Waterlogging refers to the condition where the soil’s pores become saturated with water, eventually leading to surface flooding. Urban waterlogging is characterized by its suddenness, concentration, and destructiveness, often causing injuries and property loss. The issue of waterlogging in urban areas is frequent and recurrent, both in the hills and plains, as well as in hilly urban areas (Nowak and Greenfield, 2020). Currently, waterlogging is becoming a more frequent and serious issue worldwide, particularly in urban areas. It presents numerous physical, socio-economic, and environmental challenges, particularly in small and medium-sized cities (Chattaraj et al., 2021). Waterlogging damages pavements and can lead to water pooling especially in vulnerable low-lying areas and the at the bottom of slopes. This can potentially lead to secondary flood hazards. In developing countries, it causes economic loss and affects the living environment, people’s daily life, and sustainable development (Tang et al, 2018).

Unplanned urban development and inadequate stormwater management systems are major contributors to urban waterlogging. As rainwater accumulates, surface runoff increases, transforming streets into temporary streams and resulting in widespread flooding (Tam and Nga, 2018). This issue is exacerbated when rainwater from impervious surfaces like roads and buildings collects without proper drainage, intensifying the impact of waterlogging in urban environment. Furthermore, Land Use and Land Cover (LULC) changes such as those caused by urbanization modify the Earth’s surface environment, disrupting the natural water cycle and raising the risk of waterlogging disasters (Shrestha et al., 2018).

Coastal cities face even greater risks of waterlogging due to unique challenges like sea level rise and tidal flooding. The rising sea level not only increases the frequency and severity of tidal flooding but also reduces drainage capacity, as seawater often backflows into drainage systems during high tides or storm surges (McKenzie et al., 2021). These effects make it particularly difficult for coastal cities to manage stormwater, and during heavy rainfall, the combined impact of inadequate drainage and rising tides amplifies waterlogging risks. As climate change continues to drive sea level rise, coastal areas may face a compounding effect on urban waterlogging hazards, putting them at higher risk compared to inland cities.

Effective planning for rescue and recovery operations in urban areas requires continuous monitoring of waterlogging conditions. Remote sensing (RS) techniques provide a valuable tool for tracking and assessing spatiotemporal changes in flooded areas, offering a more efficient alternative to conventional methods. Traditional approaches such as land surveys, field observations, soil sampling, and reliance on local knowledge have been widely used to map waterlogged regions. However, these methods are often time-consuming and costly particularly for large-scale regional studies (Chowdary et al., 2008; Mahmud et al., 2017). RS technologies, on the other hand, enable frequent and extensive monitoring, facilitating active benchmarking of waterlogged areas by regularly comparing data on the extent, frequency, and spatial patterns of waterlogging. Such benchmarking enables urban planners to monitor critical parameters of waterlogging performance, including the frequency, duration, and spatial extent of waterlogged areas, as well as the effectiveness of drainage capacity and stormwater management systems. By comparing actual data on these parameters against projected outcomes, planners can assess the success of current flood management strategies and make necessary adjustments. This process helps inform greener and more resilient urban design, promoting sustainable infrastructure adaptations for improved flood mitigation. Earth Observation (EO) datasets, produced by hyperspectral sensors, Light Detection and Ranging (LiDAR), or Radio Detection and Ranging (RADAR) allow for versatile input and effective synergistic sampling strategies for the development of analytics frameworks. 

Remote sensors and datasets used for urban water management and waterlogging

Remote sensors and resulting datasets play an increasingly crucial role in urban water management with implications that are significantly important in this age of climate change (Wentz et al., 2014). The sensors are mainly hosted on satellites, drones, and aircraft platforms. They provide continuous, repeatable, high-resolution images across a range of wavelengths and are thus well-suited to observe urban waterlogging, sediment transport in coastal regions, and spatiotemporal variations of surface waters in urban areas (Lu et al., 2017). Recent advancements in the application of remote sensing in the context of water management have provided a comprehensive array of datasets with a wide range of spectral, temporal, and spatial resolutions. These datasets have the potential to aid stakeholders in tackling a range of issues related to water and climate change. Specifically, Earth Observation (EO) data products such as land cover, surface reflectance, surface temperature, and vegetation indices have emerged as crucial datasets for understanding the long-term dynamics of water resources in urban areas (Dube et al., 2023).  

Currently, urban water management faces several challenges especially in relation to waterlogging, which disrupts infrastructure, degrades land, and poses risks to public health (Herslund & Mguni, 2019). Earth observation technologies have become essential tools in monitoring and managing waterlogging by providing high-resolution, near-real-time data across large urban areas. This section outlines various remote sensors and datasets used for monitoring waterlogging and supporting best management practices, which can contribute to improved environmental, social, and educational conditions in urban areas (Barbosa et al., 2012). The choice of dataset depends largely on the task’s specific requirements. For example, Landsat and Sentinel-2 provide free, medium-resolution optical data suitable for large-scale, long-term monitoring; however, they are limited by cloud cover and only capture data during the day. In contrast, Sentinel-1, which uses Synthetic Aperture Radar (SAR), is unaffected by cloud cover and can operate both day and night, making it more effective for real-time waterlogging detection under challenging weather conditions. This all-weather, all-time capability allows for consistent monitoring even during heavy rainfall or overcast conditions, which are often associated with waterlogging events. LiDAR datasets are also critical for predicting water flow during extreme events, adding precision to flood modelling by capturing detailed topographical data. The integration of these diverse datasets optical, SAR, and LiDAR significantly enhances the accuracy of waterlogging predictions and management strategies.

Algorithm and models used to model waterlogging 

The use of algorithms and models in detecting and managing urban waterlogging through EO data has become critical due to rising urbanization and extreme weather events. Accordingly, various algorithms, such as machine learning techniques, physical-based models, and statistical methods, have been developed to model waterlogging by integrating spatial and geospatial datasets.

Machine Learning models

Machine learning models, such as Random Forest (RF) and Support Vector Machines (SVM), are popular for classifying urban areas at risk of waterlogging by integrating features like vegetation cover, impervious surfaces, and hydrological data (Tang et al., 2019). For example, Random Forest models can handle the nonlinear relationships inherent in urban landscapes and are particularly useful when combined with remote sensing indices like the Soil Moisture Index (SMI) to estimate surface soil moisture a key predictor of waterlogging (Wang et al., 2021; Louloudis et al., 2023). To apply RF in waterlogging detection, a large dataset of labeled samples (indicating waterlogged vs. non-waterlogged areas) is needed. By training the RF algorithm on features such as land cover, soil moisture, and slope, the model learns to classify areas based on patterns in the input data.

Convolutional Neural Networks (CNNs) 

Deep learning techniques, particularly Convolutional Neural Networks (CNNs), have also been introduced for waterlogging detection in large-scale satellite imagery. CNNs excel at extracting spatial features automatically from images, identifying subtle changes in urban landscapes that may indicate waterlogging (Huang et al., 2020). To use a CNN, you would need a substantial set of labeled satellite images where instances of waterlogged areas are marked. The CNN can then be trained to recognize waterlogging features based on shape, texture, and color patterns in the imagery. For instance, pooling and convolution layers in the CNN extract features at various scales enabling the model to detect both fine and large-scale alterations in land cover related to water accumulation and CNNs can be structured for feature extraction and classification (Huang et al., 2020).

Hydrological models

Hydrological models such as the Soil and Water Assessment Tool (SWAT) (https://swat.tamu.edu/#/) and Hydrological Engineering Center's River Analysis System (HEC-RAS) (https://www.hec.usace.army.mil/software/hec-ras/), are also widely employed to simulate the water flow dynamics in urban catchments. These models incorporate precipitation data, soil properties, and urban infrastructure to predict flood risks and water accumulation, providing a more detailed understanding of urban waterlogging processes (Mardonez et al., 2023). When coupled with remote sensing data, these models can be calibrated and validated against observed waterlogging events, making them valuable for urban waterlogging management (Glick et al, 2023). The effectiveness of these models depends on the resolution of the RS data sources. For instance, a minimum spatial resolution of 10-30 meters, as offered by satellites like Sentinel-2 and Landsat, is generally suitable for detecting surface water patterns in urban areas and capturing details such as roadways and drainage structures (Shrestha et al., 2021). For finer detail in dense urban areas, commercial satellites such as WorldView-3 and GeoEye-1 from Maxar, and Pleiades from Airbus, offer high spatial resolutions (up to 0.3 meters) that can capture intricate urban features like small streets and drainage structures (Liu, 2023; Cisse et al, 2022; Messina & Modica, 2022). Additionally, PlanetScope from Planet Labs provides high-revisit, medium-resolution imagery (up to 3 meters) that can be valuable for frequent monitoring (Ibrahim & Balzter, 2024). Temporal resolution is also crucial, daily or sub-weekly revisit times, such as those of Sentinel-1 radar imagery, are often necessary for monitoring rapid changes, especially during heavy rainfall events or seasonal shifts. Choosing the right resolution based on the specific urban area and intended use case is essential for accurate waterlogging assessment and management.

Case studies and applications on urban water management

The advancement of remote sensing technologies has provided a set of tools for monitoring, predicting, and managing these events. This section presents case studies that illustrate how remote sensing approaches, algorithms, and models have been applied in different urban settings to detect and manage waterlogging.

Dhaka, the capital of Bangladesh, faces persistent waterlogging issues due to a combination of rapid urbanization, intensive rains, and inadequate drainage infrastructure. Alam et al. (2023) utilized remote sensing data and GIS to simulate urban waterlogging and manage flood risks. The study integrated satellite data from MERIT Hydro, Landsat images, BRAC urban slum amp and MODIS with Normalized Difference Water Index (NDWI) to detect water bodies and urban waterlogged areas. By combining these techniques, the research team successfully identified potential waterlogged areas and provided decision-makers with a comprehensive model for flood management. 

The results support the conclusion that integrating satellite data with hydrological models provides an efficient way to predict waterlogging in urban areas and suggests that this combination can be highly effective in mitigating future risks for regions like Dhaka (Alam et al., 2023).

Another study conducted in Fuzhou, China utilized a Multisource Data Fusion (MDF) approach integrated with the U-Net model, a convolutional neural network (CNN) architecture commonly used for image segmentation. Satellite remote sensing images, along with data on drainage networks, land use, and terrain, were collected for Fuzhou City. The U-Net model was specifically employed to identify and extract features such as buildings and urban infrastructure from remote sensing imagery (Yang et al., 2023). One of the critical innovations of the authors’ research was the application of coupled one-dimensional (1D) conduit drainage pipe models and two-dimensional (2D) hydrodynamic models, which allowed for a more detailed simulation of water behaviour across various urban topographies. 

The results indicated that the proposed approach greatly improved the simulation accuracy of waterlogging points by 29%, 53%, and 12% compared with the raw DEM, IDW, and MDF. This case study highlights the potential of integrating algorithm, hydrological modelling, and remote sensing to create more effective tools for managing urban waterlogging, particularly in cities with complex infrastructure like Fuzhou. The adoption of such advanced techniques offers a scalable and reliable solution for improving urban resilience to water-related disaster.

Conclusion

Urban waterlogging is a significant and growing challenge in many cities due to factors like unplanned development, inadequate drainage systems, and climate change. Remote sensing offers an effective solution for detecting and managing waterlogging, enhancing urban planning and disaster management. By integrating EO data with algorithms such as ML and CNNs, cities can improve predictions and mitigate impacts, enabling more sustainable management practices. Besides, combining remote sensing data with hydrological modelling techniques supports urban resilience and contributes to environmental sustainability.

Policy implications and recommendations

Urban waterlogging poses substantial challenges to environmental sustainability, infrastructure management, and public health. Addressing this challenge requires a combination of effective policy frameworks and technological advancements like remote sensing. Using remote sensing tools as a standard practice can improve urban water management (Liu et al., 2023). Predictive analytics and geospatial data are useful tools to design of resilient infrastructure. Data sharing and collaborative frameworks between government agencies, academic institutions, and private entities can facilitate the pooling of resources for the establishment of real-time monitoring systems that combine satellite data with ground-based sensors. Furthermore, the implementation of low-cost radar technologies for monitoring localized flood events could provide a more comprehensive understanding of the dynamics of urban waterlogging. Public and private collaborations should be incentivized, because they promote the development of advanced remote sensing technologies and ensure that urban planning departments are equipped with the necessary data and tools to respond quickly and effectively. Finally, there is also a need for policies that integrate climate change adaptation into urban water management strategies such that urban areas can better prepare for waterlogging and reduce their overall vulnerability.