Skip to Main content Skip to Navigation

Possibility and usefulness flood series reconstruction, for the study of extreme floods in small watersheds. Case study of four small tributaries of the Aude river, France.

Olivier Payrastre 
Abstract : This study deals with the estimation of flash flood frequency, in small watersheds of the French Mediterranean area. In order to improve statistical analysis of these events, the idea developped was to gather, as made as possible, historical information concerning floods over a two century period. Four small tributaries of the Aude river (south-east of France) were studied. Each of them has a catchment area of less than 200 km2. For historical information gathering, a methodology that already gave successful results in the case of larger French rivers was applied. Using this methodology it was possible to identify archive sources that were of real interest to the particular case of small watersheds. These sources are essentially local archive collections ("archives départementales", archives from hydrometry and flood warning services). On the other hand, national archive collections ("Archives Nationales", M.Pardé collection, and the ENPC's historical collection) contained little useful information, and archives from newspapers were equally disappointing. Archive documents that were finally used, come essentially from the "Services des Ponts et Chaussées" or the "Services Vicinaux" (archives located in the S and O series in the Archives Départementales collection). Nevertheless, in the four cases studied these documents gave sufficient information to build historical flood series of one to two centuries in length. The perception thresholds associated with these series, and also the peak discharge of the main historical floods, could be estimated. Considering the large uncertainty of these estimations, discharges were represented in the form of possible upper and lower bounds. The statistical study of these series first reveals the high heterogeneity of flood peak discharges among the four studied watersheds. The usefulness of historical data is then highlighted, for the selection of appropriate statistical distributions for the series, as well as for the fitting of these distributions. As only short systematic series are used, the application of such methods, which consist of fitting a statistical distribution to the observed series, can lead to very important errors. Historical data make these methods more reliable, even in the case of very limited information (number of times a threshold is exceeded, or information limited to just one historical flood). In comparison, the Gradex method appears much more robust when only systematic data is available. This method gives flood quantile estimates that are largely coherent with historical data. Moreover, the variability of flood discharges among the four case studies is well represented using this method, due to high variations of the shape factor (this factor must be carefully estimated in each case). Finally, in order to extend these results to other watersheds, especially ungauged ones, the reasons for this variability on flood discharges must first be better understood.
Document type :
Complete list of metadata
Contributor : Ecole Des Ponts Paristech Connect in order to contact the contributor
Submitted on : Thursday, June 15, 2006 - 8:00:00 AM
Last modification on : Friday, October 23, 2020 - 4:52:03 PM
Long-term archiving on: : Thursday, September 30, 2010 - 7:23:13 PM


  • HAL Id : pastel-00001792, version 1



Olivier Payrastre. Possibility and usefulness flood series reconstruction, for the study of extreme floods in small watersheds. Case study of four small tributaries of the Aude river, France.. Sciences of the Universe [physics]. Ecole des Ponts ParisTech, 2005. English. ⟨pastel-00001792⟩



Record views


Files downloads