Estimation régionale par la méthode de l'analyse canonique des corrélations: comparaison des types de variables hydrologiques

Abstract

It is often necessary to estimate extreme events at sites where little or no hydrometric data are available. In such cases, one may use a regional estimation procedure, utilizing data available from other sites in the same hydrologic region. In general, a regional flood frequency procedure consists of two steps: determination of hydrologically homogeneous region and regional estimation. This paper presents the development of a regional flood frequency procedure based on canonical correlation analysis and multiple regression: (i) the canonical correlation analysis allows us to link a set of hydrological variables and a set of physiographical and (or) meteorological variables and, therefore, to determine the hydrological neighborhood of an ungauged site; (ii) the multiple regression allows us to transfer data from sites of the homogeneous region to the target site. The developed methodology was applied to the Saint-Maurice river region (Quebec, Canada). Using a jackknife procedure, several pairs of hydrological variables were compared to identify the most appropriate grouping. Key words: regionalization, frequency analysis, canonical correlation analysis, multiple regression.