HS4.1

After having swept over western France, the ALEX storm led to an exceptional Mediterranean rainfall event which hit the “Alpes Maritimes” region during the night of the 2nd to 3rd October 2020. The rainfall accumulations observed on 12 to 24 hours durations were unique in this region, with a record of 663mm in 24h (EDF raingauge at Les Mesces).

Form West to East, several valleys, mainly those of Tinée, Vésubie, and Roya were affected by major floods, landslides, sediment transport and geomorphological changes. The hydrometric network was almost destroyed. The human and material damages were considerable, with many fatalities and missing people, several villages largely destroyed, and important destructions of communication and transport networks.

A lot of technical post-flood surveys were launched by national authorities to gather a detailed knowledge of the event characteristics, with regard to rainfall accumulations, water discharges, description the torrential phenomena, and inventory of damages. This communication is focused on the question of water discharges.

National and local authorities and organisms, universities and companies, were involved in different post-flood surveys aiming at gathering information on the peak discharges and the hydrographs of the floods, for their own needs and/or within structured programs (Administrative survey, HYMEX research project www.hymex.org).

Several kind of discharge field estimations were provided using field survey measurements, satellites images, post-event Lidar data, combined with hydraulic estimations based on hydraulic formulas, and 1D/2D hydraulic models. Several teams also applied hydrological models based on radar quantitative precipitation estimates, to calculate hydrographs at different basins outlets.

To combine and draw a uniform synthesis of all these results, a consensus exchange was launched to share the knowledge gathered by the different data providers. The objective was to compare, assess, and propose common intervals of peak discharges in the different impacted valleys. We also evaluated for each valley the return period of the final interval of discharge established by the consensus.

The final product is an official administrative document, established at the end of October 2021 by the French state authorities, providing the peak discharge values to be used for post flood studies, reconstruction, and prevention measures.

How to cite: Pons, F., Bonnifait, L., Criado, D., Payrastre, O., Billaud, F., Brigode, P., Fouchier, C., Gourbesville, P., Kuss, D., Le Nouveau, N., Martin, O., Nomis, S., Paquet, E., and Cardelli, B.: Towards a hydrological consensus about the 2nd – 3rd October 2020 ALEX storm event in the French “Alpes Maritimes” region, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7913, https://doi.org/10.5194/egusphere-egu22-7913, 2022.

Extreme weather situations are becoming increasingly frequent with devastating consequences worldwide. Heavy rainfall events in July 2021 caused severe flash floods in western Germany, Belgium and the Netherlands, resulting in a high number of casualties and material damage. The high hazard potential combined with the low reaction times, associated with these events, make it necessary to develop efficient and reliable early warning systems (EWSs) to facilitate the preparation of response strategies. As nowcast precipitation forecasts are continuously improving in both quality and spatial resolution, they become an essential input for flash flood and landslide prediction models and therefore an important component in EWSs. However, the inherent uncertainty of radar-based nowcasting systems are carried over to the output of those prediction models. Therefore, this study aims to analyze the uncertainty sources of nowcasting products of the German weather service (DWD) using the July flood Event 2021 as a case study. More specifically, the objective is to determine whether the quality of precipitation nowcast products is sufficient for usage in physics-based flood or landslide prediction models. Due to the complex nature of weather and rainfall structures as well as their spatio-temporal variability, traditional cell-by-cell comparison of predictions and ground truth is insufficient to quantify forecast quality. To overcome this issue, uncertainties in magnitude, time and space and their respective sources are identified, using techniques from various fields of science. Subsequently, error propagation in flash flood prediction models is analyzed by applying the previously determined uncertainty ranges to a hydrological model.

How to cite: Winkels, B., Hofmann, J., Yildiz, A., Edrich, A.-K., Schüttrumpf, H., and Kowalski, J.: Quantifying rainfall forecast uncertainty and error propagation in flash flood and landslide prediction models, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6170, https://doi.org/10.5194/egusphere-egu22-6170, 2022.

Mountain rivers experience channel widening as a response to high-magnitude hydrological events. Several studies indicate that the unit stream power (ω_pk) and the lateral confinement (CI) are among the most important constraints to explain channel modifications induced by a flood. That said, with the same controlling factors, a relatively broad spectrum of width ratios (WR = channel width after/before the flood) is commonly observed in real case studies. Sediment transport in mountain streams occurs via processes classified as debris flows (DFW), debris floods (DFD) and water flows (WF). This study aims to test if different flow-types are one of the drivers of channel response to floods, specifically investigating if there is a relationship between DFD (i.e. a transport condition characterized by extremely high bedload) and intense channel widening.

The case study is the Cordevole catchment (Dolomites, Italy; drainage area of 857 km²), which in October 2018 was affected by a severe hydrological event (Vaia storm). Besides the main stem of the Cordevole River, we considered four of its tributaries (Tegnas, Pettorina, Liera and Corpassa torrents). WR was determined at the sub-reach scale through aerial photographs analysis and ω_pkwas calculated considering the discharge at the flood peak provided by hydrological modelling. A post-flood survey allowed us to determine the flow-types that occurred at each sub-reach of the Tegnas Torrent during the event. The possible upheaval from WF to DFD along the other streams was determined considering the presence of conditions required for local occurrence of DFD (i.e. ω_pk exceeding 5000 Wm^-2 and/or DFW tributaries delivering large amount of sediment into a receiving stream).

DFD sub-reaches of the Tegnas Torrent experienced widenings that, at the same ω_pk, were 2-3 times larger than WR of WF sites. These processes-specific relationships were used to recognize sub-reaches of the other streams were an “anomalous widening” occurred during the Vaia event, i.e. sites where WR was significantly larger-than-expected for a specific ω_pkunder conditions of WF. Among 117 sub-reaches, anomalous widening was recognized at 13 and 6 sub-reaches of the Liera (WR up to 16) and Pettorina (WR up to 10) torrents, respectively. All these sub-reaches were characterized by the presence of conditions required for DFD occurrence during a high-magnitude flood, allowing us to infer that the process responsible for sediment transport during the Vaia event was likely DFD. Contrariwise, no sub-reaches of the Cordevole and Corpassa streams experienced anomalous widening, likely because WF occurred along their whole courses due to their morphological characteristics (e.g. wide channel before the flood) and/or lower magnitude of the flooding locally induced by the storm.  

These results suggest that an extraordinary-widening characterizes DFD channel sites, which, during a severe flood, can be affected by channel changes remarkably more intense than those occurring in response to WF. For this reason, in addition to hydraulic and morphological constraints, the different sediment–water flows possibly occurring at a sub-reach should be considered as a further controlling factor for channel modifications and, consequently, for prediction of geomorphic hazard at local scale.

How to cite: Brenna, A., Marchi, L., Borga, M., Zaramella, M., and Surian, N.: Debris floods and channel widening in mountain rivers: Examples from the Vaia Storm (October 2018) in the Cordevole River catchment (Dolomites, Italy), EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7768, https://doi.org/10.5194/egusphere-egu22-7768, 2022.

The study is geared towards the implementation of a workflow based on a Support Vector Regression Machine Learning (SVR-ML) approach which is conducive to estimates of flowrates across a given cross-section of a target stream in the presence of extreme precipitation events. The work is motivated by the observation that damages ensuing flash floods are a matter of global concern. A broad set of evidences suggests the ecosystem is experiencing changes of precipitation extremes, a causality relationship between increasing extreme floods and global climate dynamics being evidenced. In this context, practical tools associated with analyses of floods caused by extreme precipitation events can assist the design of early alert strategies across vulnerable regions. Physically and conceptually-based models have been extensively employed to link stream flowrates to precipitation events. These kinds of models are formulated and validated upon relying on continuous monitoring of flowrates as well as hydrometeorological variables associated with the area of the watershed related to a target stream. The typically high uncertainties underlying (a) the description of the physical processes governing the rainfall-runoff relationship as well as (b) monitoring and quantification of quantities and attributes characterizing the system behavior tend to propagate to outputs of interest of a given model. When considering well instrumented watersheds, data-driven modeling approaches grounded on machine learning (ML) algorithms can be an attractive alternative/complement to physically-based modeling approaches to analyze extreme flood events. Here, we rely on a Support Vector Regression ML (SVR-ML) algorithm that makes use of a linear kernel to provide estimates of hourly flowrate at a stream upon relying on observations of hydrometeorological variables across the watershed associated with the stream. The analysis encompasses three watersheds differing in size (ranging from about 25 to 250 km2) and located in the North of Italy and is structured across three steps: (i) identification of variables that are most informative to the target quantity (i.e., the flowrate in the stream), a step relying on cross-correlation and partial auto-correlation analyses; (ii) training of the SVR-ML algorithm, comprising the estimation of the optimal hyperparameters and parameters of trained models and the ensuing validation; and (iii) analysis of the anticipation time at which an early alert is effective, model performance being then quantified through the typical Mean Average Percentual Error (MAPE) metric. Our results suggest that, as expected, precipitation is the main driving force in a rainfall-runoff process, quantities such as temperature and relative humidity being least informative to the construction of the ML model considered. The predictive capability of the model (quantified through MAPE) is influenced by the desired anticipation time (i.e., the distance in time between the inputs and the output of the ML model). In general, one can note that (i) predictions of enhanced quality (MAPE smaller than 10%) are obtained for shorter anticipation times and (ii) models associated with low values of MAPE are obtained if the anticipation time is equal to or smaller than the time of concentration of the watershed.

How to cite: Sandoval, L., Riva, M., and Guadagnini, A.: Estimation of Floods Related to Extreme Precipitations through a Machine Learning Approach, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-1952, https://doi.org/10.5194/egusphere-egu22-1952, 2022.

Best allocation of ressources from stakeholders that face fast-flooding implies a dynamic representation of the risk, exposure, and danger, in a situation where parameters (roughness, infiltration, drainage network etc.), and input (bathymetry and rainfall) can be both uncertain and volatile. Ensemble strategy simulation appears as a good approach to deal with these issues.
Fast-flood event are also typically events where meteorological predictions can underestimate the actual rainfalls until very late. Urban microcharacteristics can also make models sensitive to spatial resolution, and « events » such as log jam can even modify DEM. 
At Strane Innovation, we develop a decision-support tool called BlueMapping. To be operational, that is, fast to deploy and reliable, we use this ensemble strategy together with the fastests simulation models deployed on powerful computers. It also requires quick and robust routines for the setup of the model, with proxies when data is not available at the moment, and inputs that are easy to modify if necessary.
We will ilustrate this discussion through the test case of the Alex storm that hit la Vésubie and la Roya valleys. After a quick benchmark with a standard model, we will compare the outcome between the flooding predicted by different models and the actual outcome. 
Since BlueMapping can be integrated in an alarm system, it is important to assess the values of the confusion matrix, in particular the false alarm ratio, to make sure our tool keeps its value over time.

How to cite: Wacheux, C.: Ensemble strategy for decision-support tool : a case study of the Alex storm in 2020 in la Roya and la Vésubie valleys, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9023, https://doi.org/10.5194/egusphere-egu22-9023, 2022.

Efficient pluvial flood mapping methods are needed to produce realistic flood scenarios in very small upstream catchments. The Cartino2D method was developed to launch automatically 2D models based on the Telemac2D hydraulic software. The principle is quite simple, (1) create automatically the mesh with a topography based on Lidar, (2) manage automatically the boundary conditions, (3) run the model based on rainfall input data, and (4) postprocess the results. The extent of each 2D model generally varies between 2 to 10 km² with a maximum of 20km². The only manual work consists in checking or modifying the limits of hydrological catchments.

We began to use this method on the Toulon metropole (South of France) with 66 complementary computation domains covering about 180km² and using eight statistical rainfalls. We also tested and evaluated this method on twenty other case studies in different regions of France. In this presentation, we focus on two evaluations (flood of June 2010 in Draguignan and flood in 2014/2015 around Montpellier) conducted within the ANR PICS project.

In this project, we improved the method to automatically integrate radar rainfall and to compare the results with local knowledge, observed historical floods and local hydraulic studies.

Cartino2D offers interesting results in areas with natural, rural land-use or few urban developments. The density of the mesh (less than 3m in the thalwegs) and the Telemac2D model quality are sufficient to obtain a good accuracy in these areas.  

In urban areas, the method provides a first knowledge, but more complex input data are needed to improve the accuracy of results.

We try in this presentation to describe which databases should be created to improve the accuracy of such automatic computations. At the scale of urban areas with results around buildings, the databases need to be spatially well defined. We propose some standard of databases to be integrated in computations: for example, the main underground channels or culverts, the main aerial channels (particularly very small channels not recognized by Lidar), a spatial distribution of the Curve Number and of the Manning’s coefficient.

This kind of databases, which cannot be deduced automatically from Lidar data, appears as essential to improve the results of Cartino2D automatic process. This kind of knowledge exists locally, but up to now it is not integrated in homogeneous national or regional databases.

In the same way, we need also to have well-defined databases to compare automatic results with historical floods as flood marks, gauge stations.

Automatic 2D mapping of flash floods seems to be a realizable goal at a scale of a region or country with standard 2D hydraulic models. But the current main limits appear to be a lack of good input database management, which limits the current accuracy of mapping results.

How to cite: Pons, F., Alquier, M., and Paya, E.: Automatic 2D mapping of flash floods: which possibilities and limits? An illustration based on the Cartino2D method, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7212, https://doi.org/10.5194/egusphere-egu22-7212, 2022.

Modelling river hydrodynamics in an efficient approach remains a technical challenge which limits our ability to assess river flood hazard or to use process-based erosion laws at a high-resolution in landscape evolution models. Here we present a fast iterative method, entitled FastFlood, to compute river depth and velocity in 2D on a digital elevation model (DEM). This new method solves for the 2D shallow water equation, without the inertia terms, by iteratively building the river water depth using classical flow routing algorithms based on directed acyclic graphs, including the classical single or multi-flow, applied to the water surface. At each iteration, the water depth of each cell of the DEM increases by an increment that is a function of water discharge, computed using a flow accumulation operation, and decreases based on a flow resistance equation, in a manner similar to the Floodos model (Davy et al., 2017). In the hydrostatic mode, this operation is repeated until reaching a near constant water depth over the entire DEM, which occurs after a few tens or hundreds of iterations. FastFlood can also solve for the dynamic propagation of a flood in the hydrodynamic mode. In this case, the water depth increment is only a function of the water discharge exiting the direct upstream neighbors and the iterations are replaced by a time evolution of the water depth. Water depths obtained with the hydrostatic solution were validated against an analytical solution in the case of a rectangular channel and with the Floodos model for natural DEMs. Compared to previous hydrodynamic models, the main benefits of FastFlood are its simplicity of implementation, which mainly requires a classical flow routing algorithm, and its efficiency. Indeed, for a DEM of 10⁶ cells, the algorithm takes about 2 minutes on a laptop to find the hydrostatic solution, about 10 times faster than using the Floodos model (Davy et al., 2017) that was already significantly faster than other hydrodynamic models. Moreover, the computational time scales a little more than linearly with the number of cells, which makes FastFlood a suitable solution even for DEMs larger than 10⁶ – 10⁷ cells. In the future, we expect to make progress on the numerical method by adapting graph-based solutions to the issue of flow water routing. Following Davy et al. (2017), we will also include FastFlood in a landscape evolution model to couple it to process-based laws for erosion, transport and deposition of sediments.

How to cite: Steer, P., Davy, P., Lague, D., Bernard, T., and Feliciano, H.: FastFlood: a fast and simple 2D hydrodynamic or hydrostatic numerical solution to river flow in landscape evolution models, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-3142, https://doi.org/10.5194/egusphere-egu22-3142, 2022.

Flash floods are a rapid burst of flood water that can cause extreme damage to populated areas. The European floods in France, Belgium, Germany and the Netherlands in the summer of 2021 featured a wide range of flash floods with a large number of casualties and vast financial damage. Reflection on the risk reduction strategies have reemphasized the need for early warning systems in the upstream catchments of North-Western Europe. For applications such as this, the speed of flow simulations is critical, as the quality of real-time forecasting often depends on the frequency and amount of simulations that can be carried out as new weather forecasts come in. We present a new type of flood hazard model that, in many typical cases, solves flash flood hazard a 100 times faster with similar accuracy. The developed method employs steady-state solvers for diffusive wave water flow equations to skip the dynamical process and directly estimate relevant parameters such as maximum flow height, maximum flow velocity and relative arrival time of the flood water. These paramters are often the most important for warning systems and descision making in risk reduction. Our adapted algorithm improves upon traditional steady-state flow solvers by employing inversed flow accumulation results and compensation for partial steady-state flow. We show the accuracy of the method is similar to full dynamic water flow simulation in many types of events, such as the extreme 2003 floods in the Fella Basin (Italy), Hurricane-induced flooding on Dominica and the flood impact in Limburg in 2021 (The Netherlands). On average, with highly similar accuracy, calculation time was reduced from approximately 6 hours to 2.5 minutes. We further investigate the limits of the developed methods, in particular to practical applications in different type of flood events. While the sensitivity of the model to initial conditions is similar to that of regular flood models, the sensitivity of the hydraulic aspects is lower. Finally, we discuss potential usage for early-warning, spatial descision support systems and serious gaming approaches. While further investigation is required to fully validate the method, a break-through in flood hazard assessment could be on hand.

How to cite: van den Bout, B.: Super-fast flash flood simulation using steady-state flow solvers, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6953, https://doi.org/10.5194/egusphere-egu22-6953, 2022.