This session focuses on high-resolution precipitation and hydrological measurements in cities and on approaches to improve modeling of urban hydrological response, including:
- Novel techniques for high-resolution precipitation measurement in cities and for multi-sensor data merging to improve the representation of urban precipitation fields.
- Novel approaches to hydrological field measurements in cities, including data obtained from citizen observatories.
- Precipitation modeling for urban applications, including convective permitting models and stochastic rainfall generators.
- Novel approaches to modeling urban catchment properties and hydrological response, from physics-based, conceptual and data-driven models to stochastic and statistical conceptualization.
- Applications of measured precipitation fields to urban hydrological models to improve hydrological prediction at different time horizons to ultimately enable improved management of urban drainage systems (including catchment strategy development, flood forecasting and management, real-time control, and proactive protection strategies aimed at preventing flooding and pollution).
- Strategies to deal with upcoming challenges, including climate change and rapid urbanization.
Orals: Thu, 7 May, 08:30–10:15 | Room 2.15
Artificial Intelligence (AI) is no longer a novelty in hydrometeorology. From computer vision to rainfall nowcasting, AI-based models now routinely outperform traditional approaches in many benchmark comparisons. Yet, as these tools move closer to operational use --particularly in dense and vulnerable urban environments-- it is timely to step back and ask a more fundamental question: what is AI actually good at, and how should we use it wisely?
This talk reflects on recent advances in AI for urban hydrometeorology through three interconnected research regimes: smart environmental monitoring (“eyes on the water”), short-term rainfall nowcasting, and spatial–temporal rainfall reconstruction. Rather than promoting AI as a universal solution, the talk focuses on model suitability, uncertainty, and the alignment between data-driven methods and physical processes.
In environmental monitoring, modern deep-learning computer vision models have reached an impressive level of maturity. Tasks such as object detection, classification, and segmentation can now be performed reliably using images from fixed cameras, mobile devices, CCTVs, and citizen sensors, enabling scalable monitoring of urban rivers, flooding, and water quality indicators. At the same time, these applications reveal a recurring limitation: AI performs extremely well on what it has seen before, but struggles with rare events, or poorly defined labels --often the cases of greatest societal relevance.
In rainfall nowcasting, AI is often positioned as a disruptive replacement for traditional methods. This talk argues instead for a complementary view. While classical extrapolation efficiently handles storm motion, AI’s real strength lies in learning evolution: how rainfall structures grow, decay, and reorganise across spatial and temporal scales. Deep learning models excel at capturing multiscale spatial–temporal patterns that are difficult to encode explicitly, making them particularly valuable when combined with physically informed frameworks.
A central challenge across these applications is overconfidence. Can we teach AI to say “I don’t know”? Recent uncertainty-aware learning approaches demonstrate that AI models can be trained not only to make predictions, but also to indicate when they are operating outside familiar regimes --an essential requirement for trustworthy deployment.
Finally, the talk highlights the Point-to-Image (P2I) model to illustrate AI’s ability to learn spatial–temporal structure from extremely sparse data. By reconstructing realistic rainfall fields from limited point observations, P2I demonstrates that AI can infer coherent spatial patterns and temporal consistency even when traditional methods fail. This capability challenges long-held assumptions about data density requirements and opens new possibilities for urban hydrometeorology in data-limited environments.
Overall, this talk argues that the most effective use of AI in urban hydrometeorology arises not from replacing physical insight, but from combining process understanding with models that are well matched to the questions they are asked to answer.
How to cite: Wang, L.-P., Tseng, C.-Y., Chen, C.-J., Wang, B.-Z., and Yu, Y.-C.: AI for urban hydrometeorology: insights into processes, model suitability, and challenges, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15978, https://doi.org/10.5194/egusphere-egu26-15978, 2026.
Under global warming, the intensification of extreme precipitation poses a critical threat to dense coastal metropolises like Hong Kong, where complex terrain and high population density amplify flood risks. While daily rainfall trends are well-documented, existing studies largely focus on coarse regional simulations or daily-scale metrics, leaving a gap in understanding the granular evolution of sub-daily extremes and their specific drivers within complex intra-urban environments. This study investigates the spatiotemporal characteristics of hourly extreme rainfall in Hong Kong from 1991 to 2024, utilizing continuous hourly records from approximately 80 Geotechnical Engineering Office (GEO) stations. We spatially categorize the territory into four distinct subregions—Hong Kong Island, Kowloon, New Territories, and Lantau—to examine regional heterogeneity. The analysis employs indices including hourly precipitation percentiles (95th, 97.5th, 99th, and 99.9th) and Maximum Rolling Rainfall (MRR) across 1, 3, 6, and 12-hour durations to capture both short-term intensity and cumulative event magnitude. Furthermore, a Structural Equation Modelling (SEM) framework is developed to disentangle the contributions of key drivers, specifically quantifying the impact of urbanization (e.g., built-up area, patch density) alongside large-scale climate variability (e.g., ENSO, monsoons) and socioeconomic factors. We hypothesize that short-duration rainfall intensity (1-hour MRR) exhibits a more significant upward trend than longer durations, particularly in highly urbanized sectors like Kowloon. The SEM analysis is expected to reveal that urbanization acts as a primary localized driver exacerbating extreme rainfall frequency and intensity, distinct from background climatic warming. These findings will provide essential insights for refining urban drainage standards and disaster mitigation strategies in high-density mountainous cities.
How to cite: Lu, Y., Tang, X., and Wang, D.: Spatiotemporal Variability and Attribution of Hourly Extreme Rainfall in Hong Kong: A Multi-Regional Analysis Using Structural Equation Modelling , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6258, https://doi.org/10.5194/egusphere-egu26-6258, 2026.
Rapid urbanisation is coinciding with a rising trend in the intensity of extreme precipitation. Beyond this large-scale trend, urbanisation further alters local climate by modifying land cover, energy fluxes and airflow. However, it remains unclear how much urbanisation alters heavy precipitation, and how robust different machine learning (ML) models are in uncovering its influence. This uncertainty limits our ability to design targeted adaptation measures (e.g. managing impervious surfaces, cooling hotspots). To address these gaps, we analyse extreme daily precipitation across more than 5,000 stations in Europe using gauge observations, high-resolution meteorological reanalyses and 1km land-use data. Stations are classified into four urbanisation levels (rural, suburban, urban, highly urban) based on impervious surface fraction of surrounding area, and predictors are grouped into geographic, surface, thermal and dynamic categories. We train multiple ML models (ElasticNet, RF, LightGBM, and ANN) under a unified framework and applied explainable AI techniques (SHAP and ALE) to diagnose how these models use physical information across urbanisation levels. Tree-based ensembles achieve the highest skill (R2 = 0.45, RMSE=9.28 mm), while all models systematically underestimate the most intense events (>100 mm/d). Our analysis of the ML models finds that thermodynamic variables (dewpoint temperature and heat flux) are the primary controls on extreme precipitation across all urbanisation levels, as evidenced by their larger SHAP ranges (1.36–1.66) compared with the other categories. In contrast, dynamic predictors (U/V component of wind, pressure, and vertical velocity) exert a weaker but relatively consistent influence (SHAP range: 0.74–0.88). In non-urban models, surface processes play a limited role in explaining extreme precipitation. However, in the highly urban model, increasing impervious surface fraction contributes positively to predicted rainfall intensity (net ALE change of about 8 percentage points across the data range). Hence, as urbanisation intensifies, we find impervious surfaces are becoming an increasingly significant explanatory factor in ML models of heavy rainfall.
How to cite: Zhang, Y., Anderson, B. J., Hart, N., and Slater, L. J.: Evidence from XAI for how extreme precipitation relates to urbanisation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-613, https://doi.org/10.5194/egusphere-egu26-613, 2026.
Accurate precipitation estimation is fundamental for hydrological forecasting and disaster risk management, with radar-based Quantitative Precipitation Estimation (QPE) providing high-resolution rainfall input for real-time applications. In recent years, machine learning approaches have been widely adopted to improve radar QPE, with both forest-based models (e.g., Random Forest, gradient-boosted trees) and deep learning architectures outperforming traditional reflectivity-rain rate (Z-R) relationships. Despite this progress, most studies emphasise performance comparisons, offering limited insight into how forest-based models exploit radar-derived information across different temporal scales.
In this study, we move beyond accuracy benchmarks to investigate the predictive behaviour of forest-based models for radar rainfall estimation. We conduct a systematic set of experiments in which the input feature space is progressively expanded to include three-dimensional reflectivity profiles, derived radar products (e.g. MaxDBZ, VIL and so on), dual-polarization variables (e.g. Kdp), and geographical information.Model performance and feature importance are analysed for QPE at both 10-min and 1-h timescales.
Our results reveal clear, scale-dependent patterns in model behaviour. At the hourly timescale, predictive performance is primarily governed by simplified radar intensity measures (i.e. MaxDBZ) combined with geographic information, suggesting a dependence on regional weather patterns. In contrast, at the 10-min timescale, performance is more strongly associated with three-dimensional and vertically integrated radar features, indicating a more localized and dynamic regime. These findings highlight that forest-based models adapt their effective use of radar information depending on temporal scale, motivating further diagnostic analyses of ensemble behaviour to better characterise how tree-based models balance local and aggregated information in radar QPE.
How to cite: Yang, P.-H. and Wang, L.-P.: Why do forest-based models work for radar rainfall estimation? Insights from 10-minute and hourly QPE experiments, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9148, https://doi.org/10.5194/egusphere-egu26-9148, 2026.
A critical challenge in urban hydrometeorology is achieving accurate nowcasting of storms and inundation risks, hindered by the high heterogeneity of both urban surfaces and precipitation fields, which jointly induce nonlinear and abrupt spatiotemporal patterns of inundation risk. To address this, we develop a cascaded deep learning framework that performs end-to-end, high-resolution forecasting from radar extrapolation to inundation risk. First, a ConvLSTM-UNet model is trained on a decade-long (2015–2025), highspatiotemporal-resolution (1 km, 6 min) radar echo mosaic dataset over Wuhan, China, to generate skillful short-term quantitative precipitation nowcasts, thereby capturing fine-scale rainfall heterogeneity. Second, using a large set of historical waterlogging points collected from online platforms as labeled data, another deep learning model is trained to learn the complex coupling between nowcasted rainfall and high-resolution urban features with 12.5 m DEM, 30 m Local Climate Zone maps, and fine road networks, thereby quantifying how surface heterogeneity modulates runoff accumulation and flood susceptibility. By chaining these two stages, the framework produces high spatiotemporal resolution, probabilistic inundation risk nowcasts directly from radar observations. This data-driven approach offers an effective and novel tool for real-time early warning and refined risk management in complex urban environments.
How to cite: Zhao, L. and Song, J.: Chained Nowcasting of High Spatiotemporal Resolution Urban Rainfall and Inundation Risk Using Deep Learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11991, https://doi.org/10.5194/egusphere-egu26-11991, 2026.
The extreme rainfall event of 27–28 June 2024 resulted in 228.1 mm of rain in Delhi, the heaviest in nearly eight decades, causing severe damage across the city and its surrounding regions. To understand the evolution of the spatio-temporal variability of this event and to quantify uncertainties in short-range forecasts of such urban extreme precipitation, we employ an ensemble of high-resolution, urban-aware WRF simulations. This study examines the sensitivity of WRF hindcasts to boundary condition datasets by comparing simulations forced with ERA5 and NCEP-FNL reanalyses. We also assess the model's sensitivity to microphysics parameterizations, as an accurate representation of cloud microphysical processes is crucial for forecasting extreme rainfall. Two widely used schemes, Thompson and WSM6, are evaluated. In addition to boundary conditions and microphysics schemes, we test the model performance for four different initialization times, starting from 1200 UTC on 25 June at 12-hour intervals, using both ERA5 and NCEP-FNL forcing. Preliminary results (from NCEP-FNL-driven runs) show that both microphysics schemes underestimate total rainfall. The Thompson scheme, when initialized at 0000 UTC on 26 June 2024, effectively captures the spatial structure of intense rainfall. The WSM6 scheme better reproduces the overall magnitude of the extreme rainfall but exhibits a spatial displacement bias. Among all initialization times, simulations starting at 0000 UTC on 26 June 2024 perform the best. The same experimental setup will be applied using ERA5 boundary conditions, and the outcomes will be compared against the NCEP-FNL-driven simulations to determine which boundary condition better represents the observed extreme rainfall event.
How to cite: Agrawal, A., Rajendran, V., Tewari, M., and Salamanca, F.: WRF Hindcast Sensitivity for Delhi's 28 June 2024 Extreme Rainfall: Role of Boundary Conditions, Microphysics, and Initial Time, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-327, https://doi.org/10.5194/egusphere-egu26-327, 2026.
Precipitation extremes associated with climate change are exacerbating urban flood risks worldwide, impacting rapidly urbanizing and socioeconomically vulnerable regions in the world. Addressing these challenges requires integrated approaches that link precipitation dynamics, urban hydrology, and community-centered adaptation. This work presents a multi-regional framework for precipitation-driven urban flood forecasting and mitigation aligned with the United Nations Sustainable Development Goals (SDGs), particularly SDG 11 (Sustainable Cities and Communities), SDG 13 (Climate Action), and SDG 6 (Clean Water and Sanitation). The research is led by the United Nations University (UNU) Hub at the City College of New York—the first and only UNU hub in the United States dedicated to advancing urban resilience through science–policy integration.
The presented framework integrates artificial intelligence (AI) and high-resolution hydrometeorological data across diverse urban environments. In Mumbai, India, machine-learning-based flood forecasting models are developed using high-resolution precipitation data, topography, land-use dynamics, and satellite observations to simulate real-time flood depths and extents during extreme rainfall events. These methods explicitly capture spatial variability in urban precipitation and evolving impervious surfaces, enhancing early warning capabilities in one of the world’s most flood-prone megacities.
The transferability of the methodology is demonstrated at different urban communities, including Mumbai, New York City, and the Caribbean. In New York City, precipitation-driven flash flood alert systems estimate real-time inundation risks during short-duration, high-intensity rainfall events. In Puerto Rico and the U.S. Virgin Islands, high-resolution inland flood risk maps are generated by integrating Depth–Duration–Frequency (DDF) precipitation relationships with terrain, soil, and land-use data, enabling the identification of flood hotspots under both current and projected rainfall regimes.
Beyond forecasting, the study advances a participatory framework for implementing nature-based solutions (NBS) in rural areas of Puerto Rico. By incorporating social vulnerability indicators and engaging local stakeholders, the approach supports equitable, community-driven flood mitigation strategies that enhance resilience actions.
Overall, this work demonstrates how precipitation-focused urban hydrology and AI-driven forecasting can be applied across global contexts to reduce flood risk, provide climate resilience, and facilitate the UN SDGs in both the Global South and developed urban regions.
How to cite: Khanbilvardi, R. and Goldberg, M.: Urban Flood Resilience Under Extreme Precipitation: AI-Based Forecasting and Participatory Solutions Aligned with the UN Sustainable Development Goals, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4558, https://doi.org/10.5194/egusphere-egu26-4558, 2026.
Accurate representation of radar-based rainfall inputs remains a critical challenge in urban stormwater modelling, particularly in densely urbanized environments exposed to short-duration intense storm events. While weather radar provides high spatio-temporal resolution precipitation estimates (Taromideh et al., 2025), its direct application in urban stormwater models is often affected by biases and spatial inconsistencies. Improving the integration of radar-derived rainfall information with hydrological observations is therefore essential to reliably simulate urban runoff and sewer system response.
The framework is applied to the coastal city of Portici, located within the metropolitan area of Naples in southern Italy. The study area is characterized by a highly urbanized combined sewer system serving a catchment of approximately 3.2 km², with an imperviousness of about 78% and elevations ranging from sea level to 144 m above sea level. The drainage network includes multiple regulators and combined sewer overflow structures that discharge excess stormwater to the sea during intense rainfall events (Marino et al., 2025). Discharge measurements at the main outlet are available at high temporal resolution over a multi-year period, providing a reliable dataset for model calibration and validation. While no rain gauges are installed within the catchment, nearby rain gauge stations and meteorological radar data are available. Radar precipitation is provided on a regular grid with 1 km × 1 km spatial resolution and 5-minute temporal resolution, enabling the estimation of spatially distributed rainfall fields over the study area. These data provide the necessary context for applying and evaluating the proposed optimization framework.
The objective of this study is to develop an optimization-based framework to adjust subcatchment-scale rainfall inputs in an urban stormwater model, using observed outlet discharge as an indirect reference for rainfall correction. Initial rainfall values for each subcatchment are derived from radar precipitation fields, and the optimization aims to ensure consistency with observed outlet discharges while preserving the spatial and temporal structure of radar-derived rainfall. The approach constrains rainfall adjustments to physically plausible patterns and prevents unrealistic hydrological responses, such as excessive runoff variability or flooding within the sewer network.
The proposed methodology couples the calibrated Storm Water Management Model (SWMM) with a genetic algorithm optimization scheme. Time-series rainfall values for each subcatchment are treated as decision variables over the duration of the storm event. Radar-derived precipitation fields are used both to initialize these variables and to activate them only during periods when radar precipitation is detected. The objective function integrates flow-based performance metrics, correlation measures between radar and subcatchment rainfall data, and runoff consistency indicators, while a flooding volume constraint penalizes solutions leading to surcharging or surface flooding. Model evaluations are parallelized to reduce computational cost. The proposed framework offers a robust methodology for improving the consistency between radar-based rainfall inputs and observed sewer system responses in urban environments, that can be used for the development of predictive models of rainfall-runoff transformation in urban catchments.
How to cite: Taromideh, F., Marino, P., Santonastaso, G. F., and Greco, R.: Radar Rainfall Input for Urban Stormwater Modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7034, https://doi.org/10.5194/egusphere-egu26-7034, 2026.
Characterized by high spatial heterogeneity and short response times, urban flood research has long been constrained by the scarcity of precise, high-spatiotemporal resolution observational data required to capture the complex dynamics. Without such data, it remains challenging to distinguish whether model simulation errors stem from uncertain parameters or fundamental structural deficiencies, obscuring the model's reliable physical representation. Consequently, the physical fidelity of urban hydrodynamic models in reproducing complex, spatiotemporal flood dynamics needs to be further validated. To address this, we constructed the first large-scale, 1-minute resolution urban surface inundation dataset derived from camera videos using a Large Multimodal Model (GPT-5). Using the observations, we audited the physical driving mechanisms of 1D-2D coupled hydrodynamic models by examining the spatial stratified heterogeneity of flood responses.
Focusing on 5 diverse rainfall-flood events recorded by 226 traffic surveillance cameras in Dalian, China, we utilized GPT-5 to automatically extract real-time waterlogging levels. The extracted data underwent manual verification by experts to strictly correct errors, resulting in a high-fidelity urban surface inundation dataset at a 1-minute resolution. Subsequently, we constructed and calibrated a 1D-2D coupled urban flood numerical model to obtain simulation results for the corresponding events. The Geodetector model was then employed to quantify and compare the spatial stratified heterogeneity of waterlogging derived from observations versus simulations, attributing them to 9 potential drivers including rainfall, topography, and drainage infrastructure.
Results indicate that GPT-5 achieved satisfactory extraction performance, with an average accuracy of 77%. Comparative Geodetector analysis of observations versus simulations revealed critical discrepancies. The factor detector showed low individual explanatory power (q<0.1) but distinct rankings, with simulations underestimating rainfall's role. The interaction detector revealed stronger observed synergy, where the dominant control shifted from the observed "rainfall-imperviousness" coupling (16.6%) to a simulated "pipe-imperviousness" one (14%). While the risk detector confirmed consistent trend patterns, it highlighted significant "peak-shaving" effects, with simulated depths averaging 20 cm lower. Finally, the ecological detector verified that these structural discrepancies are statistically significant rather than random errors.
Observations confirm that urban flood distribution is governed by the non-linear synergy of multiple factors, reflecting high system complexity. The model reveals a systematic structural defect: it erroneously shifts the dominant control from a dynamic rainfall-surface coupling to a static boundary condition. This bias causes the model to be insensitive to dynamic meteorological forcing and to underestimate severe localized inundation caused by micro-environments. Future improvements must move beyond parameter calibration to focus on enhancing sensitivity to rainfall fluctuations and micro-environmental representation.
How to cite: Zhou, S., Eltner, A., Zamboni, P., Lyu, H., and Zhang, C.: Auditing the Physical Fidelity of Urban Flood Model with Large Multimodal Model-Derived High-Resolution Observational Data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14345, https://doi.org/10.5194/egusphere-egu26-14345, 2026.
Urban areas can substantially modify local hydroclimate, enhancing precipitation over and downwind of cities. Yet, the urban expansion effects on rainfall remain insufficiently understood, and a quantitative relationship between urban growth and rainfall intensification remains to be established. Using the WRF model with eight urban size scenarios for Beijing, a numerical modeling framework was implemented to investigate how changes in urban extent influence rainfall during two representative summers, one relatively wet and one relatively dry. Results show that the rainfall response exhibits an approximately linear dependence on the degree of urban expansion, with the largest impacts occurring over the city center and downwind regions. In general, rainfall increases with urban area enlargement, particularly during nighttime in relatively wet summers due to higher humidity and a more pronounced urban heat island effect. In relatively dry summers, limited moisture supply leads to smaller changes in total rainfall. Changes in hourly rainfall intensity demonstrate a contrasting pattern. Heavy rainfall intensities further intensify in response to urban expansion, while light rainfall is suppressed or remains largely unchanged. Daytime and nighttime rainfall intensity respond to urban expansion in opposite ways, with daytime intensity generally weakening and nighttime intensity strengthening as Beijing expands. These contrasting diurnal behaviors ultimately lead to a reduction in rainfall intensity during relatively dry summers and a slight increase during relatively wet summers. Overall, the results highlight the dependence of urban rainfall modification on city size and background climatic conditions.
How to cite: Qiao, Z., Koukoula, M., Ni, G., and Peleg, N.: Rainfall response to urban expansion in Beijing and its local climate drivers, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2625, https://doi.org/10.5194/egusphere-egu26-2625, 2026.
Urban areas are increasingly vulnerable to extreme rainfall events, which can cause severe flooding and significant infrastructure damage. This study focuses on predicting water levels and flows in a Danish stream to support urban flood mitigation using rainfall observations and numerical weather predictions. The city of Vejle, Denmark, is vulnerable to extreme rainfall due to flooding risks associated with the stream that traverses the city; hench, a pump and sluice facility have been installed as protective measures. However, effective flood mitigation depends on timely early warning and informed decision-making, highlighting the need for accurate and reliable stream water level forecasts.
The approach presented in this study integrates observed rainfall from rain gauges and numerical weather predictions as inputs to two distinct hydrological modeling frameworks: a linear reservoir model and a neural network model. Both models aim to predict streamflow and water levels in the critical stream through Vejle. Real-time flow and water level sensors installed in the stream provide continuous measurements for model calibration and validation.
Results from this study demonstrate that the proposed models achieve high accuracy in forecasting both flow and water levels. The neural network model shows particular promise in capturing nonlinear dynamics, while the linear reservoir model offers robustness and interpretability. These forecasts are operationally significant: they enable the local utility to optimize pump and sluice operations, reducing the risk of urban flooding and minimizing potential damage during extreme events.
This work highlights the value of combining rainfall observations and weather forecasts with hydrological runoff models to enhance urban flood resilience. The proposed methodologies are computationally efficient, scalable, and adaptable, making them highly valuable in real-time applications used by municipalities for flood mitigation.
How to cite: Nielsen, J. E., Nielsen, J. M., Vester, I. K., Thorndahl, S., and Rasmussen, M. R.: Forecasting of Streamflow and Water Levels for Urban Flood Protection in Vejle, Denmark, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7653, https://doi.org/10.5194/egusphere-egu26-7653, 2026.
This study evaluates the hydrologic resilience of the Beitou–Shilin Technology Park (BSTP), a high-density newly developed district in Taipei, under intensified extreme rainfall induced by climate change. To support climate-adaptive urban water management, an integrated modeling framework combining urban drainage simulation and future climate projections was established.
A detailed urban drainage model was developed using EPA SWMM to characterize the drainage system of the study area. Subcatchment geometries were delineated through QGIS-based spatial analysis, while infiltration parameters were assigned based on land-use types and vegetation coverage. This modeling framework provides a physically-based representation of surface runoff generation and urban flood response.
To address climate uncertainty, future rainfall data were generated using the MultiWG stochastic weather generator. The projection process incorporated five Global Climate Models (GCMs) under three Shared Socioeconomic Pathway scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5). After completing data preprocessing, the daily synthetic rainfall series were temporally downscaled to an hourly resolution to enable continuous hydrologic simulations within the SWMM framework.
Finally, a systematic sensitivity analysis of Low Impact Development (LID) strategies was conducted. Incremental implementation levels ranging from 0% to 100%, at 20% intervals, were simulated to quantify their effectiveness in reducing peak discharge and mitigating urban flood risk under extreme rainfall conditions. The results reveal a clear nonlinear relationship between LID implementation scale and runoff reduction efficiency. These findings provide quantitative insights for optimizing LID configurations in compact urban developments and support long-term, ESG-oriented urban infrastructure planning.
How to cite: Chen, Y.-C. and Tung, C.-P.: Strategic Assessment of Urban Flood Risk and LID Mitigation under Climate Change: A Case Study of the Beitou–Shilin Technology Park, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7937, https://doi.org/10.5194/egusphere-egu26-7937, 2026.
Urban flood risk management requires innovative approaches to address the increasing variability of extreme meteorological events, intensified by climate change and urban expansion. This study proposes a novel and replicable methodology for the classification of rainfall events in urban areas to obtain more precise design storm events respect to the more used IDF curves. The selected study area is the hill of the municipality of Turin, characterized by steep slopes, widespread urbanization, and a dense network of minor streams, partially open-channel and partially culverted, which are not represented in standard flood hazard maps.
Extreme rainfall events exhibit complex dependencies among key attributes such as duration, intensity, and cumulative precipitation, which cannot be correctly described using univariate approaches that may cause significant over-simplification. To address this limitation, in this work a statistical framework based on an unconventional application of Peak Over Threshold (POT) theory and a trivariate copula-based dependence modeling is proposed to describe the joint behaviour of rainfall event characteristics and to estimate multivariate return periods.
Rainfall events are extracted from sub-daily pluviometric time series using the concept of the inter event time definition (IETD) and characterized in terms of duration, mean intensity, and cumulative depth. Suitable marginal distributions are identified for each variable and finally events exceeding predefined thresholds are analysed to assess their frequency of occurrence. Dependence structure among event characteristics is modelled using multivariate copula framework capable of capturing complex, non-linear relationships and tail dependences. The fitted model is then used to simulate synthetic rainfall events and to compute joint exceedance probabilities in the trivariate space. Multivariate return periods associated with compound extreme events are derived and visualized, highlighting the importance of dependence in the assessment of rainfall severity.
The proposed methodology wants to provide a robust and flexible tool for the probabilistic characterization of compound rainfall extremes and represents valuable support for flood risk assessment and hydrological design in complex urban settings.
How to cite: Paradiso, G. and Ganora, D.: A new copula-based approach for storm events analysis to support urban catchment modeling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9324, https://doi.org/10.5194/egusphere-egu26-9324, 2026.
To enhance flood risk assessment and management, particularly in catastrophe models used for estimating potential losses, generating realistic extreme rainfall scenarios is crucial for accurate flood mapping. Current probabilistic rainfall models primarily focus on either single-station analysis or the spatial dependence of static rainfall properties. However, these approaches often fail to capture the dynamic spatiotemporal characteristics of short-duration extreme rainfall events—the primary drivers of urban flooding. In this study, we propose a novel spatiotemporal stochastic rainfall generator to simulate rainfall sequences at multiple stations while preserving spatial correlations and realistic temporal dynamics at the same time.
The generator is calibrated and applied in Hong Kong, a densely urbanized and flood-prone coastal city, using hourly in-situ observations from 141 stations for 1984–2017. Although operating at hourly resolution, the model consistently reproduces rainfall statistics across 1–24 h accumulation durations. It closely matches the statistical characteristics of historical rainfall, achieving Nash–Sutcliffe efficiency (NSE) values of 0.939–0.969 for the top 10% of events, and captures the spatial patterns of extremes with a Pearson correlation of 0.831.
Hydrodynamic simulations further demonstrate that the realistic temporal variability produced by the proposed generator leads to average flood depth differences of 18.1% and 25.8% compared with the simplified exponential and constant hyetograph scenarios, respectively. Overall, the results underscore the importance of representing realistic short-term rainfall variability in stochastic rainfall modeling to support robust flood risk assessment.
How to cite: Li, Y., Zhang, L., and He, J.: A Multi-Site Spatiotemporal Stochastic Rainfall Generator for Realistic Rainfall Generation , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10586, https://doi.org/10.5194/egusphere-egu26-10586, 2026.
Climate change is expected to increase both the frequency and intensity of extreme rainfall events, which poses a particularly high risk to urban areas due to their high levels of impervious surfaces and population density. Consequently, surface flooding is likely to intensify in the future, highlighting the importance of assessing climate change impacts in urban flood risk management. Design rainfall based on Depth-Duration-Frequency (DDF) curves is commonly used to assess flood risk, while climate change effects are incorporated by applying a rainfall uplift allowance to represent future scenarios. However, this approach typically assumes spatially uniform rainfall over the simulation domain, which can misrepresent storm movement as well as the timing and location of rainfall peaks, thereby compromising the accuracy of flood risk assessment. To address this limitation, it is important to use temporally and spatially variable rainfall as input to flood risk assessments. In this study, a temporally and spatially variable rainfall generator is developed, which generates spatial-temporal design rainfall events from Depth-Duration-Frequency (DDF) curves. To ensure that the generated rainfall realistically represents observed storm characteristics, the parameters of the rainfall generator are derived from historical weather radar observations. The generated events are used to drive hydrodynamic flood models to evaluate flood impacts in the West Midlands, UK, under climate change. By producing more realistic design storms, the proposed approach provides a basis for more reliable flood mapping and risk-informed adaptation planning at city-scales.
How to cite: Xia, X., Li, Q., and Ferranti, E.: Evaluating urban flood impacts under climate change using temporal-spatially varied design rainfall, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15291, https://doi.org/10.5194/egusphere-egu26-15291, 2026.
Rapid urbanization in India has resulted in extensive conversion of natural and agricultural land into impervious surfaces, altering surface energy and water budgets and resulting in the intensification of urban heat island and flood risks. However, most land surface models either consider cities as natural vegetation or rely on simplified urban parameterizations, leading to large uncertainties in simulating urban hydro-climatic processes. In this study, we develop an integrated physics-based framework by coupling an urban module with the Variable Infiltration Capacity (VIC) model to explicitly represent urban canyons, building morphology, and associated physical processes across Tier-I and Tier-II Indian cities. The framework integrates multiple large-scale datasets, including meteorological forcing, hydrological observations, satellite products, and building-height information from the Global Human Settlement Layer and Google-Microsoft Buildings data. Model simulations are evaluated against MODIS land surface temperature and observed streamflow from the Central Water Commission, and sensitivity analyses are performed to identify key urban controls on hydro-thermal responses. The results provide a robust basis for quantifying urban-induced modifications to energy and water fluxes and offer actionable insights for assessing heat and flood risks to support climate-resilient urban planning in Indian cities.
How to cite: Kumar, M. and Mishra, V.: Quantifying Urban Impacts on Surface Energy and Water Budget over Indian Cities, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16431, https://doi.org/10.5194/egusphere-egu26-16431, 2026.
Urban hydrological systems are highly sensitive to precipitation variability at fine spatial and temporal scales, yet such variability remains poorly characterized due to limited high-resolution observations. Here, we analyze extreme rainfall in equatorial Singapore using a uniquely dense observational network, including 122 rain gauges at 5-min resolution (2020–2024), and long-term hourly gauge records (1980–2024), which are then combined with X-band radar data at 100-m and 5-min resolution to produce a high-resolution gridded rainfall reanalysis. We use this new dataset to quantify the changing space–time organization of extreme convective rainfall over 45 years. Results show that extreme convective rainfall in this tropical urban environment is more highly localized and short-lived than previously thought, with spatial and temporal correlations halving over just 1.6 km and 4 min. In response to climate warming, the total rainfall amount, spatial extent, and temporal persistence of extreme events have increased, while peak rainfall intensities remained largely stable, likely due to limitations in local humidity supply. Such compact storm structures challenge the representativeness of sparse rain gauge networks and underscore the need for high-resolution analysis in tropical regions.
How to cite: Zhuang, Q., Peleg, N., Prein, A., Babovic, V., and Fatichi, S.: High-resolution analysis of convective rainfall properties in a tropical city, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4609, https://doi.org/10.5194/egusphere-egu26-4609, 2026.
Posters virtual: Tue, 5 May, 14:00–18:00 | vPoster spot A
EGU26-7784 | Posters virtual | VPS8
Research on the Adaptation Strategies of Urban Stormwater Drainage to Increased Rainfall Due to Climate ChangeTue, 05 May, 14:00–14:03 (CEST) vPoster spot A
A 2024 climate change study in Taiwan indicated an increase in rainfall of approximately 10-35%, causing flooding in some urban areas where stormwater drainage systems exceeded their original design protection standards. Furthermore, urban stormwater drainage systems improvements in Taiwan often face complex and intertwined spatial issues related to road traffic and underground utility lines, making rapid engineering improvements difficult. Therefore, to address the threats already posed by climate change, the use of big data monitoring of urban areas and surrounding regions, along with rapid AI-powered algorithms for drainage systems, is imperative. The New Taipei City Government, in order to manage urban water information, has developed a series of adaptation strategies for its drainage system. These strategies address environmental factors such as drainage sections affected by tides and storm surges, rainfall characteristics in nearby mountainous areas, and sections with gates and pumping stations that cannot drain by gravity. The aim is to lower urban drainage levels to prevent flooding and shorten flooding duration. This includes practical operational recommendations and early flood warnings. The method is based on historical practical experience and AI-generated water level forecasts to conduct drainage system decision analysis and management value setting. It combines real-time rainfall data from the Internet of Things, road flooding sensors, road CCTV, stormwater sewer water levels, and pumping station water levels. The data used includes actual data from the past 3 hours, forecasted rainfall for the next 6 hours, tidal changes, and real-time water level information at various monitoring locations to formulate adjustment strategies. Synchronous information is released within the drainage system to systematically set stormwater sewer water levels, treating stormwater sewers as flood retention spaces for monitoring and water level control. Based on operational experience gained from the past 3 years of implementation, this method will be used in the future to address the threats posed by increased rainfall due to climate change and to formulate urban flood control strategies to reduce disaster losses.
How to cite: Yang, S., Song, D., Huang, M., Pan, J., Wang, X., Chen, C., and Yeh, K.: Research on the Adaptation Strategies of Urban Stormwater Drainage to Increased Rainfall Due to Climate Change, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7784, https://doi.org/10.5194/egusphere-egu26-7784, 2026.
