Etude comparative des différentes méthodes pour la construction des Courbes Intensité-Durée- Fréquence, en insuffisance de données et sous différents étages bioclimatiques.

Abstract

This work lies within the scope of the research project, water resources management in view of the climatic changes. The construction of the IDF curves poses problem of data lack on very scattered zones. These curves IDF are at the base of any model rain flow in the studies of the floods, therefore it became necessary to define and develop tools making it possible to predict such events, or rather to predict their consequences on the urban environment, these last cover a significant aspect in the Algerian context, considering that all Algeria is touched by these extreme events which unfortunately are very badly estimated and which cannot be known. In this respect three bioclimatic changes have be retained like field of application of this comparative study, the first stage (sub-wet climate at winter moderates) represented by the city of (Gouraya), the second stage (arid semi climate at winter moderates) represented by the zone of full with the Means Cheliff (Chlef), and the third (arid semi climate at fresh winter) represented by the mountainous zone (Ouzera). To establish these IDF curves, different methods were tested for durations of aggregation varying between 15 min and 24 H. These methods are the traditional method, the global model of Koutsoyiannis and the marginal method, the first two methods are applied to two types of series (series of the annual maxima (SMA) and series in top of a threshold (SDP) and the third is applied to all the available recordings. It consists in estimating the distribution of the maximum intensity starting from the average distribution of the intensity. For the SDP, the SERIES were built by fixing a threshold by a mathematical choice with a study of the independence. The quantities generated with each site by these methods were calculated for a confidence interval of 95% for the following periods of return (2, 5, 10, 20, 50, 100 years). For the methods (traditional and Koutsoyiannis) that are applied to the SMA and SDP, the quantities are by the square root of the average quadratic error (rRMSE). The series SMA, SDP adjust well for a choice of the function of probability distribution (normal law Log). This last is checked by the test of Khi The quantiles estimated by the marginal method are not satisfactory considering the latter is applicable that in the case of insufficiency of the data (a limiting number of years).