APPROCHE REGIONALE POUR L’ESTIMATION DES PRECIPITATIONS JOURNALIERES EXTREMES DU NORD EST ALGERIEN

Abstract

Information on precipitation magnitudes and their frequencies is needed in various projects related to water. In this study, the regional frequency analysis based on L-moments was used to improve the quality of estimating of the annual maximum daily precipitations of the north eastern area of Algeria represented by 50 measuring stations. The investigated area was divided into two regions statistically homogeneous in terms of L-moments ratios. Among the different tested distributions, (i) the Generalized Extreme Value (GEV) distribution has been identified as the most appropriate regional distribution for modeling precipitation in the northern region (region 1) characterized by a Mediterranean climate and (ii) the generalized lognormal distribution (LN3) for region 2 of southern part subject to semi-arid climate. Growth curves, derived from the regional distributions, were established for each homogeneous region. The growth curves shapes difference reflects the difference between the two homogeneous regions precipitation patterns. Therefore, to estimate the different return period’s precipitation quantiles in a given site of the region, the mean precipitation of the site has to be multiplied by the corresponding regional quantile. Comparison of estimated quantiles given by the regional and local approach showed that for 50.4 % of stations, atsite model underestimates the quantiles having high return periods.