Disaster risk financing has seen a transformative approach through Anticipatory Action (AA), designed to reduce shock and impact of multiple hazards on vulnerable population. The core of AA relies on pre-agreed triggering mechanisms, that are built around impact-based forecasts (IBF) and tailored to local contexts, determining when, where, and what interventions are required. While numerous humanitarian actors have adopted AA in the recent years, they often work in silos, employing varying definitions, methodologies, and processes, which complicates and reduces opportunities for collaboration. Additionally, the lack of standardization and transparency in trigger models limits comparability and potential for scaling efforts effectively.
The ECHO-funded project, led by the Regional Anticipatory Action Working Group (RAAWG) secretariat addresses these challenges by fostering dialogue and coordination among regional actors in Southern Africa. Through stakeholder engagement and technical assessment, the project seeks to harmonize AA trigger methodologies, by developing an inventory of existing frameworks and co-designing a knowledge management platform to enhance information sharing and operational alignment.
Initial results highlight the diverse landscape of AA in the region. The project’s first phase assembled 43 anticipatory action frameworks spanning eight countries and seven hazards, uncovering a mix of hazard-based and impact-based triggers. Funding sources for these frameworks include multilateral mechanisms, pooled funds, and bilateral arrangements, reflecting the diversity of financial arrangements to support AA initiatives. Gaps were noted in accessing comprehensive technical details and past trigger activation data, which is now being addressed through targeted surveys and forms. Stakeholder interviews highlighted growing collaboration, but also the challenges that remain in navigating the various triggers and processes and accessing timely information through an integrated platform. A prototype knowledge management platform was developed and refined based on user feedback, aiming to improve transparency and coordination at both technical and operational levels.
These characterizations and stakeholders’ insights highlight critical gaps, opportunities for harmonizing trigger methodologies, and pathways for cross-agency collaboration. Building on this work, future research will explore the global landscape of AA through systematic literature review, mapping the current frameworks, assessing the operational maturity and identifying challenges and opportunities for scaling up. These findings will provide a foundation to evaluate and align technical, operational and financial aspects of AA.
How to cite: Sedhain, S., Castellana, D., Agmon, G., Rosenthal, T., Klein Holkenborg, E., van den Homberg, M., and Kerle, N.: Challenges and opportunities in the harmonization of trigger models for Anticipatory Action; a multi-hazard and multi-agency perspective, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14500, https://doi.org/10.5194/egusphere-egu25-14500, 2025.
Floods are among the most destructive natural hazards globally, with Southeast Asia being particularly vulnerable due to socioeconomic and geographical factors. Climate change exacerbates this vulnerability, increasing the frequency and intensity of flooding events and heightening the risks to millions of people and critical infrastructures. To address these challenges, disaster risk management is transitioning from traditional hazard-based to impact-based forecasting (IBF), which focuses on predicting the consequences of flood events. IBF emphasizes actionable insights, such as the number of people affected or disruptions to essential services, enabling more targeted early actions and decision-making.
This work shows the development and implementation of an operational impact-based flood forecasting and early warning system for five pilot river basins in Cambodia and Lao People's Democratic Republic (PDR). The system integrates the use of the Continuum distributed hydrological model (see Alfieri et al., 2024) calibrated with dedicated discharge measurements, 30 m resolution inundation maps generated for seven constant probabilities of occurrence with the REFLEX model (Arcorace et al., 2024), and a risk assessment model implemented for seven asset categories including direct economic damage on built-up, population affected, crop land affected, grazing land affected, roads affected, education facilities and health facilities affected. The system is updated twice daily with four different global and limited area numerical weather predictions (NWP), enabling forecasting of flood impacts up to five days ahead of their occurrence and thus assisting hydro-meteorological forecasters and disaster managers in their daily monitoring.
A key feature of this system is a co-production platform for generating standardized warning bulletins, allowing rapid dissemination of actionable information. This automation significantly reduces the time required for decision-making and prioritization during emergencies, enhancing disaster response capabilities. By aligning with international initiatives like the Sendai Framework and the Early Warnings for All, this system represents a critical advancement in flood risk management, promoting resilience and minimizing disaster impacts in Southeast Asia.
How to cite: Alfieri, L., Bucherie, A., Libertino, A., Campo, L., D'Andrea, M., Ghizzoni, T., Gabellani, S., Massabò, M., Rossi, L., Rudari, R., Sisouphanthavong, B., Sothy, H., Trasforini, E., Tripathi, R., and Watkins, J. T.: Impact-based flood early warning in Lao PDR and Cambodia, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4063, https://doi.org/10.5194/egusphere-egu25-4063, 2025.
Roads and healthcare facilities are critical in providing populations with basic health services. However, such critical infrastructures and the services they provide can be greatly disturbed when major natural hazards hit. Knowing which roads are still functional and where population could suffer from a loss of access to basic services before the unfolding of a hazardous event could be of great help for local authorities and for actors involved in disaster preparedness and relief.
We develop an impact forecast model aiming at predicting 1) which roads and healthcare facilities become nonfunctional in the event of a flood hazard and 2) where are populations at risk of losing access to health services.
We combine the open-source weather and climate risk assessment model CLIMADA with flood forecasts to estimate the damage to roads and healthcare facilities and their resulting loss of functionality. We assess how the flood damage results in loss of access to health services for the population using a service-access model. We select several case studies of floods in the Greater Horn of Africa to illustrate the model's skill and fit of purpose. We use remote sensing data from the United Nations' disasters' charter mission and text reports from the International Federation of the Red Cross to compare modeled with observed impacts. We use an uncertainty and sensitivity quantification module available within the CLIMADA platform to study the sources of uncertainty in the impact forecasts, varying the input flood forecasts, exposures layers, impact functions, and parameters of the service-access module.
This research illustrates the potential benefits and challenges of a people-centric impact forecast in the context of flood hazard and showcases the development and calibration of an impact forecast model using open-source data and models.
How to cite: Severino, L., Mühlhofer, E., Kalladath, N., Amdihun, A., and Bresch, D. N.: People-centric impact forecasts: Predicting flood-induced loss of access to health services in the Greater Horn of Africa, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15886, https://doi.org/10.5194/egusphere-egu25-15886, 2025.
Riverine floods are among the most destructive and frequent natural hazards in Mali. To mitigate their impacts, the Mali Red Cross has implemented an anticipatory action mechanism that activates early responses when predefined triggers are met. Currently, the Early Action Protocol (EAP) relies on real-time water level observations from the National Directorate of Hydraulics (DNH) of Mali. Triggers are activated when upstream water levels exceed thresholds, which are extrapolated downstream along the river network using estimated propagation times as the lead time. The current EAP’s trigger model lacks meteorological inputs, limiting skilful lead times to less than four days. Recent advancements in global operational flood forecasting systems present opportunities to enhance Mali's EAP by leveraging increasingly skilful medium-range weather forecasts as inputs of both physically-based models, as in the Copernicus Emergency Management Service's Global Flood Awareness System (GloFAS), and artificial intelligence-based models, like in Google Flood Hub. Incorporating forecasts from these models in Mali’s EAP could improve flood anticipation. This study evaluates the performance of the latest version of GloFAS (version 4) and Google Flood Hub alongside Mali’s current trigger model for the Niger and Senegal river basins in Mali. We evaluated hindcasted triggers aggregated to administrative units, using river flow observations and flood impact data, sourced from OCHA, EMDAT, DesInventar, DRPC Mali, DGPC Mali, CatNat, Relief, and a text-mining algorithm applied to newspaper articles. Model performance was assessed using Probability of Detection (POD) and False Alarm Ratio (FAR) for different lead times and discharge return period thresholds. GloFAS and Google Flood Hub demonstrated similar skill in frequently flooded regions, suggesting that lead times can be extended beyond the four-day window. However, performance assessments are limited by the quality of impact data. This study highlights the potential and challenges of enhancing flood forecasting and anticipatory action in Mali. In the future, incorporating flood extent mapping may improve forecast value by pinpointing affected communities, and impact databases can be improved using satellite imagery, enhancing forecast assessments for early actions.
How to cite: Kuipers, E., Oldenburg, V., Sutanudjaja, E., Phung, P., Ficchì, A., and van den Homberg, M.: Assessing the riverine flood forecast skill of GloFAS and Google Flood Hub with impact data and river flow observations to support early actions in Mali, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-302, https://doi.org/10.5194/egusphere-egu25-302, 2025.
The 2021 Ahr Valley flood during storm Bernd exemplifies the severity of flash floods and the challenges in flood risk and emergency management. This event underscores the growing threat of flash floods, even in regions where they are not traditionally considered common. The sudden, localized nature of flash floods makes early warning systems (EWS) critical. Failures in EWS played a relevant role during the Ahr floods and have likely played roles in other catastrophic events, such as the 2023 Libya floods under storm Daniel and the October 2024 floods in Valencia. Effective warnings require better, actionable information from flash flood models, a challenge due to the rapid onset of such events, the high resolution needed, and significant computational demands.
This study examines the application of the fully dynamic 2D shallow water solver SERGHEI, specifically designed for multi-GPU systems in large-scale High-Performance Computing environments. The focus is on simulating the rainfall-runoff process and subsequent flooding during the 2021 Ahr floods.
In earlier work, we explored flood propagation dynamics in the lower Ahr valley using SERGHEI at a very high resolution of 1m. We showed that simulations could be performed quickly enough for early warning, but with two key limitations. Firstly, since the domain only includes the lower valley, an inflow hydrograph is required at the upstream boundary to force the model. When performing forecasts for early warning, such inflow hydrograph would need to be generated by some hydrological model for the catchment down to the point of inflow, thus requiring a modelling chain. Second, the domain of interest for flood impact modelling, and thus the location for the hydrograph generation, is a priori unknown.
To address these limitations, we scale up the simulation by simultaneously modelling runoff generation and flood propagation over the entire catchment (900 km2). We perform SERGHEI simulations informing the model with a 1m resolution DTM, and openly available land cover, land use and soil data to parametrise hydraulic roughness and infiltration processes. The model is forced using radar precipitation measurements (1km spatial resolution 5 minutes temporal resolution). The target simulation resolution is 1m, leading to a computational grid of ~900 million cells, requiring 128 A100 GPUs in the JUWELS supercomputer, running roughly 5x faster-than-real-time. To perform sensitivity analysis to the infiltration and roughness parameters, we perform simulations at 5m resolution, for which the 36 million cell domain only required 16 GPUs to perform computations ~45x faster than real time. We also explore other resolutions to understand the effects of resolution on the quality of the forecast, computational resources and attainable lead time.
The results show the tradeoffs among modelling approaches for this event and demonstrate the feasibility and advantages of this approach for early warning in flash flood events. They underscore the maturity of the technology and provide strong arguments for using it to augment existing operational flood forecasts, while still achieving excellent lead times and far better detailed flood impact forecasting.
How to cite: Caviedes-Voullième, D., Khosh Bin Ghomash, S., and Morales-Hernández, M.: Catchment scale hydrodynamic flash flood simulation for early warning: insights from the 2021 Ahr flood event. , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5862, https://doi.org/10.5194/egusphere-egu25-5862, 2025.
Short-term flood and inundation forecasts are challenging due to the short lead time of convective heavy rainfall events and the associated uncertainties of input data or model initial conditions. These uncertainties propagate along the forecast chain to uncertainties of the prediction of flooding extents, flow regimes, and, eventually, potential damages. Within the research project AVOSS (funded by the Federal Ministry of Education and Research) the aim is to quantify the contribution of the accompanying uncertainties of the individual forecast components. Hence, we focus on the uncertainties of the input variables particularly precipitation variability, soil moisture and soil properties, and urban drainage system effects as well as associated model and parameter uncertainties.
The applied forecast model chain consists of three parts. The first part is an ensemble based radar forecast of the temporal and spatial distribution of rainfall intensity. In a second part hydrological models are used to predict surface runoff formation based on the rainfall forecasts and pre-event soil moisture estimates. To capture the variety of different model approaches, two different hydrological models (RoGeR [1] and LARSIM [2]) were used. In a third step, the ensemble of surface runoff estimates from the hydrological models were then used to calculate inundation depths, flow velocities and local discharge applying a hydraulic surrogate model based on neural networks. The surrogate model was trained using a large ensemble of hydrodynamically simulated runoff scenarios generated by the 2D-hydraulic model HydroAS [3]. Uncertainties underlying the 2D-hydraulic model were considered by repeating a subset of hydraulic simulations with two additional hydraulic models.
We applied the forecast approach to an urbanized catchment at the foothills of the Black Forest, Germany, with a catchment extend of about 20 km². Based on the short computation time of the neural network model, which has been found to provide good reproductions of maximum water depths, maximum flow velocities, and maximum discharges, the setup enables the production of large forecast ensemble, suitable for a profound uncertainty estimate. In order to systematically evaluate and rank the influence of the input, parameter and model uncertainties along the forecast chain, a sensitivity analysis using Sobol Indices was carried out with the SAFE toolbox [4].
The results demonstrate which uncertainties plays the dominant role in short-term flash flood forecasting. Our study also enhances knowledge about the overall uncertainties for real events and their specific quantitative effects in pluvial flash floods. Furthermore, we identified the most relevant factors to be considered for the design of real-time flood hazard maps and subsequent damage forecasts. This ultimately has the potential to create more reliable predictions for pluvial flash floods and provide insights for decision-making under uncertainty.
[1] Steinbrich et al. (2016): Model-based quantification of runoff generation processes at high spatial and temporal resolution. Environmental Earth Sciences (2016) 75:1423.
[2] Bremicker (2000). Das Wasserhaushaltsmodell LARSIM: Modellgrundlagen und Anwendungsbeispiele. Institut für Hydrologie der Universität Freiburg.
[3] Hydrotec mbH (2021): 2D-Strömungsmodell für die wasserwirtschaftliche Praxis.
[4] Pianosi et al. (2015), A Matlab toolbox for Global Sensitivity Analysis, Environmental Modelling & Software, 70, 80-85.
How to cite: Reinecke, A., Hänsler, A., Weiler, M., Leistert, H., Schmit, M., Steinbrich, A., Haag, I., Krumm, J., Zimmer, J., Grießinger, N., Huth, B., Brendt, T., Liu, Y., Hendricks-Franssen, H.-J., and Neuweiler, I.: Quantifying Uncertainty in Flash Flood Forecasting using Ensemble Methods and Sensitivity Analysis, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11063, https://doi.org/10.5194/egusphere-egu25-11063, 2025.
Reliable warnings and forecasts of extreme precipitation and resulting floods are an important prerequisite for disaster managers to initiate flood defence measures. Thus, disaster managers are interested in extended forecast lead times, which can be obtained by employing forecasts of numerical weather models as driving data for hydrological models. Especially in arid environments, warning and forecasting systems are often missing. Challenges arise due to the short response time of watersheds and the uncertainties of the meteorological forecasts. Thus, ensemble forecasts of precipitation are an option to portray these inherent uncertainties.
This study aims to explore the usability of a global numeric weather forecast model for flash flood early warning and present our operational web-based demonstration platform for hydro-meteorological ensemble flash flood forecasting for the Wadi Al-Hawasinah in North Al-Batina region in Oman. We use the ICON-EPS product of the German Weather Service, a global weather forecast model, which provides an ensemble of 40 members each six hours. If predefined extreme precipitation thresholds are exceeded in the region, a rainfall-runoff model tailored on arid hydrology conditions is started to propagate the meteorological uncertainty into the resulting runoff, followed by statistical post processing and visualization for flash flood early warning. Different options for the visualization of the uncertainty information are presented like rainfall quantile maps, exceedance probabilities and traffic light cards. However, the current design of the web-based demonstration platform is based on an iterative stakeholder process, which is still ongoing.
Based on the current setup of the forecasting system, forecast lead times of up to 48 hours are achieved. Furthermore, due to its flexible structure the hydrologic model can be easily exchanged to more advanced 2D-surface routing and inundation modelling approaches.
Besides layout and technical issues, first experiences with the demonstration platform are presented as well as first results regarding forecast performance in this study area as a pilot study. Finally, we discuss the system’s limitations, particularly the absence of real-time observations, and propose potential solutions to address these gaps.
How to cite: Grundmann, J., Wagner, M., Wischnewski, J., Al-Jahwari, B., and Al-Rawas, G.: Towards operational flash flood early warning for an arid watershed in Oman based on hydro-meteorological ensemble forecasting, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13608, https://doi.org/10.5194/egusphere-egu25-13608, 2025.
In general, quantitative precipitation estimates from weather radars are used as input into nowcasting models to produce high-resolution accurate and timely precipitation forecasts, up to several hours ahead. However, the global distribution of high-resolution (gauge-adjusted, ground- based) weather radar products is heavily skewed, largely favoring Europe, Northern America, and parts of East Asia. In many low- and middle-income countries, predominantly located in the tropics, weather radars are largely unavailable due to high installation and maintenance costs, and rain gauges are often scarce, poorly maintained, or not available in (near) real-time. A viable and ‘opportunistic’ source of high-resolution space-time rainfall estimates is based on the rain-induced signal attenuation experienced by commercial microwave links (CMLs) in cellular communication networks. In this study we investigate whether 2D rainfall fields created by interpolating path- averaged rainfall intensities from CMLs can be used as a standalone input into a conventional nowcasting algorithm, pySTEPS.
This work is based on a CML network from Sri Lanka. The data set spans 15 months across 2019 and 2020. For each of the four monsoon seasons represented in the data set we define extreme events of different duration, ranging from 1 to 24 hours. These events are used as input to create probabilistic nowcasts in pySTEPS for lead times up to three hours. The nowcasts are evaluated spatially against the QPE at multiple catchments, and using 21 hourly rain gauges as an independent point reference source. We address challenges surrounding the nature of the input data, dealing with sparse or unequal CML coverage, and how to handle this in pySTEPS. Based on our findings we also highlight where other remotely sensed rainfall estimates, for example from geostationary satellites, can be used to complement CML based rainfall estimates to provide more accurate nowcasts.
In summary, this novel application of CMLs, essentially providing a ‘weather radar’ in the tropics, highlights the potential impact for operational early warning services in regions that lack dedicated rainfall sensors.
How to cite: Walraven, B., Imhoff, R., Overeem, A., Coenders, M., Hut, R., van der Valk, L., and Uijlenhoet, R.: Opportunities and challenges for Rainfall Nowcasting with Commercial Microwave Links in the Tropics, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21609, https://doi.org/10.5194/egusphere-egu25-21609, 2025.
The reliable prediction of drought impacts on crop yield in India poses a significant challenge due to the complex interactions of climatic variables, systems vulnerabilities and impacts propagation. Addressing this challenge requires advanced methods, such as impact-based forecasting, to account for these complexities. In this study, we leveraged remote sensing-based vegetation indicators as proxies for crop yield, along with multiple drought indices across various accumulation periods, to establish a robust indicator-impact relationship. A cluster analysis was performed to group districts, followed by a comparative evaluation of various machine-learning algorithms (Random Forest, XGBoost, Artificial Neural Network) to assess their efficacy in predicting crop yield impacts on a subseasonal-to-seasonal scale. We finally evaluated the accuracy of predicting the crop yield impacts based on drought indices computed from ECMWF’s seasonal forecast system SEAS5.
Our analysis highlights the importance of key predictors, uncovers seasonal trends and spatio-temporal patterns in indicator-impact relationships, and marks a pioneering effort in comparing diverse machine-learning algorithms for establishing an impact-based forecasting model at lead times of 1 to 6 months. As such, these findings offer valuable insights into the dynamics of drought impacts on crop yield, providing a monitoring tool and a foundational basis for implementing targeted drought mitigation actions within the agricultural sector. This research contributes to advancing the understanding of impact-based forecasting models and their practical application in addressing the challenges associated with drought impacts on crop yield in India.
How to cite: Shyrokaya, A., Uttarwar, S., Di Baldassarre, G., Majone, B., Samantaray, A., Stainoh, F., Pappenberger, F., Pechlivanidis, I., and Messori, G.: Drought impact-based forecasting of crop yield in India, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16546, https://doi.org/10.5194/egusphere-egu25-16546, 2025.
Despite the increasing availability of data from various sources, it remains difficult for humanitarians and governments to respond adequately and quickly to unfolding humanitarian crises. One of the problems that causes this, is the challenge that decision-makers face in assessing the impact of a given shock or hazard on the local population. Therefore, a major issue for the development of early warning systems for humanitarian action lies in contextualizing the data to go from a specific hazard or shock event to its impact on the local population. Given these problems and the developments in AI and data-driven modelling in the past decade, there are many hopes that AI can close this information gap.
However, many scholars and practitioners are apprehensive about using (often complex) data-driven models for actual humanitarian decision-making in practice, and rightfully so. Different documented cases from the public sector such as the Dutch child-benefit scandal or the American COMPAS case have shown what harm the irresponsible use of AI-informed decision-making can do to already vulnerable and marginalized populations. Thus, the question remains how to responsibly develop data-driven models that are useful to the humanitarian community.
In our trans-disciplinary research in collaboration with the Integrated Food Security Phase Classification (IPC), we spent a year exploring this question in a case study on how data-driven models impact the IPC's decision-making process on Acute Food Insecurity analysis updates. Using a human-centered design approach, we systematically analysed the current IPC decision-making process and their information needs and evaluated existing food insecurity models with respect to their suitability, while simultaneously conducting literature research on how to develop AI solutions in a value-driven way. These explorations indicated that the common approaches to developing data-driven models, as well as existing theoretical frameworks with regard to responsible AI implementation, have clear mismatches and shortcomings compared to what may be needed in practice. From this, we draw several lessons on how to improve, so that models for humanitarian decision-making can bring actionable insights, are understandable by its end-users, and embody the humanitarian values.
How to cite: Roelvink, M., Liem, C., and Comes, T.: Responsibly developing data-driven models for humanitarian decision-making: our research on AI for Food Security Monitoring and what we can learn from it, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20144, https://doi.org/10.5194/egusphere-egu25-20144, 2025.
Posters on site: Tue, 29 Apr, 16:15–18:00 | Hall A
In 2022, the global population reached 8 billion, with 55% residing in urban areas. Projections for 2050 anticipate a growth to 9.77 billion, with approximately 6.6 billion people (nearly 68% of the world’s population) living in cities. Urban flooding emerges as a hazardous phenomenon affecting both developed and developing nations, endangering human lives and causing damage to properties, environmental degradation, and disruptions in economic and social activities, such as transportation systems and urban mobility. Addressing this challenge, Flood Early Warning Systems (FEWSs) can play a vital role in mitigating flood risks, enhancing absorptive capacity, and minimizing the impact of hazards, ultimately reducing the loss of life.
In this context, this project aims to create a prototype for a high spatial resolution flood early warning system that will identify flooding hotspots or zones in a pilot area (São Paulo City) and provide flood lead time at the urban micro-basins scale. The project will verify flood alerts by employing artificial intelligence (AI) methods. Furthermore, innovatively, the state of the art in this context will be explored for the national scenario. The anticipated outcomes are real-time geo-information of areas with higher flood risk, offering critical insights for effective response during such events. The project will advance scientific knowledge in this domain and provide a practical support tool for Civil Defense agents, decision-makers, and policymakers. The conceptual prototype developed in this initiative is envisaged to serve as a valuable resource for São Paulo City. Providing timely information empowers authorities to make informed decisions to contain and mitigate the impact of floods, fostering resilience in the face of this pressing environmental challenge.
How to cite: Escobar-Silva, E. V. and Santos, L. B. L.: A Conceptual Prototype of an Urban Flood Early Warning System with High Spatial Resolution: A Study Case in São Paulo City, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-149, https://doi.org/10.5194/egusphere-egu25-149, 2025.
Global hydrological forecasting systems, such as the Global Flood Awareness System (GloFAS), part of the Copernicus Emergency Management Service, are operationally used to inform early warning and early action, particularly in large transboundary river basins and data scarce regions. Humanitarian organizations often integrate these global forecasting systems with local information to assist national mandated agencies in the disaster risk management chain. However, limitations in the skill of global systems restrict their operational adoption and constrain the lead times available for implementing early actions. Recent advances in AI models offer promising solutions to overcome these limitations, by complementing operational physics-based models like GloFAS with hybrid or fully data-driven systems. Despite an increasing number of studies showing the potential of such AI models, there is an urgent need of providing user-oriented evidence of the added value of these solutions in order to increase their operational uptake. Here we explore the application of a deep learning model, based on a Long Short-Term Memory (LSTM) network, to improve the forecasts of GloFAS to support humanitarian anticipatory actions. Different LSTM architectures and loss functions are tested to develop alternative post-processing models of GloFAS, using historical forecasts of river flows, past errors and catchment characteristics as inputs, to improve the prediction of daily streamflows up to a 7-day lead time. The post-processing model is developed with both a single-site and multi-site approach, showing a comparable performance in cross-validation, using streamflow observations as reference. The improvements in skill and value of the flood forecasts of GloFAS are demonstrated for anticipatory actions in Southern Africa (Zambezi River Basin and coastal areas of Mozambique), a region that is highly exposed to frequent tropical cyclones and consequent floods. Using the LSTM-based post-processing, the large biases of GloFAS in this region are substantially reduced and the skillful lead times are extended significantly. We assess the added value of the hybrid forecasts within the framework of the current Red Cross Early Action Protocol for floods in Mozambique, considering user-oriented metrics, including False Alarm Ratios and Hit Rates, and a valuation framework of early actions. Our findings highlight the critical importance of evaluating hybrid forecasting models based on user-oriented criteria and assessing their value to select the most cost-effective solution to support anticipatory actions. Finally, we discuss the potential of our hybrid approach to scale up anticipatory actions in data scarce regions and how ongoing work focusing on post-processing flood hazard maps may further improve forecast value for early actions.
How to cite: Ficchì, A., Fayaz Mir, M., and Castelletti, A.: Enhancing global flood forecasts in Southern Africa using Deep Learning: A user-oriented evaluation for anticipatory actions, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17698, https://doi.org/10.5194/egusphere-egu25-17698, 2025.
Heavy rainfall events have become increasingly frequent in recent decades, often triggering severe floodings that pose significant challenges to urban and rural communities. Consequently, robust early warning systems have emerged as a key strategy for adaptation and mitigation measures. The risks and impacts of flooding can vary significantly even within small geographic areas due to factors such as terrain, urban infrastructure, and drainage systems, as well as the sporadic nature of rainfall. Site-specific flood warning tools address local variations by providing warnings tailored to each area's unique conditions. These tools can help decision-makers and emergency responders navigate multiple challenges, improve preparedness for extreme events, and promote public awareness of flood risks.
Catalonia, located in northeast Spain and characterized by a Mediterranean climate, is occasionally affected by intense rainfall episodes. Severe flash floods have caused significant damage in recent years, leaving communities grappling with the aftermath, such as the case of Terrassa municipality in 2023 and 2024. A real-time, site-specific, flood early warning tool has already been applied in pilot locations in Catalonia within the Horizon Europe RESIST project (2023-2027). The tool integrates real-time and forecasted meteorological data to issue flood hazard warnings for vulnerable locations. In this study, we focus on a methodology for evaluating and optimizing the warning tool to minimize false alarms and missed events. Evaluation is essential to ensure the reliability and usability of the tool and build the trust of end-users, particularly emergency responders and affected communities. We present an evaluation of the tool’s different components, including the warning level thresholds, the integration of different data sources, and lead time analysis. Optimization, on the other hand, involves refining algorithms, integrating additional local data sources tailoring the tool to specific local characteristics, and incorporating feedback from end-users.
How to cite: Elhabashy, A., Park, S., and Sempere-Torres, D.: Evaluation and optimization of a site-specific early warning tool for flood hazards in Catalonia, Spain, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18842, https://doi.org/10.5194/egusphere-egu25-18842, 2025.
The EDERA project, funded by the EU Civil Protection Mechanism, has focused on the use of real-time products for forecasting and monitoring the impacts of storms, heavy rainfall and flash floods to support emergency management. The project ran a 15 months demonstration in real time with European coverage and involved the participation of several end-users (with responsibilities at regional or national level). The study presents the main results obtained during the demonstration period from the point of view of the skill of the products to identify/anticipate the occurrence of the most significant events, and the magnitude of the resulting impacts.
How to cite: Berenguer, M. and Park, S.: Evaluation of 15-months of flash flood impact forecasts over Europe, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19188, https://doi.org/10.5194/egusphere-egu25-19188, 2025.
The development of efficient urban surface water flood forecasting systems is particularly challenging with respect to accurately anticipating the space-time location of heavy precipitations, and rapidly evaluating flood probabilities to issue timely alerts. This study presents an approach based on a pre-computed catalogue of flood inundation scenarios. This approach can serve as an intermediate alternative between basic rainfall threshold-based approaches, and computationally intensive real-time hydraulic simulations. The construction of the catalogue of flood scenarios is illustrated using the Jarret River basin in Marseille, France as a case study. The methodology uses a nine-year (2014-2023) radar rainfall reanalysis with 15-minute temporal and 1-kilometer spatial resolution to define a panel of representative rainfall hyetograph shapes for two-hour convective rainfall events of different return periods. A Telemac 2D hydrodynamic model via the CARTINO 2D approach is then used to obtain the flood scenarios related to each hyetograph. Two approaches are developed to build the hyetographs: (1) a temporal pattern analysis resulting in the distinction of three characteristic hyetograph shapes (short triangle, long triangle, rectangular), and (2) a monofrequency method using triangular hyetographs with consistent return periods across 15min to 2h durations, combined with a spatial attenuation according to the drainage areas impacted by each rainfall duration (cf. concentration times). Both approaches are applied for five return periods (5, 10, 20, 50, and 100 years) under three antecedent moisture conditions, to generate flood catalogues including 45 and 15 scenarios respectively. The resulting catalogues demonstrate the significant influence of temporal rainfall variability on inundation patterns over small catchment areas. As a next step, both approaches will be integrated in an experimental forecasting chain, and be evaluated through the reanalysis of past events. These predefined flood catalogues offer a practical framework for rapid flood response in urban areas exposed to high-intensity, short-duration rainfall events.
How to cite: Kowlesser, A., Payrastre, O., Gaume, E., and Nicolle, P.: Development of surface water flood scenario catalogues for urban flood forecasting: A case study of le Jarret basin, Marseille., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5424, https://doi.org/10.5194/egusphere-egu25-5424, 2025.
Intense rainfall events with high intensities over short durations frequently result in substantial runoff and increased potential for flooding in affected catchments. The accurate assessment of flood hazards remains challenging due to the high variability of rainfall dynamics and their spatial distributions. Rain gauge stations provide precise point measurements; however, they lack information on the spatial distribution of rainfall. Conversely, weather radar offers high-resolution spatial and temporal rainfall data but is subject to biases and uncertainties that require correction.
Previous studies have predominantly focused on pointwise comparisons of rainfall data products. In contrast, this study utilizes data from 109 catchments in Lower Saxony, Germany, to evaluate the ability of station-based and radar-derived rainfall data (using the corrected Radklim product from the German Weather Service) to explain and classify observed runoff events. These events are categorized as Flash Floods, Short-Rain Floods or Long-Rain Floods and the quality of the rainfall data is analyzed in relation to these classifications. Furthermore, the study investigates whether significant runoff events can be exclusively explained by one rainfall data source.
By comparing catchment-averaged rainfall from stations and radar, this research highlights the strengths and limitations of both data types in representing rainfall-runoff relationships. The findings will contribute to improved flood hazard assessment and emphasize the importance of selecting appropriate rainfall datasets for hydrological analyses and early warning systems.
How to cite: Brandt, A. and Haberlandt, U.: Explainability of rainfall-runoff events based on radar and station based rainfall observations, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8134, https://doi.org/10.5194/egusphere-egu25-8134, 2025.
Quantification and prediction of droughts have mainly been focused on the surface and/or meteorological components of the water cycle due to the complex nature of subsurface processes and limited observational data on the hydrogeological component of the water cycle. A web-based Early Warning System (EWS) has been developed to predict seasonal hydrogeological droughts and to assess the resilience of subsurface water resources in the West Bank transboundary karst system, which encompasses the territories of Israel and the Palestinian regions of the West Bank. This innovative tool integrates the monthly-released seasonal weather prediction data from the Copernicus Climate Change Service with a surrogate hydrogeological model to predict the functioning of the karst hydrogeological system and characterize its potential drought conditions. A multi-model ensemble (MME) of daily seasonal predictions has been considered to quantify the spatiotemporal uncertainty of daily climatic variables, which subsequently translates to recharge, storage, and discharge in the subsurface, to be highlighted as the ranges of hydrogeological drought indices. The surrogate deep auto-regressive neural network model (Deep-AR-Net), is utilized to reduce the computational burden of a process-based variably-saturated double-permeability model of the region. The EWS incorporates multiple variables of the MME, including precipitation and temperature, along with flow observations on groundwater levels and spring discharges, to predict hydrogeological conditions during the upcoming six months via Deep-AR-Net. The EWS presents results through an interactive map interface and graphical displays, allowing water resource managers to visualize potential droughts and compare predictions against established drought index thresholds. The development of the EWS is a significant advancement in hydrogeological drought prediction and water resource management for karst systems in arid and semi-arid region. By providing a shared platform for data analysis and visualization, it facilitates collaborative decision-making and helps to prevent potential conflicts related to water use in this sensitive region, which has always been under significant water stress and political tension. More specifically, it will support water managers and policymakers as a powerful instrument to enhance drought preparedness, optimize water allocation, and implement timely mitigation strategies in the face of increasing climate variability and water scarcity.
How to cite: Kavousi, A., Saft, M., Maier, U., Engelhardt, I., Hochman, A., Gebel, M., Dietrich, P., and Sauter, M.: Development of a Web-Based Early-Warning System for Seasonal Hydrogeological Drought Prediction and Assessment of Water Resource Resilience in a Transboundary Karst System, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11710, https://doi.org/10.5194/egusphere-egu25-11710, 2025.
Impact-based early warning (IbEW) is a novel paradigm that aims at improving the efficacy of early warning systems by informing about potential impacts on people, assets and systems, instead of only focusing only forecasting hazards. While applications are emerging, multiple challenges still remain to develop risk-informed, impact-based warnings that are useful for triggering early actions. Conceptual risk models, such as impact chains or impact webs, are tools increasingly used in risk assessments to inform risk management and adaptation, and can provide useful guidance also for IbEW and early action. By identifying the interconnections between drivers of hazards, exposure and vulnerability, they can improve the understanding of possible impacts and risks, thus allowing for a more targeted inclusion of risk information into IbEW. Moreover, through their focus on vulnerability, they can also be used to link the warnings with early actions, highlighting capacities and barriers. Using case studies in Kenya and Ethiopia and the wider IGAD region of Greater Horn of Africa, we have constructed conceptual risk models for different risks connected to droughts and floods: the models provide detailed representations of the interaction of drivers of risk, conducive to specific potential impacts of interest in the context of impact-based early warning (e.g. risk of crop yield loss due to drought). Moreover, in the models we also introduce examples of risk profiles, i.e. characteristics of vulnerability for specific at-risk groups: these can help identifying capacities and barriers of those who need to act on the early actions that are informed by IbEW. This information is essential in order to design warnings that are understandable and actionable by people on the ground. The models were also used to inform the development of an IbEW methodology, currently being implemented at the regional level to cover multiple risks connected to droughts and floods.
How to cite: Cotti, D., Dewi, M. B. K., Pfeiffer, S., Kiptum, A., Werners, S., and Hagenlocher, M.: Bridging Risk Knowledge and Early Action: using conceptual risk models to advance impact-based early warning for floods and droughts in eastern Africa., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13867, https://doi.org/10.5194/egusphere-egu25-13867, 2025.
Accurate flood hydrograph prediction during typhoon-induced heavy rainfall events is crucial for flood risk management, particularly in critical catchments such as the Shihmen Reservoir watershed in Taiwan. The Shihmen Reservoir plays a pivotal role in flood control, water supply, and hydroelectric power generation, making reliable flow predictions essential for its effective operation during extreme weather events.
This study addresses the challenges of long-duration flood hydrograph prediction by developing a hybrid model that integrates an AI-based Rainfall-Runoff Autoregressive with Exogenous Inputs (RNARX) model and a hydrological storage function model. While the RNARX model effectively estimates flow during active rainfall periods using rainfall as the primary input, its performance diminishes post-rainfall when rainfall values drop to zero, leading to rapid underestimation of flow. In contrast, the storage function model provides reliable flow predictions during the recession phase but tends to overestimate flows during intense rainfall events.
By seamlessly combining these two models and defining conditions for model transitions, the hybrid approach ensures robust performance across the entire flood hydrograph. Applied to the Shihmen Reservoir watershed, the hybrid model demonstrates significant improvements in predicting long-duration flood flows, particularly for high-intensity typhoon rainfall events.
This integrated modeling approach enhances real-time flood forecasting, offering valuable insights for optimizing reservoir operations and mitigating flood risks in the Shihmen Reservoir watershed, a region of critical hydrological and socio-economic importance.
Keywords: Hybrid Modeling, Artificial Intelligence (AI),Storage Function Model, Flood Hydrograph Prediction, Flood Risk Management
How to cite: Hsu, C.-Y., Huang, C.-Y., Chang, F.-J., and Chang, L.-C.: Hybrid AI and Storage Function Model for Accurate Flood Hydrograph Prediction During Typhoon Rainfall Events, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15448, https://doi.org/10.5194/egusphere-egu25-15448, 2025.
Dam failures pose significant risks to life, property, and nature. Overtopping is the most frequent cause of dam failure, typically triggered by extreme rainfall events. The increasing frequency and magnitudes of such events, driven by climate change, further amplify these risks. This study investigates the effect of extreme rainfall patterns on 20 Dam Failure (DF) cases in India. Daily rainfall data are obtained from the India Meteorological Department, Pune, for 120 years and divided into four 30-year periods, i.e., ‘Epoch’ (Epoch-1: 1901–1930, Epoch-2: 1931–1960, Epoch-3: 1961–1990, Epoch-4: 1991–2020). The location-specisfic rainfall data is computed using the Inverse Distance Weighted interpolation method. The dates of DFs are sourced from the Central Water Commission, State Water Resources Departments, and other literature.
First, the 5-day Accumulated Rainfall (ACR5) prior to the date of DF is computed, and compared with the ACR5 of other years prior to DF during the same dates. Interestingly, none of the value exceeds the ACR5 of DF year in most of the locations. It denotes that these dams failed due to the accumulated effect of consecutive heavy rainfall events, which were not anticipated by the respective dam authority to prepare for safeguarding the dam through suitable operations.
Second, the trends in the rainfall distribution over each epoch are analyzed by computing the normal rainfall (30-year averaged annual rainfall). The severity of ACR5 with respect to normal rainfall (respective epoch in which DF occurred) is examined. It is noticed that the proportion of ACR5 with that of normal rainfall varied between 30%-90%. This means a huge magnitude of rainfall occurred in just 5 days. Therefore, it is indicated that the ACR5 played a crucial role in the failure of most of the dams considered in this study.
Third, the study also introduced the Efficiency Factor (EF), defined as the ratio of maximum daily rainfall to Probable Maximum Precipitation (PMP). The value of EF above 0.85 poses a severe threat to dams and could result in DF. The vital conclusion from this study is that the dam owners will be notified at least 5 days prior to the dam failure, which is sufficient to take suitable measures for safe reservoir operations. The major limitation of this study is that the date of DF is not known for existing dam locations. However, the advanced weather forecasting models are providing reliable information for 5 to 7-day rainfall estimates, which will enable us to know the critical ACR5. Moreover, the systematic analysis offers a data-driven approach to improve dam safety protocols and enhance resilience against extreme rainfall events. The findings are particularly relevant for professionals in dam engineering, supporting informed decision-making in dam design, operation, and management.
How to cite: Pichuka, S. and Roulo, D.: Extreme Rainfall vs Dam Safety: A Study on Dam Failures in India, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16556, https://doi.org/10.5194/egusphere-egu25-16556, 2025.
Posters virtual: Tue, 29 Apr, 14:00–15:45 | vPoster spot A
EGU25-3970 | ECS | Posters virtual | VPS9
Estimating critical rainfall for flash flood warning systems using integrated hydrologic-hydrodynamic modellingTue, 29 Apr, 14:00–15:45 (CEST) | vPA.16
Flash floods are one of the most severe natural hazards worldwide; they can occur within a few minutes or hours, and can move at high flow velocities, striking with violence and little warning. Early warning of flash floods is extremely important for vital risk mitigation and requires the knowledge of the critical rainfall producing flooding that is typically considered as “warning index”. The small spatial and temporal scales at which flash floods occur make the prediction of critical rainfall challenging, particularly in data-poor environments, where high-resolution weather models and advanced monitoring networks may not be available.
In this research, we present a methodology to estimate the critical rainfall for flash flooding based on an integrated hydrologic-hydrodynamic model. The model is applied in the Lilantas River catchment in Evia, Greece, considering a relatively large number of rainfall and soil moisture conditions scenario combinations in order to (1) determine inflow hydrographs used as boundary conditions for the hydrodynamic model and (2) calculate the distribution of “critical hazard” across the cells of the two-dimensional (2D) computational domain. In the present work, we define critical hazard combining the main hydrodynamic characteristics that are water depth and flow velocity, and we import all calculated “critical hazard” values into a GIS-based database.
Key findings include maximum peak discharges from all simulated scenarios, allowing a sensitivity analysis of varying Curve Number and soil moisture conditions, as well as the effects of rainfall duration and intensity combinations on flood responses. Furthermore, based on the calculated critical hazard, estimates of critical rainfall values for the selected study area are provided, along with an example of the flood warning system’s operation.
How to cite: Papoulakos, K., Mitsopoulos, G., Baltas, E., and Stamou, A. I.: Estimating critical rainfall for flash flood warning systems using integrated hydrologic-hydrodynamic modelling, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3970, https://doi.org/10.5194/egusphere-egu25-3970, 2025.
