ABSTRACT. The Nice Côte d'Azur Métropole faces significant challenges in water management, particularly with the effects of climate change. Extreme phenomena as prolonged droughts or intense rainfall, have already affected the availability and quality of water resources. The need to predict accurately future fluctuations and trends in water resources becomes imperative to better anticipate and respond to the needs of the region's inhabitants. With that aim in mind, we explored the contribution and limitations of running models using machine-learning techniques. A partnership with Laboratoire J. A. Dieudonné (Université Nice Côte d'Azur) brought expertise with time series data while Eau d’Azur provided hydrological and hydrogeological knowledge and data. The present study focuses on the groundwater level in the lower Var valley where Eau d’Azur pumps out drinking water from multiple wells located in the alluvial aquifer. To forecast groundwater level, the objectives were 1) to explore available data representing water level and associated variables (rainfall, Eau d’Azur withdrawal, groundwater flow) and 2) to model the groundwater level with time series techniques for 3 to 10 days. We considered four types of data: Hydrological, meteorological, withdrawal volumes and groundwater flow. After cleaning and pre-processing, these data became inputs in three models: SARIMA, SARIMAX and a model inspired by Kalman filtering. Kalman filter, including all four types of data, yielded the best results considering the mean square errors. However, we further explored whether all the input data is necessary and whether other data could be included in the model. We conducted tests to highlight most consistent variables. It seems that the groundwater flow did not contribute significantly to the groundwater level forecast contrary to Eau d’Azur’s withdrawal and most of weather stations. Moreover, the Var flow was not taken into account while this variable is more directly correlated to the groundwater level than rainfalls. Another drawback of this study is that groundwater level forecasting relies on MétéoFrance precipitations forecasts, which provide a 3 to 10-day forecast. As this is a timeframe too short for Eau d'Azur to take any action, the next step will be to use advanced classification methods to achieve a multi-month forecast.

ABSTRACT. Cryptosporidium is a pathogen which can cause severe and persistent gastroenteritis. Often present in surface waters in its infectious oocyst form, Cryptosporidium poses a considerable challenge to water authorities because of its tolerance to chlorine disinfection and its infectivity at low concentrations, with the potential to cause major outbreaks should oocysts break through into treated drinking water. The combined application of AI and remote sensing data has the potential to equip water providers with powerful new predictive tools, which could give early notice of periods of elevated pathogen levels in rivers and thereby aid risk management.

The objective of the present study was to develop a machine learning-based tool capable of forecasting Cryptosporidium concentrations in a river, and explore how outputs can be used to assist operational decision-making as part of a risk-based water abstraction strategy. The study makes use of an extensive Cryptosporidium dataset (on a scale not previously seen in the literature), consisting of 7+ years of approximately daily measurements made at a major UK abstraction site. A bagging-XGBoost model was trained and validated to forecast Cryptosporidium river levels using two primary features – antecedent Cryptosporidium concentrations and time-of-year – alongside meteorological and hydrological inputs, including daily rainfall, soil moisture, soil temperature, river discharge and river discharge rate-of-change. Event Duration Monitoring (EDM) data of Combined Sewer Overflow (CSO) spills was also trialled as an input.

Model variants were divided between those incorporating aggregated-only (i.e. catchment-averaged) environmental data and those with additional spatially distributed data. Travel time-based time lags were applied to capture the impact of sources mobilised at different upstream distances in the catchment and an unparsimonious approach to feature selection was adopted, with a view to providing a simple method that could be readily deployed at other sites without the need for extensive prior data analysis.

Model agnostic feature analysis, using SHapley Additive exPlanations (SHAP), was applied to interpret the model and draw inferences regarding likely sources, source areas and controlling factors in Cryptosporidium mobilisation and transport. Results showed that while the inclusion of spatially distributed data could produce modest improvements in the model, a simpler model (based on aggregated catchment data) was almost as effective as the more complicated, data-hungry models, and could therefore prove preferable in an operational setting.

Preliminary trials of the model in an operational capacity were impacted by data availability issues, highlighting the need to include redundancy, primarily in the form of backup datasets which can be drawn upon when remote sensing sources experience downtime. Different output formats were also trialled by providing operators with (for example) precise concentration predictions, or with simple red-amber-green risk classifications. Operator feedback was collected to determine which approaches were most useful for abstraction decision-making.

ABSTRACT. The volume of environmental time series data has grown significantly in recent decades, driven by advancements in sensors, IoT, and open data initiatives. This has led to big pools of data, containing valuable yet often hidden insights. However, without proper quality control (QC) methods, much of these data remain underutilized or they may lead to inaccurate conclusions, and decision-making on improper grounds (such as maladaptation). As manual checking of time series is no longer feasible, automated statistical QC tools have been developed, typically relying on threshold-based statistical tests to identify erroneous data.

Defining suitable thresholds for each parameter and station is case- and user-dependent while integrating machine learning (ML) into QC processes can accommodate such challenges. ML algorithms are well-known for their pattern recognition abilities, allowing the detection of data points that deviate from common time series trends to improve accuracy in quality checks. Given the importance of geoscience time series in hydraulic research, incorporating ML to existing statistical tools is essential, both for enhancing data quality and to support accessibility of quality-checked time series as baseline for further research and AI- ready datasets.

The QC algorithm is developed using sea level time series from five Greenland stations, with data spanning from 2007 to 2024. Due to factors such as high tide dependence, potential frozen sea, and less-maintained stations, these sea level datasets tend to be noisy, which makes them suitable for assessing the algorithm. In the QC processing, non-ML based standard procedures are applied first, including missing value detection, identification of global outliers, stability tests, date validation, and constant gradient checks. The primary focus of this research is the detection of smaller spikes and shifts. Common approaches such as moving window spline fitting, averaging of neighbouring values, and changepoint detection are explored. Hereafter, a supervised ML algorithm for spike detection is developed using manually labelled data and is compared against existing spike detection methods.

The initial QC steps are essential for identifying the most significant erroneous data. These tests are robust, efficient, and transparent, providing a solid foundation for further analysis. In contrast, the various spike detection methods show limitations in terms of performance, time efficiency and error propagation. The number of detected spikes can vary significantly, up to 100 times more, depending on the applied test, mainly due to threshold sensitivity. While the analysis of the ML step is ongoing, preliminary findings suggest that it improves accuracy compared to the best-performing statistical spike detection method.

Future efforts will focus on refining the QC algorithm and validating it using other geoscience time series.

ABSTRACT. The increasing pressure from anthropogenic activities on natural environments has highlighted the need for integrated management supported by decision-making tools capable of real-time watershed simulation and hydrological process forecasting. In the Nice Côte d’Azur metropolitan area, the Aquavar tool and its extensions have successfully combined three numerical models to account for the watershed, river, and aquifer dynamics. This integrated system has demonstrated its ability to reliably assess the impacts of development projects, extreme hydrological events, evolving hydroclimatic conditions, and point-source pollution transfers. However, challenges persist due to limitations in the quality and completeness of collected data, which hinder the tool’s real-time performance and its ability to reconstruct historical streamflow series. Incorporating artificial intelligence (AI) techniques to enhance simulation results addresses these challenges by leveraging observational data alongside physical model simulations. This study assesses integration of AI to monitor water resources and extreme hydrological events. It is innovative compared to existing forecasting tools, which typically lack AI integration or reanalysis of observational and modeled data. For instance, applying an AI technique to a specific watershed significantly improved simulated hydrographs at hourly intervals. Testing this approach on other watersheds will further validate its applicability for integration into the existing platform. The inclusion of AI demonstrates its value in enhancing the use of observational and simulation data, improving flow forecasts, and supporting the management of extreme events and water resources.

ABSTRACT. Machine learning (ML) models for streamflow prediction have been widely employed with remarkable potential in recent years. However, the residuals (the difference between observation and prediction) often exhibit heteroscedasticity, with the variance increasing as streamflow values grow. This study proposes a methodology to address the uncertainty in ML models for streamflow prediction using cluster methods to achieve higher homogeneity in the variance of residuals. The study area is the upper Fluvià River catchment with the outlet point at Olot located in Catalonia, Spain. The methodology involves several steps: First, precipitation and streamflow data are collected. The data corresponds to two periods, from January 2000 to December 2008 and from January 2011 to December 2021. Second, the data is split into three subsets to train the ML model for streamflow prediction, create the uncertainty estimation model, and test both previous models together. Third, the Random Forest (RF) algorithm is used to predict streamflow three hours ahead, with variable selection and hyperparameter tuning. Fourth, the RF model is used on the uncertainty estimation subset to obtain predictions. Then, a clustering model is trained by employing the predictions and variables of the RF model. For that purpose, the c-Means algorithm is applied, which fuzzily divides the data into clusters by assigning memberships to each cluster for every single set of inputs. Fifth, for each cluster (defined by the highest membership), residuals are calculated as well as their distribution based on Gaussian mixture models (GMMs). These models employ weighted Gaussian distributions to estimate the residuals’ joint distribution in each cluster. Finally, the RF and c-Means models are applied to the testing set to compute streamflow prediction and cluster membership for every single set of inputs, respectively. The uncertainty of a set of inputs is computed by multiplying the membership to each cluster by the corresponding GMM confidence interval, and then summing these products. The results show that the GMMs can acceptably describe the distribution of the residuals in the generated clusters. It can be seen that the uncertainty (90% interval of the GMM distributions) due to the prediction of the highest events is not constant and considerably varies in high values. This allows obtaining a close description of the residuals’ variance and containing the majority of the observed values in the uncertainty interval for this range of values. Considering the two highest events of testing, 85.5% of the observed streamflow values higher than 1-year return period are inside the uncertainty interval and only 8.6% are higher. The mean error of the highest values and the upper limit of the uncertainty interval is 1.86(m3/s), a significantly low value considering the highest streamflow (131.4 (m3/s)).

ABSTRACT. Effective reservoir management requires balancing flood control, water supply, and operational efficiency under changing hydrological conditions. This study applies deep reinforcement learning (DRL) models—Deep Q-Network (DQN), Proximal Policy Optimization (PPO), and Deep Deterministic Policy Gradient (DDPG)—to optimize reservoir operations at Soyang Dam, South Korea, over a 30-year period (1993–2022). Unlike traditional approaches that struggle with computational complexity, DRL leverages observed and remotely sensed hydrometeorological data, including precipitation, soil moisture, temperature, and runoff, to enhance adaptive decision-making. The models were evaluated on key operational metrics, including flood control violations, efficiency, and stability. DQN demonstrated the highest cumulative reward (51.605), the lowest flood control violations (23), and the smallest target storage deviation (759.759 million m³), making it the most effective for flood risk mitigation. PPO exhibited the highest efficiency (0.99) and the most stable decision-making (0.069), making it suitable for scenarios requiring consistent, controlled operations. DDPG maintained a balanced performance but had slightly higher flood control violations (27) and greater storage deviation (826.955 million m³), indicating a trade-off between flexibility and precision. The flood risk ratio analysis further confirmed DQN’s ability to mitigate extreme flood events while maintaining storage within operational limits. By integrating reinforcement learning with observed and remotely sensed hydrometeorological data, this study presents a scalable, adaptive approach to reservoir optimization. The findings underscore the potential of DRL-driven flood mitigation strategies, enhancing climate resilience and sustainable water resource planning.

ABSTRACT. Data acquisition and machine learning tools have jointly invested hydroinformatics, through simulation and predictive analytics in all areas of hydrology and hydraulics. In contrast, prescriptive analytics, at the last step of decision support for water systems, make little use of operational research and mathematical optimization. Some long-standing applications include reservoir management, design and control of urban networks, and hydroelectric production. Other contexts, such as managing competing uses at the basin scale, or implementing preventive and reactive strategies in response to meteorological and climatic hazards, are likely missing of reliable analytic decision models of reasonable size or computational complexity.

Still, optimization methods have evolved at the same sustained pace as the growing integration and complexity of decision-making problems. Modern methods are capable of optimizing over discrete decisions, nonconvex relations, uncertain forecasts, multiple objectives, composite systems, fine dynamics, and long time horizons. Combined with numerical models or digital twins, they provide systematic tools to compute final decisions together with certificates of feasibility and optimality.

In this work, we illustrate the capability of advanced mathematical optimization approaches in the context of the sustainable operation of pressurized water distribution networks involving demand-response strategies. The so-called deterministic Pump Scheduling Problem has been the focus of various optimization approaches, from linear programming in the 50s, to optimization-simulation using genetic algorithms in the 90s, or reinforcement learning today. The former only produce approximate solutions, while the latter offer no certificate of quality. Mixed integer nonlinear programming has since been applied to more accurate analytic models, hence computing both practical and optimal solutions, but the curse of dimensionality still limits the applicability of this approach to water networks of reduced size.

To overcome these limitations, we propose three enhancements: 1/ temporal and spatial Lagrangian decomposition to tackle problems of higher dimensions 2/ coupling with simulation to handle the fine hydraulic dynamics and derive practical solutions 3/ coupling with deep learning from historical data to initiate the search at nonfeasible but promising points. This proposition addresses different challenges of operating pressurized water networks: the nonconvex nature of the head-flow relation which reflects the bilevel structure (hydraulics/economy) of the problem, the combinatorial nature of the decision of operating pumps and valves which tends to put heuristics in trouble, the coupling nature of the reservoirs and water tanks, both temporally and spatially, and the proper dynamics of flows versus storage. These characteristics being common in hydraulic and hydrological systems, we expect that such a proposition, or parts of it, could find applications in other water management decision-making problems.

ABSTRACT. The estuary of Guadalquivir is located to southern part of Spain under Sevilla and Malaga. Its oceanic part is in the Gulf of Cadiz (Atlantic Ocean) but remains exposed to the mediterranean climatic changes. For this reason, it has been observed for years an event called hyperturbid, characterised by suspended sediment concentrations of several grammes per litre averaged over the water column (Dijkstra, 2018). This kind of events can have many ecological and anthropogenic impacts within the estuary and is marked by a large sediment plume visible by satellites. The main objective of this study is to use those hyperturbid events to test a GPU-accelerated high-resolution 2D suspended sediment transport model. This model combining the two-dimensional Shallow Water Equations (SWE-2D) with the 2D advective-diffusive transport equation for suspended sediments, is using an existing SWE-2D model from Martinez-Aranda (2023). The satellite data used by Megina (2023) were recovered from the COPERNICUS web browser from the 28th of November 2019 to the 21st of February 2020 and treated to obtain an approximation of the suspended particulate concentration of the sediment plume. Then, our model was used, at the same location and timeline as Megina (2023)study, with a triangular mesh grid of between 250m and 50m of resolution. Our model’s results were then compared to the data recovered from the satellite images as well as the efficiency of the model (computational time versus real used time).

ABSTRACT. Flood waves resulting from the failure of a hydraulic structure or extreme meteorological events can propagate through urban or industrial areas. The presence of urban structures in the flow path may generate localized 3D phenomena, such as vertical displacement of the flow and recirculation. Experimental studies remain an essential tool for investigating such flow dynamics; however, the flow unsteadiness often limits the spatial and temporal resolution of the datasets. Additionally, accurately capturing 3D phenomena near obstacles can be challenging with current measurement techniques.

To address these limitations, this study explores the use of 3D simulations to complement observations from reduced-scale experiments of transient flow through several idealized urban forms. Combining both approaches yields a more comprehensive understanding of the impact of urban structures on flow behavior.

After numerical validation on the water depth and horizontal velocity evolution, the energy head downstream of the urban area is computed. Local trends identified from the experimental dataset are spatialized using CFD simulations with the VOF module of the solver code_saturne. Integrated results in a channel cross-section confirm the trends observed experimentally. Such results emphasize the benefits of performing numerical simulations to evaluate inaccessible flow variables and trends that may be used in larger flood scale models.

ABSTRACT. Extreme rainfall events have become more frequent over the years and are on the increase with climate change. In addition, urban areas are expanding as populations migrate ever more towards the cities. As a result, current urban infrastructures are beginning to suffer from a severe lack of drainage efficiency during extreme rainfall events, which can cause underground drainage systems to overflow in certain districts, bringing sewage, waste, pollutants and excrement to the street surface. These overflows strongly impact people health (illness, death, etc.) and the infrastructures (damage to street furniture and buildings, pollution of soil and waterways). Urban areas must adapt and, in recent years, an increasing number of urban flooding works focused on sewer flow and surface runoff interactions. However, to the best of our knowledge, the experimental investigations required to validate the numerical models coupling 1D (in the sewers) - 2D (in the streets) approaches were restricted to a single street and have not covered a neighborhood scale. The aim of the present research is to measure the vertical exchanges of flood waters between a street network and the sewer network, at a district scale, using a 5.4 m long and 4.8 m wide physical model. The latter, termed MURI (Urban Model for the study of Inundation Risk), and located at INRAE, Villeurbanne, France, is a tilted platform that comprises two floors. The sewage network at floor #1 offers the possibility of supplying water via seven different inlet pipes and of evacuating it via one outlet pipe. The floor #2 comprises three longitudinal streets crossing three transverse streets, and 16 building blocks; the flow can enter the streets through three to nine upstream inlet tanks and evacuate through three to nine downstream tanks. The floor #2 is connected to floor #1 by 13 vertical drains running from the streets to the sewer network through which drainage can be imposed, or overflow by artificial saturation of the sewer network. Two types of scenarios are considered: the first one for which the sewer network overflows in the streets and the second one for which the surface flow is drained towards the sewer network. For each scenario, the flow discharge is measured along each pipe using flowmeters, and across each street by scanning the velocity field using an Acoustic Doppler Velocimeter (ADV) with a side-looking probe. Besides, the surface velocity fields in the flooded street network are investigated based on the Large-Scale Particle Image Velocimetry (LSPIV) and water surface elevation is measured using an ultrasonic probe.

ABSTRACT. The present work aims at measuring and computing a pollution released during a flood in a highly urbanized area (e.g., city center). A laboratory experiment mimics a flooded street-network including an urban block of varying porosity, permitting to vary the intrusion water discharge from the streets towards the urban block. A scalar is suddenly released locally, at different locations (one at a time). Two types of experiments are carried out: conductimetry measurements in the street network, with the scalar made of water and salt, making it possible to estimate the scalar discharge in each street; and colorimetry measurements in the urban block with the scalar made of water and food coloring, permitting to measure 2D-horizontal instantaneous scalar concentration fields in the block. The numerical calculation of these pollutant transport and mixing in the streets and the urban block is carried out by solving the unsteady 2D-shallow water equations for hydrodynamics and by computing the scalar transport based on the unsteady 2D-advection-diffusion equation. The flow field considered for the advection-diffusion computations was either time-averaged or taken periodic when oscillations remained in the flow field (e.g., jet meandering). The advection-diffusion equation involves a diffusivity tensor to be determined. In a first attempt, the simplest approach is selected using a Fickian and isotropic tensor with a single coefficient to be calibrated. This calibration is performed so that the model properly predicts the transverse dispersion of the pollutant measured in one of the inlet streets. The comparison between the computed and measured data reveals that both the hydrodynamics and the scalar transport are well reproduced numerically with this simple approach. A striking result is that while the diffusivity tensor was calibrated based on scalar mixing in a street, the value of the tensor coefficients also hold as the flow invades a building block, involving more complex and varied flow fields. Future work will be devoted to testing more advanced formulations of the 2D diffusivity tensor to assess their ability to improve the numerical simulation of pollution in urban floods. Based on a thorough review of existing approaches, several promising formulations have already been identified.

ABSTRACT. Urban floodwaters can carry a wide range of materials, from urban furniture to dissolved substances. The present work focuses on the transport of dissolved substances in urban floodwaters (e.g. chemical and bacteriological matters). The main objective is to experimentally investigate the intrusion of a pollutant into a porous city block, which is surrounded by a flooded street network. In particular, we attempt to answer the following question: how does the number of passages in the city block facades affect the temporal and spatial variability of pollutant concentrations within the block? The experiments are conducted on a tilted physical model termed M.U.R.I. (Urban Model for the study of Inundation Risk), located at INRAE, Villeurbanne. The flow conditions mimic flow intrusion into a city block surrounded by four streets during urban flooding. The city block facades have a variable porosity, comprising two passages (case C1), four passages (C2), or six passages (C3). This results in a highly variable flow pattern within the block. For instance, the latter features two main recirculating flow cells for the low porosity case (C1) and four cells in the more porous case (C3). A colored iso-dense scalar, mimicking a dissolved pollutant, is released at a fixed location in an adjacent street, on the side of the block just upstream from the intrusion passages toward the block. The time-evolution of 2D instantaneous concentration fields is then measured within the block using a colorimetry technique with a camera located above the block. For a steady release of pollutant, the 2D distributions of time-averaged concentration c(X,Y), normalized by the spatial and time-averaged concentration 〈C_(s-Ci) 〉 for case Ci with i = 1 to 3, indicate that the scalar field is highly dependent on the number of street-block passages and on the hydrodynamics within the block, notably the number and size of the shallow jets and recirculating flow cells. For instance, for case C1, the scalar rapidly spreads along the outer layer of the two cells but more slowly inside the cells and ends up as a homogeneous steady-state concentration field. For cases C2 and C3, the complex oscillating jets rapidly spread the scalar over the whole block area with maximum concentration encountered along the upstream-most region, resulting in heterogeneous steady-state concentration fields. In addition, the spatially averaged concentration for each of the three cases all reach a plateau concentrations 〈C_(s-Ci)〉, which differs from one case to another. This discrepancy can be attributed to the varying inlet concentration magnitudes of the three cases, which decrease along the X-axis.

We can thus conclude that the dynamics of the pollutant transport processes within a flooded city block is highly dependent on the number of street/block passages and on the hydrodynamics inside the block.

ABSTRACT. Urban pluvial flooding, which occurs when rainfall exceeds the capacity of stormwater drainage infrastructure, is one of the major challenges for modern cities. Driven by rapid urbanization and climate change, it often results in significant damages and socio-economic impacts. Accurate and efficient models are needed to understand and predict flooding events, thus permitting early warning of exposed population and informing risk mitigation strategies. However, modelling the coupled free-surface and drainage network flow faces considerable challenges, especially in large-scale or data-scarce areas. Complete models that solve the entire drainage network explicitly, combining two-dimensional (2-D) elements for surface flow and one-dimensional (1-D) elements for the components of the drainage network, are complex to setup and require lot of (often unavailable) data. Simplified methods, based on quite arbitrary reduction of rainfall hydrographs or increase of infiltration rates, lack a robust physical basis. This study presents a subgrid modelling approach for simulating urban pluvial flooding that integrates the contribution of the underground drainage network within a 2-D framework for surface flow. The presence of stormwater pipes is accounted for in the 2-D framework by complementing the free-surface cell conveyance with the anisotropic contribution provided by pipes. This is facilitated by the porous formulation of the Shallow Water Equations, in which the water level can represent either the free-surface elevation in ponded area and the piezometric head of the pressure-driven flow in the underground network. Only the salient features of the drainage network can be assigned to (portion of) 2-D cells. The input parameters (pipe diameter, direction, roughness, and lateral spacing) can be estimated through limited surveys or general characteristics of the area, reducing the need of detailed datasets. This also eliminates the need to explicitly couple a large number of 1-D pipe elements with the 2-D grid, making the model computationally efficient and suitable for data-limited scenarios. Some validation tests, in which the subgrid model is compared to an explicit 1-D/2-D coupled solver, confirm the accuracy and applicability of the proposed approach in schematic yet realistic scenarios. The results demonstrate that, under the hypothesis of limited capacity of the pipe network compared to that of inlets (drains, manholes), the new approach achieves comparable accuracy to traditional methods while reducing the computational time and, above all, the effort needed to setup the computational grid. These findings highlight the potential of the subgrid model as a practical tool for flood hazard assessment and management in large flood-prone urban areas.

ABSTRACT. Human exposure to urban floodwater poses considerable health risks due to the potential contamination from various waterborne pathogens. Sediments serve as carriers for pathogenic contaminants, facilitating their transportation and deposition through floodwater onto the surrounding surfaces. Contaminated sediments present in wastewater can enter surface flood flows via manhole/gully surcharges from the sewer network.

This study aims to provide new experimental datasets and enhanced understanding regarding the exchange of sediments from piped drainage networks to surface systems utilising an experimental scale model. The experimental tests utilise a 1:6 pipe/manhole/surface scale model at the water laboratory at the University of Sheffield (UK). It links a model surface floodplain to an urban drainage system via a manhole shaft. For the purposes of these experiments the system has been modified to allow the study of the transport of sediments within the pipe system and the potential transfer to surface flow, allowing the injection of sediment loads with particle sized between sized between 148 ≤ d ≤ 458 μm into the sewer pipe flow. This range of sizes aims to be representative of road sediments. A series of experiments utilising calibrated turbidity monitors were conducted to determine proportion the sediment load transferring to the surface flows under several different hydraulic conditions. Further, a Particle Tracking Velocimetry (PTV) system is deployed at the manhole to monitor the motion of individual sediment grains and capture the detailed three-dimensional velocity vector field, as the particles traverse the manhole structure. The resulting dataset serves as a high-resolution validation resource for Computational Fluid Dynamics (CFD) models, significantly enhancing their accuracy in simulating velocity fields and sediment transport processes. Moreover, the data provide novel insights into sediment particle trajectories and their distribution within manhole structures, advancing the understanding of sediment transport behaviour in urban drainage systems during flood conditions.

In addition to the in-manhole studies, an overhead camera system quantified the spatial distribution and settling patterns of sediment grains within the surface floodplain, with a focus on their positions relative to the manhole structure. This setup allowed precise measurement of settling distances from the manhole and analysis of resulting deposition patterns. The collected data provide critical insights into sediment transport and deposition behaviour in surface flows influenced by manhole structures under surcharging conditions. These empirical observations offer a valuable resource for validating sediment transport models and enhance understanding of deposition dynamics in urban drainage systems during flood events.

ABSTRACT. In the next decade, more than 2.3 billion people around the world will live in areas of water stress (Organization for Economic Cooperation and Development). Water scarcity, demographic changes, and operational efficiency are major issues amplified by climate change impact. Water utilities are facing several challenges related to distribution water networks. These challenges include the need to rehabilitate large parts of these networks in the next 10-30 years. Additionally, according to the World Bank database, the current level of non-revenue water (NRW) in the developing world is approximately 35% of the water produced.

In this situation, management and water supply networks operation become a rising challenge as cities grow, water infrastructure ages, and environmental issues increase being necessary to reduce water losses, related wasted energy and resources.

Pipe failures are often attributed to factors associated with pipe conditions, loadings, external factors, and their combined effect. In addition to these, intensive pipe cyclic pressure variations under steady and unsteady state hydraulic conditions contribute to structural degradation of pipelines, accelerating fatigue crack growth. While the impact of extreme pressure transients (water hammer) is well studied, the medium- and long-term impact of the unsteady and quasi-steady state pressure variations on fatigue-related cracks and deterioration in water supply systems have received little attention.

TRANSMITWATER (TRANSient Management & mITigation solution for WATER utilities) is a 3-year project under development by the consortium of companies CETAQUA, Aigües de Barcelona, and AQUATEC, which aims to reduce water losses, pipe failures and service disruptions, increasing the overall hydraulic efficiency and service quality of water networks, through the control and management of transit pressures.

The project consists of the development of a Decision Support System (DSS) to reduce pipe failures caused by transient pressures occurred in daily water network operation for water utilities.This DSS encompasses three different tools for analysing transient´s impact and mitigate them (a pipe failure tool, an assets life expectancy tool, and an investment planning tool), plus a methodology for locating transient pressure sources. This DSS will be integrated into a current platform that helps water utilities with their networks and its leakage management.

The processing of data collected by high-frequency pressure sensors deployed in the water supply network will allow assessing pressure transient events and implementing mitigation actions to reduce their impacts following a smart approach based on Artificial Intelligence (Machine learning). The TRANSMITWATER solution will allow water utilities to adopt the best mitigation technique for their network while reducing costs, failures, and extending asset life.

The TRANSMITWATER solution will be demonstrated in 3 operational environments: water supply networks of Eliana (Valencia, Spain) and Soutomaior (Pontevedra, Spain) with data previously gathered, and in Barcelona water supply network (Spain) where sensors are currently deployed.

ABSTRACT. A reliable estimate of reference evapotranspiration (ET0) requires several meteorological inputs, which may be unavailable in regions with limited data availability. In this regard, this study aimed to address the following objectives: First, to evaluate the effectiveness of ERA5_Land (ERA5_L hereafter) weather forecasts in providing daily agrometeorological variables for the period 2003-2021 at 10 study sites distributed over both plain and mountainous areas in a North African basin. Second, to investigate whether downscaling the ERA5_L data (10 km) to station scale (250 m) using a quasi-physical based model, MicroMet, could improve the reliability of the meteorological variables. Third, to compare the performance of the original ERA5_L reanalysis data and the disaggregated ERA5_L data (MicroMet) as potential sources for accurate estimation of ET0 on a daily time scale. Finally, to assess the long-term spatiotemporal changes in ET0 across the Tensift basin over the period 1950-2021, and the influence of climate variables and topography on ET0 variability. The findings of the study revealed that the original ERA5_L estimates of air temperature (Tair) were the most accurate among the studied variables, followed by solar radiation (Rs), relative humidity (RH), and wind speed (u2). When considering the disaggregated daily ERA5_L data, Tair exhibited the highest performance, followed by RH, u2, and Rs. Tair and especially u2 demonstrated an improvement across the plain and mountainous sites. However, Rs was generally degraded after Micromet. Then, a comparison was conducted between daily ET0 obtained considering both datasets and show similar correlations between ground and simulated data but with an overestimation of ET0 after MicroMet. Finally, the retrospective analysis of ET0 showed three main phases with a decrease of ET0 between 1950 and 1970, a nearly steady period during 1970-2000, and a significant increase from 2000-2021. This study provides a comprehensive insight about the potential and limitations of ERA5_L products in arid North African regarding irrigation and water management under climate variability.

ABSTRACT. Abstract for a Poster presentation

Soil water (SW) is essential for accurate hydrological modelling, as it influences key processes such as infiltration and surface runoff, which affect the overall water balance in a watershed. In the Soil and Water Assessment Tool (SWAT) model, integrating surface soil moisture (SSM) data into the calibration process has been proposed as a way to improve simulation accuracy. This study focuses on the Argens watershed (France) and aims to answer three questions: (1) How accurate is SWAT for simulating SW content? (2) Can SWAT’s SW simulations be improved by integrating SSM data into the calibration process? and (3) How does integrating SSM data impact streamflow simulations? To address these questions, SWAT simulations of SW and streamflow were performed from 2001 to 2020 using four calibration strategies: no calibration (No_Cal), calibration with streamflow (Q_only), calibration with SSM (SSM_only), and a combined calibration with SSM and streamflow (SSM_Q). Simulations were evaluated at the watershed and sub-basin scales using the Pearson correlation coefficient (r), Kling-Gupta Efficiency (KGE), and percent bias (PBIAS). Results to the three key questions of the study are as follows: (1) The Q_only calibration shows that SWAT’s SW simulations are well correlated with the observed SSM at the watershed scale (r = 0.7). However, performance at the sub-basin scale varies widely, with PBIAS values ranging from -22% to +25%, indicating significant underestimations in some sub-basins and overestimations in others. These differences are partly due to the restrictive minimum and maximum bounds for SW simulations in SWAT. (2) Integrating SSM data into SWAT calibration (SSM_Q) improves the representation of SW at the sub-basin scale by adjusting these bounds. However, this improvement varies across sub-basins and depends on soil texture, suggesting the need for a specific calibration to each sub-basin based on its soil characteristics. Moreover, while the SSM_Q calibration strategy improves the SW simulation at sub-basin scales, the SW accuracy at watershed-scale is unchanged (r = 0.7), and the bias performance remains comparable to the uncalibrated model (PBIAS = -11%). (3) In addition, the SSM_Q calibration strategy negatively affects the accuracy of streamflow simulations, with a KGE value decreasing from 0.75 (Q_only) to 0.63 (SSM_Q). Overall, integrating SSM data into SWAT calibration improves SW simulation at sub-basin scales, but it decreases streamflow simulation accuracy, and requires more time due to two calibration steps. While this might be advantageous for SW simulations at local scale, this approach is of smaller interest for optimized streamflow simulation.

ABSTRACT. Rain-flow modelling is an essential tool for understanding the behaviour of mountain karst systems, particularly in regions affected by a mixed precipitation regime with a significant snow component. The Dévoluy karst, located in France's Southern Alps, is a perfect example of this problem, given the high contribution of snow to the recharge of the hydrogeological system. To assess the impact of snow recharge on karst reservoir discharge, we used KarstMod, a reservoir rainfall- discharge model adapted to karstic hydrogeological systems. Our simulations show that the absence of snow recharge during modeling leads to a significant underestimation of spring discharge, resulting in an inability to accurately reproduce the system's behavior during the snowmelt period. By integrating KarstMod's snow routine, we were able to parameterize the accumulation and progressive release of water in the system, reproducing the effect of snow accumulation and melting. This enabled us to significantly improve the modeled flow, both in spring during the melt period and in winter, when the water is stored in the snow and does not contribute directly to the discharge. This approach also enabled us to assess the sensitivity of the karst system to climate variations by modifying the snow routine parameters to simulate temperature changes associated with global warming. Our results show that higher temperatures will lead to a reduction in the snowpack, causing faster recharge in winter. This results in reduced water storage in winter, lower discharge in spring and longer, more pronounced low-water periods in summer. This change in the hydrological cycle raises important questions about the resilience of the karst system to climate change and its implications for water resource management. In conclusion, this study highlights the crucial importance of taking account of snow processes in hydrological modelling of mountain karst systems. The use of KarstMod, coupled with fine calibration of its snow routine, not only improves the representation of reservoir outflows, but also enables us to explore the potential impacts of climate change on these vulnerable systems. This approach opens up prospects for improved water resource management in mountainous karst environments, particularly in a context of changing precipitation patterns and rising temperatures.

ABSTRACT. In the water engineering sector, numerical modeling is widely used to study hydraulic and hydrological phenomena, such as flow dynamics, sediment transport, and flood propagation. However, traditional physical modeling remains essential for capturing the complexities of hydraulic systems in a tangible manner.This study aims to compare the results obtained from both physical and numerical models, to evaluate the physical modeling consistency and to demonstrate how physical modeling can enhance our understanding of real hydraulic phenomena. To achieve this goal, three reduced-scale physical models were created using 3D printing technology and subjected to hydrological simulations with a rainfall simulator. The case studies include the Vésubie catchment (located in the south of France), the remnants of the Malpasset Dam, and the Vésubie canal intake. In parallel, 2D numerical models of these systems were developed using TELEMAC-2D and HEC-RAS hydraulic codes. By preserving critical parameters, such as Froude number, the study ensured the comparability of both modeling approaches despite the atmospheric tensile force factor which apply a surface shear stress approached with the Weber number.The results revealed notable differences between the physical and numerical models, particularly in discharge outputs, due to the reduced scale of the physical model and the complexity of real-world hydraulic behavior. Despite these differences, scenarios such as storm Alex and the Malpasset dam overflow were effectively simulated, providing valuable insights into hydrological and hydraulic responses with the infrastructure performance. These findings highlight and confirm the potential of integrating 3D printing into hydraulic modeling to complement numerical approaches.

ABSTRACT. PRESENTATION TYPE [C] - POSTER An accelerated implementation of regulations and projects concerning water reuse is noticeable over the last ten years in European and Mediterranean countries. For instance, France has published in four years, 2020 to 2024, seven regulations texts addressing water reuse legal framework, the double of publications compared to the last decade. This induces an increase of planed and implemented projects. Yet, in France, a consequent part of those projects are not in working order because of the regulation swift evolution and the lack of historical background in those thematics. The objective of this poster is to provide a review of the implementation of reuse to date, in various areas, on a watershed scale. This work focuses in particular on the approach of the Llobregat river in Barcelona (Spain), since it is a good illustrator of usages implementation. Since the middle XXᵉ century, various usages has been incremented at a time to fit with the reuse strategy of the territory. The proposed review will take stock of a subject that is evolving rapidly, namely water reuse, to mapping the reasons of success and failure of the various initiatives. It will serve as a basis for developing strategic planning methods at catchment scale to developpe integrated water reuse intitatives. This opens up perspectives for the use hydroinformatics modelling tools as assessment tools for water reuse projects strategic implementation. This approach of water reuse accords to non operational or functional projects to have facilities to be implemented.

ABSTRACT. PRESENTATION TYPE [C] - POSTER The French Riviera, situated at the heart of the Mediterranean, has become increasingly vulnerable to extreme climate events in recent years, such as prolonged droughts, floods, and sudden inundations. These challenges highlight the urgent need for a rethinking of what water resources are. In response, a Research and Innovation partnership was established in 2024 within the territories of the Maritim Alps Department and University Cote d’Azur to explore innovative solutions, including the potentialities of the reuse of treated wastewater (REUT). Wastewaters represent potential human uses, and also minimum flows during low flow periods. Thus, management of wastewater potentialities is crucial to face climate change impacts on the water ressource. Yet significant obstacles remain, in particular the absence of a methodology to quantify and anticipate the spatial and temporal variability of wastewater production, compromising wastewater resource management. This study seeks to develop a method for assessing and mapping the seasonal production of treated wastewater in relation to population and treatment plant capacity, by applying it to the Cagne River watershed. The Cagne River watershed was chosen for its representation of wastewater production potential in the coastal areas of the Maritim Alps Department. The proposed method combines hydrological analysis and geospatial tools to estimate wastewater availability, treatment capacity, and drought compensation potential. The objective is to estimate treated wastewater production over space and time to facilitate its reuse, especially during drought conditions, and to support low flows in rivers. The study is divided into three main parts. First, list the specifications for data collection to assess treated wastewater volumes by identifying key parameters. Second, develop a tool to estimate the monthly contribution of wastewater treatment plant (WWTP) discharges that support low-flow conditions in the Cagne River. Finally, it provides the replicability of this methodology to all the Maritim Alps Department.The outputs include a detailed map of estimated seasonal treated wastewater flows, and a demonstration of the potential for REUT considering the maintenance of low flows. Potentially this methodology could be useful for any other scarce regions facing similar challenges.(C)

ABSTRACT. [C] POSTER

Parita’s Bay, a protected Ramsar site on the Pacific coast of Panama, faces coastal flooding and erosion due to climate variability and extreme events. Refined assessments are required as a diagnosis for effective risk management and conservation efforts. Therefore, the goal of this study is to evaluate the link between coastal hydrodynamics and hydrological processes in two watersheds that discharge their water in Parita's bay. Parita River watershed (575 km2) and La Villa River watershed (1,269 km2) are in the central region of Panama. Coastal and fluvial floods were simulated using IBER Software, numerical simulations included storm events due to rainfall and coastal flooding due to tides. Rainfall measurements from three meteorological stations from 1990 to 2020 were used as a proxy for the hydrological process and numerical simulations of tidal range and wave height from 1990 to 2021 as a proxy for coastal hydrodynamics. The roughness and infiltration of the watershed were based on the land use. The terrain was generated by merging the Digital Elevation Model (DEM) with bathymetric data. The area is predominantly used for agricultural activities, in which gentle slopes (ranging from 0 to 8.5º) and modest elevations (92 m to 163 m). Four scenarios were considered: (i) largest precipitation, (ii) extreme storm surge event with the largest significant wave height, (iii) spring tide presenting the highest tide of the year, and (iv) mixed event, i.e.: storm event with extreme swell coincides with the largest precipitation. The study's results were compared to the registered water elevation flood of selected storm events and hyetograph by IDF curves based on different return periods. Results such as water flow velocities, river discharge, and flood heights will provide practical guidance for flood risk management and territorial planning in the region and help identify areas in the coastal zone prone to flood and coastal erosion. The results will contribute to assessing the coastal vulnerability of the study area, providing an initial diagnosis to improve flood risk management and territorial planning in the region.

ABSTRACT. Dakar, Senegal's capital and most densely populated city, has experienced numerous periods of intense rainfall over the past three decades. Located on the edge of the sea, the area has little relief, high groundwater levels and permeable soils that saturate relatively quickly. As a result, during the rainy season from August to October, the city has suffered severe material damages due to flooding. To highlight the consequences of these events, the methodology involves the use of several hydraulic models: SWMM to simulate the urban drainage network, TELEMAC-2D to analyse surface runoff and InfoWorks for integrated modelling. This study focuses on Yeumbeul, a watershed, located in the suburbs of Dakar. SWMM was used to modify the original network, which was unable to manage the amount of rainfall during a flood. The TELEMAC-2D model aims at modelling the extent of the overflow and the accumulation of water in the study area. Past rain events, 2009 and 2017, were used as reference. The results of these simulations were compared with those obtained using InfoWorks. These outcomes allowed to map inundation risk and to compare models in order to decide on the measures to take.

ABSTRACT. TYPE [C]-POSTER

ABSTRACT. Flood hazard prediction is essential for risk assessment and management. Flood hazard maps are crucial in identifying at-risk areas by providing information on flood depth, inundation extent, and frequency. These maps support decision-makers, emergency responders, and engineers in mitigating flood impacts by offering detailed information on flood characteristics, and their accuracy is vital for effective flood management strategies. With the rising frequency of natural disasters, there is increasing concern about pluvial (surface) floods caused by intense or prolonged rainfall. Unlike fluvial floods, which primarily affect areas near river channels, pluvial floods can impact entire catchments, highlighting the need for further research and improved modeling of these events. While fluvial flood hazard mapping along major rivers is a well-established practice, the mapping of non-fluvial floods remains limited. Traditionally, flood scenarios are modeled using separate hydrological and hydraulic models: hydrological models generate discharge hydrographs at specific catchment points, which then serve as input for hydrodynamic models that simulate water flow. However, manually integrating these models is time-consuming, requiring separate calibration and simulation. The Rain-on-Grid (RoG) technique addresses these challenges by directly applying precipitation to the grid cells of a hydrodynamic model, thus integrating hydrological and hydrodynamic processes within a single model (Costabile et al., 2021). The RoG technique can be applied, for example, in the well-known HEC-RAS software. Numerical models offer precise flood predictions but require extensive high-quality data, including high-resolution digital elevation models, design rainfall, river discharges, land use data for roughness parameters, and river network geometry. They also involve complex pre-processing and computationally intensive simulations. Given these challenges, an alternative approach—beneficial for large-scale or data-scarce contexts—is flood susceptibility analysis. This method identifies flood-prone areas based on intrinsic territorial factors, both natural and anthropogenic, without considering the temporal probability of specific flood events (Varra et al., 2024). In line with the concept of flood hazard, flood susceptibility refers to an area’s predisposition to flooding based on local conditions, estimating “where” floods are likely to occur rather than “how often” they may happen. Flood susceptibility classifications generally exploit the potential of Geographical information systems (GISs) and multi-criteria analysis (MCA) techniques. A case study in a selected portion of the Low Calore River watershed in Southern Italy demonstrates the effectiveness of both approaches in predicting the impact of an actual extreme rainfall event on the territory and critical infrastructure.

- Costabile P., Costanzo C., Ferraro D., Barca P. (2021). Is HEC-RAS 2D accurate enough for storm-event hazard assessment? Lessons learnt from a benchmarking study based on rain-on-grid modelling. Journal of Hydrology, 603. - Varra, G., et al. (2024). Flood Susceptibility Assessment for Improving the Resilience Capacity of Railway Infrastructure Networks. Water, 16, 2592.

ABSTRACT. Nature-based solutions are part of the sustainable systems included in the 2030 Sustainable Development Goals (SDGs) and have gained significant traction due to their multiple benefits. These include promoting biodiversity and green spaces, managing urban rainfall and reducing runoff, mitigating the Urban Heat Island (UHI) effect, decreasing CO₂ pollutants, and lowering energy consumption by acting as thermal insulation. The effectiveness of these systems relies entirely on their design, which must be adapted to the region's climatic conditions. In Mediterranean climates, the main challenge lies in future climate projections. Latest researchers suggested the effects of climate change on precipitation patterns and pointed to a decline in rainfall with more frequent drought periods. Green roofs require a substantial vegetation mass to function efficiently. Maintaining this vegetation can be a major challenge, as it requires significant amounts of irrigation water for survival. To address this issue, each design should be tailored to the specific region, ensuring that the different layers of the system are adapted to the local climate conditions. Recent studies suggest that the most important factors influencing peak flow retention and delay are in the following order: substrate composition, substrate depth, slope, and plant species. The vegetation cover may be sparse during the rainy season because plant selection focuses on achieving greater coverage in summer to counteract the effects of urban heat islands, making the substrate crucial for retaining runoff water. This study presents a summary of results from green roof experiments conducted at the SuDS laboratory located in Escúzar, Granada province. The project, funded by IFMIF-DONES and the University of Granada, aims to develop SuDS adapted to local climatic conditions to enhance their use and optimize their efficiency throughout the project. In this study, 11 different types of green roofs have been implemented, with three replicas of each, consisting of four different substrate types and four vegetation variations, combined in different configurations. The results provide clear evidence of the risks associated with using commercial substrates and plants in green roofs. By comparing hydrograms from commercial green roofs, modified green roofs with local substrates, and traditional roofs, it was found that commercial green roofs failed compared to modified green roofs. Besides, there was no significant difference in performance between commercial green roofs and traditional roofs. The risks also lie in the fact that infiltrated rainwater was unnecessarily contaminated with organic matter, making it unlikely to be reused, except for irrigation in other types of green areas. The study also demonstrates the improved hydrological performance of the green roofs that incorporate local site substrates.

ABSTRACT. Managing stormwater effectively, both in terms of quality and quantity, is a major challenge for urban areas. Cities have drainage systems that evacuate rainwater during storm events, conveying these flows to outlet points or treatment plants. However, during intense rainfall events, the network capacity may be exceeded, leading to flooding and uncontrolled wastewater discharges, known as Combined Sewer Overflows (CSOs). To assess and mitigate the impact of these events on flood risk, public health, and the environment, it is essential to use advanced hydrological-hydraulic models that accurately represent the interaction between surface runoff and drainage networks, commonly known as dual drainage models. These models, also referred to as 2D/1D models, provide a reliable framework for simulating pluvial flooding by integrating surface flow with sewer system dynamics. Urban layouts are extremely complex due to the presence of buildings and other infrastructure that influences water flow dynamics. Additionally, sewer network cartography contains extensive information on properties, materials, geometry, and topology. The process of importing and editing this large volume of geospatial data can be particularly challenging when using the model’s graphical user interface (GUI). Consequently, GIS tools play a fundamental role in the development of these models. While some hydraulic modeling software incorporates GIS-based functionalities, their capabilities are often limited compared to specialized GIS platforms such as ArcGIS or QGIS. To address this limitation, this study introduces a QGIS plugin that integrates the Iber-SWMM dual drainage model, enabling seamless urban flood modeling within an open-source GIS environment. This tool facilitates the management and analysis of drainage system data, allowing users to define urban catchments, configure drainage network components, and assign land-use characteristics. Additionally, it includes an advanced meshing tool based on the Frontal-Delaunay algorithm for generating high-resolution computational grids. The model engine, Iber-SWMM, couples the 2D surface model Iber with the 1D Storm Water Management Model (SWMM), ensuring full bidirectional interaction between surface runoff and the sewer network. Moreover, high-performance computing (HPC) techniques, such as OpenMP (CPU) and Nvidia CUDA (GPU), significantly enhance computational efficiency, making high-resolution simulations feasible. Finally, to demonstrate the capabilities of the tool, a case study was conducted in Écija, Spain, a city historically affected by flooding due to both heavy rainfall and river overflow. The urban drainage network includes over 5,400 manholes, 22 outfalls, 5,400 pipes, 5,200 inlets, and 3 pump systems, while the surface mesh consists of approximately 2.1 million triangular elements. The model simulated a storm event with a 10-year return period. The use of high-performance computing (HPC) techniques accelerated the simulations by a factor of 200.

ABSTRACT. Recent experimental and numerical studies have significantly improved the understanding of hydrodynamic forces exerted by dam-break flows on buildings. These forces mainly depend on the water depth and flow velocity, that can be estimated from numerical simulations of the flow. A first experimental campaign provided key insights into flow dynamics and allowed to successfully validate the numerical simulations, highlighting the best appropriate methods to accurately estimate these forces. Building on this foundation, the present study investigates the influence of additional upstream obstacles on the forces exerted on a target building.

The presence of obstacles can induce two contrasting effects: a shadowing effect, where the upstream obstacle reduces the force by deviating the flow and dissipating flow energy, and a flow acceleration effect, where flow deviation around the obstacle intensifies the impact on the target structure. To systematically explore these effects, a new experimental campaign was conducted, including both steady and transient dam-break flow experiments, with varying Froude numbers and obstacle configurations. The obstacle configurations consist of different position, size, and porosity, allowing for a comprehensive assessment of their impact on hydrodynamic loading.

These experimental configurations were numerically reproduced using a validated 2D shallow-water model. The numerical results, combined with statistical analysis using the JMP software, help identify the key parameters influencing the force variations. The findings provide valuable insights into the role of obstacle characteristics in mitigating or amplifying hydrodynamic loads, with potential applications in flood-resilient urban planning and infrastructure design.

This study contributes to the broader objective of improving flood risk assessment by offering a refined understanding of flow-structure interactions in transient flood events. The findings will support the development of guidelines for the strategic placement of barriers and structures in flood-prone areas to optimize protection while minimizing adverse effects.

ABSTRACT. The increase in the frequency and severity of urban flooding is linked to three main factors: the effects of climate change, population growth and the expansion of impervious surfaces. Between 2000 and 2019, 7348 disasters occurred, resulting in 1.23 million deaths and 4 million people affected, with a total economic loss of US$ 2.97 trillion worldwide. In Europe, floods account for two-thirds of natural hazard losses. However, the number of people living in flood-prone areas is also increasing. Flood damage is typically categorised as material or non-material and direct or indirect, with material direct damage being the focus of economic assessments. This type of damage refers to monetary losses caused by the direct physical contact of floodwater with property and its contents. In this context, it is necessary to have a robust economic assessment with a focus on improving flood risk management. Flood loss models are often based on depth-loss curves that relate flood depth to monetary loss by property type. However, these curves typically consider the depth outside buildings, but the main damage usually occurs inside. It is therefore important to know the depth of water inside and outside the building. Flood water can enter a building in several ways, including through structural joints, service ducts, doors and windows, even if these are closed. The duration of a flood —including how long the water remains within a property or how long it takes to enter or leave a building— has a significant impact on the damage assessment. The extent of infiltration will depend on the external flood depth, the type of building construction and the ground conditions. Understanding these mechanisms is essential for improving flood resilience and mitigate damage. Methodologically, two-dimensional (2D) hydrodynamic models are recognised as the most effective tools for simulating urban flooding. The IBER software, based on the Shallow Water Equations (SWE), is used to model flood behaviour, including the time taken to fill urban areas, which is useful and accurate for flood risk assessment. However, due to the complex nature of urban environments, these models require high-resolution data and long computation times. Consequently, a simplified model is used to simulate the effect of building façades, both open and closed, on hydraulic behaviour during flood events by representing them as gates. This approach aims to capture the complexity of the urban environment, with its varied building characteristics and land uses.

ABSTRACT. Urban flooding is an increasingly urgent challenge, exacerbated by climate change-driven extreme precipitation events. In rapidly urbanizing areas like Grabels, Southern France, flash floods pose significant threats to infrastructure, ecosystems, and public safety. A particularly severe flood in 2014 resulted in widespread inundation, underscoring the need for a comprehensive analysis of flood risk factors. TELEMAC-2D, a finite-element model, solves shallow-water equations to simulate water levels, flow velocities, and flood extents under various scenarios. The model employs a high-resolution computational mesh built from topography, providing an in-depth analysis of flood propagation patterns. By incorporating urban expansion, land use changes, and hydrological variability over time, this research assesses how these factors influence flood susceptibility in Grabels. This study leverages the use of very high spatial resolution (VHR) satellite imagery from Pléiades (50 cm/pixel) and Pléiades Neo (30 cm/pixel) to extract critical flood risk variables. These include runoff potential (Curve Number - CN), surface roughness (Strickler coefficient), land use evolution, and hydrological factors. A dataset of five Pléiades and three Pléiades Neo images, spanning 2012 to 2024, is analyzed to derive spatially distributed hydrodynamic modeling inputs.Using monoscopic imagery analysis, the study generates eight land cover maps (OCS) at different time intervals. These maps serve as a foundation for estimating CN values and Strickler coefficients, which are integrated into the TELEMAC-2D hydrodynamic model. The 2014 flood event is used as a benchmark for model validation, enabling a comparison between simulated and observed flood extents. This analysis provides insights into how land cover variations contribute to shifts in flood risk over time. Results indicate that integrating satellite-derived flood risk variables enhances the understanding of hydrodynamic interactions in urban settings. The temporal evolution of land use and runoff characteristics reveals how impervious surface expansion and changes in surface roughness impact flood hazards. Additionally, analyzing flood risk across multiple time intervals highlights shifts in hydrodynamic behavior driven by urbanization and changes in vegetation driven by fire. By combining archived remote sensing data with TELEMAC-2D simulations, this study presents a multi-variable hydrodynamic analysis of urban flood risks. The findings contribute to flood hazard assessment, urban water management, and climate adaptation strategies, offering a data-driven approach to mitigating evolving flood risks in rapidly changing environments

ABSTRACT. Urban environments are geometrically complex and exhibit features across multiple length scales. Although high-resolution data regarding urban structural topologies is increasingly available, integrating such data into hydraulic computations remains a significant challenge, specifically, when using geometry-fitted grids. As a matter of fact, small features in the input geometry, whether they represent relevant hydraulic constraints or arise as numerical artifacts from geometry-processing algorithms, lead to arbitrarily small mesh elements. These, in turn, negatively impact the performance of numerical discretizations, either by imposing severe CFL constraints on explicit methods or causing ill-conditioning in nonlinear algebraic systems when using implicit methods.

To address the geometric complexity of urban domains, we propose a multi-scale discretization method based on the Diffusive Wave approximation. Similar to the Multi-scale Finite Element Method, our approach relies on a coarse, unfitted grid and a set of basis functions that are numerically constructed by solving local fine-scale diffusion problems. Thanks to the multi-scale basis functions, the coarse numerical model—obtained through the mass-lumped Galerkin method—is able to accurately capture relevant fine-scale geometric features. The numerical implementation employs straightforward parallel computing algorithms to construct the multi-scale basis and assemble the global coarse residual. Compared to the fine-scale Diffusive Wave model, our method delivers reasonably accurate results at a fraction of the computational cost. Depending on the level of coarsening, we observe speed-ups ranging from a factor of 10 to 100.

ABSTRACT. AquaVar is a decision support system (DSS) for water resources management in the French Mediterranean Var watershed. AquaVar DSS is based on a holistic approach chaining distributed physically-based models representing hydrological, hydraulic and hydrogeologic processes. The DSS was initially implemented in 2019 within its operational environment. The objective of the present analysis is to draw lessons from five years of operational experience and establish the subsequent steps to be taken. From 2019 to 2024, multiple hydrological events allowed for an evaluation of the tool's performance, and the results demonstrated its capacity to simulate peak flows associated with two extreme rainfall events (storms Alex and Aline). However, AquaVar showed deficiencies in its ability to simulate the intense and prolonged drought episode that occurred in 2022 and 2023. This result triggered an investigation into the physical processes involved and led to a re-examination of the modeling simplifying assumptions behind the river-aquifer interactions. This outcome underlines the need for DSS to adapt in response to changing climatic conditions, which give rise to unprecedented hydrological processes. The initial five years of implementing AquaVar under operational conditions have yielded several other insights, covering topics such as DSS modularity, data positioning, technological aspects, and governance.

ABSTRACT. The development of hydroinformatics technologies is able to support the decision-makers to have more scientific and accurate measures to defence the catastrophic flood disaster on catchment scale. Taking the “23·7” catastrophic flood occurred in Ziya river basin (47,000 km2) in 2024, a modelling-based assessment approach has been presented to understand the flood characteristics and to evaluate the implementation of certain measures. With more than 12.6 billion m3 rainfall landed in 7 days, the decision-makers for managing the flood in Ziya river basin faced large pressure and have to find an operational measure to defence the flood. By using the big reservoirs located at upper and middle parts of the catchment, the flood amount enter to the Xianxian detention area had been reduced 68%. Based on a distributed hydrological model calibrated with averaged NSE value higher than 0.7, the flood progress has been represented by the numerical tool and the effects of the big reservoirs has been also evaluated. The different ratio of the Xianxian flooded area with and without the reservoirs impacts is more than 70%. The modelling-based approach presented in this study shows high applicability to be promoted to any other large catchment with complicated flood characteristics and hydraulic structures.

ABSTRACT. Recent extreme weather events, exacerbated by rapid climate change, demonstrated the necessity to enhance river management. This management must occur on a watershed scale, as water does not care about administrative boundaries. Strategies can encompass various approaches: river renaturation, flood protection measures, runoff reduction or people education. Large rivers benefit from substantial resources to mitigate the adverse effects of past anthropogenic modifications, but smaller rivers are frequently overlooked due to limited budgets. However, their importance in the global system, both in terms of length and ecosystem services, is well established. The potential added value of working at the scale of small rivers and catchments is considerable, including improved low-water flows, resilience to flooding, increased biodiversity (quality and diversity of the environment), reduced sediment accumulation, and reduced bank erosion. Nowadays, the ecological quality of watercourses is monitored using indicators that assess one component each, such as hydromorphology, biology, or physico-chemistry. Although these components strongly interact, current research hardly considers those interactions and separates hydraulic performances and ecological quality. Moreover, research on small river renaturation projects predominantly utilizes empirical approaches, which are insufficient for evaluating inter-component effects and are limited in terms of precision and ability to integrate complete renaturation scenarios. This complicates the development of a prioritized plan of action. While several specialized models address this need, they don’t consider cross-effects. One notable example is the Habby application (INRAE, France), which models the hosting potential of target species based on hydraulic data and a substrate survey. To enhance the efficiency of such tool, it is imperative to extend the scope to encompass physico-chemical water quality and further morphodynamical changes in the substrate. The Rivialis project aims at offering a novel approach to address these challenges. This ambitious project integrates high-resolution hydraulic simulations based on accurate spatial and temporal data and detailed eco-hydraulic models to define a small-river quality indicator (indice petit cours d’eau – PCE) encompassing a wide range of characteristics of the river. This will be achieved through the development of a digital river twin, i.e. a digital replica of a river, that is specifically designed to study renaturation projects. The tool encompasses the following capabilities: •The assessment of watercourse quality employing a novel multi-component PCE index that incorporates hydromorphology, biology, hydraulics, watershed, hyporheic processes, and interactions between the hyporheic area and surface waters; •The evaluation of the impact of work scenarios on watercourse quality through the previously defined index and the projected changes in quality, modelled using the Watlab hydraulic simulation framework developed at UCLouvain (https://sites.uclouvain.be/hydraulics-group/watlab/); •The prioritization of interventions according to environmental constraints and the optimization of budgets by reducing study costs; •The promotion of sustainable management adapted to small watercourses for users with varying levels of expertise.

ABSTRACT. With the development of sensors and image analysis technics, the image velocimetry has attracted the attention of hydrologists because of its low cost and convenient maintenance compared with the traditional flow velocity detection method. However, current open-source datasets only have a small number of samples and lack complete and accurate annotations, such as ground reference points (GRP) and true velocity. Therefore, laboratory experiments provide useful complementary information for testing and studying the robustness of image velocimetry and models of flood. Based on a series of physical experiments under different scenarios with various flow velocities, tracer states, illumination conditions, and resolutions, a high quality database has been constructed. Specifically, in order to collect videos with different resolutions and perspectives, three sets of identical experimental devices were built, each equipped with two cameras, which were used to collect videos of the top view and side view of the open channel. In the open channel pipeline experimental section, a water flow regulating device for the open channel pipeline was designed to adjust the flow to obtain a stable flow velocity within a specific interval, thereby controlling the flow velocities; the illuminaion conditions were controlled by building a light-shielding shed and using high-power incandescent lamps as alternative light sources; a funnel was placed upstream of the open channel pipeline, and foam balls were sown as tracers to control tracer states. When measuring real data, a millimeter water level gauge is installed on the inner wall of the open channel to accurately measure the water level data. After the flow velocity in the open channel is stable, the center flow velocity of the pipeline is measured using a Pitot tube and a radar tester, and the flow velocity data is obtained through cross-validation. In the camera calibration phase, a series of GRPs were marked with black tape every 200 mm on the outer edge of the open channel. From the collected video, several frames with high definition and clearly visible GRPs were selected as calibration images, and the pixel coordinates of the GRPs were extracted and used as reference points for subsequent calculation of camera parameters. Using the correspondence between the known coordinates of the GRPs in the three-dimensional world and their two-dimensional coordinates in the image, the Zhang calibration method was used to estimate the internal camera parameters such as focal length and principal point position, and then the image distortion was corrected in combination with the polynomial distortion model. This new open-source dataset with accurate data has high applicability to train and validate advanced data-driven image velocimetry.

ABSTRACT. The aim of digital twins (DTs) is to solve many tasks - which have been done empirically in the past, or with a high level of expertise or experience - more intuitively, quickly, and reliably using a digital world. The use of numerical simulations to run scenarios for managing large dams and river basins is now a routine. However, in the context of future projection and coupled soil-atmosphere simulations, the current ability of far-field models to incorporate water needs and, above all, to represent them accurately, is not sufficient. Far-field hydrological models pose considerable demands in terms of memory allocation and CPU time, particularly when assessment of modelling uncertainty is required. The water management sector, like others, is experiencing a wave of digital innovation. Despite the numerous examples of DTs in various fields, it is worth noting the scarcity of DT applications in the specific field of water resources management. While various initiatives have been taken to create and test more efficient platforms using advanced digital technologies, few attempts have been made to apply DT technology to the whole of a large dam or river basin or a large lake, for digital management of water resources and decision-making based on abundant and diverse data.

ABSTRACT. As extreme weather and human activities continue to strengthen, strong interference factors have a significant impact on the formation of floods in the Haihe River Basin. Evolution mechanism of runoff generation and concentration in Haihe River Basin under the unsteady hydrological pattern. The unsteady hydrological sequence caused by climate change and human activities is an important characterization of the change of runoff generation and concentration mechanism in the basin, which is also the core factor that hinders the improvement of flood forecasting accuracy in the basin. The flood forecasting of Haihe River Basin is difficult, and the underlying surface of the basin is complex. It is difficult for a single model to achieve accurate simulation of complex systems. For the first time, a 6D matrix hydro-model framework integrating 'time, space, business, model, parameter and boundary' is proposed, which realizes the nested parallel computing of 'scheme-model-parameter-boundary' in two dimensions of time and space. It focuses on the distributed storage and infiltration spatio-temporal dynamic runoff generation and confluence and flood evolution model with the characteristics of Haihe River Basin, and the simulation efficiency of million-level water conservancy elements is improved from ' hour level ' to ' minute level '. The algorithm optimization based on artificial intelligence interaction, the agile combination construction of heterogeneous models and the parameter adaptive method are proposed. A multi-paradigm flood forecasting and dispatching matrix 'Haihe model' is created to match the flood control business needs of the Haihe River Basin, give full play to the advantages of multi-model joint forecasting, and further improve the accuracy of simulation and forecasting. The coupling model is constructed from four aspects of rain, water, industry and disaster, and is based on geographic information system ( GIS ) platform and special database. After the development and debugging of each model, the model coupling and feedback are carried out, and finally the development and debugging of the multi-dimensional model are realized. This model has achieved breakthroughs in many key technologies such as fine precipitation forecast, flood forecasting and dispatching, and decision support platform. It has been fully applied in the '23.7' catastrophic flood prevention process, providing key technical support for flood control decision-making in Beijing, Tianjin and Hebei Province. It has achieved remarkable social and economic benefits and has broad application and promotion prospects, which provides demonstration application reference for flood control and disaster reduction in other basins.

ABSTRACT. In recent years, the frequency of extreme rainfall events obviously increased which often lead serious flood disasters. In order to dynamically trace the flood process, the satellite monitoring integrated with numerical modelling has been considered as one of new operational approaches applied for supporting real-time decision-making. Taking the example of “23•7” catastrophic flood disaster in Haihe River Basin, the comprehensive application of real-time satellite monitoring and hydrodynamic modelling are presented and discussed in 8 flooded detention areas. For instance, in the Yongding River flood detention area, the monitored maximum flooded area is 154.75 km2 with 5.88 % difference to the simulated maximum flooded area (163.85 km2). The main reason caused that difference is the high stem crops in this detention area. The real-time satellite monitoring data shows obvious advantages in flood detection but need to be improved for some special land use types like crops. Besides that, all 8 flood detention areas have been well monitored and represented in satellite images that validated by the modelling simulation. The work presented in this study is already applied in the flood defense process and promoted to other basins with improvement for the decision-making support in the flood management.

ABSTRACT. In recent years, as a result of climate change, the occurrence of extreme rainfall event shows an obvious increasing trend. Moreover, the urbanization and other human activities have strongly modified the natural underlying condition which often leads to catastrophic flood characterized with high flow velocity and serious economic damages. For instance, the “7·20” rainstorm in Zhengzhou caused serious urban flood, flash flood and landslides. The “23·7” catastrophic flood in Haihe river basin seriously threatened the safety of people's lives and properties. These extreme events address new challenges to watershed management. The progress of digital technologies brings new opportunities for improving the management capacity of the watershed. Since 2021, the China Ministry of Water Resources (MWR) has promoted the construction of digital twin watershed in China. For the flood defense, the China MWR requests the local decision-maker to have integrated flood management based on the catchment scale and to set up the system consisted with “4 advance” process including forecast, early warning, rehearsal and plan. After several years of discovering and practicing, the concepts, strategies and technologies of implementation of digital twin on large catchment scale becomes more and more mature. Many applications within watershed scale have been set up and operationally working in the catchment management. By systematically reviewing the digital twin applications in 7 large catchments of China, this study has presented the achievements of China in last several years: with the wide application, flood disaster loss has been significantly reduced under more intensive flood. Expectation and challenges of the future digital twin development in China are also given.

ABSTRACT. Floods are among the most frequent and devastating natural disasters, significantly impacting urban areas, causing economic losses, infrastructure damage, and displacement of populations. Rapid urbanization, ineffective drainage systems, and unregulated river flows exacerbate flood risks, particularly in major cities near large river basins. In Gujarat, major river systems such as the Sabarmati (approx. 29,000 km²), Narmada (approx. 96,000 km²), Tapi (approx. 63,000 km²), and Vishwamitri (approx. 5,000 km²) pose recurrent flood threats, affecting densely populated regions and critical infrastructure. Assessing flood risk at the micro-watershed level is essential for effective disaster preparedness and mitigation. This study aims to develop a comprehensive flood risk assessment and mapping framework by analyzing 50 micro-watersheds in the Sabarmati basin, 150 in the Narmada basin, and others across the Tapi and Vishwamitri rivers. The study employs AHP (Analytical Hierarchy Process) and MCDM (Multi-Criteria Decision-Making) techniques to systematically evaluate flood risk factors are employed within a Geographic Information System (GIS) environment to integrate key hazard parameters, including soil characteristics, terrain elevation, gradient, water accumulation patterns, and precipitation levels, are analyzed alongside vulnerability aspects such as land utilization patterns, population density, and accessibility to healthcare services. Using satellite-derived data from Google Earth Engine (GEE), each factor is assigned a weighted priority to rank flood-prone micro-watersheds. The results reveal that a significant portion of the study area falls under high to very high flood risk zones, with 20 micro-watersheds categorized as highly susceptible. The generated flood risk maps provide critical insights for policymakers and urban planners to develop targeted flood mitigation strategies, enhance early warning systems, and implement sustainable land-use planning. Aligning with Sustainable Development Goals (SDGs), particularly SDG 11 (Sustainable Cities and Communities) and SDG 13 (Climate Action), this study contributes to building resilient urban landscapes and reducing flood vulnerability, ensuring sustainable and adaptive flood risk management in Gujarat.

ABSTRACT. According to the World Meteorological Association, floods are the deadliest natural hazards, with more than $ 40 billion of dollars of damage each year and 2.5 billion people affected worldwide between 1994 and 2013. Different approaches, such as advanced or simplified numerical methods, can be adopted to simulate these events. Cutting-edge hydrodynamic models, such as those based on 2D shallow water equations, offer fine-scale insights but are time and computationally demanding, while rising climate variability and increased extreme weather events have amplified the urgency for fast flood prediction models capable of supporting rapid decision-making. Here, a simplified model that capitalizes on digital elevation models (DEMs) and simplified flood approximations is tested in different case studies within the Sardinia Region (Italy). The results obtained from the models are compared with official hazard maps provided by the regional agency within the Flood Directive (2007/60/EC) framework, and corresponding to three return periods. In order to quantify model performance and assess predictive accuracy, standard indices based on contingency tables (such as Hit Rate (HR), False Alarm Ratio (FAR) and Critical Success Index (CSI)) are computed. Results are promising and support the feasibility of the model to support initial flood hazard assessment and planning phases, such as identifying high-risk zones and evaluating potential mitigation measures, also in data-scarce regions.

ABSTRACT. Enhancing Urban Flood Risk Management through UAV-Aided Hydrodynamic Modeling and Geospatial Data Integration This study focuses on improving urban flood risk management by integrating UAV-assisted hydrodynamic modeling with geospatial data. By leveraging open-source datasets, satellite imagery, and climate information, the research employs advanced geospatial techniques, including GIS, remote sensing, and Synthetic Aperture Radar (SAR) data, to accurately map urban waterlogging. The study evaluates the effectiveness of UAV-based photogrammetry techniques, such as Digital Surface Models (DSM) and Digital Terrain Models (DTM), in enhancing hydrodynamic flood simulations. Additionally, it explores the role of utility networks—including stormwater drainage and sewer systems—in managing surface runoff and mitigating urban flooding. The research provides a comprehensive review of existing flood management frameworks, including standard operating procedures (SOPs) and response mechanisms. A temporal analysis of Ahmedabad, Gujarat, India, will be conducted to understand historical flood patterns and identify key vulnerabilities. By integrating high-resolution topographic data with hydrological and meteorological datasets, the study aims to simulate realistic flood event scenarios, enabling more precise flood risk assessments. A critical aspect of the study involves developing a decision matrix to support stakeholders in strengthening flood resilience. This matrix will assist in identifying actionable interventions and formulating both short-term and long-term strategies for flood risk reduction. Short-term measures may include optimizing drainage system performance and improving early warning mechanisms, while long-term strategies will focus on sustainable urban planning, green infrastructure, and resilient water management solutions. The research emphasizes the need for data-driven decision-making by combining UAV-derived topographic models with existing hydrodynamic simulation techniques. The integration of utility infrastructure data will provide insights into network vulnerabilities and potential system failures during extreme weather events. Additionally, the study explores cost-effective approaches to flood mitigation, ensuring optimal investment in urban resilience initiatives. By adopting a multidisciplinary approach, this research aims to bridge the gap between technological advancements and practical urban flood management. It seeks to develop a replicable framework that can be applied to other urban areas facing similar flood risks. The outcomes of this study will contribute to policy recommendations, urban planning guidelines, and improved flood preparedness strategies. Ultimately, this research supports sustainable urban development by enhancing resilience against future waterlogging and flood events, ensuring safer and more adaptive cities.

ABSTRACT. In order to improve the intelligent level of flood control management in small watersheds, this paper takes digital twin technology as the core, constructs a flood control “four-prevention” system integrating forecasting, warning, previewing, and preplanning, and carries out the application validation with the Yonghan River Basin in Huizhou City, Guangdong Province as a typical demonstration area. The system builds a three-dimensional digital twin scene by integrating multi-source monitoring data, adopts deep learning algorithms to optimize the rainfall forecasting model, and couples hydrological and hydrodynamic models to realize the dynamic simulation of flood evolution; based on multi-source data fusion and risk dynamic assessment technology, it establishes a progressive early warning mechanism, develops forward scheduling and inverse deduction functions, and intelligently generates emergency plans and risk avoidance schemes. The application results show that the system significantly improves the early warning capability, realizes the precision, dynamization and intelligence of flood control management in small watersheds, and provides technical references and practical examples for the construction of digital twin flood control system in small watersheds across the country.

ABSTRACT. The Yarlung Zangbo River Basin, characterized by complex hydrological processes due to high-altitude terrain and climate variability, requires accurate water dynamics modeling. This study enhances the Variable Infiltration Capacity (VIC) model by refining its unfrozen water algorithm, which previously struggled to accurately simulate frozen soil hydrology. Using comprehensive datasets—including meteorological records, soil properties, and land cover—the modified model was calibrated and validated for the basin. Results demonstrate significant improvements in runoff generation and soil moisture dynamics simulations, particularly in frozen soil regions. This advancement enhances hydrological understanding in the Yarlung Zangbo Basin and offers a more reliable tool for water resource management and climate change impact assessments in similar high-altitude frozen environments.

ABSTRACT. It is a major issue of concern for international organizations and governmental departments in supporting global climate governance. As floods account for 44% of all natural disasters in recent 20 years, it is of great significance to propose countermeasures by reviewing global extreme flood events in recent years. The International Conference on Flood Management (ICFM) and Flash Flood Program (FFP), have hosted a series of international academic webinars since 2020. Drawing upon the outcomes of the webinars and reviewing the latest flood mitigation strategies of major countries advanced in water management, this paper focuses on the global experience in response to watershed flood, flash flood, urban flood, extreme flood under climate change, and comprehensive mitigation measures. It can be learned that flood control strategies need combine both structural and non-structural measures, and draw upon nature-based solutions that integrate green, blue and gray spaces within basins or cities through multi-purpose and targeted programs, with a view to achieving flood management goals and promoting social, economic and environmental sustainability in a holistic way.

ABSTRACT. Flood modeling plays an essential role in hydraulic risk assessment and management, particularly in the case of dike breaches where the presence of local hydraulic singularities strongly influences flow dynamics. This study analyzes the consequences of the 1990’s dike failure on the Reno River near Malalbergo city in Italy, an event which caused major flooding in the Emilia Romagna agricultural plain, threatening infrastructures and residential areas. This study is distinguished by the detailed integration of complex hydraulic structures, including a side channel, bridges and culverts located in the immediate vicinity of the flood zone. These structures were modeled on the basis of field surveys and measurements to ensure an accurate representation of their hydraulic behavior. By combining field data and numerical simulations, the study shows how these singularities directly influence flood propagation dynamics and flood extent. Through an in-depth comparison of two modeling tools, namely TELEMAC-2D and HEC-RAS, this work highlights differences between the models in terms of flood propagation, water level distribution and the extent of flooded areas. The simulated inundation maps emphasize on the differences between the calculation methods and help to define a more reliable inundation map compared to the one obtained with observations. Thus, the results are sensitive to different modeling parameters, and particularly on the interaction between flows and surrounding hydraulic structures. It underlines the importance of integrating hydraulic singularities into flood simulations to refine accuracy and improve decision-making in crisis management. Therefore, the inherent uncertainties in modeling approaches and the influence of site-specific conditions highlight the need for continuous refinement of methodologies. As flood risks evolve due to climate change and land-use modifications, future studies should explore adaptive modeling approaches and multi-model ensembles to enhance predictive capabilities.

ABSTRACT. Numerous dams worldwide are around 50 years old or older, and therefore, their useful life is nearing its end or has already been exceeded. The behavior of inflow volumes to the reservoirs experiences seasonal variability, which has also been modified due to changes in land type and use in upstream areas. Changes in precipitation events also contribute to variations in runoff, making it necessary to periodically review and update the design floods of the reservoirs, as well as to review the operation of their spillways. In this study, the method of the Institute of Engineering, UNAM, based on an analysis of maximum average flows for different durations, was applied to the updated data of the El Infiernillo dam, Michoacán, Mexico, to calculate its design floods for two seasons of the year and different return periods. Additionally, these floods were routed through the reservoir to evaluate different operation policies for the spillway to review safe operating conditions, as the tunnels of this reservoir are often subject to cavitation phenomena when high flows occur. Open-source software of simple application was used in most of the processes. It was found that for a flood with a return period close to 20 years, these tunnels suffered severe damage in the flood that occurred in 2024, with total discharges of around 6000 m³/s. However, if extractions are proposed well before the peak flow occurs, these discharge effects can be avoided or mitigated. On the other hand, the inflow hydrographs to the La Villita dam, located downstream, are subject to discharges from the El Infiernillo dam spillway, making their review necessary to ensure the safety and proper operation of the last dam in the Balsas River hydroelectric system.

ABSTRACT. The increasing intensity and frequency of extreme events driven by climate change necessitate advanced modeling approaches for accurate flood risk management, effective damage prediction, and the design and implementation of protective measures for vulnerable assets, particularly in urban environments. High-resolution flood modeling in urban settings requires detailed information about the location and characteristics of sewer networks. However, such data are often limited, incomplete, or unavailable in digital formats, especially outside advanced GIS systems. In the absence of detailed sewer network information, bidimensional surface flood models can still be employed. However, neglecting the role of the sewer system can result in significant inaccuracies in estimating flooded areas and surface flow volumes. To address this, simplified approaches have been proposed to approximate sewer system behavior. Two common methodologies include: (i) reducing the rainfall volume by a predetermined fraction or (ii) increasing infiltration into the subsoil (or imposing a non-zero infiltration rate for impermeable surfaces) by a predefined amount. Both approaches introduce substantial simplifications and assumptions, the implications of which depend heavily on the magnitude of the rainfall event. These simplified methodologies and their associated limitations are gaining increasing attention in the scientific community, particularly in the context of climate change, which is expected to exacerbate the severity of local extreme rainfall events. The effect of climate change on extreme rainfall is typically assessed through downscaled Regional Climate Models (RCMs) such as those from the EURO-CORDEX initiative. However, RCMs often underestimate rainfall intensities due to their inability to resolve atmospheric convection processes responsible for the most extreme events. Additionally, their spatial resolution, typically in the order of several kilometers, is mismatched with the fine-scale resolution required for urban flood studies, which commonly rely on Digital Elevation Models with 1–5 m resolutions and rainfall data with hourly or sub-hourly temporal resolutions. This study presents a case study in Naples, Italy, where hydraulic simulations are conducted using the CAFlood software, a 2D Cellular Automata model optimized for urban applications. Rainfall events with varying return periods and a one-hour duration are simulated under different simplified scenarios to approximate urban drainage system behavior. To assess the impacts of climate change, rainfall intensities are derived from a large ensemble of EURO-CORDEX models and a high-resolution Convection-Permitting Model (the only freely available dataset for Italy). The analysis focuses on: (i) the influence of different sewer system approximation methods on flood modeling outcomes; (ii) the effect of climate model resolution and accuracy; and (iii) the combined impact of sewer system performance and climate change in a more realistic urban flooding scenario. This research is sponsored by the “INSPIRING” Project, funded by the Italian Ministry of University and Research (MUR) under the National Program PRIN2020 (Grant No. CUP-I65F21001820006).

ABSTRACT. The Amfilochia Pumped Storage scheme represents Greece's largest investment in energy storage and will play a key role in ensuring a stable and sustainable energy supply for decades to come. This transformative project comprises two upper reservoirs, "Pyrgos" and "Agios-Georgios", with a total installed capacity of 680 MW for generation and 730 MW for pumping. Optimizing the design and performance of critical hydraulic structures is critical to achieving the project's operational objectives. This study uses advanced Computational Fluid Dynamics (CFD) simulations with Flow 3D HYDRO to evaluate the hydraulic performance of the intake/outlet structures for the Pyrgos and Agios Georgios reservoirs under both pumping and generation scenarios. The main objectives of this study are to assess the hydrodynamic behavior of flow, including velocity fields, profiles, and patterns along the intake/outlet structures, analyze flow velocities and characteristics in critical sections, evaluate shear stresses on surrounding surfaces to identify potential erosion zones, and detect and mitigate vortex formation, which could impact operational efficiency. The study aims to evaluate the hydrodynamic behavior of the flow, including velocity fields, profiles, and patterns along the inlet and outlet structures. It analyzes flow velocities and characteristics at critical sections, evaluates shear stresses on surrounding surfaces to identify potential erosion zones, and detects and mitigates vortex formation that could affect operational efficiency. Several simulation scenarios are undergoing to achieve these goals. Complex velocity source matrices were implemented at the input levels to ensure accurate representation of the flow conditions. These matrices, derived from simulations provided by the turbine designers, established precise boundary conditions, increasing the reliability and fidelity of the results. The simulations provide valuable insights into flow dynamics and guidance for structural optimization.

ABSTRACT. This work presents a stabilized Variational Multiscale (VMS) formulation for the two-dimensional Shallow Water Equations (SWE). The classical SWE system often exhibits instability in the presence of sharp gradients or discontinuities, necessitating stabilization techniques to ensure reliable numerical solutions.

In the proposed VMS approach, the solution is decomposed into large and small scales, where the small-scale effects are explicitly modeled and incorporated as stabilization terms in the large-scale problem. The method is implemented within a finite element framework, with stabilization parameters adaptively computed based on local residuals and analyzed through Fourier analysis.

Numerical benchmarks, including dam-break scenarios and steady-state flows over variable topography, validate the proposed method. Results demonstrate enhanced stability and accuracy compared to classical Galerkin methods, particularly in shock propagation and complex flow patterns. By explicitly modeling fine-scale effects, the approach improves numerical robustness, enhances shock capturing, and reduces spurious oscillations, leading to better mass conservation and more physically meaningful solutions in complex hydrodynamic settings.

Beyond theoretical improvements, the method’s flexibility makes it well-suited for real-world applications such as flood modeling and river hydraulics in large-scale domains. Its ability to reduce numerical artifacts and better capture flow discontinuities contributes to more accurate hazard mapping and early warning systems—critical for risk assessment and disaster prevention, including dam failures, storm surges, and river overflows.

ABSTRACT. Over the past decade, European Union governments have encouraged to collect data on the volume and quality of all urban water effluents with a significant environmental impact on receiving aquatic environments. Methods for monitoring these flows require improvements, particularly for combined sewer overflows, which are complex flows that contribute in significant proportion to the degradation of the quality of the receiving waters. The DSM-flux (Device for Stormwater and combined sewer flows Monitoring and the control of pollutant fluxes) is a new pre-calibrated and pre-designed device that guarantees the appropriate hydraulic conditions for measuring discharged flows and volumes as well as the concentrations and mass loads of pollutants carried in suspension by the flow. a relationship allowing to measure the flow rates conveying through the DSM-flux was determined thanks to an experimental study on a small-scale physical model, and then validated for several flow configurations upstream of the device. Whatever the upstream hydraulic conditions are, the relative uncertainties are less than 15% and 2% for the flow rates and volumes studied, respectively, which is equivalent to the uncertainties of the most reliable current methods. The monitoring methodology was validated in situ in a large-scale device installed at the field and operating in real conditions, which shows the robustness of the method. Apart from its performance, from an operational point of view, the device has other advantages compared to other current devices: (i) it is pre-calibrated and can be installed downstream from existing combined sewer overflows, (ii) it can be directly installed through a manhole in the discharge channel, (iii) its hydraulic performance is independent from the flow conditions upstream, and (iv) it is an integrated monitoring device, measuring flow rates and their quality simultaneously, in addition to intercepting a part of the particulate pollutants. CFD simulations were also performed, and the results align with the experimental data.

ABSTRACT. Computational Fluid Dynamics (CFD) has become an essential tool in the design, evaluation, and optimization of hydraulic structures, offering detailed insights into flow behavior, turbulence, and energy dissipation. By simulating complex hydraulic phenomena, CFD helps engineers refine designs, improve efficiency, and mitigate risks, ultimately ensuring the reliability and sustainability of infrastructure. This presentation highlights five case studies where OpenFOAM has been used to address various hydraulic challenges in dam structures. The examples show the diversity of approaches and the parameters to be respected, particularly the numerical ones. The applications concern: (1) a jet behaviour assessment for the Num Phak dam to aid in plunge pool design, (2) a verification of the flow conditions on the Lauch dam spillway after its height increase and the design of the dissipation basin, (3) an optimization of the upper intake geometry at Saut Mortier to reduce head losses and for the lower intake geometry to minimize construction costs, (4) a discharge capacity verification of the flood evacuation system at Artouste dam, and (5) an optimization of the intake design for the future Sibundong dam to mitigate sedimentation risks. The results demonstrate the effectiveness of CFD, sometimes combined with physical modelling, in improving hydraulic performance and supporting decision-making for dam and hydroelectric infrastructures.

ABSTRACT. This study investigates the complex hydrodynamic and thermal interactions at the confluence of the Ebro and Segre Rivers within the Ribarroja Reservoir. Using the Finite Element Method (FEM) within the Kratos Multi-physics framework, the coupled Navier-Stokes and energy equations were numerically solved, with the Variational Multi-scale (VMS) method employed for stabilization. Model validation was performed using field data, including thermal imaging and in situ measurements, collected in March and October 2011. The intricate mixing and stratification dynamics, driven by a combination of regulated and unregulated river flows, seasonal temperature fluctuations, and the reservoir's geometry, are highlighted by the results. Specifically, the model demonstrates its effectiveness in capturing these thermal stratification and mixing processes, showcasing the influence of the interplay between river regulation, natural temperature variations, and the reservoir's shape. The VMS stabilization technique proved crucial for maintaining numerical stability, particularly given the convection-dominated nature of the flow, effectively preventing spurious oscillations and ensuring a robust solution. Furthermore, the staggered coupling approach employed contributed to both numerical stability and flexibility, enabling the efficient solution of the coupled fluid-thermal problem.

While the model successfully captured the overall system behavior, some limitations were identified. Discrepancies between simulated and observed temperature fields were observed, particularly in regions characterized by sharp temperature gradients. These deviations are attributed to the one-way coupled solution of the energy equation, wherein the temperature field is calculated based on the previously solved Navier-Stokes equations. The current model simplifies the representation of buoyancy effects and does not fully couple the energy equation with the Navier-Stokes equations. This simplification, while computationally efficient, potentially limits the accuracy of temperature predictions, especially in areas where buoyancy forces play a significant role in the flow dynamics. Future work will therefore focus on incorporating a fully coupled energy equation directly into the Navier-Stokes equations. This enhancement will allow for a more detailed and accurate solution of the temperature field and is expected to significantly improve the model's predictive capabilities, particularly in capturing the complex feedback mechanisms between fluid flow and thermal dynamics. Ultimately, this research contributes to a deeper understanding of coupled fluid-thermal processes in reservoirs, with relevance for water quality management and ecological conservation.

ABSTRACT. As part of the general overhaul of the ‘Seine Aval’, Europe's largest wastewater treatment plant, setec hydratec, as part of a large consortium, carried out the hydraulic design of the new primary settling (DPSAV) on behalf of the Syndicat d'Interdépartemental pour l'assainissement de l'agglomération Parisienne (SIAAP). This plant, which covers 800 ha, serves 5 million inhabitants and has a dry weather capacity of between 0.5 and 1 Mm3/d. The flow rates in full treatment are increased to 45m3/s in wet weather and 70 m3/s in incident flow rates. Among the challenges of this overhaul: adapting to the existing situation by inserting downstream of the pre-treatments at the head of the plant, and towards several exit points (secondary treatments, bypass); right-of-way and water line constraints that prohibit the construction of normative channels (for self-monitoring measures) and impose hydraulic dimensions adapted to compact geometries. After the first design draft based on the state of the art and a ‘general’ 1D model, a series of models was produced: a detailed 1D wireframe model with Hydra software made it possible to explore all the operating phases. This software, interfaced with QGIS, makes it possible to solve the non-stationary Barré de Saint Venant equations solved implicitly for free surface flows, taking into account singularities by modelling: change of direction, weirs, convergent/divergent, etc. As the plant has to handle large flows with high space constraints, the many singularities of this plant model have been calibrated with 3D unsteady VoF URANS modelling carried out in Fluent: siphon passages coupled with changes of direction to ensure protection of the spillway when operating without water, evaluation of equal distribution in batteries of 10 upward flow screens downstream, control of valve openings leading to 14 settling tanks to define reliable liquid levels, study of long spillways, etc. In addition, a physical model of the entire head of the plant was designed on a 1/12 scale with the aim of modelling passive protection weirs, ensuring the correct geometry of the pumping chamber and finally modelling the disconnection/deviation of the 16 pumps of 2m3/s each and raising the water by about 11m. This physical model made it possible to make the 3D model more reliable on the same scale and then to make a digital twin that could be reused after the destruction of the latter. Finally, 3D modelling made it possible to optimise pressure losses in certain inlets to the plant, design a network of sensors to carry out self-monitoring of discharges into the natural environment on an 82-metre non-standardised weir, and produce a wide variety of feed scenarios to minimise the uncertainties in volume measurement made by height/velocity sensors.

ABSTRACT. Hydraulic modelling is often an essential part of the flooding risk characterization. Improving the hydraulic model remains a subject of active research. This is particularly true for the model calibration stage, which is one of the most important issues for modelers. The calibration process involves adjusting several of the model's internal parameters to obtain simulation results as close as possible to the reference measurements used for the study. In hydraulics, which englobes flooding modelling, the calibration is generally made to the friction coefficient, which represents the roughness of the soil, to reproduce the behaviour measured during reference floods. Iterative manual calibration, a classic approach to this conventional adjustment, leads to an empirical search for the optimum value of the coefficient of friction. Nowadays, numerical optimisation methods such as gradient descent methods have also been developed and can be used to determine the optimum value of input parameters by seeking to minimise the output value of a so-called ‘objective’ function. So, the use of these methods is investigated in order to calibrate a two-dimensional (2D) hydraulic model of the Loire River, for a stretch of approximately 50 km between Gien and Jargeau, upstream of Orléans, implemented using Telemac-2D. The model calibration is made on Strickler coefficients (Ks), based on the 2003 flood and validated on two other floods (2001 and 2008). The first calibration of this model, carried out manually, resulted in the definition of 8 successive sections for the river bed and a single value for the floodplains. This initial calibration is considered here as a baseline for the calibration process. The aim of this work is, firstly , to apply the gradient descent method to automatically calibrate the model to the current definition of the sections, in two ways: a global method in which, for each simulation, all the Strickler coefficients are adjusted at the same time and a sequential method in which the calibration is forced from the most downstream sections to reproduce the manual calibration scheme for fluvial flows. Secondly, the aim is to use these same numerical optimisation methods to propose a new definition of the model's calibration sections. All the results are finally compared with the reference obtained by manual calibration and with in-situ measurements. This comparison focuses not only on the accuracy of the results but also on the operational aspect of the time required for the calibration process.

ABSTRACT. Pesticides are frequently applied to protect vineyard from diseases. The high vineyard slopes favour surface runoff and the washing-off of a fraction of pesticides, which can pollute watercourses or groundwater after infiltration. Nature-based solutions have long been built to control both flood and pollution risks caused by stormwater, especially infiltration ponds and constructed wetlands (vegetated retention ponds). In order to combine the benefits of both types, hybrid retention ponds, composed of both dry and wet compartments, are less common (sedimentation of particles and adsorbed micropollutants, phytodegradation of dissolved micropollutants, infiltration of treated water). Their pollution treatment efficiency is less documented, and even their more complex hydraulic functioning, with different, variable flow rates and bypasses between compartments. The objective of this work is to characterise the hydraulic functioning of such a hybrid retention pond, by computing the water balance of each compartment from measurements. The study site is a 1ha hybrid retention pond collecting surface runoff from a 176ha, half vine-planted, catchment in the Champagne area, Northeast France. The pond is composed of four compartments, respectively for sedimentation, sedimentation and photochemical degradation, phytodegradation and infiltration. The winegrower union provided local meteorological observations. The water levels of the four compartments, the inflow rate and the flow rates between compartments were monitored at 5min time step for five years (2016-2021). High culvert slopes made inflow measurements very difficult. A careful measurement analysis and a full 30min water balance, from the downstream compartment up to the inlets, provided reliable inflow and outflow time series for the different compartments, including infiltration and evapotranspiration. This simple model is based on hydraulic structure laws between compartments. We then computed the catchment runoff coefficient and characterised the pond hydraulic functioning pattern for each of the 85 fillings, identified by at least 5cm water height in the first compartment. The results show highly variable but lower catchment runoff coefficients than expected: 2% in summer, 3% in winter. Even a rain of 500-year return period did not exceed the pond capacity, which seems to be oversized. Even if infiltration in the first, non-waterproof, compartment was neglected during the design due to its short retention time, results show a considerable infiltration through the compartment slopes, around 45% of the total inflow. Moreover, due to a short retention time in the second compartment, around 20% of the total inflow bypasses the third compartment by overflow into the last compartment. Consequently, only around 20% of the total inflow follows the whole treatment process, which should limit the treatment efficiency of the pond. The presentation will also deal with the influence of macrophytes and permanent pools on the hydraulic retention time in the different compartments, since longer retention times favour sedimentation and micropollutant degradation.