Orals: Wed, 30 Apr | Room 2.31
Real-time ensemble flow forecasting systems are increasingly employed to manage hydrological risks associated with extreme weather events and a changing climate. The effectiveness of flow forecasting systems hinges on their ability to deliver accurate, robust, and timely predictions, despite inherent uncertainties in their components. The assessment of the system’s performance after a long-term application is essential to establish its value as a reliable reference tool for stakeholders in hydrological management and decision-making.
This study presents an approach to evaluate the performance of a regional ensemble flow forecasting system operational in real time since June 2020 over the region of Catalonia (NE Spain). The system generates flow forecasts for all gauging stations managed by the Catalan Water Agency using a modified version of the HBV rainfall-runoff model, with rainfall inputs combining QPE (blending radar and rain gauge observations) with the 52-member ensemble precipitation forecasts produced by the European Centre for Medium-range Weather Forecast (ECMWF). The analysis spans a four-year period and focuses on significant rainfall events, enabling a comprehensive analysis of the system’s accuracy, robustness, and timeliness as a function of lead time. Both deterministic metrics and probabilistic scores are applied to evaluate the quality of the flow forecasts. The evaluation focuses on assessing the impact of different sources of uncertainty using different discharge references: (1) observed flow; (2) flow simulated with the full series of rainfall observations; (3) flow simulated with no-rainfall forecasts. Additionally, key sources of uncertainty, such as rainfall variability, catchment response and forecast initialization errors, are identified to inform targeted system enhancements and improve forecast reliability.
How to cite: li, X., Berenguer, M., Corral, C., Park, S., and Sempere-Torres, D.: Performance Evaluation of a Real-time Regional Ensemble Flow Forecasting System in Catalonia: Accuracy, Robustness, and Timeliness, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-837, https://doi.org/10.5194/egusphere-egu25-837, 2025.
Flood frequency analysis is crucial for understanding and mitigating the risks of extreme flood events. However, traditional methods often assume stationarity and fail to account for the complex physical phenomena driving changes in flood behaviour. This study addresses the challenge of nonstationary multivariate flood frequency analysis by incorporating multiple covariates that represent key physical processes influencing flood variables, such as peak discharge, volume, and duration. By leveraging advanced statistical methods, including copula-based modelling and covariate selection techniques, we provide a robust framework for analysing the dependencies and dynamics of flood variables under changing climatic and hydrological conditions. Applied to the Barakar River Basin, our framework identifies significant nonstationary trends influenced by covariates such as precipitation intensity, land-use changes, and soil moisture. Results reveal that the 100-year joint return period of extreme flood events has decreased significantly from (219,247m3 8264.58m3/s) in the stationary case to (175,881m3 7,241.52m3/s) in the nonstationary case for the volume-peak pair. Similarly, for the duration-volume pair, the stationary 100-year return period supersedes the nonstationary 100-year return period from (21.41days 149,776m3) to (21.44days 159,189m3). Furthermore, the 100-year return level under stationary conditions (8150.44m3/s 30.23 days) is notably higher than the nonstationary equivalent (6995.5m3/s 25.09 days). The proposed methodology enhances the reliability of flood risk assessments by addressing the temporal evolution of key flood variables. Although not directly focused on early warning systems, the insights from this study can inform the development of probabilistic flood forecasting models and improve decision-making processes for disaster preparedness. By integrating physical drivers into multivariate flood frequency analysis, this work contributes to a deeper understanding of nonstationarity in flood regimes. The findings provide valuable implications for designing more adaptive and region-specific flood forecasting and warning systems, ultimately supporting global efforts to mitigate the impacts of extreme hydrometeorological events.
How to cite: Ankush, A., Goel, N. K., and Vinnarasi, R.: Advancing Multivariate Flood Frequency Analysis Under Nonstationarity: Implications for Flood Forecasting Systems, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1102, https://doi.org/10.5194/egusphere-egu25-1102, 2025.
Effective monitoring and forecasting of flood events are key aspects of early warning systems, especially in areas susceptible to frequent floods. In this context, Artificial Intelligence (AI) techniques have proven to be a strong tool for enhancing such systems because they can capture non-linear processes of flood genesis. AI models make accurate predictions with minimum processing time, thus providing a strong alternative to nowcasting. However, the quality and quantity of monitoring stations and the black-box nature of machine learning (ML) models hamper the development of efficient and adaptable Early Warning Systems (EWS). Accordingly, this study aims to investigate the impact of data quality and quantity on the performance of data-driven flood forecasting models built upon eXplainable AI (XAI) and Physically Informed (PI) approaches. Intending to develop a predictive analysis of stage level in this flood-hit city of Piracicaba, the Piracicaba River catchment had 18 gauging stations with 10-minute time-step rainfall and stage monitoring used for model resiliency checks based on the MLP network. This included defining the structure of the model and the input variables determined by previous studies on flood wave propagation times prior to training and testing the model. This approach considered such hydrological aspects as incorporation into the machine learning framework. A deterioration algorithm was developed to simulate the gradual introduction of gaps in the stage and rainfall time series (10% to 100%), designed to assess the impacts of failures and monitoring errors on model prediction. This methodology provides a way to assess how the ML model would be able to handle misinformation and failures in flood predictions, while at the same time, drawing a line of priority regarding the stations to maintain the effectiveness of EWS. XAI enabled us to assess the hydrological aspects behind the models’ performance and the stations’ importance, which are crucial pieces of information for planning maintenance campaigns and allocating budget. Moreover, improvement in the monitoring network is possible by defining places for installing new sensors based on the physical aspects behind runoff onset and flood propagation. Therefore, PI and XAI are central to enhancing EWS under changing climate because they incorporate knowledge of hydrological dynamics into station selection and ML model development.
How to cite: Solha, P., Perdigão, R., Brentan, B., Menapace, A., Eleutério, J., and Rodrigues, A.: Optimizing Flood Monitoring Networks Using eXplainable AI and Physical Informed approaches: A Case Study of the Piracicaba River Basin in Brazil, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1842, https://doi.org/10.5194/egusphere-egu25-1842, 2025.
In April 2022, the city of Durban in South Africa experienced one of its most devastating floods in history, with 300 mm or more recorded in the 24-hour period on 11-12 April. This extreme event led to approximately 430 fatalities and infrastructure damages estimated at 1 billion US dollars.
The primary objective of this study was to calibrate the hydrological model (HEC-HMS) and hydraulic model (HEC-RAS) by using observed precipitation data, high-resolution Digital Elevation Model (DEM) of 10 m, soil maps, land use and landcover (LULC) maps, and hydraulic structures characteristics of one of the affected areas - Inanda. The methodology involved simulating flood extent, inundation levels and flow hydrographs and subsequently comparing these outputs with observed data obtained from rainfall stations, river gauges and weirs.
The simulated results indicated a discharge of 570 m3/s in the main channel of the Mngeni River. The flood peaked on April 12, 2022, between 11:00 PM and 12:00 PM, aligning closely with observed peak flow discharges recorded by local authorities. The study identified that areas around tributaries of the Mngeni River were more severely impacted than the main channel, with flood inundation depths reaching up to 5 metres. Based on aerial imagery, the Durban Harbour experienced the most extensive flooding in terms of spatial coverage, with water depths of approximately 1 metre. This flooding led to significant disruptions, including the closure of major highways and the port.
This study successfully calibrated critical parameters required for the development of a Flood Early Warning System (FEWS), reducing the discrepancies between observed and simulated data. To further enhance the accuracy and reliability of future flood prediction models, the study recommends the creation of high-resolution soil maps and more detailed LULC maps specifically tailored for flood-prone. Such advancements could prove to be crucial for strengthening flood preparedness and mitigating risks in similarly vulnerable regions.
How to cite: Byaruhanga, N. and Kibirige, D.: From Chaos to Preparedness: Recreating the Durban, April 2022 Floods using Enhanced Hydrological and Hydraulic Modelling for Flood Early Warning Systems., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4827, https://doi.org/10.5194/egusphere-egu25-4827, 2025.
Observed runoff is, in principle, an ideal observation to be used for updating the moisture states in a hydrological model because runoff is 1) integrated catchment scale information (unlike precipitation and snow water equivalent, 2) usually well measured 3) frequently measured and 4) a direct measurement of what the model is supposed to predict. However, the model structure needs to be such that the discrepancies between observed and simulated runoff can easily and unambiguously be translated into altered moisture states. In this study, we have altered the subsurface moisture state in the Distance Distribution Dynamics (DDD). If the model underestimates runoff, more water is added as an extra precipitation event and if the model overestimates runoff, we subtract water by having an extra evapotranspiration event. The magnitude of the precipitation/evapotranspiration event is the sum of small increments (+- 0.5 % of the subsurface storage) which are added or reduced from the models’ subsurface storage until observed and simulated runoff are equal. In the DDD model, precipitation and evapotranspiration are distributed in time according to unit hydrographs (UH) estimated using the calibrated subsurface celerities and the distance distribution describing the distances from points in the hillslopes to the river network. The UHs can be seen as sets of weights distributing the input in time. In such a way the added and subtracted water influences the simulated runoff for a period of time determined by the temporal scale of the UHs which vary from catchment to catchment. The immediate correction on runoff is only due to a fraction of the added/subtracted water which is determined by the UHs. We have tested the updating procedure for 25 Norwegian catchments of different sizes and located all over Norway and in different climatic zones. The model is run on 3h temporal resolution and we tested the efficiency of updating for two levels of runoff; i) if observed runoff is higher 2x mean annual discharge (MAD) and the discrepancy between simulated and observed is more than 20% and ii) if observed runoff is higher than the mean annual flood (MAF) and the discrepancy between simulated and observed is more than 20%. On average for the 25 catchments, updating had a positive effect on the root mean square error for lead times less than 33 hours for events higher than MAF and for lead times less than 42 hours for events higher than 2xMAD. For lead times less than 18 hours, 69 % of the updates improved the runoff forecasts for events higher than MAF and 70 % for events higher than 2xMAD.
How to cite: Skaugen, T. and Bakke, S. J.: Improving runoff forecasts by using observed runoff to update the subsurface moisture state in the DDD model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6984, https://doi.org/10.5194/egusphere-egu25-6984, 2025.
The heavy rains and floods that struck southern Brazil in May 2024 highlighted the vulnerability of the population to extreme hydrological disasters, resulting in 176 confirmed fatalities, around 40 people missing, and over 422,000 individuals displaced. Flood Early Warning Systems (FEWS) are crucial tools for reducing flood-related damage and fatalities. The accurate prediction of flood peaks and their timing is essential for effective evacuation planning. This study proposes an enhanced Long Short-Term Memory (LSTM) model for runoff prediction, incorporating novel loss functions that prioritize flood periods (e.g., peak flow and peak timing) , while reducing the importance of normal and low flow periods. Additionally, the study evaluates the model’s performance across various forecast horizons (0 – 24 hours), aiming to understand how forecast accuracy varies with increasing forecast horizons. Using Piracicaba City in Brazil as a case study, 10-minute flood stage and rainfall data from 18 upstream stations (2018–2023) were utilized to predict flood stages at the target station using the enhanced LSTM model. The results compare LSTM predictions with traditional loss functions (e.g., mean absolute error) to those using the newly designed loss functions, evaluated through several metrics to assess improvements in flood prediction accuracy. By analyzing the model’s accuracy across different forecast horizons, the study provides valuable insights into the optimal lead time for issuing warnings in Flood Early Warning Systems.
How to cite: Bezerra, R., Solha, P., Rodrigues, A., Brentan, B., and Eleutério, J.: Enhanced LSTM Model for Flood Forecasting Systems: A Case Study of the Piracicaba River Basin in Brazil, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12740, https://doi.org/10.5194/egusphere-egu25-12740, 2025.
The increasing frequency and intensity of extreme weather events highlight the urgent need for more accurate flood forecasting to mitigate the devastating impacts on communities and ecosystems. The DestinE programme aims to address this challenge by enhancing the accuracy of meteorological forecasts, particularly for extreme weather events, through higher resolution, which in turn is expected to improve flood forecasting capabilities. Nearly one year of Global Extremes Digital Twin (G-EDT) simulations, providing high-resolution meteorological data, has been generated as part of the programme. These simulations drive ECMWF’s Land Surface Modelling System (ecLand) producing runoff generation, which is then routed through the Catchment-based Macro-scale Floodplain model (CaMa-Flood) to simulate river flow. A key challenge, however, is evaluating the performance of these new high-resolution prototype systems, especially given the limited availability of long-term hindcast data for evaluation. With only a short data period available, it becomes challenging to robustly assess the predictive skill of these models in forecasting flood events.
To overcome this limitation, we have developed a methodological framework that facilitates a rigorous evaluation of these high-resolution systems, enabling meaningful assessments of their forecasting skill despite the constrained data period, focusing on the potential of the G-EDT to improve hydrological forecasting in comparison with the forecasts produced by ECMWF’s operational system at lower resolution. Specifically, the framework investigates:
- Significance Testing: Statistical testing (e.g. Student t-test) are employed to assess whether observed differences in forecast skill are statistically significant and not a result of sampling variability.
- Multi-Metric Evaluation: Beyond traditional performance scores like Kling-Gupta Efficiency (KGE), the analysis incorporates flow duration curves, bias decomposition, and regional variability assessments to capture a broader range of hydrological behaviours and extremes.
- Threshold Definition: Using six years of historical simulations, thresholds for different return periods are calculated to enable flood characterisation of different severity.
- Global and Regional Assessments: The framework evaluates performance at both global and regional scales, considering spatial variability in hydrological processes and the availability of ground-truth observations for validation.
Moreover, the results lay the foundation for a continuously evolving evaluation system for the G-EDT, designed to adapt as longer datasets become available.
The results of this analysis offer valuable insights into the benefits and limitations of using high-resolution simulations for global hydrological forecasting, particularly in the context of extreme events, and will inform future improvements of the global Extremes Digital Twin in DestinE (G-EDT).
How to cite: Tanguy, M., Arduini, G., Chevallier, M., Denissen, J. M. C., Dueben, P., Gascon, E., Haiden, T., Mazzetti, C., Mastrantonas, N., Matthews, G., Morrison, O., Prudhomme, C., Rüdiger, C., Sandu, I., Vannière, B., Wortmann, M., and Zsoter, E.: Enhancing Global Flood Forecasting: A Methodological Framework for Assessing High-Resolution Simulations in the DestinE G-EDT, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17195, https://doi.org/10.5194/egusphere-egu25-17195, 2025.
Reliable seasonal rainfall forecasts are essential for improved early warning of large-scale droughts in West Africa but remain a major challenge for national meteorological services in the region. This study presents a statistical post-processing approach for improved probabilistic forecasting of seasonal rainfall amounts for the West Africa region. The novel approach relies on a circulation pattern approach to incorporate seasonal and interannual dynamics of West African monsoon processes, such as the Saharan Heat Low or the Tropical Easterly Jet, in combination with a simple logistic regression to predict rainfall amounts. The approach was tested in a reanalysis mode (1960 to 2010) for several climatic regions in West Africa using ERA5 in parallel to a regional station-based precipitation dataset and state-of-the-art global precipitation products such as CHIRPS. In addition, the statistical approach was applied to a hindcast period (1981 – 2023) for the peak monsoon period in West Africa and compared with the raw precipitation forecasts of ECMWF's SEAS5 and the real-time forecasts of the national weather services in West Africa subjectively produced as part of the West African Regional Climate Outlook Forum (WARCOF, 1998-2023). The study shows that the circulation pattern model outperforms both WARCOF and the raw rainfall forecasts of SEAS5. While WARCOF and SEAS5 show some forecasting skill for above and below normal conditions, both models show common limitations often observed in seasonal forecasting, such as lack of sharpness and a strong over-forecasting of near-normal conditions due to a risk aversion of the WARCOF experts. The circulation pattern-based approach provides much more accurate precipitation forecasts with greater reliability. Furthermore, a theoretical assessment of the economical value shows that the circulation pattern approach provides positive economic values for a wide range of potential users making it more suitable for forecasting rainfall amounts in drought situations in this region compared to SEAS5 and WARCOF. This study therefore provides a basic statistical post-processing approach for producing more accurate operational seasonal rainfall forecasts in the long-term for this challenging region.
How to cite: Bliefernicht, J., Rauch, M., Sawadogo, W., Sy, S., Waongo, M., and Kunstmann, H.: Seasonal rainfall forecasts for drought situations in West Africa, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13758, https://doi.org/10.5194/egusphere-egu25-13758, 2025.
The Alps are a crucial water reservoir for over 170 million people in Europe, storing water as snow and ice during the cold season and releasing it during the warm season. However, climate warming is causing earlier snowmelt and glacier shrinkage, leading to a mismatch between water availability and peak demand, thereby increasing the risk of water shortages and conflicts among key users such as agriculture, energy, and tourism. In this framework, skillful seasonal climate predictions combined with snow-hydrological modeling might help in the early warning of water shortages. Recent studies show advancements in the seasonal prediction of mountain snow water equivalent (SWE); however, it remains uncertain whether these improvements translate into accurate predictions of streamflow and water availability.
Within the PRIN-2022 SPHERE1 project, we developed a novel modeling chain which takes advantage of the state-of-the-art bias-corrections methods, downscaling techniques, and snow-hydrological modeling tools to model snowpack evolution, river discharge, and water availability in Alpine river basins. The modeling chain comprises the S3M snow model and the HMC Continuum hydrological model. S3M is a spatially distributed cryospheric model that simulates snow and glacier mass balance, while HMC is a spatially distributed hydrologic model that solves the mass and energy balance of vegetation and soils. The modeling setup, validated in previous projects, operates on a regular grid with a 1 km spatial resolution and an hourly time step over the Po River basin. For this domain, ERA5 reanalysis meteorological variables were used to generate the forcing for the baseline run of the modeling chain. ERA5 inputs, originally at 0.25° spatial resolution, were appropriately bias-corrected and downscaled to 1 km. A 30-year baseline simulation was then generated to reconstruct the historical evolution of mountain snowpack (in terms of SWE and snow depth), meltwater runoff, and streamflow.
We present here an initial evaluation of the snow-hydrological modeling chain over the study area for the period 1991-2020. Simulations of snow depth, SWE and river discharge were compared against various available observations from snow gauges and Italian regional hydrological networks, in terms of bias, RMSE and correlation. As a next step, the modeling chain will be extended to run using seasonal forecasts from the Copernicus seasonal prediction systems (ECMWF S51, MF S8, CMCC S35, DWD S21) to generate retrospective seasonal forecasts of snow and hydrological variables over the same study domain. The forecast skill of the modelling chain will be evaluated for starting dates November 1st and May 1st, to assess the possibility of anticipating water availability several months in advance for the winter and summer seasons, respectively.
1Progetto di Ricerca di rilevante Interesse Nazionale (PRIN-2022): Seasonal Prediction of water-availability: enHancing watER sEcurity from high mountains to plains (SPHERE)
How to cite: Lorenzo, M., Pourjavad Shadbad, E., Avanzi, F., Libertino, A., von Hardenberg, J., and Terzago, S.: Modeling snowpack evolution and water discharge in the Po River basin at 1 km resolution: a retrospective analysis (1991-2020), EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12802, https://doi.org/10.5194/egusphere-egu25-12802, 2025.
Flooding has increasingly posed significant challenges in the Western Cape, South Africa, with the September 2023 floods in Franschhoek underscoring the vulnerability of the region to extreme rainfall events. During this event, the area received over 220 mm of rainfall within 48 hours, resulting in extensive flooding that inundated approximately 500 hectares, displaced over 1,000 residents, and caused substantial damage to infrastructure. This study developed an integrated Flood Risk Information System (FRIS) designed for flood-prone regions in the Western Cape, utilizing Earth Observation (EO) technologies, hydrological modelling, and Geographic Information Systems (GIS).
The system integrated historical flood data, municipal hydrological observations, and real-time environmental variables, including rainfall, river discharge, and soil moisture, to enhance flood risk prediction, monitoring, and response. Hydrological modelling was conducted using the HEC-HMS (Hydrologic Engineering Center’s Hydrologic Modeling System) and SWAT (Soil and Water Assessment Tool) models. Machine learning algorithms, including Random Forest (RF) and Gradient Boosting Machine (GBM), were implemented to predict flood probabilities. Model outputs were validated against observed data from the local municipality, which included flood extent maps and river discharge measurements.
The system demonstrated high accuracy in predicting flood extents, with the HEC-HMS model achieving a Nash-Sutcliffe Efficiency (NSE) of 0.88 and a Root Mean Square Error (RMSE) of 12% compared to observed discharge data. The machine learning models yielded flood prediction accuracies of 87% (RF) and 91% (GBM) when compared to observed flood extents. Google Earth Engine (GEE) was used to process large EO datasets, allowing for real-time flood mapping and risk analysis.
The FRIS proved instrumental in being able to model the September 2023 floods by providing accurate predictions and mapping, enabling disaster management agencies to target evacuation efforts and allocate resources effectively. However, further improvements are planned, including incorporating finer-resolution rainfall and topographic data, expanding the system’s spatial coverage, and integrating socio-economic indicators to assess community vulnerability better.
This study highlighted the potential of combining EO, GEE, GIS, and advanced hydrological models in improving flood risk management. The FRIS provides a powerful framework for mitigating flood impacts and protecting vulnerable communities, aligning with broader efforts to enhance climate adaptation and resilience in the Western Cape.
How to cite: Kibirige, D.: Integrated Flood Risk System for the Western Cape: Lessons from the September 2023 Floods, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-9855, https://doi.org/10.5194/egusphere-egu25-9855, 2025.
Floods pose a significant threat to urban areas due to their high population densities and extensive infrastructure, a vulnerability exacerbated by climate change, rapid urbanization, and the proliferation of impermeable surfaces. While traditional flood prediction efforts have focused on maximum inundation depths, dynamic flood inundation mapping has gained prominence for its ability to provide detailed insights into flood timing, duration, and progression, which are critical for effective emergency response, infrastructure planning, and resilience-building. The integration of high-resolution Digital Elevation Models (DEMs) has improved modeling accuracy by capturing intricate urban topographies, but this advancement has introduced substantial computational challenges, particularly for large-scale, fine-resolution simulations using physics-based hydrodynamic models. Convolutional neural networks (CNNs), particularly U-Net, have shown promise in flood prediction due to their ability to handle complex segmentation tasks and varying input sizes; however, scaling these models to handle large datasets with meter-scale resolutions remains computationally intensive. Addressing this challenge, this study develops a novel approach to predict dynamic flood maps for a large urban area at 10-meter resolution (~10⁶ cells) by dividing the area into smaller tiles for U-Net training, leveraging a comprehensive rainstorm-inundation database (200 cases) created through 2D hydrodynamic simulations, and integrating results into a surrogate model. This innovative framework delivers accurate and rapid predictions of flood dynamics, including spatial extent, depth, and temporal evolution, providing essential tools for urban flood risk management and mitigation strategies. For training, validation, and testing of the U-Net model, 160, 30, and 20 cases were used, respectively. The RMSE, CSI, and POD metrics were used to evaluate the model's performance on the validation and test datasets. The results show high performance for validation with RMSE, CSI, and POD values of 0.015, 0.92, and 0.74, respectively, and for testing with values of 0.017, 0.90, and 0.81, respectively.
How to cite: Nakhaei, P., Li, R., and Ni, G.: High-Performance Full-Scale Urban Flood Prediction: A Scalable Solution for Dynamic Inundation Mapping , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2425, https://doi.org/10.5194/egusphere-egu25-2425, 2025.
In recent years, China has frequently been affected by urban flooding especially in megacities with populations exceeding ten million, which are concentrated in the East Asian monsoon climate zone. Hydrodynamic models are effective tools for predicting flood disasters. However, large urban areas with hundreds of square kilometers result in a high computational burden for modeling. This study develops a high-resolution pluvial flood forecasting model based on a national supercomputing center. The model effectively leverages the heterogeneous architecture of supercomputer and can access precipitation forecast data from ECMWF to achieve real-time predictions for mega cities. Hydraulic modeling is based on the diffusion wave equation, discretized by finite difference method. The model divides the study area into several equal rectangular partition depending on preset spatial parameters. Structured grids are defined. Employing MPI (Message Passing Interface) as parallel tool, one CPU core and one DCU (Deep Computing Unit) are used for calculation of each partition. The model is validated using two rainfall and water depth datasets collected from Tsinghua Campus. Taking Chengdu, with an area of approximately 600 km2 and a resolution of 1 m, as the study area, five partitioning schemes are set up to compare the computing time for CPU-only and CPU+DCU computation. Model performance is tested by simulating 3-hour surface runoff process with over 600 million grids. The results show that when over 6000 CPU cores and 6000 DCUs are used, the model can complete the simulation in 10 minutes. It represents a speedup of about 5 times compared to equal number of CPU cores computation without DCUs, and approximately 500 times faster than using 64 CPU cores. The model demonstrates near-linear speedup when using only CPUs, suggesting that it is approximately 30,000 times faster than single CPU core computation. By analyzing computational time of each process during model execution, hydrodynamic calculation is faster by tens of times on DCUs than on CPUs. Message passing and input/output time on CPUs will impact the scalability of the model.
How to cite: Chen, T., Sun, J., Zhang, Z., and Lin, B.: Real-time pluvial flood forecasting model for mega cities based on a heterogeneous supercomputer, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1851, https://doi.org/10.5194/egusphere-egu25-1851, 2025.
The frequency of unprecedented localized torrential rainfalls, such as the 2024 heavy rainfall event in South Korea, has been increasing due to climate change. Simultaneously, urbanization has intensified the development of underground spaces and excavation sites due to population concentration. This study assesses the risks associated with evacuation routes in various types of underground spaces during extreme rainfall events and develops a real-time flood prediction and early warning system that incorporates evacuation lead time analysis.
A testbed was established in a watershed that experiences chronic flooding caused by the combined effects of seawater intrusion, external runoff, and internal drainage issues. Using Arc-GIS, a detailed topographic model was constructed for the region. To analyze dynamic flood risks under real-time rainfall conditions, novel rainfall scenarios were created by combining observed rainfall data from rain gauges with predicted rainfall data from radar systems. Observed rainfall from rain gauges within the watershed, measured up to one hour prior, was distributed according to Huff’s 4th quartile pattern, while predicted rainfall for the subsequent 30 minutes was distributed using Huff’s 1st quartile pattern. These patterns were combined to simulate the worst-case scenario, representing the most challenging evacuation conditions.
These datasets provided the foundational framework for conducting two-dimensional flood simulations using XP-SWMM. The risks along evacuation routes were quantified by calculating the product of flood depth and flow velocity(hv). Furthermore, a flood risk nomograph was developed, and alert levels were defined based on the timing of risk escalation.
The real-time flood prediction and early warning system proposed in this study has the potential to be applied to flood-prone disaster zones across the country. By evaluating evacuation route risks under various rainfall scenarios, this system enables the timely transmission of evacuation alerts and warnings to minimize disaster impacts.
How to cite: Lee, S. I., Kim, J., Kim, K., Kang, J., and Jun, H.: Development of a Real-Time Flood Prediction and Early Warning System for Underground Spaces, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4004, https://doi.org/10.5194/egusphere-egu25-4004, 2025.
Climate change has significant impacts on water resources, making the study of hydrological cycle alterations essential for understanding regional climate dynamics. Hydrological models are crucial tools for quantifying these changes and assessing their implications.The Weather Research and Forecasting Hydrological Model (WRF-Hydro), is widely used to simulate regional hydrological processes. This study evaluates the performance of an enhanced version of WRF-Hydro, incorporating an overbank flow module, in simulating runoff for the Myanmar region from 2010 to 2012. The model was driven by offline forcing datasets, specifically Integrated Multi-satellite Retrievals for GPM (IMERG) and ECMWF Reanalysis 5th Generation (ERA5) precipitation products, and the simulated results were compared with observed runoff data. The findings indicate that simulations driven by IMERG precipitation data outperformed those driven by ERA5 in terms of accuracy, likely due to IMERG’s superior representation of regional precipitation patterns. Model performance was assessed by comparing simulated runoff with measurements from seven hydrological stations, where the modified model showed consistent improvements over the default model. NSE improved from −0.27 to 0.49 (default) to 0.51 to 0.62 (modified), indicating enhanced accuracy and reliability. A more detailed analysis of the water cycle reveals that the incorporation of the overbank flow module initially reduces surface runoff, which is offset by an increase in soil moisture storage, accompanied by a slight rise in underground runoff and evapotranspiration. Toward the end of the season, surface runoff increases, which can be attributed to the higher soil storage at the start of the season. These results highlight the significant impact of the overbank flow module on hydrological processes, particularly in flood-prone areas, and suggest that the modified model enhances hydrological forecasting capabilities.
How to cite: Sun, Q., Arnault, J., Laux, P., and Kunstmann, H.: Assessing the Role of Overbank Flow in Hydrological Modeling: A Case Study of Myanmar River basin Using WRF-Hydro, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6214, https://doi.org/10.5194/egusphere-egu25-6214, 2025.
Effective flood forecasting and warning systems depend on robust hydrometric networks tailored to meet the functional, geographical, and operational demands of real-time monitoring. This study outlines the ideal requirements for hydrometric data collection and transmission systems to support the National Flood Forecasting and Warning Service (NFFWS). Key considerations include spatial and temporal coverage, gauge placement in flood-prone areas, and sufficient resolution to align with hydrological models.
Recommendations emphasize the need for durable, reliable infrastructure designed for extreme conditions, adherence to international standards, and regular maintenance protocols. Priority gauges—those critical for public safety, hydrological model calibration, and flood risk management—should be safeguarded, with enhancements in accuracy, redundancy, and telemetry for real-time data transmission. Network resilience is bolstered by backup power, dual communication systems, and fault alerts.
Another critical component for flood forecasting is the sub-daily rainfall network. High-resolution, high-quality rainfall data at sub-daily intervals are essential for capturing short-duration, intense rainfall events that can lead to flash flooding. The design of a sub-daily rainfall network should prioritize strategic placement of gauges in areas with high rainfall variability and known flood risk areas. Real-time data collection and transmission are paramount to ensure timely updates for flood forecasting models. Integration of automated quality control processes can further enhance the reliability of sub-daily rainfall data.
We propose systematic reviews of network performance, integration of historical and real-time data archives, and automated quality control to improve data reliability. Advanced metadata management and API-based data dissemination enhance usability for stakeholders. These standards ensure hydrometric networks remain integral to flood forecasting, minimizing flood risks and improving public safety. The findings provide a blueprint for developing resilient, effective hydrometric networks that address the evolving needs of flood forecasting and warning services.
How to cite: Roberts, M., Canavan, J., Broderick, C., and Lawlor, R.: Optimizing Hydrometric Network Design for National Flood Forecasting and Warning Services (NFFWS), EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11597, https://doi.org/10.5194/egusphere-egu25-11597, 2025.
Hydro Tasmania is Tasmania’s main power generator and the guardian of many of its waterways with unique natural values. The water that powers the island state, as well as part of the mainland Australia, is a valuable shared resource between all Tasmanians environmentally, culturally, and economically. Hydro Tasmania needs to navigate several different objectives of providing the right amount of affordable energy to its residents, while meeting farmers’ irrigation requirements, maintaining lake levels appropriate to various types of recreational users, and protecting Tasmania’s vulnerable environment and endangered fauna.
Tasmania’s hydro power stations are mostly located in areas that historically receive high rainfall and inflows, but this is becoming less frequent with the climate change. The magnitude of the inflows our catchments collected over the last year was smallest in decades. Climate change also brings other extreme events such as flooding after drought and bushfires. The only way to successfully mitigate the impact of reduced inflows is by having an accurate inflow forecasting system. When water is scarce, optimisation and planning become an essential asset. It is required for scheduling of the power stations, for ensuring dam safety and for protecting the communities and environment. To complicate the problem further, Tasmania is located in the middle of different weather systems, and has a challenging topography with mountains and valleys, forests and plains, and it is surrounded by sea. This makes inflow forecasting at specific locations extremely challenging.
Until recently, our inflow forecasts were hydrographs based on deterministic models that were used as the best estimate across all the applications and end-users. These models are robust, and easy to interpret, but they do not provide information about uncertainties and probabilities of the extreme inflows. Our answer to the challenge is an all-purpose hydrological forecasting system. It consists of three timescales; short-term, outlook and long term, but more importantly, instead of resulting a single forecast, it provides an ensemble of 200 forecasts. This supplies the decision makers more visibility to the probability of different events which enables more optimised planning and using of water.
In this talk, we describe our new inflow forecasting system, and share experiences of how this has been a valuable asset in navigating one of the driest years in recorded history of Tasmania.
How to cite: Matala, A., Hunter, J., and Robinson, K.: Navigating low inflows – experiences of a hydro power generator, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19082, https://doi.org/10.5194/egusphere-egu25-19082, 2025.
Posters virtual: Tue, 29 Apr, 14:00–15:45 | vPoster spot A
EGU25-14431 | ECS | Posters virtual | VPS9
Data-Driven Flood Forecasting Using ANN: A Resource-Efficient Approach for High-Risk RegionsTue, 29 Apr, 14:00–15:45 (CEST) vPoster spot A | vPA.17
Flood forecasting is essential for hydrological assessment and catastrophe mitigation, particularly in flood-prone areas such as Bangladesh. Nonetheless, the direct measurement of water levels (WL) and discharge frequently encounters obstacles related to time, technological limits, and economical constraints. This study posits that flood levels can be accurately predicted utilising accessible data during flood events, employing a trained Artificial Neural Network (ANN) model. The complexity of hydrological systems, exacerbated by transboundary contributions from significant rivers like the Brahmaputra-Jamuna, hinders accurate forecasting. To tackle these problems, the study employed Artificial Neural Networks (ANN), a flexible and data-driven methodology adept at modelling non-linear relationships, to predict flood water levels with a lead time of up to seven days in Sirajganj, a district particularly susceptible to river flooding and bank erosion. Daily Data on water levels and rainfall were collected from the Bangladesh Water Development Board (2002–2015) for the monsoon season (May–October) were analysed, utilising information from four rainfall stations and six water level stations located 62–237 km upstream. The ANN model, employing a Sigmoid activation function with one to three hidden layers, indicated that augmenting the number of hidden layers provided only negligible enhancements in performance. Performance metrics, such as the goodness-of-fit (R²: 0.985–0.554), Root Mean Square Error (RMSE: 0.024–0.617), and Mean Absolute Error (MAE: 0.087–0.604), demonstrated a marginal improvement when rainfall and water level data were combined. This study highlights the efficacy of Artificial Neural Networks (ANN) in tackling hydrological prediction issues, confirming its ability to utilise readily accessible datasets to provide reliable and effective flood forecasts, thus aiding disaster preparedness and mitigation efforts in resource-limited areas such as Bangladesh.
How to cite: Das, P. and Mohib, K. M.: Data-Driven Flood Forecasting Using ANN: A Resource-Efficient Approach for High-Risk Regions, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14431, https://doi.org/10.5194/egusphere-egu25-14431, 2025.
EGU25-21199 | Posters virtual | VPS9
Devastating Flooding Despite Early Warning: Lessons Learned from the Nepal and Kenya FloodsTue, 29 Apr, 14:00–15:45 (CEST) | vPA.18
The increasing frequency and intensity of climate hazards, as emphasized by the IPCC’s Sixth Assessment Report, underscore the urgent need for effective disaster risk reduction strategies. Using the devastating floods of September 2024 in Nepal’s Kathmandu Valley, and the April 2024 floods in Kenya’s Nairobi, this study examines the persisting gaps in flood resilience despite early warnings using disaster forensics techniques. The Kathmandu floods, which were triggered by an extreme rainfall event resulting from the convergence of a low-pressure system from the Bay of Bengal and a cyclonic circulation from the Arabian Sea, caused extensive loss of life, property damage, and economic disruption in the Nakhu Khola watershed, despite timely early warnings issued by the government. In Kenya, a notable gap exists in the warning systems, whether in their issuance, dissemination, or uptake, despite the presence of advanced operational forecasting systems. Encroachment on floodplains, unplanned urbanization, and land-use changes have exacerbated vulnerability, while weak governance and poor enforcement of disaster risk management legislation has left populations and assets exposed. Additionally, risk assessment efforts are scarcely integrated into government plans or those of other stakeholders, highlighting a critical area for improvement in disaster preparedness and management.
Using the UNDRR’s forensic disaster analysis framework, this research investigates the underlying causes, risk drivers, and lessons from these events. The populations most affected are identified, including those living in floodplains, including marginalized communities, and critical infrastructure. Local investments in disaster preparedness are also critically examined for efficacy. Results highlight that while early warnings were disseminated through various channels, gaps in risk communication and community-level preparedness persisted. The findings emphasize the need for education and awareness and integrated approaches to disaster risk management that address root causes such as unplanned urban growth and environmental degradation. Empowering youth and fostering leadership in disaster risk reduction are critical to ensure climate resilient societies of tomorrow. This research contributes actionable insights to reduce vulnerability, enhance preparedness, and minimize losses in future climate hazard events in the Kathmandu Valley and Kenya, as well as similar rapidly urbanizing regions.
How to cite: Kettner, A., Gupta, A., Singh Shrestha, M., Trigg, M., Cohen, S., Hawker, L., Prades, L., Rudari, R., Salamon, P., Tellman, B., Sperna Weiland, F., and Wu, H.: Devastating Flooding Despite Early Warning: Lessons Learned from the Nepal and Kenya Floods, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21199, https://doi.org/10.5194/egusphere-egu25-21199, 2025.
