Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Titre : Estimation Of Turkish Precipitation Data Using Artificial Neural Networks And Wavelet Transformation Methods

Türkiye Yağış Miktarlarının Yapay Sinir Ağları Ve Dalgacık Dönüşümü Yöntemleri İle Tahmini

Auteur : PARTAL Y. Müh. Turgay

Université de soutenance : İstanbul Teknik Üniversitesi

Grade : Doctor of Philosophy (PhD) 2007

Résumé
In this study, wavelet transforms and Artificial Neural Networks (ANN) have been applied to estimate the daily precipitation of Turkish meteorological data. Each of the meteorological data considered as input for the ANN model was decomposed into wavelet sub-series (DWs) by Discrete Wavelet Transform (DWT). Then, ANN configuration is constructed with appropriate DWs as input, and original precipitation time series as output. Wavelet transform, which can produce a good local representation of the signal in both the time and frequency domains, provides considerable information about the structure of the physical process to be modeled and has positive effects on ANN modeling ability. Because of these reasons, coupling wavelets with ANN can provide significant advantages. In this study, wavelet transforms and artificial neural networks have been applied to estimate the daily precipitation on Turkish meteorological data for the first time. Meteorological data obtained from DMİ (Turkish state meteorological services) were investigated for this study. Homogeneous analysis was applied by five different homogeneous tests to data. Then 14 stations were selected for precipitation estimation. In this study, firstly, precipitation estimation was applied with three different algorithm of artificial neural networks. Later wavelet transforms and ANN has been applied to estimate the daily precipitation. The estimation performance of the wavelet-ANN model is more superior to comparing with the performance of a classically trained ANN model and multi linear regression model.

Présentation

Version intégrale (5,7 Mb)