Study on Urban System Relationships and Resilience Promotion Strategies in Underdeveloped Mountainous Areas Based on Social Network Analysis: A Case Study of Qiandongnan Miao and Dong Autonomous Prefecture

Abstract

Urban systems are the spatial carriers of social and economic relations at the regional level, and their relational and structural resilience are key to regional coordination and sustainable development, attracting widespread attention from scholars. In order to analyze the internal relationships of urban agglomerations in underdeveloped mountainous regions and optimize their spatial resource allocation and resilience, this study takes the urban agglomeration of Qiandongnan in China as an example and researches their internal relationships, development potential, and influencing factors based on quantitative methods such as social network analysis. The results show that the urban cluster in Qiandongnan presents “large dispersion and small aggregation” distribution characteristics, with the karst landscape as the main influencing factor; the spatial network exhibits a scale-free morphology with an obvious core–periphery structure, demonstrating moderate stability but poor completeness, weak equilibrium, and low overall resilience; only 15.61% of nodes demonstrate high competitiveness; urban units with functional roles serve as critical network nodes; urban units’ development potential is divided into three tiers (with 47.31% being medium-high), although overall levels remain low; and the development potential, overall network, individual network, and network resilience of urban units are all positively correlated, with economic and transportation development conditions being the main influencing factors. Based on the abovementioned findings, this study proposes a “multi-level resilience promotion path for network structure optimization”, which provides a theoretical basis and optimization control methods for the reconstruction and synergistic development of urban agglomerations. It also serves as a reference for the development planning of urban systems in other underdeveloped mountainous regions.

1. Introduction

Under the dual challenges of rapid urbanization and global environmental change, urban resilience has become a core issue in regional sustainable development. The resilience of an urban system depends not only on its ability to cope with shocks but also on the functional attributes and cooperative relations of key nodes in its network structure. As a complex system composed of multilevel and multifunctional nodes, the resilience of the urban system reflects the functional embodiment of its network structure, which determines the system’s ability to cope with, absorb, and recover from external shocks and risks, as well as to restore, maintain, or improve the original structure and key functions [1]. Therefore, rational planning and resilience enhancement of the urban system are crucial for the sustainable development of the region.

As an organic overall system composed of cities of different scales and functions, the resilience of an urban system depends on three key elements: the pivotal role of key nodes, the complementarity of functional nodes, and the robustness of network structure. Additionally, adjusting and optimizing the urban system structure is a key measure for promoting regional coordination, urban–rural integration, and rural revitalization [2]. Reasonable urban system structure deployment and resilience promotion not only improve the coordination and development potential of urban functions but also enhance the efficiency of resource allocation and promote the balanced development of the regional economy. Therefore, quantitative and visual research on urban relations based on network topology is particularly important. Existing research has several limitations: At the theoretical level, although it has surpassed the traditional center-place theory [3] and introduced the floating space theory [4] and the global urban network perspective [5], it still pays too much attention to the economic dimension [6]. There is a lack of systematic analysis of the internal relationships within urban agglomeration networks, and there is a lack of research on resilience. In terms of research methods, although the topological measurement of the social network analysis method and the system resilience index from network science have been applied to urban research [7,8] and infrastructure resilience assessment [9], respectively, they rarely achieve method integration in the study of mountain town systems and fail to deeply explore the synergistic mechanisms between functional attributes and network topology. In terms of research geography, Chinese studies have mainly focused on developed plains such as the Yangtze River Delta [6] and Pearl River Delta [10], focusing on economic linkages and spatial structure optimization, while international studies have focused more attention on the global urban network [5] and the synergistic development of metropolitan areas [11]. Research on underdeveloped mountainous areas is rare, and the institutional needs for policies for less developed regions are more urgent due to economic and complex terrain constraints. At the same time, the emerging research on urban resilience mainly focuses on regional urban safety [12], urban climate disaster prevention and control [13], urban disaster resistance [14], and so on. Although it provides a new idea of “anti-interference-adaptation-transformation” for urban network analysis, most of the achievements are still focused on the resilience evaluation of single cities, lacking in-depth research on the structural resilience of urban system networks. Therefore, this study innovatively constructs a multidimensional analysis framework. Theoretically, it integrates the spatial hierarchy idea of central place theory [3,15] and the network thinking of flow space theory [4], and at the same time, it introduces the resilience concept of “resistance to disturbance-adaptation-transformation”. Methodologically, Tyson’s polygon and kernel density analysis are used to analyze the spatial distribution characteristics of urban units (defined as county-level cities, cities, townships, and towns), integrating the topological metric tools of social network analysis [7] and the systematic thinking of network science [16], and adapting the methodology to the unique characteristics of the mountainous towns and cities network.

Qiandongnan Miao and Dong Autonomous Prefecture (hereinafter referred to as Qiandongnan), located in southwest China, is a typical province in China’s poverty alleviation efforts. It is economically underdeveloped, and lacking plains within its territory, it features a typical karst topography. Therefore, in this study, taking the urban cluster in Qiandongnan as an example, Thiessen polygon and kernel density analyses are used to analyze the spatial distribution characteristics of the urban cluster. A network model of the urban cluster is constructed based on topological relations and the gravity model, and the network relationship between urban clusters is quantitatively and visually analyzed. On this basis, the development potential of urban clusters is evaluated, and the relationship between their network and development potential is analyzed. Based on the relationships among urban spatial distribution characteristics, network relationship structure characteristics, and development potential characteristics, a “network structure optimization multi-level resilience promotion path” is proposed by constructing the spatial structure planning of “hierarchical identification-functional coordination-structural optimization” and the development planning of “resilience evaluation-resilience promotion-differentiated development”. This provides a new analytical perspective for deeply understanding the structural characteristics and resilience mechanisms of urban systems in underdeveloped mountainous areas. The multi-level resilience promotion of “core node leading, axis radiation driving, differential development of functional subgroups, and function improvement of individual towns” has effectively promoted the systematic optimization of the urban system from macro-group coordination to micro-individual management and control. This study not only expands the application of social network analysis in the study of urban resilience but also provides a theoretical reference and a practical paradigm for the sustainable development of towns in similar underdeveloped mountainous areas, effectively making up for the shortcomings of existing research in the study of urban network resilience in underdeveloped mountainous areas.

2. Research Object and Research Method

2.1. Research Object

Qiandongnan is located in southeastern Guizhou Province, which is an underdeveloped province (Figure 1a,b). This region has a typical karst landform with no plains, a low land reclamation rate, a large forest area, high forest coverage, and prominent ecological functions; however, the total amount of construction land is low. Additionally, 87% of the region is mountainous terrain, with an average slope of 28° and a relative height difference of 800–1200 m, leading to an average increase in transportation time costs between towns and cities by 2.8 times compared to straight-line distances [17]. This extreme topographical condition magnifies the constraining effect of network structure on town development, and at the same time, as a less-developed region, Qiandongnan has a higher sensitivity to planning interventions. It is the largest of China’s 30 autonomous prefectures in terms of total population and ethnic minority population and is a national comprehensive pilot area for new urbanization. Therefore, Qiandongnan serves as an “ideal case study” for examining urban network systems in underdeveloped mountainous regions. This research selects the Qiandongnan urban cluster as the study object, which comprises 1 county-level city, 15 counties, 60 townships, and 129 towns (Figure 1c). The findings will not only contribute to refining theories of urban network systems but also provide practical guidance for the sustainable development in underdeveloped mountainous areas, demonstrating significant academic value and practical implications.

2.2. Research Method

2.2.1. Overall Research Framework

Using GIS spatial analysis and social network analysis to quantify and visualize the multidimensional interactions between urban elements (economy, culture, and ecology) as a network structure model, this study analyzes the relationships between individuals and the overall network within the internal network based on indicators such as degree of centrality, core–periphery structure, structural holes, etc. Finally, by combining the resilience theory with the requirements of the upper planning level, this study proposes a planning path of “hierarchical reconstruction-functional coordination -structural optimization-resilience enhancement”, which realizes a closed-loop research logic from theoretical modeling to policy implementation. Many research results show that there is an inherent and inevitable relationship between the whole network and the individual network, as well as their node attributes, of the same research object [18,19,20]. The relationship among the whole network, the individual network, and network resilience is that changes in the whole network structure will cause changes in some individual network structures. Accordingly, the change in the individual network structure will also cause a change in the overall network structure. At the same time, the attributes of network nodes (the attributes of nodes in this study are the development potential of urban units) are controlled by the network structure relationship, the network controls the performance of node attributes as a whole, and the optimization of node attributes will affect the morphological and structural characteristics of the network as a whole. Therefore, based on the principles of multi-discipline, multi-scale, and quantification, this study proposes a “multi-level resilience promotion path for network structure optimization” in a more comprehensive and scientific manner, as shown in Figure 2. This path can accurately capture and effectively address the complex interaction between the overall network and individual networks and promote the sustainable development of urban clusters.

2.2.2. Urban Spatial Distribution Analysis Method

(1)

Thiessen polygon method

The Thiessen polygon (also known as Voronoi) is a spatial planar sectioning method that reflects the distribution pattern of point elements through its geometric features. This study uses the Thiessen polygon to analyze the spatial distribution type of urban units. Since the area of the Thiessen polygon changes with the distribution of the point set, the coefficient of variation $Cv$ value can be used to measure the relative degree of change in the area of the convex polygon, thereby allowing for the analysis of the spatial distribution type of the sample points [21]. $Cv$ is equal to the ratio of the standard deviation of the area of the Thiessen polygon to the mean value, and its calculation formula is as follows:

$$R=\sqrt{\sum {\displaystyle \frac{{\left({\mathrm{s}}_{\mathrm{i}}-\mathrm{s}\right)}^2}{\mathrm{n}}}}(i=\mathrm{1,2},\dots ,n)$$

(1)

$$C_v=R⁄S$$

In the formula, $R$ represents the standard deviation of polygon areas, $Si$ denotes the area of the Thiessen polygon, S is the average area of the polygons, and n stands for the number of polygons. Duvckaerts proposed 3 recommended values to measure the relative degree of change in the region and, on this basis, categorized the distribution type of the sample points. When $Cv$ < 33% is a uniform distribution, the human factor is the main influence; when $Cv$ = 33–64% is a random distribution, the natural factor is the main influence; and when $Cv$ > 64% is a cluster distribution, the natural factor is the main influence [22].

(2)

Kernel density analysis method

Kernel density analysis using the kernel function can calculate the density of the point or line elements in the neighborhood, converting the discrete data into a continuous surface that reflects its spatial clustering degree and distribution features [23]. In this study, nuclear density analysis is used to analyze the distribution density relationships among urban units. The greater the nuclear density value, the denser the distribution of urban units. The calculation formula for nuclear density is as follows [24]:

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

(2)

In Formula (2), $n$ 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 $point(x,y)$.

2.2.3. Evaluation Method for Urban Development Potential

This study employs a mixed-methods approach, combining a literature review and field research (covering all 205 urban units in Qiandongnan). The Delphi expert consultation method was utilized to optimize and refine the initially constructed evaluation system for urban development potential. Subsequently, a comprehensive weighting method was applied to determine the index weights, resulting in the establishment of a scientifically robust evaluation index system.

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

(3)

In Formula (3), $X_i$ represents the numerical value computed for indicator $i$ after the assignment processing of its original data, and $W_i$ 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.

2.2.4. Analysis Method for Urban System Relationships

(1)

Social Network Analysis Method

The social network analysis approach examines complex systems by reducing them to a network structure of nodes and edges. Nodes represent the basic constituent units of the system, while edges reflect the interrelationships between these units [25]. In this study, the urban unit is a “point”, the gravitational relationship between urban units is a “line”, and the value of the gravitational force between urban units is greater than the average value of the gravitational force, mark 1, and less than the value of the gravitational force, mark 0. The relationship matrix is obtained to construct the “network” of urban units, and the precise quantitative calculation of various relationships is carried out. The calculation indices are shown in Table 1:

Table 1. Definitions of indicators related to the social network analysis method.

Indicator	Definitions
Cut-point	A “cut-point” is a special node whose removal leads to a break in the network structure and the formation of independent subgraphs. The number of such nodes is positively correlated with the vulnerability of the network structure, and the greater the share of cut-points, the lower the stability of the network as a whole [26].
Small-world characteristics	If the network is large but sparsely connected, with no core hub nodes, and exhibits a high degree of clustering characteristics, then such a network is called a small world. The shorter the average path length and the higher the clustering coefficient, the more significant the small-world characteristics are, and such small-world characteristics tend to be negatively correlated with network completeness [27].
Network density	Network density reflects the tightness of connections between nodes and is calculated as the ratio of the actual number of connections to the maximum possible number of connections. The lower the density, the looser the network, the weaker the connections between nodes, and the weaker the network’s completeness [26].
Cohesive subgroups—Block model	Block model analysis achieves spatial clustering by investigating the internal and external correlation characteristics of subgroups, dividing multi-level cohesive subgroups, and identifying the network function roles of each plate based on information transmission and reception [27].
K-core	The k-kernel indicates that each node in the subgraph is connected to at least k other nodes. By calculating the k-core with different K values (1, 2, 3…), the local stability of the network can be evaluated [28].
Core-periphery	The “Core-periphery” model can distinguish between the core area and the edge area of the network structure. Nodes with an average core value or above can be used as core nodes, which are in a dominant position in the network structure [29].
Degree centrality	Degree centrality is a measure of nodes’ ability to control network resources, and its value reflects the importance level of the nodes. When the number of nodes is evenly distributed in space, the network structure tends to be balanced [30].
Structural holes	A structural hole refers to the connection gap between non-redundant nodes, and the nodes occupying this position have competitive advantages. Through the quantitative evaluation of three dimensions (efficiency, efficiency, and constraint), when a node has high efficiency and low constraint, its structural hole advantage is more obvious, and its development potential is greater [31].
Broker	The broker refers to the actors in the middle position, which are divided into five roles according to their positions: coordinator, goalkeeper, representative, consultant, and liaison (Figure 3). The coordinator refers to the strong information dissemination ability only within the subgroup. The goalkeeper and stakeholders inside and outside the subgroup have strong information connection abilities. The representatives and stakeholders inside and outside the subgroup also have strong information connection abilities. The consultant does not communicate directly with the stakeholders inside the subgroup but maintains a good connection relationship with the stakeholders outside the subgroup. Additionally, the liaison officer maintains strong information communication abilities with multiple external subgroups [32].

Table 1. Definitions of indicators related to the social network analysis method.

Indicator	Definitions
Cut-point	A “cut-point” is a special node whose removal leads to a break in the network structure and the formation of independent subgraphs. The number of such nodes is positively correlated with the vulnerability of the network structure, and the greater the share of cut-points, the lower the stability of the network as a whole [26].
Small-world characteristics	If the network is large but sparsely connected, with no core hub nodes, and exhibits a high degree of clustering characteristics, then such a network is called a small world. The shorter the average path length and the higher the clustering coefficient, the more significant the small-world characteristics are, and such small-world characteristics tend to be negatively correlated with network completeness [27].
Network density	Network density reflects the tightness of connections between nodes and is calculated as the ratio of the actual number of connections to the maximum possible number of connections. The lower the density, the looser the network, the weaker the connections between nodes, and the weaker the network’s completeness [26].
Cohesive subgroups—Block model	Block model analysis achieves spatial clustering by investigating the internal and external correlation characteristics of subgroups, dividing multi-level cohesive subgroups, and identifying the network function roles of each plate based on information transmission and reception [27].
K-core	The k-kernel indicates that each node in the subgraph is connected to at least k other nodes. By calculating the k-core with different K values (1, 2, 3…), the local stability of the network can be evaluated [28].
Core-periphery	The “Core-periphery” model can distinguish between the core area and the edge area of the network structure. Nodes with an average core value or above can be used as core nodes, which are in a dominant position in the network structure [29].
Degree centrality	Degree centrality is a measure of nodes’ ability to control network resources, and its value reflects the importance level of the nodes. When the number of nodes is evenly distributed in space, the network structure tends to be balanced [30].
Structural holes	A structural hole refers to the connection gap between non-redundant nodes, and the nodes occupying this position have competitive advantages. Through the quantitative evaluation of three dimensions (efficiency, efficiency, and constraint), when a node has high efficiency and low constraint, its structural hole advantage is more obvious, and its development potential is greater [31].
Broker	The broker refers to the actors in the middle position, which are divided into five roles according to their positions: coordinator, goalkeeper, representative, consultant, and liaison (Figure 3). The coordinator refers to the strong information dissemination ability only within the subgroup. The goalkeeper and stakeholders inside and outside the subgroup have strong information connection abilities. The representatives and stakeholders inside and outside the subgroup also have strong information connection abilities. The consultant does not communicate directly with the stakeholders inside the subgroup but maintains a good connection relationship with the stakeholders outside the subgroup. Additionally, the liaison officer maintains strong information communication abilities with multiple external subgroups [32].

Figure 3. Schematic diagram of broker roles.

Figure 3. Schematic diagram of broker roles.

(2)

Network Resilience Assessment Method

Network resilience refers to a network’s ability to maintain its functionality, structure, and performance when subjected to external disturbances or internal failures. It measures the network’s stability, resilience, and adaptability when facing unexpected events such as attacks, node failures, or traffic surges [33]. Network resilience evaluation typically employs destructive testing methods, analyzing network performance by simulating various failure scenarios, including node failures, traffic surges, or malicious attacks. In this study, we adopt a node removal approach to progressively eliminate key nodes while observing changes in network density, connectivity, stability, integrity, and vulnerability, thereby identifying fragile links within the network. Through quantitative analysis of these indicators, we can evaluate the network’s overall resilience and subsequently provide a basis for network optimization and robustness enhancement. Building upon the cascading failure model [34] and dynamic equilibrium theory [35], this method quantifies the degree of performance degradation during sustained attacks. When performance decline remains below the 30% threshold, it indicates strong resilience, thus revealing fragile components and elucidating the resilience mechanism [36].

(3)

Minimum Spanning Tree Analysis (MST)

Minimum spanning tree is an algorithm in graph theory that is used to find a tree that contains all vertices in a weighted undirected graph, where the sum of the edge weights of this tree is the smallest [37]. In this paper, the Kruskal algorithm is used to construct the MST, find the optimal path connecting all towns, ensure connectivity between towns, and thus determine the axis of development.

2.2.5. Correlation Analysis Method

(1)

Analysis of Urban Gravity Relationship

By using the gravity model method, according to the field investigation and analysis, it can be shown that the better the comprehensive conditions of urban areas, the greater their influence on their surroundings, and the longer the commuting time, the weaker the influence. In line with the purpose of this study, in order to explore the comprehensive attraction between urban units, the formula of the gravity model is revised, and the algorithm formula is expressed as follows:

$$F_{ij}={\displaystyle \frac{r_i\times M_i\times M_j}{T_{ij}^2}}$$

In the formula, the value of $M$ is calculated using the index system in Table 3, $M_i$ is the comprehensive condition value of the urban area, $M_j$ is the comprehensive condition value of the urban area, T is the shortest commuting time between two places, calculated based on the Baidu map API, and ${ r}_i$ is the regional adjustment coefficient. In this study, 10,000 is used to facilitate the comparison of gravity values. For the calculation results, the average gravity value is taken as the breakpoint value. If it is greater than the average value, the data is considered valid; if it is less than the average value, the stress value of the gravity-free relationship between the urban units is considered to be zero.

At the same time, the commuting time between urban units is more than two hours, which is defined as an unattractive relationship. Considering the actual situation, the gravity results are corrected, and a gravity matrix (where the attractive relationship between two towns is defined as 1 and the unattractive relationship is defined as 0) is constructed to reflect the relationship between urban units in Qiandongnan.

(2)

Relationship Analysis of Related Elements

SPSS software (version 29.0) is used to calculate the development potential of urban units, the overall network and individual network structure of urban clusters, and the correlation coefficient method is used for statistical analysis. The correlation coefficient accurately reflects the strength of the linear relationship between the two variables in a numerical manner.

2.2.6. Data Sources and Processing

Data sources include the Statistical Yearbook of Southeast Guizhou (2022) and the Urban System Planning of the Miao and Dong Autonomous Prefecture of Southeast Guizhou (2015–2030). Using Arc GIS10.3, the spatial distribution vector analysis and calculation, along with the visual display of spatial network relations and the application of the natural breakpoint method are carried out. The statistics and calculation of data are carried out using Excel. The weight of the evaluation index system for towns in Qiandongnan is calculated using Yaanp, and the shortest commuting time between urban units is automatically calculated using the driving route planning (lightweight) service function on the Baidu map development platform. The spatial network structure is calculated using Ucinet (version 6.767), and the correlation is calculated using SPSS.

3. Results

3.1. Distribution Analysis of Urban Units in Qiandongnan

Through calculation, the standard deviation of polygon area in Thiessen is 68.68 km², and the average polygon area in the Qingshui River Basin is 158 km², while that in the Duliujiang River Basin is 213 km², which confirms the development mode of “mountain blocking and valley gathering”. The average is 185.09 km², which is significantly higher than that in plain areas (such as the average of 53 km² in the Yangtze River Delta in China), indicating that the cost of infrastructure services is high. The $Cv$ value = 37.01%, which indicates that the distribution of urban areas in Qiandongnan is random, and natural factors are the main influencing factors (Figure 4a). As can be seen in

Table 2. Spatial distributions of urban units in Qiandongnan and their influencing factors.

Types	Cv Value	Type of Spatial Distribution	Main Influencing Factors
County-level city and City	35.86%	Randomized distribution	Natural factor
Township	35.83%	Randomized distribution	Natural factor
Town	37.62%	Randomized distribution	Natural factor

, urban areas of different administrative levels are randomly distributed and dominated by natural factors. This indicates that natural conditions play a key role in the layout of urban areas, limiting the systematic and synergistic nature of town planning, which is not conducive to the efficient integration of resources and balanced development of the region and is the main reason for the underdevelopment of the region.

As can be seen in Figure 4a–c, Qiandongnan urban units are mostly distributed in the watersheds of the Qingshui River and the Duliu River, forming a “corridor-type” distribution of river valleys. The mountainous areas are lagging behind in terms of development, exposing a dependence on a single geographic condition and the lack of regional coordination. Additionally, being cut by the karst landforms, the urban units exhibit characteristics of “large dispersion and small aggregation”. The average elevation of the urban space is 590.25 m, which is lower than the average elevation of Qiandongnan (181.75 m) and far below the highest elevation in the region. This choice effectively avoids the potential threat of extreme climatic conditions and the geographical environment and is beneficial to people’s living and economic activities. This shows that convenient travel conditions and certain terrain construction are the foundation of settlement development.

3.2. Evaluation Index System of Towns in Qiandongnan

Combined with the comprehensive index system table of urbanization development in China’s urban system planning [38] and according to the actual situation of the Qiandongnan urban cluster and the collectability of data, the index is adjusted and optimized, and a scientific and applicable evaluation index system of urban comprehensive conditions is constructed. Table 3 shows that the comprehensive condition evaluation index system for urban units in Qiandongnan consists of four first-level indicators, namely, urbanization level, basic public service level, infrastructure level, and resource and environment level, along with 19 s-level indicators, including the urbanization rate of the permanent population and the urbanization rate of the registered population. The level of infrastructure (0.31) is the highest, the level of basic public services (0.28) is the second-highest, and the levels of urbanization (0.22) and resources and environment (0.19) are the lowest. This shows that the levels of infrastructure and basic public services are the main elements of urban development in Qiandongnan. Among the weights of the secondary indicators, the urbanization rate of the resident population (0.12) is the highest, and the proportion of green buildings in new buildings (0.04) is the lowest, which indicates that the most important factor influencing the development of urbanization in Qiandongnan is the speed and scale of population concentration in urban units.

Table 3. Evaluation index system of comprehensive conditions of urban units in Qiandongnan.

Primary Index	Weight	Secondary Index	Weight	Unit
Urbanization Level	0.22	Permanent resident urbanization rate	0.12	%
		Registered population urbanization rate	0.10	%
Basic Public Services	0.28	Proportion of migrant workers’ children receiving compulsory education	0.06	%
		Coverage rate of free basic vocational training for urban unemployed, migrant workers, and new labor force	0.06	%
		Coverage rate of basic pension insurance for urban permanent residents	0.05	%
		Coverage rate of basic medical insurance for urban permanent residents	0.06	%
		Coverage rate of subsidized housing for low- and middle-income urban households	0.05	%
Infrastructure Level	0.31	Proportion of public transportation in motorized travel	0.04	%
		Public water supply coverage rate	0.07	%
		Sewage treatment rate	0.05	%
		Domestic waste harmless treatment rate	0.06	%
		Household broadband access capacity	0.05	Mbps
		Coverage rate of community comprehensive service facilities	0.04	%
Resource and Environment Level	0.19	Per capita urban construction land area	0.05	M2
		Proportion of renewable energy consumption	0.03	%
		Proportion of green buildings in new constructions	0.02	%
		Ace ratio in built-up areas	0.03	%
		Proportion of days with air quality meeting national standards	0.05	%

Table 3. Evaluation index system of comprehensive conditions of urban units in Qiandongnan.

Primary Index	Weight	Secondary Index	Weight	Unit
Urbanization Level	0.22	Permanent resident urbanization rate	0.12	%
		Registered population urbanization rate	0.10	%
Basic Public Services	0.28	Proportion of migrant workers’ children receiving compulsory education	0.06	%
		Coverage rate of free basic vocational training for urban unemployed, migrant workers, and new labor force	0.06	%
		Coverage rate of basic pension insurance for urban permanent residents	0.05	%
		Coverage rate of basic medical insurance for urban permanent residents	0.06	%
		Coverage rate of subsidized housing for low- and middle-income urban households	0.05	%
Infrastructure Level	0.31	Proportion of public transportation in motorized travel	0.04	%
		Public water supply coverage rate	0.07	%
		Sewage treatment rate	0.05	%
		Domestic waste harmless treatment rate	0.06	%
		Household broadband access capacity	0.05	Mbps
		Coverage rate of community comprehensive service facilities	0.04	%
Resource and Environment Level	0.19	Per capita urban construction land area	0.05	M2
		Proportion of renewable energy consumption	0.03	%
		Proportion of green buildings in new constructions	0.02	%
		Ace ratio in built-up areas	0.03	%
		Proportion of days with air quality meeting national standards	0.05	%

Considering the research of domestic and foreign scholars on the development potential of cities and towns, aiming at the actual situation of the urban cluster in southeastern Guizhou and the collectability of data, the indicators are adjusted and optimized, and a scientific and applicable evaluation index system for urban development potential conditions is constructed.

Table 4. Evaluation index system for urban development potential in Qiandongnan.

Primary Index	Weight	Secondary Index	Weight	Unit
Social Development Conditions	0.16	Administrative area	0.05	Km2
		Number of village committees	0.02	Units
		Number of residential committees	0.02	Units
		Registered population	0.03	Persons
		Permanent resident population	0.04	Persons
Economic Development Conditions	0.30	Per capita disposable income	0.08	CNY
		Per capita GDP	0.06	CNY
		Total fiscal revenue	0.06	CNY
		Number of stores or supermarkets with ≥50 m2 business area	0.05	Units
		Per capita gross output value	0.06	CNY
Production Development Conditions	0.18	Cultivated land area	0.03	Hectares
		Total grain production	0.03	Tonnes
		Gross agricultural output value	0.03	CNY
		Gross industrial output value	0.05	CNY
		Number of industrial enterprises	0.04	Units
Transportation Conditions	0.20	Fixed-asset investment in transportation	0.10	CNY
		Commuting time to major central cities	0.11	Minutes
Natural Terrain Conditions	0.15	Elevation	0.05	Meters
		Slope gradient	0.07	Degrees
		Slope aspect	0.03	Degrees

shows that the index system of urban development potential in Qiandongnan consists of four first-level indicators: social development conditions, economic development conditions, traffic development conditions, and natural terrain conditions, along with 20 s-level indicators, including administrative area and the number of villagers’ committees. The first-level index indicates that the economic development condition (0.30) is the highest, the traffic development condition (0.28) is the second-highest, the social development condition (0.16) and the production development condition (0.18) are lower, and the natural terrain condition (0.15) is the lowest. This shows that the economic development condition is the core driving factor of the development potential of towns in Southeastern Guizhou, the traffic condition is an important factor, and the social development condition and the production development condition serve as important foundations, while natural topographical conditions have a relatively minor impact on the development of towns.

3.3. Analysis of the Urban Network in Qiandongnan

3.3.1. Network Model and Spatial Relationship of Urban Units in Qiandongnan

As can be seen in Figure 5, the urban network in Qiandongnan exhibits a scale-free feature of “overall looseness-local cluster”: a few hub nodes are connected in large numbers, while most nodes are sparsely connected. This structure is robust to random failures, with obvious characteristics. Once a hub node is damaged, it can easily lead to systematic collapse, indicating that the structural resilience is low. The existence of 10 isolated urban units in the region highlights shortcomings in transportation, economy, and public services, which increases the difficulty of uncoordinated development and planning and regulation.

Figure 6 shows that the urban cluster in Qiandongnan is mainly distributed in the relatively flat valley areas and regions with developed traffic conditions, indicating that the scale-free characteristics of the urban network are derived from the innate advantages of the nodes in the flat dam area of the valley and the positive feedback effect of traffic investment. This is manifested in a spatial polarization pattern, in which a few hub nodes monopolize the road network connections, and most marginal urban units are limited in their development. The existence of strong connection nodes further demonstrates that the connections between urban regional divisions and urban units are not obvious.

3.3.2. Analysis of the Overall Network of Urban Units in Qiandongnan

(1)

Analysis of the Overall Structure of the Urban Network in Qiandongnan: Figure 7a shows that there are three levels of cohesive subgroups consisting of eight clusters in the urban units of Qiandongnan, with subgroups of varying sizes, and four clusters of the second level can be formed by two closely related clusters.

Figure 7b shows that the eight cohesive subgroups indicate the existence of multiple functionally or spatially connected groups in the urban network of Qiandongnan, with significant differences in the strength of connections between different subgroups. The third-level subgroups belonging to the same second level are adjacent to each other in terms of location. The type of urban units in Qiandongnan has little influence on subgroup relationships, with location, internal connections, and geographic proximity being the main influencing factors. Subgroup 3 is loosely distributed. Although it is close to subgroup 5, with Kaili as the core, there is no gravitational connection between them, and the gravitational relationship with other subgroup nodes is very weak. The contrast map shows that its essence is the phenomenon of “geographical proximity and functional alienation”, which is caused by the triple constraints of alpine terrain barriers, traditional closed ethnic settlement forms, and lagging transportation infrastructure, This situation shows that administrative divisions are adjacent, but the gravitational connection intensity is weak, highlighting the coordinated development of urban units in mountainous multi-ethnic areas.

Figure 8a shows that the core–periphery structure is obvious in the overall structure, and the core nodes are usually those with multiple connections and strong connection attraction in the network. Compared with Figure 8b, it can be seen that the core nodes are located in all subgroup areas except subgroup 3, which reflects the uneven distribution of core urban units and the clustering of important core urban units. The core node urban units are dominated by county-level cities (accounting for 100%) and cities (accounting for 100.00%), which highlights the status of county-level cities and cities as the primary carriers of regional economic, cultural, transportation, and other foreign exchanges. This reflects their strong resource-gathering ability and comprehensive functions, causing them to occupy an important position in the regional network. However, the number of township nodes in the network is relatively high (accounting for 55.81%), indicating that townships have important potential to become sub-centers within the whole network, which can effectively connect urban and rural areas and promote the balanced flow of resources. These characteristics reveal the dominant effect of administrative hierarchy in the urban network of Qiandongnan, which may cause an imbalance in urban network structure, exacerbate the uneven distribution of resources, inhibit the development of small and medium-sized towns, and weaken regional synergy.

(2)

Stability Analysis of the Urban Network in Qiandongnan

Figure 8b shows that there are 109 K-cores with an average value of more than 7.7, accounting for 53.17% of the whole network. The urban hierarchy in this region is predominantly composed of county-level cities and cities (accounting for 100.00% of their own types), indicating that the whole network is stable. Additionally, county-level cities and cities can effectively interact with other urban unit nodes in the urban network, showing stronger stability. The size of K-core exhibits characteristics of regional distribution, and the same K-core gathers nearby, leading to the process of high-K-core settlements spreading outward toward low-K-core settlements. This suggests that regional development exhibits obvious spatial dependence, and the development of core areas can drive the surrounding areas; however, the diffusion effect decreases with distance. At the same time, nodes with high K-core values are highly concentrated at the intersection of traffic trunk lines, which shows that traffic convenience helps to enhance the connectivity of urban nodes. Meanwhile, the nodes with low K-core values are distributed in an “island” way, which reveals that marginal towns are restricted by factors such as terrain barrier, lack of investment, and cultural closure and highlights the fragile faults of the network structure. Therefore, it is vital to build a multi-level resilient network and formulate innovative cross-domain collaborative mechanisms.

(3)

Analysis of the Completeness of the Urban Network in Qiandongnan

In the urban network relationship in Qiandongnan, the network density is 0.0643, and the spatial network density is low, indicating that the connections between urban nodes are sparse and the network organization is loose as a whole. The average distance of the network is 2.979, and the connectivity between networks is poor, which further proves that the connection between urban units is not convenient enough, the cost of transportation or information transmission is high, and the overall coordinated development of the region is insufficient. This indicates that the efficiency of information and resource flow needs to be improved. The cluster coefficient is large (0.685), which reflects the close connection between nodes in local areas. However, the small-world characteristics of the whole network are not obvious, which indicates that efficient connectivity across regions is insufficient and the completeness of the network structure is poor.

(4)

Analysis of Urban Network Balance in Qiandongnan

The degree centrality potential of the urban network in Qiandongnan is 21.23%, which indicates that the balance of the network is weak. The control and influence of the core nodes on the whole network are concentrated, but the overall coordination efficiency is low, resulting in an insufficient resource allocation ability and weak anti-risk ability. Figure 8c shows that the nodes with higher degree centrality are concentrated along the traffic trunk lines and cities and counties, forming a phenomenon of “hub polarization”. This suggests that resources are excessively concentrated in high-level administrative nodes, resulting in weak connections between marginal towns and highlighting the triple constraint mechanism of “terrain constraint-administrative dominance-economic cluster” in mountainous areas.

(5)

Vulnerability Analysis of the Urban Network in Qiandongnan

As shown in Figure 9a, there are 11 cut-points in the network relationship, accounting for 5.37% of the network. The maximum cut-point effect is No. 57 (Liping County), which can cause 32 urban units to lose contact with the whole network at most. Additionally, the types of cut-point urban units are mainly cities (accounting for 6.67% of their own types), indicating that although the overall vulnerability of the network is low, the cut-point influence is significant, especially for city nodes, which show stronger cohesion and control in the network. Compared with Figure 9b, we can see that the cut-point is an important regional hub, and it is mainly concentrated in the administrative division of Liping County, which further reflects that natural terrain conditions, traffic advantages, and administrative resource allocation are the key influencing factors of the cut-point.

3.3.3. Analysis of Individual Networks of Urban Units in Qiandongnan

(1)

Analysis of Individual Competitiveness of the Urban Network in Qiandongnan

As shown in Figure 10, there are 32 urban units with high or above-average competitiveness. They are distributed across the center of the highly concentrated cluster of urban units, are not limited to a certain area, and have connections with many surrounding urban units, accounting for 15.61% of the whole network. At the administrative level, county-level cities (accounting for 100% of their own types) and cities (accounting for 86.67% of their own types) are the main types of urban units, which fully proves that the administrative level is positively correlated with competitiveness. Further analysis shows that there is obvious “administrative gradient differentiation” in the current network, where the nodes with high competitiveness are all in the traffic advantage area. This shows that county-level cities and city nodes gain competitive advantages by virtue of institutional advantages and infrastructure investments.

(2)

Analysis of the Individual Nature of the Urban Network in Qiandongnan.

Figure 11a shows that there are 19 urban units with increasing network coordinator functions, accounting for 9.27% of the whole network. Additionally, the main types of urban units are county-level cities (accounting for 100.00% of their own types) and cities (accounting for 46.67% of their own types), which are mainly distributed among urban units near the Shanghai–Kunming Expressway, Shanghai–Kunming High-speed Railway, and Hunan–Guizhou Railway. These are distributed among subgroups 5 and 7 (see Figure 7b for details). This shows that county-level cities are more prominent in the network coordination function, while a considerable number of cities also undertake important network coordination functions. This reflects that such nodes have strong information dissemination abilities within the network, which is more conducive to promoting exchanges between nodes. Additionally, urban units with convenient transportation are more likely to become key nodes in the regional network. It also shows that these two subgroups exert a high influence on the logistics and information flow within their respective subgroup networks and may form a specific functional cluster or economic belt in the regional network.

Figure 11b shows that the functional roles of gatekeeper and representative are the same in this study. There are 15 urban units with increasing functional roles as gatekeepers and representatives, accounting for 7.23% of the whole network, which is a small number. Compared with Figure 7b, it is found that these urban units are distributed in subgroups 2, 4, 5, 6, and 7, indicating that they are mainly concentrated in specific subgroups. This reflects their special status or functions in the regional network, and there are urban units that guarantee strong links inside and outside the subgroups.

As shown in Figure 11c, there are six urban units with increasing consultant roles, accounting for 2.93% of the overall network, which is a very small number. The types of urban units are mainly dominated by cities (accounting for 20.00% of the urban unit types), with fewer townships (accounting for 2.33% of the town types). This reflects that cities, as administrative centers, dominate the consultant function, while townships and towns, due to limitations in resources and scale, are single-functional. This results in the phenomenon of “concentration at the county level” and a lack of towns, highlighting the high correlation between administrative hierarchy and functional differentiation. The phenomenon of the hierarchical polarization of “city-level concentration and town-level scarcity” further highlights this correlation. Compared with Figure 7b, it is found that these urban units are distributed in subgroups 4, 5, and 6. Although these urban units cannot strengthen their own subgroup contact, they are beneficial for connecting with and strengthening external subgroup connections.

As can be seen in Figure 11d, there are nine urban units with increasing liaison roles, accounting for 4.39% of the whole network. The types of urban units are mainly county-level cities (accounting for 100.00% of their own types), while cities (accounting for 33.33% of their own types), towns (accounting for 0.00% of their own types), and townships (accounting for 2.33% of their own types) account for a very low proportion. This phenomenon reveals the scarcity of urban units that can influence or restrict communication and resource circulation between different subgroups as independent third-party forces, indicating that the overall network structure does not rely too much on a few key nodes to maintain its connectivity, thus showing low vulnerability.

3.4. Analysis of Urban Development Potential in Qiandongnan

Figure 12 shows that the number of urban units with low development potential in Qiandongnan is much higher than the number with low comprehensive conditions. This indicates the characteristic of “the status quo is better than the potential”, reflecting that the market has invested heavily in the selection and construction of urban areas in Qiandongnan. Furthermore, it indicates that most urban units in this area have poor long-term development conditions and insufficient overall sustainable development momentum. Compared with geographical location, it is found that urban units with a high comprehensive evaluation of urban conditions are more concentrated, urban units with high regional development potential are unevenly distributed and the distribution range is smaller. This proves that market selection is more inclined toward the cluster effect of core locations and key resources than toward regional balance. At the same time, it is found that urban units with high comprehensive conditions are basically urban units with high development potential, which shows that the development conditions of towns themselves form the basis of their development potential. From the perspective of spatial distribution, these high-potential urban units exhibit a “point axis” development model; that is, with Kaili as the core, they radiate outward along the main traffic trunk lines (such as the Shanghai–Kunming Expressway, Shanghai–Kunming High-speed Railway, and Hunan–Guizhou Railway), forming multiple development axes. This shows that traffic is an important support for urban development. It can be seen in

Table 5. Qiandongnan urban units: comprehensive evaluation vs. development potential.

Grade	High	Average	Low
Comprehensive evaluation	36	106	63
Development potential	14	83	108

that the development potential of urban units in Qiandongnan is divided into three levels, with 14 towns with high potential (accounting for 6.83%), 83 urban units with medium potential (accounting for 40.49%), and 108 urban units with low potential (accounting for 52.86%), indicating that the overall development potential of urban units is low.

3.5. Relationship Between the Development Potential and Network Structure of Urban Areas in Qiandongnan

Figure 13a shows that the development potential of urban units, the overall network of urban units, and the individual network of urban units are positively correlated. The development potential is significantly positively correlated with the competitiveness of the individual network and is also positively correlated with the functional roles of the network. This indicates that the development potential is closely related to the key nodes in the network structure, the close connections, and the efficient transfer of resources. Therefore, the network structure can be optimized to improve the roles of the key nodes and the effective scale of the network. Therefore, the development potential of the region as a whole can be improved by optimizing the network structure and improving the role function and effective scale of key nodes. The development potential is highly positively correlated with degree centrality, K-value, and core periphery in the overall network relationship, suggesting that the overall development potential of the entire network can be increased by improving its balance, stability, and completeness.

Figure 13b shows that the network structure of urban units in Qiandongnan is highly negatively correlated with transportation conditions, indicating that transportation conditions are a key factor restricting the synergistic development of urban units. The complex mountainous and hilly environment increases the cost of infrastructure, resulting in transportation-advantaged urban units becoming the “growth poles”, while marginalized towns exacerbate the network imbalance. This vulnerability can be addressed by utilizing a “dual-path resilience strategy” for natural disasters and economic crises [39,40]. At the same time, the network structure shows a significant positive correlation with social, production, and economic conditions, revealing that the network structure can be effectively optimized by upgrading these elements to improve the development level of urban units and achieve the sustainable development of towns. It further reflects that there is a positive correlation among the whole network, individual networks, and development potential.

3.6. Resilience Analysis of Urban Networks in Qiandongnan

In this study, the destructive test method is used to evaluate the resilience characteristics of the urban network in Qiandongnan. Through systematic node removal experiments, the response mechanism of the network under targeted attacks is simulated, which provides a quantitative basis for identifying key vulnerable links.

3.6.1. Key Node Identification Result

Key node identification is filtered from both partial and whole parts. Partial aspect: Through the analysis of degree centrality, K-core, effective scale, efficiency, and functional role, the top ten nodes are selected as key nodes, and all tangents in the network are also selected as key nodes. Overall: Based on the adoption of the social network analysis method and considering the connection strength between towns, important nodes are determined by combining multi-dimensional indicators such as degree centrality, K-kernel value, effective scale, efficiency, functional role, and the comprehensive conditions of urban units themselves, as well as by considering the development potential of urban social, economic, production, and traffic development. The recognition results are shown in

Table 6. Identification results for multi-index key nodes of networks in Qiandongnan.

Index	Key Node
Degree centrality	96, 103, 110, 111, 126, 130, 140, 149, 152, 199
K-core	106, 120, 130, 136, 140, 145, 148, 149, 150, 151, 152, 162, 169, 175, 193, 196, 199, 202
Effective scale	96, 103, 110, 111, 126, 130, 134, 140, 149, 162
Efficiency	0, 1, 2, 3, 12, 43, 52, 68, 79, 105, 123
Functional role	96, 103, 110, 111, 126, 130, 134, 140, 149, 199
Cutpoint	0, 2, 3, 6, 8, 11, 16, 52, 54, 68, 112
Comprehensive assessment	10, 28, 57, 88, 98, 113, 144, 175, 194, 196

.

3.6.2. Simulation Analysis of Node Failure

Through the node removal method, the key nodes of different dimensions are removed, and the changes in network topology indicators are compared. The experimental design adopts a multi-scene simulation strategy: (1) remove a single node and delete various key nodes (including the top 10 nodes by degree centrality, K-core nodes, cutpoints, etc.) in turn; (2) collaboratively remove multiple nodes to simulate cascading failure scenarios; (3) remove random nodes as a control experiment.

From

Table 7. Comparison of the influence of key node removal on the network topology indices.

	Before and After Key Nodes Are Removed	Network Density	Degree Central Potential	Network Average Distance	Aggregation Coefficient
Degree centrality node	Front	0.0670	13.03%	3.591	0.640
	After	0.0577	7.70%	2.272	0.626
K-core node	Front	0.0670	13.03%	3.591	0.640
	After	0.0588	13.62%	4.768	0.646
Effective scale node	Front	0.0670	13.03%	3.591	0.640
	After	0.0570	8.30%	3.875	0.630
Efficiency node	Front	0.0670	13.03%	3.591	0.640
	After	0.0740	13.47%	3.280	0.671
Functional role node	Front	0.0670	13.03%	3.591	0.640
	After	0.0570	7.78%	3.873	0.629
Tangent node	Front	0.0670	13.03%	3.591	0.640
	After	0.0723	13.11%	3.245	0.664
Comprehensive evaluation node	Front	0.0670	13.03%	3.591	0.640
	After	0.0628	12.92%	3.668	0.635
Random node	Front	0.0670	13.03%	3.591	0.640
	After	0.0672	13.00%	3.608	0.629

, it can be seen that the network of urban clusters in Qiandongnan shows significant dependence on centralization. The failure of degree centrality and functional role nodes leads to a decrease in network density by 13.9–14.9% and a sharp drop in degree centrality potential by 40.3–40.9%, which is far beyond the fluctuation range of the random node control group (<2%), confirming the systematic risk of a single-center structure. The failure of K-core nodes caused the average distance to surge by 32.8%, exposing the defect of long path dependence. It is worth noting that the network density rises abnormally by 7.9–10.4%, and the clustering coefficient improves after the failure of efficiency nodes and cutpoints, which shows the potential for network self-organization optimization. Additionally, the comprehensive evaluation node shows the best resilience (index fluctuation < 6.3%), which verifies the effectiveness of multi-dimensional collaborative screening for robust hub identification.

As shown in

Table 8. Failure risk and optimization strategy analysis of key nodes.

Indicator Node	Network Density Change	Degree Centrality Change	Average Distance Variation	Variation of Cluster Coefficient	Main Risks	Optimization Strategy
Degree centrality	↓ 13.9%	↓ 40.9%	↓ 36.7%	↓ 2.2%	Global fragmentation, power structure collapse	1. Increase redundant connection of core nodes.2. Decentralized and centralized function
K-core	↓ 12.2%	↑ 4.5%	↑ 32.8%	↑ 0.9%	Long-path fracture reduces the stability of the hierarchy	1. Strengthen cross-level short paths2. Optimize the core-periphery connection
Effective scale	↓ 11.8%	↓ 8.2%	↑ 15.4%	↓ 1.2%	Non-redundant connection fracture	1. Enhance alternative paths between nodes.2. Optimize connection diversity
Efficiency	↑ 10.4%	↑ 3.3%	↓ 8.7%	↑ 4.8%	May trigger network reconfiguration.	1. Optimize the network with reconfiguration potential.
Functional role	↓ 14.9%	↓ 40.3%	↑ 7.8%	↓ 1.7%	Multifunctional service interruption	1. Establish a functional backup center.2. Improve the functional substitution
Point of tangency	↑ 7.9%	↑ 0.6%	↓ 9.6%	↑ 3.8%	Locally isolated but global optimization	1. Dynamic monitoring + preset emergency link
Comprehensive assessment	↑ 6.3%	↓ 0.8%	↑ 15.4%	↓ 0.8%	Need to prevent cumulative effects	1. Multi-index collaborative planning
Random node	↑ 0.3%	↓ 0.2%	↑ 0.5%	↓ 1.7%	The influence is weak, and the benchmark stability is verified.	As a reference value of the control group

, the removal of different nodes will have a differentiated impact on network indicators. Among these, the failure of height-centered nodes and key function nodes is particularly significant, leading to a decrease in the degree centrality index of more than 30%. This indicates that the network has been greatly impacted and its resilience is low. This proves that network resilience is positively related to the importance of nodes: the failure of core nodes directly weakens the stability of the network, which is manifested in a decrease in network density, an increase in average path length, and the interruption of function. Therefore, key nodes play a decisive role in maintaining the stability of the network’s structure and function.

At the same time, this study also reveals the unique phenomenon of the coexistence of the “core-edge” fragile mode and local redundancy characteristics within mountain urban networks. In order to improve network resilience, it is necessary to optimize the redundant connections of core nodes and enhance the connectivity of edge nodes to mitigate the global impact caused by the failure of core nodes. In addition, the multi-dimensional collaborative screening method for nodes can identify the most resilient nodes more accurately, which provides a quantitative basis for future resilience promotion strategies.

3.7. Planning and Construction of Resilient Urban Systems Based on Social Network Analysis

In the sections above, it can be seen that the urban units in Qiandongnan exhibit the distribution characteristics of “large dispersion and small cluster”, The balance, stability, and completeness of the overall network are positively related to the development potential of urban units, the competitiveness of nodes in individual networks, and their roles. The overall network’s morphological structure plays a regulatory role in the competitiveness, role, and urban development potential of individual networks. Therefore, based on the spatial layout characteristics of the current urban cluster, the hierarchical characteristics of the urban cluster network, the resilience characteristics of the urban network, and the development needs of the upper planning, we emphasize the coordinated development of point-belt line and line-area and build a resilient urban development system characterized by “core leading—circle interaction—axis radiation—area coordination”. By strengthening the leading role of core nodes in county-level cities and cities to drive circular interaction, improving linear infrastructure such as traffic corridors to strengthen axial radiation efficiency, and improving the functional capacity of township units to promote the coordinated development of the district, the regional network structure can be optimized, and development potential can be released. Differentiated resource allocation and precise policy supply are implemented, and while guaranteeing the core driving effect, the resilient nodes at the grassroots level are fostered to enhance the network’s risk-resistant capacity. This ultimately forms a new pattern of urban development that is coordinated at all levels and resiliently adapted to the situation. The system not only highlights the polarized leadership of high-level towns but also emphasizes the synergistic response of grassroots units, achieving the organic unity of core traction and territory-wide linkage.

3.7.1. Development Strategy of Differentiated Resilience and Optimization of Spatial Functional Structure of Urban Units in Qiandongnan

Based on the characteristics of current network relations, regional urban units are identified via functional zoning. Based on the distribution characteristics of cohesive subgroups in the network structure, the functional partition is determined by combining administrative requirements, regional geography, functional relationships, resource advantages, superior plannings and future development trends. According to the importance of network nodes in individual networks, and considering the functional roles, competitiveness, development potential, and regional balance requirements of nodes, the important nodes for regional development are screened and delineated. Meanwhile, based on the results of the network resilience analysis, the differentiated development strategy of a “structure-function-system” trinity is innovatively proposed. According to the distribution of cluster subgroups, management scale, and administrative boundaries, the whole region is reorganized into four collaborative development areas (Figure 14a). Additionally, a three-level structure of “core nodes-important nodes-general nodes” (Figure 14b) is constructed, with an emphasis on enhancing the functional redundancy of core hubs such as Kaili (by adding sub-centers) and the alternative path connection of Liping nodes.

Based on the results of the minimum spanning tree analysis, three strategic development axes are planned and constructed (Figure 15a,b). Based on the core-periphery theory and the regional development status, the spatial layout structure of “one core, two circles, three axes and four regions” is constructed in this study (Figure 15c). This structure takes Kaili as the core of comprehensive service and forms a hierarchical network with 15 important functional nodes. In terms of spatial layout, Sansui and Zhenyuan counties in the north form a logistics hub circle, and Rongjiang and Congjiang counties in the south form a tourism hub circle, forming a double-circle structure with complementary functions. In terms of functional zoning, the area is divided into four characteristic development zones: the Miao Ling Eco-tourism Zone, which focuses on developing the eco-health industry, the Dongxiang Cultural Heritage Zone, which focuses on building a cultural experience demonstration zone, the Duliujiang River Basin industrial zone, which is designed with green manufacturing and characteristic agriculture, and the Qingshuijiang Ecological Conservation Zone, which focuses on ecological protection and popular science research. In terms of spatial connection, the construction of three development axes, including the Shanghai–Kunming high-speed rail corridor (Kaili–Sansui), the North–South high-speed rail development axis (Tianzhu–Liping), and the Miao–Dong customs–Guiguang high-speed rail development axis (which includes both the Miao–Dong customs development axis and the Guiguang high-speed rail corridor), has effectively strengthened communication inside and outside the region. This spatial structure fully integrates the ecological, cultural, and industrial resources of Qiandongnan Prefecture and provides scientific spatial support for regional coordinated development.

The spatial planning innovatively constructs a multi-dimensional development model of “core guidance-circle interaction-axis radiation-area coordination”, which not only fully respects the natural and humanistic characteristics of the mountainous multi-ethnic areas in southeastern Guizhou but also achieves a dynamic balance between ecological protection and high-quality development through scientific spatial organization. Additionally, it explores a sustainable development path for similar areas, taking into account ecological security, cultural heritage, and industrial revitalization.

3.7.2. Path and Planning Implementation of Improving Urban Network Resilience in Qiandongnan

Based on the analysis of complex network topology and network resilience, this study proposes a “multi-level resilience promotion path for network structure optimization”, which is “structural optimization-functional coordination-resilience improvement”.

(1) Network hierarchy identification and structure optimization: Firstly, the core-periphery model is used to identify the hierarchical structure of the network, the nodes are divided into core nodes, important nodes, and general nodes, and the strategic position of each node in the network is determined according to degree centrality, K-core, and other indicators. Moreover, different functional groups are identified by agglomerating subgroup distribution, and the connection relationships between groups are optimized based on indicators such as individual competitiveness in structural holes, with an emphasis on strengthening the bridging function of middlemen (that is, nodes with special role functions). Next, the vulnerability of nodes is evaluated according to the degree that the network tangent affects the disconnection between nodes and other nodes in the network, and the node’s own resilience is improved based on the dimensions of comprehensive development conditions and the development potential of urban nodes.

(2) Collaborative development of multi-scale functional groups: Based on the framework of collaborative governance of urban agglomerations [39], the “cohesive subgroup analysis” is expanded into a three-dimensional identification system of “culture-economy-ecology”. A multi-scale collaborative mechanism is constructed, and the radiation-driven mechanism of the core nodes of the “annual assessment of pairing assistance” is implemented, with the driving effect of the core nodes on the edge nodes included in the official performance indicators. A complementary development mechanism for the functional groups of the “cultural resources trading platform” is established, allowing groups to deduct part of the cost of infrastructure construction through flexible indicators, such as the number of non-legacy performance days and handicraft quotas. The circulation mechanism for the key axis elements of the “resilience adjustment tax” is set up, and 3% is extracted from the towns along the axis whose tourism income exceeds 100 million yuan to compensate for the development restrictions in ecologically sensitive areas, so as to form a resilient urban network system with adaptive ability.

(3) Resilience Differentiated Development Planning: This study innovatively adopts a hierarchical and differentiated strategy, incorporating the concept of agricultural biodiversity from Greece to construct a composite resilience system [40]. Firstly, for the core structural nodes, a “1 + N + 3” functional redundancy model is implemented. In addition to setting up sub-centers and backup facilities, it is mandatory for each core node to cultivate more than three unrelated industries to ensure the continuous supply of critical functions in the event of an emergency. Secondly, a “dual-path optimization” strategy is applied to intermediary nodes: while improving alternative path connections, industry diversity evaluation indicators are embedded, and the intermediary function is based on at least two alternative industries. For the upgrading of edge nodes, a “resilience cultivation package” model is adopted, linking infrastructure improvements with the incubation of distinctive industries. Quantitative simulations show that after the implementation of this plan, the network density is expected to increase from 0.0643 to 0.1394, representing an increase of 116%; the average path length is shortened by 16.9%; and the proportion of high-order K-core nodes is expected to double, increasing from 40.49% to 80.98%.

4. Discussion

4.1. Characteristics of Urban Distribution and Network Morphology and Structure in Qiandongnan

Scholars have found that the network of villages and towns in the Funing District of Qinhuangdao exhibits a significant core–periphery structure, and the core towns have an obvious radiation driving effect on the surrounding areas [29]. This study found that this feature also exists in Qiandongnan, where the core urban units occupy a dominant position in resource allocation, while the development of the marginal areas is relatively underdeveloped. It is worth noting that the unique mountainous terrain in Qiangdongnan further exacerbates the imbalance in spatial development and makes the hierarchical differentiation characteristics of urban networks more prominent.

Li Na and other scholars believe that the comprehensive functions of transportation, communication ability, regional unique factors, and geographical location promote the formation of a “small cluster and large dispersion” structure in urban areas [41]. This study confirms that the urban cluster in southeastern Guizhou is characterized by “large dispersion and small cluster”, which is mainly influenced by natural terrain conditions and population activities. Natural restrictions promote the formation of small-scale clusters along traffic lines, highlighting the coupling effect between physical geography and human activities. Further research shows that this natural dominant structure leads to low network density and limited radiation of core nodes. Therefore, it is necessary to strengthen transportation and communication facilities to break through geographical barriers, thus supplementing the applicability of this theory in underdeveloped mountainous areas.

Wei Jianfei and other scholars have pointed out that Zhengzhou, Luoyang, and other cities with outstanding location and transportation advantages frequently communicate within the related network and are at the core of this network, which can drive surrounding cities to improve the efficiency of land use [42]. This study shows that urban units with good comprehensive conditions have higher development potential, and the higher the degree of centrality, the easier it is for a city to become the core of the network and drive the development of surrounding towns. In addition, this study also found that the overall network and individual networks of cities and towns are positively correlated with development potential. Adjusting the network relationships can enhance this development potential, which confirms that there is an inherent and inevitable relationship among the overall network, individual networks, and node attributes.

Xia and other scholars have found that the network density of the Greater Bay Area urban cluster in Guangdong, Hong Kong, and Macao reached 100% based on Baidu migration data, forming a multi-center network structure. The core cities promoted regional cooperation through the flow of high-frequency elements [43]. The research shows that the overall network of the urban cluster in southeastern Guizhou is generally stable but incomplete, and the network density is lower than that in developed areas, reflecting that the flow of factors in underdeveloped areas is not active. Both of them exhibit the radiation effect of core nodes, but Greater Bay Area demonstrates multi-center collaboration, while Qiandongnan shows one-way dependence, which verifies the influence of economic level on network resilience. It has also been found that the development potential of urban units in Qiandongnan is positively related to economic and traffic conditions. However, high-potential towns only account for 15.61%, which highlights the difficulty of resource integration in underdeveloped mountainous areas and provides a new basis for the differentiated development of similar regions.

Through gravity model and passenger flow analysis, Liu Zhengbing and other scholars have found that the Central Plains urban cluster presents a “dual-center” network structure, and the centrality of the central city is significantly higher than that of other cities. Additionally, the network cluster effect is obviously affected by administrative divisions [44]. It is also observed in this study that the towns with superior comprehensive conditions in the Qiandongnan urban cluster exhibit high degree centrality and assume the function of regional hubs. However, the difference is that the key nodes in Qiandongnan rely more on characteristic functions than on simple economic scale. Additionally, the positive correlation between urban development potential and network centrality is more significant, which proves that underdeveloped areas can enhance their node status by excavating unique factors and provides a new perspective for the cultivation of “functional central towns”.

Based on the development of the regional village system, some scholars have suggesteed that we should focus on functional zoning and the cultivation of important nodes [19]. This study also agrees with this view. Therefore, by improving the urban network structure and the development potential, a multi-center resilient network has been constructed, which significantly improves the adaptability and stability of complex systems under uncertainty, interference, or impact. The research approach is more rigorous, and the research methods and content are more comprehensive.

4.2. Multi-Level Resilience Promotion of Network Structure Optimization and Its Popularization Adaptability

The concept of the “Multi-level resilience enhancement of network structure optimization” proposed in this study refers to the cooperative adaptability of the urban network system at three different levels: overall network, individual nodes, and development potential. This concept draws lessons from the “multi-layer network resilience theory” introduced by Smith and other scholars [45], which emphasizes that the resilience of complex systems needs to be improved from three dimensions: macro-structure, meso-connection, and micro-elements. In addition, it forms a theoretical echo with the research results on the coordinated governance of urban agglomerations [39]. This research shows that there is an internal relationship between the whole network of the same research object and the individual networks of its nodes. This correlation makes it possible to achieve the matching of space demand and supply by adjusting the network relationships [46]. This study finds that there is an inevitable internal relationship among individual networks, overall networks, and comprehensive development potential, and the three interact through the two-way mechanism of “overall adjustment of individuals and individual influence on the whole”. This mechanism is similar to the multi-level synergistic effect of “genetic diversity-species diversity-functional diversity” in agricultural ecosystems [40], which shows that system resilience depends on the complementarity and feedback of functions at all levels.

Based on the abovementioned theoretical logic, this study proposes to achieve the hierarchical reconstruction and coordinated development of cities and towns by adjusting the internal relationships of networks. In terms of empirical research, this study takes the towns in Qiandongnan as the object and adopts multi-dimensional research methods. First, it uses Tai Sen polygon and other urban spatial distribution analysis methods to understand the current distribution characteristics and influencing factors. Secondly, based on field investigations and data crawling, the commuting times between urban units are quantified, and the working relationships between urban units are quantified and visualized as a network relationship using a gravity model. Finally, the characteristics of the urban network are studied by using the relationship analysis method of urban system, and the relationship among the overall network, individual network, and comprehensive development potential is analyzed by using the correlation analysis method, which proves the internal relationship of “overall network-individual network-development potential”. Based on the abovementioned research findings, this paper analyzes the internal relationships through the network structure and constructs a research logic chain of “status analysis-network quantification-correlation verification-structure regulation” based on the internal functional relationships of the network and the requirements of upper planning. From the perspective of spatial functional structure optimization, this paper explores ways to improve the resilience of the urban network in Qiandongnan and provides theoretical support for the reconstruction and coordinated development of the urban hierarchy. This method encompasses three levels of “whole network-subgroup-individual node” on the following scale: it combines cross-regional connections at the macro level, delineates functional partitions at the meso level, and improves node resilience at the micro level. This method of multi-level overall planning is not only suitable for systematic monitoring and early warning but also provides a reproducible solution for “network structure optimization-multi-level resilience improvement” in underdeveloped mountainous areas because its “relationship-oriented” planning paradigm breaks through traditional limitations. Therefore, because of its systematic nature, operability, and regional adaptability, the multi-level resilience promotion path proposed in this study provides a scalable solution for the network optimization of mountain urban units.

5. Conclusions

Qiandongnan, with its typical mountainous terrain, multi-ethnic cultural symbiosis, and characteristics of post-development urbanization, has become an ideal location for studying the development of urban systems in underdeveloped mountainous areas in China. This study found that the urban system in Qiandongnan exhibits three prominent features: First, the spatial distribution of urban units shows a random pattern, with natural geographical conditions as the main influencing factors. Second, the network structure exhibits typical scale-free characteristics, forming a multi-center system of “three levels and eight subgroups”, with an obvious core–periphery structure, and the strength of regional connections is a key factor in the differentiation of subgroups. Third, there is a significant positive correlation between the development potential of urban units and the network structure, economic and transportation conditions are the main influencing factors, and network resilience is positively correlated with the importance of nodes. Furthermore, the importance of nodes is positively correlated with network stability and their roles within the system. Based on these findings, this study proposes the “Multi-level Resilience Enhancement Path for Network Structure Optimization”, which follows the steps of defining spatial distribution characteristics, revealing spatial network structures, exploring spatial development relationships, delineating spatial structures, implementing spatial individual positioning, and designing hierarchical strategies. The path emphasizes multi-level collaborative optimization from the macro to micro scale. At the macro level, the regional driving role of core nodes is strengthened. At the meso level, internal coordination among functional subgroups is promoted. At the micro level, the specialized functions of individual towns are enhanced. Through this multi-scale collaborative planning approach, the goal is to achieve the multi-level resilience promotion of “core node radiation-driven, functional subgroup differential development, and individual town function promotion”, effectively promoting the systematic optimization of the urban system from macro-cluster coordination to micro-individual management and control. The research results not only provide a scientific plan for the development of the urban system in Qiandongnan but also have greater value. Firstly, the applicability of the network analysis method in the planning of underdeveloped towns in mountainous areas is verified. Secondly, a new mode of coordinated development for towns in multi-ethnic areas is proposed. Thirdly, the technical framework for “diagnosis-optimization-implementation” can serve as a reference for similar regions around the world. These findings provide new ideas and solutions for the urbanization process in underdeveloped mountainous areas, ranging from theoretical methods to practical applications, and promote regional sustainable development.

In addition, the development of urban systems is influenced by dynamic external factors, such as policy changes and population movement. Future research will incorporate these factors by conducting dynamic experimental simulations to analyze the evolution process and predict future development trends.