Community resilience to urban flooding across comparative neighborhoods in China

Abstract

As global climate change escalates and urban populations expand, natural disasters, including urban flooding, pose significant risks to rapidly urbanizing cities. Against this background, enhancing community resilience is widely recognized as one of the crucial approaches to mitigating the impact of disasters. Based on a survey in Chengdu, the Communities Advancing Resilience Toolkit (CART) assessment was employed to measure the perceived community resilience across three community types: urban, exurban, and suburban. 387 valid questionnaires were collected in this study, and the data were analyzed by using multiple linear regression analysis to explore the main influencing factors. Results indicate that: (1) Urban center residents reported the highest resilience perception, while those of the suburban community exhibited the lowest level. (2) Built environment characteristics contributed to intra-community variations in resilience perception. (3) Socio-demographic factors (e.g., age, income, education) and participation in disaster risk reduction (DRR) activities (particularly community volunteering and disaster drills) were significantly correlated with higher resilience scores. These findings highlight the significance of incorporating social capital and community involvement into urban resilience planning. Therefore, a “multi-sectoral approach” integrated framework is proposed for enhancing community resilience to cope with flooding and other disaster risks.

Introduction

Flooding is a pervasive and increasingly severe natural hazard, driven by the combined effects of climate change and rapid population growth¹. Crucially, extensive urbanization, especially in low- and middle-income countries, is compounding this risk. For example, from 2000 to 2020, global urban flood exposure rose dramatically that calls for urgent action. East Asia, notably China, and other rapidly urbanizing regions are experiencing the most severe increases^2,3. By 2022, urbanization in China reached 65% of the population, concentrated in a few mega-regions, but approximately two-thirds of China’s land area is flood-prone⁴. One of these high-risk areas is the Chengdu–Chongqing metropolitan area, which occupies only 5.5% of China’s land yet contains 25% of its people and 34% of GDP. Chengdu is a megacity with over 22 million people that has undergone rapid industrial and residential expansion in recent decades, growing into a logistics, cultural, and transportation hub of southwest China. This expansion has manifested as dense urban core areas spilling into suburban and exurban zones, with land development extending into formerly rural areas.

Given this context, it is critical to understand community-level resilience in Chengdu. Urban flood events first hit individual neighborhoods and communities, so local adaptive capacity and preparedness greatly affect outcomes. However, the majority of Chinese research conducted thus far has concentrated on flood resilience assessment at the city scale even larger scale^5–7. We argue that Chengdu, with its mix of urban, suburban, and exurban neighborhoods, is an ideal representative setting to study community flood resilience. Moreover, residents’ social responses – preparedness, risk awareness, trust, and local networks – have not been thoroughly examined in the context of community planning in Chengdu.

Literature review

Research from various academic fields has enhanced our understanding of disaster, vulnerability, and risk management in urban areas, contributing to the advancement of urban resilience in recent years^8,9. Originally coined in ecology by Holling¹⁰ to denote a system’s ability to absorb disturbance, resilience in disaster and urban planning has come to encompass multiple domains. Resilience concepts further differentiate urban versus community resilience. Over the last decade, scholars have increasingly recognized “community resilience” as distinct, emphasizing local social assets and adaptive capacities. Norris et al.¹¹ described community resilience as an emergent property of networks of individuals and institutions, and Pfefferbaum et al.¹² emphasized that how community-level engagement and resources underpin recovery. A systematic review by Patel et al.¹³ identified nine core elements common in definitions of disaster resilience at the community scale: local knowledge, community networks and relationships, communication, health and mental wellbeing, governance and leadership, resources (economic and material), and preparedness. Thus, urban flood resilience can be defined as the capacity of an urban community or system to withstand flood impacts, minimize damage, and recover or adapt quickly while maintaining essential functions. This concept encompasses not only robust physical infrastructure but also strong social networks, effective communication, and responsive governance that help communities cope with and learn from flood events^14,15.

In recent years, a variety of frameworks and tools have been developed to assess community resilience, ranging from city-scale index systems to community-focused surveys. City-level indices such as the Baseline Resilience Indicators for Communities (BRIC) gauge infrastructure, socioeconomic, and institutional capacities¹⁶, but may overlook local dynamics and community perceptions. Alternatively, survey-based measures such as the Conjoint Community Resilience Assessment Measure (CCRAM) capture resilience across psychosocial factors (e.g., leadership, preparedness, social trust)¹⁷, providing insight into community cohesion while focusing on a narrower set of attributes. Other approaches include field checklists like the Analysis of the Resilience of Communities to Disasters (ARC-D), which provides quick appraisals of community vulnerabilities¹⁸, and participatory frameworks like the Community-Based Resilience Analysis (CoBRA) that engage communities in defining resilience indicators¹⁹. Comprehensive toolkits have also emerged (e.g., COPEWELL) to guide communities and public health agencies in self-assessing disaster readiness²⁰. However, many of these tools tend to catalogue resilience components descriptively, with less emphasis on critical comparison or on actively engaging community perspectives.

When developing assessment tools, some researchers understand community resilience as a community resident’s ability to cope with risks. Therefore, the assessment methods they propose focus more on evaluating individual perceptions through bottom-up and community-participatory methods, which could bring many benefits to the community^21,22. In this study, we selected the Communities Advancing Resilience Toolkit (CART) as the basis for our assessment, a survey-based method using a bottom-up questionnaire to measure residents’ perceptions of their community’s resilience²³. There are several reasons to favor CART over other tools in the Chinese community setting. CART is a survey-based method centered on community members’ subjective evaluations of local resilience, aligning with our goal of capturing residents’ sense of preparedness and capacity. This bottom-up focus sets CART apart from top-down indices like BRIC and expert-led checklists like ARC-D, which lack direct resident input. Moreover, CART’s design is both holistic and concise: its questionnaire covers key resilience domains, from basic resources to information access, in a manageable number of items. By contrast, CCRAM and similar surveys are longer or limited in scope, and composite indices demand extensive secondary data, making them less practical and potentially less reflective of on-the-ground realities. CART is also practical to deploy across multiple neighborhoods. In contrast, tools like CoBRA require facilitated workshops, which are difficult to scale across many communities, whereas CART’s simple survey format enables broad participation and consistent comparisons.

In summary, the literature indicates a growing awareness that community-level, social factors are central to flood resilience, yet empirical work is needed to apply these concepts locally in China. Existing Chinese studies tend to focus on technical and infrastructural indices of urban flood resilience, leaving the social processes and intra-community variations underexplored^24–26. In contrast, qualitative and survey-based data at the neighborhood level, such as residents’ perceptions of flood risk, social networks, and organizational capacity, are seldom incorporated into urban resilience assessments. Although community resilience assessment tools have gained traction internationally, their deployment in China’s fast-growing cities remains largely unexplored.

Our study addresses this gap by conducting a neighborhood-scale comparative resilience assessment in Chengdu. The study makes two key contributions: (1) It applies a survey-based resilience framework to Chinese flood-prone communities, demonstrating the tool’s applicability outside its original context. (2) It offers the empirical comparison of community resilience to flooding in urban, suburban, and exurban neighborhoods within a Chinese megacity. These findings will fill a critical knowledge gap and guide planners and policymakers in China and other flood-prone regions to tailor community resilience strategies.

Methodology

Study areas

Chengdu (30°05′–31°26′ N, 102°54′–104°53′ E) is a megacity that serves as the capital of Sichuan province in China. However, as one of China’s most populous and economically advanced regional cities, Chengdu is regularly hit by a range of disasters, the most common of which are floods, geological hazards, and earthquakes. We selected three neighborhoods representing distinct community types—urban core, suburban, and exurban—to examine community resilience in this context (Fig. 1). These types are defined by each neighborhood’s location within the metropolis, administrative status, and development stage. Specifically, the urban community (UC) is an old, dense downtown neighborhood with the highest population density and a fully built-up environment; the suburban community (SC) is a newer middle-class residential area in the peri-urban zone (recently urbanized with moderate density on the city’s edge); and the exurban community (EC) is a rural-fringe neighborhood (an outlying satellite community with lower density and mixed urban–rural land use). This typology captures clear differences in population density and built environment, factors that likely influence community flood resilience (in Fig. 2, 3 and 4).

Fig. 1.

The location and study area of three communities in Chengdu City.

Fig. 2.

The built-environment map of A.Lianguixi (UC).

Fig. 3.

The built-environment map of B.Jangjunbei (SC).

Fig. 4.

The built-environment map of C.Xinlian (EC).

Importantly, all three communities were chosen for their comparable exposure to flood risk. Each site is situated near a major waterway and has a history of flooding, which helps ensure that baseline hazard levels are similar across the cases. This selection allows us to attribute differences in residents’ perceptions of resilience more to social and institutional factors than to unequal risk exposure. Additionally, all three neighborhoods are National Disaster Reduction Demonstration Communities (NDRDCs), meaning they meet baseline standards for disaster preparedness while each employs a different resilience-building approach. Studying an urban, a suburban, and an exurban community under similar flood-risk conditions thus provides a strong basis for comparing how community context and engagement strategies influence perceived resilience.

Assessment tools

The measurement tool used in this study to assess perceived community resilience was adapted from the Chinese version of the CART, which was recently validated with data from multiple studies, confirming that the adapted instrument reliably measures community resilience in China^27,28. Originally, it encompasses four interrelated domains—Connection and Caring, Resources, Transformative Potential, and Disaster Management—operationalized through 21 core items. We modified the instrument’s content and structure with expert guidance to better reflect local conditions. Specifically, the original “Resources” domain was refined into Basic Resources & Services to emphasize critical community infrastructure and services relevant to the Chinese community in coping with food shortages. We introduced a new Information & Communication domain to capture the crucial role of information access and dissemination during frequent short-duration disasters, particularly floods. In the Disaster Management domain, an item on community disaster education and drills was added to gauge residents’ preparedness, recognizing that community-led training programs. The Community Capital & Connecting domain (analogous to Connection and Caring) gauges social cohesion and mutual support, factors particularly pertinent as rapid urbanization alters traditional social ties. Finally, Community Governance & Transformative Potential assesses collaborative leadership, civic engagement, and learning from past events, reflecting communities’ capacity to self-organize and adapt even within a top-down governance structure. Together, the five domains (with 24 survey items in total; see Table 1) provide a comprehensive framework that covers the physical, social, institutional, and informational facets of resilience.

Table 1.

Core community resilience items by domains of community resilience.

Domains and Items of Communities Advancing Resilience Toolkit (CART)
Domains 1: Basic Resources & Services
1. Food supply	My community has accessible and affordable food.
2. Health services	My community has accessible holistic health.
3. Housing	My house in the community is good and safe.
4. Emergency shelters	My community has sufficient, well-located, safe shelters.
5. Facilities	People in my community can access the services they need.
Domains 2: Community Capital & Connecting
6. Civic engagement	People in my community concern and engage in local public issues.
7. Sense of place	People in my community feel like they belong to the community.
8. Well-being	People in my community are committed to the well-being of the community.
9. Community inclusion	My community treats people fairly, regardless of their background.
10. Community connection	People in my community help each other.
Domain 3: Disaster Management
11. Prevention	My community tries to prevent disasters.
12. Preparedness	My community actively prepares for future disasters.
13. Response capacity	My community can provide emergency services during a disaster.
14. Recovery capacity	My community has services and programs to help people after a disaster.
15. Training & drilling	My community provides many opportunities for disaster education and emergency drills.
Domain 4: Information & Communication
16. Information transparency	Disaster maps and emergency information are accessible and posted quickly/broadly in my community.
17. Communication channel	My community keeps people informed through diverse tools, such as television, radio, the Internet, phones, and neighbors.
18. Information credibility	People in my community trust public officials.
19. Information usefulness	I get information/communication from my community to help with my home.
Domain 5: Community Governance & Transformative Potential
20. Multi-stakeholders’ cooperation	My community works with local residents, markets, social organizations, etc., to get things done.
21. Community co-governance	People in my community work together to develop solutions so the community can improve.
22. Leadership	People in my community communicate with leaders who can help improve the community.
23. Community empowerment	My community develops skills and finds resources to solve its problems and reach its goals.
24. Learning capability	My community looks at its successes and failures to learn from the past.

Two authors of this work are proficient in English and Chinese, so we translated the original CART items into Chinese and back-translated them to ensure consistency, and then had a panel of disaster management and community resilience experts review the items for accuracy and cultural relevance. We used Cronbach’s coefficient alpha to assess the reliability of the questionnaire in measuring the items. The overall Cronbach’s alpha for the scale was 0.88, indicating high consistency. Specifically, the Cronbach’s alpha test values for the five specific domains were 0.91 (basic resources & services), 0.86 (community capital & connecting), 0.86 (disaster management), 0.88 (information & communication), and 0.89 (community governance & transformative potential). All Cronbach’s alpha values for the five domains exceeded 0.85, indicating that our adapted tool is an effective and credible measure of community resilience in this context.

Participants and sampling

A stratified sampling method was employed to select participants. Each of the three case communities was subdivided into several sampling zones based on a specific built environment. This stratification ensured that distinct neighborhood conditions were represented in the sample. In total, 400 questionnaires (20 per sub-area) were distributed across the three communities (Fig. 5), of which 387 were completed and returned (97% response rate, see Table 2). Participation was voluntary with no incentives, and surveys were conducted from July to September 2024.

Fig. 5.

The Sampling area division of three communities.

Table 2.

The number of participants and sampling.

Community	A. Lianguixi (UC)	B. Jiangjunbei (EC)	C. Xinlian (SC)	Total
Sample	120 (20*6)	140 (20*7)	140 (20*7)	400
Validity	117 (98%)	134 (96%)	136 (97%)	387 (97%)

Variables and data analysis

In this study, the final scores and the five domains of CART were used as predicated variables. Furthermore, variables related to disaster risk reduction (DRR) and resilience-building activities were treated as explanatory variables, along with related socio-demographic variables, including gender, age, education, marital status, and years of residence in a community. Quantitative analysis was conducted in the first step to assess the impact of the explanatory variables on the outcome variables. CART was first used to analyze the overall community resilience score descriptively. Secondly, a multiple linear regression model was used to examine the factors influencing community resilience. The analyses primarily utilize the SPSS Statistics Software. The expression is as follows:

1

In the formula (1), represents the perceived community resilience and its domains; , ,…, represents the dependent variable, and represents the regression coefficient corresponding to , and represents the intercept of the model.

Results

Socio-Demographic characteristics of the participants

Among them, more than half are aged 30–60, and the proportion of married respondents is higher than that of unmarried. The overall educational level of the respondents is relatively high, with about 76% holding a high school degree or higher. According to the China Statistics Bureau income level division in 2020, as for a month’s disposable income, less than 2,000 yuan is the low-income group, and more than 5,000 is the high-income group. Thus, the income level of the respondents is mostly middle-income group, with about 58% reporting an income of 2000–5000. To verify the comparability of the three community samples, we conducted statistical tests on key socio-demographic variables. One-way ANOVA and chi-square tests indicated no significant differences among the groups in age distribution, education, or income levels (p > 0.05), confirming that the sampled populations were equivalent across communities (see Table 3).

Table 3.

Descriptive analysis of socio-demographic characteristics.

Variables	Total (n = 387)		A. UC (n = 117)		B. SC (n = 134)		C. EC (n = 136)
	n	%	n	%	n	%	n	%
Age (years)
<30	96	24.8	23	19.7	39	29.1	34	25.0
30–60	229	59.2	66	56.4	79	59.0	84	61.8
>61	62	16.0	28	23.9	16	11.9	18	13.2
Gender
Male	206	53.2	61	52.1	74	55.2	71	52.2
Female	181	46.8	56	47.9	60	44.8	65	47.8
Marital status
Married	240	62.0	79	67.5	81	60.4	80	58.9
Unmarried or others	147	38.0	38	32.5	53	39.6	56	41.1
Education
Junior school or below	91	23.5	29	24.8	24	17.9	38	27.9
High school	164	42.4	47	40.2	54	40.3	63	46.4
College or above	132	34.1	41	35.0	56	41.8	35	25.7
Income (RMB/Yuan)
<2000	68	17.6	16	13.7	25	28.7	27	19.9
2000–5000	225	58.1	71	60.7	68	50.7	86	63.2
>5000	94	24.4	30	25.6	41	30.6	23	16.9
Years of residence in a community
<5	139	35.9	33	28.2	62	46.3	44	32.4
5–10	153	39.6	48	41.0	54	40.3	51	37.5
>10	95	24.6	36	30.8	18	13.4	41	30.1

Participation in disaster risk reduction activities

The participants’ involvement in DRR and resilience-building activities in the community is reported in Table 4. Among the 387 participants, more than half of the respondents (59.4%) had emergency supplies prepared at home in case of disasters. About merely 19.6% had volunteered for community-based DRR activities. Among the three communities, the UC (A) located in the urban center has a relatively high percentage of 29.9%. The proportion of respondents who had received disaster education and participated in evacuation drills is less than half, at 48.8% and 30.0%, respectively. EC (C) has a relatively high percentage in terms of disaster education and evacuation drills. The proportion of respondents who had received disaster warning information from the community is high at 63.0%, and SC (B) in the suburbs has the highest percentage at 73.1%.

Table 4.

Descriptive analysis of participation in community disaster risk reduction activities.

Variables	Total (n = 387)		A. UC (n = 117)		B. SC (n = 134)		C. EC (n = 136)
	n	%	n	%	n	%	n	%
Having emergency supplies
Yes	230	59.4	68	58.1	77	57.5	85	62.5
No	157	40.6	49	41.9	57	42.5	51	37.5
Being a volunteer
Yes	76	19.6	35	29.9	13	9.7	28	20.6
No	311	80.4	82	70.1	121	90.3	108	79.4
Attended disaster education
Yes	189	48.8	55	47.0	52	38.8	82	60.3
No	198	51.2	62	53.0	82	61.2	54	39.7
Participated in evacuation drills
Yes	116	30.0	39	33.3	21	15.7	56	41.2
No	271	70.0	78	66.7	113	84.3	80	58.2
Received disaster warning information
Yes	244	63.0	71	60.7	98	73.1	75	55.1
No	143	37.0	46	39.9	36	26.9	61	44.9

Overall score of community resilience

As shown in Table 5, community resilience perceptions are reflected in the means and standard deviations (SD) for each of the 24 core community resilience items, as well as for the five CART domains and the overall community resilience score. the overall community resilience score is 3.56. Among the five domains, Information & Communication scores the highest at 3.67, while Community governance & Transformative potential rank among the bottom at 3.45. UC performed well overall, scoring highest particularly in Basic resources & Services. SC scored lowest, however, notably in Community capital & Connecting. Disaster management stood out in EC (Fig. 6).

Table 5.

Descriptive analysis of core community resilience items by domains of CART.

Domains and Items	Overall	A. UC	B. SC	C. EC
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
Domains 1: Basic resources & Services	3.53 (0.61)	3.96 (0.50)	3.23 (0.61)	3.43 (0.44)
1. Food supply	4.01 (0.92)	4.64 (0.51)	3.52 (0.87)	3.91 (0.94)
2. Health services	3.41 (1.14)	4.25 (0.62)	3.06 (1.09)	2.94 (1.15)
3. Housing	3.29 (0.99)	3.02 (0.82)	3.51 (1.09)	3.33 (0.97)
4. Emergency shelters	3.46 (1.05)	3.77 (0.71)	3.09 (1.15)	3.52 (1.14)
5. Facilities	3.50 (1.09)	4.13 (0.57)	2.96 (1.20)	3.44 (1.05)
Domains 2: Community capital & Connecting	3.48 (0.55)	3.81 (0.43)	3.09 (0.47)	3.56 (0.47)
6. Civic engagement	2.76 (0.93)	3.22 (0.61)	2.21 (0.97)	2.87 (0.88)
7. Sense of place	3.39 (1.37)	4.11 (0.85)	2.33 (1.16)	3.76 (1.39)
8. Well-being	3.60 (0.88)	3.76 (0.76)	3.74 (0.77)	3.30 (1.02)
9. Community inclusion	3.54 (0.65)	3.65 (0.56)	3.49 (0.69)	3.47 (0.68)
10. Community connection	4.12 (1.00)	4.31 (0.73)	3.69 (1.14)	4.38 (0.95)
Domain 3: Disaster Management	3.64 (0.46)	3.57 (0.49)	3.57 (0.45)	3.79 (0.40)
11. Prevention	4.58 (0.72)	4.73 (0.47)	4.41 (0.92)	4.60 (0.66)
12. Preparedness	4.11 (0.71)	4.04 (0.64)	4.00 (0.82)	4.30 (0.60)
13. Response capacity	3.43 (0.66)	3.48 (0.62)	3.42 (0.64)	3.40 (0.71)
14. Recovery capacity	3.24 (0.74)	3.12 (0.70)	3.36 (0.81)	3.23 (0.68)
15. Training & drilling	2.85 (1.23)	2.46 (1.19)	2.68 (1.09)	3.41 (1.23)
Domain 4: Information & Communication	3.67 (0.40)	3.65 (0.40)	3.75 (0.40)	3.59 (0.39)
16. Information transparency	3.19 (0.82)	3.17 (0.79)	3.51 (0.68)	2.88 (0.87)
17. Communication channel	3.65 (1.03)	3.68 (1.08)	3.87 (0.91)	3.39 (1.05)
18. Information credibility	4.35 (0.70)	4.36 (0.70)	4.08 (0.72)	4.62 (0.59)
19. Information usefulness	3.47 (0.78)	3.39 (0.78)	3.56 (0.71)	3.46 (0.84)
Domain 5: Community governance & Transformative potential	3.45 (0.39)	3.42 (0.40)	3.32 (0.39)	3.62 (0.33)
20. Multi-stakeholders’ cooperation	3.04 (0.87)	2.81 (0.57)	2.78 (0.68)	3.52 (1.06)
21. Community co-governance	3.38 (0.93)	3.45 (0.92)	3.27 (0.98)	3.42 (0.86)
22. Leadership	4.55 (0.72)	4.61 (0.59)	4.55 (0.70)	4.50 (0.80)
23. Community empowerment	3.73 (0.76)	3.65 (0.65)	3.58 (0.71)	3.98 (0.58)
24. Learning capability	2.57 (0.86)	2.59 (0.89)	2.44 (0.88)	2.68 (0.82)
Overall Community Resilience	3.56 (0.29)	3.68 (0.30)	3.39 (0.27)	3.60 (0.23)

Fig. 6.

Community resilience by domains of CART for three community types.

Resilience score within the community

To explore intra-community variation in perceived resilience, Neighborhood sections within each community were classified using the Natural Breaks (Jenks) method into high, intermediate, and low resilience categories, with thresholds defined in Table 6. Figure 7 visualizes this spatial pattern. UC (urban core) exhibited the highest mean resilience and the greatest internal variation (CV: 10.19%), while EC (exurban) showed the most uniform distribution (CV: 6.42%). This variation correlates with distinct patterns in the built environment. Resilience hotspots in UC align with areas of mixed land use and newer high-rise housing with better infrastructure. In contrast, low-resilience zones overlap with older walk-up buildings lacking open space or service access. These spatial disparities likely reflect ongoing gentrification and redevelopment, which have produced coexistence between high- and low-capacity zones. In EC and SC, resilience is more evenly distributed, with fewer extremes, likely due to more homogeneous housing types and land uses.

Table 6.

The classification of resilience score in the three communities.

Classification	A. UC	B. SC	C. EC
High (number)	3.98–4.25 (2)	3.97 (1)	3.95–4.11 (2)
Intermediate (number)	3.49–3.61 (3)	3.39–3.64 (3)	3.48–3.72 (3)
Low (number)	3.20 (1)	3.01–3.12 (3)	3.13–3.19 (2)
Mean (SD)	3.68 (0.375)	3.39 (0.270)	3.60 (0.231)
CV	10.19%	7.96%	6.42%

Fig. 7.

Spatial distribution of resilience at the community scale.

Analysis of the influencing factors

The regression results of the correlation between the socio-demographic variables, DRR activities, and community resilience scores are reported in Table 7. The regression met all key assumptions: residuals were approximately normally distributed (Shapiro–Wilk p > 0.05) and homoscedastic, and variance inflation factors (VIFs) for all predictors were low (all VIFs < 2), indicating no multicollinearity of concern. The model was statistically significant overall (F-test p < 0.001), explaining about 22.6% of the variance in resilience (adjusted R² = 0.226). Key findings include: (1) Community resilience declines from urban to suburban areas due to reduced access to resources and weaker social cohesion; (2) Elderly and female residents are more vulnerable, with lower access to services and weaker flood preparedness; (3) Higher education and income levels significantly enhance access to information and governance participation; (4) Participation in DRR activities, especially volunteering and drills, is strongly associated with greater perceived resilience.

Table 7.

Linear regression results for community resilience.

Variables	Overall	Basic resources & Services	Community capital & Connecting	Disaster Management	Information & Communication	Community governance & Transformative potential
Area	−0.041* (0.022)	−0.273*** (0.000)	−0.108** (0.001)	0.107*** (0.000)	−0.035 (0.194)	0.143* (0.116)
Age	−0.111*** (0.000)	−0.231*** (0.000)	0.039 (0.485)	−0.210*** (0.000)	−0.171*** (0.000)	0.017 (0.697)
Gender	−0.153** (0.001)	−0.189** (0.002)	−0.000 (0.999)	−0.378** (0.001)	−0.101* (0.021)	−0.099* (0.021)
Marital status	−0.118* (0.010)	−0.199** (0.008)	−0.164* (0.014)	−0.117* (0.035)	−0.094 (0.075)	−0.018 (0.727)
Education	0.097*** (0.000)	0.195** (0.008)	−0.003 (0.944)	0.098** (0.004)	0.098** (0.003)	0.095** (0.003)
Income	0.070** (0.002)	0.134** (0.005)	−0.003 (0.936)	0.070 (0.051)	0.154** (0.001)	−0.006 (0.866)
Years of residence in community	0.107** (0.005)	0.025 (0.581)	0.318*** (0.000)	−0.014 (0.671)	0.025 (0.441)	0.180** (0.001)
Emergency supplies	0.147 (0.091)	0.233* (0.104)	0.088 (0.104)	0.254** (0.001)	0.139 (0.104)	0.117 (0.100)
Being a volunteer	0.163*** (0.000)	0.234 (0.126)	0.178*** (0.000)	0.147** (0.001)	0.261* (0.126)	0.313* (0.122)
Disaster education	0.178** (0.009)	0.285* (0.110)	0.142* (0.032)	0.217** (0.007)	0.307** (0.110)	0.242* (0.107)
Evacuation drills	0.236*** (0.000)	0.020 (0.147)	0.305* (0.119)	0.248*** (0.000)	0.149* (0.028)	0.420* (0.115)
Disaster information	0.304** (0.002)	0.095 (0.137)	0.020 (0.147)	0.142* (0.032)	0.246*** (0.000)	0.147 (0.142)
adj. R2	0.226	0.277	0.288	0.315	0.173	0.167

* p < 0.05, ** p < 0.01, *** p < 0.001.

Discussion and conclusion

Comparison of resilience-building approaches among three communities

Based on a review of community documents and interviews with community leaders, the main resilience-building approaches and models in the three communities are summarized in Table 8; Fig. 8, which are also the three common approaches for DRR in China.

Table 8.

The summary of resilience-building approaches among three communities.

Community	Main DRR and resilience-building policy	Approach
A. UC	(1) Recruiting and training community volunteers to focus on urban disasters, such as flooding; (2) Educating the residents about disasters through various community events.	Community-based (collaborative)
B. SC	(1) Building disaster warning information systems; (2) Constructing a sponge city and other engineering measures to cope with heavy rainfall and flooding.	Government-led (engineering)
C. EC	(1) Encouraging disaster social work services provided by the social sector and organizations; (2) Organizing residents to participate in disaster education and drills by NGOs, especially for flooding.	Social organization-driven (collaborative)

Fig. 8.

The model of resilience-building approaches among three communities.

The three neighborhoods in Chengdu demonstrated distinct resilience-building strategies that correspond closely with community resilience theory. UC and EC both leveraged community-based collaboration (through volunteers and NGOs), whereas SC relied primarily on top-down engineering projects (e.g., sponge-city infrastructure). Notably, residents in UC and EC reported higher perceived flood resilience than those in SC, suggesting that multi-actor engagement enhanced adaptive capacity. In contrast, SC’s government-centric approach tended to build physical capacity but limited the development of social networks. Aldrich & Meyer (2015)²⁰ likewise emphasizes that resilience emerges from interdependent social structures (trust, networks) alongside infrastructure. Thus, theory helps explain why Community-based and social-organization-driven strategies (UC and EC) yielded higher resilience perceptions: they tapped into local knowledge and social supports, whereas a siloed government approach (SC) did not mobilize these adaptive capacities. It is worth noting that the current approaches to building community resilience still place the government in a primary position, while the sectors involved are secondary and singular. The effect of relying solely on government and engineering solutions to build resilience is limited; furthermore, the multi-sector approach, including the private, social, and market sectors, should be integrated and collaborate in community co-governance.

A particularly salient finding is the inequality of resilience within communities. The urban center (UC) exhibited the highest average resilience score but also the widest internal variation, while the exurban area (EC) showed a moderate mean resilience with a relatively narrow spread. This pattern reflects the classic dynamics of social stratification that resilience frameworks predict. Norris et al. (2008)¹⁰’s “rule of relative advantage” holds that better-resourced communities and subgroups naturally accrue more resilience benefits. In UC, affluent or well-connected residents likely drive up the mean resilience through access to resources and networks, whereas disadvantaged residents do not benefit equally, creating a large disparity. By contrast, EC’s more homogeneous socio-economic profile means everyone shares similar capacities, yielding more uniform but lower resilience. The theoretical implication is clear: high average resilience is insufficient if it coexists with vulnerability among subpopulations. Our findings thus highlight the need to unpack not just how much resilience a city has on average, but who benefits, as critical for equitable planning.

The policy implications for resilience building at the community level

After carefully comparing different community resilience-building models, this study proposes that it is urgent to establish a “multi-sectoral approach” integrated framework, namely “community-based, government-organized, and society/market-involved”, for flooding resilience management and planning to respond to natural disasters, which is based on the five domains (see Fig. 9 below for details).

Fig. 9.

Integrated framework for improving resilience at the community level.

Firstly, enhancing the accessibility and regional equity of resources, with a priority for vulnerable groups and areas, is the basic guarantee for improving community resilience. Secondly, embed resident engagement into daily governance by leveraging grid-based management, encouraging participation in neighborhood councils or urban “community planner” pilot programs to foster shared responsibility and local knowledge. Thirdly, institutionalize community involvement in flood risk assessment and contingency planning by organizing seasonal preparedness workshops, jointly led by street offices, property management companies, and NGOs, especially before the monsoon. Fourthly, improve early warning and disaster information systems by integrating WeChat mini-programs, community bulletin boards, and loudspeakers to ensure all residents, including the elderly and digitally excluded, receive timely alerts. Lastly, establish cross-sector coordination platforms (e.g., “resilience alliances”) involving neighborhood committees, social workers, local enterprises, and residents to co-develop emergency response protocols and co-manage resources such as volunteer teams and equipment stockpiles. Ultimately, by embedding these mechanisms into China’s existing grassroots governance, flood resilience building becomes a tangible, equitable, and community-driven agenda.

To conclude, this study contributes to the growing body of research on urban flood resilience by applying a localized and survey-based framework to assess community resilience across urban, suburban, and exurban neighborhoods in Chengdu, China. Theoretically, the findings reinforce the central role of social capital, demographic factors, and community participation in shaping resilience, thereby extending resilience research beyond infrastructure-centered approaches. Socially, our work suggests that Chinese cities should operationalize multi-sector flood resilience by explicitly integrating community voices and targeting social vulnerability. Ultimately, this study extends resilience research beyond hardware solutions to emphasize that social processes and equity are equally critical. By marrying survey-based evidence with community resilience theory, we provide both a deeper academic understanding and clear policy pathways to strengthen flood resilience in Chengdu and similar global South megacities.

Author contributions

Conceptualization, Y.W.; Validation, F.S.; Investigation, Y.W. and Y.Y; Data curation, Y.W.; Writing, Y.W.; Supervision, T.K. and F.S.; Funding acquisition Y.W and Y.Y. All authors reviewed the manuscript.

Funding

Project of MOE (Ministry of Education) Foundation on Humanities and Social Sciences (No. 24YJCZH332). Project of Key Research Center of Philosophy and Social Sciences of Sichuan Province (No. MD24E017). Scientific Research Fund of Zhejiang Provincial Education Department (No. Y202351341). Construction Scientific Research Fund of Zhejiang Provincial Housing and Urban-Rural Development Department (No. 2024K013).

Data availability

The datasets generated and analyzed during the current study are not publicly available due to the consideration of high accuracy of administrative village survey data involves privacy and trust issues for interviewees, but are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.