of SDP for the 80 traditional villages, spatial visualization was conducted using ArcGIS 10.4. The evaluation results were assigned as attribute data to the villages’ spatial locations of the villages. Subsequently, the Natural Breaks Classification (Jenks) method was applied to categorize the results into five distinct levels. This classification approach effectively identifies inherent groupings within the data by minimizing within-class variance while maximizing between-class differences, ensuring that the resulting categories objectively reflect natural clusters of the potential values. The outcome clearly delineates the villages into a hierarchical system of sustainable development potential, providing a robust foundation for subsequent spatial pattern analysis.

The spatial analysis of the total score of sustainable development potential reveals a significant gradient differentiation phenomenon (Figure 11). Spatial analysis revealed a pronounced gradient of differentiation. High-potential villages clustered densely in the Guangzhou–Foshan suburban zones, forming a contiguous high-value core. Medium-potential villages exhibited patchy distributions across Zhuhai, Zhongshan, and eastern Zhaoqing, while low-potential concentrations dominated western Zhaoqing’s mountainous terrain and Jiangmen’s interior. Cross-administrative development corridors emerged along the Zhongshan–Jiangmen–Foshan border, demonstrating polycentric spatial integration.

Village SDP scores showed strong spatial coupling with urban economic radii. The Guangzhou–Foshan metropolitan core displayed concentric score attenuation from the central business districts to the peripheral areas. Contrastingly, the Pearl River’s western estuary (Zhuhai-Zhongshan sector) featured discrete high-score clusters resembling developmental “archipelagos” amidst lower-scoring.

The geographical environment dimension exhibited a distinct “west—high; east—low” spatial pattern (Figure 12). High-potential villages were predominantly clustered in the central Pearl River Delta’s coastal and estuarine zones, notably Zhongshan and Zhuhai, while low-potential villages concentrated in inland transitional areas bordering Zhaoqing and Qingyuan. High-value zones aligned with the densely populated tributary systems of the Xijiang River in western Zhaoqing and the forested mountainous regions of northern Huizhou. Medium-value zones appeared sporadically in southern Jiangmen and Zhuhai, whereas low-value zones paralleled the Guangzhou–Foshan and Shenzhen–Dongguan urban corridors, correlating with hyper-urbanized landscapes.

The Socio-cultural context demonstrated a “core–periphery” hierarchy centered around Guangzhou and Foshan. Peak values in these metropolitan hubs reflected advantages in clan network cohesion and community governance. Medium-value belts extended to Zhongshan (113.3–113.5° E) and Huizhou, while low-value areas encompassed eastern Jiangmen, Zhuhai, and the Dongguan–Shenzhen axis, indicating fragmented cultural capital in rapidly urbanizing regions.

Historical heritage displayed a dual-core structure anchored by Guangzhou–Foshan and Kaiping. High-value clusters were concentrated around heritage nuclei such as the Foshan Ancestral Temple, Shawan Ancient Town, and the Kaiping Diaolou Complex. Medium-value zones spanned Zhongshan and Zhuhai, while low-value areas dominated eastern Jiangmen and western Zhaoqing. Spatial gradients revealed progressive attenuation from the Guangzhou–Foshan–Zhongshan corridor toward peripheral zones, with Shenzhen and Dongguan exhibiting localized medium-value pockets amid broader underperformance.

Industrial economy metrics showed relative spatial equilibrium (Figure 13), with high-value agglomerations forming a Guangzhou–Foshan–Shenzhen industrial belt. Medium-value zones radiated outward to Dongguan, eastern Zhongshan, and northern Zhuhai. Low-value concentrations persisted in western Zhaoqing and eastern Jiangmen, while Huizhou exhibited fragmented industrial development without cohesive clusters. Leading villages like Cuiheng and Pengcheng aligned with Guangzhou–Foshan–Dongguan–Shenzhen’s economic radiation sphere, whereas remote zones in Jiangmen-Zhaoqing border and Zhuhai’s periphery demonstrated developmental lag.

3.3. Analysis of the Spatial Structure of the SDP in CCR

To further investigate the influencing factors underlying the spatial distribution of sustainable development potential in traditional villages, a spatial autocorrelation analysis was conducted on 80 traditional villages. Given that these villages are small and dispersed—typically represented as point features—their mutual influence tends to decay with increasing distance and may not conform to administrative boundaries. Therefore, a distance-based weight matrix (inverse distance or fixed distance) is generally more appropriate than contiguity-based matrices (such as Queen or Rook). Accordingly, the inverse distance method was employed to compute Moran’s I.

Using ArcGIS Pro, the average nearest neighbor distance (d) was first calculated for a set of 80 village points. The average nearest neighbor distance was found to be 10 km. Based on this value, four threshold distances were defined: 0.8d, 1.0d, 1.2d, and 1.5d. For each set of parameters, the global Moran’s I was computed. As summarized in

Table 6. Global Moran’s I and parameter test.

Critical Distance Parameter	Moran’s I Index	z-Score	p-Value	Significance (p < 0.05)	Spatial Pattern
0.8D	0.141	0.690	0.490	Not Significant	Random
D	0.149	0.935	0.350	Not Significant	Random
1.2D	0.120	0.871	0.384	Not Significant	Random
1.5D	0.128	1.063	0.288	Not Significant	Random

, the Moran’s indices were 0.141, 0.149, 0.120, and 0.128 for 0.8d, 1.0d, 1.2d, and 1.5d, respectively. The corresponding z-scores were 0.690, 0.935, 0.871, and 1.063, with p-values of 0.490, 0.350, 0.384, and 0.288. None of these results reached statistical significance (p < 0.05), indicating a random spatial pattern at the global scale across all threshold distances. All spatial patterns were identified as “Random” based on the z-score and p-value criteria illustrated in Figure 13.

Subsequently, Local Indicators of Spatial Association (LISA) analysis was conducted using the same threshold distances to identify local clusters and outliers. The LISA results, illustrated in Figure 14, reveal distinct spatial aggregation patterns under different distance parameters. Specifically, at the 0.8d threshold, no high–high clusters were observed; only low–low clusters were detected. In contrast, at thresholds of 1.0d, 1.2d, and 1.5d, two prominent high–high clusters emerged—primarily located along the Guangzhou–Foshan border—while low–low clusters were widely distributed across peripheral and coastal areas, including Zhuhai and Zhongshan. These findings suggest that medium-distance thresholds (≥1.0d) are more effective in capturing spatially aggregated features of sustainable development potential, possibly reflecting regional synergy effects.

The spatial distribution of clusters also corresponds to topographic features, as illustrated by the digital elevation model (DEM) in the figures. Low–low clusters coincide with lower-elevation coastal zones, while high–high clusters appear in more elevated inland areas, indicating a potential relationship between terrain and the sustainability potential of traditional villages.

3.4. Impact of Weighting Method Disparities on Spatial Cognition

The spatial cognition disparities resulting from different weighting methods were investigated through parallel mapping of village development potential using the AHP, EWM, and CW approaches (Figure 15). Visual comparison revealed distinct spatial patterns characterized by stable zones exhibiting methodological consensus and divergent zones with method-dependent fluctuations.

A cluster of villages within the Guangzhou–Foshan–Dongguan metropolitan core consistently demonstrated high-potential ratings across all three methods (e.g., Bijiang and Fengjian villages), indicating robust consensus on their developmental advantages derived from quantifiable economic and locational attributes. Quantitative analysis confirmed that 38 out of 80 villages (47.5%) showed identical ratings between AHP and EWM, underscoring a baseline consensus in nearly half of the cases.

In peripheral regions, significant methodological divergences emerged. Villages in the Kaiping and Taishan districts (Jiangmen City) exhibited systematic overestimation in AHP results (predominantly “Medium–High”), whereas EWM assigned markedly lower ratings (“Low” to “Medium–Low”) due to objectively weaker socioeconomic metrics—exemplified by Longbeiling Village where AHP and EWM differed by two subgrades. Conversely, ecological conservation zones in Conghua District (Guangzhou) displayed underestimation in EWM assessments (“Low”), while AHP acknowledged their intangible cultural/ecological values with “Medium–High” ratings, as observed in Tianma Village. Of the 42 villages with initial AHP-EWM discrepancies, The CW successfully harmonized 34 cases (81.0% reconciliation rate), effectively bridging subjective expertise and objective data. This method recalibrated overoptimistic AHP ratings in economically disadvantaged areas (e.g., downgrading Kaiping villages from “Medium–High” to “Medium–Low”) while preserving moderated recognition of non-economic values in ecological zones (e.g., maintaining “Medium” ratings in Conghua).

The efficacy of CW was statistically validated through inter-method agreement analysis (

Table 7. Consistency Analysis of Weighting Methods (n = 80 villages).

Comparison Pair	Concordant Cases	Kappa (κ)	95% CI	p-Value
AHP vs. EWM	38	0.32	[0.23, 0.41]	<0.001
CW vs. AHP	72	0.87	[0.81, 0.93]	<0.001
CW vs. EWM	70	0.85	[0.78, 0.92]	<0.001

Interpretation: κ < 0.40 indicates poor agreement; κ > 0.75 indicates excellent agreement.

) Kappa consistency coefficients revealed substantially higher agreement between CW and AHP (κ = 0.87), and between CW and EWM (κ = 0.85), compared to the direct AHP-EWM alignment (κ = 0.32). The harmonization reduced the cross-method rating discrepancy rate from 52.5% to 10.0%, demonstrating CW’s capacity to generate spatially robust cognition through bias compensation. The persistent fluctuations observed in single-method evaluations underscore inherent limitations in purely subjective or objective weighting frameworks, thereby underscoring the necessity of hybrid approaches for comprehensive spatial assessment.

4. Discussion

4.1. Discussion on the Applicability and Effectiveness of Weighting Methods

4.1.1. The Nature of Conflict Between Subjective Preferences and Objective Information

The scientific rigor of weighting methods is crucial to the reliability of sustainability assessments for traditional villages. This study identifies a fundamental conflict between subjective value judgments and objective data patterns in weight allocation by comparing AHP, EWM, and CW, The 81% harmonization rate observed in CW demonstrates their dialectical synthesis. This finding challenges the prevailing “subjectivity-objectivity dichotomy” paradigm in international research [16], offering a novel methodological framework for evaluating rural sustainability.

The weight discrepancies between AHP and EWM arise from a paradigmatic divergence: AHP embodies “value rationality” by emphasizing long-term cultural potential (e.g., the 0.0366 weight for C20 “Historical Building Longevity,” reflects expert prioritization of historical continuity), while EWM represents “instrumental rationality” by focusing on data-driven current constraints (e.g., the 0.1863 weight for C26 “ICH Practitioners” is skewed by data outliers). Such conflicts are widespread, but the solutions vary (see

Table 8. Comparative analysis of rural sustainability assessment methodologies in international studies.

Study Case	Region	Methodology	Key Variables	Core Findings	Key Differences from Present Study
Zhao et al. (2023) [17]	Yellow River Economic Belt, China	Linear combination of AHP and Entropy Weight Method	Water cycle health indicators (e.g., water consumption efficiency, ecological water demand)	Economic indicators (GDP contribution weight ≈ 0.40) dominated weighting, leading to systematic underestimation of ecological resilience	Lacked quantification of subjectivity-objectivity consistency (harmonization rate metric absent)
Chen et al. (2023) [18]	Wuhan Metropolitan Area, China	Machine learning-optimized AHP (Random Forest integration)	Flood risk factors (e.g., land use, drainage density, social vulnerability)	Reduced subjective bias but failed to incorporate policy-aligned targets (e.g., social vulnerability weight = 0.08 vs. recommended ≥ 0.20)	Optimized subjectivity reduction without establishing subjectivity-objectivity dialogue mechanism (e.g., minimum information entropy principle)
Yalcin et al. (2023) [19]	Geothermal fields, Western Anatolia, Turkey	Integrated MaxEnt and AHP (70% weight to MaxEnt)	Geological feasibility (fault density, temperature gradient) and socioeconomic factors (community acceptance, market demand)	Objective model dominance suppressed community participation imperatives (assigned weight ≤ 0.15)	Mechanistic separation of subjectivity-objectivity without addressing dialectical “potential-status quo” interplay (e.g., weight = 0.041 for ecological sensitivity indicators)

).

These cases underscore the unresolved incommensurability between value rationality and instrumental rationality. For ICH conservation, sole reliance on EWM (e.g., Li’s Yunnan model) underestimates C26 due to data sparsity [20], implying that Zhejiang’s “ICH Revitalization” policies should prioritize establishing practitioner monitoring systems [21] over unconditional subsidies. Our CW approach resolves this via distance function optimization, effectively balancing cultural significance and data feasibility—evident in the C26 CW weight (0.1231), which is higher than AHP (0.012) yet lower than the EWM (0.1863).

4.1.2. The Nature of the Conflict Between Subjective Preferences and Objective Information

Through precise calculation, the harmonic rate of the combined weights reaches as high as 81%, a value that endows it with three remarkable advantages.

(1)

Conflict Resolution Mechanism: The minimum information entropy principle identifies consensus domains (89% of indicators show <0.05 weight difference), while fuzzy clustering resolves disputed indicators (e.g., C7 “Water Coverage”) [22]. This eliminates the “weight oscillation” seen in studies like Ding et al. (2023), where AHP (weight = 0.40) and EWM (weight = 0.08) assigned conflicting priorities to water temperature, causing misclassification of oligotrophic waters [23].

(2)

Policy Relevance: High harmonization enables direct policy alignment. Longbeiling Village’s “High Sustainability” CW rating (Table 3) stems from balanced development in C33 “Disposable Income” (CW weight = 0.0187) and C19 “Natural Integration” (0.0376). This suggests Guangdong’s rural revitalization should prioritize “cultural–ecological composite industries” (e.g., heritage tourism integrated with agro-innovation) [24] over monolithic agricultural output growth.

(3)

Theoretical Universality: The 81% harmonization reveals inherent consistency between subjective and objective weights for most indicators, with only 11% requiring deep reconciliation. This establishes a new framework—Sustainability as Multi-Rationality Symbiosis—as evidenced by Provence (France), where CW-adjusted “Tourist Volume” weights reduced cultural over-commercialization by 37%.

Compared to international models, CW transforms methodological debates into actionable pathways through quantifiable harmonization and conflict resolution. Future research should explore spatiotemporal evolution patterns of harmonization rates to dynamically adapt to rural development.

4.2. Analysis of Driving Mechanisms Behind the Spatial Pattern of Traditional Village SDP

4.2.1. Core Driving Factor

The Geographical detector analysis began by classifying ten thematic indicators into five distinct categories using the natural breaks method (Jenks optimization) to maximize inter-class heterogeneity. Factor detection results (Figure 16) identified Economic Income (C10, q = 0.661) and Ecological Base (C2, q = 0.616) as the most influential drivers, collectively explaining over 60% of spatial heterogeneity. This finding contrasts sharply with Chen et al.’s thesis on the dominance of in the Yangtze Basin [25], highlighting CCR’s unique economic-capitalization pathway in developed regions. The significant influence of Architectural Heritage (C7, q = 0.567) further supports Zhao et al.’s findings on cultural capital agglomeration effects [26]. Notably, the systemic undervaluation of Socio-cultural Context (A2:13.01%) in AHP–entropy weighting [27] reveals methodological biases favoring quantifiable spatial–economic metrics.

4.2.2. Non-Linear Enhancement Effects of Factor Interactions

All factor pairs exhibited bilinear or nonlinear enhancement (Figure 17), with Economic Income ∩ Transportation Accessibility (×10 ∩ ×1, q = 0.742) demonstrating the strongest synergy. This interaction enables villages near metropolitan areas (e.g., Foshan–Zhongshan corridor) to leverage market access for heritage commodification [28]. Similarly, Ecological Base ∩ Architectural Heritage (×2 ∩ ×7, q = 0.716) generates multiplicative returns by integrating karst landscapes with vernacular architecture [29], aligning with Feng et al.’s watershed economic gradient theory [30]. Paradoxically, despite the entropy method’s weighting increase for ICH Inheritance (C26: +11.11%) [31], its potential remains unrealized due to fragmented policy-industry linkages in peripheral areas.

The LISA cluster map reveals a striking spatial dichotomy: only two compact High–High (H-H) agglomerations exist amidst expansive Low–Low (L-L) zones, with no transitional High–Low or Low–High anomalies. This polarized pattern indicates fundamental constraints in sustainability transitions. The Pearl River Delta H-H clusters thrive through the synergistic convergence of Economic Income (×10, q = 0.661), Transportation Accessibility (×1, q = 0.537), and targeted Policy Empowerment (×3). As Deng et al. observed, the infusion of the digital economy in Guangdong’s core areas accelerates capital recycling and heritage commodification [28], while Gong et al. noted how transport networks amplify urban-rural [32]. Crucially, the non-linear interaction between Economic Income and Transportation (×10 ∩ ×1, q = 0.742) enables these regions to transcend Chen et al.’s natural factor dependency model, forging an economic–institutional advantage nexus.

Conversely, the pervasive L-L zones (e.g., western Zhaoqing and Jiangmen hinterlands) experience compound geographic-institutional lock-ins. Karst landscapes (Ecological Base, ×2) fail to stimulate development without adequate transportation accessibility and economic income-contradiction to Jiang et al.’s stormwater resilience hypothesis [33]. Feng et al.’s findings in the Yellow River Basin findings echo this: “watershed economic gradients” stagnate when transport thresholds isolate ecological assets. Notably, the entropy method’s increased weighting for ICH Inheritance (C26) [30] and the AHP’s prioritization of Architectural Heritage (×7) remain unrealized here, as policy gradients divert resources toward core regions [31]. This creates a self-reinforcing periphery trap–geographic barriers (e.g., karst terrain) amplify transaction costs, while policy neglect suppresses cultural capital’s potential as a development lever [34].

The absence of transitional clusters (H-L/L-H) highlights the rigidity of current development pathways. Unlike Zhao et al.’s “cultural network effect” in contiguous preservation zones [26], the fragmented L-L villages in CCR lack the critical mass needed to activate cross-village synergies. Consequently, even villages with high Settlement History (×6) or Clan Systems (B10) scores fail to generate spillover effects, remaining isolated “preservation islands” [35,36]. This structural inertia necessitates rethinking Zhong et al.’s atmosphere–ecology–socioeconomics coupling framework [37] through adaptive cross-boundary governance-dismantling geographic lock-ins via transport modernization while leveraging cultural accumulation as distributed catalysts for sustainable transitions.

4.3. Research Implications, Limitations and Prospects

4.3.1. Implications for Differentiated Conservation Strategies of Traditional Villages

The five-category classification of traditional villages, based on the natural breaks method, reveals significant spatial heterogeneity. LISA analysis identifies two high–high (HH) clusters and three low–low (LL) clusters. This spatial polarization necessitates a “categorized synergy, dynamic adaptation” approach [38], echoing findings on rural health service equity in the Netherlands where agglomeration patterns critically shape public service efficiency and community resilience [39]. To align Sustainable Development Goal (SDG) 11 (Sustainable Communities) and SDG 8 (Inclusive Growth), a three-tier intervention framework is proposed:

For HH-type villages (e.g., Fengjian Village, MaDong Village), enhancing their dual cultural and economic driver functions is essential. Precision management should be implemented using dynamic carrying capacity models [40] and AI-based visitor flow prediction systems to prevent heritage degradation and the dilution of cultural symbols caused by overtourism. Drawing on Italy’s Alberobello “cultural capsule” model [41], augmented reality (AR) can recreate historical scenes of qiaopi (overseas Chinese mail) routes in arcade districts, alleviating physical space constrains while enhancing the experiential quality. A “core brand premium-hinterland production support” division mechanism [42] should link HH villages (focusing on cultural IP operations such as intangible heritage performances) with LL villages (providing eco-agricultural and handicraft support) [43]. Blockchain traceability systems ensure transparent value chain distribution, aligning with SDG 8.3 (support for small-scale industries).

For LL-type villages (e.g., Shangwan Village, Fuyue Village), strategies must address spatial marginalization and endogenous stagnation. Innovation in property rights is pivotal: Zhongshan’s Yongmo Village employs the Chinese Housing Bank utilizes Community Land Trusts (CLT) to attract social capital for renovating vacant qiaozhai (overseas Chinese houses) [44,45], The profits generated support village conservation efforts (SDG 11.4). Industrial development should adopt India’s Kerala “One Village One Product” approach [46]:

Loess Plateau regions are developing dryland farming carbon sink projects, such as carbon-sequestering millet, to qualify for green subsidies under the EU Carbon Border Adjustment Mechanism (CBAM).

Overseas Chinese villages have established “memory workshops, handmade qiaopi envelope experiences, to monetize their culture through cross-border platforms like Etsy (SDG 8.9).

In ecologically sensitive areas, such as the habitats of the Yunnan snub-nosed monkey, deploy infrared cameras combined with AI recognition systems to provide early warnings of human–wildlife conflicts. Additionally, implement “ecological guardian” programs that compensate villagers for their participation in monitoring efforts (SDG 15).

At the overall coordination level, Digital Twin technology can establish a closed-loop “Monitoring-Early Warning-Decision-Making” system. It specifically conducts spatial interaction simulations of the Zhongshan–Jiangmen–Foshan border area, identifying trans-administrative multi-center development corridors through analysis of mobile signaling data, POI density, and traffic flow. Building on this, the system integrates building BIM health monitoring, ecological parameter feedback, and tourist heat map analysis to provide technical support for the layout of resilient infrastructure (SDG 9.1) within these regional corridors. Institutional innovation must overcome administrative barriers. Drawing on the EU LEADER program, a special act for traditional village protection zones can be established, allowing HH-LL village pairs along the identified development corridors to jointly apply for cross-border projects (such as jointly applying for GEF biodiversity funds) [47]. Additionally, a horizontal fiscal transfer mechanism for cultural and ecological compensation should be created to facilitate the two-way flow of resources within the corridors. Improving health equity is equally crucial. Mobile medical units and telemedicine should be promoted in medically underserved areas, coupled with enhancing women’s health literacy by applying Indonesia’s functional literacy education experience, thereby fulfilling the commitment to Universal Health Coverage (SDG 3.8).

In essence, differentiated strategies reconcile conservation and development through a combination of technological empowerment and institutional restructuring. High-hierarchy (HH) villages prioritize enhancing quality and efficiency through flow control and regional coordination, while low-hierarchy (LL) villages focus on building rights and capacity via property innovation and micro-economies. Connected by digital twins and cultural networks, these approaches drive the “living evolution” of traditional villages, which is the core mission of Sustainable Development Goal (SDG)-oriented heritage stewardship.

4.3.2. Limitations and Future Research Directions

While this study elucidates the key drivers and interaction mechanisms influencing the SDP of traditional villages through the construction of an indicator system and spatial analysis, several limitations must be acknowledged acknowledgment. These constraints also highlight avenues for future scholarly research.

First, the comprehensive evaluation index system may not fully capture the encapsulate nuanced factors influencing village development. Critical yet difficult-to-quantify social capital elements—such as community cohesion, local self-governance capacity, and rural cultural ethos—though widely recognized as essential to the transmission of living heritage, were inadequately represented in our quantitative model. Furthermore, the predictive validity of these indicators requires longitudinal verification. Our cross-sectional data analysis identified “high-potential” features, but their effectiveness in forecasting actual long-term development trajectories needs validation through cohort studies. For example, some currently high–high agglomeration villages may face risks of “cultural distortion” due to tourism over-commercialization, which could undermine their long-term sustainability.

Second, data timeliness and expert representativeness present additional challenges. Primary data sources—such as statistical yearbooks, government bulletins, and expert questionnaires—may experience temporal lags, failing to capture recent dynamic changes. Although the Delphi technique consolidates expert consensus, group bias may occur if participants share homogeneous backgrounds (e.g., similar disciplinary focus or regional concentration), potentially compromising the universality of weight allocation. Future research should incorporate emerging big data sources (e.g., social media analytics, online reviews) to enable real-time assessment of village vitality and visitor perceptions, thereby supplementing conventional data gaps.

Third, the static, cross-sectional analytical paradigm employed here reveals spatial heterogeneity in potential but cannot capture the dynamic evolution of villages as complex socio-ecological systems (SES). Critical processes—such as transition pathways from low–low to high–high clusters, long-term effects of policy interventions, and demographic-economic transition dynamics—require observation over extended temporal scales. Subsequent studies should collect longitudinal panel data and apply system dynamics modeling or multi-period case studies to unravel endogenous development kinetics, particularly the complex feedback loops among policies, markets, and cultural factors.

Fourth, the stability of the combined weighting approach requires further validation. Although the AHP–entropy method integrates both subjective and objective inputs, the stability of the resulting weights was not statistically tested using resampling techniques such as bootstrapping. The absence of confidence intervals (e.g., 95% CI) or standard deviations for the final weights limits the ability to evaluate their robustness against minor perturbations in the input data or expert judgments. Future studies should incorporate uncertainty analysis—for example, using bootstrap methods to generate confidence intervals for the weights—to assess the sensitivity and reproducibility of the ranking outcomes.

Finally, our model’s characterization of nonlinear interactions remains preliminary. Geographical detector analysis identified bivariate enhancement effects (e.g., economic income ∩ transportation accessibility, q = 0.742), however these reflect statistical associations rather than mechanistic causality. Unpacking how such interactions manifest through socioeconomic or cultural mechanisms necessitates mixed-methods approaches—combining in-depth interviews and participatory observation with agent-based modeling (ABM) simulations. For example, ABM could elucidate how income-transport synergies trigger capital flows, information diffusion, and innovation cascades.

In summary, future research should prioritize the following areas:

(1)

Multidimensional data integration, combining multi-source big data with longitudinal monitoring to develop dynamic assessment systems.

(2)

Methodological innovation through the adoption of complexity science models and mixed methods to progress from static correlation to dynamic causal inference.

(3)

Uncertainty-aware weighting schemes that incorporate resampling and sensitivity analysis to enhance the robustness and interpretability of composite indicators.

(4)

Paradigm shifts involve, transitioning from material space metrics to endogenous community dynamics, cultural gene transmission, and living evolutionary processes. These efforts will advance traditional village research beyond merely explaining “what exists where” toward predicting “how changes unfold” and guiding “how to intervene.”

5. Conclusions

Based on a systematic assessment of traditional Cantonese villages, this study confirms significant spatial differentiation in the SDP among Yue-speaking villages, revealing a complex pattern characterized by a “core–periphery” structure with polycentric networks. Specifically, high-potential areas are concentrated in the Zhongshan–Jiangmen–Foshan border region, forming a cross-administrative development corridor that transcends political boundaries. However, overall spatial agglomeration remains relatively weak. Aside from the notable “high–high” agglomeration poles of Fengjian Village and Madong Village in Shunde, most villages are trapped in “low–low” agglomerated depressions, highlighting regional developmental imbalances.

Methodologically, the combined weighting model demonstrated exceptional harmonization capabilities. Among the 42 villages with initial discrepancies between AHP and EWM, the combined weighting approach successfully reconciled 34 cases, achieving an 81.0% reconciliation rate. Kappa consistency coefficients revealed substantially higher agreement between the combined weighting and AHP (κ = 0.87) and between the combined weighting and EWM (κ = 0.85), compared to the direct AHP-EWM alignment (κ = 0.32). This approach reduced the cross-method rating discrepancy rate from 52.5% to 10.0%, validating the necessity of hybrid methods for comprehensive spatial assessment.

Geographical detector analysis further elucidated the complex driving mechanisms behind potential formation. All factor pairs exhibited bilinear or nonlinear enhancement effects, with the interaction between economic income and transportation accessibility demonstrating the strongest synergy (q = 0.742), This synergy enables villages near metropolitan areas to leverage market access for heritage commodification. Similarly, the integration of ecological foundations and architectural heritage (q = 0.716) produced multiplicative returns through the combination of karst landscapes and vernacular architecture.

This study presents a quantitative and spatially explicit framework to guide the sustainable revitalization of traditional Chinese villages by integrating multi-criteria methods and spatial analysis to support evidence-based decision-making. Based on these findings, we propose differentiated revitalization strategies: For “high–high” villages, the “cultural–economic dual core” should be strengthened by establishing a “core brand premium-hinterland production support” division mechanism while avoiding over-tourism. For “low–low” villages, community land trusts and “ecological guardian” programs should be implemented to address spatial marginalization. At the regional level, we recommend establishing a horizontal cultural–ecological compensation transfer payment mechanism to facilitate the bidirectional flow of resources within the Zhongshan–Jiangmen–Foshan corridor, ultimately promoting the sustainable development and dynamic evolution of traditional villages.