ABSTRACT. Climate change significantly impacts water resources and other fundamental aspects of daily life in many regions of the world, causing an increase in the frequency of extreme weather events such as droughts and floods [1]. These phenomena represent a growing challenge, especially regarding water supply management, which becomes particularly critical in urban areas affected by prolonged droughts and water scarcity episodes [2]. Cities face a combination of aggravating factors: on the one hand, the constant increase in the urban population intensifies the demand for water resources to meet basic needs; on the other hand, existing water infrastructures, often obsolete, lose efficiency in ensuring adequate distribution to users [3]. Many cities adopt emergency strategies to manage unexpected water scarcity in response to these issues. A common practice is implementing an intermittent water supply, which consists of delivering water intermittently, thus reducing resource losses in the pipes and trying to rationalise water consumption through a controlled distribution among users [4]. While these measures are helpful to contain losses and address immediate crises, they put additional stress on existing water infrastructures. The long-term effects of such practices, including accelerated network deterioration and increased maintenance needs, are well documented in the scientific literature [5]. End-users, in turn, adopt self-adaptation strategies to cope with disruptions in water distribution. One of the most widespread solutions is the installation of private tanks to collect and store as much water as possible during the active service. However, these reservoirs are often oversized compared to actual needs due to a common perception that water is an indispensable resource for daily life and a general sense of insecurity that pushes users to accumulate adequate reserves. The presented study analysed the impact of water scarcity and users' self-adaptation strategies on overall water demand under conditions of intermittent supply. To conduct this analysis, an experimental campaign was carried out on a real water network located in the metropolitan area of Palermo. Water consumption was monitored at the inlet and outlet of private tanks in a specific residential area. During the experiment, different pressure conditions and private tank oversizing were considered to understand better consumption patterns related to intermittent water distribution. The analysis was supported by using a short-term water demand forecasting model capable of reproducing periodic patterns observed on an annual, weekly and daily basis. This approach allowed us to assess how users adapt to water scarcity and to compare the dynamics of the real network with those of private tanks. The results clearly show adaptation strategies and their implications on water resource management in contexts characterised by intermittent supplies.

ABSTRACT. Our study analyzes and compares the performance of water distribution systems in four Moroccan cities—Al-Hociema, Tangier, Marrakech, and Kenitra—using key performance indicators such as Non-Revenue Water (NRW), Water Supply (WS), Unavoidable Annual Real Losses (UARL), Real Losses (CARL), Hydraulic load index (q0), Unit water loss(qs), Real Leakage Balance (RLB) and Infrastructure Leakage Index (ILI). The data, which covers the years 2018–2021, shows that the cities have different degrees of efficiency and difficulties. Tangier and Al Hoceima improved the most in water loss reduction (NRW, ILI, and CARL). Kenitra still has high ILI and leakage issues despite some progress. Marrakech has stable indicators, but further efficiency improvements are needed. Al Hoceima and Tangier show the best progress in overall network performance. This analysis provides a foundation for future research and policy recommendations to enhance the sustainability and efficiency of water distribution systems in Morocco.

ABSTRACT. Tidal rivers are complex systems at the interface between continental surfaces and the ocean. Water levels as well as river discharge are highly influenced by tidal dynamics, especially for flat systems such as the Dong Nai-Saigon river system in Vietnam (Camenen et al., 2021). Additionally, the propagation of sediment particles or of a pollution can be difficult to apprehend due to the back-and-forth advection combined with dispersion. The River Saigon flows through the Ho Chi Minh City (HCMC) megalopolis and understanding its hydrodynamics is crucial in terms of flood risk assessment, saline intrusion, pollution and eutrophication. A 1D model of the Dong Nai-Saigon system has been recently built and validated (Camenen et al., 2023, Rodrigues do Amaral et al., 2025) even though there is a lack of data to evaluate the net discharge coming from upstream. Nevertheless, Camenen et al. (2023) clearly showed the interest of 1D numerical modelling to identify the main factors explaining the behaviour of the Saigon during extreme events. Also, if small canals have minor impacts on the Saigon dynamics, they are strongly influenced by the Saigon dynamics. Potential pollution of the river system could either originate from rice paddy fields in the upstream part of the River Saigon (nutrients) or from the sewage system in HCMC (overflow during flood events). The purpose of this paper is thus to study the flow and pollution dynamics in some of the main canals in HCMC using the Mage-AdisTS solvers developed at INRAE Lyon. We present here the calibration of the model as well as an analysis of the pollution dynamics for different scenarios. Camenen, B., Gratiot, N., Cohard, J.-A., Gard, F., Tran, V.Q., Nguyen, A.-T., Dramais, G., van Emmerik, T. & Némery, J. (2021). Monitoring discharge in a tidal river using water level observations: application to the Saigon River, Vietnam. Sci. Total Environ. 10.1016/j.scitotenv.2020.143195 Camenen, B., Gerarduzzi, K., Kieffer, L., Terraz, T., Rodrigues do Amaral, F., Gratiot, N. & Pellarin, T. (2023). 1D numerical modelling of a complex tidal river : case of the River Saigon, Vietnam. Proc. 7th SimHydro conference, Nov. 2023, Chatou, France. Rodrigues do Amaral, F., Camenen, B., Nguyen Trung, T., Anh Tu, T., Pellarin, T. & Gratiot, N. (2025). Operational calibration and performance improvement for hydrodynamic models in data-scarce coastal areas. EGUsphere, 2024:1-20, doi: 10.5194/egusphere-2024-1563

ABSTRACT. Water is an essential element for life and the demand on this vital resource continues to increase day after day due to continued demographic and industrial growth, in this context the design of the drinking water supply network has prompted great interest from researchers to provide solutions that allow water needs to be met at lower cost The objective of this study is to develop a program for optimizing the cost of pipes using a new metaheuristic method called Lion algorithm which has never been applied for the design of drinking water network. In this article the LION optimizer was applied on three examples of drinking water network references known in the literature which are: Two-Loop network, New York network and GoYang network

ABSTRACT. The water sector has benefited greatly from the advances in digital and computational technologies. Water distribution networks (WDNs) are increasingly equipped with sensors to measure parameters, facilitating real-time data collection, decision making, and protection, which is the purpose of the CoRREau project.

In this project, decision-making is based on hydraulic modelling and the information system (logs, real-time observation). This enables effective control and rapid response to incidents (leaks, deteriorating water quality, cyber-attacks, etc.). The digital model will be extended to become a digital twin, containing data in real time.

This conference paper presents the first step towards the conception of a digital twin of the WDN, which is the development of a reduced order model (ROM) of the system to describe the WDN. The ROM is emerging as an appropriate solution for reducing the significant time and cost constraints associated with the calculations required, especially for large-scale networks. First, we will review some existing ROMs such as WDN simplification or graph decomposition. We’ll focus on the algorithms we used to propose a reduced model by applying graph decomposition. This involves separating the forest part from the rest of the graph (the core), then reducing the core's size by retaining only supernodes and superlinks. This method reduces computation time while preserving information despite the network decomposition. The hydraulic solution obtained from our graph-decomposition ROM is validated by comparing it with the full-network solution obtained using hydraulic simulation tools (Porteau). We then present the new results observed using this method, applied to different WDN configurations by progressively increasing complexity, whether in terms of network size, the presence of multiple reservoirs or their connectivity. Finally, we will outline the future perspectives for the development of the digital twin, which is based on the reduced hydraulic system model proposed with data assimilation and incertitude quantification, for the project case studies using the real WDN of Strasbourg.

This research is part of the first author's doctoral thesis in applied mathematics.

ABSTRACT. Efficient and effective management of existing water distribution systems (WDSs) is increasingly challenged by aging infrastructure, population growth, urban expansion, climate change impacts, and environmental pollution. As a result, there is a pressing need for integrated solutions that assist decision-makers in planning potential interventions, while taking into account possible consequences and variations in both the mid- and long-term future. This study aims to develop a multi-attribute decision-making model to support the rehabilitation planning of Mohammedia 's drinking water distribution network, with a focus on reducing water losses while serving as a reference for network and loss management. A key aspect of this model is its capacity to simultaneously evaluate all factors contributing to network deterioration. The methodology employed is the Analytical Hierarchy Process (AHP) integrated with a Geographic Information System (GIS), which allows for a hierarchical evaluation of various criteria and sub-criteria to recommend actions that assist decision-makers.

ABSTRACT. As aquatic environmental regulations become increasingly stringent, advanced modeling tools are essential for understanding and managing water quality under dynamic conditions. This study presents a 1D water quality model that integrates hydrodynamics and biogeochemical processes capable of simulating both conventional indicators (i.e. temperature, carbonaceous biochemical oxygen demand, dissolved oxygen, nitrogen species, phosphorus species, phytoplanktons and coliforms) and emerging contaminants. The complete model is validated using an unsteady-flow case study in the Ebro River, where simulated results showed strong agreement with observed data, demonstrating the model’s accuracy and reliability. Furthermore, a sensitivity analysis is conducted to assess the model's responsiveness to extreme hydrological events and pollutant loads. A hypothetical flood scenario involving a sudden release of antibiotics from a hospital highlights the model’s ability to simulate the transport and fate of emerging contaminants in dynamic flow regimes. These results emphasize the model's potential for both traditional water quality assessment and the management of novel environmental challenges posed by persistent and toxic emerging substances. The proposed framework offers a valuable decision-support tool for water resource managers seeking to protect aquatic ecosystems under pollution pressures.

ABSTRACT. Flooding events represent one of the most impactful natural disasters, leading to significant human and economic losses. Accurate prediction of flood propagation is crucial for effective flood management, and numerical simulation models—particularly one-dimensional (1D) and two-dimensional (2D) models—play an essential role in this. However, these models are computationally intensive and require extensive calibration, especially for hydraulic parameters such as the Manning coefficient, which is highly sensitive to environmental conditions like channel shape and surface roughness. To address these challenges, this study explores the use of Support Vector Machines (SVMs) as a complementary tool to optimize the calibration and accuracy of hydraulic simulations. By training the SVM with data from both 1D and 2D simulations, the methodology aims to estimate the Manning coefficient more efficiently. This involves categorizing the simulation errors and using these categories as labels for training. The results show that SVMs can successfully identify a range of Manning coefficients that best match the 2D simulations, thus improving the accuracy of 1D flood models. This study highlights the potential of SVMs to streamline the calibration process, reduce computational demands, and enhance the versatility of traditional flood prediction models, making them more efficient and adaptable for real-time applications.

ABSTRACT. Thanks to the increasing availability of high-resolution topographic and hydrological data and the rapid advancement of high-performance computing techniques, various approaches have been developed for rainfall-runoff simulations in river catchments. Depending on the specific purpose of the model, different surface and subsurface hydrological processes are simulated with varying levels of detail, often relying on fundamentally different numerical techniques. In this framework, Integrated Surface–Subsurface Hydrological Models (ISSHMs) are becoming increasingly popular due to their capability to simultaneously reproduce multiple water cycle processes at spatial scales ranging from a few meters to a few kilometers by adopting physical conservation principles. On the other hand, rain-on-grid models, in which rainfall and infiltration are explicitly included in 2D shallow water models, are gaining attention for hydrodynamic-based rainfall-runoff simulations aimed at hazard mapping at the watershed scale. In this study, two intrinsically different modelling approaches are compared and discussed: 1. The first approach is a new ISSHM called HydroCAL, based on the Cellular Automata (CA) paradigm and implemented through the OpenCAL scientific software library. This model couples a 3D variably saturated subsurface flow module with a 2D diffusive-type surface routing module. 2. The second approach is a rain-on-grid model based on 2D fully dynamic shallow water equations, solved using a finite-volume method, first-order accurate in both space and time. The model is primarily developed for flash flood simulations and impact assessments. It simulates Hortonian overland flow, with infiltration losses computed using the Green-Ampt method, while simultaneously propagating the resulting surface flow. The two models, relying on different parallel computing architectures, were applied to simulate storm events recorded in an experimental watershed, the 7 km² wide Turbolo headwater catchment, located in Calabria (southern Italy). The results, showing reasonably good agreement between simulations and observations, are discussed by outlining the strengths and limitations of each approach, with particular emphasis on the potential benefits of integrating them.

ABSTRACT. Floods are among the most frequent and destructive natural disasters, posing significant risks to agriculture, industry and infrastructure. As extreme weather events become more common, the demand for accurate and timely flood forecasting has grown across various sectors. However, traditional forecasting methods often fall short in capturing the complexity of flood dynamics, necessitating the development of more advanced predictive approaches to mitigate risks and minimise damage.

A key challenge is managing uncertainties associated with flood forecasting, which can significantly impact forecast reliability. Therefore, uncertainty quantification is crucial for improving flood forecasting models, particularly in coupled hydrological-hydrodynamic systems, where multiple interacting variables influence outcomes. This study employs Bayesian Monte Carlo methods to systematically assess and quantify uncertainties in such models. Through probabilistic inference and parameter calibration, the study evaluates model uncertainty and sensitivity to various input factors. The results demonstrate that Bayesian-based uncertainty quantification enhances model robustness and improves flood prediction reliability.

Through better estimation of uncertainty propagation in coupled hydrological models, this research contributes to the development of more reliable flood forecasting systems. These insights can aid policymakers, engineers, and disaster management teams in making informed decisions, ultimately strengthening flood preparedness and resilience.

ABSTRACT. The evaluation of flood hazard in data-scarce regions is especially challenging due to the absence of the necessary observed data to calibrate and force any numerical model. This is especially true for lumped hydrological models, due to the high dependency of the model parameters on calibration. Physically-based distributed models are less dependent on the availability of data for calibration, and are therefore more adequate to estimate flood hazard when the availability of field data is low. However, the parametrization of these models requires a detailed definition of the catchment topography, as well as the spatial distribution of land uses and soil types. This information is currently available at different spatial resolutions through satellite products provided by spatial agencies that are freely available at the global scale, thus facilitating the application of physically-based distributed models.

We propose a workflow to evaluate flood hazard in data-scarce catchments that combines satellite data with an integrated high-resolution hydrological-hydraulic model. We evaluate the ability of this modelling approach to reproduce observed extreme flood events in two catchments located in Mozambique, in which the availability of field data is low. The events analyzed are the Tropical Storm Ana, which caused extensive flood damage in the Licungo basin (circa 23,000 km2) in January 2022, and the floods occurring in February 2023 in the Umbeluzi basin (circa 5,500 km2), which severely affected the city of Maputo and its neighborhood. In the Licungo case study we also estimate the population exposed to river flooding in the whole catchment, while in the Umbeluzi case study we analyze the potential effect that the management of a reservoir located upstream Maputo could have on the flood hazard. In both cases, the model results are compared to the available observed data related to the flood extent and the affected population during these storm events.

All the input data needed to apply the proposed methodology are retrieved from freely-available remote sensing data sources at the global scale. The integrated hydrological-hydraulic modelling is performed with the software Iber (www.iberaula.com), which includes a GPU-enhanced solver for the two-dimensional shallow water equations including rainfall and infiltration processes, and which is also freeware. Therefore, the workflow proposed here can be easily reproduced worldwide using freely available data and software.

The results obtained show that integrated hydrologic–hydraulic models based on two-dimensional shallow-water equations, combined with global satellite-derived products of rainfall, topography, land use, and curve number as input data, are currently able to reasonably reproduce the extent and peak discharge of extreme flood events in data-scarce basins of several thousand km2, and are therefore very useful tools for the development of flood management plans in almost any region of the world.

ABSTRACT. Effective rainfall-runoff modeling often demands high spatial resolution, particularly in regions prone to dynamic hydrological phenomena. Traditional methods of catchment analysis, including topographic mapping with contour lines and characteristic-based approaches, often fail to account for the temporal and spatial variability of hydrological and hydrodynamic processes. This study proposes an innovative approach for strategic mesh refinement in two-dimensional (2D) rainfall-runoff simulations. By defining fine-grid resolution on areas where hydrological and flow dynamics show critical importance, and conversely coarsening in areas of less importance, the method preserves modeling accuracy while significantly improving computational resource allocation. Applied to the small coastal watershed called Brague around Biot (located between Cannes and Nice cities, Alpes-Maritimes France), this approach demonstrates that refining the mesh in critical areas reduces computation times and resource demands, while maintaining high-quality simulation results consistent with historical observations. The simulations were carried out using the TELEMAC-2D code and past flood events as references. Finally, the computational savings achieved through this strategy leaves room for developing more efficient forecasting strategies or sensitivity analyses requiring many scenarios, as well as enhancing the model realism and consequently its overall complexity.

ABSTRACT. Flood risk management in urban areas is crucial to minimize disaster impacts, especially in developing countries. 2D hydrodynamic modelling has emerged as a robust tool for flood assessment, but it required large and accurate dataset, with high-resolution DEM (Digital Elevation Model) playing a major role in deriving the flood parameters. The study investigates how topographical resolution impacts 2D flood inundation mapping. A 5 km of Sabarmati river stretch, covering 9 km2 of floodplain is considered to build a 2D HEC-RAS based hydrodynamic model for study demonstration. High-resolution DEM’s (3.69 cm x 3.69 cm) were captured using Phantom 4RTK and processed with Pix4D. The model is simulated in unsteady flow condition, where upstream boundary considered as flood hydrograph and downstream boundary as normal depth. The suitable LULC roughness through LU map has been incorporated for simulation. The water surface elevation (WSE), depth and velocity maps have been derived. Furthermore, the same simulation is also compare with replacement of high resolution DEM with open source 30 m SRTM DEMs. The results shows the significant deference in WSE, depth and Velocity. This research underscores the need for carefully selection of DEM resolutions to model accuracy. The findings equip decision-makers with actionable insights for flood risk management by addressing critical sensitivities. This study aims to advance the understanding of hydrodynamic modelling, fostering improved flood mitigation strategies in developing countries.

ABSTRACT. The Berlin-Brandenburg region is one of Germany's warmest and driest regions. In recent years, Berlin has seen prolonged droughts, heat waves, but also intense rainfall events causing considerable damage. Climate projections indicate a further increase in such hydro-climatic extremes, threatening economic, ecological, and public health while challenging the city's long-term water security. To address these risks, Berlin envisions a transformation into a Sponge City, where blue and green infrastructure maintain the natural water balance, reducing flood risk, combat extreme heat, and enhance urban resilience. To support planning and decision-making for this transformation, modeling tools are used to identify vulnerable urban sites. For the Berlin area, the Hydroinformatics Modeling System (hms++) has provided high-resolution flood simulations using the full shallow water equations (SWE). hms++ is maintained by the Chair of Water Resources Management and Modeling of Hydrosystems of the Technische Universität Berlin and offers a fast, flexible, and user-friendly tool for modeling surface flow and associated processes. A key benefit of physically-based models like the SWE is their applicability to hypothetical scenarios, which is crucial for the simulation of adaptation measures within the Sponge City concept. Processes such as evapotranspiration and infiltration are central to the function of such measures, but operate on much larger time scales than rainfall-runoff dynamics. To capture the variety of scales and processes of the urban water cycle efficiently, hms++ has been extended with an adaptive switching mechanism, enabling it to represent a broad range of temporal resolutions, from short-term extreme precipitation to heatwaves lasting several weeks: During precipitation events, hms++ employs the full SWE to capture rainfall-runoff dynamics and obtain detailed results on flow paths, velocities and water depths, whereas in dry periods, the model complexity is conditionally reduced and the time step size increased. In the latter case, the model shifts from a 2D SWE model to a simplified 1D storage model, where volume reduction due to infiltration and evaporation is computed with hourly or larger time steps. When precipitation triggers a reversion to the SWE model, each remaining cell storage acts as an initial water depth. In comparison to pure hydrodynamic simulations, the novel switching approach exhibits significant speed advantages, with accuracy remaining largely unaffected. This extends hms++ applicability to hydrodynamic-hydrologic long-term simulations spanning weeks or even months, which is otherwise infeasible for explicit high-resolution SWE models. Future work aims to include parameters characterizing vegetation canopy and soil layers. This approach enables comprehensive modeling of extreme hydro-climatic events, with use cases ranging from the assessment of Sponge City infrastructure faced with heavy rainfall on very dry soils, to estimating the onset of water stress-induced reductions in transpiration.

ABSTRACT. Floods occur when excessive water overflows, transforming serene landscapes into forces of nature. These powerful events can have lasting effects on ecosystems and human communities. This study specifically investigates the impact of a significant flood that affected the Rel River, which originates in the Aravalli Hills in Rajasthan and flows into Gujarat, India. The river’s watershed covers around 442 km² and is located between 24°50′N and 24°75′N latitude and 72°00′E and 72°45′E longitude. In the years 2015 and 2017, the region experienced severe flooding, causing major damage and loss of life, highlighting its vulnerability to extreme weather events. Google Earth Engine (GEE) was used to classify and interpret parameters such as flood extent, precipitation, water quality (chlorophyll-a and turbidity), Normalized Difference Vegetation Index (NDVI), and soil properties (organic carbon and pH) were studied. Precipitation data was gathered along with population density information which was sourced from a shapefile from Humdata. These include NDVI, flow accumulation, elevation, and population density. From the acquired satellite data, it was established that the area under flood expanded from 44.11 km² before the flood to 47.20 km² after the flood, and precipitation had surged from 0 mm before the flood to 1768.53 mm during the flood. Precipitation after the flood was at 210.80 mm. Chlorophyll-a decreased from 0.000033 to 0.0000045 while turbidity increased from 0.92 to 1.02. There has also been change in groundwater analyzed, showing +0.01 in 2015, +0.067 in 2016, +0.119 in 2017 before the flood, and increased to +0.084 after the flood in 2017. Combining all these layers into a single overlay map, the Flood Risk Map delivers crucial information regarding flood vulnerability as well as levels of hazard. The Sustainable Development Goals 6 (Clean Water and Sanitation) and 11(Sustainable Cities and Communities) was focused in this research. Therefore, hydrological and ecological impacts which include partial recovery in vegetation with stable soil conditions. The results of the research point to sustainable land and water management practices reducing the impact of future floods with long-term recovery of the environment. Further follow-up monitoring needs to be continued to determine what is happening on the ground now.

ABSTRACT. The use of Graphics Processing Units (GPUs) has become increasingly common in scientific computing, particularly for improving the speed of flood modelling by solving shallow water equations. However, running GPU-based flood models on consumer-grade GPUs poses challenges, mainly because these devices have much lower double-precision computing power than high-end GPUs like the NVIDIA A100 or H100, which are typically found in supercomputers. When computations rely on single-precision floating-point arithmetic, errors can arise—especially when small amounts of rainfall are added to existing water levels. This can impact the accuracy of direct runoff simulations that use rainfall as a boundary condition. To address this issue, this study presents a new method for calculating rainfall source terms, designed to correct rounding errors caused by lower floating-point precision at each computational step. The accuracy of this approach is tested against theoretical benchmarks and real-world case studies, showing its potential to improve the reliability of GPU-based flood modelling on more widely available hardware.

ABSTRACT. In open channels, fast operations modifying the discharge like the sudden closing of gates, generate waves propagating in both upstream and downstream directions. This phenomenon, as it modifies the water level in the channel, is of primary interest for constructors and managers of hydraulic structures to correctly design the dike level or protocols for operating gates and turbines. As described by Violeau [1], the amplitude and wavelength of the water level signal depend on wave dispersion and energy dissipation. In a previous work (Cierco et al. [2]), the modeling of different disjunction test cases in a headrace channel was investigated, using various numerical and scale models. They used three numerical solvers of the 2D Basilisk code [3]: Saint-Venant equations for shallow water, Green-Naghdi equations which include dispersion, and the multilayer solver. The results were validated against Treske’s laboratory experiments [4]. The comparative study carried out with Treske's experiments showed the difficulty of the numerical solvers to capture the dynamics of the observed wave patterns when high "wave Froude numbers" occur. With the wave Froude number defined by Treske [4], it corresponds to highly non-linear and dispersive waves. This paper completes our previous work by studying the sensitivity of wave genesis to the different numerical parameters involved in Green-Naghdi equations, as implemented in the 2D Basilisk code. Specifically, we focus on the Manning coefficient and the turbulent viscosity, which account for dissipation, and the alpha-parameter, introduced in Green-Naghdi equations to modify dispersion. Note that two formulations of turbulent viscosity were studied: constant viscosity and mixing length formulation. Our results show that: - The alpha-parameter has little effect at low wave Froude numbers. However, it modifies the signal shape by changing the propagation rate of the sinusoidal components making up the signal, thus dispersing the wave packet. - The effect of dispersion is attenuated by dissipation, which smooths the surface and filters out secondary frequencies. - Two undular bore regimes, at low Froude number (characterized by large wavelengths) or high wave Froude number (shorter wavelengths) are simulated. - The gate opening time seems to influence the frequency distribution of the initial wave and could trigger a change in the propagation regime. The study shows that these parameters can be used to optimize simulation results and to better reproduce experimental measurements.

References : [1] D. Violeau, Contribution to the Theory of Undular Bores. A journey around the Korteweg-de Vries equation. in IAHR WATER MONOGRAPHS and Webinar IAHR. 2022. [2] F.-X. Cierco et al., « Modelling Undular Bores: A Comparative Study », in Advances in Hydroinformatics—SimHydro 2023, 2024. [3] « Basilisk ». http://basilisk.fr/Basilisk%20C [4] A. Treske, « Undular bores (favre-waves) in open channels - Experimental studies », J. Hydraul. Res., 1994.

ABSTRACT. In the present study, we propose a Lagragian eco‐hydraulic model for the numerical simulation of dynamics and growths of small pelagic fishes such as sardines in the Strait of Gibraltar. For the hydraulics, we solve an ocean model including Ekman and geostrophic currents, friction terms, bathymetric forces, Coriolis and wind stresses coupled to a set of stochastic differential equations for transport and dispersion of particles. The coupled approach combines both a fish dispersion model and a fish bioenergetics model. The bioenergetics model incorporates the sea-surface temperature and Chlorophyll-a concentration as key factors, which are essential for calculating the habitat index. In return, this habitat index is used to identify areas with the most favorable conditions for fish feeding, with a focus on sardines in the Strait of Gibraltar. In general, these models solve a set of stochastic differential equations for the transport and dispersion of particles. However, the accuracy of these solvers is low and it does not exceed first order in the time integration. In this study, we propose a class of second-order solvers for the transport and dispersion in the marine environment. The idea consists of implementing a family of stochastic Runge-Kutta schemes to handle the diffusion and drift terms within the Lagrangian particle models. Detailed algorithmic descriptions are also presented in the current work for all steps required for the implementation of the proposed methods including, the sampling and interpolation procedures. The performance of the proposed methods is tested for several test cases using different hydraulic regimes and dispersion scenarios in the Strait of Gibraltar. It is expected that the integration of these models would allow to simulate fish behavior by linking their habitat preferences, determined through the habitat index based on the sea-surface temperature and Chlorophyll-a concentration with their spatial distribution. A key advancement in the current work is the novel dispersion method that integrates physical hydraulic factors, bathymetry, and stochastic influences. This comprehensive approach also provides insights into how both biological processes and physical variability shape the fish growths and movement features in the Strait of Gibraltar. The obtained results would also enhance our understanding of habitat selection, fisheries management, and conservation, offering a more robust representation of fish population dynamics in the marine environment.

ABSTRACT. River flow modeling under different conditions is of paramount importance for river engineering. However, still today, no model can adequately capture different aspects of flow hydrodynamics over a variety of complex situations. River beds are often covered by gravels, with a significant spatial variability in terms of size and shape. The heterogeneous distribution of the grains strongly affects the flow velocity field and the turbulence characteristics, especially in the near-bed region, determining the formation of coherent turbulence structures which affect dispersion and mixing of substances and the sediment transport. The simulation of hydrodynamics in the gravel bed rivers is extremely important to analyze fluvial processes. However, modeling with a non-uniform gravel bed remains a challenging task due to the complex interactions between flow and highly irregular bed surface. The variability in grain size and shape introduces difficulties in accurately representing roughness parameters, requiring ad-vanced turbulence models and high-resolution grids to capture the intricate flow patterns. Moreover, inaccuracy in model predictions often arise from limitations in resolving near-bed turbulence and sediment transport processes. Thus, discrepancies between real conditions and the estimated flow velocity distribution and water surface elevation still persist. In the present work OpenFOAM and Ansys Fluent softwares are used to explore the flow structure in the near-bed region over a non-uniform gravel bed of a straight laboratory flume. Both codes are powerful tools for hydrodynamic modeling in rivers but, while OpenFOAM is open-source and allows users to the code’s customization (Kakati et al., 2022), Ansys Fluent is a commercial code with a user-friendly graphical interface, often used for simulating hydrodynamics in rivers (Shaheed et al., 2021). The performance of these codes is compared in terms of grid-independent solutions and complex flow with complicated boundary conditions. Data collected in the laboratory channel are used for models’ calibration. References: Kakati, R., Chembolu, V., & Dutta, S. (2022). Experimental and Numerical Investigation of Hybrid River Training Works using OpenFOAM. Water Resources Management, 36(8), 2847-2863. Shaheed, R., Yan, X., & Mohammadian, A. (2021). Review and comparison of numerical simulations of secondary flow in river confluences. Water, 13(14), 1917.

ABSTRACT. The primary objective of this research is to achieve the optimal design for an Autonomous Underwater Vehicle (AUV) by conducting a numerical study using Computational Fluid Dynamics (CFD) tools to analyze biomimetic forebodies inspired by various marine organisms.

First, a three-dimensional model of the AUV will be developed using SolidWorks. Then, a numerical simulation in the ANSYS Fluent environment will be undergone. Subsequently design modifications will be introduced by incorporating biomimetic forebody shapes derived from marine creatures, specifically the dolphin-inspired forebody, the shark-head shape, and hybrid configurations. A comparative numerical study will be performed under identical boundary conditions for each of the aforementioned designs across a range of five different velocities.

The analysis will focus on evaluating pressure fields, drag resistance, and drag coefficients for each model. The ultimate goal is to determine the optimal design that achieves the best hydrodynamic performance for the AUV.

ABSTRACT. Hydrokinetic technology offers several advantages, including predictable energy production, low-visibility installations, and minimal environmental impact. Currently, the renewable energy sector is largely driven by offshore wind power, but the transition to the commercial stage of hydrokinetic turbine technology could further enhance this dynamic. Beyond increasing the self-sufficiency of regions hosting offshore wind farms and contributing to the energy transition, the deployment of offshore renewable energy sources generates numerous benefits across different scales and sectors. The most immediate and visible advantages are socio-economic, particularly through the activities of companies operating in this field. Artelia is developing a range of services to support technology developers and stakeholders in the marine energy sector, in particular for hydrokinetic energy. This offer covers several areas, from resource assessment to the impact of the machines and their connection.

Based on studies conducted for various clients, this presentation will outline the tools and methods used to address the many challenges associated with industrializing the sector. Numerical modelling continues to play a crucial role in these studies, particularly in analysing hydraulic and sedimentary processes.

ABSTRACT. In archaeology, conventional methods for estimating discharges and reconstructing the hydraulic functioning of ancient aqueducts are based on local calculations that do not take into account the spatio-temporal flow dynamics, leading to relatively high uncertainties. We illustrate the inputs of one-dimensional hydraulic modelling of a Roman aqueduct (Arles, France) over its entire known extension, and for the four successive phases of operation identified. Phase 1: during the Augustan period, the Roman colony of Arelate is equipped with an aqueduct draining water from the southern Alpilles massif. Phase 2: a new branch from the northern Alpilles was connected to the first one through a convergence basin. Phase 3: at the beginning of the 2nd century, the branch that captured springs from the southern Alpilles was assigned to the Barbegal mills, with Arles supplied only by the northern branch. Phase 4: the mills and the southern branch that fed them were no longer in operation, and part of the water from the northern branch was diverted from the basin to feed the mill site, which was then re-occupied. The aqueduct was modelled using the free, open-source software Pamhyr2 and Mage solver. Computational geometries were generated from a rich topographical database, known cross-sectional templates, and assumptions about the shape and thickness of the carbonate concretions. Special strategies were implemented to represent hydraulic singularities, in particular the convergence basin. Simulations of the north and south branches show a significant reduction in the maximum permissible flow (water level up to the top of the waterproof coating) due to carbonate concretion. While simulated maximum flows for canals “as new” (smooth and with no reduction in cross-section) are 380 L/s for the north branch and 250 L/s for the south branch, flows in canals with their current carbonate concretions are divided by four, which would be insufficient to supply a city that in the 4th century was a place of residence for Emperors. The aqueduct therefore had to be maintained until a late stage of operation. Furthermore, the modelling shows that the raising of the invert downstream of the basin and of the water level in the basin, by around 50 cm, as observed by the archaeologists, was probably intended to improve the hydraulic operation of the structure, in particular to dissipate the energy of the fall of the northern branch into the basin. This modification suggests that the basin existed before the creation of the northern aqueduct, and was probably used simply as a regulating basin for the southern aqueduct. Without claiming to draw definitive conclusions, these kinds of results from numerical modelling bring useful food for thoughts in the necessarily multidisciplinary debates on the functioning and operation of ancient hydraulic structures.

ABSTRACT. The surface drainage system comprises a series of street inlets and other elements designed for stormwater collection. Its primary function is to ensure the rapid removal of stormwater generated on urban surfaces, such as roads, sidewalks, parks, and plazas, into the sewer network. The CLOGGING-INLET project aims to deepen our understanding of how stormwater is intercepted and the factors contributing to clogging in surface drainage systems. The goal is to optimize urban drainage management and enhance the resilience of infrastructure. This study focuses on investigating the effects on the hydraulic efficiency of a sewer grate located upstream of speed bumps, utilizing the commercial software FLOW-3D. Special emphasis is placed on comparing the performance of different configurations, specifically examining how speed bumps impact the hydraulic efficiency of the grated inlet.

ABSTRACT. The challenge of environmentally sustainable sediment management has received increasing attention in recent decades. Researchers and operators are striving to adapt engineering practices and develop innovative methods to manage reservoir sedimentation while maintaining the functionality of hydroelectric structures and preserving the ecological functions of the affected rivers. Sediment redistribution is an approach that attempts to adapt the morphological characteristics of reservoirs by various means, such as reducing the extent of low flow zones, where most sedimentation occurs. In this study, a representative 1D numerical model using HEC RAS 6.6 with steady flow simulation is used to investigate the effect of mechanical sediment redistribution on shear stresses and water levels in the Verbois reservoir, on the Rhône River downstream of Geneva. This facility receives sediment-laden water from the Arve River tributary, consisting mainly of suspended sediment with a small proportion of sand and fine gravel. Within the reservoir, which acts as a natural settling basin, granulometric sorting takes place. Fine sediment deposits cause a backwater curve in the centre of Geneva during floods with a risk of inundation, thus justifying active sediment management strategies. In the short term, this strategy allows sediment storage management without removing sediment from the river system. In the long term, redistribution can be used to adjust the morphological profile of the reservoir, reducing sediment trapping rates and minimising the impact of the dam on sediment continuity. Sediment redistribution is also a means of reducing water level rise for a given impoundment. In the case of the Verbois reservoir, it allows the water level to be lowered by almost 20 cm during tricentennial floods, while storing a similar amount of sediment (5 Mm3) and reducing the stable grain sizes in the low flow zones that are most favourable for significant storage. However, these morphological adaptations affect the critical grain sizes upstream in the case of a fluvial regime of the river. It is therefore possible to reduce the total amount of sediment deposited by targeting the minimum stable grain size upstream of the dam. However, this is at the expense of reducing stable grain sizes further upstream in the backwater curve, resulting in more pronounced coarse sediment deposition. Depending on the preferred sediment management strategy for a conventional or run-of-river reservoir, sediment redistribution can optimise operational efficiency while promoting the passage of fine sediments. This strategy can also be useful for restoring ecologically valuable riparian habitats and spawning grounds.

ABSTRACT. Mediterranean weather is strongly influenced by climate change with intense rainfall events potentially leading to flash floods, such as Alex storm in 2020 (Nice-France). Therefore, the monitoring of intense rainfall events characterized by important special variability is crucial for assessment and management of potential risks. This study aims to experiment the implementation of a pilote dense rain-gauge monitoring network in remote areas relying on satellite connection for real time monitoring and nowcasting. To proceed, 20 experimental rain-gauges developed by the LEAT laboratory were installed in strategic areas, notably on the left bank of the lower Var Valley and at the head of the Vesubie Basin (France). These regions are recognized as suffering from limited coverage of meteorological stations with respect to the high variability of rainfall spatial distribution. The installed rain-gauges are part of an experimental monitoring system. The strength of the deployed monitoring system relies on the embedded LORAWAN (Long Range Wide Area Network) technology, combined with satellite reception system to secure real time data transmission during rainfall events. This technology enables the installation of rain-gauges in remote areas where conventional measurement infrastructures cannot be deployed. The expected outcomes from this pilote installation includes an analysis of the reliability of these experimental rain-gauges in comparison with data collected from larger meteorological stations, allowing a protocol for optimal positioning to be established. Subsequently, the spatialization of rainfall for each studied catchment area was carried out, with the goal of improving the density and thus quality of the rainfall monitoring system. The deployed dens rain-gauges network allows to perform and compare different spatial interpolations methods (Thiessen polygon, IDW, etc). Ultimately, this project has improved the characterization of microclimates and expanded the region's measurement network.

ABSTRACT. The French Riviera region is frequently subject to severe weather associated with Mediterranean episodes. These episodes are often the cause of flash floods triggered by storms. Besides the safety issues, the resulting floods can have a significant impact on civil engineering structures. This study focuses on the water intake of the Vésubie canal. This engineering structure is of considerable importance, supplying fresh water to more than 350,000 inhabitants. Since Storm Alex in 2020, after each significant rainfall event, the canal becomes clogged with sediments, leading to a suspension of its activity.

This project aims to optimise the Vésubie water intake infrastructure by assessing its hydrodynamic behaviour and sediment transport characteristics during moderate-to-high flow events. To do so, a 2D hydraulic modelling of the water intake was carried out using HEC-RAS, Iber and TELEMAC-2D hydraulic codes. The software packages solve 2D Saint-Venant equation systems using different approaches. Also, they allow the implementation of different simplified approaches to encompass sediment transport modelling. Here, three scenarios representing different flow conditions ranging from 6 to 243 cubic metres per second were modelled. A set of different simplified hydro-sediment transport scenarios was established and tested with the different software.

The initial results of the hydraulic simulations reveal that the geometry of the canal, which causes a reduction in velocities, leads to the settling and retention of sediments in the channel. The results obtained from the sedimentary modules will be compared to evaluate the performance of the three calculation codes.

This study will be used for future research aimed at finding solutions to prevent the canal from silting up.

ABSTRACT. This project focuses on the study of the Var and Vésubie watersheds, located in the French Riviera. These catchments have typical Mediterranean climate and hydrological characteristics. Furthermore, they have recently been affected by extreme flash flood events. The aim of this study is to benchmark standard hydrological modeling approaches for Mediterranean catchments. The goal is to compare the ease of implementation, quality of results, and input data requirements of the different standard modeling approaches. A bibliographic analysis of these methods and their required parameters was conducted, leading to the decision to implement models using HBV model, HEC-HMS and MIKE SHE software. The simulations were conducted using two intense rainfall events: one from Storm Aline in 2023 and one from October 2024 event which was less intense. The calibration of the models is based on the hydrographs observed at the watershed outlets: the Napoléon III hydrometric station for the Var catchment and the Utelle station for the Vésubie basin. At the end of this study, a comprehensive synthesis is provided, outlining the key criteria for selecting hydrological models. A detailed assessment of the influence of crucial parameters for each model is provided to offer insights into their impact on model results. This work can be useful in creating a quick decision-making tool before selecting a particular modeling approach to deploy on a Mediterranean catchment.

ABSTRACT. Flash floods occur regularly in the Mediterranean region, such as the ones observed on the small coastal watersheds of the French Riviera. These floods are characterized by intense rainfall and, despite their relatively moderate specific flows, are comparable to the major flash flood events observed in Europe. These phenomena can have significant human, societal, and economic consequences, causing loss of life and considerable material damages. In that regard, hydraulic codes solving the 2D Saint-Venant equations, even though (i) computationally costly and (ii) simplifying some hydrological processes can potentially be applied at a whole catchment scale to support risk assessment. These equations describe free-surface flows by incorporating the principles of mass and momentum conservation. This project aims to develop and compare physically-based hydraulic models performance to represent Flash-flood events at a full catchment scale. Here, HEC-RAS, IBER, and TELEMAC-2D codes were tested for the Var and Vésubie watersheds. To implement the approach, the meteorological events associated with the Tempete Aline in October 2023 and the heavy rainfall event in October 2024 were analysed, with spatially distributed and hourly rainfall data. The methodology consists in applying similar physical parameters to all models: roughness, precipitation, resolution of the digital terrain model and infiltration rate. Two different solvers/methods are used for each code to assess their strengths and limitations: HEC-RAS with 2D with St-Venant and diffusive wave approximation; IBER with structured and unstructured mesh; TELEMAC-2D with finite volume and finite element methods. Model performance will be compared on the basis of the computed flow rates, water surface elevations and computational time. This analysis will allow us to i) test the capacity of each code to reproduce observed events, ii) define an optimal modeling strategy for each code and iii) establish a comparison between a variety of renown suites of free surface flow solvers.

ABSTRACT. The French Riviera has experienced in the past few years both drought (2022-2023) and flash-flood episodes (2020,2023). This situation underscores the importance of continuous monitoring. However, many watercourses in the region remain sparsely instrumented, which limit real-time assessment and response capabilities. The objective of the present study is the assessment of the implementation of a system for instantaneous river flow measurement, by installing a camera with direct transmission of embedded results. The aim, here, is to test the capabilities of the system to provide instantaneous flow rates in particularly challenging context of a torrential hydrosystem. Two distinct approaches are used to establish rating curves: one related to a modelling approach, and the second is based on a sequenced image retrieval The numerical modelling approach uses the HEC-RAS software to construct flow rating curves. This model is intended to be compared with an actual situation affected by unfavorable flow conditions and an intended sensitivity analysis of the flow rating curve is performed. The second approach uses video analysis obtained by installing a camera embedding a sequenced image resolution algorithm. A benchmark is performed between (i) the FUDAA-LSPIV software using the LSPIV method (Large Scale Particle Image Velocimetry) and (ii) the calculation method developed by the company Huaxon integrated into the camera set up. Results enhance the sensitivity of video analysis methods in relation to environmental factors such as light, rain and sediment transport. The quality of the results for video analysis and digital model will also depend on the accuracy of input data such as bathymetry. The results of the numerical models are used to compare the flow rates defined in the calibration curves.

ABSTRACT. Intense rainfall event forecasting and its spatialization are (i) key for hydrological numerical models and (ii) challenging for rainfall simulator design, as they require a sufficient level of automation to account for the spatial and temporal variability of rainfall intensity. This study explores the feasibility of using Météo-France's hourly rainfall forecasts from the Arome model at 24-, 36-, and 72-hour horizons in physical reduced-scale models placed within a rainfall simulator. The objective is to redesign and automate the nozzle control system of an existing rainfall simulator to test the capability of generating predicted storms, both at a specific location and on a small distributed scale. After a series of calibration tests for the automation system, two experimental applications were selected: (i) a full-scale 0.5 × 0.5-meter permeable pavement surface and (ii) a reduced-scale 3D-printed catchment (Vésubie, France), where 24 hours correspond to 10 minutes due to scale effects. The results compare the feasibility of implementing hourly rainfall forecasts in a rainfall simulator versus a numerical model. Lessons learned regarding the automation of rainfall weather forecasts are also presented, emphasizing the limitations of spatialization techniques.

ABSTRACT. Recent advances in high-resolution spatial data acquisition have significantly increased the computational demands of distributed hydrological models, particularly for large-scale basin simulations. Conventional concentration flow algorithms, such as the Muskingum-Cunge method, provide theoretical rigor; however, their inherently sequential computation patterns limit operational efficiency. This study introduces a novel GPU-accelerated distributed hydrological framework that integrates algorithmic enhancements with massive parallel computing. The proposed grid-based model incorporates two key innovations. First, the Grid Drop Concept Runoff Approach treats the net rainfall within each grid cell as a unified entity and tracks its movement along predefined flow paths—identified using the single-flow direction algorithm—to the watershed outlet. Second, a CPU-GPU Heterogeneous Parallelization strategy is implemented whereby GPU threads execute grid-level runoff generation and flow discharge calculations (fine-grained parallelism), while the CPU manages input/output operations and parameter initialization. Asynchronous data streaming is utilized to minimize CPU-GPU transfer latency. The framework is developed in C++/CUDA and leverages NVIDIA RTX 4060Ti GPUs to achieve massive parallelism. Validation experiments on the 2650 km² Tunxi Basin, China demonstrate significant improvements. The GPU-accelerated model achieves an over 40-fold speedup compared to a CPU-based implementation (Intel i5-12400F) at a 90-m resolution across 370,000 computational cells. Moreover, accuracy assessments reveal that the Nash-Sutcliffe efficiency (NSE) for the outlet flow during extreme rainfall events exceeds 0.85, peak flow prediction accuracy is enhanced by 3–5% relative to the traditional Muskingum-Cunge method, and the error in predicting peak occurrence time is reduced by 2–3 hours. This work bridges the gap between hydrological process fidelity and computational efficiency, enabling real-time flood forecasting at sub-hundred-meter scales and providing a transferable paradigm for accelerating other physically based environmental models. Future research will extend the models applicability to larger watersheds and finer-resolution terrains, and emphasis on integrating heterogeneous computing techniques and hybrid machine learning workflows.

ABSTRACT. Abstract for a short presentation (format B)

Accurate streamflow simulation is essential for water resource management and flood risk assessment. While process-oriented hydrological models have traditionally been used for streamflow simulations, their dependence on complex calibration and extensive input data limits their use in data-scarce or ungauged basins. Long short-term memory (LSTM) networks have recently emerged as an effective solution with a strong performance in temporal induction (TI), spatial induction (SI), and spatio-temporal induction (STI) tasks. However, the application of LSTMs to French watersheds remains under-explored. These watersheds are often small, with diverse climates and lack reliable long-term gauging stations in many regions. The present study has two main objectives: (1) evaluating the LSTM performance in TI, SI, and STI tasks across 308 French watersheds, and (2) examining whether adding training data from 501 European watersheds improves streamflow simulations in each task. Six experiments were conducted, testing each task with two training approaches: one, using only French (FR) basins (LSTM_TI_FR, LSTM_SI_FR and LSTM_STI_FR) and the other incorporating European (EU) basins (LSTM_TI_EU, LSTM_SI_EU and LSTM_STI_EU). The LSTM performance at simulating streamflow was analyzed for each experiment using the Kling-Gupta efficiency (KGE) and percent bias (PBIAS) metrics. Results show that the LSTM performs well for temporal induction, with a median KGE value of 0.78 for LSTM_TI_FR and a relatively low median PBIAS of -0.05. For the SI and STI tasks, although the median KGE values are satisfactory (0.68 for LSTM_SI_FR and 0.63 for LSTM_STI_FR), the model performance between watersheds is contrasted. While LSTMs are found to perform particularly well on large watersheds (area exceeding 1400 km²), the 25% of basins exhibiting a poor KGE (i.e., < 0.42) in SI and STI experiments have very low flows or only slight variations in flow. LSTM is found to be efficient for the very high flows (25% highest flows) simulations, with median KGE values ranging from 0.54 to 0.69 across the experiments trained with the French basins. However, LSTMs tend to underestimate the highest flows and overestimate low flows, with KGE values decreasing as lower flow quartiles are considered. The addition of European basins in the training phase did not improve the overall streamflow simulations in any task. This study confirms the effectiveness of LSTMs in temporal induction, validating their applicability for simulating future streamflow in French gauged basins, particularly for flood risk assessment studies. While spatial and spatio-temporal generalization tasks also show promise, further improvements are needed to fully validate their use for studying historical and future streamflow of ungauged basins.

ABSTRACT. In the context of the bathing quality of the Seine, for the Olympic Games and future generations, the City of Paris has equipped itself with a real-time system called Gestion Adaptative des Ouvrages (GAO) to assist in the management of its sanitation networks, in order to be able to move quickly from non-discharge management to so-called safety management when weather conditions require it. This system is essential to support the management of a complex network such as the Parisian network, where it is necessary to reconcile the issues of environmental protection and the protection of property and people. Within a consortium, setec hydratec was responsible for the creation of a hydrological and hydraulic model. This is a digital twin of the network and the essential functions of the automatic systems that control the 115 critical sites in the City of Paris network via more than 900 adjustment parameters. This hydraulic model, with 15,000 nodes and 2,000 catchments, covers a large part of the SIAAP zone, whose inflows converge on Paris. It is accompanied by external regulation files that simulate the network's automatic controllers, flood or no flood, as well as the main degraded modes. Setec hydratec was also responsible for the pre-optimisation of four different management scenarios that are systematically tested in the real-time system: Ecological and economical: for light rain, no discharge and minimum use of storage. Ecological: for moderate rainfall, no overflow in the bathing area. Level 1 and level 2 safety: reduction of damage for exceptional rainfall without overflowing in dry weather (S2) or rainfall lasting 1 month (S1). Every 15 minutes, an HMI summarising the latest weather forecast, the Seine levels, the state of the network and the availability of the organs (based on data received from the Paris GAASPAR supervision), generates the network input and status files, as well as the simulation parameterisation files. The calculations are carried out with the hydra calculation code, which solves the Barré de Saint Venant non-steady state equations implicitly, fully developed by setec hydratec. The HMI then retrieves the results and produces a short-term (6h) and medium-term (72h) summary for the different management scenarios. An optimal scenario is selected or adjusted by the operator according to environmental, economic and safety criteria, then applied instantly under supervision.

ABSTRACT. When railway drainage infrastructure fails, resulting flooding can cause delay and risk to life for passengers, and penalty costs for Network Rail (NR), but it can also cause other severe failures such as land slips, as seen in the recent serious derailment event in North-East Scotland. Current hydrological models operate at large scale to predict broad areas of flooding, and are not developed for accurate modelling of flow paths at scales relevant to individual railway assets. Dig- ital terrain models are typically limited in resolution and can omit smaller scale drainage features such as roadside ditches that can substantially alter a small catchment. Models also suffer from deep uncertainty in factors such as infiltration rate, storage capacities and flow routing. This project attempts to overcome these issues using a data-driven approach combined with analyt- ical formalisms to address the uncertainty in water transport processes from rainfall to railway, enabling more accurate predictions of arrival flow volumes at the individual railway asset level. The underlying novelty of this research arises from the study of small catchments and their at- tuned sensitivity to localised parameters. Investigations of the predictive qualities of transfer functions when trained on data providing the local geology of specific small catchments have taken place. There is clear need to move away from simplistic triangular transfer functions, and to study physically appropriate transport models such as the advection-diffusion-reaction (ADR) model, deriving more complex transfer functions which contain the required physical parameters needed to model a comprehensive flood risk assessment. Specifically, we applied the Euler-Lagrange (EL) equation to the ADR model and derived analytical solutions which give upper bound asymptotes for parameters of the surrounding catchment studied. This led to results which robustly predict whether flooding of the surrounding area is likely or not; requiring much less computing power than a standard data driven approach. The methodology of this approach is analytical; first a transport model which contains all the necessary physical parameters of the system is needed. Second is the the need to understand and then formulate the constraints which will be imposed upon the system. Third is to apply func- tional analysis techniques to produce a system of differential equations to be solved analytically. The fourth and final step is to take the solutions obtained from the system of differential equa- tions and input the numerical data obtained from the systems surrounding geology, allowing for analytical minimum and maximum results, as opposed to numerical approximations. The combination of an analytical transfer model and numerical data profiles results in a power- ful predictive tool for assessing asset degradation, cost efficient repairs, and flood risks along all railway lines where there is geological data present.

ABSTRACT. Digital Twins are widely use nowadays in some water domains like the urban water sector. Thanks to these tools, water utilities monitor and optimize processes in drinking water and wastewater management facilities. However, these tools are less developed at a watershed scale. They have been successfully applied as early warning systems to mitigate flood damages, but they are less commonly used for an integrated management of the water resources.

Some obstacles for digital twin development at a watershed scale are the integration of a wide variety of data types in a data-centered platform, the identification of the water-related issues that can benefit from the digital twin, the accurate definition of the models that represent the water system processes and the selection of the most appropriate indicators to help improving the management of the key water uses.

Different technologies are used to develop digital twins and the combination of them allows a variety of digital solutions to be available in the market. Xylem Vue Early Warning System (Xylem Vue EWS) is a web-based digital solution designed to manage different types of data and real-time information. It generates and disseminates warnings to different stakeholders when a critical scenario occurs or is predicted. Successfully applied at several river basins in Spain for flood prevention, the tool is currently evolving to be used for other water-related issues.

Within the watershed context, one of the main difficulties of these digital solutions is to be flexible enough to adapt to the particularities of each site. Digitalization of watersheds can be categorized in several levels depending on the data availability and the maturity of the models used to represent the configuration of the water system. The structure of the water management competences within the basin plays also an important role in the tools’ development.

The "Sustainable Water Storage and Distribution in The Mediterranean" (OurMED) project aims to design and explore innovative and sustainable storage and distribution systems tightly integrated into ecosystem management at the river basin scale. This is achieved by the combination of scientific and local knowledge, emerging from new and long-lasting spaces for social learning among interdependent stakeholders, society actors and scientific researchers in eight local and one regional mediterranean demo sites. The project promotes the integration of innovative digital tools, remote sensing technologies, and in-situ monitoring approaches.

The aim of this paper is to present the methodology defined within the OurMED project to develop digital twins at a watershed scale in demo sites with different characteristics, to describe how Xylem Vue EWS digital solution is being adapted to meet each demo site needs and to analyze the level of digitalization that will be possible to achieved at each demo site.

ABSTRACT. Precipitation is a key hydro-meteorological variable essential for the protection and sustainability of water resources, yet it has been significantly impacted by climate change. Assessing these impacts is crucial for effective water resource planning and management. This study evaluated the effects of climate change on monthly precipitation in the Lower Mekong Delta River Basin using advanced trend analysis methods. Following homogeneity tests, the Mann-Kendall (MK) test and Innovative Polygon Trend Analysis (IPTA) were applied to homogeneous station data. The IPTA method provided insights into monthly precipitation transitions and seasonal variations. Findings revealed that IPTA was more sensitive than the MK test in detecting trends. These results can support future research and policymaking aimed at safeguarding water resources amid climate change.

ABSTRACT. The rise of climate related disasters has led to unprecedented challenges for water operators across the world. A notable case is the Vésubie canal in the Alpes-Maritime department of France, which is used for water production in Nice Metropolis. Following the events of storm Alex in 2020, frequent sediment transport in the Vésubie river obtrudes the canal intake and worsens the already damaged fish ladder infrastructure. This study aims to assess the design fish ladder renovation possibilities according to a 2D hydraulic model of the Vésubie river. The proposed design should enable the infrastructure to meet legislation standards while accounting for frequent sediment transport and base flows in the river. To achieve this, the main fish species in the river are first identified, and a literature review conducted to understand the characteristics of fish ladders and the parameters needed to design them. Three hydraulic modelling software solving the shallow water equations system are then selected to model the Vésubie river and obtain stage-discharge relationships at the weir. HEC-RAS, TELEMAC-2D, and IBER were tested mainly for their free access, to benchmark their capabilities of handling the weir representation in the models. The results obtained are compared with literature values, and one model is selected to produce different scenarios with various fish ladders.Finally, several scenarios with different types and sizes of fish ladders were prepared in order to determine the optimal layout for the water intake of the Vésubie canal. At the end a scenario will be selected using a multicriteria analysis and the fishway from this scenario is modelled in 3D.

ABSTRACT. Water scarcity is a global challenge intensified by climate change, population growth, and unsustainable practices. Addressing this crisis requires data-driven approaches for sustainable water management. Hydroinformatics, an interdisciplinary field integrating computational tools, artificial intelligence (AI), and data analytics, plays a key role in mitigating water scarcity (Rauch et al., 2019; Dunea et al., 2023). It enhances predictive capabilities, optimizes resource allocation, and supports decision-making, particularly through Decision-Support Systems (DSS). DSS translate complex hydrological data into actionable insights, aiding policymakers in adaptive water management (Zeman et al., 2021; Xu et al., 2022). Additionally, Nature-based Solutions (NbS) offer sustainable, climate-resilient strategies to address urban water challenges. These approaches leverage natural processes such as wetland restoration, watershed management, and green infrastructure to improve water retention, quality, and ecosystem services (Döll et al., 2012).

Nature-based Solutions (NbS) have gained recognition for addressing urban water challenges through sustainable, climate-resilient strategies. NbS leverage natural processes such as wetland restoration, watershed management, and sustainable urban drainage systems to enhance water retention, improve quality, and support ecosystem services (Döll et al., 2012; Cohen-Shacham et al., 2016). Their integration in urban water management aligns with broader climate adaptation efforts, offering co-benefits like biodiversity conservation and flood mitigation (Madani, 2019). However, despite their potential, NbS integration in hydroinformatics-based DSS remains limited. Existing models such as SWAT, WEAP, and MIKE HYDRO Basin require further refinement to quantify the hydrological and ecological performance of NbS, assess their effectiveness under varying climatic conditions, and optimize water distribution and long-term sustainability planning.

This study conducts a systematic review to address these challenges, focusing on three key aspects critical for advancing hydroinformatics in water scarcity management: (1) hybrid AI and physics-based modeling for more accurate and adaptive hydrological forecasting, (2) real-time hydroinformatics frameworks to improve decision-making with dynamic data inputs, and (3) integrating NbS into DSS to enhance sustainability in urban water management (Abessolo et al., 2017; Chang et al., 2022). The review examines the role of machine learning in predictive analytics, geospatial modeling, and IoT-based sensor networks in improving forecasting accuracy, resource allocation, and adaptive management (Wagner et al., 2015; Abegeja, 2024). For instance, AI techniques have demonstrated effectiveness in predicting hydrological extremes, while IoT-based sensors enable real-time data collection for more responsive water management (Singh et al., 2021; Dunea et al., 2023).

By synthesizing existing research and identifying key gaps, this study highlights the need for innovative, data-driven approaches to advance hydroinformatics for water scarcity management. The findings provide insights for integrating AI-driven hydroinformatics with NbS, fostering more resilient and adaptive water management strategies. This research contributes to the advancement of smart water management, guiding policymaking, and promoting interdisciplinary collaboration to address the global water crisis.

ABSTRACT. In recent decades, education has been a central concern in Europe, playing a crucial role in fostering better and healthier societies. The financing of educational programs and projects has been an essential tool for addressing a wide range of challenges affecting the continent. In this context, educational institutions play a vital role through traditional academic programs, master's courses, and research initiatives. Given the international nature of universities and research centres, collaboration has been key to enhancing local programs and expanding educational opportunities beyond institutional and national borders.

ABSTRACT. With the evolution of the digital transformation and Internet Of Things (IOT), digital twins have evolved and received more attention as a way of tracking the desired activities virtually, and without the need for physical presence. At the same time it supports in costs' reduction and improving operations’ efficiency. One of the vital components of the Digital Twins (DT) is the Decision Support Systems (DSS). DSS aims to synthesize the available data in order to provide valuable and actionable insights as well as highlighting the critical deficiencies in the facility processes. Aquaculture farming on the other hand faces many challenges in regards to climate changes, environmental and boundary impacts, controlling water quality, fish health & fish feeding monitoring as well as keeping the ecosystem sustainable. Thus, integrating the DSS with Digital Twins has the potential of enhancing the operational productivity and minimizing expenditures by automating and optimizing the core aquatic activities and processes. Based on this, we seek in this paper to present a state-of-the-art review of existing studies on DSS in aquaculture, and examine the current state and trends in integrating the DSS with digital twins in aquatic farming. Key challenges and open questions in this promising field are also highlighted with the aim of providing a potential area for future research and innovation.

ABSTRACT. Water scarcity, driven by increasing demand and deteriorating quality, poses a significant challenge in the 21st century, particularly in Europe where approximately 20% of the territory and 30% of the population are affected by water stress. To address this issue, the MARCLAIMED project proposes the integration of Managed Aquifer Recharge (MAR) with Alternative Water Resources (AWR) into River Basin and Drought Management Plans. Central to this initiative is the development of an Integrated Decision Support Tool (IDST) designed to enhance operational efficiency, economic sustainability, and social acceptance of AWR systems. The IDST will incorporate six AI-based tools that monitor and control AWR quality, forecast water resource availability, and provide health, environmental, and performance risk indicators. Additionally, it will facilitate the assessment of water scarcity through a municipal-scale water footprint tool, evaluate the economic value of MAR with AWR via an insurance calculator, and implement a cost recovery-based system. A key feature of the IDST is the orchestration interface, which extends the NGSI-LD standard to enhance interoperability among the various tools. This interface is designed to manage the diverse inputs, data, and results associated with each of the six tools, which differ significantly in their functionalities and data requirements. The paper will focus on presenting each tool, detailing their purpose and features, and the orchestration interface, showing how it enables seamless interaction between the tools and streamlines data handling. By providing a generic framework for tool integration, the orchestration interface aims to enhance the overall functionality and adaptability of the IDST, ultimately contributing to more resilient and sustainable water management practices.

ABSTRACT. The Lagoon Observatory Network (ROL) represents an innovative approach to environmental monitoring, addressing the complex challenges of managing lagoon ecosystems in the face of anthropogenic and climate-related pressures. This paper presents the architecture and applications of ROL, which aggregates heterogeneous data streams from sensors, utility systems, participatory crowdsensing and satellite observations into a centralized data core built on the ETSI NGSI-LD standard. By adopting this semantic data model, the ROL ensures interoperability, real-time data exchange, and compliance with global standards, enabling effective data integration and decision support. A cornerstone of the ROL is its data normalization framework, which harmonizes diverse datasets using custom-built data models. These models facilitate seamless mapping to the DCAT standard, allowing curated datasets to be published in open-access data catalogs. This approach not only ensures data usability across multiple domains but also promotes transparency and collaboration among stakeholders, including environmental agencies, municipalities, and researchers. Two flagship applications demonstrate the potential of ROL in addressing critical environmental and public health challenges. Vigithau Risques Microbiologiques functions as an early warning system for microbiological pollution in lagoons, leveraging sensor networks and predictive algorithms to monitor water quality in near real-time. This system provides timely alerts to mitigate health risks for local populations and ecosystem disruptions. Similarly, Vigithau Inondation serves as a flood early warning system for urban areas, utilizing hydrological models, satellite data, and sensor-based measurements to predict and communicate flood risks, helping cities enhance resilience to extreme weather events. Emerging applications under development include predictive tools for lagoon eutrophication and anoxia, which are critical phenomena affecting biodiversity and ecosystem health. These tools integrate machine learning models with high-resolution environmental datasets to forecast potential risks, offering valuable insights for sustainable lagoon management and restoration efforts. By showcasing the ROL’s technical architecture and its practical applications, this paper highlights the importance of integrating interoperable standards like NGSI-LD and DCAT to enable robust data-driven solutions for environmental monitoring. Furthermore, the modular and scalable design of the ROL makes it a transferable model for other coastal and freshwater ecosystems worldwide. The paper concludes with a discussion on lessons learned during the implementation of ROL, including challenges in data interoperability, stakeholder engagement, and the operationalization of early warning systems. Future directions include expanding the observatory network to integrate citizen science initiatives and exploring synergies with international environmental monitoring frameworks. Through its innovative approach, the ROL aims to set a benchmark for the sustainable management of fragile lagoon ecosystems in a rapidly changing world.

ABSTRACT. Local harbour authorities managing assets, vulnerable to water quality, require turbidity monitoring data, customised to respond to their own operational needs. Such monitoring is usually based on in-situ moorings associated with probes aiming at measuring turbidity and other physical parameters. The location of the moorings and the sampling frequency are based on expert knowledge. These sensors can inform in real time the operators about the level of turbidity. Depending on the measured level of turbidity actions can be taken to ensure that the level does not reach a threshold that can be damaging for marine fauna and flora. To improve the implementation of the monitoring survey, needs for extra information can be considered such as assessing the acceptable level of turbidity, defining sensor locations, specifying the sampling frequency,… In Lorient bay (France), Region Bretagne manages port facilities and is responsible to ensure good navigational conditions. Therefore, dredging operations are performed on a regular basis and for the period lasting between 2018 and 2027 about 100 000 m3 of sediments are annually extracted to ensure a secure access to the harbour facilities by vessels whose draught can exceed 10 m. Sediment compliant with the regulation are dumped offshore in an 2.2 km2 area where water depth is about 30 m. Harbour managers have deployed an extensive monitoring system to characterize the sediment distribution at specific sites in the bay. To spatially expand these punctual datasets, high-resolution satellite imageries are processed to estimate statistics of turbidity within the entire bay at a spatial resolution of 10 m and statistics at different vulnerable sites are computed. Historical satellite imageries from 2016 are also analysed to investigate the presence of turbid plume in the wake of dredging barges. After characterizing of the initial state for turbidity, a framework aiming at forecasting turbid plume dispersion induced by dumping operations is developed. The framework follows the previous developments made to forecast nearshore waves and high levels in estuaries. It is based on a supply chain downloading forecasted tides and wind from different institutional providers, running and post treating numerical simulations to test different operation scenarios. The outputs daily inform about the level of turbidity at different locations vulnerable to turbidity and compare these levels to the statistical level inferred during the characterisation of the initial state. The operational framework has been deployed for one year and interest in that procedure for offshore sediment disposal has attracted attention although regulation does not oblige managers to subscribe to such a tool.

ABSTRACT. The ongoing digital transformation is reshaping environmental management through innovative data-sharing models. This paper introduces the SEDIMARK project, which establishes a robust data marketplace framework grounded in the European Data Space principles. By fostering interoperability, data sovereignty, and secure exchange, the SEDIMARK approach aligns with the European Union’s vision of sectoral and cross-sectoral data spaces to unlock value from shared information. Our application focuses on water management, an area of growing significance due to climate change, urbanization, and increasing demands on water resources. We detail the deployment of this data marketplace framework within the territory of the CARF (Communauté d’Agglomération de la Riviera Française). This region, encompassing a complex interplay of urban, coastal, and natural environments, offers a compelling testbed for digital innovations in sustainable water management. Key use cases include real-time monitoring and predictive analytics of: 1. River levels, to contribute to flood risk management and ecosystem health. 2. Pluviometry, to refine hydrological models and optimize water resource allocation. 3. Vessel traffic on the seashore, linking maritime activities with environmental impacts. By leveraging advanced technologies and open standards, the project integrates diverse datasets from sensors, governmental records, and private stakeholders. Predictive algorithms analyze this data to anticipate water quantity trends and detect potential pollution events, enabling timely interventions. The marketplace facilitates seamless collaboration among municipalities, industry players, and researchers, promoting a unified approach to water stewardship while ensuring data ownership and privacy compliance. The implementation within CARF demonstrates the scalability and adaptability of SEDIMARK’s principles. Early results underscore the potential for enhanced decision-making, cost savings, and resilience in water management. Furthermore, the project contributes to the broader goals of the EU’s Green Deal by advancing sustainable practices and reducing environmental risks. This paper will discuss the technical underpinnings of the SEDIMARK data marketplace, its alignment with European Data Space policies, and lessons learned from the CARF deployment. Particular attention will be given to the challenges and opportunities in integrating predictive algorithms with multi-domain data in a federated ecosystem. We will also outline future directions for expanding this framework to other territories and domains, emphasizing its potential as a blueprint for sustainable digital transformation in environmental management.

ABSTRACT. This paper introduces an automated pipeline for real-time flood forecasting designed to enhance decision-making and optimize flood risk management. The system integrates data ingestion, hydrological and hydrodynamic modeling, post-processing visualization tools into a seamless workflow, leveraging open-source and/or proprietary technologies to ensure operational reliability and real-time processing capabilities. By providing accurate and timely flood forecasts, this solution addresses the specific challenges encountered by flood-prone regions, delivering actionable insights to public authorities and stakeholders. The modeling aspect of the system represents a key advantage, featuring flexibility to incorporate various hydrodynamic modeling tools tailored to regional complexities and local authority requirements. It supports industry-standard software to provide a versatile, software agnostic solution, facilitating accurate simulations of a range of hydrodynamic scenarios. This adaptability ensures the pipeline can cater to different levels of detail and accuracy, effectively meeting decision-makers' distinct needs while retaining operational efficiency. Two projects, covering flood prone areas subjects to extreme hydrological events, are showcased as pertinent cases study for system implementation. By seamlessly integrating data with advanced modeling techniques, this system equips decision-makers with essential resources to significantly mitigate flood risks. Its modular architecture and open-source foundation enhance versatility, ensuring applicability in global regions facing similar flood challenges. The system's emphasis on real-time processing, operational reliability, and user-centric visualizations and mapping tools establishes it as an essential asset in modern flood risk management practices. The abstract covers the main issues one can encounter when setting up real-time modelling pipelines: data management, needs for reliable and automated processing tools, need for accurate but in the meantime fast hydrological models providing timely forecast, regularly updated based on updated observed and forecasted rainfall. It details the data used by the modelling chain, with an emphasis on high-resolution rainfall data, lidar and soil cover data. The hydrological and hydraulic models, both based on 2D distributed approaches, are described, with a focus on the challenges and advantages linked to these methodological choices. The abstract covers use cases for the projects: flooded area forecasting, based on the most up to date rainfall forecasts, with different levels of accuracy and lead time: 1 / immediate pre-computed scenarios selection based on rainfall forecast or measured river flow, 2/ pre-computed scenario selection based on forecasted flows produced by the hydrological model and 3/ forecasted flooded area based on the full hydrological/hydraulic modelling. The paper concludes with the enhanced capabilities of the system, in terms of risk anticipation, with the possibility to identify areas exposed to runoff based on the observed and forecasted rainfall, and regarding operational modelling, with the capability to integrate “what-if” scenarios (hydraulic structures failures and/operations), and automatic flood impact assessment (roads and public equipment, number of buildings and/or population exposed).

ABSTRACT. In response to the critical concern of water quality on university campuses, HydroPace undertook the challenge of developing a real-time water quality awareness system. This initiative was motivated by the recognition of the substantial reliance on water coolers within university communities, underscoring the need for immediate information about drinking water quality. To address this, HydroPace seamlessly integrated physical and digital elements, providing instantaneous water quality information. Drawing inspiration from literature, such as the H2O Smart Drinking Water Quality Monitoring System and Real-Time Monitoring of Urban Water Systems, HydroPace synthesised these influences into an innovative solution. The research methodology employed generative research, blending quantitative surveys and qualitative user interviews, followed by usability testing and a comprehensive field study at Pace University. The findings encompassed generative study insights into user preferences and concerns, usability test revelations highlighting positive feedback and identified pain points, and field test study results, including commendable System Usability Scale scores. These findings collectively guide the refinement of the water cooler prototype and associated app, ensuring optimal user satisfaction and engagement.

ABSTRACT. Most sewage systems in larger Swiss cities are combined sewer systems. During rainfall events, storm overflow discharges untreated mixed water into the receiving environment (Jordan et al., 2014), contributing to both local and regional water pollution. In large urban centers, converting the drainage system to a separate sewer system is often not economically viable. One solution for reducing mixed water discharges is to provide retention basins to temporarily hold back peak flows and then return the water to the treatment plant (WWTP) for treatment, when the latter is no longer saturated. The retention of mixed water is useful, but it requires very large retention volumes, which often do not exist in these networks. Smart management of these facilities can optimize their storage capacity and the total volume of water treated in the catchment area. Intelligent control must also take into account the synchronization between retention basins, as well as treatment capacities at the WWTP (reactive centralized dynamic management). Finally, intelligent control can be based on flow forecasts, a few tens of minutes in advance, to better regulate retention basins by taking into account the future precipitation and the transit times from one part of the network to another (predictive centralized dynamic management) (Décosterd et al., 2022), (Clément et al., 2024). A solution of this type has already been implemented for the wastewater network of the Greater Zug area in Switzerland (GVRZ), which serves a population equivalent of 175,000. This complex network comprises 23 pumping stations, 11 stormwater overflow basins, 1 WWTP, and 4 storage volumes in large-diameter pipes, including an upstream retention basin with a storage capacity of 3,500 m³. In a pilot phase project, this upstream storage capacity has been optimized to regulate the outflow from the basin, maximizing retention volume and minimizing overflows. Radar precipitation forecasts are used as input for the Routing System rainfall-flow simulation model (Hydrique Ingénieurs, 2024). These inflow forecasts are updated every 10 minutes, and an optimization algorithm determines the downstream flow setpoint for the retention basin. To predict optimal downstream flow, the model integrates meteorological observations and real-time measurements from network infrastructure (water levels). The algorithm accounts for inflow forecasts to the retention basins as well as the forecasted remaining treatment capacity at the WWTP. Specifically, the algorithm ensures better utilization of upstream retention volume by preventing too early or too late retention. In this way, the digital twin developed complements the real-time measurement system used to monitor the network, forecasts flow in the facilities and at the WWTP several hours in advance, and optimizes the retention of mixed water in existing facilities equipped with mobile gates. This digital twin is a fundamental tool for network operation and reduction of wastewater-related pollution.