Designing a traditional village cluster protection-utilization system via complex network analysis: Qiandongnan case study

Abstract

This study develops a Cluster Protection and Utilization System (CPUS) through complex network analysis to enhance conservation strategies for traditional villages in Qiandongnan, China. Results reveal naturally shaped clustered distributions forming multi-centered radial networks with a distinct core-periphery hierarchy. The network demonstrates moderate structural integrity combined with low stability (27% node failure tolerance) and high vulnerability. Competitiveness stratification identifies five tiers, where only 9% of settlements (primarily core nodes) exhibit high/moderate competitiveness. Role-based classification pinpoints critical connectivity hubs. Development potential shows a strong correlation (r = 0.82, p < 0.01) with network metrics, attributed to government support, transport access, and architectural value. The proposed framework integrates spatial optimization algorithms with adaptive reuse protocols, achieving 23% greater resource efficiency than conventional approaches while balancing preservation and sustainable tourism. This study innovatively applies complex network analysis to build a Cluster Protection and Utilization System (CPUS), offering a feasible approach for the collaborative development of traditional villages and cultural heritage clusters globally.

Introduction

The protection and utilization of cultural heritage has always been a key focus for scholars worldwide, and the holistic protection and coordinated utilization of heritage clusters are of utmost importance¹. Traditional villages hold immense value in history, culture, science, art, society, and economy, making their protection and utilization an international focal point of interest². In order to protect traditional villages, the Chinese government began assessing and protecting them as early as 2012. By August 2024, the number of traditional villages had reached 8155, showing a significant clustering effect³. However, due to their numerous quantity, wide distribution, and diverse types, the predominant mode of individual village protection has proven inefficient⁴. Moreover, driven by local interests, traditional villages have operated independently, hindering the formation of complementary and prosperous open collaborative systems among them. Therefore, the key innovation in the path towards clustered protection and development of traditional villages lies in establishing community frameworks based on their cluster relationships. This involves analyzing the theoretical relationships between village entities and their regional groups, and scientifically constructing the most suitable clustered structural protection and utilization systems. This system emphasizes the interdependence of villages in terms of space, culture, and function, advocating for holistic conservation, complementary development, and dynamic utilization. It treats cultural heritage as an interconnected system rather than isolated entities, constructing a multi-level “Core–Linkage–Periphery” structure⁵. This approach avoids the superficial splicing of spatial boundaries defined by local government policies, which is key to innovating the conservation and development pathways of traditional village clusters. Furthermore, it provides the urgently needed theoretical and technical foundation for the conservation planning of traditional villages, holding significant theoretical and practical implications for the systematic preservation and developmental utilization of their historical and cultural heritage.In recent years, research on traditional villages has predominantly focused on individual village scales, with limited studies addressing village clusters. These studies primarily explore concentrated protection methods such as cluster protection strategies⁶ and centralized development modes and strategies⁷. They often summarize current characteristics⁸ and patterns⁹, yet discussions on theoretical frameworks and specific implementation paths remain relatively scarce. While some scholars have examined cluster spatial structures, these efforts often rely on subjective planning or geographical relationships, such as subjective planning based on current conditions and balanced development¹⁰, or clustering based on spatial, cultural, and administrative elements⁴. There is a lack of quantitative and visual research treating villages as complex entities, and a deficiency in scientifically aligning internal village cluster connections with market development potentials. Consequently, surface-level cooperation among traditional village clusters often contrasts with ineffective regional collaboration. In the limited number of visualization studies, complex network analysis is less frequently employed compared to traditional clustering methods that rely on spatial proximity. However, complex networks can reveal the intrinsic logical relationships among research subjects, thereby uncovering the dual structure of spatial organization and functional linkages, and facilitating the formulation of appropriate conservation and utilization strategies¹¹. Moreover, research is constrained by difficulties in data collection, typically limiting studies to small groups of up to 100 traditional villages, with a severe lack of strategic coordination on a larger scale. Although there are collective studies on cultural heritage abroad, with research samples exceeding 100 historical sites, the subjects of these studies primarily focus on historical relics such as fortresses and churches within towns, with little attention given to historical villages^12,13,14,15. Qiandongnan Miao and Dong Autonomous Prefecture (hereinafter referred to as Qiandongnan), rich in natural resources and divided by the Miaoling Mountains, serves as an ecological source area for both the Yangtze River Economic Belt to the north and the Pearl River Economic Zone to the south. It also forms a crucial part of the contiguous impoverished areas of the Wuling Mountains in China. Known for its rich indigenous cultural heritage, Qiandongnan boasts 415 national-level traditional villages, ranking first among prefectures and cities, and is recognized as one of the world’s 18 circles for protecting indigenous cultures. Therefore, Qiandongnan holds significant exemplary significance for China and the world in terms of ecological security, ethnic unity, cultural prosperity, and socio-economic development. This study focuses on the typical cluster demonstration area of traditional villages in Qiandongnan, employing methods such as Thiessen polygons and kernel density analysis to explore the spatial distribution characteristics of traditional villages. By establishing a complex network model of village interactions based on topological relationships and leveraging a traditional village gravity model, the study quantitatively and visually analyzes complex network relationships among village clusters. Building upon this analysis, it evaluates the development potentials of traditional villages, dissecting the relationship between village cluster network structures and development potentials. Integrating spatial distribution features of traditional villages, characteristics of complex networks, development potentials, and their interrelationships, the study proposes methods for constructing a clustered protection and utilization system for traditional villages. This endeavor aims to provide theoretical, methodological, and practical foundations for the clustered protection and development of traditional villages in concentrated areas.

Methods

Research object selection and overview

Based on the principles of typicality, authenticity, and scale (Table 1), all nationally designated traditional villages in Qiandongnan Prefecture are selected as the research subjects. As early as 1998, Qiandongnan was designated as one of the eighteen original ecological cultural protection zones recognized by the United Nations, showcasing significant authentic characteristics. It ranks first in China as the municipality (prefecture) with the most traditional villages (415 traditional villages), demonstrating considerable scale (Fig. 1). Additionally, in 2020, the Ministry of Housing and Urban-Rural Development of China listed it as a demonstration project for the concentrated protection and utilization of traditional villages, highlighting the typical demonstrative nature of its clustered development of traditional village groups.

Fig. 1: Spatial distribution of traditional villages in Qiandongnan Prefecture.

This figure illustrates the geographic distribution of traditional villages across Qiandongnan Prefecture. Village locations are represented as points within administrative boundaries. The base map is derived from official geographic data (http://www.qdn.gov.cn/zjqdn/zqgk_5935839/xzqh_5871567/).

Table 1 Characteristics of traditional villages in Qiandongnan

Overview of research objects

Qiandongnan Miao and Dong Autonomous Prefecture is located in the southeastern part of Guizhou Province, China, with geographic coordinates ranging from 107°17'20″E to 109°35'24″E longitude and 25°19'20″N to 27°31'40″N latitude. It borders Hunan’s Huaihua to the east, Qiannan Buyei and Miao Autonomous Prefecture to the west, and connects with Guangxi’s Liuzhou and Hechi regions to the south. To the north, it extends to Zunyi and Tongren cities. The total area of the prefecture is ~30,337 square kilometers. The region features a subtropical humid monsoon climate, characterized by mild winters and moderate summers, with an average annual temperature ranging from 14 to 18 °C and annual precipitation between 1000 and 1500 mm¹⁶. Located in the transitional zone from the Yunnan-Guizhou Plateau to the hilly basin of Hunan and Guangxi, Qiandongnan boasts lush forests and numerous peaks, earning its reputation as the “emerald on the Guizhou Plateau.“Most of Qiandongnan’s land is used for slope cultivation, highlighting its ecological significance despite limited construction land. The prefecture is home to 33 ethnic groups including Miao, Dong, Han, and Tujia, with ethnic minorities accounting for 81.6% of the total population, making it one of China’s autonomous prefectures with the highest population of ethnic minorities¹⁷. Due to its historical and geographical isolation, Qiandongnan has preserved a relatively intact original ethnic cultural ecosystem. It boasts a rich variety of indigenous cultures, with preserved cultural relics dating back to the Wei and Jin dynasties, Tang dynasty hairstyles, Song dynasty costumes, Ming and Qing architecture, among others. Dong Chorus is recognized as a UNESCO Intangible Cultural Heritage, while 72 items such as Miao costumes and ancient Miao songs are listed in China’s national intangible cultural heritage. Qiandongnan is also one of the 18 global ecological cultural protection areas recognized by the World Foundation for Village Culture Preservation (Fig. 2).

Fig. 2: Display of architecture, activities, and daily life scenes in traditional villages of Qiandongnan.

a Panoramic view of traditional village; b Dong Ethnic Customs Park; c Hmong Welcome and Toast; d Hmong New Year Activities; e County Hmong New Year Activities; f Residence and living scenes of the Dong people.

Overall research framework

Using network analysis, this research represents complex and variable relational forms as specific network configurations. Based on these configurations and their changes, the study explains their significance for both individuals and the overall structure. Furthermore, the development of a conservation and utilization system based on the research objectives constitutes the theoretical principle of this study. Numerous research outcomes indicate an inherent and necessary connection between the overall network of the same research object, individual networks, and their node attributes^18,19,20. The relationship between the overall network and the individual network manifests as the overall network serving as a contextual factor that exerts top-down influence on the behaviors of individual network nodes. This influence is realized through two mechanisms: one is the direct impact of the overall network structure across levels on the actions of individual network nodes, while the other is the overall network acting as an environmental variable that plays a cross-level regulatory role in the relationships of individual network nodes, thereby indirectly affecting their behaviors. Accordingly, individual networks reflect certain general features of the overall network based on the relational information formed by self-connection. At the same time, the attributes of network nodes (in this study, the attributes represent the development potential of villages) are controlled by both the overall network and the individual network. The network as a whole governs the expression of node attributes, while the optimization of all node attributes collectively influences the structural characteristics of the overall network. Therefore, a conservation and utilization system for clustered traditional villages (CPUS) is constructed based on the principles of intersectionality, multi-scale, and quantification.

The construction of the Traditional Village Clustered Conservation and Utilization System (CPUS) follows a multi-level, quantitative analytical framework that integrates three key components: spatial analysis, network modeling, and planning application. First, spatial distribution analysis is used to identify the clustering characteristics of villages, establishing the foundation for understanding the spatial pattern. Second, a comprehensive evaluation system is developed to assess village development potential across social, cultural, economic, and ecological dimensions. Third, a complex network model is constructed based on the gravitational relationships among villages, revealing the hierarchical connections and the “core–periphery” structure within the system. Finally, the results of the network analysis are combined with the comprehensive evaluation to classify villages into functional groups and hierarchical zones, enabling the formulation of differentiated conservation, coordinated development, and sustainable utilization strategies. This framework establishes a logical progression from spatial analysis to network interpretation and finally to planning application, clearly illustrating the entire process from data modeling to the construction of a clustered conservation and utilization system (Fig. 3).

Fig. 3: Conceptual framework for the conservation and utilization system of clustered traditional villages.

This framework systematically elucidates the interrelationships among spatial distribution, network structure, and development potential, thereby establishing an integrated system for the conservation and development of traditional village clusters.

Spatial distribution analysis method of traditional villages

Spatial distribution and cluster characteristics are crucial analytical components supporting the construction of a clustered conservation and utilization system. Through the analysis of spatial distribution and cluster characteristics, the objective relational structure among villages is revealed, providing fundamental parameters and node relationships for complex network modeling. This, in turn, supports the establishment of a clustered conservation and utilization system and the formulation of related strategies.

Thiessen polygon method

Thiessen polygon method (also known as Voronoi diagrams) are a result of spatial plane partitioning. Because the areas of Voronoi polygons vary with the distribution of points, the coefficient of variation (Cv) is used to measure the relative variability of polygon areas, thereby analyzing the spatial distribution patterns of sample points²¹. Cv is equal to the standard deviation of Voronoi polygon areas divided by the mean area, and its formula is:

$$\begin{array}{c}{\rm{R}}=\sqrt{\sum \frac{{\left({{\rm{s}}}_{{\rm{i}}}-{\rm{s}}\right)}^{2}}{{\rm{n}}}}({\rm{i}}=1,2,\cdots ,{\rm{n}})\\ {C}_{V}=R/S\end{array}$$

(1)

In the formula: R represents the standard deviation of polygon areas, Si denotes the area of the ith polygon, Si the average area of the polygons, n stands for the number of polygons. When ${Cv}$ = 33–64%, the point set exhibits a random distribution. When ${Cv}$ > 64%, the point set shows a clustered distribution. When ${Cv}$ <33%, the point set demonstrates a uniform distribution²².

Kernel density estimation

Kernel density estimation calculates the density of point or line features in their surrounding neighborhoods. It uses a formula to convert discrete point data into a continuous surface in space, thereby assessing the concentration and distribution pattern of spatial point data²³. A higher kernel density value indicates a denser distribution of traditional villages, and its formula is:

$$f(x,y)=\frac{1}{n{\lambda }^{2}}{\sum }_{i=1}^{n}k/\left(\frac{{d}_{i}}{n}\right)f(x,y)=\frac{1}{n{\lambda }^{2}}{\sum }_{i=1}^{n}k/\left(\frac{{d}_{i}}{n}\right)$$

(2)

In the formula: $n$ stands for the sample size; $\lambda$ represents the bandwidth; ${k}$ is the kernel function; $d$ denotes the spatial distance to sample $i$; and f(x,y) is the kernel density estimation value at coordinate point(x,y). Using ArcGIS, the research findings related to traditional villages are integrated into their attribute table. Kernel density mapping is performed across different dimensions to visualize the spatial distribution of traditional village features.

The evaluation method for traditional villages

Through literature review and field visits to 415 traditional villages in Qiandongnan, a preliminary comprehensive evaluation index system for traditional villages in Qiandongnan and a development potential evaluation index system were constructed based on comprehensive opinions. The Delphi method was employed to screen and refine the two preliminary evaluation systems mentioned above. A total of 15 experts were invited to participate in the evaluations using both the Delphi and Analytic Hierarchy Process (AHP) methods, including specialists in cultural heritage conservation (6), urban and rural planning (5), and geography and environmental studies (4). Consistency checks were conducted on the AHP pairwise comparison matrices, and all consistency ratios (CR) were below 0.1, thereby establishing the formal evaluation indicator system.

$${ADP}={\sum }_{i=1}^{n}{X}_{i}\cdot {W}_{i}$$

(3)

In Eq. (3), ${X}_{i}$ represents the numerical value computed for indicator i after assignment processing of its original data, ${W}_{i}$ represents the weight parameter associated with indicator $i$. ADP denotes the total value of the evaluation index system. A higher ${ADP}$ value indicates a larger value for either the comprehensive evaluation index system or the development potential evaluation index system of traditional villages, while a lower value indicates the opposite.

Complex network analysis method

Complex network analysis simplifies complex systems into a collection of nodes connected by edges, where nodes represent basic units of the complex system and edges denote interactions between these units²⁴. In this study, nodes represent traditional villages, and edges represent attraction relationships between villages. An edge is assigned a value of 1 if the attraction value between two villages exceeds the average attraction value, and 0 otherwise, forming a relationship matrix to construct the “network” of traditional village clusters. Various quantitative measures are computed as follows:

Cut-point

Cut-points are nodes in a network that, once removed, split the entire graph into two or more disconnected subgraphs (components). A higher number of cut-points indicates a more fragile network²⁵.

Small-world characteristics

A network exhibits small-world characteristics if it is large but sparse, lacking a central core but highly clustered. This is characterized by a small average path length and a large clustering coefficient. In the network structure of traditional village clusters, poorer completeness corresponds to more pronounced small-world characteristics²⁶.

Network density

Network density measures the closeness of connections between nodes in a network. It is calculated as the ratio of the actual number of connections to the maximum possible connections. A lower density indicates a looser network structure and weaker completeness²⁷.

Cohesive subgroups - block model

Block model analysis is a primary method for spatial clustering analysis. It identifies different levels of cohesive subgroups by analyzing relationships between internal and external members of subgroups. It determines the role of blocks in the network by analyzing the exchange of information between different blocks²⁸.

K-core

K-core refers to a subgraph where each node is adjacent to at least k other nodes within that subgraph. K-core analysis (k = 1, 2, 3, …) measures the local stability of the network²⁵.

Core-periphery

The core-periphery model distinguishes between the core and periphery of a network structure. Actors within the core have dense connections, forming cohesive subgroups. Nodes with an average core value above a certain threshold are identified as core nodes, which hold a dominant position in the network structure^28,29.

Degree centrality

Degree centrality measures the extent to which an actor controls other resources and nodes in the network, indicating its importance. Higher values of degree centrality indicate more evenly distributed node importance and a balanced network structure³⁰.

Structural holes

Structural holes describe brokerage positions that connect otherwise disconnected actors in the network, giving them competitive advantages. Efficiency, effectiveness, and constraint are typically used to measure structural holes. Nodes with higher efficiency and lower constraint exhibit more prominent structural hole advantages, indicating superior developmental advantages³¹.

Broker

Brokers are actors positioned in the middle of networks. They can be categorized into five roles based on their specific positions: coordinator, gatekeeper, spokesperson, advisor, and liaison. Each role signifies different capabilities in terms of information dissemination and connectivity within and between subgroups²⁹.

Correlation analysis method

Village gravity relationship analysis

Using the gravity model, based on field surveys, it is found that villages with better overall conditions exert greater influence on their surroundings, with the influence weakening as distance increases. In accordance with the research objectives to explore the mutual comprehensive attractiveness between traditional villages, modifications were made to the gravity model formula. The algorithmic formula is as follows:

$${F}_{{\rm{ij}}}=\frac{{r}_{i}\times {M}_{i}\times {M}_{j}}{{D}_{{ij}}^{2}}$$

(4)

In the formula, M values are computed based on the indicator system in Table 1, where ${M}_{i}$ represents the comprehensive condition value of traditional village $i$, and ${M}_{j}$ represents the comprehensive condition value of traditional village $j.$ ${D}_{ij}$ denotes the distance between the two locations calculated using the Gaode API.${r}_{i}$ is the regional adjustment coefficient, set to 10,000 for ease of comparing gravitational values. For computed results, the average gravitational value serves as the breakpoint: values greater than the average are considered valid data, while values below indicate no gravitational relationship between the two villages, resulting in a stress value of 0. Simultaneously, commute times exceeding two hours between villages are defined as lacking attraction relationships³². Considering practical circumstances, gravitational results are adjusted to construct a gravitational matrix (where villages with gravitational relationships are defined as 1, and without as 0 to reflect the interrelationships among traditional villages in Qiandongnan.

Although a binary threshold was used in the topological analysis to determine village-to-village connections, differences in gravitational strength were still preserved and utilized for analyzing spatial interaction relationships. This representation reflects the weighted strength of inter-village connections while ensuring the comparability and interpretability of network metrics such as degree centrality and modularity.

Analysis of related factors relationships

Utilizing SPSS software to compute village development potential, village-related attribute factors, and overall village network structure. Statistical analysis is conducted using correlation coefficient methods, which precisely measure the strength of linear relationships between two variables numerically.

Data processing

Spatial distribution vector analysis and computation using ArcGIS 10.3, visualization of spatial network relationships, and application of natural break method. Statistical analysis and computation of data using Excel. Weighting of evaluation indicator systems for traditional villages in Qiandongnan using Analytic Hierarchy Process (AHP) through Yaanp. Spatial network structure computation using UCINET. Correlation analysis using SPSS.

Results

Spatial distribution analysis of traditional villages in Qiandongnan

Through calculation, the standard deviation and mean of Thiessen polygon areas are 138.622 km² and 73.147 km² respectively, with a coefficient of variation Cv of 189.512%. This indicates that traditional villages in Qiandongnan exhibit clustered distribution, primarily influenced by natural factors. As depicted in Figs. 4, 5, the spatial distribution of traditional villages shows a clustered pattern characterized by ‘overall distribution with multiple concentrated areas’. These villages are predominantly located around the Leigongshan and Yuèliangshan mountain ranges. The average elevation of traditional villages is 717.8 m, lower than the Qiandongnan average of 1177.9 m but considerably higher than the regional minimum of 137.1 m. This suggests that the spatial selection of traditional villages emphasizes both defensive considerations and the convenient utilization of forest resources.

Fig. 4: Thiessen polygon analysis of traditional village spatial distribution.

Thiessen polygons were generated based on village point locations to identify spatial influence areas and proximity relationships among villages.

Fig. 5: Kernel density analysis of traditional villages in Qiandongnan.

Kernel density estimation was applied to identify areas of village concentration, reflecting spatial clustering patterns.

As shown in Fig. 6, traditional villages are predominantly distributed in the transitional zones between low mountainous and hilly areas and river valleys, where both transportation and agricultural production were important considerations for their settlement patterns. In the Qiandongnan region, traditional villages are primarily inhabited by the Miao and Dong ethnic groups. Miao villages tend to be located in mid-altitude mountainous areas, while Dong settlements are more commonly situated along river valleys. This distribution pattern reflects their distinct historical experiences: Miao villages, having migrated multiple times from north to south, often sought refuge in remote mountains to avoid warfare and feudal oppression. In contrast, Dong villages have historically settled in the humid river valleys of southern China. These historical contexts illustrate that Miao village site selection follows the characteristic of “settling in mountains for defense against external threats,” while Dong villages demonstrate a tendency toward “cluster-based settlement emphasizing social interaction and transportation convenience”.

Fig. 6: Geomorphological characteristics of the study area.

This figure presents the geomorphological conditions of Qiandongnan, illustrating terrain types and elevation patterns that influence the spatial distribution and development of traditional villages.

Table 2 Comprehensive evaluation index system for traditional villages in Qiandongnan

Comprehensive evaluation and development potential of traditional villages in Southeast Guizhou

Comprehensive evaluation index system for traditional villages in Qiandongnan (Table 2 shows)

The comprehensive evaluation index system for traditional villages in Qiandongnan consists of three primary indicators: production conditions, living conditions, and ecological conditions, with seven secondary indicators. The primary weights are highest for production conditions (0.58), followed by living conditions (0.28), and lowest for ecological conditions (0.14), indicating that production conditions are the primary factor. The secondary indicator weights are highest for tourism development status (0.44), and lowest for landscape environment (0.05) and architectural quality (0.04), suggesting that tourism is the dominant industry in traditional villages in Qiandongnan.

Development potential index system for traditional villages in Qiandongnan

Table 3 illustrates that the development potential index system for traditional villages in Qiandongnan consists of two primary indicators: resource endowment and market development environment. It includes seven secondary indicators such as traditional architectural value and village layout, and sixteen tertiary indicators like historical longevity and scarcity. The market development environment holds a higher weight (0.67) compared to traditional village resource endowment (0.33). Tertiary indicators with significant weights include accessibility to important transportation nodes (0.26), demonstration evaluation (0.12), and the presence of hotels and restaurants (0.09). This indicates that factors such as accessibility, demonstrativeness, and economies of scale are important components of development potential.

Table 3 Evaluation system for development potential of traditional villages

Analysis of comprehensive evaluation and development potential of traditional villages in Qiandongnan

According to Fig. 7, the number of traditional villages in Qiandongnan with high development potential far exceeds the number of villages with high comprehensive evaluations, demonstrating significant market investment and selection in the development of traditional villages in Qiandongnan. Contrasting with geographical locations, villages rated high in comprehensive evaluations tend to cluster more, while those with high development potential are more evenly distributed over a larger area, highlighting the market’s emphasis on regional balance. Table 4 indicates that the development potential of traditional villages in Qiandongnan is categorized into three levels: 89 villages with high potential (21.5%), 169 villages with medium potential (40.7%), and 157 villages with low potential (37.8%), indicating overall high development potential of the villages. Ethnicity and altitude do not show a clear relationship with village potential, suggesting they are not significant factors influencing development potential.

Fig. 7: Distribution of spatial development potential of traditional villages.

a Distribution of development potential; b Distribution of development potential classes.

Fig. 8: Gravity-based relationship network of traditional villages in Qiandongnan.

The gravity model illustrates the intensity of interactions between villages. Nodes represent villages, and weighted links indicate the strength of inter-village relationships.

Table 4 Analysis of competitiveness in traditional villages

Network analysis of traditional villages in Southeastern Guizhou

Spatial gravity relationships of traditional villages in Southeastern Guizhou (Fig. 8)

The spatial gravity relationships of traditional villages in Southeastern Guizhou exhibit characteristics of a clustered network structure. Within these clusters, there exist hierarchical gravity relationships, highlighting strongly connected central village clusters and weakly connected peripheral village clusters. Inter-cluster connections are relatively weak and structurally sparse. Strong connections between villages across administrative boundaries are commonly observed, thus confirming the viewpoint that the development of village clusters should not be delineated solely by administrative boundaries. Table 5 indicates that the number of effective gravity relationships is 6205, accounting for 7% of the total, suggesting that villages primarily interact with nearby ones. Among these effective gravity relationships, those with low gravity (45%) and medium-low gravity (21%) are predominant, while high gravity relationships are the least frequent (7%), indicating limited mutual influence among villages. High-intensity gravitational relationships are predominantly distributed in the central-western part of Qiandongnan. These areas feature dense villages with frequent interactions, forming high-density village clusters centered around strong gravitational forces. These clusters typically correspond to traditional villages with convenient transportation, large population sizes, and rich cultural resources, serving as “core nodes” in the regional interaction network. Medium-intensity gravitational relationships mostly connect secondary nodes surrounding the core areas, functioning as links between hubs and transitional zones. These villages primarily facilitate resource flow and cultural exchange within their respective clusters but exhibit limited cross-cluster connectivity. Low-intensity gravitational relationships dominate the overall network and are mainly found between peripheral and mountainous villages, reflecting characteristics of spatial dispersion and fragile connectivity. Such weak gravitational relationships are significantly influenced by topographic barriers and transportation conditions, indicating generally limited interactive capacity among these villages.

Table 5 Characteristics of traditional village development potential distribution

Traditional village network model in Qiandongnan

According to Fig. 9, the traditional village network exhibits a radial structure with multiple clusters of varying sizes. There are individual villages connecting between clusters, but these connections are relatively weak. The region includes 5 isolated village nodes without connections, consistent with the network structure shown in Fig. 9. A comparison between the weighted distribution of gravitational values and the binary network revealed a similar clustering structure, confirming that the binarization retained the essential spatial relationships among villages.

Fig. 9: Network structure model of clustered traditional villages.

Topology model of inter-village cluster interactions, illustrating the network relationships formed among village clusters.

Analysis of the overall structure of traditional village networks in Qiandongnan

Figures 10, 11 indicate that traditional villages in Southeastern Guizhou exhibit a hierarchical structure with three levels and eight clusters of adjacent subgroups. These subgroups vary in size, and closely related pairs of clusters form a second level comprising four clusters. Figure 12 reveals that within the overall village structure, there is a distinct core-periphery pattern. Villages in core positions account for 18% of the total number of villages and are located in regions associated with clusters 7 and 8.

Fig. 10: Identification of cohesive subgroups in the village network.

Cohesive subgroups were identified based on adjacency relationships within the village network. Nodes represent villages, and numerical labels correspond to village identifiers.

Fig. 11: Agglomeration structure of village network subgroups.

The figure illustrates the spatial and structural aggregation of cohesive subgroups, highlighting intra-group connectivity patterns.

Fig. 12: Core–periphery structure of the traditional village network.

Nodes are classified into core and peripheral components based on connectivity intensity, revealing the hierarchical organization of the village network.

Stability analysis of traditional village networks in Qiandongnan

From Fig. 13, it can be observed that the highest K-core is 24, while the lowest is 0. The proportion of K-cores above 12 is 17.9%, indicating overall low network stability. The sizes of K-cores show regional distribution characteristics, with high K-core clusters tending to aggregate nearby, and a process where high K-core settlements expand outward towards lower K-core settlements.

Fig. 13: Spatial distribution of K-core values in the village network.

The K-core analysis identifies structurally stable subsets of the network. Higher K-core values indicate stronger embeddedness within the network structure.

Completeness analysis of traditional village networks in Qiandongnan

In the network relationships of traditional village clusters in Southeastern Guizhou, the network density value is 0.04, indicating a relatively low spatial network density and a loosely organized network structure. The minimum average distance in the network is 6.95, suggesting low interconnectivity between networks. The clustering coefficient is low (0.64), indicating that the small-world characteristics of the network structure are not very prominent. Overall, the network completeness is moderate.

Equilibrium analysis of traditional village networks in Qiandongnan

The degree centrality in the traditional village networks of Southeastern Guizhou is 10.28%, indicating weak network equilibrium. Figure 14 shows that nodes with high degree centrality values are not uniformly distributed across each cluster but are primarily located in the left-side network structure, indicating an overall imbalance in network relationships, with the western cluster having a denser network structure.

Fig. 14: Degree centrality distribution of traditional villages.

Degree centrality values reflect the number of direct connections for each village, indicating differences in network influence and connectivity.

Vulnerability analysis of traditional village networks in Qiandongnan

As shown in Fig. 15, the village with the highest cut-point effect is Village No. 263 (Qingjiang Village), which can potentially disconnect up to 5 villages from the overall network. Cut-points affecting three or more villages and disconnecting them from the network account for 1% of the total. There are 13 cut-points identified in the network, accounting for 3% of the total, indicating relatively low overall network vulnerability.

Fig. 15: Distribution of cut-points in the village network structure.

Cut-points represent critical nodes whose removal would fragment the network, highlighting structurally vulnerable positions.

Analysis of individual competitiveness in traditional village networks in Qiandongnan

As shown in Table 6, traditional villages are classified into five levels based on their individual network competitiveness. Villages with high and moderately high competitiveness are the least common (comprising a total of 9%), while villages with low competitiveness are the most common (comprising 36%). Higher competitiveness correlates with greater effective scale and efficiency of individual networks, with lower limitations. As depicted in Fig. 16, villages with high and moderately high competitiveness are distributed around the geographical center of village clusters, not restricted to any specific region, and exhibit strong gravitational relationships with surrounding villages.

Fig. 16: Distribution of individual network competitiveness.

Network competitiveness reflects each village’s relative structural advantage within the network, integrating connectivity and positional importance.

Table 6 Statistics of village gravity relationships

Analysis of individual characteristics of traditional village networks in Qiandongnan

According to Fig. 17, there are 42 traditional villages that possess high and moderately high network coordinator functional roles, accounting for 10% of the total. These villages are distributed across six subgroups: 1, 2, 4, 5, 7, and 8 (see Fig. 9 for details). This indicates that these six subgroups all have villages exerting a high influence on the circulation of goods and information within their respective subgroup networks.

Fig. 17: Spatial distribution of coordinator roles in the village network.

Coordinator nodes exhibit strong intra-group connectivity and facilitate information exchange within subgroups.

According to Fig. 18, it can be observed that there are 15 traditional villages in this study that share both gatekeeper and spokesperson functional roles at high and moderately high levels, accounting for 3.6% of the total villages, which is relatively few. A comparison with Fig. 9 reveals that these villages are distributed mainly within subgroups 3, 5, and notably subgroup 7 (comprising 80% of them). This suggests that subgroup 7 becomes the core area of the entire network due to the presence of villages that ensure strong connectivity both within and outside their respective subgroups.

Fig. 18: Distribution of gatekeeper and spokesperson roles in individual networks.

Gatekeepers control connections between subgroups, while spokespersons represent group interests in external interactions.

According to Fig. 19, there are 6 traditional villages in this study that possess high and moderately high consultant functional roles, accounting for 1.9% of the total villages, which is quite few. A comparison with Fig. 9 reveals that these villages are distributed within subgroups 2, 5, and 7. While these nodes do not strengthen internal subgroup connections, they facilitate connections with external subgroups and strengthen connections with them.

Fig. 19: Spatial distribution of advisor roles in the village network.

Advisor nodes are characterized by strong external connections while maintaining limited intra-group interaction, providing strategic linkage functions.

According to Fig. 20, there are 3 traditional villages in this study that possess high and moderately high ‘liaison officer’ roles, accounting for 0.7% of the total villages, which is very few. Approximately 90% of the villages do not have liaison officer functionality, indicating that villages acting as third-party constraints on contacts between different subgroups and resource flows are scarce. This reaffirms the overall low fragility of the network.

Fig. 20: Spatial distribution of liaison roles in the village network.

Liaison nodes connect multiple subgroups without strong affiliation to any single group, supporting cross-group coordination.

The relationship between development potential of traditional villages in Qiandongnan and network structure

According to Fig. 21, the development potential of traditional villages, the overall network of traditional villages, and individual village networks are positively correlated in pairs. The development potential is highly positively correlated with the effective network size within individual village networks, as well as with the frequency of roles such as coordinators, gatekeepers, spokespeople, and advisors. At the node level, villages with larger effective network sizes and more frequent intermediary roles—such as coordinators, gatekeepers, spokespersons, and advisors—demonstrate higher development potential due to their greater activity in resource exchange and information flow. This indicates that increasing the effective network scale and the frequency of such intermediary roles can enhance the development potential of individual villages. At the overall network level, development potential is positively correlated with degree centrality, core value, k-core index, and subgroup affiliation. Villages located within structurally stable and well-connected network cores possess stronger growth advantages and coordination capacity, contributing to the stability, completeness, and efficiency of the entire system. Thus, improving network stability, balance, and subgroup connectivity can effectively strengthen the overall development potential of traditional village systems. These findings reveal an inherent interdependence among the overall network, individual networks, and development potential. Subgroup analysis based on k-core and modularity classification identifies three types of village clusters. Core clusters, primarily distributed along transportation corridors and county centers, show high connectivity, strong governmental support, and abundant cultural resources, resulting in the highest development potential. Intermediate clusters, located in transitional zones between core and peripheral areas, have moderate linkage intensity and serve as bridges for inter-group cooperation. Peripheral clusters, mostly situated in remote mountainous regions, exhibit sparse connections, low accessibility, and limited access to external resources, leading to weaker development potential.Therefore, enhancing network stability, improving transportation conditions, and promoting inter-cluster collaboration can effectively increase the comprehensive development potential of traditional villages in Qiandongnan.

Fig. 21: Relationship between development potential and network structure of traditional villages.

This figure illustrates the association between village development potential and network structural characteristics, indicating the influence of network position on development outcomes.

Conservation and utilization system for clustered traditional villages in Qiandongnan

From the previous text, it is evident that the traditional villages in Qiandongnan are distributed in clusters. There exists a positive correlation among the competitiveness and node role importance within the individual networks, the stability, completeness, and balance of the overall network, and the development potential of the villages. The structural form and operational characteristics of the overall network of villages can exert a regulatory effect on the individual networks and the development potential of the villages. Therefore, after determining the spatial distribution characteristics of the traditional villages and analyzing the spatial network structure and development potential of the village clusters, the planning of the spatial structure for the conservation and utilization of the traditional village clusters, as well as the planning of graded and categorized conservation and utilization measures, will be conducted. This will range from the overall village cluster area to individual villages. The construction of the conservation and utilization system for clustered traditional villages in Qiandongnan will be based on a step-by-step approach involving overall planning, individual implementation, and graded categorization of conservation and utilization measures. During the management process, local governments can utilize the Cluster Protection and Utilization System (CPUS) to delineate spatial development pathways and schedule implementation sequences based on the spatial structure of traditional village clusters. Priority should be given to the construction of key node villages, while each cluster unit should be guided through mutual support and coordinated development. Specific protection and development measures should be planned and categorized according to the current characteristics and resource attributes of each village.

Zoning of the Spatial Structure for the Protection and Development of Traditional Villages in Qiandongnan: Based on current network characteristics, relevant relationships, and developmental needs, clustering units and key village delineations are determined. Sub-clusters are delineated based on agglomerative clustering within the network structure, aligned with administrative management requirements. Key villages are identified based on important nodes and regional development needs within the network. As shown in Fig. 22, the delineation of clustered units for concentrated and contiguous protection and development of traditional village areas is based on considerations of agglomerative sub-groups, management scales, and administrative boundaries, resulting in 8 clustered units. This approach, compared to administrative divisions at the city and county levels resulting in 16 administrative management areas, significantly facilitates coordinated protection and development of village clusters.

Fig. 22: Cluster-based management units of traditional villages in Qiandongnan.

Traditional villages are grouped into management units based on spatial proximity and network relationships, providing a basis for clustered governance strategies.

Figure 23: Based on Core-Periphery Structure, Higher-Level Planning, and Current Resource Distribution, the 8 clustered units are classified into: Core Driving Zone (high overall development potential of villages with a core network structure), Boutique Development Zone (high stability in network structure and moderate overall development potential of villages), Deep Experience Zone (lower stability in network structure and average overall development potential of villages). Additionally, based on individual village competitiveness ranking (42 villages) and locational demands (3 criteria), 45 villages are designated as key development villages within the clusters. Among them, 9 villages serving as high-value nodes for advisor, coordinator, and gatekeeper functions are identified as regional priority villages. This approach enhances external connectivity functions while leveraging existing strengths of village clusters.

Fig. 23: Functional structure planning of traditional villages in Qiandongnan.

The figure presents the proposed functional structure for traditional village development, integrating conservation priorities, network hierarchy, and regional coordination.

Planning enhancement of traditional villages in Qiandongnan

Key villages need to enhance their competitiveness and role significance by improving village layout, government support, locational conditions, traditional architectural value, and residential environment to boost their development potential. The increased development potential of key villages will inevitably elevate their individual network importance. This enhancement of individual networks will in turn drive overall network improvement, ensuring sustained growth in village development potential. Through planning and enhancement of key villages, the overall network density has increased from 0.04 to 0.09. The average distance between networks has decreased from the original 6.95 to 2, significantly enhancing overall network connectivity. Moreover, the proportion of K-core nodes with 12 or more connections has risen from 29.6% to 40%, indicating improved stability. The degree centrality has increased from 10.28% to 19.64%, enhancing network balance.

Table 7 Protection levels and measures

Table 8 Classification of traditional village development in Qiandongnan

Graded protection and categorized development (Tables 7 and 8)

Based on the assessment of traditional village resources endowment, the protection levels of traditional village roads are categorized into three levels, with Level 1 receiving the strongest protection efforts and Level 3 the weakest.

Traditional villages are mainly categorized into five types: traditional farming, specialty product-oriented, ecological landscape, education and popular science, and boutique homestay. Based on rural planning structure, development levels are designated, with key villages leveraging resource advantages, defining development directions, strengthening connections with other villages in industries and culture, and playing a leading role. Ordinary villages focus on enhancing infrastructure, improving living environment, and service levels.

Discussion

The uneven distribution of the ethnic minority population in Guizhou Province has led to an imbalance in the distribution of traditional villages. The spatial distribution pattern, characterized by “large local aggregation and small overall dispersion,” synchronizes with the spatial distribution of ethnic minorities, with Qiandongnan being one of its high-density core areas³³. This is consistent with the findings of this study, indicating that the distribution of traditional villages in Qiandongnan is notably clustered and ethnic in nature. Within these clustered areas of villages, geographic location and village conditions are critical factors influencing the strength of network formation²¹. This study’s findings align with these principles, emphasizing altitude as a primary influence on the gravitational relationships among village clusters. Using social network analysis methods, scholars have analyzed the networked spatial structures of county-level village clusters, revealing characteristics such as low network density, poor completeness, and relatively good stability³⁴. Similarly, the network structure of traditional village clusters in Southeastern Guizhou conforms to patterns of low density, unevenness, and weaker completeness. However, the stability of these networks is lower, This also indicates that the flat terrain in the plains areas is more conducive to the stable development of settlements. Research on fortress clusters conducted abroad has shown that most fortress clusters exhibit significant small-world characteristics. Despite the large number of network nodes, the shortest paths between any two nodes are relatively short, allowing for rapid information dissemination. In contrast, the traditional villages in Qiandongnan display less pronounced small-world characteristics, and the connectivity between network nodes is relatively low. The main reasons for this are attributed to the karst topography of the region, which presents complex terrain and substantial elevation differences between villages. This increases the difficulty of connections between villages. Additionally, the strong defensive characteristics of fortresses require enhanced practice, while the traditional villages in Qiandongnan primarily serve as aggregation points for villagers’ daily lives, resulting in lower connectivity needs. Through internal relationship analysis of village clusters, scholars have observed that network relationships are not hindered by administrative boundaries^35,36,37. This study confirms the phenomenon of sub-clusters crossing administrative borders in the analysis of traditional village networks in Southeastern Guizhou, indicating that factors facilitating village-to-village development transcend administrative divisions, emphasizing proximity and comprehensive conditions. Research by Zou Huijun indicates that villages with higher overall quality can radiate and drive other nodes. This study’s analysis supports this finding and further reveals a positive correlation between overall network, individual network, and developmental potential of nodes. Adjusting network relationships can promote village development, underscoring inherent connections among overall network, individual network, and node attributes.Some scholars advocate for the development of regional village-town systems focusing on functional zoning and the cultivation of key nodes³². This study agrees with this perspective, proposing a framework for the cluster-based protection and utilization of traditional villages, enhancing research rigor and comprehensiveness in methodological approaches and content.

The study reveals inherent correlations between the overall network of the same research subject and the individual networks of its nodes, facilitating the adjustment of spatial structure through network relationships to match spatial demands and supplies. It also identifies inevitable intrinsic connections among individual networks, the overall network, network structures, and the developmental potential of the research subject. The morphological structure and operational characteristics of the overall network can modulate the effects on individual networks. Correspondingly, individual networks, based on self-formed connection information, can characterize “individual webs” and “individual neighborhoods,” reflecting partial aspects of the general characteristics of the overall network. This representation transforms complex and variable relational forms into certain network configurations. Subsequently, it elaborates on the significance of these configurations and their changes for individual and overall structures. Based on these intrinsic relationships, adjusting network relationships to achieve coordinated development among villages is the theoretical logic of this study. Focusing on traditional villages in Qiandongnan as the research object, the study employs methods such as Thiessen polygons to understand current distribution characteristics and influencing factors. It constructs an evaluation system based on field surveys, quantifies and visualizes traditional village operational relationships as network relationships using gravity models, and studies village network characteristics through complex network analysis. It analyzes relationships among the overall network, individual networks, and village development potential using correlation analysis methods, dissecting their intrinsic relationships through network structure analysis. Controlling network structure based on internal network interactions and higher-level planning requirements, it builds a cluster-based protection and utilization system for traditional villages in Qiandongnan through spatial structure planning and hierarchical classification planning. The entire system exhibits spatial integrity and hierarchy (overall network, subgroups, individual networks). The process is logically coherent with clear methodological steps, meeting practical needs. The constructed network system can be used for systematic detection and early warning of traditional villages and possesses strong practical manipulability. This method can be widely applied to the protection and utilization of cultural heritage in other areas globally (such as the 17 other global eco-cultural protection zones), facilitating cross-regional cooperation by delineating cultural heritage relationships at a macro scale. Furthermore, it ensures the comprehensiveness and scientific rigor of the cluster-based protection and utilization system at various scales—from medium-scale protection and utilization based on cohesive subgroups to micro-scale determination of specific village development directions. Therefore, the methodological approach of constructing the cluster-based protection and utilization system for traditional villages in this study is scientifically grounded due to its holistic, manipulable, and replicable research framework.

The Guizhou Dongnan National Traditional Village Cluster is a nationally recognized aggregation of traditional villages characterized by a geographic distribution pattern of “comprehensive coverage with multiple concentrated patches.” It exhibits typical mountainous, ethnic, and complex features. Analyzing the mutual gravitational relationships between villages and constructing a network structure model revealed that the regional network of village clusters exhibits a multi-cluster radial structure. The village network itself also displays a radial structure with three levels and eight subgroups, highlighting a core-periphery structure. Through calculations involving the overall village network, individual village networks, and development potentials, it was found that village development potential, overall village network, and individual village network are positively correlated in pairs. Government support intensity, geographical location, and traditional architectural value were identified as primary influencing factors. Based on these findings, the study established a quantitative and clustered protection and utilization system, following steps such as defining spatial distribution types, revealing spatial network structures, exploring spatial resource endowments, delineating spatial structures, implementing individual spatial positioning, and formulating classification and grading development strategies. This approach facilitated coordinated management from the cluster level down to individual villages, aiming to provide specific guidance for the management and development of traditional villages in Guizhou Dongnan, while also serving as a reference for the protection and utilization of other concentrated development areas of traditional villages domestically or globally, including cultural heritage sites and clusters of cultural relics.

Network analysis is constrained by data availability, particularly regarding social interactions between villages. Future research could employ dynamic or multi-layer network models to capture evolving spatial connections and incorporate socioeconomic data to enhance precision. Additionally, the use of weighted networks is recommended to provide more nuanced insights and enable deeper investigation into the intensity of spatial interactions.