L'estimation des volumes des laves torrentielles: méthodes disponibles et application au torrent du Poucet (Savoie)

En ingénierie des risques torrentiels, la connaissance des volumes des laves torrentielles est indispensable. Cet exercice est difficile du fait notamment de la méconnaissance des processus de de´clenchement et de la complexité de ces écoulements. Par contre, les matériaux transportés constituent fréquemment des dépôts caractéristiques dont le volume peut être évalué et archivé. Un important gisement de données peut ainsi être valorisé par des approches globales. Cet article présente, de façon détaillée, trois types de méthodes globales et leur application à l'estimation des volumes de laves du torrent du Poucet (Savoie, France): à partir d'une analyse fréquentielle d'une chronologie d'événements historiques, à l'aide de modèles sommaires, et enfin au travers d'une approche géomorphologique. Comparés, les résultats montrent leur grande dispersion. S'appuyant sur les fondements de chaque méthode et des travaux antérieurs de validation, la conclusion préconise pour l'ingénierie une démarche les associant dans un ordre privilégiant analyse fréquentielle et approche géomorphologique.

The engineers dealing with the risks associated with torrential rainstorms often need to estimate debris flow volumes. This is a difficult task, notably because of the poor knowledge of triggering mechanisms and the complexity of flow processes. However, some debris flows form ''typical'' deposits whose volumes may be estimated and recorded, from which a significant amount of data may prove available for developing and testing methods for the assessment of debris flow volumes. This paper presents three methods and their application to a casestudy - the assessment of debris flow volumes for the Poucet torrent (Savoy, France): using frequency analysis carried out on a historical time-series; with the help of basic models; and using a geomorphological approach. For frequency analysis, a time-series of historical events is established. A volume is estimated for each debris flow, based on reports from archives when available, which may allow it to be placed in one of three classes of magnitude from qualitative descriptions of the events. Stationarity tests for the year of the occurrence of the debris flows are carried out and relevant time-intervals and volume thresholds selected for the frequency analysis. A Poisson distribution is carried out for the number of debris flow per year and an exponential analysis above threshold debris flow volumes. A value of 36,000 m3 was obtained for a 100-year debris flow volume in the Poucet torrent. Global models are empirical relationships for debris flow volumes computed from the morphometric and lithological parameters which influence debris flow generation. All of these are calibrated outside the French Alps, except for a recent tentative method. Results are widely scattered. Geomorphological approaches are based on the recognition of areas prone to high sediment accumulation. A less subjective approach has been suggested with a tentative standardization based on a classification of channel reaches according to their erodibility. Applying this to the Poucet torrent results in a volume of 110,000 m3 for an extreme event. The comparison between the results shows considerable variation. On the basis of the characteristics of the methods considered and previous validation work, it is concluded that, for engineering purposes, a combined approach with an emphasis on frequency analysis and geomorphological methods is used.