Spatiotemporal Evolution Characteristics and Influencing Factors of Urban Flood Resilience: The Case of Yangtze River Delta, East China

Abstract

Urban flood management is pivotal to the construction of resilient cities. However, investigation into the spatiotemporal evolution of urban flood resilience (UFR) and its influencing factors is insufficient. Aiming to address the challenge, this study establishes a multidimensional UFR indicator system grounded in the disaster resilience of place (DROP) model. Following the calculation of UFR through the entropy weighted technique for order preference by similarity to ideal solution (EW-TOPSIS) method, spatiotemporal evolution evaluation and factor detection are conducted. With panel data from the Yangtze River Delta Urban Agglomeration (YRDUA) over the period of 2012–2021, the results demonstrate overall UFR growth from a dominance of lowest-level and low-level cities to a more balanced distribution. Moreover, significant spatiotemporal heterogeneity is observed, with UFRs in cities adjacent to the Yangtze River higher than peripheral ones. Spatial clustering is significant until 2019, primarily manifested as High-High clusters along the Yangtze River and Low-Low clusters in northern Jiangsu and Anhui. Finally, factor detection identifies economic status, population size, environmental regulation, and drainage infrastructure as key influencing factors. These findings not only advance the understanding of UFR in urban agglomerations but also provide targeted recommendations for policymakers to enhance UFR.

1. Introduction

Urban floods are emerging as an imminent global challenge, exacerbated by climate change in an unprecedented way [1]. According to the Emergency Events Database, floods impacted 48.8 million people and caused economic losses of USD 32.8 billion in 2024 alone [2]. Especially in urban agglomerations, dense populations and overloaded infrastructure aggravate the damage of flood events [3]. For instance, urban flooding caused by extreme rainfall in July 2023 led to 107 fatalities and missing persons in the Beijing-Tianjin-Hebei urban agglomeration [4]. For deltaic areas worldwide, recurrent coastal storm surges and river floods threaten urban systems [5,6,7]. Against this backdrop, conventional disaster management has proven inadequate for such climate-related risks and therefore the concept of urban flood resilience (UFR) has been proposed. It is defined as a city’s capacity to resist flood impacts, sustain essential functions, and recover normal operations [6,8,9]. Although the Chinese government has taken a range of actions to enhance UFR, such as the sponge city pilot policy [10] and climate-adaptive city pilot policy [11], urban flood governance remains limited to a specific city. Given the complex socioeconomic interactions across cities, it is crucial to evaluate the UFR and explore its influencing factors from the perspective of urban agglomeration.

In recent decades, UFR has garnered substantial attention in academia. Several classic theoretical models have been developed to evaluate UFR. One of them is rooted in the stages of a flood event [8,12,13]. For example, the robustness-resistance-recovery (3Rs) model divides the UFR into pre-flood robustness, during-flood resistance, and post-flood recovery [7,14]. Likewise, the PSR model evaluates the UFR from pressure, state, and response dimensions, corresponding to disaster stages [15,16,17,18]. Another category is grounded in components of urban systems, including social resilience, economic resilience, and environmental resilience [9,19]. The most representative are the disaster resilience of place (DROP) model [20] and the baseline resilience indicators for communities (BRIC) model [21]. While both of them were originally developed for community resilience, they have been widely applied in the UFR evaluations [7,19]. Based on these aforementioned models, the spatiotemporal evolutionary characteristics of UFR have been preliminarily investigated. However, their research areas are primarily confined to geographical units and provinces [14,22]. There is a dearth of UFR evaluations at the urban agglomeration scale.

Identifying the influencing factors of UFR is fundamental to thoroughly understanding it and formulating improvement strategies. Currently, relevant analyses predominantly rely on traditional methods. The decision-making trial and evaluation laboratory (DEMATEL) method is utilized to identify the key influencing factors of UFR [23]. On this basis, the interpretive structural modeling (ISM) method displays the relationships between these factors as a hierarchical structure [24,25]. Due to the superiority of structure equation modeling (SEM) in large-sample causal pathway analysis, it is also employed to explore the influence mechanism of UFR [26]. Nevertheless, these methods cannot detect spatial differentiation of variables. Consequently, they possess a limited capacity to elucidate the spatiotemporal evolution of UFR.

In light of the above literature review, three key research questions need to be addressed: (1) How might a robust indicator system be developed to UFR? (2) How can the spatiotemporal evolution characteristics of UFR within urban agglomerations be comprehensively analyzed and visualized? (3) What factors significantly influence the spatiotemporal evolution of UFR in urban agglomerations?

To answer these questions, this study investigates the spatiotemporal evolution of UFR, with the Yangtze River Delta urban agglomeration (YRDUA) as a case study. First, a UFR indicator system based on the DROP model was established and the entropy weighted technique for order preference by similarity to ideal solution (EW-TOPSIS) method was employed to qualify it. Subsequently, this study evaluated the spatiotemporal evolution characteristics of UFR from 2012 to 2021 through spatial autocorrelation analysis. Finally, the key influencing factors were identified by the factor detector. This study extends the investigation of UFR and its influencing factors to urban agglomerations, highlighting the importance of collaborative flood governance. The spatiotemporal analysis could inform policymakers of notable disparities in UFRs across cities, thereby aiding in the formulation of targeted interventions. The practical recommendations proposed in this study merit consideration in future urban agglomeration development planning.

2. Materials and Methods

2.1. Study Area and Data Source

The YRDUA (114°54′–122°12′ E, 27°12′–35°20′ N) is situated in eastern China, spanning three provinces (Jiangsu, Zhejiang, and Anhui provinces) and a municipality (Shanghai), as illustrated in Figure 1. It is one of the most densely populated and economically vibrant regions in China. Covering merely 358,000 km² of land area, the YRDUA accommodates a permanent population of approximately 213 million. Moreover, as one of the world’s largest urban agglomerations, it generated a GDP of 28.57 trillion CNY in 2024, accounting for over 20% of China’s national economic output. These advantages are attributable to the efficient economic collaboration, integrated industrial system, unimpeded factor circulation, and siphoning resource aggregation in the YRDUA [27]. However, dramatic climate change has greatly heightened the risks of flood disasters in mid-latitude and subtropical monsoon regions [28,29], which has become a prevalent threat to urban agglomerations worldwide. Infrastructure overload and impervious surfaces resulting from urbanization, coupled with low-lying terrain and dense fluvial networks, further worsen the flood vulnerability of the YRDUA. On the whole, the YRDUA was selected as a typical case to explore the spatiotemporal differentiation of UFR in urban agglomerations.

In this study, the panel data of 41 cities in the YRDUA from 2012 to 2021 were derived from China Urban Statistical Yearbooks, China Urban-Rural Construction Statistical Yearbooks, China Urban Construction Statistical Yearbooks, China Statistical Yearbooks on Environment, and Yearbooks of China Insurance. Missing data were supplemented using the moving average interpolation method.

2.2. Developing the UFR Indicator System

Given that the DROP model is widely recognized in the field of urban resilience and can enhance the systematicity of indicators, this study established an indicator system for UFR (Table 1) based on the DROP model [20]. This indicator system consisted of five dimensions: ecology, economy, infrastructure, institution, and society. A total of 15 indicators were selected through extensive literature review, with detailed explanations as follows.

(1)

The ecology dimension is defined as the capacity to mitigate flood impacts through environmental governance, ecological quality, and pollution control, which encompasses three specific indicators [21,30]. The rate of domestic waste harmless disposal is a common metric to represent environmental governance and is contained in the indicator system [31]. Urban landscapes, functioning as green infrastructure, constitute nature-based flood solutions [32]. Moreover, excessive pollutants are deteriorating urban water ecology [10]. Thus, the harmless disposal rate of domestic waste, the area of parks and green spaces per capita, and the volume of sewage discharged per capita are included.

(2)

The economic dimension, in the context of urban flood management, refers to the capacity to reduce property losses and sustain economic stability through a variety of fiscal measures [20,21]. The income of employed workers largely determines the amount of property. Aside from that, social security and employment expenditures constitute a composite metric that reflects both government revenue and personal income. Insurance is also taken into account, as it transfers the financial burden on cities and losses borne by residents to capital markets [33]. Hence, average wages of employees, social security and employment expenditures per capita, and insurance premiums per capita are chosen to capture the economy dimension of UFR.

(3)

The infrastructure dimension is defined as the resilience of urban physical systems to withstand flood impacts and sustain their essential functions [20,21,30]. It includes urban public utilities, particularly water supply and wastewater treatment infrastructure. Public facilities are carriers of lifeline services and their vulnerability is amplified by extreme weather events. Particularly, municipal infrastructure withstands disaster stress and regulates water resources [34]. Thus, this indicator system incorporates the area of public facility land per 10,000 people. Per capita daily water use and per capita sewage treatment are selected to represent the supply and treatment capacity of the urban water system, respectively.

(4)

The institution dimension refers to the flood management capacity of government departments, manifested in effective planning and comprehensive policies [20,21]. According to the “Sustainable Cities and Communities—Indicators for Resilient Cities” published by the National Administration Standardization of the People’s Republic of China, both the disaster management plan and the continuity plan take public sanitation into account. Hence, the number of municipal sanitation vehicles per 10,000 people is chosen. Recently, the government has encouraged a range of techniques to be introduced into urban flood management, such as permeable pavement materials and digital twin-based flood response systems [35,36]. Therefore, the number of patents is also selected to represent the innovation outputs.

(5)

The social dimension is the capacity to enhance social cohesion and maintain community services, with a specific focus on education, employment and healthcare [30,37]. Cities with greater social resilience are able to provide emergency medical assistance and reduce casualties. Meanwhile, local employment and education should be minimally disrupted during the long-term recovery process. Additionally, public services and social welfare directly affect the life quality of the residents [38]. Therefore, this study selects the number of college students per 10,000 people, the number of medical personnel per 10,000 people, the proportion of employees in residential services and other services, as well as the proportion of employees in health social security and social welfare.

Table 1. Evaluation indicator system of UFR.

Dimension	Indicator	Unit	Attribute	Reference
Ecology	Harmless disposal rate of domestic waste (X1)	%	+	[23,30,39,40,41]
	Area of parks and green spaces per capita (X2)	m2	+
	Volume of sewage discharged per capita (X3)	m3	-
Economy	Average wages of employees (X4)	CNY	+	[8,23,40,41,42]
	Social security and employment expenditures per capita (X5)	CNY	+
	Insurance premiums per capita (X6)	CNY	+
Infrastructure	Area of public facility land per 10,000 people (X7)	km2	+	[9,30,41,43]
	Volume of daily water use per capita (X8)	Liter	+
	Volume of sewage treatment per capita (X9)	m3	+
Institution	Number of municipal sanitation vehicles per 10,000 people (X10)	Vehicle	+	[16,31,44]
	Number of green patents (X11)	Number	+
Society	Number of college students per 10,000 people (X12)	Person	+	[8,39,40,41,42,44]
	Number of medical personnel per 10,000 people (X13)	Person	+
	Proportion of employees in residential services and other services (X14)	%	+
	Proportion of employees in health social security and social welfare (X15)	%	+

Table 1. Evaluation indicator system of UFR.

Dimension	Indicator	Unit	Attribute	Reference
Ecology	Harmless disposal rate of domestic waste (X1)	%	+	[23,30,39,40,41]
	Area of parks and green spaces per capita (X2)	m2	+
	Volume of sewage discharged per capita (X3)	m3	-
Economy	Average wages of employees (X4)	CNY	+	[8,23,40,41,42]
	Social security and employment expenditures per capita (X5)	CNY	+
	Insurance premiums per capita (X6)	CNY	+
Infrastructure	Area of public facility land per 10,000 people (X7)	km2	+	[9,30,41,43]
	Volume of daily water use per capita (X8)	Liter	+
	Volume of sewage treatment per capita (X9)	m3	+
Institution	Number of municipal sanitation vehicles per 10,000 people (X10)	Vehicle	+	[16,31,44]
	Number of green patents (X11)	Number	+
Society	Number of college students per 10,000 people (X12)	Person	+	[8,39,40,41,42,44]
	Number of medical personnel per 10,000 people (X13)	Person	+
	Proportion of employees in residential services and other services (X14)	%	+
	Proportion of employees in health social security and social welfare (X15)	%	+

2.3. Quantifying the UFR Through the EW-TOPSIS Method

On the basis of the above indicator system, the EW-TOPSIS method was employed to calculate the UFR of each city in a given year. This method combined the entropy weight (EW) method and the technique for order preference by similarity to ideal solution (TOPSIS) method. As an objective weighting approach, the EW method determines weights from the degree of data dispersion, thus avoiding estimation biases caused by subjective scoring [45]. The TOPSIS method is a multi-criteria decision analysis method that can reduce the potential biases and enable accurate ranking by calculating the Euclidean distances to ideal solutions [46]. This hybrid approach has been widely applied in the domain of urban safety and resilience [47,48]. The specific calculation steps are as follows:

(1)

Data normalization

The equation for positive indicators is:

$$x{\prime }_{ij}=(x_{ij}-\mathrm{m}\mathrm{i}\mathrm{n}(x_{ij}\left)\right)/\left(\mathrm{m}\mathrm{a}\mathrm{x}\right(x_{ij})-\mathrm{m}\mathrm{i}\mathrm{n}(x_{ij}\left)\right)$$

(1)

The equation for negative indicators is:

$$x{\prime }_{ij}=\left(\mathrm{m}\mathrm{a}\mathrm{x}\right(x_{ij})-x_{ij})/\left(\mathrm{m}\mathrm{a}\mathrm{x}\right(x_{ij})-\mathrm{m}\mathrm{i}\mathrm{n}(x_{ij}\left)\right)$$

(2)

where $x_{ij}$ is the original value of the i-th indicator in the j-th city, and $x{\prime }_{ij}$ is the normalized value of the i-th indicator in the j-th city ($0\le i\le m,0\le j\le n$). m is the number of indicators. n is the number of samples.

(2)

Calculation of entropy values of indicators

$$k=1/\mathrm{l}\mathrm{n}\left(n\right)$$

(3)

$$p_{ij}=x{\prime }_{ij}/\sum _{j=1}^{n}x{\prime }_{ij}$$

(4)

$$e_i=-k\sum _{j=1}^n{p}_{ij}\mathrm{l}\mathrm{n}\left(p_{ij}\right)$$

(5)

where $e_i$ is the entropy value of the i-th indicator.

(3)

Calculation of indicator weights

$$w_i=(1-e_i)/\sum _{i=1}^m(1-e_i)$$

(6)

where $w_i$ is the weight of the i-th indicator.

(4)

Calculation of standardized indicators

$$r_{ij}=w_{i}x{\prime }_{ij}$$

(7)

where $r_{ij}$ is the standardized indicator.

(5)

Determination of positive and negative ideal solutions

The equation for positive ideal solutions is:

$$S_i^{+}=\mathrm{m}\mathrm{a}\mathrm{x}(r_{i1},r_{i2},\dots r_{in})$$

(8)

The equation for negative ideal solutions is:

$$S_i^{-}=\mathrm{m}\mathrm{i}\mathrm{n}(r_{i1},r_{i2},\dots r_{in})$$

(9)

where $S_i^{+}$ and $S_i^{-}$ represent the positive ideal solution and the negative one, respectively.

(6)

Calculation of the distance to ideal solution

$$D_j^{+}=\sqrt{\sum _{i=1}^m{(S_i^{+}-r_{ij})}^2}$$

(10)

$$D_j^{-}=\sqrt{\sum _{i=1}^m{(S_i^{-}-r_{ij})}^2}$$

(11)

where $D_j^{+}$ and $D_j^{-}$ represent the distances to positive ideal solution and the negative one, respectively.

(7)

Calculation of the relative proximity to ideal solution

$$C_j={\displaystyle \frac{D_j^{-}}{D_j^{+}+D_j^{-}}}$$

(12)

where $C_j$ is the relative proximity to ideal solution. In this study, $C_j$ represents the level of UFR.

2.4. Analyzing the Spatiotemporal Evolution of the UFR

Following the evaluation of UFR, a spatiotemporal autocorrelation analysis was conducted to investigate its spatial distribution characteristics. The Moran’s I index is a statistical metric for the degree of spatial clustering [49,50]. In this study, the global Moran’s I was utilized to assess the overall clustering degree of UFRs across the study area, with values ranging from −1 to 1. When the global Moran’s I approaches 1, the spatial correlation is regarded as positive and clustered [51]. If its value is close to 0 or −1, this indicates a random or discrete distribution [52]. The local Moran’s I was adopted to identify the spatial distribution patterns of UFR among a specific city and its surrounding cities, including high-high cluster, high-low outlier, low-high outlier, and low-low cluster [45]. The global Moran’s I and local Moran’s I are computed by Equation (13) and Equation (14), respectively.

$$I={\displaystyle \frac{n}{{\sum }_{i=1}^n(y_j-\overline{y}{)}^2}}\cdot {\displaystyle \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{ij}(y_i-\overline{y})(y_j-\overline{y})}{{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{ij}}}$$

(13)

$$I_i={\displaystyle \frac{n(y_i-\overline{y}){\sum }_{j=1}^n{w}_{ij}(y_j-\overline{y})}{{\sum }_{i=1}^n(y_j-\overline{y})}}$$

(14)

where $y_i$ and $y_j$ are the UFR of cities i and j, respectively. $w_{ij}$ is the spatial weight based on inverse geographical distance between cities i and j. n is the total number of samples.

2.5. Identifying the Influencing Factors of the UFR

This study employed a geographical detector to identify the influencing factors behind the above spatiotemporal evolution of UFR. First, economic status, population size, environmental protection, and drainage infrastructure were selected as potential influencing factors, as shown in

Table 2. Potential influencing factors of UFR.

Factor	Description	Unit	Reference
Economic status (X1)	Retail sales of social consumer goods per capita	CNY	[23,26,54]
Population size (X2)	Ratio of population to urban area	%	[54,55]
Environmental regulation (X3)	Ratio of environmental protection expenditures to fiscal expenditures	%	[54,56]
Drainage infrastructure (X4)	Length of drainage pipelines per capita	Km	[23,26,56]

. Second, the K-means cluster method was adopted to classify the sample data of these factors, as expressed in Equations (15) and (16). This method can enhance the similarity of data within the same cluster by reducing the distance between sample data and the corresponding centroid [53]. Finally, the factor detector (Equation (17)) was utilized to examine their explanatory degree on the spatial distribution of UFR.

$$J=\sum _{j=1}^k\sum _{x_i\in C_j}{\left(x_i-c_j\right)}^2$$

(15)

$$c_j={\displaystyle \frac{1}{n}}\sum _{x_i\in C_j}{x}_i$$

(16)

where J is the sum of the distances between the sample points and their respective centroids in all clusters. $x_i$ is the i-th sample data. $C_j$ is the j-th cluster. $c_j$ is the centroid of the j-th cluster. n denotes the number of sample data. k denotes the number of clusters.

$$q=1-{\displaystyle \frac{{\sum }_{j=1}^k{N}_j{\sigma }_j^2}{N{\sigma }^2}}$$

(17)

where q represents the explanatory degree of a specific factor. N and ${\sigma }^2$ represent the total number of geographical units and the total variance throughout the study area, respectively. $N_j$ and ${\sigma }_j^2$ represent the number of geographical units and the variance in the j-th cluster, respectively.

3. Results

3.1. Evaluation Results of the UFR in the YRDUA

The UFR values were first computed using the aforementioned EW-TOPSIS method. Detailed calculation results of UFR values can be found in Supplementary Table S1. The data were winsorized to eliminate the effect of extreme values. Subsequently, to visualize the spatiotemporal evolution characteristics, this study selected the years 2012, 2015, 2018, and 2021 as the investigation periods. Then, these values were categorized into five levels: lowest, low, moderate, high, and highest, using the natural breakpoint method.

Table 3. Proportion of cities with different UFR levels.

Year	Lowest	Low	Moderate	High	Highest
2012	53.7%	22.0%	14.6%	4.9%	4.9%
2015	39.0%	29.2%	17.1%	9.8%	4.9%
2018	14.6%	29.2%	34.1%	9.8%	12.2%
2021	2.4%	17.1%	36.6%	24.4%	19.5%

and Figure 2 display the proportion of cities with different UFR levels and their changes. It is observed that cities with the lowest UFR continuously decrease, while cities at high level and above significantly increase. In particular, the growth in the number of cities classified as high and highest UFR levels is quite significant between 2018 and 2021. Owing to the emphasis placed by the Chinese government on flood control, other regions such as the Beijing-Tianjin-Hebei urban agglomeration and the Yangtze River economic belt have also achieved notable UFR improvements in recent years [40,57].

3.2. Spatiotemporal Evolution Characteristics of the UFR in the YRDUA

3.2.1. Spatiotemporal Distribution Results of the UFR in the YRDUA

By associating the UFR values with their respective cities on the map, its spatial distribution is shown in Figure 3. The distribution has shifted from being dominated by low-level UFR cities in 2012 to a relatively even pattern in 2021. Notably, cities in Jiangsu Province and Zhejiang Province exhibit overall higher UFR levels compared to those in Anhui Province. Furthermore, the UFR of cities along the Yangtze River is significantly higher than that of peripheral cities in the YRDUA. Presumably, this discrepancy is due to the concentration of cities with more developed economies and higher administrative hierarchies in these regions [8,23], such as Shanghai, Nanjing, and Hangzhou. Inequitable flood management between core cities and their peripheral counterparts is prevalent across many urban agglomerations globally [57]. Based on the core-periphery theory [58], the resource aggregation from peripheral cities to these core cities has also accentuated the spatial heterogeneity of UFR.

3.2.2. Spatial Autocorrelation Analysis Results of the UFR in the YRDUA

On the basis of Equations (13) and (14), the global Moran’s I and local Moran’s I for UFR were obtained using ArcGIS 10.8 software.

Table 4. Global Moran’s I of UFR.

Year	Moran’s I	z-Score	p-Value
2012	0.19	2.07	0.04
2013	0.22	2.24	0.03
2014	0.25	2.53	0.01
2015	0.17	1.81	0.07
2016	0.22	2.20	0.03
2017	0.24	2.44	0.01
2018	0.25	2.47	0.01
2019	0.18	1.86	0.06
2020	0.04	0.84	0.40
2021	0.03	0.80	0.43

reports the annual global Moran’s I indices from 2012 to 2021. It is evident that the global Moran’s I remains positive and the p-values are less than 0.1 until 2019, suggesting significant spatial clustering in the YRDUA at a 90% confidence level. However, the results in 2020 and 2021 are no longer statistically significant. As the Chinese government prioritizes the management of urban flood disasters [34], the improvement in UFR across cities has narrowed the disparities among them. Consequently, the balanced UFR distribution diminishes spatial clustering effects over the study area, resulting in the reduced and insignificant global Moran’s I.

Figure 4 presents the local indicators of spatial autocorrelation (LISA) clustering maps of UFR during the investigation years. The total number of cities in the four clusters exhibits a trend of initial increase followed by a decrease, with the peak occurring in 2018. This trend is basically consistent with the significance of the global Moran’s I. In terms of clustering types, High-High clusters and Low-Low clusters are the predominant patterns. The former are primarily located along the Yangtze River, characterized by their economic strengths and complete infrastructure. It is worth noting that Zhenjiang was found in the High-High clusters in both 2015 and 2018, which may be associated with the implementation of the sponge city pilot policy [35,59]. On the contrary, the latter are concentrated in the northern parts of Jiangsu and Anhui provinces, where the economic foundation is relatively weak. Moreover, in other cases, low-UFR regions often exhibit spatial clustering due to similar hydrological characteristics [60].

3.3. Results of Factor Detection of UFR

Table 5. Results of factor detection.

Factor	X1	X2	X3	X4
q-value	0.27	0.04	0.05	0.20
p-value	0.00	0.00	0.02	0.00

presents the results of the influencing factor analysis using a geographical detector. Their explanatory power ranks from highest to lowest as X1, X4, X3, and X2, with all significant at the 5% confidence level. Therefore, all of them can be identified as key influencing factors. Notably, X1 and X4 also have a substantial impact on UFR, which confirms the findings of Ref. [7] and Ref. [40] that identify economic development and drainage pipelines as the primary drivers of UFR heterogeneity. On the one hand, pre-disaster prevention, during-disaster relief, and post-disaster reconstruction require considerable financial support. On the other hand, properly designed drainage facilities are vital for controlling surface runoff and mitigating urban waterlogging [61]. Likewise, the result of X3 underscores the necessity of environmental protection for urban flood preparedness. Environmental regulation, including policies and measures aimed at reducing pollution and enhancing ecological quality, has been shown to be highly effective in improving urban flood resilience. However, despite its proven effectiveness, the widespread adoption of environment-based approaches remains limited. This is partly due to the complexity of implementing such policies and the varying levels of commitment from different regions [59]. As a result, the strong explanatory power of X3 in our analysis underscores the critical role that environmental regulation can play in enhancing UFR, even though its full potential has not yet been realized across all regions. The result of the demographic factor X2 indicates that population density elevates the risk of urban exposure to floods. However, the explanatory degree of X2 is relatively lower, which can be attributed to several factors. The limited time span of our research data (2012–2021) may not fully capture the long-term impacts of population density on flood resilience. Moreover, the spatial distribution of population density within the study area may not be uniform, leading to variations in its influence on UFR [62]. Furthermore, the effectiveness of flood management strategies in densely populated areas may mitigate some of the negative impacts of high population density [63], thus reducing its explanatory power in our analysis.

4. Discussion

4.1. Comparison and Analysis of UFR Results

The empirical results validate established findings and advance the understanding of UFR. First, the trend of overall UFR growth evidenced by the decrease in lowest-level cities and the increase in high-level cities is consistent with systemic national policies that prioritize flood control and the comparable patterns documented in other urban agglomerations [40,57]. Second, the findings on the spatiotemporal evolution of UFR confirm the pervasive spatial heterogeneity driven by the core-periphery theory [58]. The UFR is significantly higher in core cities due to concentrated economic resources and administrative hierarchies, a phenomenon widely reported in global urban agglomerations where flood management is often fragmented and unequal [57]. Finally, the factor detection results are also validated by established knowledge, identifying X1 and X4 as the most substantial drivers. These factors are universally recognized as primary determinants of UFR, dictating the capacity for effective prevention, relief, and recovery [7,40,61]. Moreover, the X2 and X3 underscore the critical role in enhancing ecological flood preparedness and population exposure [62,63].

4.2. Advantages of the Proposed Method

The method proposed in this study embodies the following improvements compared to existing research. Its advantages lie in addressing critical gaps through a comprehensive methodology that evaluates UFR and analyzes its spatiotemporal evolution.

First, a UFR indicator system comprising five dimensions was established by adapting the DPOR model and overviewing relevant literature. On the one hand, this multidimensional indicator system provides a more comprehensive perspective, integrating socioeconomic interactions, urban ecosystems, physical infrastructure, and institutional measures. On the other hand, unlike ad hoc models applied to specific cases [43], it can serve as a scalable and general framework.

Second, this study not only expands the spatial scope from individual cities or provinces to urban agglomerations but also provides a temporally dynamic outcome [22]. The findings from spatial autocorrelation analysis offer a more nuanced understanding of the clustering effects on the UFR within urban agglomerations. The interpretable results could inform the formulation of balanced flood prevention strategies and the promotion of equitable development among cities.

Third, the geographic detector enables the identification of factors influencing the spatiotemporal evolution of UFR, while quantifying their explanatory degree. Unlike traditional methods such as DEMATEL and SEM [23,26], which overlook spatial variations in influencing factors, the geographic factor detector excels in spatial causation analysis and explicitly investigates the underlying mechanisms of spatial clustering in this study.

4.3. Strategies for Enhancing UFR

Although the UFR has experienced overall growth in the past decade, a number of challenges have been exposed, such as inter-city disparities in UFR and the distribution characteristics of various clusters. Combining the identified influencing factors, this study proposes the following strategies.

First, given the significant role of environmental regulations and drainage facilities, developing integrated green and grey infrastructure is an effective measure to enhance UFR. This not only necessitates drainage networks capable of rapidly reducing runoff, but also requires ecological facilities such as green roofs, permeable pavements, and urban wetlands to preserve the natural water cycle [64,65].

Second, it is advisable to advance coordinated urban agglomeration governance and regional planning to jointly enhance UFR. Considering the UFR disparities within the YRDUA, policymakers ought to establish inter-city collaboration mechanisms. This includes early warning systems for urban agglomerations, infrastructure construction plans that transcend administrative boundaries, and equitable resource sharing, particularly the resource transfer from more resilient core cities to less resilient peripheral areas [58,66].

Finally, whereas the UFR levels of Shanghai, Nanjing, and Hangzhou have surpassed those of their surrounding cities, they have not exhibited spatial clustering. Conversely, in spite of the relatively underdeveloped economy, Zhenjiang remains in the High-High clusters. Therefore, governments could formulate context-specific policies such as sponge city initiatives to enhance UFR.

4.4. Limitations and Future Research Directions

Although this study provides a robust framework for evaluating the spatiotemporal characteristics of UFR in urban agglomerations, the following limitations and future research directions ought to be considered. When applying this indicator system to other cases, contextual adjustments for climate variations, indicator granularity, and data availability are necessary. Moreover, the proposed spatiotemporal evolution evaluation method relies on city-level data, smoothing out internal variations within cities, although it is suitable for the urban agglomeration scale. Last but not least, other geographic detection techniques such as interaction detection and risk detection can be employed to complement and validate the findings of factor detectors [7].

5. Conclusions

This study conducted a comprehensive analysis on spatiotemporal evolution characteristics of UFR. Specifically, a multidimensional indicator system for UFR was developed. This study carried out an empirical analysis regarding the UFR across 41 cities in the YRDUA over the period 2012–2022. The primary findings are highlighted as follows: (i) The UFR evaluation results indicate that a substantial portion of cities have experienced the transition from lower to higher levels. (ii) The UFR in the core areas of YRDUA was consistently higher than that of peripheral cities. Spatial clustering of UFR was significant until 2019 but subsequently decreased, dominated by High-High and Low-Low clusters. (iii) Economic status, population size, environmental regulation, and drainage infrastructure were identified as key influencing factors.

This study enriches the theoretical knowledge in the realm of urban flood management and provides practical guidance for policymakers. Theoretically, the UFR evaluation and influencing factor detection in this study can serve as a scalable paradigm for related research and enlighten urban agglomeration governance. By expanding the research scope, this study emphasizes synergistic UFR enhancement of urban agglomerations rather than a single-city level. Practically, the proposed recommendations could equip policymakers with effective measures to strategically allocate resources and coordinate actions.

Several limitations warrant further exploration in future studies. Although significant spatiotemporal variations were observed during the study period, longer-term analyses could provide deeper insights into climate change and policy interventions. Moreover, the scope of identified influencing factors remains limited. Hence, future research could focus on longer time spans and a broader range of influencing factors. Further, despite these practical insights, the motivations and potential obstacles of local governments in practice may affect the effectiveness of UFR improvement, which also deserves in-depth exploration.