Formation Mechanism and Response Strategies for Urban Waterlogging: A Comprehensive Review

Abstract

With the intensification of climate change and the continuous advancement of urbanization, the pressure on urban drainage systems has increased, leading to the growing prominence of urban waterlogging issues. Besides the destruction of infrastructure, urban waterlogging also affects environmental quality, economy, and residents’ daily lives. Researchers have recently analyzed the causes of urban waterlogging from multiple perspectives, including land-use changes driven by urbanization, the inadequacy of urban drainage systems, and extreme rainfall events resulting from climate change. Various strategies have been proposed to address waterlogging, including optimizing urban green spaces, establishing forecasting systems, and creating effective emergency management systems. Additionally, some scholars highlight the significance of integrated urban planning and interdepartmental collaboration, suggesting that multi-party cooperation can help mitigate the risks of waterlogging. This paper conducts a comprehensive literature review to summarize the current research status of urban waterlogging, focusing on theoretical, experimental, numerical simulation, and artificial intelligence approaches. The review aims to provide a clearer understanding of the existing knowledge, identify gaps for future research and propose ideas that combine advanced technologies and interdisciplinary approaches.

1. Introduction

Waterlogging refers to the phenomenon of heavy rainfall or continuous precipitation exceeding the drainage capacity of the city, resulting in water accumulation disasters in the city. As the pace of urbanization accelerates and climate change intensifies [1,2], urban waterlogging issues are becoming increasingly severe, posing significant challenges for numerous cities worldwide. Waterlogging not only threatens urban infrastructure and public safety but also raises new demands for urban economic development and resilience building. For instance, waterlogging events can lead to traffic paralysis and disruptions to commercial activities, consequently affecting residents’ quality of life [3]. Furthermore, the environmental impacts of waterlogging cannot be overlooked, as they may lead to water pollution and the destruction of biological habitats. Previously, Berndtsson et al. [4] proposed a framework to analyze the driving factors of urban flood risk variation, emphasizing the need to develop long-term policies to address high-impact but low-manageable driving factors. Therefore, understanding the mechanisms and management strategies of urban waterlogging is crucial for enhancing urban resilience and achieving sustainable urban development.

Heavy rainfall or extreme precipitation events within a short time often exceed the design capacity of urban drainage systems, making it difficult for rainwater to be discharged quickly and leading to water accumulation. Simultaneously, the accelerated urbanization process has increased the area of impervious surfaces, reducing natural infiltration and further exacerbating runoff. Additionally, the aging of urban drainage systems, inadequate design, and poor maintenance are also significant reasons for the frequent occurrence of waterlogging. The characteristics of urban microtopography, sunken areas, and unreasonable layout of drainage facilities can lead to water accumulation in specific regions. The combination of multiple factors makes urban waterlogging increasingly complex, affecting not only transportation and daily life but also potentially causing significant losses to urban economic activities, highlighting the importance of researching urban waterlogging management.

Waterlogging mainly occurs in urban areas, usually caused by heavy rain or drainage system issues. It is related to hydrological disasters such as flooding, groundwater rise, and coastal inundation, focusing on the problem of water accumulation within the city.

Currently, research on urban waterlogging mainly focuses on four aspects: theoretical analysis, experimental research, numerical simulation, and the application of artificial intelligence technology. At the theoretical level, researchers have deeply analyzed the formation mechanisms of waterlogging, revealing the complex interactions among precipitation, topography, land use, and drainage systems. Experimental research explores the mechanisms of waterlogging on urban streets, underground drainage systems, and buildings through physical models and hydrodynamic experiments. For example, Velickovic M et al. [5] conducted an idealized experiment using square blocks to represent buildings, exploring the effects of street width and direction on flow. Numerical simulation, as a key tool, utilizes one-dimensional, two-dimensional, and three-dimensional coupled models to predict and assess the processes and spatial distribution characteristics of urban waterlogging [6,7,8,9,10,11]. The rapid development of artificial intelligence has breathed new life into waterlogging research; through deep learning, reinforcement learning, and data mining technologies, it has not only improved the accuracy of waterlogging predictions but also promoted the intelligent processes of emergency response and drainage scheduling [12,13,14,15,16]. Existing studies have revealed the complexity of urban waterlogging formation, indicating that urbanization processes, rainfall intensity, drainage system capacity, and microtopographic features are key factors influencing waterlogging.

Despite significant progress in research on urban waterlogging, current analyses still face numerous challenges. For instance, the dynamic characteristics of waterlogging involve complex interactions between natural and social systems, and traditional models have limitations when handling high-resolution data and dynamically coupled scenarios. Furthermore, how to efficiently integrate multidisciplinary technologies (such as hydrological modeling, artificial intelligence, and remote sensing) to meet the prediction demands of extreme rainfall events remains a critical issue that needs to be addressed. The research objectives include the following: (i) Reviewing the current theoretical research results on urban waterlogging, analyzing the formation mechanisms of waterlogging, and revealing the impact of multi-factor coupling on waterlogging. (ii) Reviewing experimental studies on urban waterlogging, analyzing experimental methods related to waterlogging, systematically summarizing the application status of physical models, scaled experiments, and hydrodynamic experiments, and discussing the critical role of experimental research in validating numerical models, optimizing drainage system design, and revealing the mechanisms of waterlogging damage. (iii) Evaluating numerical simulation methods for urban waterlogging: clarifying the theoretical support role of numerical simulation in urban waterlogging risk assessment and management, and proposing model frameworks to address increasingly complex waterlogging issues. (iv) Exploring the application of artificial intelligence in urban waterlogging research, particularly the use of algorithms such as deep learning and reinforcement learning for data analysis and risk assessment. (v) Discussing potential strategies to enhance urban disaster resilience through the integration of existing technologies.

The structure of this paper is as follows: The second Section presents the research methodology. The third Section introduces the relevant theories of urban waterlogging, the current status of experimental research on urban waterlogging and the latest advancements in numerical simulation in complex waterlogging scenarios. The fourth Section analyzes the innovative applications of artificial intelligence technologies in waterlogging research, highlighting the advantages of methods such as deep learning, random forests, and reinforcement learning in waterlogging prediction and risk assessment. Finally, the fifth Section concludes the research findings and proposes directions for future research. The flowchart is shown in Figure 1.

2. Research Methodology

This comprehensive review employs a systematic methodology to explore the formation mechanisms, influencing factors, prediction models, risk assessment, and response strategies for urban waterlogging. The research integrates theoretical analysis, experimental studies, numerical simulations, and artificial intelligence (AI) applications to provide a holistic understanding of urban waterlogging dynamics and management approaches. The methodology is structured into two main phases: literature retrieval and selection and methodological framework development, each designed to ensure rigor and comprehensiveness [17].

2.1. Phase 1: Literature Retrieval and Selection

To capture the breadth of urban waterlogging research, a thorough literature search was conducted across multiple academic databases, including Scopus [18], Science, IEEE Xplore, and SpringerLink, covering publications from 1990 to 2024. This timeframe was chosen to reflect significant advancements in urban waterlogging research, particularly in the integration of emerging technologies such as AI and advanced numerical modeling. Journals spanning disciplines like environmental sciences, civil engineering, urban planning, and climate change were prioritized to capture interdisciplinary perspectives.

The search utilized keywords such as “urban waterlogging”, “flood risk assessment”, “drainage systems”, and “artificial intelligence” to identify relevant studies. Inclusion criteria focused on peer-reviewed journal articles, conference papers, and technical reports to ensure academic rigor, while book chapters and non-peer-reviewed materials were excluded. After deduplication and screening, 115 high-quality articles were selected for in-depth analysis.

2.2. Phase 2: Methodological Framework

The analysis was organized into three interconnected stages to systematically map the research landscape.

(a)

Literature Analysis

The initial stage focused on categorizing the selected literature to identify overarching trends. This included examining the temporal distribution of publications to track evolving research priorities, geographical patterns in study locations, and methodological approaches such as experimental, numerical, or AI-driven techniques [15]. Key themes like theoretical frameworks, drainage system optimization, and climate change impacts were also highlighted.

(b)

Descriptive Analysis

Based on the initial categorization, a detailed overview of the literature was provided at this stage. Through statistical methods, we obtained the distribution of articles in terms of year of publication and research methodology. It can be seen that the selected articles were mainly distributed in the last decade and various research methods were involved.

(c)

Scientometric Analysis

To visualize the scientific landscape, we used VOSviewer software (version 1.6.20) to perform the scientometric analysis [19]. The software created a distance-based visual representation of the network based on the proximity of any two nodes. In the scientometric analysis, key areas of urban flooding research, AI approaches in flooding research, and applications of AI in flooding research and mitigation were identified using this software.

3. Analytical Framework

This Section systematically integrates theoretical, experimental, and numerical approaches to investigate urban waterlogging mechanisms. The focuses of research on urban waterlogging are presented in Figure 2. Theoretical analysis establishes the foundational understanding of waterlogging dynamics, focusing on the multi-factor coupling mechanisms involved in its formation, including the impacts of precipitation intensity, land-use changes, urban drainage system capacity, and climate change. Experimental research validates theoretical hypotheses through scaled physical models and hydrodynamic tests, capturing flow behaviors under controlled scenarios to refine urban morphology and infrastructure design, etc., and validating some of the theoretical research [20,21,22,23]. Numerical modeling bridges theory and practice by simulating flood propagation, risk distribution, and mitigation strategies using advanced computational tools [10,24,25,26]. The percentage of citations for each section is shown in Figure 3. Together, this tripartite framework enables a comprehensive assessment of waterlogging processes, offering insights for urban resilience planning. The distribution of research contents is shown in Figure 4.

3.1. Theoretical Research

The issue of urban waterlogging is becoming increasingly severe, and with the acceleration of urbanization, both the frequency and intensity of urban waterlogging are rising steadily. Relevant scholars have primarily focused their theoretical research on urban waterlogging mechanisms [27,28,29,30,31,32,33,34], prediction models [35,36,37,38,39], risk assessment, and response strategies [28,40,41,42,43,44,45]. The following Sections will summarize the relevant theoretical research from these aspects.

3.1.1. Mechanisms and Influencing Factors of Waterlogging

The mechanisms are complex, involving various dynamic interactions, and are closely related to the process of urbanization. According to existing studies, the formation mechanisms of waterlogging can be analyzed from aspects such as land-use changes and drainage system capacity.

During urbanization, changes in land use, such as the increase in impermeable surfaces, have significantly altered regional hydrological characteristics, leading to an increased risk of waterlogging. Weng et al. [27] combined remote sensing and GIS technology to study the impact of urban growth on surface runoff in the Pearl River Delta. Their findings indicated that from 1989 to 1997, the annual runoff depth in the Pearl River Delta increased by 8.10 mm, exacerbating the risks in the region. Muis et al. [28] found that urban expansion significantly increases the risk of urban waterlogging, emphasizing the importance of adaptive measures and spatial planning. Guillén et al. [29] demonstrated that land occupancy has a substantial impact on water flow resistance, quantifying the effect of increased obstacles on resistance coefficients. In highly urbanized areas, dense buildings lead to increased flow resistance coefficients and reduced water flow speeds, further worsening waterlogging conditions. Changes in flow characteristics may cause rainwater accumulation, especially in areas with insufficient or poorly designed drainage facilities. This phenomenon is particularly pronounced in urbanized regions. The mechanisms of urban waterlogging caused by rainfall and flooding factors are illustrated in Figure 5. Zhang et al. [30] and Ning et al. [46] analyzed the impacts of global climate change and urbanization on urban waterlogging in China, proposing prevention and control strategies, such as low-impact development and sponge city construction.

The increase in waterlogging events is not only influenced by urban expansion but is also a result of climate change. Ji et al. [31] assessed the impact of climate change on urban waterlogging in Beijing, with results indicating that waterlogging will worsen, emphasizing the need to consider the uncertainties of climate change in the design of prevention and control measures. Zhang et al. [32] used the regional climate model RegCM4.6 and the annual maximum method to simulate rainfall changes, concluding that climate change will exacerbate waterlogging and recommending that the design of urban drainage systems take extreme rainfall into account. Additionally, factors influencing waterlogging include the coverage and layout patterns of green infrastructure. Chang et al. [33] analyzed urban governance documents, noting that cities are gradually adopting socio-ecological elements, such as floodplain restoration and green infrastructure, to enhance resilience against future extreme floods. Fahy et al. [34] studied the effects of green stormwater infrastructure on reducing urban rainwater runoff and found that distributed green infrastructure systems are more effective than centralized structures, significantly lowering runoff ratios, peak flow rates, and flicker indices, thereby providing new ideas for urban waterlogging management.

The mechanisms of urban waterlogging formation are summarized in Figure 6. The occurrence of waterlogging is the result of multiple factors. Management and adaptive measures addressing these factors should be integral components of urban planning and disaster prevention strategies to mitigate the risks and losses caused by waterlogging.

3.1.2. Waterlogging Prediction Models

Predicting the occurrence and development of urban waterlogging is an important direction in theoretical research. In recent years, with the acceleration of urbanization and the intensification of climate change, traditional hydrological models and laboratory simulation methods have faced increasing challenges. Researchers have enhanced the accuracy and practicality of waterlogging predictions by analyzing and refining existing models, and by integrating various advanced technologies [48,49,50].

Waterlogging prediction models simulate and analyze urban waterlogging phenomena, trying to find out the impacts of factors such as rainfall, runoff, and model distortion to provide scientific evidence and effective governance plans, thereby strengthening urban resilience against waterlogging. Cea et al. [35] directly applied rainfall intensity to models for experimental validation. Mignot et al. [36] reviewed 45 laboratory-based urban flood studies, categorizing and discussing different flow processes and recorded flow variables. By identifying existing datasets and research gaps, they contributed to the application of computational and experimental models in urban flood research. Additionally, Li et al. [37] assessed the impact of proportional distortion in urban flood experimental models on water depth and flow, indicating that within a reasonable range of distortion, experimental models can still provide effective references for waterlogging predictions when simulating small-scale physical processes of urban flooding.

Remote sensing technology and Geographic Information Systems (GIS) are vital tools in modern urban waterlogging prediction. Weng et al. [27] proposed a model combining remote sensing and GIS to analyze the impact of urbanization on runoff in the Pearl River Delta, demonstrating how urbanization exacerbates urban flood risks. High-resolution data enable more precise urban flood modeling. Dottori et al. [38] explored how to improve flood modeling using high-resolution data and reviewed methods such as sub-grid parameterization and roughness upsampling to balance efficiency and reliability, thereby enhancing the reliability of flood predictions. Muste et al. [39] reviewed the application of large-scale particle image velocimetry technology in river environments, improving the simulation capacity for complex hydrological processes and providing new insights for urban waterlogging prediction models, which are beneficial for capturing the interactions of different factors.

As the impacts of urbanization and climate change on waterlogging become increasingly complex, researchers have adopted ensemble modeling methods to assess waterlogging risks. Noh et al. [42] proposed an ensemble method for urban flood modeling, considering the uncertainty of parameters related to wells, sewage pipes, and surface flow. Their results indicated that ensemble simulations based on overflow and orifice formulas can accurately reproduce experimental data. Ravazzani et al. [43] combined the FEST-WB and WRF models to test two hydrological integrated forecasting systems, finding that optimizing integration strategies helps improve flood warning efficiency. Rahmati et al. [51] hybrid integrated two novel models (Wavelet-SVR-Bat and Wavelet-SVR-GWO) to assess spatial modeling of urban flood inundation, with these hybrid models significantly outperforming traditional support vector regression in performance. Figueiredo et al. [52] proposed using a multi-model ensemble approach to improve flood loss modeling, evaluating 20 flood loss models and demonstrating that a well-designed multi-model ensemble can provide more accurate loss estimates and reliable probability distributions of uncertainty.

By combining multiple models, researchers not only consider the predictions of waterlogging itself but also integrate various factors such as urban layout and drainage conditions, providing more reliable prediction results.

3.1.3. Waterlogging Risk Assessment and Management

The risk assessment of urban waterlogging is an important step in responding to extreme weather events and the urbanization process. To achieve a comprehensive assessment, it is essential to integrate factors such as climate change, urban expansion, and hydrological characteristics, employing a holistic risk analysis approach. As urbanization accelerates, the issue of waterlogging becomes increasingly complex, necessitating a thorough evaluation of various influencing factors and the formulation of management strategies [4].

Waterlogging risk assessment relies not only on traditional analytical management models but also on the integration of multiple data sources and technical means to ensure more accurate early warnings and decision support. Da Silva et al. [40] systematically reviewed 52 papers on flood risk management and found a lack of formal methods for climate modeling and the use of GIS tools. They proposed decision-making guidelines to support the effective implementation of urban flood management, providing new technological pathways for waterlogging risk assessment. Chen et al. [41] proposed an urban flood risk early warning method based on a multi-indicator fuzzy assessment using the DPSIR model, incorporating five assessment indicators. They validated the effectiveness of the urban flood forecasting model using Dongguan City as a case study. Additionally, Muis et al. [28] introduced a probabilistic analysis method through global-scale flood hazard and land change modeling, systematically assessing the impacts of climate change and urban expansion on waterlogging risk. Their research indicated that as urbanization intensifies, both the frequency and intensity of waterlogging events significantly increase.

With the intensification of global climate change, the uncertainty of waterlogging risk will be greater in the future. To address these challenges, waterlogging management strategies need to be more flexible and adaptive. Da Silva et al. [44] explored the impacts of climate change and urbanization on flood risk management, introducing climate prediction models and their effects on rainfall patterns. They provided long-term decision-making guidance for decision-makers in a changing climate through multi-attribute utility theory and multi-dimensional decision analysis models. Furthermore, Duan et al. [45] proposed an urban waterlogging risk assessment method that combines multi-criteria decision analysis (MCDA) with geographic information systems (GIS), using Changchun City as a case study. By calculating weights through the Analytic Hierarchy Process (AHP), their results indicated that this model is more reliable for waterlogging risk assessment. This multi-dimensional perspective aids in formulating more comprehensive urban flood prevention and emergency response strategies.

Urban waterlogging risk assessment and management is a complex issue interwoven with multiple layers and factors. By integrating advanced technical means (such as hydrological models, GIS, remote sensing data, etc.) with green infrastructure, it is possible to mitigate the occurrence and losses associated with urban waterlogging. As climate change and urbanization accelerate, future waterlogging risk management will need to emphasize adaptability and flexibility, establishing scientific assessment and forecasting systems to respond to increasingly complex risk environments.

3.2. Experimental Research

Experimental research is a crucial approach for investigating urban waterlogging phenomena and their impacts. It enables researchers to validate various hydraulic models and the effects of urban morphology on flood flow and inundation depth through simulations and actual data [53,54,55,56,57,58]. By employing a range of experimental techniques, researchers can uncover the specific mechanisms behind urban flooding and summarize effective prevention and control strategies.

3.2.1. Experimental Design

Experimental research is a crucial method for investigating the mechanisms of urban waterlogging, primarily through scaled models, physical experiments, and parameter measurements to simulate real flood scenarios, revealing the complex relationship between fluid behavior and urban morphology.

Most studies employ scaled models to simulate urban waterlogging scenarios. For instance, Finaud-Guyot et al. [53] utilized a 1:200 scaled urban geometry experimental setup to measure hydraulic variables under different flood conditions. Tomiczek et al. [54] conducted wave experiments using a 1:20 scaled model to measure water surface heights and extreme pressures around idealized structures. Sturm et al. [59] measured the impact forces of flood events on buildings through a 1:30 scaled model experiment.

To enhance the accuracy of flow and water level monitoring experiments, many studies have integrated high-resolution monitoring technologies. Legout et al. [55] employed large-scale particle image velocimetry (LSPIV) and laser scanning techniques to obtain spatial distribution data of surface water flow depth and velocity, providing a high-precision basis for flow calculations and demonstrating strong robustness. In response to urban flood flow fields caused by extreme rainfall or levee breaches, LaRocque et al. [56] recorded flow depth and velocity using point flow meters and ultrasonic velocity profilers, with their dataset available for numerical model validation.

The details of experimental research are shown in

Table 1. Experimental research methods for waterlogging.

Reference	Year and Analysis	Experimental Methods	Contents
Finaud-Guyot et al. [53]	2018 Experimental insight for flood flow repartition in urban areas	Scaled Experiment	Measure hydraulic variables under different flood conditions by using a 1:200 scale urban geometric experimental setup.
Martínez-Gomariz et al. [60]	2016 Experimental study of the stability of pedestrians exposed to urban pluvial flooding	Prototype Experiment	Investigate the stability limits of pedestrians in rapidly flowing water by using a full-scale model.
Liu L et al. [61]	2018 Building performance in dam-break flow—an experimental study	Non-Intrusive Measurement	Employ non-intrusive pressure and ultrasonic measurements to explore the impact of dam breaches on residential structures.
Leandro J et al. [58]	2010 Experimental Scaled model as a benchmark for validation of Urban Flood models	Fluid Dynamics Experiment	Collect flow rate and water level distribution to validate the correlation of the inlet discharge coefficient.
Legout et al. [55]	2012 High spatial resolution mapping of surface velocities and depths for shallow overland flow	Field Monitoring	Combine large-scale Particle Image Velocimetry (PIV) and laser scanning technology to measure water flow depth and surface velocity over complex terrain.

. These experimental design methods enable researchers to simulate urban waterlogging under both normal and extreme conditions in the laboratory, providing essential experimental evidence for a deeper understanding of urban hydrological responses. Selecting appropriate experimental design strategies helps improve the accuracy of experimental results, making them closer to actual conditions, thereby providing valuable scientific support for urban planning and flood prevention measures.

3.2.2. Model Applications

The application of experimental models plays a crucial role in urban flooding research, often used to predict flood dynamics, assess flood risks, and support urban planning.

Cea et al. [35] conducted experimental validation of two widely used urban flood inundation numerical models, applying rainfall intensity and measuring surface runoff under various rainfall conditions. The experimental results demonstrated the advantages of the dynamic wave model in predicting peak flow and hydrological response. Martins et al. [62] further validated the applicability of the finite volume model under the influence of sewer overflow, confirming the relationship between model sensitivity and overflow ratio, and proving that the finite volume-based model can effectively simulate the local flow interactions between the sewer system and floodplain. Chen et al. [63] explored the sensitivity of 1D and 2D models to input parameters by combining experimental data with variance decomposition methods. The study revealed that upstream inflow and roughness significantly affect water level changes, while downstream water levels are entirely determined by boundary conditions, providing a solid foundation for model parameter calibration.

In response to complex scenarios such as intricate terrain, multi-branch flow fields, and areas of air concentration, researchers have developed more adaptable models. Mignot et al. [64] analyzed the velocity distribution at sewer junctions, assessing typical errors when calculating flow near junctions, and proposed that the local flow structure at junctions significantly affects the accuracy of flow measurements, offering guidance for practical engineering design. Kim et al. [65] investigated the impact of urban heterogeneous porosity conditions on flood model errors, finding that heterogeneous porosity models outperform traditional models in overall predictive accuracy, although further enhancement of local flow characteristics is still needed. Lopes et al. [66] validated an air entrainment model based on the volume of fluid method through experimental data, combining it with the k-ω turbulence model to simulate air entrainment phenomena within ditches, confirming that this model demonstrates good accuracy in predicting areas of air concentration.

3.2.3. Hydrological and Hydrodynamic Experiments

Hydrological and hydrodynamic experiments are essential tools for exploring the causes and evolution of urban flooding. In recent years, they have achieved widespread application and significant progress in simulating flood behavior, validating numerical models, and optimizing drainage design.

The complexity and diversity of urban morphology lead to highly nonlinear characteristics of flood flow patterns, where factors such as the arrangement of street networks and the size of gaps between buildings play crucial roles in determining the speed and extent of flood propagation [67]. Urban morphology significantly influences flow characteristics during flooding events. Li et al. [68] analyzed the impact of urban morphology on urban flood flow variables through experimental and computational modeling, revealing that the conductivity porosity of the primary flow direction significantly affects flood severity, and proposed important considerations for the design of new projects in flood-prone areas. Nanía et al. [69] experimentally investigated the diversion phenomenon at steep street intersections by establishing a scaled model of the intersection, finding that the inflow power ratio at the entrance can serve as a dimensionless parameter to predict flow distribution and patterns, thus developing a one-dimensional formula to forecast flow diversion at intersections.

Streets and drainage facilities play important roles in the flow behavior of urban flooding, and studying their influence mechanisms provides significant theoretical support and practical guidance for optimizing urban drainage system design. Nanía et al. [70] investigated the subcritical flow diversion phenomenon at an intersection with two inflows and two outflows of equal width, proposing and validating a linear relationship among five dimensionless parameters to successfully predict flow distribution at the intersection. Rubinato et al. [71] focused on the interaction between sewage pipelines and floodplain flow, analyzing flow exchange under steady and unsteady flow conditions through physical model experiments. They validated the effectiveness of existing formulas and proposed a new drainage coefficient, providing improved tools for optimizing urban drainage network design.

Research on the interaction between buildings and floodwaters emphasizes the characteristics of impact forces and flow field distribution. Liu et al. [61] investigated the impact of floodwaters from dam breaches on buildings through physical model experiments, finding that when the flood wave reaches the house, the upstream water level rises rapidly and oscillates, with significant effects from the orientation and distance of the house. Tomiczek et al. [54] measured the water surface height and extreme pressures around idealized structures during wave experiments, validating the effectiveness of shielding structures in reducing impact pressures, and providing a basis for optimizing urban flood prevention design. Sturm et al. [59] measured the impact forces on buildings during flood events through model experiments, analyzing factors affecting impact forces, concluding that flow height is positively correlated with impact forces, and clarifying the influence of different flow patterns on impact forces. Furthermore, Martínez-Gomariz et al. [60] expanded the experimental perspective to the field of pedestrian safety in urban flood environments, simulating human stability experiments under different flood conditions, identifying critical thresholds for pedestrian stability under combinations of shallow water depths and high flow velocities, and providing experimental evidence for flood risk assessment and urban disaster prevention and mitigation planning.

It is recommended to conduct a more in-depth analysis of experimental results under different urban conditions, focusing on the impacts of urban morphology, drainage systems, and topographical features. For instance, high-density urban areas as a priority may need to emphasize the flow characteristics and the space between buildings to mitigate flood propagation. While in low-lying areas, attention should be given to the design of drainage systems and water flow distribution. Such analyses will provide targeted recommendations for urban flood prevention strategies to assist decision-makers in formulating more effective measures.

3.3. Numerical Simulation

Numerical simulation plays a crucial role in the study of urban flooding, effectively predicting flood behavior and providing a theoretical foundation for urban planning and disaster prevention and mitigation [1,72,73,74]. In recent years, with advancements in computational power and simulation accuracy, numerical models have become indispensable tools for analyzing urban flooding, addressing various aspects from rainfall simulation and flow simulation to flood damage assessment [8,47,75,76,77,78]. Despite significant progress in inundation simulation and risk assessment with existing models, parameters’ sensitivity, accuracy, and computational efficiency still require further optimization. Paquier et al. [76] indicated that rainfall distribution and urban representation methods significantly impact simulation results, and both water levels and urban representation techniques introduce substantial uncertainties into flood flow patterns. Since urban flooding is primarily caused by heavy rainfall and issues with drainage systems, our discussion is divided into two parts: hydrodynamic models and simulation models.

3.3.1. Hydrodynamic Models

Hydrodynamic models are mathematical tools that describe the flow characteristics and behavior of water bodies, primarily used to simulate the spatiotemporal distribution of flood flow, velocity, and water depth. Depending on the dimensionality and complexity, hydrodynamic models can be categorized into one-dimensional (1D), two-dimensional (2D), and three-dimensional (3D) models.

One-dimensional models are the earliest type used for flood simulation, primarily based on the Saint-Venant equations to describe water flow along river channels. One-dimensional models are suitable for modeling flow in river networks [62]. Luo et al. [75] proposed an improved 1D nonlinear dynamic model, finding that the approximation provided by the 1D model significantly deviates from the actual flow field. One-dimensional models are suitable for long-distance simulations of large water systems but perform inadequately in capturing lateral flow and local turbulence characteristics. This limitation motivated the adoption of 2D/3D frameworks, calibrated using experimental LSPIV velocity datasets to enhance local turbulence prediction.

Two-dimensional models describe the planar motion of water flow by solving the shallow water equations, allowing for better representation of flood propagation characteristics in plains or urban areas. They are commonly used for studying urban flooding and effectively simulate changes in flow and water depth during inundation events. For instance, Bruwier et al. [8] used an improved anisotropic porosity model to analyze the impact of urban morphological features on flood flow, revealing a significant correlation between urban characteristics and flood depth. Paquier et al. [76] employed a two-dimensional depth-averaged model to simulate urban flooding in Marseille and Ulin, concluding that the spatial distribution of rainfall input significantly affects water depth.

Three-dimensional models (3D) extend into the vertical direction, simulating the three-dimensional movement characteristics of water flow by solving the Navier–Stokes equations. Three-dimensional hydrodynamic models have also been applied to more complex urban flow conditions. For example, Luo et al. [75] compared 1D and 3D numerical models to analyze flow characteristics at the junctions of open channels, finding that traditional 1D models may not accurately reproduce flow field characteristics in certain cases, and parameterizing 3D features can enhance the accuracy of large-scale water network models. Different types of models and related studies are shown in Table 2.

Using models in isolation has its limitations [79]. Thus, the combined application of different models can leverage the advantages of each dimension model, such as coupling 1D and 2D models. Luan et al. [47] employed a dynamic bidirectional coupled model, integrating 1D–2D hydrodynamic and water quality simulations to explore the formation process of urban waterlogging in Changzhi City, China. Li et al. [77] proposed a dynamic 1D–2D coupled model to simulate urban flooding, with results indicating that the physically-based 2D overflow diffusion model outperforms geological statistical models, suggesting their combined use. Naves et al. [80] investigated the application of Structure from Motion (SfM) photogrammetry technology for high-resolution topographic surveys in urban drainage physical models to simulate runoff under different rainfall intensities, with results showing that the 2D model using SfM topographic data outperformed traditional measurement methods in replicating experimental data.

Table 2. Numerical models for waterlogging research.

Reference	Year and Analysis	Numerical Models	Contents
Chen et al. [41]	2015 Urban flood risk warning under rapid urbanization	Prediction Model	Establish an urban flood prediction model and propose a multi-indicator fuzzy assessment early warning method based on the DPSIR model for flood risk warning in rapidly urbanizing areas.
da Silva et al. [81]	2022 A novel spatiotemporal multi-attribute method for assessing flood risks in urban spaces	Decision Model	Combine climate change and population growth scenarios for flood risk perception and optimized management decision-making.
Liang et al. [82]	2021 Modeling the dynamics and walking accessibility of urban open spaces under various policy scenarios	Simulation Model	Integrate cellular automata and system dynamics to simulate the dynamic development of open spaces under different construction time lag scenarios.
Luu et al. [47]	2021 GIS-based ensemble computational models for flood susceptibility prediction in the Quang Binh Province, Vietnam	Prediction Model	Develop a flood susceptibility prediction model using the PART classifier and various ensemble learning techniques.
Bruwier et al. [1]	2020 Influence of urban forms on surface flow in urban pluvial flooding	Surface Flow Model	Use an efficient porosity-based surface flow model to calculate and analyze the impact of nine urban characteristics on surface flow during heavy rainfall events.
Beg et al. [7]	2020 CFD modeling of the transport of soluble pollutants from sewer networks to surface flows during urban floods	Numerical Model	Establish a three-dimensional CFD model to quantitatively analyze the interaction between pipeline flow and surface flow, as well as solute transport.
Galuppini et al. [83]	2020 A unified framework for the assessment of multiple source urban flash flood hazard	Unified Urban Flash Flood Model	Analyze the flood risk in the city of Monza and develop a unified urban flash flood model to capture flood sources from rivers and drainage systems.

Table 2. Numerical models for waterlogging research.

Reference	Year and Analysis	Numerical Models	Contents
Chen et al. [41]	2015 Urban flood risk warning under rapid urbanization	Prediction Model	Establish an urban flood prediction model and propose a multi-indicator fuzzy assessment early warning method based on the DPSIR model for flood risk warning in rapidly urbanizing areas.
da Silva et al. [81]	2022 A novel spatiotemporal multi-attribute method for assessing flood risks in urban spaces	Decision Model	Combine climate change and population growth scenarios for flood risk perception and optimized management decision-making.
Liang et al. [82]	2021 Modeling the dynamics and walking accessibility of urban open spaces under various policy scenarios	Simulation Model	Integrate cellular automata and system dynamics to simulate the dynamic development of open spaces under different construction time lag scenarios.
Luu et al. [47]	2021 GIS-based ensemble computational models for flood susceptibility prediction in the Quang Binh Province, Vietnam	Prediction Model	Develop a flood susceptibility prediction model using the PART classifier and various ensemble learning techniques.
Bruwier et al. [1]	2020 Influence of urban forms on surface flow in urban pluvial flooding	Surface Flow Model	Use an efficient porosity-based surface flow model to calculate and analyze the impact of nine urban characteristics on surface flow during heavy rainfall events.
Beg et al. [7]	2020 CFD modeling of the transport of soluble pollutants from sewer networks to surface flows during urban floods	Numerical Model	Establish a three-dimensional CFD model to quantitatively analyze the interaction between pipeline flow and surface flow, as well as solute transport.
Galuppini et al. [83]	2020 A unified framework for the assessment of multiple source urban flash flood hazard	Unified Urban Flash Flood Model	Analyze the flood risk in the city of Monza and develop a unified urban flash flood model to capture flood sources from rivers and drainage systems.

3.3.2. Simulation of Urban Flooding

The impact of urban morphology on flooding has been a key research focus in recent years, with numerous studies utilizing numerical simulations to explore the role of urban characteristics on flood flow [84,85,86]. Bruwier et al. [8] generated 2000 synthetic urban patterns and utilized an improved anisotropic porosity model, revealing a significant relationship between urban morphology and flood depth. El Kadi et al. [87] simulated flood propagation in urban areas using a two-dimensional shallow water model, finding that buildings have a notable impact on both flood depth and wave velocity. Du et al. [88] assessed the role of concave green spaces in mitigating urban flooding in Shanghai through urban flood simulations and scenario analyses, indicating that concave green spaces can significantly reduce runoff and inundation areas, thus enhancing community resilience. Mustafa et al. [89] simulated flood risks under various urban expansion policies, discovering that strict control of development can significantly alleviate flood impacts, while excessive densification may exacerbate risks. Macchione et al. [90] proposed a method for reconstructing the hydrodynamic dynamics of urban flooding events using non-traditional information sources, such as videos, photographs, and news reports. By combining the data with hydrodynamic models, they successfully reconstructed a numerical model of the flooding event in Crotone, Italy, in 1996.

Numerical models not only simulate the effects of flooding on urban areas but also evaluate flood risks and potential damage under different scenarios. Dottori et al. [91] proposed a synthetic flood damage model based on component-wise analysis, designing damage functions by combining existing literature and expert opinions. This model was validated using flood events in northern Italy, with results indicating that its damage estimates are comparable to or better than existing models. Additionally, the multi-criteria fuzzy evaluation and early warning method (DMFEW) based on the DPSIR framework proposed by Chen et al. [41] provides effective support for flood risk management in rapidly urbanizing areas. This method combines multiple indicators to more accurately predict and assess urban flooding risks, thereby providing decision-making support for urban managers.

Conducting in-depth analysis of flood behavior under different urban conditions using numerical simulation methods is essential. For old urban areas, emphasis should be placed on the impact of building structures and urban morphology on flood flow and water accumulation depth, utilizing three-dimensional hydrodynamic models or 1D–2D-coupled models to enhance the accuracy of flood simulations. In newly developed areas, particularly those where rainwater management is crucial, GIS models could be used to assess the impact of urban planning schemes on flood risk. Such analyses will provide scientific evidence for urban planners to optimize flood prevention strategies.

4. AI Methods for Urban Waterlogging

In recent years, artificial intelligence (AI) methods have played an increasingly important role in urban waterlogging management. These methods leverage technologies such as machine learning, deep learning, and data analysis to predict, monitor, assess, and manage urban flooding, providing more precise and efficient solutions [92,93,94,95,96,97,98]. The application of AI has been widely explored in existing research to enhance emergency response capabilities for flood disasters [99,100,101,102], and support the sustainable management of urban flooding in long-term planning [103,104,105]. Figure 7 illustrates the artificial intelligence methods employed in current research on urban waterlogging, as examined in our paper. Furthermore, the artificial intelligence methods used in urban waterlogging research are categorized, and the related studies are listed in Table 3.

4.1. Application of AI in Urban Flooding Prediction

The application of AI technologies in urban flooding prediction primarily involves data integration, risk assessment, model optimization, and intelligent decision-making. Using methods such as deep learning and data mining, AI can efficiently process multi-source data from various sensors, remote sensing images, weather stations, and other sources to identify and predict potential flooding risks in real time.

Urban flooding prediction relies on various data sources, including accurate precipitation data, land-use information, the status of urban drainage systems, and ground water levels. AI technologies can effectively integrate monitoring data from ground-based meteorological stations and water level sensors. Through model learning, AI can establish spatial and temporal patterns, providing more precise data for urban flooding predictions. Breiman et al. [106] proposed the Random Forest algorithm as a predictive tool, which demonstrates significant superiority in classification and regression tasks. The introduction of randomness provides optimized algorithmic support for urban flood prediction. Guo et al. [92] used an automated machine learning (AutoML) approach based on genetic algorithms to develop a rapid early warning model for urban flooding, combining meteorological information for quick flood depth predictions. Lin et al. [93] used the Random Forest model to study the impact of three-dimensional buildings on urban flooding, finding that building density, coverage, and congestion significantly affect the occurrence of flooding events. Derakhsha et al. [94] used Artificial Neural Networks (ANN) to refine precipitation data for the Sydney area, compensating for missing data from unrecorded rain gauge stations. Ke et al. [95] employed machine learning methods to classify urban flooding events in Shenzhen based on rainfall thresholds. Yan et al. [107] proposed using stacked models and hybrid models to assess urban flooding susceptibility in Beijing. Based on data from 266 flooding points and 532 non-flooding points, the study found that stacked models outperformed traditional machine learning models and identified five key indicators highly correlated with flooding, providing important references for urban planning.

Additionally, AI models can be combined with remote sensing data and GIS technologies for prediction. Tang et al. [16] used a weighted Naive Bayes (WNB) classifier integrated with GIS to assess urban flooding risk in Guangzhou. The results showed that six spatial factors significantly enhanced prediction efficiency, providing valuable insights for urban flood management. Chakrabortty et al. [96] used remote sensing and GIS platforms, combined with particle swarm optimization, artificial neural networks, and deep learning neural networks, to develop a flood susceptibility map and predict flood risks in the Kangsabati River Basin in India. The study found that particle swarm optimization had a high AUC value (0.942). Chen et al. [97] employed a machine learning-based REPTree model with Bagging and Random Subspace (RS-REPTree) ensemble framework, using GIS for spatial flood susceptibility prediction. The results indicated that the RS-REPTree model had the best predictive ability in flood susceptibility assessment, outperforming the models used individually. Tran et al. [98] used regression models in ArcGIS to predict urban flooding risks in Hanoi, Vietnam, from 2012 to 2018, with an explanatory power of 67.6%, providing significant references for urban planning.

4.2. Application of AI in Flood Risk Assessment

Another significant application of AI technology in the context of flooding is risk assessment and emergency response. AI can conduct quantitative assessments of flood disaster risks based on real-time data and historical information, spatially assess flood susceptibility, evaluate the severity of flooding in different urban areas, and propose emergency measures such as traffic diversion, drainage strategies adjustments, and the mobilization of rescue resources [83,108,109].

By analyzing meteorological, geological, and hydrological data with artificial intelligence techniques, it is possible to predict the likelihood and impact of flooding, providing a scientific basis for urban planning and risk assessment, thereby enhancing urban resilience and response efficiency. Lyu et al. [99] used Shanghai’s subway system as an example, employing a combination of Fuzzy Analytic Hierarchy Process (FAHP) and Fuzzy Clustering Analysis (FCA) to assess flood risks in a subsiding environment. The results indicated that over 30% of subway lines face high flood risk, suggesting the need for more detailed consideration of subsidence factors to mitigate risks. Cai et al. [100] analyzed the risk assessment of urban stormwater systems using neural networks and dynamic hydraulic models, introducing a fuzzy comprehensive evaluation method to establish a flood risk assessment framework, supporting drainage planning and emergency management in sponge cities and mapping urban stormwater and flood risks. Zhou et al. [101] predicted urban flood risks in Shenzhen by integrating hydrometeorological data, urban surface characteristics, and building configuration factors, with results showing that LightGBM performed best, with the main influencing factors being rainfall and building density. Li et al. [102] constructed a flood risk assessment framework using the MaxEnt model to analyze flood risks in Tianjin, revealing that population density and impervious surfaces are the primary influencing factors, with high-risk areas concentrated in the city center, providing theoretical support for urban planning and disaster warning.

In addition to short-term risk assessments, AI is also utilized in the long-term planning of urban drainage systems, identifying bottlenecks and potential risk points within the system and proposing corresponding optimization solutions. For example, AI-based simulation and optimization models can assist urban planners in designing new drainage networks, allowing for the rational layout of drainage facilities and enhancement of drainage capacity to avoid future flooding issues. Chapi et al. [103] proposed a new artificial intelligence model, Bagging-LMT, for flood susceptibility mapping. By generating a spatial database that includes flood inventories and flood condition factors, the model’s assessment results showed it outperformed four other benchmark models, making it suitable for sustainable management in flood-prone areas. Zhang et al. [104] used machine learning algorithms to study the dominant factors of flooding in different regions, integrating spatial heterogeneity theory to propose regionally adaptive disaster reduction strategies. Lin et al. [105] systematically explored the dynamic characteristics of the urban integration process through spatial clustering techniques, landscape metrics, and circular buffer analysis, with the successful experience of the Guangzhou–Foshan integration providing references for similar cases.

Table 3. Artificial intelligence methods for waterlogging research.

Reference	Year and Analysis	AI methods	Contents
Lin et al. [93]	2021 Investigating the influence of three-dimensional building configuration on urban pluvial flooding using random forest algorithm	Random Forest	Discussed the impact of three-dimensional building configuration on urban waterlogging, finding that building density, congestion, and coverage significantly affect waterlogging events.
Kornejady et al. [110]	2017 Landslide susceptibility assessment using maximum entropy model with two different data sampling methods	Maximum Entropy Model	Used the maximum entropy model and two sampling strategies to model landslide susceptibility.
Derakhshan et al. [94]	2011 Rainfall disaggregation in non-recording gauge stations using space–time information system	Artificial Neural Network	Refined precipitation data to compensate for missing data from unrecorded rain gauge stations and evaluated the best data refinement model.
Ahmed et al. [111]	2021 Deep learning hybrid model with Boruta-Random forest optimizer algorithm for streamflow forecasting	Deep Learning	Improved flow level prediction accuracy using a deep learning-based feature selection algorithm.
Guo et al. [112]	2020 Evaluation of spatially heterogeneous driving forces of the urban heat environment based on a regression tree model	Regression Trees	Analyzed the spatial heterogeneity impact of urban biophysical components on surface temperature using a regression tree model.
Tang et al. [16]	2018 A spatial assessment of urban waterlogging risk based on a Weighted Naïve Bayes classifier	Naïve Bayes Classifier	Combined a weighted Naive Bayes classifier and GIS to assess waterlogging risk in Guangzhou.
Yan et al. [107]	2024 Urban waterlogging susceptibility assessment based on hybrid ensemble machine learning models	Stacking Model and Hybrid Model	Evaluated the susceptibility to urban waterlogging in Beijing.
Aydin et al. [113]	2023 Predicting and analyzing flood susceptibility using boosting-based ensemble machine learning algorithms	Adaptive Reinforcement Learning	Analyzed flood susceptibility in Adana Province, Turkey.
Guo et al. [92]	2022 Construction of rapid early warning and comprehensive analysis models for urban waterlogging	Automated Machine Learning	Constructed an urban waterlogging early warning model.

Table 3. Artificial intelligence methods for waterlogging research.

Reference	Year and Analysis	AI methods	Contents
Lin et al. [93]	2021 Investigating the influence of three-dimensional building configuration on urban pluvial flooding using random forest algorithm	Random Forest	Discussed the impact of three-dimensional building configuration on urban waterlogging, finding that building density, congestion, and coverage significantly affect waterlogging events.
Kornejady et al. [110]	2017 Landslide susceptibility assessment using maximum entropy model with two different data sampling methods	Maximum Entropy Model	Used the maximum entropy model and two sampling strategies to model landslide susceptibility.
Derakhshan et al. [94]	2011 Rainfall disaggregation in non-recording gauge stations using space–time information system	Artificial Neural Network	Refined precipitation data to compensate for missing data from unrecorded rain gauge stations and evaluated the best data refinement model.
Ahmed et al. [111]	2021 Deep learning hybrid model with Boruta-Random forest optimizer algorithm for streamflow forecasting	Deep Learning	Improved flow level prediction accuracy using a deep learning-based feature selection algorithm.
Guo et al. [112]	2020 Evaluation of spatially heterogeneous driving forces of the urban heat environment based on a regression tree model	Regression Trees	Analyzed the spatial heterogeneity impact of urban biophysical components on surface temperature using a regression tree model.
Tang et al. [16]	2018 A spatial assessment of urban waterlogging risk based on a Weighted Naïve Bayes classifier	Naïve Bayes Classifier	Combined a weighted Naive Bayes classifier and GIS to assess waterlogging risk in Guangzhou.
Yan et al. [107]	2024 Urban waterlogging susceptibility assessment based on hybrid ensemble machine learning models	Stacking Model and Hybrid Model	Evaluated the susceptibility to urban waterlogging in Beijing.
Aydin et al. [113]	2023 Predicting and analyzing flood susceptibility using boosting-based ensemble machine learning algorithms	Adaptive Reinforcement Learning	Analyzed flood susceptibility in Adana Province, Turkey.
Guo et al. [92]	2022 Construction of rapid early warning and comprehensive analysis models for urban waterlogging	Automated Machine Learning	Constructed an urban waterlogging early warning model.

4.3. Ongoing Challenges and Future Prospects

In the study of urban flooding issues, artificial intelligence (AI) has demonstrated exceptional performance in areas such as computational speed, accuracy, and handling complex information. However, AI still faces ongoing challenges when analyzing urban flooding, including issues related to adaptability, data acquisition, model accuracy, and real-time processing capabilities, all of which require continuous improvement.

AI models have shown excellent precision and efficiency in predicting flood depth and assessing flood risk in urban areas. For example, a dual-model approach combined with numerical models has significantly outperformed traditional methods in computation speed according to Liu et al. [114], while video IoT-based sensing systems with high real-time capabilities can transmit flood information in a timely manner [115]. AI algorithms excel at processing complex multi-source data from fields such as remote sensing, meteorology, hydrology, and geographic information systems (GIS), and enabling comprehensive flood risk assessments through efficient modeling. For instance, the Random Forest model based on the Boruta algorithm effectively integrates climate patterns and flow levels, improving prediction performance [111]. AI models also quantify the main driving factors of urban flooding, providing a scientific basis for optimizing urban building configurations and formulating disaster reduction strategies. Key factors such as building density, vegetation coverage, and rainfall are critical to flooding. By adjusting these variables, flood risk can be effectively reduced [93]. Additionally, AI supports the development of rapid early warning systems, enhancing both the coverage and response speed of flood risk predictions. Early warning models based on AutoML can adapt to flood prediction needs in areas lacking monitoring data, such as water level stations [92]. Based on the Sections discussed earlier in the article, the framework for the warning mechanism of urban waterlogging is illustrated in Figure 8. The application of artificial intelligence in flood forecasting can accelerate data analysis of predictive models, thus speeding up the overall iterative process and enhancing the rate of generation for this emergency response increment.

The application areas of artificial intelligence in current research on urban waterlogging, as explored in our review article, are presented in Figure 9.

Despite the success of AI models in specific regions, their adaptability in different areas, scenarios, and climate conditions still needs improvement. For example, flood-driving factors may vary significantly across different geographical conditions, requiring regional optimization of models [107]. Furthermore, AI models, particularly deep learning models, are often criticized for their “black-box” nature, which increases the difficulty of interpreting results. This becomes a significant barrier when it comes to policy decisions and public trust. While interpretability tools like SHAP analysis can partially mitigate this issue, their scope and applicability are limited [113]. High-precision AI models typically require complex computational resources, which may exceed the infrastructure capacity of some cities. In real-time applications, the cost of large-scale data processing can be high, requiring further optimization of model efficiency [115].

Looking to the future, the application of AI in urban flooding can focus on the following directions: developing low-cost, high-performance models by improving algorithms and computational architectures to reduce the resource demands of models, enhancing model adaptability across various scenarios; promoting interdisciplinary research by integrating AI with knowledge from meteorology, hydrology, ecology, and other fields to explore more comprehensive solutions; developing AI tools with higher transparency and interpretability to make them more acceptable to policymakers and the public; and enhancing regional adaptability by optimizing AI models for the specific needs of different regions, improving their generalization capabilities.

The research methods under different urban conditions could be analyzed in greater depth. For instance, integrating GIS with deep learning models can lead to better predictions of the effectiveness of stormwater management measures in areas with abundant green spaces. In transportation hubs, combining real-time data and machine learning algorithms to optimize traffic diversion strategies can mitigate the impact of flooding on transport. Finding suitable models for various scenarios helps urban flooding management personnel in making informed decisions.

5. Conclusions and Prospect

Urban flooding not only threatens urban infrastructure and public safety but also has significant negative impacts on the ecological environment, economic activities, and the overall quality of life of residents. This paper provides a comprehensive review and analysis of the formation mechanisms, experimental research methods, and numerical simulation techniques of urban flooding, offering insights to address this global urban issue.

Currently, the main research directions in urban flooding include theoretical analysis, experimental research, numerical simulation, and the application of artificial intelligence (AI) technologies. Theoretical studies have revealed the complex mechanisms of urban flooding, such as the interactions between precipitation, land-use changes, and drainage system capacity. Experimental research, through physical models and hydrodynamic experiments, simulates flood scenarios and helps deepen the understanding of flooding processes, providing experimental evidence for further studies. Numerical simulation techniques, by constructing and testing numerical models, predict and evaluate the flooding occurrences and their impacts, providing algorithmic support for flood management. Additionally, various AI technologies play a key role in flood prediction and emergency response systems.

Despite the significant achievements in urban flooding research, many challenges remain. For example, traditional models are limited in processing high-resolution data and dynamic coupled scenarios, especially when confronted with complex urban terrain and climate change. Furthermore, efficiently integrating multidisciplinary technologies, such as hydrological modeling, AI, and remote sensing technologies, to address extreme rainfall and complex flooding scenarios remains a critical issue. Future research needs to explore more innovative governance strategies, optimize drainage system designs, develop green infrastructure, and integrate advanced prediction models with real-time data analysis to offer more accurate support for urban planning and emergency response. Based on this, the future research directions for urban flooding can be outlined as follows:

(i)

Prediction Models: There is a need to further improve the accuracy of existing models, particularly in how to efficiently combine high-resolution data with dynamic simulations.

(ii)

Flood Management Strategies: Optimizing urban drainage design, utilizing green infrastructure, and implementing innovative urban planning to enhance the city’s resilience to disasters.

(iii)

Multidisciplinary Integration: Combining hydrological modeling, AI, remote sensing technologies, and other methods to improve the comprehensive effectiveness of flood prediction, assessment, and management.