ABSTRACT. In the week of July 12, 2021, Wallonia experienced record levels of rainfall, leading to exceptional flooding and causing considerable loss of life and damage to property, furnishings, and personal effects. Citizens, businesses, and public authorities were severely affected. From the very first hours of the crisis, questions were raised about the quality of the information circulating, the decisions taken and the resilience of infrastructures. The very next day, questions were put to the public authorities about the predictability of the phenomenon and the prevention of the consequences of flooding. The study carried out by Gruner and published in autumn 2021 first established that the July 2021 event could be considered rare to very rare, both in terms of the intensity and geographical distribution of precipitation and the hydrology of the rivers. The flows recorded on the Vesdre exceed the 100-year return period. In terms of precipitation, it rained almost twice as much as the maximum recorded in the available historical series, and 4 peaks were observed between July 13 and 14. The analysis showed, among other things, that the Eupen dam played its role in cushioning the flood, and that operations were carried out in accordance with handling instructions. Without the presence of the dam, the situation downstream would have been even more dire, but the velocity and relative unexpectedness of the flood did not lead the managers to create an additional holding reserve in the reservoir through preventive releases, which, in any case, would not be warranted by the operating instructions of the dam. The situation has highlighted the importance of reservoirs in terms of their rainwater storage functions. Reservoir dams also play a major role in ensuring the production of drinking water in Wallonia, a role that is becoming more important in the context of climate change, with increasingly frequent droughts. Analysis of the July 2021 event has emphasized the need to optimize the management of dams under public ownership, involving a trade-off or weighing of interests between the risk of flooding and other issues (low flow/drought, drinking water, energy production, etc.). It was with this background that the public authorities launched a study aimed at reorienting and optimizing existing methods and tools by moving towards dynamic management of reservoir dams, integrating data on inflows, outflows, past trends and short- and medium-term forecasts into the models. This paper will cover the events of July 2021 and illustrate a proposed way forward for operational management of reservoirs in Wallonia. The proposed system integrates innovative artificial intelligence techniques and advanced hydraulic models to provide reliable probabilistic forecasts of inflows in a context of climatic and meteorological change.

ABSTRACT. On the evening of october 3, 2015, the southern coast of the Alpes Maritimes region was hit by a particularly intense and rapid flash flood episode. The municipalities of Biot, Antibes, Cannes, Mandelieu (non-exhaustive list) were extensively flooded and suffered significant damage. The death toll form these floods reached 20, with substantial material and economic losses (approximately 600 million euros in insured damages according to FFSA). The three main rivers that caused these floods were the Brague (70 km²), the Grande Fraère (22 km²) and the Riou de l'Argentière (48 km²). Among these rivers, only the Brague had a hydrometric station, which was damaged during the flood and did not provide a complete record. This flood event was studied using hydrological approaches or 2D hydraulic modeling tools. Here, we propose to model and simulate this event in the downstream area of the Brague river, using a 1D-2D approach enabled by the HEC-RAS modeling software.

ABSTRACT. The Brague watershed (Alpes-Maritimes) is recurrently subjected to flash floods known as "cévénoles", which result in flooding and damages. In October 2015, the municipalities of Biot and Antibes were particularly severely affected. The hydrometric station of the Brague river was damaged during the flood and did not provide a complete record. This flood event was previously studied using hydrological approaches or 2D hydraulic modeling tools with adaptive mesh refinement. Here, we propose to model and simulate this event in the downstream area of the Brague, based on a 2D aproach on an unstructured mesh using the TELEMAC modeling software. A special attention is dedicated to hydraulics structures such as bridges and culverts, which are showed to impact the flood dynamics.

ABSTRACT. In the past and now, flooding has been a serious problem, causing significant property damage and a high number of fatalities despite the extensive risk mitigation measures and billions invested in flood defenses globally. The Loire River, which is the longest in France, has a considerable history of flooding. In order to prevent floods, levees have been built since the Middle Ages. For this study, a 2D hydraulic model was built using HEC-RAS. The model represents a 50-kilometer section of the river that goes from Gien to Jargeau. Considering that the IRSN had previously developed a numerical model of the study area with TELEMAC-2D, the HEC-RAS model was created according to its data and assumptions. The aim was to validate the model and assess the performance of both software by comparing the results obtained, including flood areas, water levels, velocities, and computational time. The numerical model has been calibrated for the largest flood event recorded, which occurred in 2003, and validated on two other major floods. This enables to estimate the flooding's effects focusing in particular on the simulation of a 1,000-year return period flood and to incorporate the analysis of multiple levee breach scenarios. Furthermore, the results obtained with the hydrodynamic models implemented using HEC-RAS and the existing TELEMAC-2D model were summarized including the performance, capabilities and limitations of the flood study, but also of the levee breaches analysis that test various breach parameters and modes of rupture. This analysis provides a solid basis of criteria, to select in further studies, the more suitable tool according with the objectives of each project.

ABSTRACT. Seine estuary experiences large tidal ranges (e.g., around 7 m during spring tide) and a long high-tide slack time (roughly three hours). As an embayed estuary located in the English Channel, its dynamics is highly affected by the tidal transformation and mass/momentum exchanges between the English Channel and Seine Bay. However, these mechanisms are not well studied yet. In this investigation, a Telemac2D hydrodynamic model of Seine Bay and Estuary (abbreviated as SBE in this paper) is developed and calibrated for both moderate and extreme conditions. First, the main parameters affecting the tidal dynamics of SBE are identified. Then, the model is enhanced to correct the simulation of tidal propagation in Seine Bay in moderate conditions, thereby reproducing the length and shape of the high-tide slack in the Seine estuary as observed by the in-situ measurements. The enhancement procedure is further confirmed by applying it to another marine-boundary-condition type (i.e., time series of total water level as well as harmonic constituents). Moreover, the results in the bay are validated using remotely-sensed, altimetry data. Then, the model is adapted to simulate the storm-surge events in SBE, focusing on the Johanna storm which caused considerable damage and casualties in March 2008. The calibrated model is used to investigate the tidal dynamics as well as the effects of meteorlogical forces in SBE and can be a useful tool to further study the hydro-eco-morphology of the SBE. This work aims at providing a reliable map of surface water evolution throughout the SBE in stormy conditions. The model will be used in the Cal/Val phase of the SWOT (Surface Water and Ocean Topography) mission to calibrate and validate the SWOT simulator.

ABSTRACT. Floods are, in general, a natural phenomenon that might cause serious economic, social and environmental damage. However, floods can also be induced by anthropic actions, such as improper operation or terrorism on hydraulic structures such as off-stream reservoirs, dams, irrigations channels, levees, etc. In some cases, the flood risk becomes greater due to the increase in vulnerability and exposure when these structures are placed near of populated areas or essential services that might cause significant material or environmental damage. Off-stream reservoirs are human-made structures that can storage of any kind of fluid, either Newtonian or non-Newtonian. They are commonly located outside the basin’s channel network and can be totally or partially delimited by a retention dyke or excavated in the natural terrain. The hydraulic studies that consider the break of the dyke of the structure, and the subsequent flood wave propagation, are utilised for the inundation extent mapping. In such cases, one of the main issues for practitioners is the selection of the break point, which is not trivial in off-stream reservoirs for several reasons: a) because they are not directly associated to a river bed; b) they might be built with an irregular geometry; and c) they can be placed on a highly modified terrain (smoothing slopes, changing the natural drainage network, conforming terraces, etc.). As a result, the flood wave propagation process is complex and, at first instance, not easily determinable. The deterministic approach is traditionally applied by manually selecting one break point per scenario. However, this kind of analysis could lead to an underestimation of the maximum flood extent because the break point that would cause the worst consequences is generally unknown. By contrast, a stochastic definition of the break point, together with an automatic break point definition and a probabilistic assessment of basic hydraulic variables (e.g. water depth), allows for the direct determination of the envelope of the flood extent. In this case, from a purely hydraulic point of view, the direct determination of the global envelope of the flood extent is obtained, the worst flood scenario being already included in this probabilistic approach.

ABSTRACT. In global sensitivity analysis, ubiquitous Sobol indices are used to quantify the effect of the variation of some input parameters of a model on the variance of the model outputs at each time-step and at each spatial point of the model. This paper introduces the coupling of the 1D numerical solver for transient open-channel flows Mage with Melissa, a framework for large scale in-transit sensitivity analysis. This framework is fault tolerant and can manage the specific cases in which some sets of parameters can lead to a divergence in the numerical solver that causes an early stop of the simulation. In addition, as the Sobol index estimators are computed iteratively and in-transit, the analysis is run in parallel to the simulations and results are available as soon as the last simulation ends. Moreover, as the simulation results are not stored on the file-system, it allows to run large scale sensitivity analysis without extra cost on disk usage and read/write overhead. The Lower Seine River, a tide river, is being used to demonstrate the quantification of the influence of a set of Strickler coefficients on the water elevation and the flow rate over a tidal cycle. The Lower Seine River model is segmented into 14 zones, each with independent constant Strickler coefficients. Tidal transient models are challenging to calibrate because the Strickler coefficients of a single zone influences the water elevation and the flow rate in different cross-sections at different time-steps of the model. In our study, the Strickler coefficient values are sampled from a uniform distribution using a Monte Carlo sampling method, the corresponding simulations are run on a desktop computer and the results are aggregated iteratively and in-transit by the Melissa server to generate ubiquitous Sobol index maps. Melissa provides first order and total order Sobol indices, as well as their 95% confidence interval. The Sobol indices are useful to identify the most influential Strickler coefficients for each zone at each time-step, leading to a better understanding of the model. The modeler can then adjust the model by either defining more relevant zones for the Strickler coefficients, or fine-tuning the most influential coefficients to fit the model results to field data, with setting the other ones once and for all. This approach can be applied to any other uncertain input parameter.

ABSTRACT. To overcome the challenge of flooding, urban planning must consider innovative and simple strategies to achieve resilient and sustainable cities. The vulnerability assessment is considered key to reduce flood risk, and the most widespread methodology for assessing vulnerability is the creation of an index. However, the indices that have been obtained so far are fragmented or follow a very complicated methodology making its interpretation arduous for stakeholders, limiting theirs operational uses for decision making. This document addresses these gaps and proposes a methodology by developing a multidimensional index based on indicators that take into account five dimensions: social, economic, physical, environmental, and institutional. This method uses easy-to-use interactive maps to assess flooding and multidimensional vulnerability. These maps help decision makers understand and effectively address these challenges by allowing them to intuitively visualize risks and assess vulnerability in different areas. This index is comprehensive and rigorous, as well as complementary in its ease of use and application for decision makers in urban planning.

The proposed methodology allows us to identify the sources of vulnerability to flooding and to formulate strategies to reduce risk and promote urban resilience, enabling us to face present and future challenges in urban planning. By its simplicity of construction and interpretation, the added value of this vulnerability index is to allow to comprehensively support adaptation and mitigation strategies implementation in each of the single evaluated dimension.

ABSTRACT. The use of free software as a support tool for engineering problems has been increasing for just over two decades; free software from internet sites of recognized international institutions are frequently used in simulation of pressure or free-surface conduits, rainfall runoff or floodplain and urban drainage modeling. The graphical representation of the result of the different analyzes is usually done using spreadsheets and with computer programs such as visual basic, matlab, and python. Artificial intelligence (AI) has presented a revolution in recent months as a tool to help numerous professions with some cases of inaccuracies in their results . Openai's chatgpt has released new versions in a short time and new chats with AI elements ; Bing AI, Google's bard, among others, seek to compete in information reporting . AI tools can play an important role in providing support in solving practical problems. First, AI can be used to analyze large volumes of hydrological data, such as precipitation records, water levels, and river flows. This data can be processed and analyzed more efficiently by AI algorithms, allowing the identification of patterns, trends and anomalies that can help to better understand the hydrological behavior of a basin or water system. Furthermore, AI can be used to develop more accurate predictive models for water resources management. By training machine learning algorithms with historical data, AI can learn to predict water availability, river flow, or the behavior of hydrological systems in different scenarios. Regarding specific AI tools, such as Bing AI and Google Bard, they may be using advanced AI algorithms to improve efficiency in finding and presenting relevant information in the field of hydraulic and hydrological engineering. This research highlights the use of the free software Gnuplot for the elaboration of data graphs whose interpretation helps to define solutions in practical cases. The cases of the results of a reservoir flood routing are presented to review a discharge policy of a dam; the graphic representation of a storm hyetograph and the runoff produced; the representation of graphs of the historical series of average daily maximum annual expenses for a day duration and its theoretical distribution function of the Gumbel type and the representation of a digital model of elevations, the latter is also drawn from a program generated by Bing AI chat. Examples of use of the free software R for statistical data analysis are also presented. In this case, the program that was developed by the Bing AI chat must be polished more to give a higher quality 3d surface.

ABSTRACT. Discharge prediction is a crucial aspect of water resources management and flood forecasting, requiring accurate modelling of the complex dynamics of hydrological processes. Several AI models like Convolutional Neural Networks (CNNs), Artificial Neural Networks (ANNs), have shown promises in capturing spatial and temporal dependencies, which can be advantageous for mimicking complex hydrological processes. Moreover, the Long Short-Term Memory (LSTM) model is particularly promising in rainfall-runoff modelling due to its recurrent nature and ability to handle sequential data. Its proficiency in capturing long-term dependencies makes it a popular choice in hydrological modelling. While the LSTM model has demonstrated promise in this domain, its performance in relation to the length of the training data remains poorly explored.

This study aims to explore the influence of the learning dataset size and the hydrometeorological typology on the performance of the prediction accuracy. In this respect, two experiments were conducted respectively using one and three consecutive years for model training. The purpose is to analyse whether the type of year (i.e., wet, dry, standard) used for the training influences the LSTM model prediction performance. To classify the years based on the hydrometeorological characteristics, a clustering of the dataset was made, based on the annual mean and standard deviation of daily rainfall.

The study is carried out using the CAMELS-GB dataset at Saxons Lode on the Severn River (UK), spanning from 1975 to 2015. The LSTM model is first trained on each year independently, and then validated on the entire dataset via the sequential computation of the Nash-Sutcliffe Efficiency (NSE) for each validation year.

The NSE of discharge predictions have been grouped together in a matrix containing each training and validation years in column/row respectively. This heatmap based on the NSE values visually revealed heterogeneous and unsymmetrical patterns. Notably, the years preceding or following high water events (2009, 2006, 2013) showed poor performance only in the training phase, while the model in year 2004 performed poorly in the validation phase. When training the model with only one year, 23.03% of the NSE values are above 0.7, indicating a relatively good agreement between predicted and observed discharge. However, 84.42%, of the NSE values exceed 0.7 when training the model with three consecutive years. Furthermore, the model's performance varies based on the typology of the dataset, particularly in distinguishing humid and dry years.

In conclusion, this study enhances our understanding of the LSTM model's sensitivity to different training periods. The findings of this research contribute to optimizing the duration of the training period and improving the accuracy of discharge predictions in hydrological forecasting models. Future experiments are to be conducted to characterise for example the influence of the choice of training years among the three clusters.

ABSTRACT. In both Hydraulics and Fluid Mechanics, the transition to turbulence and the identification of stable coherent structures are essential to identify and control the mechanism of flow development and its properties. Free-surface flows can often show one particular kind of coherent structure in the form of the so-called roll wave instability, identified as long waves propagating downstream at a constant velocity on the free surface. Mathematical, numerical, and experimental works have been widely reported in the literature as an attempt to mathematically model the phenomenon and allow the prediction of wave properties as a way of better controlling free-surface flows in mountain regions, surface runoff, and flooded regions. However, still, methods are either time costly, expensive, or inaccurate for practical applications, thus remaining a matter of study by the scientific community. One of the main difficulties arises from the fact that natural environments offer more variability of parameters and properties of the flow due to its natural aspects. In this aspect, machine learning techniques can significantly aid engineers and scientists to detect relevant parameters of the phenomenon and even model it. As an attempt to generate additional mathematical tools to explore the roll-waves problems, the present work brings an evaluation of the performance of Physics-Informed Neural Networks when used to model roll waves for a classical laminar flow of a Newtonian fluid. The objective of this paper is to understand the operation and challenges when using PINNs to model the spatial behavior of roll waves based on a dataset of high-resolution numerical results obtained via OpenFOAM while using a loss function that also employed differential equations from an analytical model of roll waves. To that end, seven different PINNs were defined and trained using a small subset of numerical results for a 2D transient open-channel flow of a Newtonian fluid, in favorable conditions for roll waves generation, while employing a set of Saint-Venant-like differential equations that can be used to model the roll waves phenomenon. Then, the PINNs were fed the whole domain points and predicted the wave interface heights and average streamwise velocities, which were compared to the reference numerical data to estimate the prediction errors. It was found that the PINNs can accurately predict roll wave properties such as frequency and wavelength. The PINNs also predicted wave heights and average velocities with some accuracy but performed poorly in the regions of wave peaks, having an overall better performance predicting flow height than velocity. The performance of the PINNs was also shown to decrease with the increase of the Froude number, an effect attributed to the mathematical hypothesis considered during the theoretical development of the system of equations considered.

ABSTRACT. This article reviews the literature on the application of Machine Learning (ML) to identify flood-prone areas, covering studies published since 2013. The review focuses on data considerations, such as the specifics of the study area and conditioning factors, as well as the ML algorithms used to identify flooding areas. 100 scientific articles were analyzed through a wide scope of geographical areas, ranging from arid to tropical climates and from small catchments to large river basins, to evaluate the influence of geographical features, historical flood occurrences, climatic patterns, urbanization, and data availability on flood susceptibility modeling (FSM). Iran, India, China, and Vietnam are the most frequently studied locations. The slope of the land, topographic wetness index, land use and land cover, rainfall levels and distance to rivers were key conditioning factors in at least 61% of the reviewed articles. Furthermore, the employed ML algorithms can be categorized into various types: statistical, kernel-based, tree-based, Neural Network (NN)-based, ensemble, and hybrid approaches. NN-based models, such as Long Short-Term Memory (LSTM) and Recurrent Neural Networks (RNNs), excel in solving high-dimensional problems but face challenges related to reliability and overfitting. Kernel-based approaches require optimal configuration through a trial-and-error process, while tree-based models offer simplicity and are less prone to overfitting, although they may be less precise. Among these, ensemble and hybrid models generally outperform traditional ML methods, despite their own limitations. These methods primarily focus on event-based historical floods, limiting their ability to make real-time predictions due to the lack of time-series data. Additionally, most models face restrictions given data consistency and validity. They often use inconsistent data, where flood conditions and input parameter values are not aligned in time and space. This discrepancy undermines the models' reliability. Consistent and valid datasets are imperative for accurate model development.

ABSTRACT. More than 40% of Germany´s waterways are impounded rivers. The water level in the im-poundment needs to be adjusted within very narrow tolerances to ensure the safety and the ease of shipping. This task is resolved by the two main components of an impoundment, namely the weir and the hydropower station. Real-time water level control is a short-time (minute-by-minute) challenge that still leaves room for new approaches and technologies. Recent efforts promote the establishment of model predictive control (MPC) on German waterways, like for example at river Moselle. The basic idea behind MPC is to use a prediction of future water levels over a finite time span in order to improve the control input for the impoundment. This paper investigates on using machine learning algorithms to provide the water level predic-tions needed for the MPC. The current state of research shows that classical hydrological water level forecast models can already be outclassed by machine learning algorithms regarding accu-racy, simplicity and computation time. These models usually allow predictions in an hourly reso-lution. The aim of this work is to present the challenges faced to minute-by-minute water level predic-tions. Influences like positive and negative surge waves, as well as the operation of the ship locks disturb the water level and make it difficult to handle for a data-driven machine learning ap-proach. Beside the data pre-processing, preliminary results on using machine learning for water level prediction regarding impounded waterways will be shown and discussed.

ABSTRACT. A methodology of automatic identification of openings for urban flood has been developed within Deufi ANR project about Detailing Urban Flood Impact. That is the combination of two techniques: photogrammetry and image recognition.

Photogrammetry is based on slicing that is a method for finding the openings from a recess in the main plane corresponding to the plane of the facade. The detection of openings by slicing is done by analysing the holes that are formed in the main obtained plane.

The image recognition is based on deep learning which uses artificial intelligence algorithms based on learning such as RCNN. Once trained, the algorithm is used to detect several types of opening and therefore works by classification to identify windows, shutters, windows wells, doors, garage doors or sliding windows.

The last step named cross analysis consists in comparing the results of the two methods to assign a confidence score between 0 and 1 according to the results of the two methods and the architectural rules.

The data come from street video like Google street view. The workflow of the numerical treatment and the details of each task are discussed. The results of the treatment with the dedicated interface of the developed application provide several conclusions.

Among 210 selected buildings in Nimes, the average ratio of detection of openings is 90 % with the combined methods of slicing and deep learning. The identification of opening types has a lower success ratio of 80 %. The detection ratio of openings fall from 90 % to 72 % with the use of cross analysis.

We deduce standard architectural rules for window wells, windows and doors. For window wells, the width and height are around 0,5 m. The sill is around 0,5 m. For windows, the width is around 1 m and height around 2 m. The sill is around 1 m for the ground floor. For doors, the width is around 1 m and the height is around 3 m. The sill is around 0,2 m but can be higher when there are several steps. For windows and doors there is correlation between width and height. There is no correlation between ground level and sill, unfortunately for damages in flood event.

Hydraulic applications in case of homogeneous run-off show that water passes mainly through the doors up to 1 m but the contribution of the windows exceed doors’ one rapidly. The contribution of window wells is also not negligible. The hypothesis of accumulation zone gives the same kind of conclusions but the phenomenon is less strong because the filling of openings is not as simultaneous as in the first hypothesis. The openings are in water successively and more progressively.

ABSTRACT. Man-made soil and water loss have lots of types and very complex itself. Measures combined with remote sensing images, drone data and ground tools are new pattern of supervision used nowdays. Change detection of the disturbed land of the ground surface is essential to the supervised work of soil and land loss. At present, human-machine interaction method are the main way of land extraction with low efficiency and high cost.In ordet to serve the work of man-made soil and water loss supervision caused by production and construction projects, In this paper, high-quality semi-automated sample labeling technology is used to build high-quality labels of key elements and disturbed lands. Deep learning based algorithms based on ground object classification and change detection technology are used to develop intelligent interpretation models based on high-performance of distributed parallel GPU clusters. Transformer framework is also added to optimize and improve the above algorithms. Models and methods are used in some provinces of china, according to the arrangement of bureau of soil and water conservacy of MWR, both in 2022 and 2023. Results showing that the algorithms have strong adaptability and the model accuracy is improved by about 10% compared with traditional one. The supervision efficiency is improved and the overall ability of integrated supervision is improved.

ABSTRACT. Despite ongoing efforts to develop high performance hydrodynamic numerical models accounting for complex flood processes, the main challenge still remains the computer power required to run multiple simulations with a chain of models of increasing resolutions (from a hundred meters for coastal waves and surges to a few meters for marine flooding). A possible approach to overcome this problem is to rely on the statistical analysis of existing databases of results that were pre-calculated using full process high-fidelity numerical simulations using for instance some machine learning methods like neural networks or Gaussian process regression. This approach named “metamodelling” (also referred to as “surrogate”) has shown promising results in multiple coastal contexts. However, one major difficulty is related to the high dimension of the flooding numerical model’s output that generally varies over space; a typical example being a flood map representing the maximum value (over time) of water depth (i.e. difference between the water level and the topographic elevation) as a function of the spatial location. In practice, the locations are discretized, and the output is represented by a high-dimensional vector of length corresponding to the number of discretizations that can typically reach several thousands. An efficient approach is to combine the metamodelling approach with a dimension reduction (DR) technique whose objective is to extract a much smaller number of latent variables to represent the very high-dimensional output. Solving this problem is however complicated by the complexity of the flood maps, which have discontinuous heterogeneous patterns (especially related to land covers) and non-negative and zero inflation data (many cases lead to the absence of flooding). To address these difficulties, we use a set of simulation results of about 200 flood maps computed for the ten districts of Arcachon Lagoon (Gironde, France). On this basis, an in-depth analysis of the influence of modelling choices in the combined DR – kriging metamodeling procedure is carried out in particular regarding two factors: (1) the choice in the DR method, namely the popular principal component analysis PCA or a non-linear DR method based on neural-network-based autoencoders; (2) the parametrization of the DR, in particular the size of the reduced domain (i.e. the number of latent variables), and the architecture of the network (number of layers, number of nodes, type of activation function). Using a cross-validation-based procedure, we show the higher performance of the autoencoders with respect to the reconstruction error of the flood maps (with a reduction by a factor ~10), and to the prediction error (with a reduction by a factor up to 3). The disadvantage, however, is the need for careful optimization of the influence of the autoencoder architecture, particularly with regard to the selection of the activation function.

ABSTRACT. Flood hazard mapping is an important step to address the risks associated with floods. In recent times, large improvement has been seen in accuracy and efficiency of flood mapping through the combination of high resolution DTMs and advanced hydraulics. Machine learning techniques also become widely proposed for flood susceptibility mapping. The present work aims to assess how the flood hazard maps obtained with hydraulic approaches can be retrieved with machine leaning techniques. The presented case study is the “Argens basin” of France. The reference flood hazard maps were obtained through the simulation of water depth for a 1000-year return period, using the FLOODOS 2d hydraulic model at a 5-m resolution. The database for the application for machine learning methods include several geo-environmental factors (GeFs), combining high-resolution topographic datasets (derived from the 5-m lidar Digital Terrain Model (DTM) of; the average annual rainfall on a 1x1km grid for the period of 1997-2022; geology maps and a soil classification dataset for France. Height Above Nearest Drainage (HAND) and the river discharge datasets used for the reference hydraulic modeling have also been prepared and included in the GeFs database. Further, the flood inventory used to apply machine learning was developed by combining the equal number of non-flood locations and the flood locations. The multi-collinearity analysis was used to opt out the GeFs having a collinearity issue. Since the selected region has complicated geographic factors, we divided the flood inventory through random sampling method into training and validation datasets with 70:30 ratio. Advanced artificial neural network, random forest, and XGBoost models of machine learning techniques including frequency ratio, a bivariate model, were trained on the data sets. To validate these models, we have considered the area under the receiver operating characteristic (AUROC), a widely used validation technique, and other accuracy assessment techniques such as kappa index, seed cell area index, accuracy, and precision. The results illustrate the performances that can be achieved with machine learning for flood mapping in areas with complicated geo-topographic factors such as the Argens basin.

ABSTRACT. The Mediterranean coastal city of Cagnes-sur-Mer is increasingly threatened by flooding due to the increased frequency of extreme weather caused by global climate change. Optimization and modification of certain infrastructural elements such as bridges can effectively prevent floods. In this study, a 2D hydrodynamic model based on MIKE 21 Flexible Mesh (FM) was utilized to simulate the effects of a small bridge removal on the flood dynamics within the downstream urban area of the Cagne catchment. The Digital Elevation Model (DEM) was sourced from a high-resolution 25cm DEM. The model domain mesh was constructed using two different resolutions: a 1m resolution for the riverbed and a 5m resolution for the floodplain. The small bridge was represented using the long culvert function. Input data for the hydraulic model was generated by a deterministic hydrologic model based on MIKE SHE model and which encompasses the entire Cagne catchment. Calibration of the model was carried out using data from the October 2020 event, and validation was achieved with data from the November 2021 event at the Cagne Aval observation station. The model was run to simulate water levels both before and after the removal of the small bridge. A reduction in flood risk was noted upon removal of the bridge, as demonstrated by the changes in simulated water levels. The findings of this study underscore the potential of using a 2D hydrodynamic model for accurate hydraulic simulations. Furthermore, the results suggest that strategic modifications in infrastructures, such as bridge removal, can reduce urban flood risk. Future work should expand on these findings by exploring the effects of other structural modifications on flood dynamics in diverse hydrological conditions.

ABSTRACT. With the development of computer science, the digital technologies are widely applied in many tradition industries. For the water industry, especially for the flash flood defense, the ministry of water resources in China proposed “four-pre” approach for guiding the flash flood defense in China. The core of the “four-pre” strategy consists by four components: forecast, early warning, rehearsal and contingency plan. By using the rainfall forecast with different pre-periods, several hydrological/hydraulic models are built according to the user requests. By comparing with the warning threshold values, the model integrated in the system is able to determine the future progress under natural and artificial condition. The system will send the early warning information to the stakeholders and show the results to the decision makers, to let all of them aware the future condition. Then based on the rehearsal results, the system will produce a contingency plan for the system user to help them optimize their working process. This paper presents the “four-pre” approach with one application in Yunnan province in China. The system has covered 20 small –scale catchments (less than 200km2) at the mountainous area in Yunnan province. By clearly understanding the manager’s requests and the difficult in the flash flood defense in those places, the system was designed with an operational framework consisted by two kinds of models and many digital twin technologies such as BIM and 3D simulation. The system is able to represent the flash flood process. And with the forecast rainfall, the system can automatically send the early warning information to the stakeholder to information to escape from their home when the flash flood may happen in their places. Moreover, by using the system, the view of decision makers is also improved by digital twin representation of the flooding impacts on the rural and urban area. The system is current used in the Yunnan province and continuously testing by the flood managers. The “four-pre” approach of flash flood defense in China could also be applied in other countries who suffered the flash flood disaster every year.

ABSTRACT. The famous Guilin scenery is located on Lijiang River with 164km river length and nice landscape along the river. The Lijiang River is characterized with sudden flow fluctuations and long flood season. In order to improve the digital ability of this river and increase the capacity of flood defense of this catchment, the Chinese government decided to launch the project of digital Lijiang River basin. By utilizing the new digital technologies such as cloud computing, big data, digital mapping, digital twins, and artificial intelligence, this project plan to realize the representation of physical watersheds. Integrated considering the current condition and characteristics of the Lijiang River, three kinds of model including the hydrological forecast model, hydraulic computation model and reservoir operation model have been set up in this system in order to support the real-time decision-making process. The design of the system framework has considered the extensibility which is able to apply in province scale to integrate more catchment sub-systems in the future. By validating the system with one passed flood disaster happened on 18th June 2022, the system is able to produce accurate simulation with different pre-period rainfall and to optimize the reservoir operations according to different purposes. The measure made by the decision makers can be simulated by the hydraulic models and let the user clearly understand their decision’s impacts on the catchment under the forecast condition. The successful experience obtain from this system can be reference in other system design in flood management in China or any other countries.

ABSTRACT. Visualization plays a crucial role in assessing the risk associated with rare events such as coastal or river floodings. Typically, political authorities base their decisions on flood risk management using flood maps generated through hydraulic simulations and probability studies. Traditionally, these probability studies focus on the exceedance of a given water level at specific locations, making it challenging to connect them to flood map analyses. To facilitate the interpretation of the overall probabilistic distribution of flood maps, a simple yet effective approach involves providing a small set of representative flood maps, each associated with probability. These prototype maps can illustrate the most typical flooding scenarios that may occur in a given area, and represent the probability distribution of the flooding events. The process of transforming a continuous distribution into a discrete one is known as quantization, which becomes particularly challenging when data generation is expensive and critical events are rare, as is the case with extreme natural hazards like flooding. In the context of floodings, each event relies on an expensive-to-evaluate hydraulic simulator. In this study, we adapt Lloyd's algorithm, commonly used for data quantization, to the context of rare and costly observations. We address low-probability events through importance sampling, while utilizing Functional Principal Component Analysis in combination with Gaussian Processes to handle the expensive hydraulic simulations, creating a metamodel specifically designed for spatial outputs. Our case study focuses on a section of the Loire River near Orléans, France, which is protected by levees on both banks and has a history of significant flooding during the 19th century, including historical breaches. To simulate the river flow along a 50 km stretch of the Loire River between Gien and Jargeau, the Institute for Radiation Protection and Nuclear Safety (IRSN) developed a hydraulic model using open-source TELEMAC-2D computational codes. The model incorporates an upstream hydrograph and a downstream calibration curve as boundary conditions, and has been calibrated using well-known flood events by adjusting the roughness coefficients (Strickler coefficients). To simulate breaches, the model considers two parameters: erosion rate and breach initiation parameter, also known as the control level, which represents the water level above the levee. Breaches occur when the water level above the levee reaches the control level. In the end, we vary 8 different input variables: the maxmimum flow rate, the rise time of the flood, four Strickler coefficients and the two breach parameters.

In our work, we have developed the R package "FunQuant" to perform quantization within a probabilistic framework, specifically tailored for rare events. This package provides various pre-built functions to fine-tune the importance sampling density and metamodel hyperparameters, making it readily usable for a wide range of users.

ABSTRACT. The numerical modelling of hydro- and morphodynamics in watercourses is of great importance for both the advancement of scientific understanding of underlying processes and for design in engineering practice. For this purpose, and especially adapted to alpine and subalpine conditions, the modelling software BASEMENT has been developed in the last two decades. The simulation environment is currently composed by different tools, either freeware or open-source. BASEMENT supports the use of multi-core processors or general-purpose graphics processing units (GPGPUs) to increase calculation efficiency. This can speed up simulations many times over, depending on the problem at hand. BASEMENT version 4 consists of two main modules, namely BASEMD (MD stands for Multi Domain) and BASEHPC (HPC stands for High Performance Computing, based on the possibility to use GPGPUs). The BASEMD module contains the three simulation submodules BASEchain (1D open-channel flow), BASEplane (2D open-channel flow), BASEsub (3D subsurface flow) and allows for submodule coupling in different ways. Further, there is the BASEextern submodule for simulation data exchange during runtime. Furthermore, a PID (proportional integral derivative) controller for automatic control of various process variables is implemented. The BASEHPC module contains only the BASEplane submodule which has a simpler spatial discretization but improved performance compared to BASEMD. The software provides a unified simulation workflow and graphic user interface for setting up numerical models. In both modules, the hydrodynamics are calculated based on the shallow water equations and for the simulation of morphodynamics, the bed- and suspended load transport can be calculated. In the present contribution the various application possibilities of the software will be illustrated by means of several examples, i.e. lateral flood diversion, fine sediment deposition on flood plains, progressive dam breaching due to overtopping and tsunami wave propagation. The steadily growing number of users confirms that BASEMENT is a useful tool for the efficient investigation of complex and extensive questions in river engineering and hydraulic engineering.

ABSTRACT. Located in the Alpes-Maritimes (France), the Espartes watershed (1.6 km2) is characterized by a strong artificialization of both its surface and its hydrographic network. Nearly 50% of the surface of the basin is urbanized, and nearly 50% of the length of the hydrographic network is either canalized or artificialized. By decreasing infiltration and increasing flow velocities, such artificialization leads to frequent flooding in different parts of the basin. In this context, decision-makers need effective tools to assess the current flood risk, as well as the impact of potential urban developments on it (e.g. enlarging the diameter of pipes, creating retention basins, or facilitating infiltration). We model the Espartes watershed using MIKE+ (DHI, 2023). MIKE+ solves the 2D and 1D Saint-Venant equations for respectively overland flow and storm water flow in the collection system. First, we calibrate the model on measured discharges, available in November and December 2017. Then, we run simulations with design-rainfall forcings, for the return periods of 10 and 100 years. These simulations allow us to evaluate the risk of flooding under the current state of urban planning. Finally, by implementing different scenarios of urban development into the model, we are able to quantify their influence on the flood risk. The obtained results provide the technical knowledge necessary for decision-making.

ABSTRACT. The Tlautla hydrometric station is located in Jasso, Municipality of Tula, State of Hidalgo, Mexico, 300 meters northwest of the Central Railroad station. The station measures runoff to Tlautla river and is and is very important because it measures runoff towards to Zimapán dam in the Tula river basin. In September 2021, the city of Tula, located in the state of Hidalgo, Mexico, was affected by heavy flooding due to intense rains in the region. These floods caused extensive damage to infrastructure and affected thousands of residents. Torrential rains began on September 7 and continued for several days. The Tula River, which runs through the city, reached critical levels, flooding numerous urban areas. The flooding mainly affected the municipalities of Tula, Tepetitlán and Tetepango. The floods left significant human and material losses as a result. Dozens of people were reported dead and many more were injured or missing. Thousands of homes were damaged or destroyed, and an estimated thousands of people had to be evacuated from their homes. Unfortunately, there is no record of the return period of the flood that occurred in 2021. The return period refers to the time in years that, on average, a flood of a certain magnitude can occur.To estimate the floods in the Tlautla river basin, studies have been carried out using rain-runoff relationships. However, this article proposes the use of the Method of the Institute of Engineering, which is based on an analysis of average daily discharges to obtain design hydrographs with different return periods. The Institute of Engineering Method uses a frequency analysis to determine the average maximum annual flows at different durations. Then, a disaggregation process is carried out at a daily level of these flows and a process of alternating blocks to build the shape of the design hydrograph. This approach makes it possible to compare the magnitude of the flood that occurred in 2021 with the design hydrographs obtained using the Institute of Engineering Method. The disaster in Tula highlighted the need to improve the flood prevention and response infrastructure in the region. Proposals have been put forward to strengthen early warning systems, build hydraulic works, and strengthen prevention measures in flood-prone areas. The floods in Tula 2021 were a tragedy that left a deep mark on the city and its inhabitants. Reconstruction and recovery will take time and effort, but it is hoped that with the support and solidarity of the community, Tula can recover and strengthen itself against future natural disasters.

ABSTRACT. Hydrodynamic models which solve the 2D Shallow Water Equations are considered to be mechanistic simulators, however they still incorporate grey-box parameters which should be calibrated. One of the major limitations until now, except the lack of data, was the computational cost. In our era, parallel coding and the boost of supercomputing made feasible to calibrate the required parameters. In this work, we discuss this potential using the UniCal simulator in the Mandra (Greece) 2017 flood event.

ABSTRACT. The Midwestern United States experienced a prolonged flood event in the spring of 2019 causing significant damage to communities. Reconstruction of such a major flood event can help develop long-term mitigation and resilience strategies. To this end, a tool that can efficiently and accurately simulate large-scale flood inundation dynamics while ensuring a high spatial resolution is desired. We reconstruct this rare flood event using a GPU-accelerated 2D hydrodynamic model, TRITON (https://triton.ornl.gov/) driven by the simulated runoff and streamflow from a calibrated VIC-RAPID hydrologic modeling framework, forced with the NCEP Stage IV hourly precipitation. TRITON’s baseline terrain information is obtained from the National Elevation Dataset at 30 m resolution leading to roughly 3.3 billion grid cells representing 1.37 million km2 of the Missouri River Basin. To initiate TRITON simulations, we utilize long-term climatic mean runoff and streamflow and develop a steady-state channel stage and velocity. We benchmark the simulated flood inundation outputs against high-water marks, US Geological Survey stage data, remote-sensing-based flood maps, and other high-fidelity inundation model simulations. Additionally, we demonstrate the computational efficiency and scalability of our TRITON simulations on the Summit supercomputer at Oak Ridge National Laboratory. Finally, we discuss the challenges of large-scale hydrodynamic inundation modeling and their implications for real-time inundation forecasting.

ABSTRACT. Nowadays, one of the most destructive natural disasters is flooding. Flood events can be mathematically represented by complex systems of equations that require computational solutions. In fact, numerical simulation is a valuable technology for analysing, understanding, and preventing natural hazards, enabling risk-assessment under various hypotheses and scenarios.

Natural geophysical flows, such as those caused by floods, often involve complex multi-physics models. The adoption of two-dimensional modeling and the use of high-resolution Digital Terrain Models support an accurate computation of water depth and velocity both in the channel and in floodplain areas. To further improve the accuracy of flood risk estimation, additional issues that occur during floods should be included in hydraulic risk modeling, such as the transport of sediments, pollutants but also wood, vehicles, bed rocks and boulders. The latter are of particular interest as they can obstruct infrastructures such as bridges, reducing the conveyance capacity and originating inundations that may reach levels expected for higher return period events. In fact, the destructive power of flooding events is related not only to the high velocity of flow, but also to the transport of debris and large materials.

The present work is a first step to study the spatial and temporal evolution of the transport of debris during a flash flood using a Lagrangian model. Material elements are entrained and transported based on the hydrodynamic forces, computed with appropriate coefficients, or following a kinematic approach. This model is driven by an Eulerian model based on the Shallow Water Equations to describe the evolution of the flow. The implementation of these new techniques is part of an ongoing initiative to develop a portable and future-proof open source HPC modeling framework based on physically-based surface hydrodynamics: the SERGHEI (Simulation Environment for Geomorphology, Hydrodynamics and Ecohydrology in Integrated form) model. The Lagrangian Particle Tracking module has been seamlessly integrated into SERGHEI, which empowers researchers and practitioners to comprehensively study the spatial and temporal evolution of debris transport during flash flood events, enhancing the ability to effectively manage and mitigate the impacts of flooding.

The Lagrangian model undergoes initial validation through rigorous testing in academic and laboratory settings, both with and without particle interactions. The intricate aspects of the implementation, specifically designed to handle computations for HPC environments, are described. Subsequently, the module is deployed to analyze a specific flooding event that took place in Central Europe in July 2021. This event tragically led to the loss of more than 180 deaths in Germany alone and inflicted significant economic damage in the order of billions of euros. By leveraging the Lagrangian model within this context, a comprehensive investigation of the event can be conducted, contributing to a deeper understanding of its dynamics and consequences.

ABSTRACT. The advances of data sciences including artificial intelligence and big data is defining how clients' need are met in water consultancy industry and how water-related research is done in academia. In a nutshell, the paradigm is shifting to data and service, where new tools are applied more creatively to address increasingly complex challenges exacerbated by climate change. Modern high-performance hydrodynamic models for problems such as flooding and sediment transport is key part of the tool chain. However, existing hydrodynamic models lack the flexibility and transparency that can be easily integrated into a data analysis toolchain. Moreover, developing complex hydrodynamic models for large scale and industry-level applications requires collaborations from a wide range of organisations, which require an inclusive framework. This work outlines our efforts to build such a platform and create such tools to support both industry and academia to meet their challenges.

ABSTRACT. The design of a hydrodynamic modelling tool called hms++ is detailed, with an emphasis on addressing key concerns for research software. These are identified as high performance on heterogeneous hardware, flexibility and extensibility of the model, ease of use for non-developers, and low maintenance requirements. By employing the linear algebra library ‘Eigen’, equations and solution algorithms in hms++ can be expressed as succinct, readable source code, while compiling to highly optimised operations on a variety of platforms. This serves as the base for a flexible framework for the finite volume method with multiple mesh types, and a solver for the shallow water equations (SWE). This solver is shown to be both accurate and highly scalable, with test cases involving more than a billion cells on 256 compute nodes, using hybrid MPI/OpenMP parallelisation, as well as vectorisation. Finally, a plugin mechanism is described, which loads program extensions at runtime, providing extensibility while minimising the need for modifications to the core code. Several existing plugins are presented, which fulfil functions ranging from implementing additional source terms, streaming data to other programs, or facilitating two-way couplings with other hydrological models

ABSTRACT. In probabilistic flood modelling, uncertainty manifests in frequency of occurrence, or histograms, for quantities of interest, including the flood extent and hazard rating. Such modelling at the field-scale requires the identification of a more efficient alternative to the Standard Monte Carlo (SMC) method that can reproduce comparable output probability distributions with a relatively reduced sample size, including detailed histograms of quantities of interest.

Latin Hypercube Sampling (LHS) is the most evaluated alternative for fluvial floods but yields no considerable sample size reduction. Potentially better alternatives include Adaptive Stratified Sampling (ASS), Quasi Monte Carlo (QMC) and Haar-Wavelet Expansion (HWE), which are yet unevaluated for probabilistic flood modelling. To fulfil this gap, LHS, ASS, QMC and HWE are compared to quantify sample size reduction to reproduce output detailed histograms – for flood extent, and average and maximum hazard rating – while keeping the difference below 10 % to the reference SMC prediction.

The comparison is done for two test cases with two (i.e. inflow discharge and Manning’s coefficient) and three (i.e. including the ground elevation) input random variables, and a real case with five input random variables. With two random input variables, all four alternatives yield sample size reductions, with QMC and HWE considerably outperforming the others; with three and more random input variables, HWE becomes inflexible and LHS underperforms. QMC is therefore the best alternative to SMC to boost sample size reduction for the real case and shall be preferred in probabilistic flood modelling. Accompanying model software is openly available online.

ABSTRACT. Circular economy initiatives like the EC funded iWAYS project (grant agreement: 958274) promote the reuse of waste heat in industrial sites. This presents opportunities and challenges for technological adaptation. Membrane distillation (MD) is a thermally driven process for water treatment that can use waste heat. However, effectively treating complex industrial wastewater requires adapting MD units to achieve reliable and efficient performance. Filament spacers within the MD units play a key role in structural maintenance and flow mixing. CFD simulations can help to characterize, filament spacer configuration impacts on the hydrodynamic of feed and permeate channels, which affects both trans-membrane temperature gradient and membrane fouling control. Here, we performed a CFD study on a direct contact membrane distillation (DCMD) sub-unit with the goal of evaluating impact on robustness and performance of a set of designed filament spacer configurations. The modeled membrane distillation system has an overall length of 200 mm, width of 10 mm and height of 4.1 mm, containing two layers of filaments in each of the feed and permeate channels. The diameter of the filaments was 0.499 mm, and they had a 45◦ degree angle to the flow direction in the channels. Variations of this standard filament configuration were also tested and simulated to optimize their mixing performance. The numerical simulations to approximate in a 3D solution of Navier-stokes equations for steady state conditions were performed using OpenFOAM code. The computational domains were meshed to an overall number of 6 million cells using OpenFOAM snappyHexMesh utility, and finite-volume based simulation relying on the chtMultiRegionFoam solver were executed in parallel over 40 CPU cores. Comparing the CFD analysis of different filaments´ configurations lead to an assessment of an improved spacer structure. The selected configuration is to be 3D printed for laboratory-scale experimental confirmation of the validity of the CFD model and the optimal configuration finding.

ABSTRACT. Correct determination of antecedent moisture conditions (AMC) in a catchment is one of the key elements in effective rainfall-runoff modeling. This aspect is often neglected in hydrological modeling considerations yet has a key impact on model calibration. Typically, simple methods of assigning precipitation to sub-catchments from the nearest meteorological stations are used to estimate AMC. This approach is a considerable simplification, and a significant limiting factor is the availability of a ground-based measurement network and the timing of data availability. Meanwhile, alternative sources of measuring precipitation, such as weather radars and satellites, can also be used to make AMC estimates. In the case of meteorological radar, a significant limitation in mountainous areas is the limited visibility of the radar through mountain ranges and the difficulty in distinguishing, for example, orographic precipitation. In recent years, the use of satellite data, which covers significant areas of the world with its coverage, has become increasingly common. The biggest drawback of these data is the relatively low spatial resolution. However, they are available in areas without a ground-based monitoring network. In the analyses presented in the paper, calculations were based on IMERG GPM satellite data, which is currently one of the world's most widely used sources of satellite precipitation data. The paper attempts to verify how different precipitation data sources (rain gauge data, weather radar, and satellite) contribute to the estimation of AMC moisture conditions. Particular emphasis was placed on the use of satellite data since it may be the future in terms of data acquisition for hydrological models. The analysis area covers a mountainous catchment in southern Poland, where flash floods have been frequent in recent years. The results of the simulations show significant potential for using satellite data to estimate AMC vs. traditionally used rain gauge data.

ABSTRACT. Rainfall Intensity Duration Frequency (IDF) relationships usually play a major role in the designing and building of water infrastructure. The IDF curves describe rainfall intensity as a function of duration for a given return period which are important for the design of storm water drainage systems and hydraulic structures. In the present daily rainfall data for 33 years was collected from state water data centre, Gandhinagar for Morwa raingauge station and analysed for development of different IDF curves using three different frequency methods i.e., Gumbel, Log Pearson III, and Log normal. Moreover, three empirical formula used to estimate the intensity at various duration from daily data for different return periods. The most suitable IDF equation is, i=xt_d^(-y) with coefficient of correlation with 0.99 and also having minimum percentage absolute difference between observed and estimated values of rainfall intensities in present study area. It was also found that intensity of rainfall decreases with increase in rainfall duration. According to the obtained results, one can notice that the intensity of rainfall increased with the increment in return periods, but decreased with the increment in duration. The results showed that rainfall intensity reduced as the duration of the storm increased, and if the return period of the rainfall was large, rainfall of any specific duration showed a higher intensity. This study will be helpful in many design problems related to watershed management, such as runoff disposal and erosion control, it is necessary to know the rainfall intensities of different durations and different return periods. The results of the rainfall IDF curves can provide useful information for policymakers to make the right decisions in controlling and minimizing flooding in the study area. The IDF relationship developed can be used as valuable information for the planning and management of hydraulic structures in the region.

ABSTRACT. One of the major challenges of the day is urban sprawl. Urban sprawl is the cause of a number of urban environmental problems, including deteriorating water quality, increased runoff and subsequent flooding, lower air quality, and increasing local temperatures. For the purpose of this study, Vadodara City is used as a case study to examine the urban growth and alteration of the land use/ land cover (LULC) over the course of 38 years, from 1977 to 2015. To investigate the spatial land use changes that had place throughout the research period, remote sensing technology is used. Two Landsat satellite images of the year 1977 and 2015 have been used to apply these techniques. Pre-processing of these images using the Hybrid classification. Total four land use classes have been identified by image classification based on satellite images and Google Maps, which are Water Bodies, Vegetation area, Built-up area and Uncultivated area. The results from 1977 to 2015 showed a substantial increment in built-up area from 4.205 sq. km to 74.80 sq. km., agriculture area from 4.29 sq. km to 41.27 sq. km., water bodies from 0.176 sq. km. to 2.203 sq.km. and uncultivated area increased from 7.75 sq. km to 31.11 sq. km. respectively. In general, flooding is caused by inadequate urban infrastructure and the fast, unrestrained expansion of cities into marshes and other areas that typically absorb surplus rainwater. This study thus highlights the significance of considering land use/ land cover changes when assessing the worst-case scenarios for flooding of the city. Increased built-up area causes frequent flooding and water logging. Therefore, it is important to regularly review the LULC and then, the necessary measures can be taken to prevent floods.

ABSTRACT. The primary catastrophe impeding urbanization in underdeveloped nations is flooding. Lead productivities, characteristics, and losses due to flooding. Additionally, it resulted in life and limb loss. Therefore, in order to lessen the threat to a certain extent, countries that are developing must priorities flood assessment and risk management. The hydrodynamic modelling is essential for calculating and estimating flood risk, but it requires a large amount of input data to simulate flood events. DEM is one of the crucial variables that have a big influence on flood estimation and risk assessment. DEMs are created using RL points, satellite techniques, and contour data. The most common and valuable DEM used for modelling currently is satellite-based DEM, such as SRTM and ALOS. Nevertheless, current developments in RS and AI have made the UAV (Drone) based DEM a very important tool to use for flood simulation in 1D and 2D. The major goal of this research is to compare the results quality by using a DEM based on UAV and a DEM based on satellite data. Due to money and resource constraints, freeware DEM is better suited for analysing flood disasters in developing countries, but it could prove challenging for those making decisions to assess the accuracy of the simulation results. This research attempted to compare the results of two free satellite-based DEM images and one UAV-based DEM image for the modelling of flooding over an 8 km2 stretch of the Sabarmati River in Gujarat, India. For satellite-based DEM (Cartosat and ALOS) and UAV-based DEM (10x10 grid), the discharge, E.G. elev., and velocity parameters simulated model are compared. For hydrodynamic modelling, the HEC-RAS 6.0 unsteady flow conditions is employed. The U/S boundary is bounded by flood hydrograph and D/S boundary is considered as normal depth. For each segment, it has been noted that Carrtosat and ALOS vary with some variation with UAV-based DEM. The performance of the simulation's R2 of the parameter demonstrates 0.9996 and 0.9999 for the discharge valve. Freeware satellite extracted parameters that are substantially similar to UAV-based DEM include discharge, E.G. elev., and velocity. Therefore, in developing nations, using a satellite-based DEM for wide areas (Catchment) is extremely promising; nevertheless, the UAV-based high-resolution DEM will be used for urban areas. Therefore, using numerous DEM platforms for flood simulation will be economical for decision-makers.

ABSTRACT. Increasingly erratic climatic conditions cause sudden water surges from the Ukai dam in the river Tapi, flooding Surat city. There were disastrous floods in 1883, 1884, 1942, 1944, 1945, 1949, 1959, 1968, 1994, 1998, 2002, 2006, 2007, 2008, 2009, 2010, 2011, 2012, and 2013. The river's carrying capacity is estimated to be around 4.5 million cubic meters. Due to silting and encroachment caused by urbanization, the effective waterway of the river Tapi is shrinking in width and depth, reducing the river's carrying capacity. The Hydrologic Engineering Centres River Analysis System (HEC-RAS) software is used in this study to assess the carrying capacity of the Lower Tapi River reach, which is 12.25 km long, and to conduct an unsteady flow analysis for the tidal reach in order to understand the effects of high and low tides. As a study reach, the Tapi River reach was taken from Weir cum Causeway to Magdalla bridge. There are 49 cross-sections in the study's reach. The variations of water surface level due to upstream flow and high tidal backflow conditions in the channel reach are considered in the study. For the years 2010, 2011, and 2012 daily discharge data, unsteady flow analysis was performed using HEC-RAS, taking into account the tidal surges and backflow. To improvise the effect of minimum and maximum variations of tidal effect, three years of daily discharge data were considered and studied. The effect of changes in tidal level variations due to various peak discharges has been studied. The effects of the tidal effect on flooding are depicted in this study. To prevent flooding at various critical cross-sections levees are proposed to be constructed to store and mitigate the high inflow. This study may be used as a basis for disaster management, flood management, early warning systems, and infrastructure development decisions.

ABSTRACT. Flood events pose significant challenges to communities and agricultural areas, necessitating the flood modelling and mitigation strategies. Check dams are widely recognized as effective structures for flood mitigation and groundwater quantity and quality enhancement. Araniyar river is of 108km flows through the states of Tamil Nadu and Andhra Pradesh. This study analyze the hydrological impact of the Reddipalyam Check dam, Araniyar River on flood mitigation and groundwater quality enhancement. This study focuses on the application of HEC-RAS flood routing modelling for generating flood inundation map of Araniyar River. The research aims to investigate the effectiveness of the check dam in reducing flood risks and improving the quality of groundwater used for irrigation purposes. The study comprehensive of two main objectives: 1) flood routing modelling to simulate inundation scenarios with and without check dam, and 2) groundwater analysis to compare the groundwater quality before and after the rehabilitation of a check dam. To visualize the potential extent of flooding, flood inundation mapping is performed using Google Earth images. Additionally, a farmer survey is conducted to validate the effectiveness of the check dam in reducing flood risks and improving water quality. The research findings contribute to a better understanding of the benefits and effectiveness of check dams in flood mitigation and groundwater management.

ABSTRACT. Hydrodynamic conditions and bedload transport in river bends are characterized by three-dimensional effects and still in the focus of research. In bends secondary flow leads to typical cross sections with a cut bank at the outer bend and a point bar at the inner bend. Hydraulic structures (e.g. groynes) increase the complexity additionally. At the Federal Waterways Engineering and Research Institute (BAW), the effects of river training measures in bends are investigated using a laboratory model and a two-dimensional numerical model as its digital twin. A 60 m long and 12.5 m wide laboratory scale model with bends was built in order to investigate different hydrodynamic and morphodynamic phenomena at river bends. The mobile bed is modelled using lightweight sediments made of polystyrene. Different steady-state discharges (low, mean and bankfull) were operated as well as artificial hydrographs. High-quality measurements allow the analysis of the phenomena in river bends and can be used to calibrate and compare the results to the numerical model. The open source software TELEMAC-2D was used to numerically reproduce the laboratory model at the same scale. The numerical model was calibrated using the measured water levels, longitudinal and lateral bottom slopes and transport rates. A good agreement between the results of the numerical model and the measurements of the scale model could be found. In addition, the numerical model was run in natural scale with sand instead of polystyrene. The results of the scaled models with the artificial material were compared with the full-scale numerical model with sand. The hybrid use of a laboratory model and the numerical model as its digital twin can overcome the drawbacks of both types of models. The results from both types of models provide better reliability, which is important for the evaluation of river training measures.

ABSTRACT. In river bends, sediment transport is affected by secondary currents. The near-bottom current transports sediment towards the inner bank, where it is partially deposited. Such a phenomenon can be observed, for example, on the Middle Rhine, in the area of the Jungferngrund gravel for-mation. An ecological valuable gravel bar has formed here on the inner bank of a sharp bend. The problem is that parts of the sediment material deposit within the navigation channel, which re-quires regular dredging.

In this area, the sediment transport processes are significantly influenced by the fractured bed-rock in addition to the three-dimensional flow effects. However, the transport processes of sed-iment over partially alluvial covered bedrock are poorly understood at present. In order to better understand these processes and to develop river training measures to reduce sediment deposi-tion within the navigation channel, a hybrid modelling approach was chosen. This consists of a laboratory scale model with a partially movable bed and a three-dimensional hydro-numerical model. In the scale model sediments particles with low density were used to model the sediment transport. An extensive database with data from nature was available for setting up the two models. The models were calibrated using natural data. Both models reproduce the measured water levels and velocity distributions well. It was also shown that the sediment depositions in the laboratory model correspond well with the positions of past dredging campaigns.

In order to determine the transport path of the sediment across the bed, extensive investigations were carried out in the laboratory model at different discharges. The comparison with the near-bottom streamlines in the 3D model shows that the sediment particles very strongly follow the near-bottom flow over the bedrock. The transport path of the sediments was also confirmed by comparison with a high-resolution digital terrain model where the transport path was estimated from the dominant dunes.

The results show a very good representation of the near-bottom flow in the three-dimensional hydro-numerical model. The model is therefore very well suited for the design of river engineer-ing measures to influence the bedload path in order to reduce landfalls in the navigation channel, , even if the bedload transport is not directly represented. The hybrid modelling approach has proven to be very powerful, both in terms of the ability to investigate a large number of measures and in terms of the accuracy of the results for physically extremely complex processes.

ABSTRACT. The study of the boundary layer around the hull of submarines is considered as one of the most important research projects because it is closely related to the drag force in addition to the effects related to the hydrodynamic compounds on submarines.

In this study, a numerical model based on computational fluid dynamics (CFD) is proposed to determine the boundary layer thickness on the hull of a three-dimensional submarine on its three axes (x, y, z). Furthermore, the relation among the thickness of this boundary layer and the dimensions of the submarine in addition to the flow velocity is also determined. In order to achieve this result, five three-dimensional models of a submarine with different scales have been considered for ten different values of velocity. Hence, the thickness of the boundary layer could be determined for 50 study cases. These calculations have been made by using a non-structured mesh of the type "Fluent Mosiac". Whereas, the boundary layer area consists of a number of layers calculated as a function of velocity and the value of y^+. The mesh in this area is "Poly-prisms" and the mesh density is calculated to capture all points of flow on the hull of the submarine. This study does not use the functions of the wall to solve the boundary layer, but the value of y^+≈1 is fixed in order to solve the flow equations in the viscous layer. Then, the maximum thickness of the boundary layer is calculated. Moreover, the hydrodynamic forces on the lifting wings of the submarine, and the point of separation of the boundary layer, and its effect on both the lifting and drag forces are also determined.

The importance of this research lies in the role played by the boundary layer in the development of submarines by injecting fluids of low viscosity around the bodies of these vehicles to reduce the frictional resistance, and the principle result of this study is the calculation of the minimum volume of the fluid which should be injected for the purpose of generating a stable bubble that completely envelops the hull. This minimum volume should be equal to the volume which takes into account the thickness of the boundary layer.

ABSTRACT. CNR is the first producer of 100 % renewable energy in France, operating and managing 19 hydroelectric power plants on the Rhone River. When a sudden shut-down of turbines occurs (because of mechanical or electrical breakdown), the sudden disruption in discharge at the power plant generates waves that raise the headrace channel water level and propagate upstream over several kilometers. Theses waves are made of primary wave which amplitude is proportional to the incoming discharge and secondary waves, called Favre waves, which can be generated because of the vertical acceleration, that superimpose on the primary waves. These waves must be investigated with care since they may overflow and erode the dike crest, hence possibly damaging the dike core and even harming people. Traditional 1-D Shallow Water Equation numerical models based on the assumption of hydrostatic pressure distribution fail to accurately reproduce the complex patterns of the waves system. Improving the knowledge of this phenomenon and predicting it, is therefore something essential for CNR. In the frame of the project of the refurbishment of Châteauneuf-du-Rhône hydro-powerplant (1,6 billion KWh, nominal discharge 1,850 m3.s-1, six Kaplan turbines), a dedicated study has been implemented to assess the consequence of increasing the nominal discharge on the waves amplitude and to anticipate any mitigation in case of adverse effects would be pointed out. Hybrid modeling of these phenomena has been initiated on the Montélimar run-of-river scheme, including a 1/35 undistorted physical model representing 1200 m of the upstream stretch of the powerplant and an overall numerical model of the 13 km headrace channel. In order to calibrate more accurately the large-scale numerical approach by 2D Basilisk code [1], tests have been conducted on a physical model at CNR hydraulic laboratory. The results obtained thanks to an innovative way of modelling the turbines shutdown process in fully similitude (water hammer, initial conditions [2]), completely match with the field observations and the measurements recorded during the onsite tests. This paper aims at exposing the numerous practical difficulties encountered during the carrying out of transient tests, including the generation of a wave in a pressure flow conduit as well as its propagation at the free surface of the physical model in complete similitude. The findings contribute to advance and extend the knowledge and the management of disjunction phenomena within the field of hydraulic engineering.

ABSTRACT. The Saut Mortier hydropower plant, located in the city of Cernon (39 - Jura), is part of the cascade forming the Ain hydroelectric complex. It is surrounded upstream by the Vouglans dam and downstream by the Coiselet, Cize-Bolozon and Allement dams. It is composed of a spillway, made up of 3 gates and a powerhouse housing two production units. The powerhouse is located on the right bank of the river. EDF intends to add a pumping/turbining unit on the left bank. This new facility will partly reused the old diversion tunnel, built for the construction of the dam. Numerical and physical modellings (scale model) of the new water intakes are being carried out in order to understand how it will operate and, if necessary, propose technical improvements to ensure better hydraulic behaviour. In particular, the following issues will be studied: risk of vortexes, flow pattern in front of the intakes, velocity criteria at the screen planes, homogeneity of the velocity field and its distribution through the intakes and head losses. The specific features of the different approaches, with their advantages and limitations, will be discussed.

ABSTRACT. In penstock pipes upstream of hydropower stations, a safety valve is located at the top in order to cut the flow rate before the downward slope if necessary. During its closing, the pressure downstream reduces whereas the upstream one remains constant. To avoid low pressures in the penstock pipe, air-venting valves are present and open when the pressure inside the pipe is lower than outside. Once opened, the air from the atmospere flows into the penstock pipe and counters the pressure decrease. The correct design of air entry systems is of primary interest for hydraulic engineering. The present work is focused on the numerical simulation of air entry in a penstock pipe during the safety valve closing. To close or open valves, the CFD model uses a discrete forcing method (Immersed Boundary) to allow solid motion without any need for dynamic re-meshing. For the air-venting valve, its mouvement depends directly on the pressure forces on its two sides (the atmosphere outside, the penstock pressure inside). The two-phase air-water flow is modeled with an eulerian-eulerian approach with a multi-regime model solving large gas structures and modeling bubbles. A dedicated reduced-scale closed-loop experiment is used to characterize the flow in terms of pattern and to reach pressure evolution in the pipe. The numerical model is validated for different flow rates and for different number of air-venting valves (1 to 3 with different diameters). Moreover, operating condition data are used to validate the model at full-scale. The numerical model is in a satisfactory agreement with measurements, and show its ability to increase our knowledge on the phenomenon of air entry in such configuration.