Performance of automated methods for flash flood inundation mapping: a comparison of a digital terrain model (DTM) filling and two hydrodynamic methods

Abstract

Abstract. Flash floods observed in headwater catchments often cause catastrophic material and human damage worldwide. Considering the large number of small watercourses possibly affected, the use of automated methods for flood inundation mapping at a regional scale can be of great help for the identification of threatened areas and the prediction of potential impacts of these floods. An application of three mapping methods of increasing level of complexity is presented herein, including a digital terrain model (DTM) filling approach (height above nearest drainage/Manning–Strickler or HAND/MS) and two hydrodynamic methods (caRtino 1D and Floodos 2D). These methods are used to estimate the flooded areas of three major flash floods observed during the last 10 years in southeastern France, i.e., the 15 June 2010 flooding of the Argens river and its tributaries (585 km of river reaches), the 3 October 2015 flooding of small coastal rivers of the French Riviera (131 km of river reaches) and the 15 October 2018 flooding of the Aude river and its tributaries (561 km of river reaches). The common features of the three mapping approaches are their high level of automation, their application based on a high-resolution (5 m) DTM, and their reasonable computation times. Hydraulic simulations are run in steady-state regime, based on peak discharges estimated using a rainfall–runoff model preliminarily adjusted for each event. The simulation results are compared with the reported flood extent maps and the high water level marks. A clear grading of the tested methods is revealed, illustrating some limits of the HAND/MS approach and an overall better performance of hydraulic models which solve the shallow water equations. With these methods, a good retrieval of the inundated areas is illustrated by critical success index (CSI) median values close to 80 %, and the errors on water levels remain mostly below 80 cm for the 2D Floodos approach. The most important remaining errors are related to limits of the DTM, such as the lack of bathymetric information, uncertainties on embankment elevation, and possible bridge blockages not accounted for in the models.

Hybrid Surrogate Model for Timely Prediction of Flash Flood Inundation Maps Caused by Rapid River Overflow

“…One of the strategies developed for obtaining IMs in a time interval compatible to the update cycles of real-time flood forecasting systems is based on the development of simplified 2D models such as the Height Above Nearest Drainage [9], AutoRAPID [10] and Safer_RAIN [11], which adopt the approach of "spilling and filling" digital elevation models at specific water heights. Despite providing realistic results for flood events of large scale with applicability demonstrated in real-time scenarios [12], the lack of physics for representing momentum results in limited applicability for rapidly-evolving flash flood events [13,14]. Other modelling approaches developed for reducing the computational burden of high-resolution models that do not neglect the momentum component include sub-grid techniques [15,16], the use of cellular automata frameworks [17], and the development of new systems and models specifically designed to explore the processing power of graphic processing units [18,19].…”

Comparison of Flood Inundation Modeling Frameworks within a Small Coastal Watershed during a Compound Flood Event

“…HAND and/or AutoRoute perform well as first order approximations of fluvial flooding (Afshari et al, 2018;Johnson et al, 2020). However, these low complexity models do possess less skill when compared to higher fidelity hydraulic models (Hocini et al, 2021). One of the more notable limitations of steady-state inland models such as HAND and AutoRoute is their limitations in coastal watersheds.…”

Inondations sur la Côte d’Azur : bilan hydro-météorologique des épisodes de 2015 et 2019

“…Ces études, couplant analyse météorologique, hydrologique et hydraulique, permettent, en combinant mesures sur le terrain et modélisation, d'évaluer les débits de pointe ayant transité sur des tronçons jaugés et non-jaugés. Différentes crues marquantes des dernières années ont ainsi été analysées : crues de l'Aude de novembre 1999 par , crues du Gard et du Vidourle de septembre 2002 par Gaume et Bouvier (2004) et Delrieu et al (2005), crue de juin 2010 dans la région de Draguignan par Payrastre et al (2019), crue d'octobre 2014 à Montpellier par Brunet et al (2018), crue de septembre 2015 dans l'Hérault par Brunet et Bouvier (2017), crue de l'Aude en octobre 2018 par Lebouc et al (2019) et Hocini et al (2021), etc. La compilation de l'ensemble des données produites lors de ces REX a permis, durant la dernière décennie, de constituer des bases de données des débits de pointe observés/estimés sur de petits bassins versants méditerranéens (Amponsah et al, 2018 ;Gaume et al, 2009 ;Marchi et al, 2010) et de les relier aux caractéristiques des bassins et à celles des épisodes météorologiques générateurs (e.g.…”

“…De plus, les débits de pointe estimés peuvent être utilisés dans le cadre de travaux de modélisation hydraulique d'emprise d'inondation, à la fois comme entrée de modèle et comme données de validation (e.g. DDTM 06, 2018 ;Kirstetter et al, 2020 ;Hocini et al, 2021).…”