Urban Pluvial Flood Resilience Evolution and Dynamic Assessment Based on the DPSIR Model: A Case Study of Kunming City, Southwest China

Abstract

The increasing frequency of extreme weather events and rapid urbanization has exacerbated pluvial flood risks, underscoring the urgent need to strengthen the assessment of pluvial flood resilience in China’s southwestern mountainous regions. Kunming—a plateau basin city—was selected as a case study, and an urban pluvial flood resilience assessment system was developed based on the DPSIR model. The analytic hierarchy process (AHP), entropy method, and game theory-informed combination weighting were applied to determine indicator weights, while the extension cloud model was utilized to quantitatively assess resilience evolution from 2013 to 2022. The results reveal that: (1) Kunming’s pluvial flood resilience experienced a clear three-stage evolution—initial construction (Level II), resilience enhancement (Level III), and resilience reinforcement (Level IV)—reflecting a transition from rudimentary resilience to advanced adaptive capacity; (2) the ranking of primary indicator weights is as follows: Driving Forces > Pressure > State > Response > Impact, with Flood Disaster Risk (P6), Flood Disaster Early Warning Capability (R1), and Topographic and Geomorphological Characteristics (P7) identified as key influencing factors; (3) marked disparities exist across the five dimensions: the Driving Forces dimension demonstrates increasing economic support; the Pressure dimension reflects structural vulnerabilities and climate variability; the State and Impact dimensions advance incrementally through policy implementation; and the Response dimension has substantially improved due to smart city technologies, although persistent gaps in inter-agency emergency coordination remain. This research offers a scientific basis for enhancing pluvial flood resilience in southwestern mountainous cities.

1. Introduction

In recent years, the increasing frequency of extreme weather events has exerted profound global impacts, most notably through intense precipitation and urban pluvial flooding. Coupled with rapid urbanization, pluvial flooding has become an increasingly critical impediment to sustainable urban development [1]. The relentless expansion of impervious surfaces has disrupted natural hydrological cycles, thereby increasing the frequency and intensity of flood events and exposing the inherent capacity limitations of conventional drainage systems [2]. In response, China has prioritized urban water security and resilience governance, issuing a series of pivotal policy documents—such as the Notice of the General Office of the State Council on Strengthening the Construction of Urban Drainage and Waterlogging Prevention Facilities, the 13th Five-Year Plan for Ecological and Environmental Protection, and the 2018 Notice on Strengthening Urban Drainage and Waterlogging Prevention to Ensure Safe Passage Through the Flood Season. Collectively, these policies underscore a paradigmatic shift in urban stormwater management: from the traditional “rapid drainage” approach to an integrated framework characterized by “natural retention, systematic regulation, and collaborative governance.”

Despite continuous policy advancement and practical exploration, the current urban pluvial flood governance system remains inherently vulnerable to sudden pluvial flood risks and systemic threats [3]. On the one hand, most cities continue to over-rely on “grey” infrastructure, while the regulatory capacity of natural ecosystems remains underutilized [4]. On the other hand, because of the far-reaching cascading impacts and intricate systemic interlinkages of pluvial flood events, conventional static, single-dimensional risk analysis methods are inadequate for characterizing temporal evolution processes and systemic feedback mechanisms [5]. Thus, it is imperative to establish a comprehensive, dynamic, and adaptive urban pluvial flood resilience assessment framework—capable of systematically identifying key risk factors across temporal stages and spatial units.

Urban pluvial flood resilience, the core capacity of urban systems in addressing pluvial flood risks, is defined as a system’s ability to resist, absorb, and recover from external disturbances. In recent years, the Driving Forces–Pressure–State–Impact–Response (DPSIR) framework has been extensively applied in environmental management [6] and risk assessment [7], owing to its systematic architecture, clear hierarchical structure, and chain-like causal representation. The DPSIR model enables a comprehensive understanding of the complex causal pathways and regulatory mechanisms underpinning urban pluvial flood systems across five dimensions: Driving Forces, Pressure, State, Impact, and Response. It thereby provides a solid theoretical foundation for multi-dimensional coupled analysis and comprehensive resilience assessment.

To date, scholars worldwide have made significant advances in assessing urban pluvial flood resilience. For example, Qiu et al. analyzed 13 cities in Jiangsu Province, selecting 15 indicators within three dimensions: social vulnerability, resource endowment, and disaster-prevention capacity. By applying the entropy weight method to derive indicator weights, a composite-index evaluation model was constructed, and the spatiotemporal evolution and drivers of resilience in the region were analyzed [8]. Zhu et al. combined the Pressure–State–Response (PSR) model with the Social–Economic–Natural Complex Ecosystem (SENCE) theory to establish a PSR–SENCE framework, which enabled static–dynamic assessments and multi-scenario simulations across 27 cities in the Yangtze River Delta [9]. Schmitt et al. systematically assessed pluvial flood risk in German cities and examined challenges in risk evaluation and communication [10]. Tayyab et al. integrated GIS, remote sensing, and the Analytic Hierarchy Process (AHP) to develop a resilience model, with Peshawar, Pakistan, used as a case study to evaluate flood sensitivity and adaptive capacity [11]. Internationally, Cutter et al. proposed a disaster-resilience indicator system grounded in disaster-risk theory and developed a regional resilience evaluation framework spanning six dimensions, including ecological, economic, and infrastructure aspects [12]. Sharifi reviewed urban resource-resilience assessments and proposed an evaluation framework consisting of six thematic dimensions, including economy, population, and environment [13]. Yuan et al. developed a 17-indicator framework for Yingtan City covering the socioeconomic, ecological, and infrastructure subsystems; with FAHP–EWM applied for weighting, TOPSIS for evaluation, and GRA for driver identification, resilience evolution from 2010 to 2022 was traced, and the main causes of ecological decline were identified [14]. Notwithstanding these advancements, existing research continues to exhibit notable limitations. Geographically, existing case studies have largely concentrated on cities in China’s eastern and central plains, while neglecting geographically distinctive contexts such as southwestern mountainous cities. Methodologically, although models such as PSR and DSR are widely applied, the DPSIR framework—despite its proven efficacy in vulnerability assessments—has remained underutilized in systematically evaluating urban pluvial flood resilience.

Accordingly, Kunming—a representative plateau-basin city—was selected as the case study, and an urban pluvial flood resilience assessment indicator system was constructed based on the DPSIR framework. A hybrid weighting approach, combining the Analytic Hierarchy Process (AHP) and the entropy weight method, was employed for indicator weighting and further refined using game theory to enhance the scientific rigor and stability of weight determination. Subsequently, the extension cloud model was applied for quantitative assessment of Kunming’s urban pluvial flood resilience during 2013–2022, thereby revealing temporal evolution trends, identifying key influencing factors, and supporting the formulation of targeted enhancement strategies. The findings aim to provide theoretical underpinnings and actionable insights for pluvial flood risk governance and resilience planning in mountainous cities, thereby contributing to the advancement of China’s urban disaster prevention and mitigation capacities toward greater efficiency, coordination, and resilience.

2. Construction of the Urban Pluvial Flood Resilience Index System

2.1. Overview of the Study Area

Kunming, located in central Yunnan Province, China, lies between longitudes 102°10′–103°40′ E and latitudes 24°23′–26°33′ N, covering approximately 21,012.54 km² and accounting for 5.3% of the province’s total land area (Figure 1). As the capital of Yunnan Province and a pivotal city in Southwest China, Kunming is located within the Dianchi Basin, characterized by undulating terrain and significant hydrological convergence. The city has a typical monsoonal climate, with a mean annual precipitation of approximately 1003.5 mm, over 85% of which occurs between May and October—coinciding with the peak season for urban pluvial floods. Historically, Kunming has experienced recurrent pluvial flood events, with more than 100 documented incidents dating back to 277 AD, including 39 major events between 1950 and 2020. In recent years, rapid urban expansion and intensifying extreme weather have resulted in several episodes of heavy rainfall—such as in 2013 and 2018—triggering severe pluvial flooding and secondary disasters, and exposing vulnerabilities in the city’s drainage system and flood resilience. These challenges underscore the urgent need for a systematic assessment and enhancement of urban pluvial flood resilience [15]. Against this backdrop, the entire administrative region of Kunming—including central urban districts and subordinate counties—was selected as the study area, with the aim of comprehensively evaluating the overall level, spatial disparities, and driving factors of urban pluvial flood resilience. The goal is to provide scientific insights and decision support for disaster preparedness and urban planning. This research contributes to both theoretical advancement and practical implementation, strengthening urban disaster risk reduction systems and promoting integrated regional resilience.

2.2. Concept of Urban Pluvial Flood Resilience

Urban pluvial flood resilience is an essential component of overall urban resilience and refers to a city’s capacity to resist, adapt to, and recover from water-related disasters such as floods and pluvial inundation. With intensifying global climate change and accelerating urbanization, extreme rainfall events have become more frequent, posing major challenges for many cities. A clear definition of urban pluvial flood resilience is crucial for effective assessment, yet academic perspectives on the concept remain divergent. Gou et al. defined urban pluvial flood resilience as a city’s ability to resist, adapt to, and recover from pluvial flood disasters [16]. Mayer highlighted the importance of “adaptive capacity”, particularly in reducing systemic impacts of pluvial flooding, shortening recovery times, and enhancing future resilience [17]. Zhou and colleagues defined urban pluvial flood resilience as a city’s ability to prepare for foreseeable disasters, adapt to changing conditions, absorb disturbances, and recover rapidly from disruptions [18]. Mabrouk et al. argued that urban pluvial flood resilience entails the development of sustainable, adaptive, and environmentally friendly cities capable of withstanding and recovering from flood events [19]. Overall, scholars have generally agreed that urban pluvial flood resilience encompasses not only resistance but also adaptation and recovery. Accordingly, this study defines urban pluvial flood resilience as the comprehensive capacity of urban systems, whereby the coordinated functioning of infrastructure, ecosystems, and socio-economic subsystems enables cities to withstand pluvial flood impacts, adapt dynamically to environmental changes, sustain essential functional continuity, and restore normal operations rapidly following disruptions.

2.3. Construction of the Indicator System

The DPSIR model—“Driving Force–Pressure–State–Impact–Response”—represents an extension and refinement of the PSR (Pressure–State–Response) and DSR (Driving Force–State–Response) models. This framework is designed to provide a comprehensive understanding and analysis of pluvial flood issues and their impacts on socio-economic systems and ecosystems across five dimensions: Driving Force, Pressure, State, Impact, and Response [20] (Figure 2). As the “Spring City” and a regional international tourist destination, Kunming is noted for its favorable climate and strong ecological endowments. However, the city is also exposed to significant pluvial flood risks, resulting from natural hydrological convergence in its mountainous valley terrain and the increasing burden on its drainage system due to rapid urbanization.

The Driving Force dimension is intended to reveal the fundamental causes of urban pluvial flood risks and reflects the long-term pressures imposed on the environment by socio-economic development, thereby serving as a fundamental basis for constructing a resilience framework. Regarding the Regional Economic Condition, GDP serves as a key indicator for measuring infrastructure investment as well as disaster prevention and mitigation capacity, with higher GDP levels generally associated with stronger post-disaster recovery and response capabilities [21]. The Household Economic Condition is assessed using the proportion of disposable income remaining after expenditures, with a higher proportion indicating stronger economic resilience and greater ability to withstand flood-related losses [22]. Data on per capita disposable income are sourced from the nationwide household income and expenditure survey conducted by the National Bureau of Statistics. Economic Diversity is reflected by the share of the tertiary industry, as an optimized industrial structure contributes to enhanced public service provision, strengthened information dissemination, and improved recovery capacity. Greater industrial diversification reduces the likelihood of economic activities being severely affected by shocks to a single sector, thereby fostering stronger resilience and recovery capacity during disasters [23]. The Employment Situation is indicated by the unemployment rate, with higher rates signifying greater social vulnerability and heightened impacts from pluvial flood disasters [24]. The Urbanization Rate serves as an indicator of urban development, with higher urbanization generally associated with more advanced pluvial flood management systems, thereby enhancing a city’s disaster response capacity [25]. Collectively, these indicators, spanning macroeconomic to social structural dimensions, systematically reveal the socio-economic driving mechanisms underlying pluvial flood disasters and provide essential references for strengthening urban pluvial flood resilience.

The Pressure dimension is employed to assess the stress state and degree of exposure of cities in the context of natural disasters, focusing on the direct and indirect pressures imposed on pluvial flood systems by population structure, rainfall intensity, and topographic conditions. The Aging Population Ratio and Child Population Ratio indicate the spatial distribution of two vulnerable groups with relatively weak coping and recovery capacities, thereby serving as key indicators for assessing population vulnerability [26]. Population Exposure Density significantly increases exposed areas and the number of affected people, serving as an important driver of disaster diffusion, and is particularly relevant in flood-prone or high-risk regions [27]. In terms of natural rainfall, the Long-Term Precipitation Pattern and Short-Duration Precipitation Intensity indicate long-term climate change trends and extreme event impacts, which often place excessive burdens on drainage systems [28]. Flood Disaster Risk is measured by the proportion of extreme rainfall in total precipitation, with higher values indicating greater risks of sudden flooding and weaker system resilience [29]. Topographic and Geomorphological Characteristics are assessed through the integrated analysis of elevation, slope, and aspect, which directly affect water convergence and inundation locations. Steep-slope areas tend to intensify surface runoff, whereas plains often experience poor drainage, making topography one of the core variables for assessing natural stress and water-storage capacity [30]. Collectively, these indicators formulate an external load profile of urban pluvial flood systems from the perspectives of population, climate, and terrain, thereby providing essential support for identifying vulnerability hotspots and devising resilience enhancement strategies.

The State dimension is intended to capture the current conditions of urban pluvial flood-bearing capacity within specific spatial patterns and infrastructure configurations. It primarily evaluates drainage capacity, building density, and ecological regulation, thereby serving as a fundamental component of the structural basis for urban pluvial flood resilience. The Drainage Network Condition, regarded as a key infrastructure indicator, determines the efficiency of urban drainage under heavy rainfall, with higher density generally associated with a more effective reduction of short-term inundation risks [31]. The Urban Road Condition reflects both traffic efficiency and emergency accessibility during flooding events [32]. The Building Exposure Density, measured by the proportion of built-up area to the total urban area, serves as an indicator of the extent of building and population exposure in flood-prone zones [33]. The Green Coverage Ratio reflects the ecological buffering capacity of cities, as green spaces enhance rainwater infiltration and storage, mitigate surface runoff, and strengthen ecosystem resilience [34]. The Water Resource Regulation and Storage Capacity is assessed through river network density, with higher density associated with stronger drainage and diversion capacity, thereby facilitating rapid redistribution of floodwaters under extreme pluvial stress and serving as a core indicator of natural buffering capacity and system elasticity [35]. Collectively, these indicators, spanning “infrastructure–construction intensity–ecological buffering–hydrological regulation,” systematically characterize the current carrying capacity of cities, thereby providing a structural foundation for evaluating flood response mechanisms and formulating resilience enhancement strategies.

The Impact dimension is designed to assess the social, economic, and ecological disruptions caused by urban pluvial flood events, thereby serving as an essential dimension in the evaluation of system exposure and vulnerability. The Affected Population and Affected Towns and Subdistricts are used to reflect the population scale and spatial distribution of impacts, which indicate both the capacity to absorb shocks and the extent of protective infrastructure coverage [36]. Direct Economic Loss is used to quantify the financial burden and recovery challenges, thereby reflecting the economic cost of disasters and the city’s overall capacity to withstand floods [37]. The Proportion of Excellent Surface Water Quality reflects the capacity of aquatic environments to maintain and restore functions after flooding, thereby serving as an indicator of ecological resilience to disturbances [38]. The Wastewater Treatment Rate is used to indicate the effectiveness of water environment management; higher rates can mitigate post-flood water quality deterioration and enhance the resilience of the urban water cycle [39]. Collectively, these indicators are used to capture flood impacts across four dimensions—human exposure, spatial vulnerability, economic loss, and ecological disturbance—thereby providing a multi-dimensional basis for resilience diagnosis and the formulation of targeted improvement strategies.

The Response dimension is designed to evaluate a city’s capacity for early warning, emergency management, and post-disaster recovery, thereby reflecting its adaptability, coordination, and recovery mechanisms. The Flood Disaster Early Warning Capability is employed to assess the effectiveness of meteorological information acquisition, risk assessment, and timely alert dissemination, which collectively underpin proactive disaster prevention and exposure reduction [40]. The Emergency Management Capability reflects the government’s capacity for post-disaster coordination, resource mobilization, and recovery implementation [41]. It encompasses command systems, emergency plans, public participation, and policy execution, which directly influence response efficiency and recovery progress. The Public Response Capability is proxied by the proportion of residents with at least primary education, which reflects collective judgment, self-rescue, and recovery capacity; higher proportions are associated with stronger resilience [42]. The Communication Capability is assessed by mobile and broadband user penetration, thereby reflecting the infrastructure supporting information dissemination and emergency coordination during disasters [43]. The Medical Rescue Capability is measured by the hospital beds-to-population ratio, which indicates a city’s ability to provide medical services during floods; greater capacity is indicative of stronger protection for residents. The Level of Social Security is assessed by the coverage rate of basic medical insurance, which reflects the city’s ability to alleviate post-flood medical burdens, ensure service accessibility, and support recovery; the number of insured individuals comprises both employee and resident participants [44]. Collectively, these institutional, social, and technical indicators are employed to provide a comprehensive assessment of urban disaster response throughout the flood cycle, thereby revealing the key pathways through which cities withstand external shocks and restore functionality. As such, they constitute a fundamental dimension in evaluating urban pluvial flood resilience.

Based on the systematic analysis of the five dimensions of the DPSIR framework, this study constructs an urban pluvial flood resilience evaluation system encompassing indicators of Driving Forces, Pressures, State, Impact, and Response, as illustrated in Figure 3.

3. Data Sources and Methodology

3.1. Data Sources

This study designates Kunming as the case study area and evaluates its urban pluvial flood resilience from 2013 to 2022 using an indicator system grounded in the DPSIR framework. Indicators are categorized into three types: Statistical Category, Computational Category, and Qualitative Evaluation Category. Data for each indicator are sourced primarily from the following channels, differentiated by indicator type:

Urban Statistical Yearbooks and Specialized Bulletins: Statistical Category indicators (D1, D3, D4, D5, P1, P2, P3, S1, S2, S4, I1, I2, I3, I5) and Computational Category indicators (D2, S3, I4, R3, R4, R5, R6) are sourced from the Kunming Statistical Yearbook [45], China Urban Statistical Yearbook [46], China Flood and Drought Disaster Prevention Bulletin [47], and Kunming Water Resources Bulletin [48].

Meteorological Data Platforms: Precipitation-derived indicators (P4, P5, P6) are acquired from the China Meteorological Data Service Center and Kunming Meteorological Bureau.

Geospatial Data Platforms: The topographic and geomorphological indicator (P7) and water resource regulation and storage capacity indicator (S5) are retrieved from the Geospatial Data Cloud.

Government Official Websites: Qualitative Evaluation Category indicators (R1, R2) are accessed from the Official Website of the People’s Government of Kunming Municipality.

3.1.1. GIS-Based Analysis of Topography, Geomorphology, and River Network

(1)

P7 Topographic and Geomorphological Characteristics Indicator

The Topographic and Geomorphological Characteristics indicator (P7) under the Pressure dimension is calculated by integrating three surface features: elevation, slope, and aspect. Landsat imagery (30 m resolution) from the Geospatial Data Cloud was used to generate an elevation map for Kunming, from which the mean elevation was computed. Slope and aspect data were extracted using the Spatial Analyst tool in ArcGIS 10.8, and their mean values were calculated to describe terrain variability. These three metrics were then combined to produce the P7 indicator, as shown in Figure 4.

(2)

S5 Water Resource Regulation and Storage Capacity Indicator

The Water Resource Regulation and Storage Capacity indicator (S5) under the State dimension is represented by the river network density of the study area. Landsat imagery (30 m spatial resolution) for Kunming was first obtained from the Geospatial Data Cloud to generate a digital elevation model. Hydrological analysis was then conducted using the Spatial Analyst tools in ArcGIS, and we applied Fill to obtain a depressionless DEM, computed Flow Direction (D8) and Flow Accumulation, delineated streams by Con/Reclassify on the accumulation surface using a calibrated threshold, converted the stream raster to polylines with Stream to Feature, and assigned hierarchical orders via Stream Order, yielding the final vector river network [9]. Subsequently, line density analysis was applied to compute river network density and generate the corresponding density layer. Finally, the average river network density was calculated based on the raster data, serving as the indicator value for S5, as shown in Figure 5.

3.1.2. Calculation of Evaluation-Type Indicators

The R1 and R2 indicators under the response dimension are assessed using the CMMI method, mainly by collecting early warning information, emergency plans, and relevant regulations issued by provincial, municipal, and local authorities. Evaluation results are classified according to the five-level CMMI standard, with a comment set V = {Very Poor, Poor, Fair, Good, Excellent} and a corresponding score set S = {1, 2, 3, 4, 5}, as shown in

Table 1. Description of urban pluvial flood resilience evaluation indicators.

Evaluation Category	Dynamic Description of Indicators (Score)					Data Source
	1	2	3	4	5
R1 Flood Disaster Early Warning Capability	No disaster warning information acquisition channels	Disaster warnings released on government websites, but with low accuracy	Disaster warnings available on government websites, radio, and TV, with general accuracy	Disaster warnings disseminated on government websites, WeChat, radio, TV, mobile apps, and SMS, with high accuracy	Multiple channels for disaster warning dissemination, with authoritative forecasts from provincial meteorological bureaus and very high accuracy	Provincial/Municipal Meteorological Bureau
R2 Emergency Management Capacity	No emergency preparedness or disaster-related plans/regulations	General emergency preparedness plan, but lacking disaster-specific plans or regulations	Both general and specialized emergency plans, with 1–2 disaster-related plans or regulations	Both general and specialized emergency plans, with 3–5 disaster-related plans or regulations	Comprehensive emergency plans, dedicated emergency management departments, and more than five disaster-related plans or regulations	Provincial/Municipal Government Website

.

3.2. Research Methods

Based on the indicator system constructed using the aforementioned DPSIR model, and combined with multi-source data collection and processing, the annual values for each indicator from 2013 to 2022 for Kunming City are compiled and presented in

Table 2. Statistical data of pluvial flood resilience evaluation indicators for Kunming city (2013–2022).

Indicator	Unit	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
D1	billion yuan	3515.31	3712.99	3970.00	4300.43	4857.64	5206.90	6475.88	6733.79	7222.50	7541.37
D2	%	38.22	41.59	39.29	36.23	34.42	33.53	23.54	22.71	20.94	26.2
D3	%	49.9	50.7	55.3	56.7	57.3	56.6	63.7	64.2	64.2	63.7
D4	%	2.67	2.34	3.14	3.12	3.00	3.09	3.44	4.22	3.86	4.07
D5	%	68.05	69.05	60.06	57.03	72.05	72.85	73.60	79.67	80.50	81.10
P1	%	12.84	13.06	13.27	13.46	13.59	13.72	13.76	14.39	14.24	13.99
P2	%	15.86	15.96	16.03	16.09	16.10	16.09	16.60	14.98	14.70	14.20
P3	persons/km2	2137	2117	2170	2208	2206	2244	2282	2293	2311	2639
P4	mm	786.0	947.0	1077.0	957.0	1050.0	883.0	797.2	867.8	885.0	830.0
P5	mm	41.8	110.2	73.1	112.9	65.9	54.4	126.8	84.2	96.3	101.2
P6	%	2.63	2.80	4.10	2.14	2.22	0.79	1.90	4.04	2.78	0.71
P7	-	2214.384	2214.384	2214.384	2214.384	2214.384	2214.384	2214.384	2214.384	2214.384	2214.384
S1	km/km2	2.10	10.31	10.46	13.93	11.49	12.01	12.53	10.83	11.53	10.90
S2	m2	9.24	19.55	12.91	15.42	8.07	8.95	9.83	12.58	13.56	11.81
S3	%	10.60	10.89	10.94	8.41	7.72	7.45	7.20	7.79	7.31	7.80
S4	%	39.40	40.36	40.64	41.88	41.92	41.94	41.95	42.15	44.99	43.40
S5	km/km2	1.137436	1.157365	1.193679	1.421647	1.332355	1.307347	1.314488	1.329255	1.221491	1.326640
I1	10,000 persons	9.7600	2.4500	4.6200	10.240	42.110	1.6900	5.7600	0.0592	8.2800	2.7400
I2	units	58	37	52	24	115	44	39	52	65	62
I3	billion yuan	3.710000	2.540000	2.054900	0.394300	7.130000	0.410900	0.545002	2.590000	0.714360	0.081198
I4	%	22.64	24.49	34.69	33.33	34.00	40.82	50.98	61.82	61.10	60.38
I5	%	98.00	94.89	95.38	94.07	94.88	95.70	96.52	98.89	98.76	99.46
R1	-	2	2	2	2	3	4	3	4	5	5
R2	-	3	2	4	3	3	3	3	3	4	5
R3	%	22.51	23.65	21.16	21.43	22.03	22.91	23.95	20.77	21.17	21.42
R4	%	156.71	162.84	165.04	170.93	186.20	215.77	206.47	171.03	180.50	190.28
R5	%	0.73	0.77	0.83	0.93	0.90	0.91	0.92	0.78	0.77	0.77
R6	%	80.13	79.03	77.44	78.62	79.69	80.81	93.29	76.81	76.67	77.61

.

Weighting Methods

Weighting is central to evaluation frameworks. Scientific rigor requires matching the scheme to the objective: Subjective, expert-based methods suit data-scarce, policy-driven settings, whereas objective methods derive weights from data dispersion and fit data-rich contexts where bias should be minimized [49]. Internal validity further requires consistency checks for subjective weights and information preservation for objective weights so that variability is faithfully captured [50]. To enhance stability, combination weighting reconciles subjective and objective inputs and typically outperforms single methods [51]. Otherwise, ill-chosen schemes distort scores and weaken conclusions. This study adopts a hybrid weighting strategy that integrates both subjective and objective approaches: The Analytic Hierarchy Process (AHP) is first employed to determine subjective weights, followed by the Entropy Weight Method (EWM) for objective weights. Finally, to balance the strengths of both methods, a game theory-based combination weighting approach is introduced to derive the final composite weights, as illustrated in Figure 6.

(1)

Analytic Hierarchy Process (AHP)

The Analytic Hierarchy Process (AHP) is a decision-making method that quantitatively analyzes qualitative problems. The core principle is to decompose a complex system into a goal layer, a criterion layer, and an alternative layer, enabling both qualitative and quantitative analysis to determine the indicator weights [52]. This approach structures various factors in complex issues into interconnected and ordered hierarchical levels. Through pairwise comparisons, it assesses the relative importance of each factor within the hierarchy. Mathematical methods are then applied to calculate the weights that reflect the relative priorities of each element at each level, and an overall ranking across all levels is performed to determine the final weights of all elements. The main steps include ① establishing the hierarchical structure model; ② conducting pairwise comparisons of indicators at the same level to construct the comparison matrix; ③ solving for the eigenvector corresponding to the maximum eigenvalue of the judgment matrix; ④ checking the consistency ratio (CR value). When CR < 0.1, the judgment matrix is considered to meet the consistency requirement. The calculation formulas are shown in Equations (1) and (2):

$$CR={\displaystyle \frac{CI}{RI}}$$

(1)

$$CI={\displaystyle \frac{{\lambda }_{max}-n}{n-1}}$$

(2)

In the formulas, CR denotes the consistency ratio. When CR < 0.1, it indicates that the consistency of the judgment matrix is within the acceptable range. RI is the random consistency index, determined from standard reference tables. ${\lambda }_{max}$ is the maximum eigenvalue of the judgment matrix, and n represents the order of the matrix.

An expert panel of five (two university scholars, two government technologists, and one engineering specialist) spanning urban rain–flood resilience and emergency management completed an anonymized, independent AHP survey using Saaty’s 1–9 scale with standardized examples. Pairwise-comparison matrices were built at the criterion level for DPSIR (D, P, S, I, R; 5 × 5) and at the sub-criterion level for D (5 indicators), P (7), S (5), I (5), and R (6). Individual matrices were aggregated by element-wise arithmetic means to form the group matrix; weights were derived via the principal eigenvector method, and ${\lambda }_{max}$, CI, and CR were computed. Consistency was acceptable at all levels (CR < 0.10): criteria = 5.3343, CR = 0.0746; sub-criteria–D: 5.1925, CR = 0.0430; P: 7.7700, 0.0944; S: 5.1605, 0.0358; I: 5.3276, 0.0731; R: 6.5431, 0.0862.

(2)

Entropy Weight Method (EWM)

The Entropy Weight Method (EWM) is a weighting approach that objectively reflects the differences among evaluation indicators based on data [53]. During the weighting analysis, it intuitively highlights the informational content of each indicator, ensuring clear differentiation among them and avoiding ambiguous weight results caused by minimal differences. The entropy value reflects the uncertainty of an event; the greater the internal variation within a particular indicator, the more information it contains, leading to lower uncertainty, a smaller entropy value, and thus a higher weight. Conversely, higher entropy indicates less information and results in a smaller weight. The processing steps of the entropy weight method are as follows:

① Data normalization is performed on the raw data to obtain a standardized matrix $V_{ij}=v_{ij}$, where $v_{ij}$ represents the standardized value of the j-th urban pluvial flood resilience indicator in the i-th region. The calculation formulas for $v_{ij}$ are shown in (3) and (4):

$$\mathrm{Positive} \mathrm{Indicator}: v_{ij}{\displaystyle \frac{x_{ij}-min\left\{x_{ij}\right\}}{max\left\{x_{ij}\right\}-min\left\{x_{ij}\right\}}}$$

(3)

$$\mathrm{Negative} \mathrm{Indicator}: v_{ij}{\displaystyle \frac{min\left\{x_{ij}\right\}-x_{ij}}{max\left\{x_{ij}\right\}-min\left\{x_{ij}\right\}}}$$

(4)

② To calculate the weight of the j-th indicator in the i-th regional project as shown in Formula (5):

$$p_{ij}={\displaystyle \frac{v_{ij}}{{{\displaystyle \sum }}_{i=1}^n{v}_{ij}}}$$

(5)

where: i = 1, 2, 3, …, n; j = 1, 2, 3, …, m. n represents the number of urban pluvial flood resilience evaluation indicators, and m represents the number of urban pluvial flood resilience evaluation regional projects.

③ To calculate the entropy value of the j-th indicator as shown in Formula (6):

$$e_j=-k{{\displaystyle \sum }}_{i=1}^n{p}_{ij}\ln \left(p_{ij}\right)$$

(6)

where:$k = 1/ln\left(n\right) > 0$, ensuring that $e_j$ ≥ 0.

④ To calculate the information redundancy as shown in Formula (7):

$$d_j=1-e_j$$

(7)

⑤ Finally, to calculate the weights of each indicator as shown in Formula (8):

$$w_j={\displaystyle \frac{d_j}{{{\displaystyle \sum }}_{j=1}^m{d}_j}}$$

(8)

(3)

Game Theory-Based Combined Weighting

The game theory-based aggregation model aims for a Nash equilibrium, comprehensively considering the conflicts and contradictions between subjective and objective weights, seeking consensus and compromise amid opposition. In the game process, dynamic comparisons and mutual coordination are used to integrate data and information in a combined weight model [54]. As previously described with the entropy weight method, when analyzing urban pluvial flood resilience levels, it is necessary to consider the unique natural environment. This is reflected in the pressure disturbances (P) in the established indicator system. For example, the P7 topographical and geomorphological characteristic indicator remains relatively static for a long period, and even if the selected measurement surface varies, the relative elevation does not undergo significant changes. However, elevation differences in different areas have a considerable impact on the city’s system, maintaining relative stability and providing an appropriate response to stormwater disturbances. Therefore, when combining the entropy weight method with AHP weight results, it is essential to consider both weight data and information data. The optimal linear combination of subjective and objective weights should be achieved based on the game theory aggregation model.

① First, use Formula (9) to construct the linear combination:

$$w={\alpha }_1{\omega }_1^T+{\alpha }_2{\omega }_2^T$$

(9)

In the equation, $w$ represents the comprehensive weight, ${\alpha }_1$ and ${\alpha }_2$ are the coefficients of the linear combination, and $\omega =\{{\omega }_1, {\omega }_2\}$ is the vector set where ${\omega }_1$ represents the subjective weights determined by AHP, and ${\omega }_2$ represents the objective weights determined by the entropy weight method.

② Next, the objective function is determined. Based on the game theory aggregation model, the optimal linear combination of subjective and objective weights is performed by finding the minimum deviation to optimize the linear combination coefficients ${\alpha }_1$ and ${\alpha }_2$, thus obtaining the optimal weights. The objective function is defined as shown in Equation (10):

$$min\Vert W-{\omega }_{\mathrm{k}}{\Vert }_2$$

(10)

In the equation, k = 1, 2.

③ By utilizing the properties of matrix differentiation, Equation (10) can be equivalently transformed into a system of linear equations representing the first-order optimality condition, as shown in Equation (11), thereby obtaining the optimal linear combination coefficients ${\alpha }_1$ and ${\alpha }_2$.

$$\left(\begin{array}{cc}{\omega }_1{\omega }_1^T& {\omega }_1{\omega }_2^T\\ {\omega }_2{\omega }_1^T& {\omega }_2{\omega }_2^T\end{array}\right)\left[\begin{array}{c}{\alpha }_1\\ {\alpha }_2\end{array}\right]=\left[\begin{array}{c}{\omega }_1{\omega }_1^T\\ {\omega }_2{\omega }_2^T\end{array}\right]$$

(11)

Based on the above equation, the minimum deviation optimal linear combination coefficients ${\alpha }_1$ and ${\alpha }_2$ are obtained and subsequently normalized. By substituting these results into Equation (10), the model achieves an optimal linear combination of subjective and objective weights using game-theoretic aggregation, yielding the final combined weight vector W as shown in Equation (12):

$$W={\alpha }_1^{\ast }{\omega }_1^T+{\alpha }_2^{\ast }{\omega }_2^T$$

(12)

In the equation, ${\alpha }_1^{\ast }$ denotes the normalized weight coefficient of ${\alpha }_1$, while ${\alpha }_2^{\ast }$ denotes the normalized weight coefficient of ${\alpha }_2$.

$${\alpha }_1^{\ast }={\displaystyle \frac{{\alpha }_1}{{\alpha }_1+{\alpha }_2}}$$

$${\alpha }_2^{\ast }={\displaystyle \frac{{\alpha }_2}{{\alpha }_1+{\alpha }_2}}$$

The final calculation of the combination coefficients is shown in

Table 3. Optimal combination coefficients.

Coefficient	α1	α2	α1*	α2*
Result	0.754	0.393	0.657	0.343

.

The weight calculation results are shown in

Table 4. Weight values of Kunming’s pluvial flood resilience evaluation indicators based on the DPSIR model.

Target Layer	Criteria Layer	Weight	Scheme Layer	Attribute	AHP Weight	Entropy Weight	Combined Weight
Urban Pluvial Flood Resilience Assessment	D	0.3234	D1 Regional Economic Condition	+	0.0273	0.0489	0.0347
			D2 Household Economic Condition	+	0.0111	0.0396	0.0209
			D3 Economic Diversity	+	0.0076	0.0347	0.0169
			D4 Employment Situation	−	0.0144	0.0337	0.0210
			D5 Urbanization Rate	+	0.0222	0.0272	0.0239
	P	0.2194	P1 Aging Population Ratio	−	0.0192	0.0325	0.0238
			P2 Child Population Ratio	−	0.0168	0.0407	0.0250
			P3 Population Exposure Density	−	0.0266	0.0158	0.0203
			P4 Long-Term Precipitation Pattern	−	0.0555	0.0291	0.0464
			P5 Short-Duration Precipitation Intensity	−	0.0621	0.0339	0.0524
			P6 Flood Disaster Risk	−	0.1242	0.0367	0.0942
			P7 Topographic and Geomorphological Characteristics	−	0.0996	0.0000	0.0655
	S	0.1663	S1 Drainage Network Condition	+	0.0407	0.0151	0.0319
			S2 Urban Road Condition	+	0.0256	0.0460	0.0326
			S3 Building Exposure Density	−	0.0207	0.0411	0.0277
			S4 Green Coverage Ratio	+	0.0255	0.0291	0.0267
			S5 Water Resource Regulation and Storage Capacity	+	0.0536	0.0352	0.0473
	I	0.1259	I1 Affected Population	−	0.0446	0.0148	0.0344
			I2 Affected Towns and Subdistricts	−	0.0152	0.0148	0.0157
			I3 Direct Economic Loss	−	0.0322	0.0175	0.0272
			I4 Proportion of Excellent Surface Water Quality	+	0.0182	0.0423	0.0265
			I5 Wastewater Treatment Rate	+	0.0135	0.0404	0.0227
	R	0.1644	R1 Flood Disaster Early Warning Capability	+	0.0871	0.0806	0.0849
			R2 Emergency Management Capacity	+	0.0236	0.0276	0.0250
			R3 Public Response Capability	+	0.0207	0.0459	0.0293
			R4 Communication Capability	+	0.0056	0.0429	0.0184
			R5 Medical Rescue Capacity	+	0.0418	0.0403	0.0413
			R6 Level of Social Security	+	0.0488	0.0918	0.0635

Note: “+” denotes that the indicator is positively correlated with urban pluvial flood resilience, where a larger value indicates better performance; “−” denotes a negative correlation, where a smaller value indicates better performance.

.

3.3. Cloud Model Based on Extenics

Extensional theory can address the issues of contradiction and incompatibility within evaluation systems, while cloud models can account for both the randomness and fuzziness of indicators. The extensional cloud model combines the advantages of both, effectively handling the multidimensional indicators and uncertainty issues in urban flood resilience assessments, thereby providing more comprehensive and accurate evaluation results [55].

3.3.1. Construction of the Object to Be Evaluated

The extensional cloud model is manifested in the transformation between the classical domain and the standard cloud, using the cloud membership function to replace the relevance function in the extensional element model. Additionally, considering perceptibility, the grade characteristic values from the extensional element evaluation are used to represent the grade bias. For the evaluation object, the results can be represented by an extensional element matrix, as shown in Equation (13):

$$R=\left(N,C,V\right)=\left[\begin{array}{ccc}N& C_1& V_1\\ & C_2& V_2\\ & ⋮& ⋮\\ & C_m& V_m\end{array}\right]=\left[\begin{array}{ccc}N& C_1& \left(E_{x_1},E_{n_1},H_{e_1}\right)\\ & C_2& \left(E_{x_2},E_{n_2},H_{e_2}\right)\\ & ⋮& ⋮\\ & C_m& \left(E_{x_m},E_{n_m},H_{e_m}\right)\end{array}\right]$$

(13)

In the equation, R represents the urban pluvial flood resilience evaluation grade result; N is the object to be evaluated; $C_i$ represents the urban pluvial flood resilience evaluation indicators (i = 1, 2, …, m); $V_i$ is the specific value of the object to be evaluated; ($E_{x_i},E_{n_i},E_{e_i}$) are the cloud parameters corresponding to the evaluation indicator $C_i$.

3.3.2. Construction of the Standard Cloud for Evaluation

After reviewing and organizing the relevant materials, the risk level indicators for assessing the current state of urban pluvial flood resilience are classified into five levels: I, II, III, IV, and V. The evaluation criteria and definitions for each level are presented in

Table 5. Classification of comprehensive urban flood resilience evaluation levels.

Evaluation Level	Form of Expression
I	The city’s ability to resist stormwater disasters is low, with poor resilience.
II	The city’s ability to resist stormwater disasters is relatively low, with poor resilience.
III	The city’s ability to resist stormwater disasters is moderate, with medium resilience.
IV	The city’s ability to resist stormwater disasters is relatively high, with good resilience.
V	The city’s ability to resist stormwater disasters is high, with excellent resilience.

.

After dividing the urban flood resilience into comprehensive levels, it is necessary to categorize each indicator and construct corresponding interval values. Then, the cloud model is used to fuzzify the boundaries of the resilience levels, allowing the calculation of the expected value $E_x$, entropy $E_n$, and hyper-entropy $H_e$ for the normal cloud model of the urban pluvial flood resilience comprehensive evaluation. The calculation formulas for the numerical characteristics of the standard cloud are shown in Equation (14):

$$\left\{\begin{array}{c}Ex={\displaystyle \frac{T_{max}+T_{min}}{2}}\\ En={\displaystyle \frac{T_{max}-T_{min}}{6}}\\ He=b\end{array}\right.$$

(14)

In the equation, $T_{max}$ and $T_{min}$ represent the upper and lower limits of each indicator evaluation interval; $E_x$ is the expectation of the standard cloud; $E_n$ is the entropy of the standard cloud; $H_e$ is the hyper-entropy; and $b$ is a random number that can be adjusted according to the actual situation [56].

Each indicator is classified into five grades: “I, II, III, IV, IV;”. The quantitative values for each evaluation indicator are determined based on expert opinions, relevant standards, and literature. In this study, the standard boundaries for the grades of urban pluvial flood resilience evaluation indicators are shown in

Table 6. Grade boundaries of urban pluvial flood resilience evaluation indicators.

Indicator Number	Indicator Value Ranges for Different Evaluation Grades
	I	II	III	IV	V
D1	(792.379, 3073.029)	(3073.029, 4593.464)	(4593.464, 6113.898)	(6113.898, 7634.333)	(7634.333, 9914.983)
D2	(8.807, 20.237)	(20.237, 27.857)	(27.857, 35.477)	(35.477, 43.097)	(43.097, 54.527)
D3	(41.748, 49.989)	(49.989, 55.483)	(55.483, 60.977)	(60.977, 66.471)	(66.471, 74.712)
D4	(5.107, 4.201)	(4.201, 3.597)	(3.597, 2.993)	(2.993, 2.389)	(2.389, 1.483)
D5	(46.811, 59.103)	(59.103, 67.298)	(67.298, 75.494)	(75.494, 83.689)	(83.689, 95.981)
P1	(15.120, 14.376)	(14.376, 13.880)	(13.880, 13.384)	(13.384, 12.888)	(12.888, 12.144)
P2	(17.947, 16.804)	(16.804, 16.042)	(16.042, 15.280)	(15.280, 14.518)	(14.518, 13.375)
P3	(2704.958, 2482.828)	(2482.828, 2334.743)	(2334.743, 2186.657)	(2186.657, 2038.572)	(2038.572, 1816.442)
P4	(1205.762, 1056.881)	(1056.881, 957.627)	(957.627, 858.373)	(858.373, 759.119)	(759.119, 610.238)
P5	(169.336, 128.009)	(128.009, 100.456)	(100.456, 72.904)	(72.904, 45.351)	(45.351, 4.024)
P6	(5.837, 4.124)	(4.124, 2.982)	(2.982, 1.840)	(1.840, 0.698)	(0.698, 0.000)
P7	(2767.98, 2546.5416)	(2546.5416, 2325.1032)	(2325.1032, 2103.6648)	(2103.6648, 1882.2264)	(1882.2664, 1660.788)
S1	(1.072, 5.841)	(5.841, 9.020)	(9.020, 12.199)	(12.199, 15.378)	(15.378, 20.146)
S2	(1.749, 7.233)	(7.233, 10.452)	(10.452, 13.932)	(13.932, 17.413)	(17.413, 22.635)
S3	(13.276, 10.944)	(10.944, 9.389)	(9.389, 7.833)	(7.833, 6.278)	(6.278, 3.946)
S4	(37.165, 39.515)	(39.515, 41.080)	(41.080, 42.646)	(42.646, 44.212)	(44.212, 46.561)
S5	(1.001, 1.137)	(1.137, 1.229)	(1.229, 1.320)	(1.320, 1.411)	(1.411, 1.547)
I1	(45.425, 27.098)	(27.098, 14.880)	(14.880, 2.662)	(2.662, 0.100)	(0.100, 0.000)
I2	(128.540, 91.670)	(91.670, 67.090)	(67.090, 42.510)	(42.510, 17.930)	(17.930, 0.000)
I3	(8.530, 5.274)	(5.274, 3.103)	(3.103, 0.931)	(0.931, 0.100)	(0.100, 0.000)
I4	(0.000, 19.788)	(19.788, 34.879)	(34.879, 49.971)	(49.971, 65.062)	(65.062, 87.698)
I5	(90.772, 93.714)	(93.714, 95.675)	(95.675, 97.636)	(97.636, 99.596)	(99.596, 100.000)
R1	(0.5–1.5)	(1.5–2.5)	(2.5–3.5)	(3.5–4.5)	(4.5–5.5)
R2	(0.5–1.5)	(1.5–2.5)	(2.5–3.5)	(3.5–4.5)	(4.5–5.5)
R3	(18.770, 20.436)	(20.436, 21.545)	(21.545, 22.655)	(22.655, 23.765)	(23.765, 25.430)
R4	(122.743, 151.661)	(151.661, 170.938)	(170.938, 190.216)	(190.216, 209.493)	(209.493, 238.411)
R5	(0.600, 0.716)	(0.716, 0.793)	(0.793, 0.869)	(0.869, 0.947)	(0.947, 1.062)
R6	(65.388, 72.700)	(72.700, 77.573)	(77.573, 82.447)	(82.447, 87.320)	(87.320, 94.632)

. The parameters of the normal cloud model, $E_x$, $E_n$ and $H_e$, are calculated using Equation (14). With these three characteristic parameters, the cloud matter-element corresponding to each grade of urban pluvial flood resilience can be determined. In this study, the value of b is set to 0.5. This value is chosen to achieve a moderate “cloud thickness”such that adjacent classes overlap enough to accommodate randomness without excessive overlap that would erode separability, consistent with the fuzziness required by our indicator grading and normalized scale [57].

3.3.3. Calculation of Cloud Membership Degree

(a)

Calculation of the Membership Degree for Secondary Indicators

The data of each indicator in the urban flood resilience evaluation is treated as a single cloud drop, which is substituted into the normal cloud generator to produce a random number $E_n^{\prime }$ with an expectation of $E_n$ and a standard deviation of $H_e$. The cloud membership degree $\mu$ is then calculated using Equation (15).

$$\mu =exp\left(-{\displaystyle \frac{{\left(x-E_x\right)}^2}{2{\left(E{n}^{\prime }\right)}^2}}\right)$$

(15)

The cloud membership degree between the actual score value of each indicator and the standard normal cloud of the resilience evaluation level is calculated. The above steps are repeated n times, and the median of all results is taken as the final membership degree. The overall membership degree matrix K is then obtained as shown in Equation (16).

$$K=\left[\begin{array}{cccc}{K}_1\left(c_1\right)& K_2\left(c_1\right)& \dots & K_m\left(c_1\right)\\ K_1\left(c_2\right)& K_2\left(c_2\right)& \dots & K_m\left(c_2\right)\\ ⋮& ⋮& \ddots & ⋮\\ K_1\left(c_p\right)& K_2\left(c_p\right)& \dots & K_m\left(c_p\right)\end{array}\right]$$

(16)

In the equation, $K_m\left(C_p\right)$ represents the cloud membership degree between the evaluation indicator and the level; mmm is the level; and $p$ is the number of evaluation indicators.

(b)

Calculation of the Membership Degree for First-Level Indicators

The membership degree for first-level indicators is calculated by weighted aggregation of the second-level indicators’ membership degrees, as shown in Equation (17):

$$K\left(b_i\right)=\left({\omega }_{c1},{\omega }_{c2},\dots ,{\omega }_{cp}\right)·K$$

(17)

where ${\omega }_{bp}$ is the weight vector of the second-level indicators, and $K$ is the second-level membership degree matrix.

(c)

To calculate the membership degree of the target level

The weighted membership degree of the primary indicators is used, as shown in Equation (18):

$$K\left(a_i\right)=\left({\omega }_{b1},{\omega }_{b2},\dots ,{\omega }_{bp}\right)·K\left(b_i\right)$$

(18)

where ${\omega }_{bp}$ is the weight vector for the first-level indicators, and $K\left(b_i\right)$ is the membership degree matrix for the first-level indicators.

3.3.4. Determination of Resilience Grade

According to the principle of maximum membership degree, if $K_j\left(N\right)=ma{x}_{j=\left\{1,2,\dots ,m\right\}}{K}_j\left(N\right)$ (19), then the resilience level of evaluation object $N$ is considered to be level $j$ (20).

$$\overline{K_j\left(N\right)}={\displaystyle \frac{K_j\left(N\right)-\underset{j}{\min } K_j\left(N\right)}{\underset{j}{\max } K_j\left(N\right)-\underset{j}{\min } K_j\left(N\right)}}$$

(19)

$$j^{\ast }={\displaystyle \frac{{{\displaystyle \sum }}_{j=1}^{m}j\cdot \overline{K_j\left(N\right)}}{{{\displaystyle \sum }}_{j=1}^m\overline{K_j\left(N\right)}}}$$

(20)

In the equation, $j^{\ast }$ represents the level of urban flood resilience.

4. Results and Analysis

4.1. Annual Variation in Pluvial Flood Resilience Grades in Kunming

The comprehensive membership degrees and grade characteristic values for each year were calculated, and the annual urban pluvial flood resilience grades for Kunming from 2013 to 2022 were determined according to the maximum membership principle. The detailed results are presented in

Table 7. Evaluation results of annual urban pluvial flood resilience grades in Kunming.

Comprehensive Resilience	I	II	III	IV	V	Evaluation Result
2013	0.102760	0.339464	0.303803	0.189265	0.064107	II
2014	0.052962	0.377563	0.313619	0.170835	0.084421	II
2015	0.108413	0.388869	0.329149	0.151203	0.021766	II
2016	0.090548	0.296046	0.372618	0.201089	0.039076	III
2017	0.104462	0.146300	0.588385	0.132895	0.027358	III
2018	0.019670	0.091550	0.557753	0.246666	0.083762	III
2019	0.034342	0.165366	0.362520	0.355857	0.081315	III
2020	0.064482	0.199399	0.320750	0.352979	0.061790	IV
2021	0.048079	0.170395	0.314985	0.363478	0.102463	IV
2022	0.045853	0.164374	0.232913	0.384117	0.172142	IV

Note: The maximum value is shown in bold.

.

Based on the maximum membership principle, the evaluation results of urban pluvial flood resilience in Kunming from 2013 to 2022 reveal a phased evolution from fluctuation to steady improvement in resilience levels (Figure 7).

(1)

Initial Construction Stage (2013–2015)

During this stage, Kunming’s resilience in the Driving Forces and Pressure dimensions remained low, reflecting that economic and policy support was at an early stage and that significant challenges persisted in urban pluvial flood management. Resilience in the State dimension improved from Level II (2013) to Level III (2015), primarily due to the release of the Kunming Urban Drainage (Rainwater) and Waterlogging Prevention Comprehensive Plan in 2015. The plan enhanced urban disaster-prevention infrastructure and ecological conditions by strengthening the Drainage Network Condition, improving Urban Road Condition, regulating Building Exposure Density, increasing the Green Coverage Ratio, and enhancing Water Resource Regulation and Storage Capacity. These measures strengthened the city’s capacity to cope with pluvial floods, laying the foundation for subsequent development. Resilience in the Impact dimension decreased to Level II in 2014 but recovered in 2015. This fluctuation was primarily attributed to the heavy rainfall in July 2014, during which precipitation in some areas reached torrential levels and caused severe waterlogging in low-lying zones and regions with weak Drainage Network Conditions, thereby exposing deficiencies in the city’s drainage capacity. With increasing attention to long-term disaster impacts and environmental protection, resilience in the Impact dimension improved in 2015. Resilience in the Response dimension declined from Level III to Level II between 2013 and 2015, indicating that post-disaster emergency management and recovery systems remained unstable. Although emergency response capacity gradually improved, the system was still constrained by insufficient resource allocation and institutional deficiencies during this stage.

(2)

Resilience Enhancement Stage (2016–2019)

During this stage, significant progress was made in advancing policies, economic support, and infrastructure development for pluvial flood management in Kunming, particularly reflected in improvements in the Driving Forces dimension. Although improvements were observed in drainage systems and ecological construction, fluctuations in the resilience level of the State dimension indicated ongoing challenges in infrastructure development and environmental protection. In 2017, two rounds of heavy rainfall, including an exceptionally severe storm in July, were experienced in Kunming, causing extensive urban waterlogging and infrastructure damage. Although emergency responses were promptly activated and partial progress was achieved, deficiencies in post-disaster recovery and impact management were revealed by the aging drainage system and the city’s mountainous terrain, resulting in a sharp decline of resilience in the Impact dimension to Level I. The resilience level in the Response dimension began to rise after 2016, reaching Level IV by 2018. This reflected notable progress in strengthening emergency management capacity, social security, and resource allocation, thereby demonstrating a gradual enhancement of post-disaster recovery capacity.

(3)

Resilience Reinforcement Stage (2020–2022)

During this stage, significant improvements in resilience were recorded across several dimensions, most notably within the Driving Forces, Impact, and Response dimensions. However, in 2021, the resilience level of the Response dimension fell to Level II, largely attributed to shortcomings in emergency management plan execution, insufficient responses to extreme weather, inadequacies in the social security system, and breakdowns in information dissemination and coordination. Although emergency plans and a “smart flood control” system had been established, the heavy rainfall in June and the cold wave in December 2021 exposed governance weaknesses, such as limited rescue efficiency, delayed responses to extreme weather, postponed disbursement of relief funds, and inadequate infrastructure. Overall, resilience in pluvial flood management was advanced during this period through strengthened policy orientation, optimized infrastructure development, and enhanced emergency management capacity, thereby establishing a stronger foundation for future disaster response and post-disaster recovery. Collectively, Kunming’s pluvial flood management has shifted from a moderate response toward higher resilience, with the urban system exhibiting greater stability and adaptability when subjected to extreme climatic shocks.

4.2. Pluvial Flood Resilience Analysis Based on the DPSIR Model

Across all dimensions, Kunming’s urban pluvial flood resilience demonstrated significant annual fluctuations from 2013 to 2022, as illustrated in Figure 8.

The resilience of the Driving Forces dimension remained low, consistently at Level II from 2013 to 2016, reflecting limited policy and socio-economic support. Resilience rose after 2017, reaching Level IV in 2019 and remaining stable, driven by economic growth and stronger social policies, particularly diversification and social security. The Pressure dimension remained stable (2013–2022), indicating balanced flood pressures. This resulted from weighting, indicator traits, and the Extension Cloud Model classification. P6 and P7, with higher weights, anchored the structure. P7, static, stabilized the grade, while P6 varied narrowly with monsoonal rainfall, keeping results near Level III. Lower-weighted P1–P5 fluctuated but had little effect. Under the Extension Cloud Model, maximum membership kept Level III dominant. The State dimension stayed at Level II before 2014, indicating weak infrastructure and recovery, but fluctuated between Levels II and III after 2015, reflecting gradual flood prevention improvements. The Impact dimension was volatile, dropping to Level I in 2017 due to extreme weather and weak infrastructure. It then rose steadily, stabilizing at Level IV after 2020, showing effective recovery and ecological gains. The Response dimension mostly stayed at Levels II–III, showing gradual gains but remaining gaps. It rose to Level IV in 2020 and Level V in 2022, reflecting advances in emergency management, social security, and medical rescue, especially during COVID-19.

4.3. Pluvial Flood Resilience Analysis by Dimension

4.3.1. Driving Force

The Driving Forces dimension overall exhibits a dual characteristic of increased economic support but intensified pressure on residents, as shown in Figure 9. This indicates that while Kunming’s macroeconomic development remains robust, more attention should be paid to the stability of residents’ income and employment at the micro level, in order to achieve a synergistic improvement in overall urban pluvial flood resilience.

The resilience level of the Regional Economic Condition continued to rise and stabilize, indicating that Kunming’s economic risk resistance capacity was enhanced through increased investment, accelerated GDP growth, and improved fiscal revenue. In contrast, Household Economic Condition exhibited a persistent decline, as residents’ “economic reserves” and self-help capacity were diminished by lagging income growth, rising living costs, and deteriorating employment quality. The resilience of Economic Diversity steadily increased, suggesting that industrial optimization enhanced the overall economic capacity to absorb shocks and mitigate the impacts of pluvial flood disasters. Employment Situation showed a fluctuating decline, with stability eroded by economic restructuring and external shocks, including the COVID-19 pandemic. The resilience of the Urbanization Rate initially decreased but subsequently improved and stabilized, reflecting that early-stage urbanization was constrained by insufficient facilities, whereas since 2017, resilience has advanced significantly due to policy promotion and infrastructure enhancement. To further strengthen resilience in the Driving Forces dimension, priority should be placed on promoting collective wage negotiations, adjusting minimum wage standards, supporting small and micro enterprises as well as flexible employment groups, reinforcing price regulation, and expanding social security coverage to enhance household income and reduce costs. In parallel, the development of labor-intensive industries, reinforcement of vocational training related to pluvial flood resilience, and optimization of employment platforms are essential to stabilize employment among vulnerable groups. At the same time, continuous economic investment, particularly in green and high-tech industries alongside industrial restructuring, remains critical for enhancing overall economic resilience to external shocks.

4.3.2. Pressure

The Pressure dimension overall exhibits characteristics of “structural vulnerability and extreme fluctuations”, as shown in Figure 10. This indicates that, in the context of increasing population sensitivity and intensified climate uncertainty, Kunming needs to strengthen the protection of vulnerable groups and enhance adaptation strategies for rainfall risks, thereby improving the system’s pressure resistance capacity.

The resilience level of the Aging Population Ratio continued to decline, primarily due to the rising aging rate and the limited emergency self-rescue capacity of elderly individuals, which increased their exposure to pluvial flood disasters. In contrast, the resilience of the Child Population Ratio has risen significantly since 2020, supported by declining birth rates and improved drainage systems in schools and residential areas, which reduced children’s disaster risk. The resilience of Population Exposure Density declined markedly during urban expansion, as population concentration in low-lying flood-prone areas and outdated drainage systems in older communities increased risks. The resilience levels of Long-Term Precipitation Pattern, Short-Duration Precipitation Intensity, and Flood Disaster Risk fluctuated considerably, primarily due to climate change and abnormal monsoon patterns, which destabilized rainfall regimes. By contrast, the resilience of Topographic and Geomorphological Characteristics remained stable, since its natural attributes are difficult to alter. To address these challenges, targeted measures should be implemented in Kunming, such as developing a “Flood Risk Map for the Elderly Population”, renovating infrastructure in aging communities, and providing emergency bracelets and shelters. In high-risk areas, “population resettlement + housing reinforcement” should be advanced, high-risk no-construction zones designated, and GIS with big data applied to trigger early evacuation warnings. For rainfall and flood risk management, the “Sponge City 2.0” initiative should be accelerated, prioritizing drainage upgrades in high-risk areas. River dredging, channel widening, and intelligent early warning systems are essential to strengthen drainage and emergency response capacity, thereby enhancing resilience in the Pressure dimension.

4.3.3. State

The state dimension exhibits an overall trend of “fluctuating increase with occasional declines”, as shown in Figure 11. This indicates that, although Kunming has achieved phased progress in infrastructure development and ecological retention, further efforts are needed to enhance system maintenance and integrated management in order to consolidate the foundation of pluvial flood resilience.

The resilience of the Drainage Network Condition initially improved after renovation projects but declined after 2017 owing to frequent extreme rainfall and delayed pipeline repairs. The resilience of Urban Road Condition fluctuated significantly: While 2014 sponge city pilots briefly enhanced capacity, the absence of citywide rollout and facility failures led to a gradual decline. The resilience of Building Exposure Density rose steadily due to flood-control renovations in older communities and policies avoiding flood-prone zones, with retrofitting sustaining high resilience. The resilience of the Green Coverage Ratio increased steadily, reaching Level V with wetland parks and ecological corridors, but declined slightly in 2022 under extreme weather. The resilience of Water Resource Regulation and Storage Capacity showed large fluctuations: Facilities built or restored in 2016–2017 boosted resilience temporarily, but mismanagement and extreme rainfall eroded stability. To enhance State-dimension resilience, a digital twin platform should be established for real-time drainage monitoring, with density increased in aging districts. Citywide sponge city retrofitting should be accelerated with reinforced facility maintenance. Coordinated operation and intelligent management of storage facilities should be optimized to ensure effective functioning within design thresholds, strengthening capacity to cope with pluvial flooding.

4.3.4. Impact

The impact dimension exhibits an overall trend of “frequent disaster shocks—gradual manifestation of governance effectiveness”, as shown in Figure 12. This suggests that, although Kunming faces high volatility risks caused by extreme weather events, the effectiveness of its governance measures is becoming increasingly evident.

The resilience of the Affected Population showed sharp fluctuations. In 2017, extreme rainfall overloaded drainage systems and sharply increased affected residents, while in 2020, reduced rainfall and efficient evacuation lowered impacts. Persistent pipeline defects in old districts caused recurrent flooding, keeping resilience between Levels III and IV. The resilience of Affected Towns closely mirrored I1, with valley towns heavily impacted in 2017 due to weak flood defenses. Direct Economic Loss fluctuated markedly: Industrial and agricultural areas suffered severe losses in 2017, but by 2022, improved defenses and wider insurance coverage had reduced damages. Excellent Surface Water Quality remained at Level II until 2017, then improved with stricter wastewater standards and storage expansion, reaching Level IV in 2022. Wastewater Treatment Rate fell during 2014–2017 but rose steadily after 2018 through diversion and smart scheduling, stabilizing at Level V. To strengthen Impact-dimension resilience, a flood vulnerability map and household evacuation plans should be developed, with orders issued via smart devices and township emergency teams. Industrial and agricultural areas require risk assessments, retrofitting, resilient farming techniques, and subsidized flood insurance. Ecological measures should include bioretention zones, rain gardens, river restoration, and continued rain–sewage diversion. A sewer system digital twin should ensure efficient wastewater treatment.

4.3.5. Response

The response dimension overall exhibits a pattern of “technology-driven improvement with disaster-induced fluctuations”, as shown in Figure 13. This indicates that Kunming has achieved phased progress in early warning, governance, and security systems. However, the city still encounters challenges related to insufficient system coordination and stability in responding to sudden pluvial flood events.

The resilience level of flood disaster early warning capability has gradually increased with the adoption of meteorological radar and AI-based rainfall prediction models. The resilience level of emergency management capability fluctuated but has steadily improved to Level V via the implementation of the digital twin platform. Public response capability improved from 2014 to 2016, driven by educational poverty alleviation measures and expanded vocational education, but declined during the COVID-19 pandemic. Communication capability rose briefly due to full deployment of 5G base stations and emergency satellite communication vehicles; however, extreme rainfall and mountainous terrain reduced its resilience to Level III–IV. Medical rescue capability declined due to the COVID-19 pandemic. The Level of social security improved via promoting electronic medical insurance certificates and government-subsidized contributions but remains limited by fiscal constraints. To further enhance Response dimension resilience, Kunming should establish an “educational information register for the floating population” to ensure accurate educational statistics, promote “accelerated adult primary education programs”, and implement a “vocational education return plan”; for flexible workers, offer “deferred medical insurance payments plus subsidies” and enhance medical insurance service mobility.

5. Discussion

5.1. Implications of the Evolutionary Characteristics of Urban Pluvial Flood Resilience in Kunming

The evolution of urban pluvial flood resilience in Kunming from 2013 to 2022, which followed the stages of “initial construction—resilience improvement—resilience enhancement”, closely mirrors the systemic policies and engineering practices implemented in the city during this period. The implications of this process can be deeply analyzed from three perspectives: policy-driven mechanisms, the synergy of ecological and technological measures, and the distinctive adaptive path of mountainous cities.

5.1.1. Phase-Based Leap Driven by Policy Initiatives

The promulgation of the Kunming City Urban Drainage and Stormwater Flood Prevention Comprehensive Plan in 2015 marked a transition from passive response to proactive planning in pluvial flood management. The plan established strict targets—limiting the annual runoff rate to 85% and requiring pipelines to meet at least a 5-year recurrence interval—and initiated short-term projects totaling CNY 900 million, including 64 km of stormwater pipelines and remediation of 44 rivers, thereby providing a solid foundation for resilience enhancement. Since 2016, sponge city construction has been advanced through the formulation of plans, technical standards, and leadership mechanisms. In 2018, the Kunming Municipal Regulations on Urban Drainage and Wastewater Treatment (2023) were enacted, mandating stormwater–sewage separation, low-impact development, and maintenance supervision, thus providing a legal basis for integrated engineering and institutional safeguards. In May 2022, Kunming was designated as a national demonstration city for sponge city construction. Through integrated flood prevention and Dianchi Lake governance, a comprehensive system was established featuring “upper interception, midstream diversion, downstream discharge, and low-elevation drainage”, along with stormwater–sewage separation projects. To date, 188.51 km² of sponge city areas have been completed, covering 160 of the 251 planned projects. In 2021, a 24 h collaboration mechanism was established between Kunming Drainage Company and the Meteorological Bureau, and a smart drainage system integrating online water-level monitoring was developed. Emergency plans were updated and mobile pumping capacity strengthened, reinforcing flood prevention. In 2022, the Implementation Guidelines for Dianchi Lake Three-Zone Management defined ecological zones, enforced “outflow-only” population policies, and prohibited new commercial residential development, thereby reducing density-driven flood risks. In 2024, Yunnan’s first hydrological digital twin project was approved, enabling real-time “four-preparation” flood control in the Panlong River Basin and reducing response times to the minute scale. This stage represents the integration of policy and technology, establishing a composite resilience system characterized by ecological retention, intelligent response, and spatial optimization.

5.1.2. Synergistic Effects of Ecological Foundations and Technological Innovation

In recent years, Kunming has adhered to the principle that “Dianchi Lake is for protection rather than development”. A suite of “water supplementation and pollution interception” measures has been adopted, including watershed diversion, pollution interception and treatment, and lakeside wetland construction. Measures such as “four retreats, three returns, and one protection”, the delineation of ecological red and yellow lines, and the “lake-to-city retreat” strategy have promoted green agriculture and reduced non-point source pollution. Through technology-enabled ecological restoration, the long-standing inferior Class V water quality of Dianchi Lake has been reversed. With improved governance, the lake’s regulation and purification capacity have been enhanced, supporting an increase in urban pluvial flood resilience from Level II in 2013 to Level IV in 2022. In 2022, the first phase of the hydrological digital twin project was launched in Kunming, piloted in the Panlong River Basin, with foundational models of river topography, remote sensing, and related parameters developed to enable “four-preparation” flood control and real-time visualization. By leveraging this platform, the impacts of heavy rainfall can be dynamically simulated and precise data provided, facilitating the advancement of smart water management. These digital initiatives have improved flood forecasting accuracy, strengthened early warning, and enhanced emergency deployment efficiency, thereby further reinforcing urban pluvial flood resilience.

5.1.3. The Kunming Paradigm of Urban Pluvial Flood Resilience in Mountainous Cities

Unlike the traditional drainage models of plain-based cities that rely on network density, a site-specific approach has been adopted in Kunming to accommodate its unique mountain–basin topography. Since 2013, an urban pluvial flood resilience system has been gradually developed, centered on “mountain flood interception—urban retention and regulation—lake buffering”. In the northern mountains and hilly areas, flood interception ditches and rainwater retention facilities have been constructed to capture upstream floodwater. In urban built-up zones, sponge city practices such as green roofs, sunken green spaces, and rain gardens have been applied to manage surface runoff. Dianchi Lake and its surrounding wetlands function as the ultimate ecological buffer, absorbing stormwater from the multi-level retention system. In 2017, the Kunming Sponge City Planning and Construction Management Measures were promulgated, explicitly integrating the principles of “infiltration, retention, storage, purification, utilization, and drainage” into urban renewal, old neighborhood renovation, and city expansion. This policy established a spatial coordination mechanism of “multiple-plans-in-one”. Sponge city development was also incorporated into municipal performance assessments and the river–lake chief responsibility system, thereby facilitating cross-departmental collaboration. From 2016 to 2020, while urban drainage safety was ensured, steady improvements in Dianchi Lake water quality and groundwater recharge were achieved, providing a “low-cost, high-resilience” reference model of pluvial flood governance for plateau and mountainous cities.

5.2. Strengths and Limitations

5.2.1. Strengths

(1)

Breakthrough in the Research Paradigm for Plain Cities

Existing studies on urban pluvial flood resilience in plain-based cities have largely focused on the prevailing logic of “engineering facility density and economic support capacity”, positing that drainage network density and GDP are the primary drivers of resilience improvement. However, through analyzing Kunming’s distinctive “plateau basin–mountain valley” geomorphology, this study demonstrates that ecological foundations play a more prominent role in shaping urban pluvial flood resilience. Kunming’s green coverage ratio (S4) increased from 39.4% in 2013 to 43.4% in 2022, and its water resource regulation and storage capacity (S5) improved by ~1.5% annually—together accounting for nearly 45% of the State dimension’s resilience indicator weight. The ecosystem, by enhancing stormwater retention capacity, substantially mitigates urban waterlogging risks arising from convergent topography. This study is the first to quantitatively identify the synergistic resilience pathway of “ecological retention–engineering drainage” in plateau and mountainous cities, proposing that the integration of retention and drainage constitutes a distinct mechanism—differing from the traditional drainage-dominated model. This finding enriches the literature—previously dominated by plain-based city studies—by revealing the differentiated effects of topographic heterogeneity on resilience drivers, and provides a new theoretical basis for pluvial flood resilience assessment in special geographic settings such as the southwestern plateau.

(2)

Application of the DPSIR Model

Compared to the traditional PSR model, featuring a linear “pressure—state—response” structure, the DPSIR model adopted in this study uses a chain logic of “driving force—pressure—state—impact—response”. This framework systematically captures the nonlinear feedback paths among social development, environmental change, and governance response within urban pluvial flood systems. For example, Kunming’s urbanization rate (D5) increased from 68.05% in 2013 to 81.1% in 2022, leading to a higher proportion of impermeable surfaces and a rise in population exposure (P3) from 2137 to 2639 people/km²—thus intensifying urban waterlogging risks. Concurrently, the proportion of excellent surface water quality (I4) declined to 34% in 2017, reflecting mounting pressure on the water environment. With subsequent accelerated sewage treatment facility construction, the wastewater treatment rate (I5) reached 99.46% by 2022, demonstrating a “development—risk—governance” closed-loop response mechanism. This model not only reveals the relationships between individual indicators but also identifies the dynamic linkages among “socio-economic development—natural system pressure—ecological feedback—governance response”, aligning with the social–ecological–engineering composite attributes of urban pluvial flood systems. For plateau cities facing the dual pressures of urbanization and ecological conservation, the DPSIR framework provides a robust tool for comprehensive analysis from risk sources to response targets and enjoys strong potential for broader application and replication.

5.2.2. Limitations

(1)

P7 Topographic and Geomorphological Characteristics as Constant

Treating P7 as constant fails to capture microtopographic disturbances from urban construction and can underestimate human influence on the relationship between topography and runoff. In Kunming’s plateau basin setting, rapid urbanization introduced cut and fill earthworks that changed local slopes, lake infilling around Dianchi that reduced depressional storage, and road and slope engineering that rerouted natural drainage and triggered a cascade from local ponding to rapid overland flow. These processes are not represented by a constant P7. Future analyses should parameterize P7 dynamically using time series remote sensing.

(2)

Public Response Capability Measured Only by Education

Assessing R3 solely by educational attainment, expressed as the share of residents with at least primary schooling, captures only information-reception potential and omits practice-oriented and behavioral capacities such as participation in disaster prevention drills, disaster risk awareness, and self-rescue skills. This design biases the assessment of the Response dimension. A composite R3 integrating basic, practical, and cognitive capacities is recommended.

5.3. Resilience Enhancement Measures

Based on the analysis results of this study, and taking into account Kunming’s geographical characteristics, current policy framework, and development needs, the following three targeted measures are proposed to effectively enhance the city’s urban pluvial flood resilience and its capacity to cope with heavy rainfall and flood disasters:

(1)

Strengthen Ecological Restoration and Water Quality Improvement in the Dianchi Lake Basin

The Dianchi Lake Basin serves as both Kunming’s water source and a critical ecological buffer. Enhancing its water quality and ecological environment is critical to boosting urban pluvial flood resilience. Priority should be given to wetland restoration and vegetation recovery in key areas—particularly the eastern and northern lake shores—to enhance the natural retention and storage capacity for pluvial floods. Wetland and vegetation restoration will enhance stormwater infiltration and retention, thereby reducing runoff and pollutant loads. Accelerating the construction and upgrading of sewage treatment facilities will improve treatment capacity, cut pollutant discharge, and avoid adverse impacts of water quality degradation on the ecosystem. Strengthening governance of major inflow rivers—including the Panlong and Jinzhihu Rivers—will also pursue comprehensive basin-wide water quality improvement.

(2)

Advance Smart Drainage System Construction and Improve Emergency Response Capacity

Modern urban drainage systems demand efficient, intelligent management and rapid emergency response mechanisms. Kunming should accelerate the development of a smart drainage system, integrating IoT, big data, and AI to enable real-time monitoring and management of drainage networks, rainwater retention facilities, and waterlogging hotspots. Intelligent early warning systems will enable prompt identification and mitigation of flood risks. Enhancing emergency response capacity requires refining flood emergency plans, conducting regular flood response drills, and raising public awareness of disaster prevention and mitigation. Strengthening disaster education in communities and schools will improve public participation in emergency management, thereby boosting the city’s overall resilience and flood control capacity.

(3)

Enhance Public Participation and Disaster Education

The improvement of urban pluvial flood resilience relies not only on engineering and technical measures but also on broad-based societal engagement. Kunming should enhance public disaster awareness and self-rescue capabilities via community drills and public education campaigns. Prioritizing regular dissemination of flood risk prevention knowledge via schools, media, and other platforms will foster a culture of disaster preparedness—particularly in high-risk areas. Expanding public training in emergency evacuation and risk identification will further enable residents to respond to and recover from flood disasters.

6. Conclusions

This study developed an urban pluvial flood resilience evaluation indicator system based on the DPSIR model, adopted a game theory-based aggregation weighting method, and constructed an urban pluvial flood resilience evaluation model using the extensional cloud model. Membership degree analysis yielded Kunming’s overall pluvial flood resilience levels from 2013 to 2022. The main conclusions are as follows:

(1)

The game theory-based aggregation weighting method indicated that among primary dimensions, the weight ranking is Driving Force dimension > Pressure dimension > State dimension > Response dimension > Impact dimension. Among secondary indicators, the top three are P6 Flood Disaster Risk, R1 Flood Disaster Early Warning Capability, and P7 Topographic and Geomorphological Characteristics—the most influential factors for urban pluvial flood resilience.

(2)

Extensional cloud model-based assessment of Kunming’s urban pluvial flood resilience (2013–2022) reveals a phased evolution from fluctuation to enhancement. Resilience improved significantly over the decade: 2013–2015 saw it stay at Level II (weak urban infrastructure and emergency response capabilities); 2016–2019 saw it rise to Level III, though 2017 pluvial flood-related extreme rainfall exposed gaps in post-disaster recovery capabilities; 2020–2022 saw it reach Level IV, driven by advances in disaster prevention and infrastructure upgrades. Despite 2021’s Response dimension resilience decline (due to the COVID-19 pandemic and extreme weather), the overall trend shifted from “moderate resilience” to “higher resilience”—reflecting substantial improvement.

(3)

Characteristic value analysis shows the resilience of each dimension in Kunming: The driving force dimension demonstrates continuous economic strengthening but increasing pressure on residents, highlighting challenges such as lagging income growth and unstable employment. The pressure dimension exhibits “structural vulnerability and extreme fluctuations”, as aging and population exposure heighten vulnerability, though child population resilience has improved. The state dimension shows gradual improvements in infrastructure and ecological retention but still faces resilience setbacks from extreme weather and outdated facilities. The impact dimension reflects gradual progress in post-disaster recovery; although significant losses occurred during the 2017 storms, subsequent flood control and environmental measures have steadily reduced disaster impacts. The response dimension has benefited from technological empowerment, especially in early warning and emergency management, yet still faces challenges in system coordination during sudden events. Overall, resilience across all dimensions in Kunming shows an upward trend, indicating phased progress in urban pluvial flood management, while further enhancement is still needed in emergency response, infrastructure maintenance, and the protection of vulnerable groups.