The Impact of Urban Networks on the Resilience of Northwestern Chinese Cities: A Node Centrality Perspective

Abstract

Urban networks are a key force in reshaping regional resilience patterns. However, existing research has not yet systematically elucidated, from a physical–virtual integration perspective, the underlying mechanisms through which composite urban networks shape multidimensional urban resilience in regions confronted with severe environmental and infrastructural challenges. Northwest China, characterized by its extreme arid climate, pronounced core–periphery structure, and heavy reliance on overland transportation, provides an important empirical context for examining the unique relationship between network centrality and the mechanisms of resilience formation. Based on the panel data of 33 prefecture-level cities in northwest China from 2011 to 2023, this article empirically examines the impact of the composite urban network constructed by traffic and information flows on urban resilience from the perspective of network node centrality using a two-way fixed-effects model. It is found that (1) the spatial evolution of urban resilience in northwest China is characterized by “core leadership—gradient agglomeration”: provincial capitals demonstrate significantly the highest resilience levels, while non-provincial cities are predominantly characterized by medium resilience and contiguous distribution, and the growth rate of low-resilience cities is faster, which pushes down the relative gap in the region, but the absolute gap persists; (2) the urban network in this region is characterized by a highly centralized topology, which improves the efficiency of resource allocation yet simultaneously introduces systemic vulnerability due to its over-reliance on a limited number of core hubs; (3) urban network centrality exerts a significant positive impact on resilience enhancement (β = 0.002, p < 0.01) and the core nodes of the city through the control of resources to strengthen the economic, ecological, social, and infrastructural resilience; (4) multi-dimensional factors synergistically drive the resilience, with the financial development level, economic density, and informationization level as a positive pillar. The population size and rough water utilization significantly inhibit the resilience of the region. Accordingly, the optimization path of “multi-center resilience network reconstruction, classified measures to break resource constraints, regional wisdom, and collaborative governance” is proposed to provide theoretical support and a practical paradigm for the construction of resilient cities in northwest China.

1. Introduction

In recent years, the accelerated superposition of global climate change and urbanization has led to the frequent occurrence of compound disasters such as public health events [1], extreme weather (e.g., high temperatures [2], torrential rains [3]), urban waterlogging [4], and infrastructure accidents (e.g., gas explosions [5]), which pose a serious challenge to the sustainable development of urban systems. In this context, resilient cities, as a core vehicle for realizing the Sustainable Development Goals (SDGs), have been included in the important agenda of global urban governance [6].

At the international level, enhancing urban resilience has become a core strategy for addressing global climate change and complex risks. Different countries and regions have developed distinctive practices: Europe has taken the lead in systematic approaches, such as the “water squares” in Rotterdam, the Netherlands, and the blue–green infrastructure systems in Copenhagen, Denmark, which integrate flood prevention and urban spatial renewal. North America emphasizes technology-enabled solutions, exemplified by New York’s OneNYC resilience strategy, which leverages big data and ecological restoration to strengthen coastal protection and community response capacities. Asia, meanwhile, focuses on innovation in high-density environments, such as Singapore’s ABC Waters Programme (Active, Beautiful, Clean Waters) and Japan’s disaster-resilient community development. These practices have not only expanded the multi-layered resilience framework—from “defense and adaptation to transformation”—but have also provided methodological and practical references for China’s ongoing efforts in building urban resilience. In China, “resilient cities” were first included in the national strategic plan at the 5th Plenary Session of the 19th CPC Central Committee in 2020, in the 20th National Congress report, which explicitly stated that China needs to “to build livable, resilient, and smart cities”, and in the 2023 General Secretary Xi Jinping’s proposal to “comprehensively promote resilient and safe cities”, which is a major step forward. In 2023, General Secretary Xi Jinping proposed to “comprehensively promote the construction of resilient and safe cities”; the evolution of policy discourse reflects the systematic upgrading of the comprehensive risk prevention and control capabilities of cities.

Urban resilience is essentially the comprehensive ability of a complex urban system to absorb external shocks, adapt, learn, and transform, and recover to the original or reach a new steady state when coping with disturbances [7]. The measurement of urban resilience predominantly relies on multi-dimensional evaluation systems, constructed based on either its constitutive dimensions [8,9,10,11,12] (e.g., economic, social, ecological) or core characteristics [13,14,15] (e.g., robustness, adaptability). The quantification and analysis of these systems are facilitated by diverse methodological tools, including those for assigning indicator weights (e.g., the entropy method [16]) and for modeling complex driver relationships (e.g., machine learning [17], DPSIR frameworks [18]), thereby advancing the empirical understanding of resilience determinants. In terms of research scale, the research object extends from a single city [12] to the cross-scale of provinces [19], urban agglomerations [20], and watersheds [21], etc., and in terms of the correlation mechanism, scholars have started to explore the interaction between resilience and external systems such as the digital economy [22], land use [23], and urbanization [24]. The current high-frequency flow of social factors is gradually dissolving the isolation of cities. Such cross-regional factor flows not only reconfigure the material and information exchange patterns between cities but also promote the transformation of cities from closed geographical units to open network nodes, giving rise to the “flow space” perspective of resilience. This has given rise to a new paradigm for resilience research from the perspective of the “flow space”.

Along with the rise of “flow space” theory, the urban research paradigm is shifting from a static local space to a dynamic network space [25]. Since Camagin, Taylor, and other scholars pioneered the measurement of inter-city connections, urban network research has gradually become an important paradigm for analyzing urban systems. Domestic and international research has mainly focused on two types of networks; “hard networks” constructed based on physical elements such as railroad flow [26], highway flow [27], aerodynamic network, and people flow [28] have achieved significant research results. On this basis, “soft networks” derived from and constructed based on virtual elements such as information flow [29] and capital flow [30] are abstract but can effectively reveal the functional division of labor and cooperative relationships between cities. In recent years, the leaping development of transportation and information technology has further weakened geographical constraints and strengthened the network connection between cities. As nodes in the network, the resilience of cities is closely related to the network structure: On the one hand, the network may amplify risks. After a node city encounters an external shock, relying on the network correlation, its failure may produce a risk transmission effect. Cities with strong resilience can effectively absorb shocks, while cities with weak resilience may fail to absorb shocks to other nodes in the network, triggering a “domino” type of cascading effects. On the other hand, the network also empowers resilience. Cities that occupy the core of the network (with high node rank and power) usually have stronger regional resource control, deployment capacity, and access channels, thus obtaining a more abundant supply of material, information, organization, and other dimensions needed for resilience, and their overall resilience level and risk resilience are also enhanced accordingly. It can be seen that the impact of urban network on urban resilience has a significant “double-edged sword” effect [31]; there is a phenomenon of potential synergistic amplification of risk and uneven distribution of resilience resources.

There is no lack of research on the relationship between urban networks and urban resilience in current studies. For example, Du Y et al. [32] found that the deeper a city’s participation in the cross-regional collaborative innovation network, the more favorable it is to improve the economy’s ability to withstand risks and promote recovery, both in the earthquake-resistant period and in the recovery adjustment period. Hui X et al. [33] employed a multi-period DID method to empirically test whether the construction of network infrastructure can significantly enhance urban economic resilience and whether entrepreneurial activity is an effective pathway through which network infrastructure construction enhances urban economic resilience. Peng D et al. [34] applied an improved gravity model to construct an industrial spatial association network for the Yangtze River Midstream Urban Agglomeration, empirically testing whether industrial spatial association promotes the enhancement of the economic resilience of the urban agglomeration. Furthermore, in cities with good financial environments and high levels of marketization and in non-industrial cities, the role of industrial spatial association in enhancing economic resilience is more significant. Su R [35] employed various econometric models, including fixed-effects models and quantile models, to determine whether information networks have a positive moderating effect on the promotion of urban economic resilience through manufacturing upgrading, exhibiting heterogeneous characteristics. The integration of manufacturing upgrading and information networks enhances urban economic resilience by improving the level of total factor productivity.

Although existing studies have explored the two and their related elements, providing a preliminary empirical basis for this study, there are still some research gaps. First, prevailing studies focus on the impact of single-type networks on resilience subsystems [36,37,38,39], overlooking the role of composite networks in shaping holistic urban resilience. Second, empirical analyses concentrate on developed regions [40,41,42], with insufficient attention paid to underdeveloped areas, particularly northwest China. Third, there is a lack of attention to the internal mechanism of the correlation between urban network structure and resilience under the specific constraints in the northwest region.

The northwestern region is located in the heart of the Eurasian continent and has seen a significant increase in its strategic importance. It has risen from its traditional status as a peripheral area to become the “geopolitical foundation” for the country’s participation in global competition. It shoulders the important mission of serving as a key hub for the Belt and Road Initiative, the core gateway for the new western land–sea channel, and a strategic link for the “land–sea linkage and east–west mutual assistance” open pattern [43]. However, its development faces multiple unique challenges: a vast geographical expanse coexists with a sparse population; the urban system exhibits a “spike-like” fragmentation, with only Xi’an, a megacity, forming a prominent core; the arid climate and complex terrain of its inland location lead to frequent disasters; ecological fragility and economic lag (gross domestic product (GDP) accounted for only 5.86% of the national total in 2023) intertwine, compounded by the social complexity of multi-ethnic cohabitation, making it a “regional stronghold” that must be overcome in the nation’s modernization process. Additionally, the infrastructure network has significant shortcomings: the density of the railway network is significantly below the national average, with interprovincial train services highly concentrated in Xi’an; road freight accounts for over 80% of total freight, but the proportion of low-grade roads is too high, and marginal cities face insufficient supply of public resources [44]. Therefore, continuously strengthening the region’s resilience to external shocks is of critical importance.

In summary, against the backdrop of the accelerated advancement of regional integration and the Western Development Strategy, this study focuses on the northwestern region of China. From the perspective of a hybrid virtual–physical cyberspace [45], it primarily explores three core questions: Does the composite urban network significantly influence regional resilience? How does network centrality influence the multidimensional resilience of cities? Can this network empower the overall enhancement of resilience in northwestern cities? Based on this, this paper takes 33 prefecture-level cities in the northwestern region as the research object, constructs a four-dimensional resilience evaluation system encompassing economy, society, infrastructure, and ecology, and measures the comprehensive resilience level. Then, based on transportation and information networks, a composite urban circulation network is constructed, and the network centrality index for each year is calculated. Finally, using tools and methods such as ArcGIS, Ucinet, and Stata, this study empirically explores the mechanisms through which urban networks in the northwest region influence urban resilience, aiming to provide important empirical support for enriching and deepening the theoretical framework and practical pathways of China’s resilient city construction.

2. Study Area and Data Sources

2.1. Study Area

As one of the seven major geographic regions in China (Figure 1a), the northwest Region is the region with the driest climate in China, covering five provinces and regions, namely Shanxi, Gansu, Ningxia, Qinghai, and Xinjiang (Figure 1b). The region has a total area of 3,072,400 square kilometers, accounting for about one-third of China’s total land area. However, the resident population in 2023 was only 103.38 million, accounting for 7.33% of the national population. Northwest China is located in the transition zone between the Qinghai–Tibet Plateau and the Loess Plateau, with obvious topographic relief. Its unique three-dimensional geomorphological pattern of “three mountains sandwiched between two basins” directly restricts the physical network connectivity between cities on the one hand and pushes up the cost of transportation infrastructure. On the other hand, it also leads to the construction of 5G base stations facing a hostile environment, a short cycle, high investment, and other challenges, which restricts the enhancement of the effectiveness of the virtual information network [46]. Meanwhile, dense active fracture zones have made northwest China a highly earthquake-prone area in China, and the frequent occurrence of secondary disasters such as landslides and mudslides have significantly increased the vulnerability of the urban system [47]. Therefore, based on the availability of data, this study targets 33 prefecture-level cities in the region (Figure 1c) and is guided by urban resilience and urban network theories to provide theoretical support for solving the problems of sustainable development in the “northwest context”.

2.2. Data Sources

Data on urban resilience indicators, road passenger and freight transport volumes, and control variables are primarily sourced from provincial statistical yearbooks and the China Urban Statistical Yearbook, with some supplemented by reviewing local statistical bulletins. Missing values are interpolated and corrected using the average value method for adjacent years or linear regression. Some of the indicators are synthesized by inter-indicator arithmetic with reference to the existing literature (

Table 1. Sources for calculating selected indicators of urban resilience.

Data	Calculation Method	Reference
Industrial Structure Advancement Index	Output value of tertiary industry/output value of secondary industry	Fu L [48]
Local Fiscal Self-Sufficiency Rate	Local general public budget revenue/local general public budget expenditure	Xie H [49]
Foreign Trade Dependency Ratio	Total import and export/GDP	Fu J, etc. [50]
Social Insurance Coverage Rate	(Number of urban workers enrolled in basic medical insurance + number of urban workers enrolled in basic pension insurance)/2/city’s resident population	Wang J [51]

). The digital financial inclusion index is from the Digital Finance Research Center of Peking University (https://idf.pku.edu.cn/, accessed on 28 February 2025). The PM_2.5 concentration data is from the National Tibetan Plateau Scientific Data Center (https://data.tpdc.ac.cn/, accessed on 10 March 2025). The Normalized Difference Vegetation Index (NDVI) is derived from the MOD13A3 dataset regularly released by NASA (https://www.earthdata.nasa.gov/, accessed on 12 March 2025). Information network data were obtained from the official website of the Baidu Search Index. Railway schedule data were obtained using historical versions of the Shengming Train Schedule software. To construct a time-series dataset and capture annual variations, we collected the daily frequency of direct passenger trains between each pair of cities on a fixed date (20 December) for each year from 2011 to 2023. All data were cross-verified against the 12306 online railway ticket booking system.

3. Research Methods

3.1. Measurement Method of Urban Resilience

As a complex system intertwined with multiple elements, urban resilience covers key areas such as society, economy, ecology, institutions, and infrastructure. Drawing on the established research framework [8,9,10,11,12], this study constructs an urban resilience evaluation index system containing 32 specific indicators from the four core dimensions of economy, society, ecology, and infrastructure (

Table 2. Indicator system for evaluating urban resilience in northwest China.

Objective Level	Criterion Level	Indicator Level	Unit	Attribute	Weighting Factor
					Entropy Weight Method	CRITIC Method	Integrated Weighting
Urban Resilience	Economy	GDP per Capita	RMB	+	0.0366	0.0237	0.0302
		Industrial Structure Advancement Index	%	+	0.0314	0.0393	0.0353
		Rural–Urban Income Ratio	%	−	0.0068	0.0291	0.0179
		Per Capita Total Retail Sales of Consumer Goods	RMB	+	0.0316	0.0293	0.0305
		Local Fiscal Self-Sufficiency Rate	%	+	0.0478	0.0445	0.0462
		Foreign Trade Dependency Ratio	%	−	0.0010	0.0237	0.0124
		Per Capita Aggregate RMB Deposits in Financial Institutions	RMB	+	0.0575	0.0192	0.0384
		Digital Financial Inclusion Index		+	0.0162	0.0374	0.0268
	Social	College Students per 10,000 People	per 10,000 population	+	0.0832	0.0369	0.0600
		Hospital Beds per 10,000 Population	per 10,000 inhabitants	+	0.0173	0.0243	0.0208
		Licensed Physicians per 10,000 Inhabitants	per 10,000 population	+	0.0242	0.0228	0.0235
		Crude Rate of Natural Increase	%	+	0.0110	0.0420	0.0265
		Public Administration Staff per 10,000 Residents	per 10,000 population	+	0.0279	0.0276	0.0277
		Average Wage of Employed Workers	RMB	+	0.0155	0.0270	0.0213
		Public Library Book Collections per 10,000 Capita	volumes per 10,000 capita	+	0.0888	0.0218	0.0553
		Social Insurance Coverage Rate	%	+	0.0453	0.0313	0.0383
	Ecology	Built-up Area Green Coverage	%	+	0.0039	0.0248	0.0144
		Industrial SO2 Emissions	ton	−	0.0019	0.0266	0.0143
		Municipal Solid Waste Treatment Rate	%	+	0.0011	0.0198	0.0105
		Urban Sewage Treatment Rate	%	+	0.0038	0.0323	0.0181
		Per Capita Park Green Space	m2	+	0.0189	0.0223	0.0206
		PM2.5 Concentration		−	0.0051	0.0302	0.0177
		NDVI	μg/m3	+	0.0378	0.0735	0.0556
		Good Air Quality Days Ratio	%	+	0.0033	0.0369	0.0201
	Infrastructure	Per Capita Urban Road Area	m2	+	0.0281	0.0329	0.0305
		Public Buses per 10,000 Population	vehicles per 10,000 population	+	0.0544	0.0305	0.0424
		Broadband Internet Subscribers	104 households	+	0.0797	0.0214	0.0506
		Drainage Pipeline Density (Built-up)	km/km2	+	0.0242	0.0322	0.0282
		Natural Gas Coverage Rate	%	+	0.0049	0.0323	0.0186
		Year-end Mobile Subscribers	104 households	+	0.0055	0.0480	0.0268
		Per Capita Daily Domestic Water Use	L	−	0.0676	0.0297	0.0486
		Per Capita Natural Gas Supply	104 m3	+	0.1177	0.0267	0.0722

). In the indicator assignment link, the entropy value method and CRITIC weighting method [52] are comprehensively used to combine the assignment to balance the objective information and the internal structural characteristics of the indicators. Finally, based on this methodology, the comprehensive index of urban resilience level of 33 prefecture-level cities in northwest China during 2011–2023 is calculated.

3.2. Methods of Constructing Urban Networks

3.2.1. Transportation Network

The railroad transportation network measures the connection intensity from one city to another by the direct railroad passenger frequency between cities. The data of passenger frequency between two cities in one day is used to calculate the railroad connection intensity. The formula for calculating the railroad connection intensity is:

$$R_{i\leftrightarrow j}=\max (R_{i\to j},R_{j\to i})$$

(1)

where $R_{i\leftrightarrow j}$ is the intensity of railroad connection between cities $i$ and $j$, $R_{i\to j}$ is the number of direct train trips from city $i$ to city $j$, and $R_{j\to i}$ is the number of direct train trips from city $j$ to city $i$.

The gravity model is used to measure the intensity of highway connections in the highway transport network. The highway connection intensity is calculated by using the passenger and freight transport volume data of each city at the end of the year. The formula for calculating the intensity of highway connections [53] is:

$$H_{ij}=K_{ij}\times (\sqrt{P_i{N}_i}\times \sqrt{P_j{N}_j})/{D_{ij}}^2$$

(2)

$$K_{ij}={\displaystyle \frac{1}{2}}({\displaystyle \frac{Q_i+Q_j}{Q}}+{\displaystyle \frac{C_i+C_j}{C}})$$

(3)

where $H_{ij}$ is the intensity of highway connections between city $i$ and city $j$, $K_{ij}$ is the highway connection coefficient between city $i$ and city $j$, $P$ is the permanent resident population at the end of the year, $N$ is the regional GDP data, and $D_{ij}$ is the sum of the road distances between the two cities. $Q_i$ and $Q_j$ represent the total passenger transport volume between city $i$ and city $j$, $C_i$ and $C_j$ represent the total freight transport volume between city $i$ and city $j$, and $Q$ and $C$ denote the average passenger transport volume and average freight transport volume of cities in the northwestern region, respectively.

Determining the intensity of urban links in a comprehensive transportation network requires considering the weights of different transportation modes. The research period of the article is from 2011 to 2023, and the passenger and freight traffic volume data of five provinces in China Statistical Yearbook of past years are selected to calculate the traffic weights of railroads and highways [54]. Specifically, the initial weight of the railway network in the five northwestern provinces for a given year was calculated as the total passenger and freight volume of railways divided by the combined total passenger and freight volume of all transport modes in that year. The initial weight for the highway network was derived in the same manner. Since this study only considers railways and highways, the initial weights of these two modes were normalized to sum to 100%. Thus, the final weight of the railway network in a given year equals its initial weight divided by the sum of the initial weights of both railways and highways. The calculated weights are shown in

Table 3. Weight distribution of urban railways and highways in northwest China.

Weight (W)	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022	2023
Railroad Weight (WR)	0.14	0.13	0.16	0.15	0.14	0.15	0.16	0.17	0.20	0.23	0.22	0.23	0.20
Highway Weight (WH)	0.86	0.87	0.84	0.85	0.86	0.85	0.84	0.83	0.80	0.77	0.78	0.77	0.80

:

Subsequently, the intensity of railroad and highway connections in previous years was standardized using extreme value analysis. Finally, the final urban transport network connection intensity was obtained by adding the weights of the different modes of transport.

$$N_{ij}=W_R\times R_{ij}+W_H\times H_{ij}$$

(4)

where $N_{ij}$ is the transportation network connection intensity between city $i$ and city $j$, $W_R$ is the weight of railroad connection, $R_{ij}$ is the intensity of railroad connection between city $i$ and $j$, $W_H$ is the weight of highway connection, and $H_{ij}$ is the intensity of highway connection between city $i$ and city $j$.

3.2.2. Information Network

Based on the official website of the Baidu Search Index, the city names of 33 prefecture-level cities in northwest China are used as keywords to obtain the overall daily average value of attention between two cities and to construct the information network between the cities. The intensity of the information connection between city $i$ and city $j$ is calculated by the following formula [53]:

$$M_{ij}=I_j\times J_i$$

(5)

where $M_{ij}$ is the information network connection intensity between city $i$ and city $j$, $I_j$ is the overall daily average value of city $i$’s attention to city $j$, and $J_i$ is the overall daily average value of city $j$’s attention to city $i$.

3.2.3. Composite Urban Network

The composite urban network is based on the transportation network and information network connection matrix. The extreme value method is used to normalize the data of each matrix over the years, and the importance of the two different types of flow elements to the city is considered to be the same, with their weights set to 0.5 [55]. The weighted value of the urban network connection for each year is calculated and used as the intensity of the comprehensive urban connection between cities in the northwest region to construct a comprehensive urban connection network matrix. The intensity of composite urban network connections between each pair of cities is:

$$U_{ij}=0.5\times N_{ij}+0.5\times M_{ij}$$

(6)

where $U_{ij}$ is the intensity of the city network connection between city $i$ and city $j$, $N_{ij}$ is the intensity of the transportation network connection between city $i$ and city $j$, and $M_{ij}$ is the intensity of the information network connection between city $i$ and city $j$.

3.3. Model Selection

In the article, the panel model selection, the F test, the BP test, and the Husman test by Stata are used to compare the ordinary least squares (OLS), random-effect (RE), and fixed-effect (FE) models, and the results show that the fixed-effect model is more appropriate. Therefore, the article refers to existing research [35] and chooses the two-way fixed-effect model for benchmark regression to explore the impact of urban network centrality on urban resilience in northwest China. Its formula is:

$${Res}_{i,t}=\alpha +{\beta }_{1}Degree+{\beta }_2{Controls}_{i,t}+{\lambda }_i+{\gamma }_t+{\epsilon }_{i,t}$$

(7)

where $i$ is the city; $t$ is the year; ${Res}_{i,t}$ is the urban resilience; $Degree$ is the city network centrality index; $Controls$ is a series of control variables; $\alpha$ is a constant term; $\beta$ is the parameter to be estimated; ${\lambda }_i$ is the city fixed effect; ${\gamma }_t$ is the time fixed effect; and ${\epsilon }_{i,t}$ is the random error term.

3.4. Variable Setting

3.4.1. Explained Variables

Based on Table 2, the composite index of urban resilience level of 33 cities in northwest China is calculated, and the index of urban resilience level is taken as the explained variable.

3.4.2. Explanatory Variables

City network centrality (

Table 4. Formulas and meanings of urban network centrality indicators and urban network power indicators.

Indicator	Calculation Formula	Meaning of the Formula	Indicator Meaning
Degree Centrality	${Degree}_{DC}={\displaystyle \sum _{j=1}^n}{x}_{ij}, i\ne j$	$x_{ij}$ indicates the numerical value of the direct connection between node city i and node city j.	Reflects the direct connectivity of city nodes within the network. The higher the degree centrality, the more central the city is within the network, indicating stronger resource acquisition and information transmission capabilities.
Betweenness Centrality	${Degree}_{BC}={\displaystyle \sum _{i=1}^n}{\displaystyle \sum _{j=1}^n}{\displaystyle \frac{b_{ij}\left(k\right)}{b_{ij}}},i\ne j\ne k,i<j$	$b_{ij}$ is the number of shortest paths between nodes i and j. $b_{ij}\left(k\right)$ is the number of paths among these shortest paths that pass through node k.	Refers to the number of shortest paths between a node and other nodes in a network, reflecting the node’s ability to act as an intermediary for information transmission in the network, regarded as “control capability.”
Closeness Centrality	${Degree}_{CC}={\displaystyle \sum _i^n}{\displaystyle \frac{1}{b(i,j)}}$	$b(i,j)$ is the shortest path between node i and node j.	Represents the average shortest path length from a node to other nodes, reflecting the node’s propagation capability.
City Network Power	$NP=w_1\times {Degree}_{DC}+w_2\times {Degree}_{BC}+w_3\times {Degree}_{CC}$	${w }_1, {w }_2, \mathrm{a}\mathrm{n}\mathrm{d} w_3$ represent the weight values of degree centrality, betweenness centrality, and closeness centrality, respectively. The study set up uses “${w }_1={w }_2=w_3=1$.”	Reflects the control and influence of node cities within the entire economic network. Cities are interconnected, so cities can “influence” other cities, meaning that cities also possess “power” [39].

), as a key indicator of node control [36], becomes the core entry point for parsing the network–resilience relationship. This article uses the Ucinet 6 software to measure the degree centrality ${(Degree}_{DC})$ of each city in the urban network of northwest China and takes it as the core explanatory variable of the article.

3.4.3. Control Variables

In order to reduce the estimation bias caused by omitted variables and to more accurately assess the impact of the urban network on urban resilience, the article refers to the existing studies in [56,57,58]. The following seven control variables are selected: (1) Financial development level (Fin), measured by the ratio of the balance of deposits and loans of financial institutions to the regional GDP at the end of the year. (2) Urban economic density (Eco), measured by taking the logarithm of the ratio of regional GDP to the land area of the administrative region. (3) Population size (Pop), measured by the logarithm of the year-end resident population. (4) Government governance level (Gov), measured by taking the logarithm of the ratio of year-end general public budget fiscal expenditure to regional GDP. (5) Urbanization rate of the resident population (Urb), measured by the ratio of the resident urban population to the total resident population of the region. (6) Informatization level (Inf), measured by the ratio of total postal and telecommunication revenues to the region’s GDP at the end of the year. (7) Per capita water supply capacity (Wat), measured by the ratio of the total water supply to the total resident population at the end of the year.

4. Empirical Results and Analysis

4.1. Characteristics of Spatio-Temporal Evolution of Urban Resilience in Northwest China

4.1.1. Characteristics of Temporal Evolution of Urban Resilience

From the overall level of analysis, as shown in Figure 2, the annual average resilience index of cities in the northwest region ranges between 0.272 and 0.397, exhibiting an upward trend over the years. This indicates that cities’ ability to adapt to external disturbances is gradually improving, reflecting to some extent the enhancement of urban planning and management capabilities. However, the overall average resilience index remains relatively low, indicating room for further improvement. The coefficient of variation for the resilience level indices of individual cities ranges from 0.15 to 0.24, showing a fluctuating downward trend during the study period. This reflects that the relative differences in resilience among cities in the northwestern region are gradually narrowing, and regional resilience development is showing a converging trend.

When examining the average resilience values (RVs) of cities within the study area (Figure 3), cities such as Urumqi, Karamay, Xi’an, Lanzhou, Yinchuan, Xining, and Jiayuguan exhibit significantly higher RVs than other cities, with the provincial capitals of the five provinces all ranking among the top. This phenomenon highlights the polarization effect of provincial capital cities, whose fundamental cause lies in their role as regional administrative centers and resource allocation hubs, endowing them with stronger capabilities for aggregating factors. In contrast, the average RVs of non-provincial capital cities (excluding Karamay and Jiayuguan) are concentrated in the range of 0.271 to 0.351, showing a significant absolute gap compared to provincial capital cities. Although the decline in the coefficient of variation indicates that the relative gap is narrowing, the absolute imbalance between provincial capital cities and non-provincial capital cities remains prominent. This reflects that while there is a trend toward narrowing regional resilience development gaps, significant imbalances still exist. In terms of growth rates, cities in the northwest region have all shown notable increases in resilience, with the entire region—particularly low-resilience cities—exhibiting a significant catch-up trend. Baiyin (+77%), Longnan (+67%), and Haidong (+67%) lead in growth rates, while Urumqi, Yinchuan, Xi’an, and Karamay, which have a high resilience baseline, rank at the bottom in terms of growth rates. This indicates that cities with lower initial resilience have achieved significantly higher growth rates than high-resilience cities due to the effects of policy support and infrastructure improvements. This growth rate disparity is the key driver behind the narrowing of the aforementioned relative gaps (coefficient of variation), clearly demonstrating the convergence effect of resilience evolution.

4.1.2. Characteristics of Spatial Evolution of Urban Resilience

To visually illustrate the spatial evolution patterns of various cities, the natural break method in the ArcGIS 10.8.1 software was employed to analyze cross-sectional data on urban resilience in the northwestern region from 2011, 2015, 2019, and 2023 (Figure 4).

Urban resilience levels were classified into five tiers: low resilience (RV < 0.2), low–medium resilience (0.2 ≤ RV < 0.3), medium resilience (0.3 ≤ RV < 0.4), medium–high resilience (0.4 ≤ RV < 0.5), and high resilience (RV ≥ 0.5). The spatial pattern of these resilience levels is illustrated in Figure 4.

Overall, during the study period, the urban resilience in the northwestern region improved significantly. The number of low-resilience cities decreased sharply from 6 to 0 by the end of the period, and the number of moderately low–medium-resilience cities decreased significantly from 22 to 3, with their proportion corresponding to a decrease to 9%. Meanwhile, moderately resilient cities experienced a leapfrog growth, increasing from 3 to 23, becoming the main category in the regional resilience level, accounting for 70%. Cities in the medium–high-resilience and high-resilience categories also made significant progress, increasing from two to seven, accounting for 21%. In terms of spatial pattern evolution, the distribution of urban resilience in the northwestern region exhibits a distinct “core leadership—gradient agglomeration” characteristic. Specifically, cities with medium–high and high resilience levels serve as core nodes for regional development, with their spatial distribution being relatively scattered and sparse. In contrast, cities with medium resilience, which dominate in terms of numbers, exhibit a large-scale, contiguous spatial aggregation pattern. This spatial differentiation pattern, characterized by a dispersed distribution of high-resilience nodes and a contiguous distribution of medium-resilience hinterlands, constitutes the typical spatial structure of urban resilience in the northwestern region during this period.

4.2. Form and Characteristics of Urban Network in Northwest China

4.2.1. Overall Network Characterization

This paper utilizes the Netdraw module in the Ucinet 6 software to generate urban network association diagrams for the years 2011, 2015, 2019, and 2023 (Figure 5). The visualization analysis shows that the urban network in the northwest region exhibits a polycentric and complex topology during the period. Although the number of network connections did not show a significant growth trend during the observation period, it can be observed that high-intensity spatial associations are highly concentrated in the core cities of the region, primarily including Xi’an, Yinchuan, Urumqi, and Lanzhou, which form the core hub nodes of the network. Network topology analysis indicates that the sample cities generally have direct or indirect connection relationships within the spatial association network, with overall good network connectivity and no completely isolated “island nodes”.

To further analyze the overall network characteristics,

Table 5. Overall network characteristics of the urban spatial association network in the northwest region.

Year	Network Connectivity	Network Density	Network Centralization	Average Shortest Path
2011	1.00	0.26	0.78	1.74
2012	1.00	0.24	0.80	1.76
2013	1.00	0.23	0.81	1.77
2014	1.00	0.22	0.83	1.78
2015	1.00	0.23	0.82	1.77
2016	1.00	0.25	0.79	1.75
2017	1.00	0.24	0.81	1.76
2018	1.00	0.24	0.80	1.76
2019	1.00	0.25	0.80	1.75
2020	1.00	0.25	0.80	1.75
2021	1.00	0.26	0.79	1.74
2022	1.00	0.25	0.80	1.75
2023	1.00	0.23	0.81	1.77

shows that the urban network connectivity is constant at 1 during the sample period, which further indicates that the nodes in the whole region are in a fully connected state during the study period, laying the foundation for regional integration. The network density shows a trend of decreasing and then stabilizing, but the overall change is not significant, indicating that the spatial connection of the associated network and its closeness between the cities need to be further improved. The network centralization is relatively high, showing a trend of continuous increase and then fine-tuning, indicating that resources are highly concentrated in the core hub cities, and the later slight decrease reflects the limited optimization of the edge nodes, but it has not changed the nature of the high degree of centrality. The average shortest path is adjusted back to the range of 1.74 to 1.77 after the initial prolongation, highlighting the efficiency gain of the hub radiation structure. Although the current urban network in northwest China has operational efficiency, it is highly dependent on the stability of the core nodes; in the future, it will be necessary to cultivate sub-hubs, build redundant paths, and use intelligent scheduling to balance efficiency and robustness to support the sustainable development of the region.

4.2.2. Individual Network Characterization

Centrality is an important indicator to measure the strength of the connection and network status of city network nodes [34]. This paper still takes 2011, 2015, 2019, and 2023 as the cross-section to show the centrality and city network power data of each city in order to portray the position and role of each city in the spatial connection network. As shown in Figure 6, as a whole, the node centrality indicators and city network power values of the cities in northwest China show significant spatio-temporal stability under the composite city network framework. The time-series cross-section radar chart shows that the folds of the indicators of each city highly overlap the trend, which confirms the strong path-dependence property of the network structure. Specifically, Xi’an, Lanzhou, and Urumqi serve as core hubs, with their node centrality and network power values significantly higher than those of other cities. Their advantages are manifested in the following ways: first, as stable leaders in regional development, they control direct connections to resources and play a “hub–gateway” role; second, they act as “central actors” within the network, possessing more efficient cooperation and connectivity capabilities; third, they occupy a core “bridge” position, with Xi’an and Lanzhou serving as the “gateway bridges” connecting northwest China to the east and Urumqi acting as the “hub” linking Xinjiang to the inland regions. During the study period, Xi’an’s “leading” position remained stable; Lanzhou’s centrality saw a slight increase; Urumqi and Yinchuan experienced a slight decline; and the remaining cities showed little change.

4.3. Impact of Urban Network on Urban Resilience in Northwest China

4.3.1. Descriptive Statistics of Variables

In this paper, the mean, median, standard deviation (SD), and minimum and maximum values of each variable are calculated by the Stata 18 software (

Table 6. Descriptive statistics.

Type of Variable	Variable Name	Variable Symbol	OBS	Mean	Median	SD	Minimum Value	Maximum Value
Explained Variables	urban resilience	Res	429	0.315	0.301	0.076	0.164	0.563
Core Explanatory Variables	degree centrality	Degree	429	7.828	6	6.993	1	32
Control Variables	financial development level	Fin	429	2.951	2.451	1.47	1.047	14.43
	urban economic density	Eco	429	6.06	6.039	1.252	2.779	9.383
	population size	Pop	429	5.156	5.375	0.805	3.149	7.176
	government governance level	Gov	429	0.297	0.257	0.16	0.071	0.872
	urbanization rate of the resident population	Urb	429	0.559	0.513	0.194	0.196	0.99
	informatization level	Inf	429	0.025	0.021	0.021	0.004	0.26
	per capita water supply capacity	Wat	429	2.98	2.892	1.213	−0.338	6.202

). As can be seen from Table 6, the observation sample size (OBS) is 429, and all variables have no missing values. The mean value of urban resilience (Res) over the sample period is 0.315, and its standard deviation (SD = 0.076) is 24.1% of the mean, indicating that the level of resilience is relatively less volatile across the cities. The core explanatory variable, degree centrality ${(Degree}_{DC})$, shows a significant right-skewed distribution (mean 7.828 > median 6) and a high degree of dispersion (SD = 6.993), reflecting the heterogeneity of cities in terms of network centrality. Among the control variables, the financial development level, urban economic density, population size, government governance level, urbanization rate of the resident population, informatization level, and per capita water supply capacity also differ significantly across cities. Overall, the SD of each variable is in a reasonable range (SD/mean < 50%), and the median has limited deviation from the mean, which meets the basic requirements of panel model analysis.

4.3.2. Multicollinearity Test

First, the Pearson correlation analysis shows that urban resilience (Res) is significantly and positively correlated with degree centrality (Degree) (r = 0.685, p < 0.001), from which it can be initially judged that there is a strong positive correlation between the two. Second, the results of the variance inflation factor (VIF) test (

Table 7. VIF test.

Variables	VIF	1/VIF
Degree	3.16	0.317
Fin	2.67	0.374
Eco	3.02	0.331
Pop	4.56	0.219
Gov	3.23	0.31
Urb	7.6	0.132
Inf	1.3	0.769
Wat	6.07	0.165
Mean VIF	VIF	3.95

) confirmed that the VIF values of all variables were less than 10 (and tolerance > 0.1), indicating that the effect of multicollinearity was small and negligible.

4.3.3. Benchmark Regression Results

Table 8. Benchmark regression results.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Degree	0.002 *** (0.000)	0.000 *** (0.000)	5.046 *** (1.068)	0.027 *** (0.006)	0.001 * (0.000)	0.001 * (0.000)
Fin	0.002 ** (0.001)	0.001 (0.001)	0.002 * (0.001)	0.002 * (0.001)	0.016 *** (0.002)	0.001 (0.001)
Eco	0.020 *** (0.006)	0.022 *** (0.006)	0.020 *** (0.006)	0.020 *** (0.006)	0.016 ** (0.006)	0.009 (0.006)
Pop	−0.083 *** (0.006)	−0.080 *** (0.006)	−0.084 *** (0.006)	−0.083 *** (0.006)	−0.053 *** (0.018)	−0.075 *** (0.008)
Gov	0.024 (0.021)	0.025 (0.021)	0.022 (0.020)	0.023 (0.020)	−0.025 (0.024)	0.001 (0.019)
Urb	−0.018 (0.014)	−0.020 (0.014)	−0.018 (0.014)	−0.018 (0.014)	−0.031 ** (0.013)	0.007 (0.014)
Inf	0.106 *** (0.032)	0.110 *** (0.032)	0.107 *** (0.032)	0.107 *** (0.032)	−0.006 (0.068)	0.116 *** (0.030)
Wat	−0.005 ** (0.002)	−0.005 ** (0.002)	−0.005 ** (0.002)	−0.005 ** (0.002)	−0.005 * (0.003)	−0.003 (0.002)
_cons	0.562 *** (0.051)	0.550 *** (0.052)	0.489 *** (0.053)	0.562 *** (0.051)	0.424 *** (0.102)	0.559 *** (0.054)
OBS	429	429	429	429	264	377
city	YES	YES	YES	YES	YES	YES
year	YES	YES	YES	YES	YES	YES
R2	0.928	0.927	0.929	0.929	0.911	0.940

Note: ***, **, and * denote passing the significance level test of 1%, 5%, and 10%, respectively, and standard errors are in parentheses.

shows the results of the regression estimation of Equation (7) using a two-way fixed-effects model. Model 1 is where a significant positive correlation can be found between point degree centrality and urban resilience, controlling for other variables (β = 0.002, p < 0.01).

Observing the control variables, the effect of the financial development level (Fin) on urban resilience is positive and passes the significance test, indicating that the improvement of the ability to supply financial resources can effectively enhance the ability of urban risk resistance and recovery. The effect of urban economic density (Eco) on urban resilience is positive and passes the significance test, indicating that the improvement of gross regional product will effectively enhance the level of urban resilience. It is worth noting that the effect of population size (Pop) on urban resilience is negative and passes the significance test, suggesting that the population pressure of megacities may weaken the resilience base through resource crowding and system complexity. The government governance level (Gov) and urbanization rate of the resident population (Urb) positively affect urban economic resilience but do not pass the significance test, implying that the current governance model and urbanization path do not sufficiently support resilience building systematically. Informatization level (Inf) positively affects urban resilience and passes the significance test, highlighting the empowering role of information technology in early disaster warning and emergency response, indicating that the higher the level of informatization, the more positive the impact on urban resilience. Per capita water supply capacity (Wat) negatively affects urban resilience and passes the significance test, which appears to be counter-intuitive on the surface but in fact deeply reflects the resilience logic of water scarcity forcing institutional innovation, technological upgrading, and systematic governance.

4.3.4. Robustness Test and Endogeneity Treatment

In order to test the stability of the core finding that city networks positively promote urban resilience and make the results more robust and reliable, this study conducts a robustness test based on Equation (7) by replacing the core explanatory variables, shortening the sample years, and excluding the core cities.

(1) Replacement of explanatory variables

In this paper, betweenness centrality, closeness centrality, and city network power are chosen to replace degree centrality, and the results are regressed again as new explanatory variables, and the regression results are shown in Models 2, 3, and 4 in Table 8. After replacing the variables, the coefficients of the explanatory variables are 0.000, 5.064, and 0.027, respectively, which all pass the significance test of 1%, and the impact of city network centrality on urban resilience is still significantly positive. It is worth noting that the coefficient of 0.000 in Model 2 is due to rounding rules, and its statistical significance indicates that the variable still has substantive explanatory power. The direction of the control variables is highly consistent with the benchmark regression, confirming the robustness of the results.

(2) Shortening of sample years

Considering the impact of urban network centrality on urban resilience under the shock of the COVID-19 epidemic, this paper adjusts the sample period to 2011–2019, excluding the impact of the shock event of the COVID-19 epidemic, and the results are shown in Table 8 as Model 5. The regression coefficients are positive at the 10% significance level, which proves the robustness of the core finding that urban network centrality positively affects urban resilience.

(3) Exclusion of core cities

As the core cities of the study area, Xi’an, Urumqi, Yinchuan, and Lanzhou not only have a higher status in terms of administrative hierarchy but also have a larger economic scale and stronger factor aggregation capacity compared with other cities; they are usually able to attract more foreign capital by virtue of their unique resource endowment and location advantages. In view of this, in order to eliminate interference, this paper conducted another regression analysis after the exclusion of these four core cities, and the results are shown in Model 6 in Table 8. The regression coefficients are positive at the 10% significance level, indicating that the conclusion that “the centrality of the urban network helps to enhance urban resilience” is robust and reliable.

(4) Endogeneity treatment

This paper adopts the two-way fixed effects of city and time as the baseline model and incorporates multi-dimensional control variables to mitigate potential endogeneity bias. However, urban resilience, as a complex system concept, is characterized by the inexhaustible nature of its influencing factors, and the problem of omitted variables may still exist. Meanwhile, urban network centrality measures are subject to data availability and methodological constraints with measurement errors, and there may be reverse causality between them and urban resilience, which together constitute an endogeneity threat. Therefore, in order to alleviate the endogeneity problem, this study chooses the lagged one-period point centrality as an instrumental variable and uses the two-stage least squares (2SLS) method for estimation, and it ensures the validity of the instrumental variable through the non-identifiable test and the weak instrumental variable test to ultimately obtain the unbiased estimation results (

Table 9. Results of endogeneity analysis.

Variables	First Stage	Second Stage
Degree	0.4976 *** (10.4279)
IV		0.0021 ** (2.3249)
Control Variables	control	control
Anderson canon. corr. LM	95.113 ***
Cragg–Donald Wald F	108.741 [16.380]
OBS	396	396
city	Yes	Yes
year	Yes	Yes

Note: The Anderson canon. corr. LM statistic is the result of the under-identification test for instrumental variables, the Cragg–Donald Wald F-statistic is the result of the weak instrumental variables test, and the critical value of the F statistic at the 10% significance level is in parentheses. *** and ** indicate passing the 1% and 5% significance level tests, respectively, with standard errors in parentheses.

).

The results in Table 9 show that the unidentifiable test (LM statistic p < 0.01) rejected the original hypothesis. The weak instrumental variable test (Wald F statistic > 10% critical value) ruled out the possibility of a weak instrumental variable and confirmed that the instrumental variable was valid. The results of the two-stage regression show that the following: the first stage shows that the instrumental and explanatory variables are significantly correlated; the second stage shows that the regression coefficient of point centrality is 0.0021, which is significant at the 1% level. The direction and values of the urban network centrality coefficients estimated by 2SLS are similar to those obtained by the benchmark regression, further validating the robustness and reliability of the benchmark results.

5. Discussion

5.1. The Dual Effects of Network Centrality and Regional Resilience

The empirical results of this study strongly support the hypothesis of the “empowerment” effect of city networks. Cities with high network centrality significantly enhance their resource acquisition capacity by virtue of their core hub position in the transportation–information complex circulation network: on the one hand, the dense railway/highway flow accelerates the cross-regional dispatch of emergency supplies; on the other hand, the timeliness and high efficiency of the information network facilitates the sharing of risk warnings and collaborative response. This “dual-flow” resource channeling effect directly affects all dimensions of urban resilience, thus systematically enhancing its comprehensive ability to cope with shocks, adapt to changes, and recover.

Although this study mainly observes the positive contribution of network centrality to resilience, the high network centralization of 0.8 reveals potential systemic risks. This structure implies that regional resilience is highly dependent on the stability of a few core nodes. In the event of a major shock to the core hub cities, the risk transmission effect of their dense network connections may be amplified, affecting the peripheral cities that rely on their resource radiation, confirming the potential “synergistic amplification of risk” in the “double-edged sword” effect mentioned in the introduction. The “double-edged sword” effect mentioned in the introduction is a potential “synergistic amplification of risk”. Therefore, in the future, it will be necessary to balance efficiency and robustness; while maintaining the function of the core hub, the cultivation of railroad sub-nodes should be accelerated (such as Baoji, Tianshui, and other cities), and redundant paths should be build so as to avoid regional paralysis triggered by “single-point failure”.

5.2. Spatial Differentiation Patterns and Multi-Factor Driving Mechanisms

The urban resilience of northwest China shows a differentiation pattern of “provincial capital highland–hinterland continuous”. Capital cities rely on administrative resource monopoly and network hub status to form the core of resilience, while the rapid growth of low-resilience cities is highly dependent on external policy interventions [59] (e.g., regional tax incentives, tilting of infrastructure for western development, and introduction of counterpart support technology). However, it is difficult to bridge the absolute gap by “blood transfusion” catching up, and it is necessary to turn to the cultivation of endogenous dynamics.

The control variable analysis further reveals the specificity of the northwest context. Population size (Pop) has a significant negative effect on urban resilience, which is different from the findings of some developed regions [60], deeply reflecting that under the resource and environmental constraints (e.g., water scarcity) and the development stage of northwest China, too large a population agglomeration may exacerbate the competition for resources, environmental pressure, and the complexity of social management, thereby weakening the redundancy and flexibility of the urban system in coping with shocks. The significant positive effect of the information level (Inf) highlights the great potential of digital technology in empowering urban resilience. In the vast and partly infrastructure-poor northwest China, information networks play an irreplaceable role in the rapid transmission of disaster warning information, the precise dispatch of emergency resources, remote medical support, and the maintenance of online public services [43], which are key levers for improving the efficiency and coverage of regional resilience. The negative impact of per capita water supply capacity (Wat) seems counterintuitive at first glance, but in fact it profoundly reveals the complexity of water resource management in the arid and semi-arid regions of northwest China and the “resilience paradox”. Higher per capita water supply capacity is often associated with a crude water use pattern or a particular industrial structure (e.g., high-water-consuming industries, agriculture) [61], which increases the vulnerability of the system under the constraints of the total amount of resources. This result implies that simply increasing supply is not the way to resilience, but promoting institutional innovation (e.g., water rights trading, ladder pricing), technological upgrading (water conservation technology), and systematic governance (water recycling, ecological restoration) to improve water use efficiency and system adaptive capacity is the core path to building water resilience.

5.3. Theoretical Contributions and Limitations

The study confirms the key enabling role of network centrality as a “resource channel” for urban resilience and enriches the application of “flow space” theory in resilience research. At the same time, the study reveals the unique role of network structural characteristics, population size, and water resource management in shaping resilience in underdeveloped and environmentally sensitive regions such as northwest China, deepening the understanding of the drivers of regional resilience at different stages of development. There are some limitations in this study: First, although the equal weighting (0.5 each) of transportation and information flows in constructing the composite network draws on the existing literature, the sensitivity and regional applicability of this approach require further validation. Dynamic optimization methods could enhance the robustness of the network representation. Second, although the resilience evaluation index system covers multiple dimensions, the characterization of ecological and cultural resilience remains insufficient. Existing ecological indicators (such as built-up area green coverage) only reflect scale but fail to effectively capture the functionality of green infrastructure [62] (e.g., ecological connectivity, application of Nature-based Solutions (NbS), etc.). Meanwhile, the cultural dimension lacks quantification of soft factors such as community identity and emergency culture. This limitation is driven by both data accessibility challenges and the gap between theoretical frameworks and practical implementation. Future efforts should introduce more granular data to establish a more sensitive and systematic evaluation framework. Third, while the static correlation analysis based on panel data reveals general patterns, it fails to capture the dynamic interaction mechanisms between network structure and resilience under external shocks (e.g., extreme climate events, public health crises, or major policy interventions). Future research could employ interrupted simulations to assess the co-evolution of network structure and resilience in response to such disturbances [63]. Fourth, this study focuses primarily on the direct effect of network centrality on resilience without sufficiently considering other important mediating or moderating variables. Subsequent studies could use methods such as moderation effect models to further elucidate the complex mechanisms involving multi-factor synergies. Fifth, the urban network constructed in this study is limited to connections within the northwest region and is not embedded in larger-scale networks at the national or global level, making it difficult to account for cross-regional external correlation effects. Future research could integrate regional urban networks into national or global networks to analyze the formation mechanisms of urban resilience under multi-scale nested network effects. Finally, the accessibility of certain data remains a constraint on the depth and local representativeness of this study. For instance, intercity highway passenger trip data and district-/county-level resource allocation data were not fully incorporated. Future efforts should aim to enrich the northwest-specific research database through multi-source data fusion.

6. Conclusions

This study takes 33 prefecture-level cities in northwest China as the object and empirically explores the impact of urban network centrality on urban resilience based on panel data from 2011–2023, and it mainly draws the following conclusions: (1) Coexistence of gradient concentration and convergence in spatial pattern of urban resilience: Regional resilience presents a typical spatial pattern of “core leadership–gradient concentration”; the growth rate of low-resilience cities is faster, and the relative gap between regions is narrowed, but the absolute resilience gap between provincial capitals and non-provincial capitals is still significant. (2) Coexistence of network structure efficiency and risk: The regional urban network has good connectivity and high operational efficiency, but its highly centralized structure makes it vulnerable to systemic risk transmission and amplification when facing core node shocks. (3) Network centrality empowers resilience: A city’s centrality in the network is the core channel for it to access key resources and enhance comprehensive resilience and has a significant and robust positive effect on urban resilience. It has a significant and robust positive facilitating effect on urban resilience. (4) Multi-dimensional factors jointly shape resilience: Financial development level, economic density, and informatization are important pillars supporting resilience, whereas the pressure brought by over-concentration of population and the rough pattern of water resource utilization constitute bottlenecks for resilience enhancement. This study provides an important basis for deepening the understanding of network-driven mechanisms and spatial differentiation patterns of urban resilience in less developed regions.

Based on the conclusions of this study, the following suggestions are made to enhance the resilience of cities in northwest China and optimize the urban network: First, a new resilience-oriented regional network pattern should be built: While consolidating the functions of the core hubs of Xi’an, Urumqi, Lanzhou, etc., the study focuses on fostering the sub-hubs of Yinchuan, Xining, and potential non-provincial capitals (e.g., Baoji, Tianshui) so that the network structure will be “multicentric and distributed” and the network structure will be “distributed” so as to form a “multicenter, distributed” network structure to enhance system robustness. Transportation and information network connections should be continuously encrypted, especially to improve the access quality and efficiency of low-resilience cities and edge cities. Second, categorized measures should be taken to enhance the resilience kernel of cities: Core cities should focus on preventing major risks (especially network dependence risks), strengthening system redundancy and smart emergency response capabilities, and promoting green transformation of industries. Medium- and low-resilience cities need to actively integrate into the network to leverage development, make up for the shortcomings of infrastructure (disaster prevention, information, public services), strictly control sprawl to ease population pressure, and develop specialty industries to strengthen urban resilience. Third, key resilience support elements should be strengthened: Digital empowerment should be comprehensively promoted, combined with the “east counts west counts” strategy, wisdom, and computing power resources, into the depth of the resilience of all aspects of construction. The strictest water resource resilience management should be implemented, and institutional innovation, technical water conservation, and recycling should be promoted. Innovative financial tools should support resilient infrastructure and green transformation. Fourth, regional coordination and common governance should be deepened: A cross-regional resilience coordination mechanism should be established, regional risks should be jointly responded to, resource sharing and mutual assistance in emergencies should be promoted, regional risk assessment and joint planning should be carried out, and an overall leap in regional resilience should be realized.