Article Preview
Top1. Introduction
Over the last two decades, flash floods posed serious risk in sedimentary areas (Boardman et al., 2003; Evrard et al., 2007), especially on several populated outlets in the department of Seine-Maritime (Delahaye et al., 2001; Douvinet et al., 2013). Generated shortly after high rains, ranging between 50 and 100 mm in less than 6 hours, and occurring in small and dry valleys (< 20km2), these flash floods are characterized by a violent onset, a rapid rising time and a surge rushing down the main valley just a few minutes after rainfall have peaked. These floods present features quite similar to others occurring, for example, in western France (Auzet et al., 1995) or in Flanders (Evrard et al., 2007), but are really different to Mediterranean floods (Barrera et al., 2006; Schmitz and Cullmann, 2008; Ortega and Heydt, 2009; Morin et al., 2009). Even though these hazards threaten human lives (11 persons died over the period 1983-2005 in Seine-Maritime) and cause significant damage to infrastructures (from 0.05 to 14 millions of euros; Douvinet, 2008), predicting their time of occurrence and intensity remains delicate at larger scales for several reasons: measurements and field-based experimentations are rarely conducted in dry valleys; these phenomena are insufficiently documented and remain difficult to monitor as they produce destructive effects to measuring devices; the rarity of events and the long recurrence intervals render obvious statistical analysis and calibration of models (Ferraris et al., 2002); the short distances between source areas (runoff production) and risk zones (i.e. settlements) frequently surprises inhabitants in a few minutes; changes in velocity, roughness and water height introduce uncertainties in the estimation of peaks of discharge (Gaume et al., 2009) and hamper classical hydrological approaches.
To anticipate the spatial occurrence and areas at risk, we propose an original approach coupling GIS-data within a cellular automaton (CA) namely RUICELLS (Delahaye et al., 2001; Douvinet et al., 2013). This model is a triangular cellular automaton driven according to a set of three simplified and deterministic hydrological rules of flow pathways (Tarboton, 1997). We choose this deterministic approach for two reasons: it is easily transferable to basically any site (Coulthard & Van De Wiel, 2006; Ménard and Marceau, 2006; Van de Wiel et al., 2007), and CA modelling considers common physical characteristics of processes, whereas the statistical models based on inventories of past events are less transferable and only implicitly represent the impacts of processes (rather than the processes themselves) through the reliance on numerous records (Kappes et al., 2012).