Flash floods occur when heavy rain causes a fast and powerful flow of water in a drainage area. In the Eastern Mediterranean region, which contains arid and semi-arid areas, the location and timing of rainfall is the most significant factor in the formation of flash floods. Predicting when and where extreme weather events such as storms, heavy rainfall, and flooding are likely to happen is a key challenge in the effort to prevent natural disasters. Here, we present an improved version of a previous work by Ziskin and Reuveni, which investigated the use of precipitable water vapor (PWV) data from ground-based global navigation satellite system (GNSS) stations, along with surface pressure measurements to predict flash floods in an arid region of the eastern Mediterranean. The previous study involved training three machine learning models to perform a binary classification task, using multiple unique flash flood events and testing the models using a nested cross-validation technique. The results showed that the support vector machine (SVM) model had the highest mean area under the curve (AUC) and the lowest AUC variability compared to random forest (RF) and multi-layer perceptron (MLP) models.  When tested on an imbalanced dataset simulating a more realistic flash flood occurrence scenario, all models demonstrated a decrease in the false alarm rate (precision) with a high hit rate (recall) performance.

In this study, we extend the previous work by integrating nearby lightning data as a new feature in our studied dataset. The inclusion of this feature is motivated by the observation that heavy rainfall, which can lead to flood events, is often accompanied before by an increase in lightning activity. The experimental results show that the adding a 24-hour vector of nearby lightning activity improves the precision score significantly.

How to cite: Reuveni, Y., Asaly, S., and Gottlieb, L.-A.: Flash flood predictions over the Eastern Mediterranean using artificial intelligence techniques with precipitable water vapor, pressure, and lightning data., EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-9979, https://doi.org/10.5194/egusphere-egu23-9979, 2023.

Flash floods (FF) represent an important part of the flood damages and fatalities in the world. Today, operational FF nowcasting and warning systems are often based on the use of precipitation weather radars, and therefore still offer limited anticipation. They also generally rather represent the intensity of the flood events than their severity in terms of impacts, which may limit the capacity of emergency services to take relevant decisions.

This contribution aims at evaluating the value of a new ensemble FF impacts forecasting chain for the decision making of an emergency service.  The case study corresponds to the Aude River flash floods that occurred on October 15 and 16, 2018, and which are among the most important FF observed in southeastern France in the recent years. This event is responsible for the death of 15 people (99 people injured), as well as particularly large material damages.

The tested FF impacts forecasting chain combines three new rainfall ensemble forecast products (provided by CNRM), specifically designed for short-range forecasting (0-6h), and a highly distributed rainfall-runoff model (Charpentier-Noyer et al., 2022). A simple impacts model is built and applied for each river reach based on a catalog of 8 inundation scenarios corresponding to return periods of 2 to 1000 years. The impacts are represented in terms of a number of inundated buildings.

The value of the ensemble impacts forecasts is finally evaluated based on the implementation of a multi-agent model, for the simulation of the field decisions taken by an emergency service. This new evaluation approach, based on simple but realistic hypotheses, allows to illustrate and measure the gains associated with a better anticipation of impacts, and the costs associated with false alarms, which lead to the unnecessary mobilization of rescue teams, to the detriment of really impacted locations. In case of extremely limited means for safety operations (low number of rescue teams), the decisions based on a naive zero future rainfall scenario may sometimes appear better than those using ensemble rainfall forecasts. Nevertheless, in all the simulated cases, the decisions taken from the ensemble rainfall forecasts appear more efficient than those based only on field observations.

Charpentier-Noyer, M., Peredo, D., Fleury, A., Marchal, H., Bouttier, F., Gaume, E., Nicolle, P., Payrastre, O., and Ramos, M.-H.: A methodological framework for the evaluation of short-range flash-flood hydrometeorological forecasts at the event scale, Nat. Hazards Earth Syst. Sci. Discuss. [preprint], https://doi.org/10.5194/nhess-2022-182, in review, 2022.

How to cite: Charpentier-Noyer, M., Nicolle, P., Payrastre, O., Gaume, E., Bouttier, F., and Marchal, H.: Relevance of using ensemble forecasts of flash-flood impacts for an emergency service: an evaluation for the October 2018 flood event in the Aude river basin, France, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-11757, https://doi.org/10.5194/egusphere-egu23-11757, 2023.

Recent large floods across Europe, including those in Belgium and Germany in 2021 or, more recently, in Italy in October 2022, showed that major obstructions of bridges due to the mobilized large wood (LW) significantly influenced the flood-related damages. However, in principle, none of the dangers posed by wood was inherent to the wood itself but to the obstacles and infrastructures that were not designed to allow the wood to pass. Understanding this legacy effect on wood in rivers due to the increased artificial trapping efficiency of river structures (bridges, dam reservoirs) still needs to be completed.

The Francolí River in Catalonia, NW Iberian Peninsula (853 Km2 area and 59 km length) underwent a major flash flood on October 22, 2019, that caused six fatalities. The rainfall recorded in the NW basin was 293 mm in 24 hours. Consequently, significant bio-geomorphological changes occurred; a large amount of sediment was eroded, transported and deposited, and many trees were damaged or uprooted with subsequent large wood (LW) supply and transport. In addition, infrastructures were severely damaged (e.g., three bridges collapsed).

The legacy effects on instream large wood related to the human infrastructures in river systems is an essential factor to consider when assessing the effects of floods and potential risks. Therefore, this study's main objective was to evaluate the influence of bridges on large wood accumulation during floods. 

We analyzed a reach of 30 km along the Francolí River in which there were 23 bridges. The reach was split into 52 sub-reaches based on their morphological characteristics (i.e., the width of the valley bottom, slope, and sinuosity), the presence of infrastructures, or lithologic and anthropic knickpoints, and the junction with tributaries. The 52 sub-reaches were grouped into four main typologies based on statistical segmentation and clustering.

Individual pieces of LW and accumulations were digitalized using post-flood high-resolution orthophotos (i.e., 0.10 m resolution). They were characterized using four attributes: orientation with respect to the channel (parallel, perpendicular, oblique), transported (yes or not), location (active channel or floodplain), and length. Average Nearest Neighbour, Spatial Autocorrelation (Global Moran's I test) and Density were computed and revealed the depositional pattern of LW along the study reach.

Preliminary results showed that morphological characteristics favoured LW trappings: wide valley bottoms and sinuous bends. In addition, the standing vegetation and other in-channel obstacles were crucial to trap wood. The most significant aspect, however, was the presence of bridges. A significantly more considerable amount of wood (i.e., the highest density observed, ranging between 33 and 101 pieces/ha) was trapped upstream from bridges, where wood was deposited at significantly higher elevations. Further analyses will explore the characteristics of the bridges and upstream sub-reaches.

This study will provide crucial information to understand large wood accumulation at bridges during floods and will inform flood-hazard assessments and river management.

How to cite: Valera-Prieto, L., Ruiz-Villanueva, V., and Furdada, G.: Bridges influence large wood trapping efficiency during large floods: insights from the Francolí River flood in 2019, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-766, https://doi.org/10.5194/egusphere-egu23-766, 2023.

In the last 20 years a variety of heavy precipitation events (HPEs) have caused severe floods and large damages in Germany. However, the impact of an HPE is not solely determined by the event itself, but also by the geomorphologic characteristics of the location where it occurs.

Previous studies have shown that HPEs can happen anywhere in Germany. To find out where in Germany historical HPEs could have caused a potential hazard, we extracted the 10 most extreme HPEs by using the cross-scale weather extremity index (xWEI) from the last 20 years of radar data (RADKLIM) and shifted these events to every mesoscale subbasin in Germany.

We use the geomorphological instantaneous unit hydrograph as a simple screening tool to investigate the runoff concentration at the mesoscale and the following flood wave propagation in these subbasins as response to historical HPEs. While this method might not be sufficient to model precise discharge, it can be used to spot rapid increase in direct runoff and shed light on the peak development further downstream, depending on the spatiotemporal characteristics of the HPE. 

By using historical HPEs as benchmarks, our method can help to identify areas in Germany which are prone to flood hazard and assist to adjust mitigation measures accordingly.

How to cite: Voit, P. and Heistermann, M.: Downward counterfactual analysis of historical rainfall events in Germany, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-1241, https://doi.org/10.5194/egusphere-egu23-1241, 2023.

Convective rainfall extremes usually trigger due to their highly localised and intense input of mass and momentum ‘hot moments’ in water and matter cycling. Terrestrial systems then respond with strong Hortonian overland flow and erosion up to the formation of flash floods. While heavy precipitation events are characterised by multi-decadal variability, it is noteworthy that the largest observed floods in many rivers of Europe have occurred in the last three decades. Similarly, flash floods have also intensified. The recent clustering of extremes likely reflects the ongoing acceleration of the hydrological cycle, with expected increasing frequencies of intense convective rainstorms and related flash flood and erosion events due to Clausius-Clapeyron scaling. This urgently calls for an improved understanding and models that allow the design of strategies to mitigate onsite and catchment-wide offsite damages of flash floods and erosion events.

Hortonian overland flow occurs when precipitation intensity exceeds the soil’s infiltration capacity. The latter depends on the soil water content, soil hydraulic properties and the density and connectivity of vertical preferential flow paths and are often biologically mediated, as in the case of worm borrow and root channels. Whether locally generated surface runoff reaches the stream depends on the generated spatial connectivity of overland flow paths to the river network.

Here we propose that land use management and soil surface preparation bear the key to reducing the formation of Hortonian overland flow and the connectivity of its flow path, e.g., through a locally elevated infiltration capacity and roughness, thereby reducing the overland flow velocity and favouring its re-infiltration. Moreover, we demonstrate that physically based hydrological models are key to quantifying how changes in landuse and surface preparation techniques (including buffer areas, vegetation barriers, and fascines) in combination with local flood defense reservoirs reduce the formation of flood runoff during convective extremes. Specifically, we use the model CATFLOW and the representative hillslope approach to investigate flash floods observed in four ungauged headwaters catchments in the Kraichgau, Baden-Württemberg (Germany) in 2016. While each catchment drains into a regulated flood defense reservoir, we inverted the flood hydrograph/ inflow into the flood reservoirs using water level measurements and reservoir geometry equations. LULC maps are derived from LANDSAT images using spectral profiles obtained from field surveys over the region. Since flash floods are often associated with localised short-duration, high-intensity rainfall of convective origin, the model is forced using commercial radar-based precipitation products. The CATFLOW model was set up separately for the four headwaters by transferring a completed hillslope setup (soil catena, soil hydraulic properties, plant roughness parameters) from a gauged Weiherbach experimental catchment in the same landscape while deriving the representative hillslope profiles from the digital elevation data. Our results indicate that physically based models perform well in capturing the dynamics of the reconstructed hydrographs, which speaks a) for the transferability of physically based model structures within the same hydrological landscape and b) the feasibility of representative hillslope approach and c) the usefulness of the radar product.

How to cite: Manoj J, A., Villinger, F., Mälicke, M., Loritz, R., and Zehe, E.: Representative Hillslope Approach for Modeling Flash Flood Generation in Ungauged Catchments, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-6096, https://doi.org/10.5194/egusphere-egu23-6096, 2023.