Spatio-Temporal Dynamic of Rural Resilience to Multiple Water-Related Hazards: A Case Study in Zhejiang Province, China

Abstract

Rural areas face increasing threats from water-related disasters yet often lack the infrastructure and resources available in urban areas for effective disaster response and recovery. Understanding and improving rural resilience—the ability to withstand and recover from disasters—is critical for sustainable rural development, especially under pressures from the climate. This study introduces a practical, indicator-based framework to evaluate rural resilience by analyzing five key aspects: stability, hazards, exposure, vulnerability, and adaptability. Using Zhejiang Province in China as a case study, we examined changes in rural resilience from 2000 to 2020. Our findings reveal that regions with a strong economic foundation, diversified livelihoods, and well-developed healthcare and education infrastructure exhibit higher resilience, while those with high exposure to hazards and economic dependency on agriculture remain vulnerable. The study highlights the crucial role of adaptive capacity in driving long-term resilience, emphasizing targeted investments in rural healthcare, education, infrastructure, and economic diversification. By providing data-driven insights, this research contributes to SDG 11 by offering practical strategies for policymakers and local communities to enhance disaster preparedness and rural sustainability. Moreover, the framework developed here can be adapted and applied to other rural regions facing similar hazards globally, enhancing disaster preparedness and promoting sustainable rural development.

1. Introduction

Coastal areas are particularly vulnerable to water-related disasters, a vulnerability that has been exacerbated by climate change and urbanization [1]. Natural hazards, such as floods, typhoons, and storm surges, are escalating in both frequency and intensity, posing severe challenges, especially to communities located along vulnerable coastlines [2]. Urban areas and centers, equipped with comprehensive infrastructure systems and a highly educated populace, demonstrate a greater capacity to manage and mitigate the impacts of such multi-hazard events [3]. These urban areas benefit from advanced utility infrastructure systems, robust transportation networks, and redundant green–blue space, which collectively enhance their resilience [4]. In contrast, rural areas, especially those in coastal regions, often lack the infrastructure and adaptive capacity necessary to effectively respond to and recover from water-related disasters [5]. This urban–rural resilience gap is particularly pronounced in developing countries, including China, where rapid urbanization has significantly exacerbated rural vulnerabilities.

Research on rural resilience has largely emphasized defining its conceptual framework [6], measurement, and related influencing factors [7]. Historically, like resilience theory in general, rural resilience was initially defined as the capacity of rural areas to absorb disturbances and reorganize while retaining their essential functions, structures, identities, and feedback mechanisms [8]. This perspective emphasizes the ability of rural communities to withstand shocks without undergoing fundamental changes. However, further studies have expanded this definition, recognizing resilience as not only the capacity to absorb and adapt to disturbances but also the potential for transformation and renewal [9]. For instance, Di Nardo et al. developed an adaptive resilience framework applied to high-capacity railway networks, demonstrating how adaptive concepts initially employed in infrastructure contexts can provide valuable insights and transferable strategies for enhancing rural resilience [10]. Disturbances can induce significant changes in rural systems, prompting a shift from mere response and adaptation to a broader process of community transformation [11]. Specifically, rural resilience refers to the capacity of social groups to reduce risks, manage disaster impacts effectively, and facilitate recovery efforts in a manner that limits social disruption while lessening the impact of future hazards [12].

Evaluations of rural resilience have employed various methodologies, which have evolved over time. Existing approaches mainly consist of indicator-based methods, model-driven simulations, and various toolkits. The indicator-based approach establishes a measurement framework by incorporating multiple factors and deriving a composite index through techniques like hierarchical analysis, principal component analysis, and stakeholder-centered methodologies [7]. These methods are advantageous due to their convenient operation and ease of comparison using socioeconomic data, making them widely applicable [13]. Model-based methods predominantly focus on the resilience of urban communities, infrastructure systems, urban ecosystems, and economic benefit–cost analyses or complex urban system models [14]. These models typically utilize recovery time or recovery efficiency to provide a detailed description of how urban systems respond to disturbances. Tools or toolkits offer ready-made mechanisms and procedures for assessing community resilience. These tools include sample procedures, survey instruments, models, and composite indicators, providing comprehensive frameworks for resilience evaluation [15].

Despite methodological advancements, there is a notable research gap in the application of quantitative methods to rural resilience. Most existing studies focus on urban resilience, often addressing single hazards with short-term impacts. The measurement of rural resilience, however, requires a more holistic approach that goes beyond monetary loss reduction. It should encompass the ability to recover during the post-disaster period, capturing long-term adaptive capacities and transformation potentials. Current research lacks comprehensive evaluations of rural resilience to multiple hazards over extended time frames. This gap is critical, as rural areas face unique challenges and vulnerabilities that are not adequately addressed by urban-focused studies. Understanding and measuring rural resilience necessitate a multi-dimensional approach that integrates various factors influencing resilience and examines their interactions over long periods.

For coastal areas in Zhejiang Province, China, two primary factors contribute to frequent and severe water-related hazards. Firstly, severe typhoons coupled with extreme precipitation often result in river flooding and storm surges in this region. The geographical location and climatic conditions of Zhejiang make it particularly susceptible to these natural events [16]. Secondly, rapid urbanization over the past four decades has led to significant disparities in infrastructure quality between urban and rural areas. The influx of educated individuals into urban centers further exacerbates the vulnerability of rural areas by depleting their human capital. This disparity undermines the adaptive capacity of rural communities, making them more susceptible to the adverse impacts of water-related hazards [17]. As a result, reducing these impacts, strengthening resilience, and pinpointing key factors influencing rural resilience have become crucial issues for promoting sustainable development in rural areas.

By integrating and adapting insights from boundary-crossing urban frameworks, such as those applied by Bixler et al. [18], this research contributes by proposing a quantitative framework specifically designed for evaluating rural resilience to multiple water-related hazards. This framework incorporates rural-specific indicators designed to capture the distinctive characteristics of rural areas in China. By examining the temporal and spatial dynamics of rural resilience in Zhejiang Province from 2000 to 2020, this study offers a comprehensive analysis of resilience trends and their underlying determinants over two decades. The extended temporal coverage enables a deeper understanding of long-term resilience trajectories and their driving forces, an aspect often overlooked in previous research. Additionally, the study investigates the differentiated impacts of various water-related hazards on rural resilience and identifies key influencing factors. Methodologically, this work contributes to the literature by developing rural-specific indicators and applying them across an unprecedented 20-year timeframe and provincial spatial scale, ultimately providing actionable strategies to improve rural resilience (Figure 1).

2. Study Area

This study was conducted in Zhejiang Province, located along China’s eastern coastline (118°01′–123°10′ E, 27°02′–31°11′ N) (Figure 2). Zhejiang features a diverse landscape characterized by extensive coastal plains in the east and mountainous terrain in the west, covering a total area of approximately 105,500 km². The province experiences a subtropical monsoon climate, with annual precipitation ranging between 1100 and 2000 mm. This climatic condition makes Zhejiang particularly vulnerable to water-related hazards, such as typhoons and flooding [19].

Zhejiang’s rural areas, home to over 17 million residents, frequently face socioeconomic challenges, limited infrastructure, and fragile livelihoods, amplifying their vulnerability and underscoring the necessity of targeted resilience strategies. From 2010 to 2019, water-related hazards resulted in direct economic losses totaling CNY 170.399 billion across both rural and urban areas [20].

3. Materials and Methods

3.1. Data Sources

Multiple datasets were combined to ensure a comprehensive assessment and high spatial–temporal resolution. Land use data (2000–2020) were obtained from the Resource and Environment Data Cloud Platform (30 m resolution). Elevation data were sourced from the Alaska Satellite Facility (12.5 m resolution). Meteorological data (rainfall and typhoon records from 2000 to 2020) were collected from the National Meteorological Center and China Meteorological Administration databases. Socioeconomic data (economic indicators, employment, infrastructure investments) were drawn from Zhejiang Statistical Yearbooks. Additional data, including population censuses, points of interest (education, medical facilities), and transportation networks, were compiled from official Chinese statistical bureaus, the Amap API, and OpenStreetMap, respectively (

Table 1. The sources of data used in this study. (The spatial distributions of each indicator are provides in Supplementary Materials, Figures S2–S6).

No.	Data Type	Description	Source	Resolution
1	Land use and land cover change	2000, 2005, 2010, 2015, 2018, and 2020 land use	https://www.resdc.cn	30 m
2	Elevation	DEM	https://search.asf.alaska.edu/#/	12.5 m
3	Rainfall	Rainfall records from rainfall weather stations from 2000 to 2020	https://www.cma.gov.cn/
4	Typhoon	The CMA Tropical Cyclone Database from 2000 to 2020	https://tcdata.typhoon.org.cn/en/
5	Social economy	Economy, employment, and utility infrastructure data from 2000 to 2020	Zhejiang County Statistical Yearbook Zhejiang City Statistical Yearbook
6	Points of interest	Businesses service, education centers, medical service, sports service, etc.	https://lbs.amap.com/
7	Population data	National census (5th, 6th, 7th)	https://www.stats.gov.cn/
8	Traffic	Guodao, Shengdao, highway, primary, secondary, and other roads	https://www.openstreetmap.org/

). To prevent the confounding effects of highly correlated variables, correlation analysis was performed prior to subsequent analytical procedures (Supplementary Materials, Figure S1).

3.2. Methodology

3.2.1. Rural Resilience Measurement

This study evaluates rural resilience to various water-related hazards by assessing the extent of losses within the rural community system [21]. To quantify resilience, we compare the community’s stable (normal) state (S) with its disrupted state (D) following hazardous events:

$$R=(S_i-D_i)/S$$

Here, R indicates resilience in year I; $S_i$ is the community’s stable state for that year; $D_i$ is its disrupted state due to disasters; and the stable state (S) is averaged across the entire study period (2000–2020).

The calculation of the disrupted state D is defined based on traditional resilience theory and the disaster risk assessment model recommended by the ISO Disaster Management Principles and Guidelines (ISO 31000:2018) [22].

The disrupted state (D) integrates hazard intensity (H), community exposure (E), inherent vulnerability (V), and adaptive capability (A), following widely accepted international disaster assessment frameworks:

$$\mathrm{D}=S\times \mathrm{H}\times \mathrm{E}\times \mathrm{V}/\mathrm{A};$$

This formulation acknowledges the multidimensional nature of resilience and provides a balanced reflection of community preparedness and vulnerability.

3.2.2. Selection of Rural Indicators

Indicators were carefully selected based on the established community resilience literature [23,24] and adapted explicitly for the rural context of Zhejiang Province. Each dimension (S, H, E, V, A) is represented through clearly defined socioeconomic, demographic, and environmental variables (see

Table 2. Indicators of rural resilience.

Dimension	Variables	Indicators	References
Stable state (S)	Social conditions	Total population (S_Soc_Pop)
		Electricity consumption per km2 (S_Soc_Elec)	[25]
	Economic conditions	GDP per km2 (S_Eco_GDP)
		Public infrastructure investment per km2 (S_Eco_Invest)
	Land conditions	% water area (S_Land_Water)	[26]
		Average elevation (S_Land_DEM)	[27]
Hazard (H)	River flood disaster risk	Slope (H_RF_Slope)
		Flood season water volume (H_RF_FloodSeason)	[27]
	Inundation disaster risk	Extreme daily precipitation (H_In_ExtremeRain)	[28]
		Average impermeability rate (H_In_Impermeability)	[28]
	Typhoon disaster risk	Distance from the coastline (H_Ty_DistCoastline)	[28]
		Tropical cyclones disaster index (H_Ty_DistCoastline)
Exposure (E)	Social conditions	Rural population (E_Soc_RuralPop)	[5]
	Economic conditions	Gross product of the primary industry per km2 (E_Eco_1stGDP)	[13]
	Land conditions	% rural residential area (E_Land_Rural)	[29]
Vulnerability (V)	Social conditions	% elderly people (V_Soc_Elder)	[13]
		% young people (V_Soc_Young)	[30]
		% women (V_Soc_Women)	[13]
	Economic conditions	Structural imbalance between agriculture and non-agricultural industries (V_Eco_Imbalance)	[31]
	Land conditions	% cropland area (V_Land_Cropland)
Adaptative capability (A)	Social conditions	% educated population (A_Soc_Educated)	[13]
		Number of medical staff in hospitals and health clinic (A_Soc_MedicalStaff)	[27]
		Number of beds in hospital and health clinic (A_Soc_MedicalBed)	[27]
		Number of employees in government (A_Soc_Employee)	[13]
	Economic conditions	Stability of rural labor force (A_Eco_Stable)	[29]
		Rural residents’ disposable income per capita (A_Eco_Income)	[13]
		Savings deposit held by urban and rural residents per capita (A_Eco_Deposit)	[13]
	Land conditions	% grassland area (A_Eco_Grass)
		% forest land area (A_Eco_Forest)

for full details). Indicators reflect data availability, relevance to resilience theory, and the specific socioeconomic realities of rural Zhejiang.

3.2.3. Weights Approach

We employed the entropy weighting method due to several critical advantages over other commonly used multi-criteria decision making methods, particularly the Analytic Hierarchy Process (AHP) and principal component analysis (PCA) (Supplementary Materials).

First, unlike AHP, entropy weighting does not rely on subjective expert judgments, which can introduce significant biases based on individual expertise or perceptions. AHP demands considerable expert input, making it potentially inconsistent when applied across different regions or contexts.

Second, compared to PCA, entropy weighting has fewer statistical restrictions and assumptions. PCA assumes that indicators are linearly correlated and normally distributed, conditions rarely fully satisfied in real-world resilience data, particularly with complex and heterogeneous socioeconomic datasets, such as those used in this study. The entropy weighting method does not require data normality or strict linear independence among indicators, providing greater flexibility for incorporating diverse data types.

3.2.4. Cluster Analysis

Cluster analysis identified distinct temporal patterns of resilience evolution across Zhejiang Province. We tested multiple clustering techniques (K-means, agglomerative, bisecting K-means, spectral clustering) and evaluated their performance using silhouette scores. Based on this comparison, bisecting K-means emerged as the preferred method due to its balance of accuracy, computational efficiency, and interoperability. By grouping units with similar resilience indicators, the clustering results provided insights into shared strengths, vulnerabilities, and evolution patterns across different rural areas.

3.2.5. Regression Model

A multiple linear regression model analyzed how selected indicators affected rural resilience. The regression quantified each factor’s relative contribution to resilience, enabling clear identification of priority areas for policy interventions.

4. Result

Using this comprehensive methodology, we conducted detailed spatial–temporal analyses to capture resilience dynamics across Zhejiang Province. The following results highlight key patterns, illustrating how resilience varies regionally and over time and revealing insights into the primary factors shaping rural resilience.

4.1. Spatial and Temporal Variations in Rural Resilience in Zhejiang Province

During the study period, rural resilience in Zhejiang Province exhibited significant temporal evolution and distinct spatial variation. As depicted in Figure 3, different regions demonstrated varying levels of resilience over time. For instance, Lishui (LS) displayed a substantial increase in resilience from 2010 onwards, maintaining the highest resilience levels among all regions by 2020. Conversely, Hangzhou (HA) and Ningbo (NB) showed more stable resilience patterns, with minor fluctuations throughout the period. The resilience trends in Wenzhou (WZ) and Taizhou (TZ) were more volatile, with noticeable peaks and troughs, reflecting their susceptibility to episodic natural hazards.

Spatially, as illustrated in Figure 4, resilience levels varied significantly across the province. Hangzhou (HA) and Huzhou (HU) exhibited relatively high resilience, consistently ranking in the medium–high to high categories. In contrast, regions like Quzhou (QZ) and Jinhua (JH) displayed lower resilience levels, indicating greater vulnerability to environmental and socioeconomic challenges. The spatial analysis underscores that resilience is unevenly distributed, as some areas near the coastline with better economic development are equipped to cope with and recover from water-related hazards.

The analysis demonstrates that the model developed in this study effectively captures the temporal and spatial variations in rural resilience with high accuracy.

4.2. Typical Temporal Pattern of Rural Resilience

As shown in Figure 5, the silhouette score decreases as the number of clusters increases, with bisecting K-means consistently delivering the best performance and thus confirming its suitability for this analysis. Applying the selected clustering approach to the temporal rural resilience data yielded seven typical temporal patterns, illustrated in Figure 6, which reveal how resilience varies across different regions.

Type I (Linan): These regions began with low resilience levels, saw a significant increase, and then saw a decline, stabilizing at a moderate level by the end of the period.

Type II (Dongyang): These areas showed a steady rise in resilience, with a peak followed by some fluctuations before stabilizing at a slightly lower level.

Type III (Wencheng): Starting with moderate resilience, these regions experienced a sharp increase, followed by a decline and fluctuations, ending at a moderate level.

Type IV (Shengsi): This unique pattern displayed a stable low resilience level initially, followed by a sharp increase late in the period, indicating substantial improvement.

Type V (Shangyu): Beginning at a high resilience level, these regions saw a sharp decline, followed by a gradual improvement and stabilization at a moderate level.

Type VI (Sanmen): These regions started with a higher resilience level and experienced a continuous decline to a lower point, followed by minor improvements and stabilization.

Type VII (Dongtou): Starting at a high resilience level, these regions showed a steady decline, ending at a much lower level by the study period’s end.

The temporal patterns, particularly Types I, II, and III, located in mountainous regions start at lower levels and, despite fluctuations, eventually stabilize at middle or low levels. In contrast, Types V, VI, and VII, primarily in coastal areas, begin at higher levels but decline due to various impacts, settling at middle and low levels by the end of the study period. These patterns demonstrate clear spatial heterogeneity, underscoring the diverse resilience trajectories across different landscapes in Zhejiang Province.

4.3. Five Dimensions Influencing Rural Resilience

The multiple linear regression model between five dimension and rural resilience shows a high R-squared value of 0.908, indicating that the model explains 90.8% of the variance in rural resilience. The results are presented in the

Table 3. Statistical testing results of the dimension variables using the regression model.

Dimension	Coefficients	Standard Error	t	p > |t|	[0.025	0.975]
Stable state (S)	0.0021 ***	0.001	3.193	0.001	0.001	0.003
Hazard (H)	−0.0172 ***	0	15.786	0	0.006	0.008
Exposure (E)	−0.0274 ***	0.001	36.169	0	0.026	0.029
Vulnerability (V)	−0.0215 ***	0.001	38.926	0	0.02	0.023
Adaptative capability (A)	0.0534 ***	0.001	81.924	0	0.052	0.055

Note: Asterisks indicate significance levels: *** 1% level.

.

The regression results highlight that each dimension has a statistically significant impact on rural resilience, with varying degrees of influence.

The dimension of stable state (S) exerts a statistically positive influence on rural resilience, as evidenced by a coefficient of 0.0021 (*** p < 0.001), which underscores its lower importance. Empirical observations from the provided maps indicate that regions like Hangzhou (HA) and Huzhou (HU), which consistently rank high in terms of stability, manifest elevated resilience levels over the evaluated periods. Regions like Taizhou (TZ) and Ishui (LS) show variations in stability over the years, with some fluctuations in the color coding that suggest changes in their stable state rankings. High stability areas were concentrated in certain zones, potentially indicating areas with better economic growth, infrastructure investment, and governance. Lower stability regions might be facing challenges, such as slower economic development, political instability, or inadequate public services (Figure 7).

Hazard (H) significantly undermines rural resilience, as indicated by a substantial negative coefficient of −0.0172 (*** p < 0.001). This dimension quantifies the likelihood and severity of water-related threats, such as rainfall inundation, typhoons, and river floods. The series of maps from 2000 to 2020 reveals a spatial and temporal variation in hazard across regions. Notably, Ningbo (NB) and Wenzhou (WZ) consistently exhibit high hazard, aligning with observed decreases in resilience during periods of severe natural disasters. These maps elucidate the dynamic nature of hazards, showing how some regions like Taizhou (TZ) and Jinhua (JH) see fluctuating levels of risk over the two decades, potentially correlating with environmental and climate changes. The visual data underscore that high-risk areas generally experience sharper declines in resilience, reflecting the critical impact of managing and mitigating hazards to bolster rural resilience (Figure 8).

The dimension of exposure (E) substantially detracts from rural resilience, as evidenced by a significant negative coefficient of −0.0274 (*** p < 0.001). Exposure encompasses the socioeconomic degree to which regions are subject to various water-related disasters. Analyzing the exposure maps from 2000 to 2020, a clear spatial and temporal pattern emerges. Regions like Taizhou (TZ) and Zhoushan (ZS), which consistently exhibit higher levels of exposure, correspondingly demonstrate lower resilience levels over the years. The exposure levels in 2010 generally seem higher across most regions compared to other years, perhaps reflecting particular environmental or meteorological conditions prevalent during that period, which indicated dramatic climate change in recent years (Figure 9).

Vulnerability (V) significantly undermines rural resilience, as evidenced by a robust negative coefficient of −0.0215 (*** p < 0.02), which underscores its critical role in regional sustainability. An examination of the vulnerability maps from 2000 to 2020 reveals both temporal fluctuations and spatial variations in vulnerability across the region. Specifically, in the northern and coastal regions, such as Zhoushan (ZS) and Taizhou (TZ), high vulnerability is more persistent, while areas like Jiaxing (JX) and Shaoxing (SX) consistently display heightened vulnerability, which correlates with their reduced resilience levels over the observed period. However, regions like Hangzhou (HA) and Ningbo (NB) consistently show lower vulnerability levels. There is a notable trend of some regions improving in terms of reducing their vulnerability by 2020, likely due to investment in the primary economy, especially in agriculture (Figure 10).

Adaptive capability (A) demonstrates the most substantial positive impact on rural resilience, with a significant coefficient of 0.0534 (*** p < 0.052), illustrating its pivotal role in sustaining regional development. This dimension reflects the ability of a region to adapt and respond proactively to varying conditions, encompassing the capacity to learn from past experiences, foster innovation, and deploy effective adaptation strategies. The series of maps spanning from 2000 to 2020 reveal a notable temporal evolution and spatial variation in adaptive capabilities across regions. In particular, regions like Hangzhou (HA) and Jinhua (JH) exhibit consistently high levels of adaptive capability, correlating with marked improvements in resilience. On the other hand, regions like Ishui (LS) show lower levels of adaptive capability, which could be due to various factors, such as limited healthcare accessibility, less effective governance management, or fewer deposit savings. Also, there is a clear trend of increasing adaptive capabilities in several regions, which correlates with an improvement in overall resilience (Figure 11).

The spatial and temporal patterns illustrated in the adaptive capability maps clearly correlate with the regression results, emphasizing the intricate relationship between regional stable state, hazard, exposure, vulnerability, and adaptive capabilities and overall rural resilience. High levels of adaptive capability consistently bolster resilience, enhancing a region’s ability to manage and recover from socioeconomic and environmental stresses. Conversely, regions marked by high exposure and vulnerability exhibit reduced resilience, underscoring the need for targeted interventions to mitigate these risks and fortify adaptive responses. This dynamic interplay highlights the critical role of adaptive capacity in sustaining and enhancing rural resilience across varied landscapes.

4.4. Factors Influencing Rural Resilience

The analysis of factors influencing rural resilience utilized a multiple linear regression model, with various indicators representing the dimensions of stable state (S), hazard (H), exposure (E), vulnerability (V), and adaptive capability (A). The statistical testing results, summarized in

Table 4. Statistical testing results of the indicator variables through multiple linear regressions.

Dimension	Indicators	Coefficients
Stable state (S)	S_Soc_Pop	−0.019 *
	S_Soc_Elec	−0.049 ***
	S_Eco_GDP	0.035 ***
	S_Eco_Invest	0.019
	S_Land_Water	0.018 ***
	S_Land_DEM	−0.012 **
Hazard (H)	H_RF_Slope	−0.011
	H_RF_FloodSeason	−0.006
	H_In_ExtremeRain	−0.009 **
	H_In_Impermeability	−0.037 ***
	H_Ty_DistCoastline	−0.015 ***
	H_Ty_TyphoonIndex	−0.066 ***
Exposure (E)	E_Soc_RuralPop	0.001
	E_Eco_1stGDP	−0.237 ***
	E_Land_Rural	−0.091 ***
Vulnerability (V)	V_Soc_Elder	−0.047 ***
	V_Soc_Young	0.006
	V_Soc_Women	−0.041 ***
	V_Eco_Imbalance	0.026 ***
	V_Land_Cropland	−0.17 ***
Adaptative capability (A)	A_Soc_Educated	0.035 ***
	A_Soc_MedicalStaff	0.012
	A_Soc_MedicalBed	0.049 ***
	A_Soc_Employee	0.109 ***
	A_Eco_Stable	−0.037 ***
	A_Eco_Income	0.001
	A_Eco_Deposit	0.451 ***
	A_Eco_Grass	0.084 ***
	A_Eco_Forest	0.015 **

Note: Asterisks indicate significance levels: *** 1% level, ** 5% level, and * 10% level.

, reveal the significance and impact of these indicators on rural resilience. The regression model shows a high R-squared value of 0.941, indicating that the model explains 94.1% of the variance in rural resilience.

The stable state dimension significantly influences rural resilience through various socioeconomic and land-related indicators. The model reveals a slightly negative effect of higher population densities on resilience (S_Soc_Pop: −0.019 *), suggesting that increased demands on resources and services may hinder resilience. Conversely, access to electricity shows a more significant negative impact (S_Soc_Elec: −0.049 ***), emphasizing the crucial role of reliable energy infrastructure. In contrast, a higher Gross Domestic Product positively correlates with resilience (S_Eco_GDP: 0.035 ***), indicating that economic strength supports resilient communities. Additionally, water resources are positively linked to resilience (S_Land_Water: 0.018 ***), while elevation diversity presents challenges, as shown by its negative correlation with resilience (S_Land_DEM: −0.012 **).

In the hazard dimension, the coefficients indicate that terrain slope (H_RF_Slope: −0.011) and flood season (H_RF_FloodSeason: −0.006) have minor, though not statistically significant, negative effects on resilience, suggesting that these factors slightly exacerbate hazards. More significant is the role of extreme rainfall (H_In_ExtremeRain: −0.009 **), which notably heightens hazards. Furthermore, the presence of impervious surfaces (H_In_Impermeability: −0.037 ***) greatly increases these risks by hindering water absorption and increasing runoff, while proximity to coastlines (H_Ty_DistCoastline: −0.015 ***) also contributes significantly to higher hazard due to increased exposure to storm surges and coastal flooding. The most substantial impact is associated with the typhoon index (H_Ty_TyphoonIndex: −0.066 ***), reflecting the severe risk increase with more frequent or intense typhoons.

In the exposure dimension, rural population (E_Soc_RuralPop) has a minimal impact (0.001), while the economic contribution of the primary sector (E_Eco_1stGDP, −0.237 ***) shows a substantial negative correlation with exposure, suggesting that higher economic output is strongly associated with reduced vulnerability to hazards. Additionally, land use patterns in rural areas (E_Land_Rural: −0.091 ***) also show a notable negative impact, where increased usage of land for rural purposes might correlate with lower exposure to risks.

In the vulnerability dimension, the presence of elderly populations (V_Soc_Elder) and women (V_Soc_Women) within a community are both associated with increased vulnerability, as evidenced by negative coefficients of −0.047 *** and −0.041 ***. This suggests that these demographics might be more susceptible to hazards, requiring targeted social support measures. Conversely, the proportion of young people (V_Soc_Young) shows a minimal and statistically insignificant effect on vulnerability, as indicated by a coefficient of 0.006. Economic imbalance (V_Eco_Imbalance) has a coefficient of 0.026 ***. However, the use of land for cropland (V_Land_Cropland) is significantly associated with a decrease in vulnerability, with a coefficient of −0.17 ***.

In the adaptive capability dimension, higher-educated people (A_Soc_Educated) positively impact adaptive capability with a coefficient of 0.035, indicating that better-educated populations are better equipped to implement and benefit from adaptation strategies. The presence of medical facilities and staff, indicated by coefficients for medical beds (A_Soc_MedicalBed: 0.049) and medical staff (A_Soc_MedicalStaff: 0.012, not statistically significant), suggests that better healthcare infrastructure also contributes to greater capacity, although the impact of medical staff alone is minimal. Employment levels (A_Soc_Employee: 0.109) strongly correlate with adaptive capability, affirming that job availability is crucial for fostering resilience. Economically, while stability (A_Eco_Stable: −0.037) shows a negative correlation, suggesting that overly stable economic conditions might reduce the impetus for adaptation, other factors, such as bank deposits (A_Eco_Deposit: 0.451) and agricultural productivity, represented by grassland productivity (A_Eco_Grass: 0.084), show strong positive impacts, highlighting the role of financial resources and agricultural viability in supporting adaptive strategies. Surprisingly, income levels (A_Eco_Income: 0.001) and forest resources (A_Eco_Forest: 0.015) show less pronounced effects while still contributing positively to adaptive capabilities.

5. Discussion

5.1. Rural Resilience Policy Recommendations

The quantitative analysis presented in this study reveals critical insights regarding rural resilience dynamics, particularly emphasizing adaptive capability’s substantial role. To provide a deeper, more humanized contextualization of our findings, we illustrate these statistical outcomes with examples from Lishui and Taizhou.

For instance, regions like Lishui have exhibited notable improvements in resilience since 2010, attributed mainly to strategic local government initiatives under rural revitalization policies aimed at enhancing adaptive capabilities. These initiatives include financial investments in healthcare infrastructure and ecological development, increased educational opportunities, improvements in environmental quality, and targeted rural employment programs [32]. Conversely, areas like Taizhou have shown fluctuating resilience, impacted by recurrent water-related hazards and limited adaptive measures due to constrained financial resources and infrastructure deficits. In the future, Taizhou’s elevated resilience could be enhanced further by addressing any subtle vulnerabilities; continuous investment in innovation and social services is needed to maintain momentum [33].

Based on the statistical patterns and these case insights, we propose several concrete policy measures to reinforce rural resilience. (1) Invest in adaptive capacity and infrastructure; increase funding for modern agricultural infrastructure and digital technology in rural areas [34]. (2) Promote economic diversification and livelihood security; encourage a shift away from single-sector dependence by diversifying rural economies [35]. (3) Strengthen social services and community resilience; targeted investments should also go to health, education, and housing infrastructure in rural communities.

These measures provide a roadmap for policymakers to design and implement effective strategies that enhance rural resilience, thus ensuring that rural communities are better equipped to adapt to and thrive in an increasingly uncertain world.

5.2. Future Directions for Data-Driven and Quantitative Research

The evolving challenges in rural resilience demand an equally advanced research toolkit. Quantitative and data-driven approaches are becoming increasingly important in resilience research, offering precision and objectivity in understanding complex dynamics. In recent years, the field has seen a surge in metrics and models that allow researchers to measure and compare resilience across communities and time periods [36]. This shift is revolutionizing how we assess resilience; instead of solely relying on qualitative case studies, scholars now leverage statistics and machine learning to uncover patterns that might otherwise go unnoticed [37]. For example, our study’s panel dataset (2000–2020) enabled us to capture trends and shifts in resilience over two decades, revealing an overall upward trajectory and significant regional disparities that a single-year snapshot could not show. Such a spatio-temporal analysis (tracking changes across both space and time) has proven valuable for identifying where and when resilience improves or declines and correlating those changes with policy interventions or external events.

One key future direction is the expanded use of spatio-temporal datasets to capture resilience dynamics in real time and at finer scales. Traditional studies often lacked granular temporal data, which limited insight into how systems respond during and after disturbances. Today, however, researchers can exploit an explosion of data sources. High-resolution satellite imagery, remote sensing indices (e.g., NDVI for vegetation health), and climate data can quantify ecological resilience and detect environmental stress or recovery in rural landscapes year by year. Likewise, socioeconomic resilience can be observed through time-series data on incomes, employment, and population movements. Big Earth Data—a concept referring to massive geospatial and socioeconomic datasets—are increasingly being applied to resilience measurement. With modern data acquisition (e.g., sensors, drones, IoT devices) and computing power, it is feasible to integrate datasets that capture human behavior, infrastructure usage, and environmental conditions simultaneously. For instance, researchers have begun using mobile phone mobility data and social media sentiment to gauge community well-being and response during disasters, alongside physical data on infrastructure outages. By combining these streams, we can obtain a more holistic, real-time picture of resilience in action [38].

Another promising avenue is the development of new analytical methods tailored to these rich datasets. Advanced spatial statistics (such as spatial autocorrelation and Geodetector models) and machine learning algorithms can help isolate the drivers of resilience from large data pools. For example, recent studies have employed entropy-weighted models and cluster analysis to identify key factors (like the urbanization rate, economic structure, fiscal resources, etc.) that explain why some regions are more resilient than others [7]. Future research could enhance this by using predictive modeling—for instance, training algorithms on past resilience outcomes to predict how a region might cope with future shocks under various scenarios. Network analysis is another tool of interest, as rural resilience is influenced by networks (transportation, social networks, supply chains), so mapping these and simulating disruptions can pinpoint vulnerabilities and cascading effects. Additionally, longitudinal case databases (compiling how different communities recovered from past floods, droughts, or market crashes) would enable meta-analyses to find common determinants of successful adaptation [39].

Emerging data sources will also enrich resilience research. Importantly, disseminating knowledge of past natural hazards to the public can significantly raise awareness in risk areas, complementing these new data approaches [40]. For example, crowd-sourced data and citizen science (reports of local hazards or coping strategies) can provide on-the-ground observations to validate quantitative indices [41]. Remote sensing and GIS allow for the creation of spatial resilience maps that show “hotspots” of vulnerability or growth, guiding targeted interventions. By leveraging these tools, future studies can move beyond static or coarse assessments and begin to capture the dynamic, interconnected nature of resilience. Ultimately, a data-driven approach—grounded in robust quantitative methods and diverse datasets—will improve not only the accuracy of resilience assessments but also our ability to test “what-if” questions. This supports policymakers in crafting timely, evidence-based strategies to bolster rural resilience under climate change and socioeconomic transformations.

5.3. Study Limitations

While our study provides valuable insights, we acknowledge several limitations that temper the interpretation of the results and point to directions for future research. First, the dataset used covers only the years 2000 through 2020. This 20-year window may be insufficient to capture longer-term trends and cyclical changes in rural resilience. Developments beyond 2020—including the prolonged impacts of the COVID-19 pandemic or new policy initiatives under China’s latest rural revitalization programs—fall outside of our analysis. As a result, our conclusions apply to past trends and require caution if extrapolated into the future.

Second, resilience, by nature, is an abstract and latent construct—there is no direct observable “resilience meter”. Therefore, our methodology cannot be conclusively proven “correct” in the same way that a physical measurement can be; it can only be assessed for robustness and plausibility. This challenge is common in resilience research, as different definitions and frameworks can yield varying results and many existing assessment tools lack empirical ground-truthing. In other words, there is no single accepted benchmark to validate a resilience index against real-world outcomes. We mitigated this by aligning our indicators with the established literature and by comparing our findings with known case realities (like Lishui and Taizhou’s qualitative resilience). Future studies could compare multiple methods or incorporate direct feedback from communities to triangulate the resilience measurements, thereby increasing confidence that the methodology captures meaningful aspects of rural resilience.

An unexpected trend observed was the negative correlation between cropland use and resilience. A possible explanation for this finding lies in the inherent vulnerability of agricultural practices in flood-prone regions. Extensive cropland areas may exacerbate community vulnerability by reducing adaptive flexibility and increasing economic dependency on agriculture, which is highly susceptible to water-related hazards. Policymakers should therefore consider diversified agricultural practices, including crop rotation, flood-resistant crops, and integrating agriculture with other resilient economic activities.

6. Conclusions

This study presents a detailed quantitative assessment of rural resilience dynamics in response to multiple water-related hazards within Zhejiang Province, China. By introducing a robust indicator-based framework that integrates stability, hazard exposure, vulnerability, and adaptive capacity, our analysis uncovered significant spatio-temporal variations and provided clear insights into the underlying determinants influencing rural resilience. Adaptive capability was identified as the most critical factor positively affecting rural resilience, emphasizing the essential roles of education, healthcare accessibility, employment stability, and economic diversification. Conversely, hazard intensity, exposure, and vulnerability—especially economic dependence on agriculture and high hazard susceptibility—emerged as significant impediments to rural resilience.

In addressing these findings, this study proposes targeted regulatory actions with clear practical implications for policymakers and regional authorities. First, enhancing rural adaptive capabilities should become a strategic priority achieved through increased investments in healthcare infrastructure, educational resources, and vocational training programs. Second, promoting economic diversification to reduce dependency on agriculture is essential, with policy support directed toward developing sustainable rural industries, such as agro-tourism, small-scale manufacturing, and rural e-commerce. Third, regions identified as highly vulnerable or hazard-exposed should receive dedicated interventions, including improved flood control systems, optimized land use planning, and enhanced early warning systems.

This comprehensive framework and its findings provide valuable methodological guidance and actionable recommendations for other rural coastal regions, emphasizing the importance of adaptive capacity and targeted resilience policies. Future studies could further refine the framework by incorporating real-time, high-resolution data, enhancing predictive capabilities, and enabling policymakers to proactively strengthen rural resilience in an increasingly uncertain climate context.