Despite advances in hydrological science and technology, our understanding of human–water interactions across scales remains limited. Challenges stem from the complexity and uncertainty in quantifying human influences, the scarcity of long-term records, and the limitations of conventional models often designed for natural catchments under assumption of stationarity. Thus, the reliability of hydrological forecasting in human-influenced systems is compromised. Given the large populations exposed to water-driven hazards, there is an urgent need for intensified research and innovation.
This session will highlight recent advances in understanding and forecasting hydroclimatic extremes in human-influenced catchments. We invite abstracts on (but not limited to):
• Development and application of statistical, process-based, machine learning, or hybrid models to forecast hydrological variables (e.g., meteorological forcings, catchment states, and responses) at multiple scales
• Advances in data acquisition capturing human activities (or proxies), including in-situ monitoring, remote sensing, and unconventional sources such as social media, with innovations in data integration and analytics
• Novel quantitative methods to assess diverse human impacts on hydrological processes and water cycle
• Coupled human-natural system modelling and scenario analysis to capture feedbacks between socio-economic drivers and hydrological processes
• Impact-based risk assessments of water-related hazards, spanning economic, health, social, and environmental dimensions
• Uncertainty quantification and risk analysis of singular and compound hydro-hazards under non-stationarity
• Enhanced visualization and communication for early warnings and short- to long-range predictions, including projections of unprecedented extremes
• Integration of nature-based solutions and adaptive management strategies into forecasting and risk reduction frameworks
Orals: Wed, 6 May, 10:45–12:30 | Room 2.31
Drought risk emerges from multi-directional feedbacks between water availability, risk perception, and adaptation decisions. Human activities both aggravate and alleviate hydrological drought. Adaptation measures to cope with drought can generate negative unintended consequences for other water users, but also unintended benefits.
A global synthesis of 28 cases reveals that water abstraction consistently intensifies drought severity, while reservoir releases can reduce deficits during dry periods but often alter seasonality, leading to wet-season shortages. Drought adaptation measures rarely offset the impacts of water abstraction, but instead shift drought effects in space or time. For example, water transfers reduce deficits in the receiving basin, but increase them in the providing basin, and groundwater-derived streamflow augmentation alleviates extreme low flows, but at the expense of reduced flows during other flow periods.
To further explore the effects of drought adaptation, we developed integrated socio-hydrological tools using system dynamics (SD) and agent-based models (ABM). SD analyses revealed that societies with homogeneous risk-attitudes implement fewer collective measures, opting for more individual measures to address drought risk. However, individual measures by specific social groups may lead to unsustainable water use and therefore larger drought damages. SD analysis in Crete shows that single-sector climate service prioritization amplifies sectoral benefits but increases systemic vulnerabilities, whereas equitable cross-sectoral prioritization fosters balanced, sustainable outcomes.
ABM applications highlight spatial trade-offs, with individual measures also affecting downstream or neighbouring water users. For example, in the Netherlands, irrigation pumping reduces neighbours’ water availability, while measures to raise groundwater levels benefit surrounding farms. In Kenya, upstream commercial abstractions amplify downstream drought risk, and water harvesting improves short-term access but reduces discharge downstream.
These findings underscore the complexity of drought management and the need to represent social diversity, institutions, and cross-scale hydrological feedbacks to design equitable and sustainable adaptation pathways.
How to cite: Van Loon, A., Mazzoleni, M., Wamucii, C., Streefkerk, I., De Graaff, L., Henao Casas, J. D., Rangecroft, S., and Matanó, A.: Human–water interactions and drought adaptation: insights from global cases and socio-hydrological modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18615, https://doi.org/10.5194/egusphere-egu26-18615, 2026.
Anthropogenic climate change is expected to pose significant challenges for European agriculture. Rising temperatures, altered precipitation patterns, and increasing frequency and intensity of extreme weather events, including prolonged and intensified droughts, are projected to substantially increase irrigation water demands. To date, a large fraction of European croplands relies solely on rainfed cultivation. Consequently, the projected increase in drought conditions will likely require an expansion of irrigated areas to mitigate climate-induced yield losses.
This study examines the hydrological impacts of enhanced irrigation across Europe with a particular focus on river discharge, using the Community Water Model (CWatM) at 5′ resolution. CWatM is a large-scale water resources model that simulates precipitation–runoff processes with river routing, capturing both natural hydrological processes and anthropogenic water demands. Its integrated approach provides a unique opportunity to investigate not only impacts of climate change but also impacts arising from the expected increase in irrigation water demand.
The model was calibrated and validated against GRDC discharge data using a regionalized calibration approach. Six irrigation scenarios, representing a stepwise transition from rainfed to irrigated cropland, were implemented in CWatM and simulated for the exceptionally hot and dry summers of 2003 and 2018, which may be considered representative of future summer conditions under climate change.
Simulation results indicate that even a moderate expansion of irrigated areas (converting 10 % of currently rainfed cropland to irrigated cropland) could lead to a substantial increase in unmet water demands and a significant reduction in summer river flows. Rivers in Central and Eastern Europe (e.g., Loire, Rhine, Elbe, Oder, Danube) with agriculturally dominated river basins are particularly affected. For the years 2003 and 2018, summer discharges of these rivers are already below the interquartile range of the 30-year average (1990-2019) under the default scenario (i.e., without additional irrigation); and decline even markedly further under irrigation expansion scenarios. These severely reduced summer low flows pose significant risks to river ecosystems and long-term river resilience. Simulations further revealed that reductions in river discharge are largely independent of the source of water abstraction (groundwater or surface waters), which is likely related to the fact that baseflow from groundwater reservoirs is one of the most important sources of river water during dry periods.
The results highlight the importance of effective water management adaptation strategies in intensively farmed regions of Central and Eastern Europe to prevent future water scarcity and to reduce the ecological risks associated with increasingly severe and prolonged low-flow periods.
How to cite: Schreiber, S. J., Schmitt, A., Clédat, A., and Greve, P.: Hydrological impacts of enhanced irrigation under drought conditions in Europe, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21169, https://doi.org/10.5194/egusphere-egu26-21169, 2026.
Quantifying long-term water withdrawals at the river basin scale remains a major challenge due to scarce direct observations and the interaction between natural hydrological processes and human activities. This study introduces a semi-distributed modelling framework to reconstruct natural discharge and infer net water withdrawals from observed streamflow records. Net water withdrawals are estimated as the difference between simulated natural and observed discharges. Natural discharge is simulated by partitioning precipitation (BIGBANG v.8 database, ISPRA) into surface runoff and groundwater recharge through an infiltration-based scheme. Groundwater contributions are represented using a linear reservoir to capture delayed baseflow response. The model is governed by three parameters, which are (i) the infiltration coefficient, (ii) the ratio between the hydrogeological and catchment area and (iii) storage coefficient of the linear reservoir model. Calibration is performed over 1954–1965, assumed minimally impacted by withdrawals, by maximizing Nash–Sutcliffe Efficiency and minimizing volume bias, with Kling–Gupta Efficiency as an additional metric. The approach is applied to the Tiber River basin closed at Ripetta station (central Italy) using data spanning 1954–2023. The calibrated model reproduces observed discharge dynamics satisfactorily. The reconstructed natural discharge series is then extended to 2023 to proceed with the calculation of net withdrawals. The resulting time series shows a clear long-term linear increasing trend, with significant interannual variability. Statistical tests (Chi-square and t-tests) on residuals confirm normality and a mean not significantly different from zero, supporting the robustness of the inferred trend. This approach enables spatially coherent reconstruction of water withdrawals using commonly available hydrological data, providing a valuable tool for assessing anthropogenic pressures where direct measurements are lacking. Results for the Tiber River basin reveal progressive intensification of human influence on water resources over seven decades, offering insights for water management, policy development, and hydrological forecasting in human-modified catchments.
How to cite: Pomarico, I., Fiori, A., Volpi, E., and Zarlenga, A.: From Natural Flow to Anthropogenic Pressure: Quantifying Water Withdrawals in the Tiber River Basin Over 70 Years, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13732, https://doi.org/10.5194/egusphere-egu26-13732, 2026.
Smallholder farmers in the global south predominantly rely on rainfed agriculture, making accurate precipitation forecasts crucial for agricultural decision-making. However, the low reliability, limited skills, and accessibility of scientific forecasts (SF) derived from Numerical Weather Prediction models in rural communities hinder the use of SF. Instead, smallholders often rely on indigenous knowledge to predict the rainfall based on observed local indicators, e.g., meteorology, animals’ behavior, and astronomy, hereafter we call it local forecasts (LF). However, the use of LF also faces challenges, including the loss of LF knowledge since it is communicated orally, not documented, not always observable, or not deemed useful. In addition, the use of local forecast also faces challenges by increasing climate variability, which undermines farmers’ confidence in their forecast. Studies conducted in Africa evaluating SF and LF skills have demonstrated that LF’s performance is comparable to or even outperforms the SF. Furthermore, these studies highlight that integrating SF and LF, known as hybrid forecast (HF), results in higher forecast performance than either SF or LF alone. In this study, we aim to develop an HF system that combines SF and LF using machine learning approaches to improve precipitation predictions in northern Ghana. Four rain gauges were installed at the field and used to evaluate the performance of SF, LF, and HF to predict precipitation events based on the Hanssen-Kuipers discriminant (HK) and accuracy skill assessment metrics. Results show that the HF achieved a HK value of 0.79, outperforming the scientific forecast (HK = 0.50), and local forecasts (HK = 0.37). In terms of accuracy, the HF also led with a score of 0.92, followed by the SF at 0.69. Similar to its HK, LF has the lowest accuracy of 0.65. Our study proved that ML approaches can be highly effective in developing a seamless forecasting system, specifically the HF, which outperforms the accuracy of individual forecasts alone. Such enhanced precipitation forecasts could enable smallholder farmers in the Global South to make better-informed agricultural decisions, ultimately enhancing their livelihoods.
How to cite: Sutanto, S. J., Bosdijk, J., Benedict, I., Moene, A., Milosevic, D., Ludwig, F., and Paparrizos, S.: Advancing Climate Services Through Hybrid Precipitation Forecasts That Integrate Indigenous Knowledge, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5968, https://doi.org/10.5194/egusphere-egu26-5968, 2026.
Urban catchments are highly human-influenced hydrological systems in which drainage networks, pumping stations, and engineered conveyance structures fundamentally modify runoff generation and flow dynamics. These anthropogenic controls introduce strong nonlinearity and non-stationarity, challenging short-term hydrological forecasting and reducing the effectiveness of reactive flood control, particularly under intensifying extreme rainfall. This study develops an Intelligent Flood Control Decision Support System (IFCDSS) that integrates data-driven hydrological forecasting with adaptive operational control to support proactive urban flood management. At the catchment scale, short-term flood inundation nowcasting is achieved by combining Principal Component Analysis (PCA) and Self-Organizing Maps (SOM) with Nonlinear Autoregressive models with exogenous inputs (NARX). This approach enables efficient extraction of dominant inundation patterns from high-resolution two-dimensional flood maps and provides reliable multi-step-ahead forecasts at 10-minute resolution up to one hour. At the infrastructure scale, hybrid deep learning models (CNN–BP) are used to generate multi-input, multi-output forecasts of sewer, forebay, and river water levels, achieving high predictive skill under rapidly evolving rainfall and operational conditions. Forecast outputs are translated into operational decisions through a decision layer integrating NSGA-III for multi-objective optimization, TOPSIS for solution ranking, and an Adaptive Neuro-Fuzzy Inference System (ANFIS) for real-time pump control. Application to a major pumping-station catchment in Taipei, Taiwan, demonstrates that the system delivers actionable forecasts and control strategies within seconds. Compared with manual operation, the IFCDSS achieves more robust trade-offs among flood mitigation, energy efficiency, and operational reliability. The results highlight the importance of explicitly representing human interventions in urban hydrological forecasting and demonstrate how intelligent decision support can enhance flood preparedness in complex, human-regulated catchments under climate change.
How to cite: Chang, F.-J., Chang, L.-C., Yang, M.-T., and Liou, J.-Y.: Proactive Hydrological Forecasting and Intelligent Decision Support in Human-Regulated Urban Catchments, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6078, https://doi.org/10.5194/egusphere-egu26-6078, 2026.
Climate change and rapid urbanisation have intensified the frequency and consequences of extreme flood events. During floods, transportation systems may fail, leading to traffic breakdowns, prolonged exposure, and cascading impacts on emergency response and wider urban functioning. Flood risk is therefore shaped not only by the physical dynamics of inundation, but also by how people perceive, respond to, and adapt their mobility under evolving hazard conditions, exemplifying a Coupled Human And Natural System (CHANS). This tight coupling between hazard evolution and human response makes it essential to represent hazard-human interactions at the event timescale, particularly for reliable flood forecasting, early warning, and emergency preparedness. However, capturing adaptive human mobility under dynamically changing flood conditions remains a major challenge, especially within a CHANS modelling framework.
Agent-based modelling (ABM) has been increasingly applied to represent human behaviour during floods, often coupled with hydrodynamic inundation models. However, most existing implementations rely on offline or weakly coupled co-simulation, in which flood dynamics and human behaviour are computed in separate platforms and synchronised through frequent data exchange. Such data-exchange-driven approaches become increasingly expensive when high-frequency updates are required, limiting their capability to represent real-time feedback between flood evolution and human mobility.
In this study, we present a CHANS modelling framework built upon the GPU-accelerated High Performance Integrated hydrodynamic Modelling System (HiPIMS) for predicting flood hydrodynamics, fully coupled with an agent-based module within the same computational framework to represent human mobility. This enables seamless simulation of interacting flood conditions and human responses. Human mobility is represented by autonomous agents within a unified architecture that supports pedestrians, cyclists, and vehicles. Mobility agents exhibit heterogeneous behavioural attributes, including risk aversion, awareness, compliance, and patience, and interact within a shared, dynamically evolving flood environment.
The framework is demonstrated through an urban case study in Newcastle upon Tyne, with data from the Urban Observatory for model validation. Further simulations are conducted for light, medium, and heavy rainfall scenarios to analyse adaptive transport responses under different flood conditions.
By supporting large numbers of agents and real-time hazard-human interactions within a single computational environment, the proposed framework enables systematic analysis of human adaptive behaviour and system-level disruption during flood events. This work provides a new methodological basis for characterising flood risk in a coupled human and natural systems context, with clear implications for early warning, emergency response planning, and integrated flood forecasting.
How to cite: Tong, X., Liang, Q., Chen, H., and Zong, Y.: A Coupled Human And Natural System (CHANS) Framework for Human Mobility during Flood Events, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10058, https://doi.org/10.5194/egusphere-egu26-10058, 2026.
Anthropogenic activities such as rapid urbanization, highly regulated dams and reservoirs, and widespread land-use changes significantly modify natural streamflow dynamics, making streamflow prediction increasingly challenging for traditional hydrological models. Deep learning approaches, particularly the long short-term memory (LSTM) network, have gained popularity due to their ability to capture long-term dependencies in hydrological time series. However, the purely data-driven nature of LSTM limits its reliability in human-influenced watersheds. The Mass-Conserving LSTM (MC-LSTM) addresses this limitation by incorporating mass-balance constraints directly into its architecture, enabling physically consistent predictions. Despite this advancement, a systematic comparison between LSTM and MC-LSTM in human-influenced hydrological systems remains limited. In this study, we evaluate the predictive advantage and hydrologic suitability of MC-LSTM across 51 human-influenced watersheds in India. The watersheds are categorized into low- and high-human-influenced categories using a composite disturbance index (CDI), derived from the number of dams, reservoir storage, cropland fraction, built-up fraction, and population density. This setup allows us to address a key question: Does incorporating mass balance constraints into LSTM improve streamflow reliability in highly regulated watersheds? The results show that MC-LSTM substantially outperforms traditional LSTM in highly human-influenced watersheds, yielding a significantly higher median NSE (MC-LSTM: 0.69; LSTM: 0.62). MC-LSTM also demonstrates several additional benefits, including improved high-flow prediction, reduced sensitivity to training data size, and slightly enhanced performance in semi-arid watersheds. In contrast, traditional LSTM tends to underestimate high-flow, depends on larger training datasets, and performs poorly in semi-arid and highly regulated basins. These findings underscore the importance of incorporating mass balance into DL-based hydrological models to enhance reliability in real-world applications.
How to cite: Sahu, G. and Sharma, A.: Assessing the Hydrologic Suitability of MC-LSTM for Reliable Streamflow Prediction in Human-Influenced Watersheds, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-617, https://doi.org/10.5194/egusphere-egu26-617, 2026.
Groundwater influences flood dynamics by modulating subsurface saturation states and engaging in complex interactions with surface water in multiple pathways. Accurately representing these processes is essential for physically consistent flood prediction, risk assessment, and mitigation strategies. However, groundwater-related processes remain poorly resolved in most existing flood modeling frameworks, which typically employ oversimplified representations of subsurface flow. In this study, we present HiPIMS-GWF, a three-dimensional, variably saturated groundwater flow module integrated into the High-Performance Integrated Hydrodynamic Modelling System (HiPIMS). HiPIMS is a GPU-accelerated, high-performance flood model capable of simulating catchment-scale fluvial flooding driven by extreme rainfall events. The HiPIMS-GWF module provides functionality to solve the three-dimensional Richards equation using both iterative and non-iterative numerical schemes, enabling explicit representation of surface-subsurface water exchanges within a unified modeling framework. Model accuracy is evaluated against a suite of standard numerical benchmark problems, and computational scalability and efficiency are assessed on a multi-GPU computing platform.
Beyond the acute phase of flooding, we are also interested in investigating the long-term impacts of flood events on groundwater and surface water systems after its recession. Because catchment-scale groundwater dynamics evolve over temporal scales that can be orders of magnitude longer than those of surface flooding, capturing the full hydrological response necessitates extended simulation capabilities beyond the time horizon of flood events. To this end, HiPIMS-GWF introduces a novel modeling flexibility: once floodwater recedes, the high-resolution, physics-based surface hydrodynamic component can be swtiched to a computationally efficient, hydrologic model tailored for long-term watershed simulation. Critically, the spatially distributed fields of water saturation and surface water depth generated by the fully physics-based simulation serve as initial conditions for the long-term mode, ensuring continuity in the representation of system states across timescales. The overall accuracy and robustness of this integrated modeling framework are validated against a real-world flood event.
How to cite: Zong, Y., Tong, X., and Liang, Q.: HiPIMS-GWF: A GPU-Accelerated 3D Variably-Saturated Subsurface Solver for Integrated Flood Modeling with Groundwater Components, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-783, https://doi.org/10.5194/egusphere-egu26-783, 2026.
High-resolution shallow-water (SW) models are critical for flood and inundation forecasting, yet their operational efficiency is often bottlenecked by the computational cost of repeatedly solving intercell Riemann problems. While emerging machine-learning surrogates (e.g., PINNs and neural operators) can accelerate PDE prediction, they often struggle to meet the rigorous requirements of hydrodynamic modelling. Specifically, these end-to-end models generally enforce physics only via soft constraints, leading to non-physical mass leakage and error accumulation over long durations. They also suffer from spectral bias, which hinders the sharp capture of discontinuities and wet–dry fronts. Furthermore, they typically lack cross-geometry generalizability, requiring costly retraining when boundary conditions or mesh resolutions change. This study proposes a structure-preserving hybrid strategy that integrates deep learning into a classical Godunov-type finite-volume (FV) solver. Rather than approximating the global solution map, we employ a neural network as a local, plug-in surrogate specifically for intercell flux evaluation. This network learns a discretization-aware operator, mapping local reconstructed interface states to normal-aligned numerical fluxes. Crucially, by embedding this learned surrogate within the standard FV backbone—retaining CFL-controlled time marching and wetting–drying treatments—the hybrid solver strictly enforces mass conservation through rigorous flux-difference assembly. Because the model learns local interface physics rather than global flow patterns, it exhibits strong cross-resolution generalization: a model trained on a specific grid can be deployed directly on different mesh densities and unseen initial conditions without retraining. This work establishes a scalable pathway for integrating deep learning into hydrodynamic solvers, combining the computational speed of machine learning with the reliability and conservation properties of numerical mechanics.
How to cite: Wu, J., Liang, Q., and Chen, H.: A Structure-Preserving Neural Flux Surrogate for Efficient Shallow-Water Modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15780, https://doi.org/10.5194/egusphere-egu26-15780, 2026.
With the increasing complexity of urban systems, traditional flood risk assessments often fail to capture the systemic vulnerability arising from infrastructure interdependencies. This study integrates high-fidelity hydrodynamic modeling (HiPIMS), physics-guided deep learning, and complex network theory to develop a novel dynamic flood chain risk assessment framework. To enhance hazard prediction, a Physically-Guided Spatiotemporal Mixture-of-Experts (PG-ST-MoE) network is constructed. By leveraging hydrodynamic outputs as guidance, this network dynamically predicts high-precision spatiotemporal flood probabilities, effectively bridging the gap between idealized physical simulations and real-world flood occurrences. Crucially, the framework transcends static hazard mapping to analyze disaster chain effects. By simulating cascading failures within infrastructure-community networks, it quantifies how localized physical damage propagates into widespread functional paralysis and identifies functional islands where critical services are severed despite the absence of direct flooding. The framework has been deployed in the San Isabel Basin, South America, demonstrating the capability to reveal hidden systemic risks in data-scarce regions. This study offers a paradigm shift from static exposure assessment to dynamic chain-reaction analysis, providing actionable insights for preventing systemic collapse and enhancing adaptive emergency management.
How to cite: Qu, Z., Kou, M., and Zhang, K.: From Inundation to Systemic Collapse: A Dynamic Flood Risk Assessment Framework Coupling Hydro-Deep Learning with Cascading Failure Analysis, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-908, https://doi.org/10.5194/egusphere-egu26-908, 2026.
Surface water flooding (SWF), when intense rainfall overwhelms drainage systems and inundates streets, homes, and infrastructure, is the most widespread form of flooding in the UK and a rapidly escalating global hazard under climate change and growing urban exposure. In England alone, around 3.2 million properties are at risk. In the extreme cases, SWF can develop rapidly with little or no warning and exhibit highly dynamic, fast-moving flow conditions, behaving like an “inland tsunami”, with debris-laden flood waves overwhelming streets, vehicles, and buildings within minutes. It can paraly transport, disrupting essential services, and, in some cases, cause catastrophic loss of life. Recent events have highlighted the deadly consequences of such flooding, including the July 2021 Zhengzhou (China) flood with nearly 400 fatalities, the 2021 floods in Germany and Belgium with over 200 deaths, the October 2024 flash floods in Valencia, Spain causing 237 fatalities, and widespread cyclone- and monsoon-driven flooding across South and Southeast Asia in 2025 causing more than 1,000 deaths and displacing millions.
At large spatial scales, SWF does not occur as isolated local events. Intense rainfall may occur simultaneously or sequentially over wide areas, and interconnected river networks, drainage systems, and infrastructure can couple multiple local flood processes into a single, spatially extensive flood system. Understanding and predicting such large-scale, interacting flood dynamics is therefore essential for both national-scale risk assessment and real-time forecasting.
Numerical modelling provides an indispensable tool for representing SWF processes. However, due to their highly transient, shock-like behaviour, hydrological or simplified hydraulic approaches are often insufficient. Fully hydrodynamic models solving the two-dimensional shallow water equations with shock-capturing capability are required, but their computational cost has historically limited their application to city or local-catchment scales. Scaling such models to regional or national extents is not a simple domain enlargement problem, but introduces coupled challenges related to computational demand, terrain resolution, and modelling strategy. As a result, fundamental questions remain regarding the feasibility of national-scale hydrodynamic modelling, the computational resources required, the sensitivity of flood hazard metrics to DEM resolution, and the trade-offs between alternative large-scale simulation strategies.
To address these questions, we conduct a national-scale hydrodynamic flood modelling experiment over England using the High-Performance Integrated hydrodynamic Modelling System (HiPIMS) accelerated by multi-GPU computing. Event-based simulations are performed over the England at DEM resolutions of 10 m, 20 m, and 40 m to systematically quantify resolution effects on flood hazard representation and associated computational costs. The experimental design also enables comparison between alternative national-scale modelling strategies, including domain-wide versus partitioned simulations.
The results delineate the practical feasibility limits, resolution sensitivity, and performance trade-offs of national-scale hydrodynamic flood modelling, and provide quantitative guidance on the computational and data requirements for moving towards national-to-street-scale, physics-based surface water flood forecasting and risk assessment.
How to cite: Liang, Q., Tong, X., Chen, H., and Jiang, J.: Towards national-scale hydrodynamic flood modelling: feasibility, resolution sensitivity, and computational trade-offs, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5294, https://doi.org/10.5194/egusphere-egu26-5294, 2026.
Urban Nature-based Solutions (NbS) are critical for flood mitigation, yet existing modelling approaches have limitations in explicitly capturing their performance during extreme events. Conventional approaches typically couple NbS modules with two-dimensional (2D) surface flow models as separate executables. This "loose coupling" fails to capture the two-way dynamic interactions between surface runoff and NbS features, especially when systems approach full saturation or exceed their design capacity.
This study introduces HiPIMS-NbS, a high-performance, GPU-accelerated modelling framework that embeds physical behaviours of multi-layer NbS directly into high-resolution 2D shallow water computations. Unlike traditional models with predetermined drainage areas, HiPIMS-NbS calculates flow directions and surface-NbS exchange rates dynamically across the 2D domain. Surface ponding depths influence NbS infiltration rates while NbS storage regulates surface water availability within the unified GPU-accelerated computational framework, achieving dynamic two-way coupling.
The model was validated against SWMM benchmarks and field-scale bioretention experiments. To demonstrate city-scale applications, HiPIMS-NbS was applied to the 2012 "Toon Monsoon" flood event in Newcastle upon Tyne, UK, to evaluate various NbS implementation scenarios. Results demonstrate that the model achieves the computational efficiency required for city-scale simulations while capturing key NbS behaviours under realistic overflow conditions. This integrated approach provides a robust modelling basis for urban planners to optimise NbS placement and design for the extreme rainfall events projected under changing climate.
How to cite: Jiang, J., Chen, H., Tong, X., Varley, D., and Liang, Q.: A Fully Integrated Hydrodynamic-NBS Model for City-Scale Flood Risk Assessment under Extreme Rainfall, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17989, https://doi.org/10.5194/egusphere-egu26-17989, 2026.
Urban flooding poses an increasing threat to lives and property in urbanized environments under climate change and growing human exposure. Digital Twin (DT) concept provides a potential framework for integrating real-time monitoring, numerical simulation, and decision support. However, DT implementations that exploit dense senor networks and high-resolution, physics-based hydrodynamic flood models to enable real-time, bi-directional information exchange between physical and virtual systems remain limited. In this study, we develop a campus-scale DT framework that couples real-time monitoring, high-resolution hydrodynamic modeling, 3D virtual representation within a unified data and computational environment supported by embedded data-analytics capabilities.
A high-resolution 3D digital campus model is reconstructed from ultra-high-resolution LiDAR point clouds to provide the geometric basis for spatial representation and data management. A dense IoT monitoring network, comprising rainfall gauges, water-level sensors, CCTV, and pipe flow meters, is deployed to acquire and transmit high-frequency hydrometeorological and hazard-related observations in real time. At the core of the framework is the dynamic coupling between real-time rainfall observations and the GPU-accelerated High-Performance Integrated hydrodynamic Modelling System (HiPIMS), which resolves surface water inundation processes at high spatial and temporal resolution. Static spatial data, real-time monitoring observations, and model outputs are ingested, harmonized, and managed within the unified data and computational environment, enabling automated model execution and coordinated system operation. Monitoring data and simulation outputs are mapped directly onto the 3D virtual environment to provide real-time visualization of spatiotemporal evolution of flooding and to support flood warning and emergency management.
The reliability of flood simulations is evaluated against historical flood records and further assessed through continuous comparison with in-situ water-level observations. The framework supports near-real-time flood forecasting and systematic identification of high-risk locations, providing information for early warning and emergency decision-making. Emergency interventions, such as deployment of temporal flood defenses and mobile pumping stations, can in turn influence flood dynamics and risk; these changes are subsequently captured by the monitoring-modelling system and reflected in updated DT outputs. This establishes a closed-loop, real-time monitoring-simulation-decision-feedback cycle, forming an operational DT framework for urban flood management.
How to cite: Guo, M., Zong, Y., Tong, X., and Liang, Q.: A Campus-scale Digital Twin Framework for Urban Flood Monitoring, Simulation and Management, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19143, https://doi.org/10.5194/egusphere-egu26-19143, 2026.
Escalating flood frequency and intensity, driven by anthropogenic land use modifications and climate variability, pose critical challenges to watershed management worldwide. This study examines the hydrological impacts of the Billion Tree Tsunami (BTT) reforestation initiative, implemented in 2014 across Pakistan's Khyber Pakhtunkhwa province, with specific focus on flood attenuation dynamics in the Swat River catchment. We integrate land use and land cover (LULC) change analysis spanning three decades (1990–2024) with Long Short-Term Memory (LSTM) neural networks to assess historical discharge patterns and project future hydrological conditions through 2050. LULC analysis reveals substantial landscape transformation, including significant forest expansion, marked reduction in barren land, and agricultural land modifications. Statistical evaluation demonstrates notable flood mitigation effects post-intervention, with 15% reduction in peak flows and decreased discharge intensification rates. The LSTM models exhibit strong predictive performance (R² = 0.87), forecasting a 20–25% reduction in peak discharge events by 2050 under continued reforestation scenarios. These findings underscore the efficacy of large-scale reforestation as a nature-based solution for flood risk reduction and demonstrate the value of integrating machine learning approaches with conventional hydrological modeling for enhanced watershed management strategies.
How to cite: Shah, L.: Quantifying Nature-Based Flood Risk Reduction Through LSTM Modeling: Evidence from Pakistan's Swat River Basin, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5559, https://doi.org/10.5194/egusphere-egu26-5559, 2026.
High-resolution (approximately 1 m) flood modelling is increasingly recognised as essential for resolving flow pathways and hydraulic connectivity in complex urban environments. At this scale, flood dynamics are strongly controlled by micro-topographic features, including hydraulically permeable elements beneath vegetation and bridge structures, as well as small-scale obstructions such as kerbs and surface discontinuities. However, conventional Digital Terrain Model (DTM) generation approaches struggle to reliably represent such features due to vegetation occlusion and sensor-specific terrain acquisition limitations, often necessitating extensive manual intervention.
This study presents a semi-automated terrain reconstruction framework that integrates Unmanned Aerial Vehicle-borne LiDAR, Unmanned Aerial Vehicle （UAV）oblique photogrammetry, handheld LiDAR Simultaneous Localization and Mapping (SLAM), and Real-Time Kinematic Global Navigation Satellite Systems (RTK-GNSS) measurements to generate flood-ready DTMs for high-resolution hydrodynamic modelling. Rather than treating multi-source datasets as interchangeable inputs, the framework explicitly exploits their differing information characteristics and spatial sensitivities to occlusion and ground accessibility. UAV LiDAR provides spatially continuous but occlusion-prone surface measurements, handheld LiDAR SLAM offers dense ground-level observations in vegetated and structurally complex areas, and RTK-GNSS provides sparse but high-accuracy elevation control.
An initial DTM is established through adaptive fusion of morphologically filtered UAV-derived DTMs and SLAM-derived ground observations, supported by vegetation indices extracted from digital orthophoto maps and SLAM point-density metrics. To address residual elevation errors arising from partial occlusion and sensor limitations, a residual learning strategy based on a U-Net architecture is employed to predict local elevation corrections relative to RTK-GNSS ground truth. The learning component is explicitly constrained to operate as a local correction mechanism rather than an end-to-end terrain predictor, thereby preserving physical plausibility and spatial consistency of the reconstructed terrain.
The framework is demonstrated over the Zhengzhou University campus (approximately 1 km²), encompassing diverse building typologies, vegetation densities, and pedestrian and vehicular infrastructure. The hydrodynamic relevance of the reconstructed DTM is evaluated using the High-Performance Integrated Hydrodynamic Modelling System（HiPIMS） through comparative two-dimensional simulations of historical extreme flood events. Results demonstrate that improved representation of micro-topographic controls significantly enhances simulated flow connectivity, inundation extent, and inundation timing relative to conventional terrain products, and provides a transferable workflow for campus- and neighbourhood-scale flood modelling and risk assessment and urban digital twin applications.
How to cite: Zhang, Y., Cheng, H., Liang, Q., Zong, Y., and Shi, B.: Multi-Sensor Terrain Reconstruction for High-Resolution Urban Flood Modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19225, https://doi.org/10.5194/egusphere-egu26-19225, 2026.
Thallium (Tl) is a toxic metal. Its contamination in aquatic ecosystems is likely to intensify in the upcoming decades. Exploding demand for Li, rare earths, Co, Ni and Cu is pushing mining into various Tl-enriched ores/deposits. Every new battery, wind-turbine and power line therefore may carry a hidden Tl footprint that conventional lime-dosing treatment fails to retain. Thus, the low-carbon transition may turn Tl into an unexpected river contaminant. Field-to-lab experiments reveal that Tl⁺ mimics K⁺ in fish gills and algal cells, inducing oxidative stress, and Na⁺/K⁺-ATPase collapse at very low concentrations. These impacts may induce a series of alterations at the physiological, biochemical, and genomic expression levels. These disruptions can, in turn, undermine the survival and demographic structure of these organisms, ultimately posing risks to human beings via complex trophic networks. Given the urgency of this situation, it is therefore suggested that Tl should be incorporated into routine river-health monitoring in mining-impacted basins and propose a cost-effective proxy-screening and forecasting protocol.
How to cite: Liu, J., Guo, H., Zhu, Z., and Zhong, Y.: Hidden River Contamination from Critical-Metal Mining Demands Proactive Monitoring and Forecasting , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16279, https://doi.org/10.5194/egusphere-egu26-16279, 2026.
Posters virtual: Fri, 8 May, 14:00–18:00 | vPoster spot A
EGU26-16207 | ECS | Posters virtual | VPS11
Data-Driven LSTM Architectures for Reservoir Inflow ForecastingFri, 08 May, 14:51–14:54 (CEST) vPoster spot A
Accurate forecasting of reservoir inflow is crucial for managing water resources, maintaining a balance between water supply and demand, preventing floods, supporting hydropower production, and planning irrigation. India, ranking third globally, with more than 5,000 dams, faces challenges in reservoir operations due to hydrological variability caused by the monsoon. While ensuring demand, supply, and flood security requires high water levels, the reservoir also needs to maintain a certain amount of free storage to accommodate high inflows. While Long-Short Term Memory (LSTM) models have been widely used for inflow forecasting, traditional LSTM models often limit their ability to capture sudden hydrological extremes and accurately represent peak timings. Therefore, a comparative evaluation of various advanced LSTM variants is necessary to identify architectures that are more reliable for modelling nonlinear inflow dynamics. Our study introduces a specialised type of recurrent neural network, specifically the LSTM framework, for forecasting daily reservoir inflow. Our methodology uses a structured feature engineering strategy that integrates hydrometeorological forcings, hydrological state variables, and outputs from the CaMa-Flood hydrodynamics model. A permutation-based feature importance analysis, in terms of the increase in mean absolute error, highlights that antecedent precipitation and lagged upstream reservoir outflow are the main influencing factors for the inflow forecast within a multivariate sequence-to-one LSTM framework. Overall, this framework provides a strong, scalable, and practical solution for inflow forecasting. By supporting timely operational decisions for water release, flood preparedness and storage optimisation, the framework serves as an effective tool for managing reservoirs.
How to cite: Mani, D. and Mishra, V.: Data-Driven LSTM Architectures for Reservoir Inflow Forecasting, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16207, https://doi.org/10.5194/egusphere-egu26-16207, 2026.
EGU26-16490 | ECS | Posters virtual | VPS11
Estimation of Ecological Flow for Major Indian River Basins under Changing ClimateFri, 08 May, 14:54–14:57 (CEST) vPoster spot A
Streamflow provides critical support for the biodiversity of aquatic and riparian ecosystems, sediment transport, and nutrient cycling. Therefore, a minimum streamflow in rivers is crucial for sustaining the proper functioning of aquatic habitats. However, lean or low-flow regimes have been significantly altered by various human activities, such as dam construction, flow diversion for irrigation purposes, industrialisation, and urbanisation. Moreover, changing climate is causing erratic monsoons, increased temperatures, and more prolonged droughts, thereby maintaining ecological flow has become increasingly challenging and urgent to preserve the riverine ecosystems. Our aim is to develop a robust, data-driven framework for estimating environmental flows (E-flows) across 55 stations in major Indian river basins. The primary objective is to assess the quantity and timing of streamflow required to sustain the various river ecosystems, utilising hydrological indicators and long-term datasets, such as temperature and precipitation from the Indian Meteorological Department (IMD). Changes in streamflow characteristics are assessed by comparing observed and machine learning-based naturalised flows, enabling the isolation of reservoir-induced impacts on the streamflow regime, magnitude, duration, and seasonal timing. The study hypothesises that, with the use of observed streamflow data, naturalised streamflow reconstruction and a multi-indicator hydrologic approach, integrating Indicators of Hydrological Alteration (IHA), the Range of Variability Approach (RVA), and Flow Duration Curve (FDC) analysis, can provide reliable E-Flow estimates at regional and national scales. By comparing indicator-based benchmarks derived from observed and naturalised streamflow, the stations are classified according to the degree of hydrologic alteration, thereby supporting scientifically informed river management and policy decisions.
How to cite: Sahil, S. and Mishra, V.: Estimation of Ecological Flow for Major Indian River Basins under Changing Climate, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16490, https://doi.org/10.5194/egusphere-egu26-16490, 2026.
EGU26-1322 | ECS | Posters virtual | VPS11
From Contamination to Forecast: Linking Anthropogenic Hydrological Change to Ecological Risk in Poyang LakeFri, 08 May, 15:03–15:06 (CEST) vPoster spot A
Anthropogenic activities including industrial and agricultural discharges, sand mining, and water regulation have drastically altered the hydrological regime and water quality of Poyang Lake, China’s largest freshwater lake. These modifications lead to non-stationary inputs of heavy metals (e.g., As, Hg, Cr, Se) and nutrients, driving eutrophication and posing significant risks to aquatic ecosystems.
This study analyses multi-year (2018–2020) water quality data from key inflow sites to quantify human impacts on contaminant regimes. Results reveal strong seasonal patterns: heavy metal concentrations (As, Hg) peak during low-flow periods, whereas nutrient loads and algal blooms intensify following high-flow events linked to agricultural runoff. This dynamic hydrological contamination directly threatens the endangered Yangtze finless porpoise (Neophocaena asiaeorientalis), with tissue analyses showing high bioaccumulation of Hg and Cu in the liver, indicating significant ecological risk.
Building on these findings, we highlight the urgent need for forecasting frameworks tailored to human-influenced catchments. We propose integrating process-based hydrological models with water quality modules and machine learning techniques to simulate contaminant transport under non-stationary climatic and anthropogenic drivers. Furthermore, we demonstrate how remote sensing and continuous sensor data can improve the monitoring of pollutant sources and algal blooms. Finally, we outline a pathway towards ecological risk forecasting by coupling hydrological-water quality predictions with bioaccumulation models for vulnerable species.
This work underscores the critical gap in forecasting tools for heavily modified systems and provides a case for developing coupled human-natural models to support early warning systems and adaptive management strategies for biodiversity conservation.
How to cite: Sabir, A., Hua, W., Hanan, A., Deng, Y., and Wu, X.: From Contamination to Forecast: Linking Anthropogenic Hydrological Change to Ecological Risk in Poyang Lake, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1322, https://doi.org/10.5194/egusphere-egu26-1322, 2026.
