Reservoir regulation reshapes high-flow dynamics, yet the event-scale limits of flood moderation remain poorly quantified at national scales. We develop a framework to compare naturalized (NAT) and regulated (DAM) high-flow responses for 223 Indian reservoirs by coupling Noah-MP runoff with vector-based mizuRoute routing (2000–2022). High-flow events are identified using scenario-specific P99 thresholds and evaluated at the inflow peak (T) and the subsequent recession day (T+1) to quantify peak attenuation and the storage states that determine remaining flood buffer. Reservoirs reduce peak magnitudes relative to NAT conditions across all storage classes, from large (>1,000 MCM) to small (<1,000 MCM) systems. However, T+1 diagnostics reveal a consistent limitation: reservoirs exit most events nearly saturated, with a median post-event storage of 99.8 % of capacity and 74 % of events exceeding 90 % capacity. The T+1 storage fraction-attenuation relationship exhibits a near-zero slope across storage sizes, indicating that larger reservoirs attenuate more but still approach full capacity, while smaller reservoirs saturate even more rapidly. Instances where strong attenuation coincides with appreciable storage headroom are rare (<5 %). These findings highlight a national-scale constraint: Indian reservoirs effectively moderate individual peaks but rapidly expend available storage buffer, increasing vulnerability to multi-peak or persistent inflow sequences.
How to cite: Sharma, P., Saharia, M., and Mizukami, N.: Event-Scale Limits of Reservoir Flood Moderation: Peak Attenuation, Storage Saturation, and Capacity Size-Dependent Constraints Across India, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-502, https://doi.org/10.5194/egusphere-egu26-502, 2026.
Intensifying hydrologic and environmental extremes are making aging dams around the world increasingly vulnerable as many of them have surpassed their design service life and are classified as high-hazard potential dams. Traditionally used deterministic dam breach analysis methods can underestimate flood risk since they do not necessarily capture an entire spectrum of possible outcomes and their associated uncertainties. This study presents an alternative probabilistic approach for dam breach analyses and flood risk assessments accommodating various modeling approaches and uncertainty quantification techniques. We integrate stochastic simulation methods with a computationally efficient flood modeling tool to enable large-ensemble analysis. We explicitly consider uncertainties in three key components: (1) breach parameters - breach geometry and timing derived from empirical prediction methods, (2) forcings - rainfall characteristics, including depth, temporal patterns, and antecedent soil moisture, and (3) downstream hydraulic conditions and hydrodynamic responses. Advanced sampling strategies, such as copula methods, are adopted to propagate uncertainties from different sources through the modeling chain while maintaining computational efficiency.
The framework is demonstrated through the application to a recent Sanford Dam failure case, where we compare different breach prediction methods and evaluate their impacts on downstream flood characteristics. The deterministic breach model is first validated against observed failure characteristics and downstream flood impacts before extending to probabilistic analysis. Multivariate analysis reveals that breach formation timing characteristics exert stronger influence on peak outflow and flood wave arrival time than geometric breach parameters. Generated probabilistic flood inundation maps showing overall probability, flood depth, and timing provide critical information for emergency response and resilience planning. Initial results that probabilistic approach provides refined confidence bounds for flood risk estimates, informing decision-making with quantified uncertainty. Incorporating future projections highlights that extreme rainfall intensification could increase dam overtopping probabilities, with the magnitude depending on projection scenarios. Our study infuses evidence-based comparison of modeling approaches and supports a more realistic dam-break flood risk assessment for aging dams under non-stationary environmental conditions, informing emergency planning, and infrastructure management strategies.
How to cite: Bista, S., Ghimire, G. R., and Talchabhadel, R.: Moving toward probabilistic dam breach modeling and flood inundation mapping, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-812, https://doi.org/10.5194/egusphere-egu26-812, 2026.
Over the past three decades, lake water level fluctuations have intensified due to climate change and increasing water demand, creating an urgent need for accurate and efficient prediction methods. However, existing deep learning-based surrogates often suffer from two major limitations: the lack of physically informed guidance for hyper-parameter selection, which increases computational costs, and the scarcity of extreme water level samples, which leads to imbalanced datasets and reduced accuracy. To address the limitations, this study proposes a novel Physics-Informed Neural Network (PINN) framework that integrates data augmentation with physically guided hyper-parameter selection. The framework employs boundary water level time series as input, incorporates mass-conservation constraints, and applies a clustering-based augmentation method to enrich extreme event samples. Its applicability was validated in the Lower Lake of Nansi Lake in China. Evaluation using Root Mean Squared Error (RMSE) and Nash–Sutcliffe Efficiency (NSE) shows that incorporating physical constraints robustly improves predictive accuracy, with performance even surpassing that of a classical LSTM model. Physically guided hyper-parameter selection further enhances both training efficiency and accuracy, and the proposed augmentation method reduces RMSE by 69.1% under extreme conditions. Compared with an existing augmentation method, the proposed method can shorten training time by 63.35% with better prediction performance. The final surrogate achieves RMSE = 0.021 m and NSE > 0.94 (against observations), requiring only 2.42% of the computational time of a traditional hydrodynamic model. These results highlight the framework’s potential for reliable real-world water level prediction and its transferability to other hydrological systems.
How to cite: lu, L., Yan, T., Chen, Y., Wang, H., Yang, T., and Liu, Z.: A lake water level prediction method based on data augmentation and Physics-Informed Neural Networks with imbalanced data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4139, https://doi.org/10.5194/egusphere-egu26-4139, 2026.
Abstract
The Manimala River Basin (Kerala, India) experienced extensive inundation during the 2018 Kerala flood, underscoring the need for robust, interpretable, and spatially explicit susceptibility assessments to support risk-informed planning. This study develops a flood susceptibility model for the basin using a Random Forest (RF) classifier and explains its behaviour using an explainable AI framework. A set of hydro-morphometric and land-surface conditioning factors was compiled, including rainfall, vertical distance to channel network (VCDN), slope angle, soil texture, distance to streams, land use/land cover (LULC), topographic wetness index (TWI), and terrain curvature metrics (upslope and downslope curvature). The RF model was trained for binary classification of flooded and non-flooded locations, and predictive skill was evaluated using both discrimination and classification metrics. The model achieved strong performance, with an area under the receiver operating characteristic curve (AUROC) of 0.90, overall accuracy of 0.82, sensitivity of 0.83, specificity of 0.81, and an F1-score of 0.83, indicating reliable detection of flood-prone locations while maintaining balanced error rates. The susceptibility map was reclassified into three levels to facilitate interpretation and application. The areal distribution shows that 12.62% of the basin falls within the high-susceptibility class, 15.43% within the moderate class, and 71.95% within the low class, providing a basin-scale overview of priority zones for mitigation and preparedness. Model interpretability was addressed using SHapley Additive exPlanations (SHAP). The SHAP summary and mean absolute contribution rankings indicate that rainfall and VCDN exert the strongest influence on RF outputs, followed by slope angle and soil texture, whereas streams, LULC, TWI, and curvature variables contribute comparatively less. These results emphasize the dominant role of hydro-climatic forcing and drainage-related controls, modulated by terrain and substrate characteristics, in shaping flood susceptibility within the Manimala basin. Overall, the proposed RF–SHAP workflow delivers a high-performing and transparent susceptibility product that can support targeted management actions and communication of drivers underlying predicted flood-prone areas.
How to cite: Rajendran, R. and Rangarajan, S.: Explainable Artificial Intelligence and Machine Learning for Flood Susceptibility Modelling in a Tropical River Basin., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5538, https://doi.org/10.5194/egusphere-egu26-5538, 2026.
Rapid urbanization combined with changing rainfall patterns has significantly increased the frequency and intensity of urban flooding in Indian cities. Pune, located along the Mula–Mutha river system, has experienced recurrent flood events in recent years, highlighting the need for a comprehensive assessment of urban flood resilience. This study aims to develop a spatial Flood Resilience Index (FRI) for Pune City to support informed urban planning and infrastructure decision-making. A GIS-based multi-criteria decision analysis framework is adopted to integrate physical, environmental, and socio-economic indicators influencing flood resilience. Key indicators include drainage density, road network density, land-use pattern, impervious surface extent, green cover, population density, and proximity to river channels. Indicator weights are derived using the Analytical Hierarchy Process (AHP), and all spatial layers are normalized and aggregated at the ward level using QGIS. The resulting Flood Resilience Index enables the classification of city wards into very low, low, moderate, high, and very high resilience categories. Preliminary results indicate significant spatial variability in flood resilience across Pune, with densely built-up and low-lying wards along river corridors exhibiting lower resilience levels. The study demonstrates the effectiveness of open-source GIS tools in developing an urban flood resilience assessment framework that can assist policymakers and city managers in prioritizing flood mitigation and resilience-enhancing interventions.
Keywords: Urban flooding, Flood resilience index, GIS, AHP, Pune City
How to cite: Deore, S. and Vinchurkar, S.: Mapping Urban Flood Resilience using GIS: A Case Study of Pune City, India, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7001, https://doi.org/10.5194/egusphere-egu26-7001, 2026.
The accuracy of urban flood inundation modelling is fundamentally limited by the scarcity of ground truth data, which impedes reliable forecasting for disaster risk reduction. This study proposes a novel framework to validate flood model simulations using insurance claim records. We applied a 2D hydraulic model (FloodMap-HydroInundation2D) to simulate a pluvial flood event in Ningbo, China. The simulated maximum inundation extent was validated against georeferenced insurance claims, revealing strong spatial agreement and demonstrating the practicability of such non-traditional datasets. A comparative assessment with social media data further contextualized the performance of the insurance-based validation. Sensitivity analysis highlighted hydraulic conductivity as a key parameter influencing model accuracy, offering valuable insights for parameter optimization. To address inherent locational uncertainties in claims data, a buffer-based validation method was developed and tested, which enhanced the robustness of the model assessment. This work provides a transferable approach for integrating non-traditional observations to improve flood model reliability, thereby supporting flood risk management and emergency decision-making.
How to cite: Gao, D.: Validation of urban flood inundation modelling with insurance claims, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8473, https://doi.org/10.5194/egusphere-egu26-8473, 2026.
Real-time control (RTC) of drainage pumping stations offers substantial potential for urban flood mitigation, yet the computational burden of physics-based models severely limits their utility in rapidly evaluating alternative control strategies during flood events. Although many artificial intelligence methods have achieved rapid urban flood prediction, they have not yet been able to model the flood response to infrastructure control. We develop a data-driven surrogate modeling framework that rapidly predicts flood dynamics (maximum flood extent and peak water depth) under varying pumping operation scenarios. We employ high-resolution finite volume hydrodynamic simulations integrated with pumping control modules to generate an extensive training dataset spanning diverse rainfall events and control configurations across a highly urbanized catchment (approximately 100 km²). Specifically, in the test set, the spatial Root Mean Squared Error (RMSE) is less than 0.05 m, and the structural similarity index (SSIM) exceeds 0.95. The prediction is completed in under one second, representing a three-orders-of-magnitude speed-up compared to the numerical model. This method provides an effective tool for the emergency management of urban flooding
How to cite: Wang, B., Cao, X., and Qin, H.: A Data-Driven Surrogate Approach for Real-Time Evaluation of Pumping Control Strategies in Urban Flood Management, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21367, https://doi.org/10.5194/egusphere-egu26-21367, 2026.
Operational flood forecasting and risk management requires high-resolution, spatially and temporally explicit predictions of inundation dynamics alongside robust uncertainty quantification. In practice, forecast skill is strongly constrained by uncertainties in meteorological forcing, boundary conditions, and human controls (e.g., reservoir releases) as well as inherent model uncertainties. While physics-based 2D hydrodynamic models provide physically consistent inundation dynamics, their computational cost make it impractical to generate several of ensemble members needed for uncertainty quantification and the real-time exploration of “what-if” intervention scenarios.
We present DeepWaive, a physics-informed Foundation Model, that translates precipitation- or discharge-driven boundary conditions and static geospatial input into transient, spatially explicit 2D inundation dynamics within seconds. By leveraging deep-learning architectures trained on synthetic 2D hydrodynamic simulations, DeepWaive achieves zero-shot transferability to previously unseen basins without the need for domain-specific re-training. Crucially, the model architecture maintains the flexibility for optional site-specific fine-tuning, allowing for further optimization using either regional hydrodynamic models or in-situ sensor data to meet localized precision requirements. Benchmarking against classical numerical solvers demonstrates high predictive fidelity, with R² values ranging from 0.85 to 0.97, achieved alongside acceleration factors of 105–106. The model maintains scalability for domains up to 40,000 km2 and event durations exceeding 24 hours.
Building on this capability, we develop an ensemble-to-probability workflow that propagates meteorological and hydrological forecast ensembles, and alternative reservoir release scenarios, through DeepWaive to generate probabilistic inundation products (e.g., spatial exceedance probabilities for depth and velocity thresholds) and impact-relevant summary metrics.
Within the Indo-German FLAIR project (Flood Forecasting using AI for Regional Sustainability, funded by BMBF), DeepWaive provides the fast dynamic core required to (i) quantify and communicate forecast uncertainty, (ii) support rapid sensitivity analyses of key uncertainty sources, and (iii) enable tight coupling to consortium modules on EO-derived flood variables, data assimilation, and reservoir operation optimization.
How to cite: Hofmann, J., Holt, A., Johnen, G., and Welten, S.: DeepWaive: A Scalable and Fine-Tunable AI Foundation Model for Probabilistic 2D Inundation Forecasting , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18141, https://doi.org/10.5194/egusphere-egu26-18141, 2026.
Urban flooding has intensified under rapid urbanization and climate change, posing a growing threat to infrastructure and public safety. Conventional flood risk assessment approaches that are based on fluid dynamics simulations and physics-driven models are computationally demanding and not well-suited real-time applications, particularly in highly dynamic urban settings. This study develops a convolutional neural network (CNN) framework that integrates historical flood inventories, hydrometeorological data, topographic information, and urban morphological characteristics for faster and more accurate prediction of flood extent and depth. The CNN model is trained on a dataset from Shenzhen, China and then applied to Hong Kong, China, demonstrating robust spatial transferability for cross-city flood risk assessments. The framework is used to simulate flood inundation for four design rainfall scenarios . The results indicate that short-duration, high-intensity rainfall events significantly increase the extent of flooding and the depth of inundation. Flood-prone areas in Hong Kong expand to 64.1 km² during a 100-year rainfall event with a 60-minute duration, accounting for 5.79% of the urban area, and the mean inundation depth reaches 0.15 m. In addition, a complementary road-level flood vulnerability analysis identifies 501 flood-prone roads, primarily located in districts with aging infrastructure and high population density. This study highlights the potential of CNNs for rapid flood prediction with strong cross-city transferability, and provide decision-makers with timely insights for targeted flood prevention and disaster mitigation strategies.
How to cite: Zhuang, J., Liu, K., and Wang, M.: A Cross-City Transferable Convolutional Neural Network Framework for Street-Scale Urban Flood Risk Assessment, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9002, https://doi.org/10.5194/egusphere-egu26-9002, 2026.
Extreme rainfall and cloudburst events are becoming increasingly frequent in Europe, placing growing pressure on urban drainage systems and local water utilities. In Denmark, current heavy-rainfall warnings are largely municipality-based binary alerts, which utilities often experience as too frequent, insufficiently localised, and difficult to translate into early operational action. Limited communication of forecast uncertainty and warning thresholds that are not aligned with drainage system design standards further reduce their practical value. As a result, many utilities primarily respond to cloudburst impacts after events occur rather than acting proactively.
This presentation presents the concepts and overall workflow of a co-created urban flood warning framework developed within the Horizon Europe CLEAR-EO project and the Danish funded initiative (VUDP) on future precipitation and flood warnings. The framework is designed to translate probabilistic rainfall forecasts into surface flood impact information that supports earlier and more confident decision-making by water utilities.
Within CLEAR-EO, we develop a modular, end-to-end workflow that links ensemble-based precipitation forecasts with near-real-time EO satellite data and high-resolution surface data. When rainfall thresholds are exceeded, the workflow activates an urban surface flood model that routes water across the urban terrain while accounting for drainage capacity and infiltration, enabling on-demand simulation of pluvial flood impacts. The modelling chain produces spatially explicit, probabilistic flood indicators, including flood depth, spatial extent, and warning levels at sub-metre resolution.
This presentation introduces the overall warning workflow, data integration strategy, and key design choices emerging from the combination of ensemble forecasting, EO-based datasets, surface flood models, and close collaboration and co-creation with end users. A key component of the framework is the generation of hydrologically conditioned, high-resolution DSM, which provides the topographic basis for urban drainage modelling and flood simulations. The workflow integrates classified airborne LiDAR point clouds with semantic infrastructure information from OpenStreetMap, improving drainage connectivity while preserving geometric fidelity. Early experiences indicate that co-developing probabilistic, impact-based warning products that explicitly communicate forecast uncertainty can strengthen utilities’ ability to act earlier and more precisely under uncertain cloudburst conditions.
Ongoing work will further refine the modelling chain, strengthen validation, and extend the approach to additional European case studies, contributing to the development of future national and local heavy-rainfall warning services.
How to cite: Thrysøe, C., Pedersen, J. W., Livia Kruse, I., Durando, E., Oliveti, M., Artù Cassin, E., Destefanis, T., Di Rita, M., and Thomassen, E. D.: Co-creating next-generation cloudburst warnings using ensemble forecasts and surface flood modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11095, https://doi.org/10.5194/egusphere-egu26-11095, 2026.
Until recently, flood early warning in Denmark was based on weather and marine forecasting for extreme rainfall and sea water levels. In 2022, the Danish Meteorological Institute (DMI) was appointed as the national authority for flood forecasting for Denmark, and was tasked with developing and implementing a national flood forecasting and early warning system. The initial goal for these developments was to provide timely, reliable and relevant data to support decision-making by local and national emergency services prior to major flood events.
Initial efforts were focused on the development of a rapid, nationwide flood extent mapping for addressing both pluvial flooding from intense summer cloudbursts and coastal flooding associated with elevated sea water levels and storm surges. In parallel, a national hydrological forecasting system has been successfully developed and the forecasted discharges now form the basis for fluvial flood early warning for the Danish emergency services (2024) and for the general public (2025). However, there remains a critical need for forecasts of flood extent, ahead of extreme events, along the river systems. The key challenge is to deliver flood extent forecasts with sufficient lead time and reliability to support decision-making by authorities during time-critical emergency situations.
In this study, we explore the trade-off between computationally efficient but approximate methods against more accurate but computationally more demanding hydrodynamic simulations for the case study area Ry River (Ryå). The Ryå catchment drains a relatively large – and low-gradient– area of 590 km², but drainage capacity has become increasingly constrained. This has become evident from prolonged and more frequent flooding of intensively cultivated agricultural areas. Ryå represents many of the river modelling and forecasting cases in Denmark but is particularly challenging because of the low slopes.
In particular, we compare and evaluate three methods of varying complexity for estimating flood extent. These are, in order of increasing complexity: 1. GIS-based, static mapping from measured or simulated water levels, 2) an approximate 1D-2D hydrodynamic coupled model obtained by simplifying the governing (St. Venant) equations (LISFLOOD-FP) and 3) a full 2D hydrodynamic modelling of the river and floodplains (HEC-RAS). Performance comparisons, for selected flood events during 2019-2024, are carried out using optical drone imagery together with both radar-based and optical satellite data. We evaluate model run time and accuracy and in the end also the usefulness of the tools to build models covering larger areas of Denmark and being run on-demand. The goal is to determine which methods can, with sufficient accuracy, be used in a semi-operational manner in DMI’s warning setup.
How to cite: Plum, C. A., Henry Madsen, M., Nielsen, A., Trolle, D., and Butts, M.: Evaluating fluvial flood mapping for flood early warning, Ry River Denmark , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10527, https://doi.org/10.5194/egusphere-egu26-10527, 2026.
Floods are the most frequent natural hazard worldwide, causing severe impacts on human populations and leading to substantial economic and environmental losses. In recent decades, the frequency and intensity of flood events have increased significantly, largely due to the growing occurrence of extreme climatic phenomena. Consequently, flood susceptibility mapping has become a crucial tool for flood hazard assessment, management, and mitigation. This study aims to predict and map flood-prone areas in Calabria region (southern Italy) by integrating historical flood data, Geographic Information Systems (GIS), and the Maximum Entropy (ME) modeling approach. A catalogue of flood events impact recorded in the study region between 2000 and 2025 by documentary sources was systematically analyzed within a GIS environment. A total of 270 flood occurrence points affected by flood damage were identified and mapped; 70% of these were randomly selected to construct a balanced training dataset and calibrate the prediction model, while the remaining 30% were used for validation. For the application of the ME method, the flood inventory was combined with fifteen flood-predisposing factors, including lithology, soil texture, land use, normalized difference vegetation index (NDVI), precipitation, elevation, local relief (LR), slope, curvature, topographic position index (TPI), sediment transport index (STI), topographic wetness index (TWI), drainage density (DD), distance to streams, and distance to roads. The validation of the flood-prone areas model was performed based on accuracy, kappa coefficient, and receiver operating characteristic curve (ROC) and its associated area under the curve (AUC). The results indicate very good predictive performance of the model, with success and prediction rates of 89.7% and 86.3%, respectively. In addition, the jackknife test highlighted the significant contribution of soil texture, TWI, precipitation, distance to streams, and land use to the spatial prediction of flood occurrence. The produced flood-prone areas map provides a valuable tool for disaster risk management and mitigation planning, offering significant support to decision-makers in reducing both economic losses and flood-related risks to human life.
How to cite: Conforti, M. and Petrucci, O.: Spatial prediction of flood-prone areas in the Calabria Region (Southern Italy) using historical flood inventories and Maximum Entropy approach, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12503, https://doi.org/10.5194/egusphere-egu26-12503, 2026.
Currently urban flooding poses a growing threat to urban and semi-urban cities across India, mainly due to climate change and rapid unplanned urbanization. Many existing urban flood models require extensive hydrological, hydraulic, and infrastructure data, which are often unavailable in developing or data scarce regions. Our work addresses this challenge by proposing a minimal input modelling approach.
In this study, we develop a high-resolution urban flood simulation model using only rainfall and Digital Elevation data, significantly reducing the dependency on detailed datasets such as land use/land cover and drainage network information. The model framework is implemented in MATLAB using a matrix-optimized water redistribution algorithm combined with flood water routing. This enables efficient simulation of water accumulation and propagation across both flat and complex urban terrains, while allowing simulations over large urban areas.
The model is currently being deployed for Bengaluru urban region in India, which is often prone to flooding during monsoon season, flood inundation maps for multiple time durations were generated using artificial rainfall inputs and a 1-m high resolution DEM. By requiring minimal input data while maintaining high spatial detail, the proposed framework provides a scalable solution for urban flood hazard assessment in data limited regions, supporting early-stage risk mapping and climate resilient urban planning.
How to cite: Singh, R. and Das Bhowmik, R.: A Minimal High-Resolution Urban Flood Model for Data-Scarce Regions, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14562, https://doi.org/10.5194/egusphere-egu26-14562, 2026.
Streamflow forecasting in anthropogenically altered river basins presents substantial challenges, particularly where dam and weir operations strongly modify natural flow regimes. In such systems, conventional hydrologic models often have limited capability to represent the non-linear effects of reservoir regulation, resulting in rapid degradation of forecast skill during extreme events. This study evaluates an ensemble-based hydrologic data assimilation (DA) framework applied to the Nakdong River Basin, South Korea, a highly regulated river system characterized by a dense network of dams and multi-functional weirs. We implement a coupled modeling framework integrating the WRF-Hydro system with the Data Assimilation Research Testbed (DART) to investigate the applicability of DA in a managed hydrologic environment. The WRF-Hydro reservoir module is used to explicitly represent storage and release processes, while DART provides ensemble Kalman filter–based assimilation. A central challenge is the mismatch between modeled natural flows and observed regulated discharges. To address this, streamflow and dam storage (or water level) observations are assimilated to update both natural hydrologic states and managed infrastructure states. The framework is evaluated for an extreme rainfall event in August 2022, demonstrating that joint updating of streamflow and reservoir states improves the ensemble representation of human-induced timing and magnitude. Remaining challenges related to error covariance specification in operationally controlled systems are discussed, underscoring the importance of explicitly accounting for anthropogenic intervention in hydrologic DA systems to improve flood forecasting in regulated basins.
How to cite: Lee, Y., Kim, B., Choi, J., and Noh, S. J.: Integrating WRF-Hydro and DART for Ensemble Streamflow Forecasting in a Highly Regulated Basin with Anthropogenic Intervention, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16399, https://doi.org/10.5194/egusphere-egu26-16399, 2026.
Posters virtual: Fri, 8 May, 14:00–18:00 | vPoster spot A
EGU26-3693 | Posters virtual | VPS11
Designing efficient rain-gauge networks for improved flood forecasting in a large river basinFri, 08 May, 14:06–14:09 (CEST) vPoster spot A
Accurate runoff estimation is fundamental to improving streamflow forecasting, particularly in large river basins with sparse or uneven rain-gauge coverage. This study investigates the identification of representative rain gauges from a densely but randomly distributed network to support reliable runoff simulation in data-limited regions. The Middle Tapi Basin (MTB), comprising 26 operational rain gauges and extensive ungauged areas, is used as a case study. Four approaches—Hall’s method, K-means clustering, hierarchical clustering (HC), and self-organizing maps (SOM)—are applied to identify key rain gauges that effectively capture the spatial variability of basin-scale rainfall. Hall’s method selected 15 representative stations, whereas the clustering-based approaches identified nine stations each. The performance of the resulting rain-gauge networks is evaluated by simulating basin runoff using a lumped hydrological model. Results indicate that the rain-gauge network derived from Hall’s method consistently produces superior runoff simulations compared to the clustering-based networks, demonstrating improved representation of rainfall inputs at the basin scale. Based on these findings, the use of 15 key rain gauges identified through Hall’s method is recommended for runoff prediction in the Middle Tapi Basin. The proposed framework is transferable and can be applied to other large basins with heterogeneous rainfall patterns and limited monitoring infrastructure, offering a practical approach for optimizing rain-gauge networks to enhance hydrological modelling and flood forecasting.
How to cite: Yadav, S. and Panchal, A.: Designing efficient rain-gauge networks for improved flood forecasting in a large river basin, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3693, https://doi.org/10.5194/egusphere-egu26-3693, 2026.
EGU26-3205 | ECS | Posters virtual | VPS11
Implementation and Evaluation of the WRF-Hydro Model for Hydrometeorological Forecasting in the Piura River Basin, PeruFri, 08 May, 14:09–14:12 (CEST) vPoster spot A
Extreme hydro-meteorological events associated with the Coastal El Niño phenomenon represent a critical threat to the socioeconomic stability of the northern coast of Peru. In particular, the Piura River basin, characterized by its complex topography and short concentration times, requires precise monitoring and modeling to address these episodes. Currently, the National Meteorology and Hydrology Service of Peru (SENAMHI) employs a semi-distributed system (ARNO/VIC coupled with RAPID) for the operational assessment of flood risk. However, the increasing intensity and frequency of recent events highlights the need for tools that explicitly represent physical processes at higher resolution. This research proposes the implementation of the fully distributed WRF-Hydro model, focusing the methodology on the reconstruction and analysis of the main extreme flood events within the period covered by the PISCOp_h product, a gridded hourly precipitation observational dataset developed by SENAMHI for 2015–2020. The methodological strategy is based on generating a hybrid meteorological forcing to feed the hydrological model. For this purpose, an atmospheric simulation is carried out with WRF, forced by initial and boundary conditions from the GFS, obtaining high-resolution distributed atmospheric fields. Given the uncertainty of the modeled precipitation, the rainfall field generated by WRF is replaced by the hourly gridded observations from PISCOp_h, ensuring controlled and realistic forcing. With this configuration, model calibration and validation are performed. Calibration prioritizes the highest-magnitude events, highlighting the 2017 Coastal El Niño episode for the adjustment of physical parameters, while validation considers a set of floods recorded between 2015 and 2020, evaluating the robustness of the system. It is expected to demonstrate that this combination of atmospheric dynamics and observational accuracy constitutes a physically consistent and operationally viable tool for predicting intense floods, strengthening flood risk management in Peru.
How to cite: Tufino, J. C., Huerta, A., Lavado-Casimiro, W., De la Cruz, G., Saavedra, D., and Ibañez, A.: Implementation and Evaluation of the WRF-Hydro Model for Hydrometeorological Forecasting in the Piura River Basin, Peru, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3205, https://doi.org/10.5194/egusphere-egu26-3205, 2026.
