Akademik Veri Yönetim Sistemi
Tezin Türü: Yüksek Lisans
Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Fen Edebiyat Fakültesi, İstatistik Bölümü, Türkiye
Tezin Onay Tarihi: 2015
Öğrenci: NEVİN YAMAN
Eş Danışman: İNCİ BATMAZ, CEYLAN YOZGATLIGİL
Özet:Estimation methods on climate changes have become increasingly popular in the world over the recent years. They are useful for making comments about the future by using the past data related to temperature and precipitation. Especially, precipitation models, which are usefull for forecasting and simulation purposes, play a crucial role in forecasting climate changes. Estimations of daily rainfall amounts and occurrences found by using precipitation models are commonly used to generate scenarios of runoff, drought, flood, and so on. The main purpose of this study is to estimate the daily occurrence of rainfall and the daily amount of rainfall. For this purpose, daily amount of rainfall data from nine stations located at East Black Sea Region, one of the wettest regions of Turkey; located at Central Anatolian Region, one of driest regions of Turkey and vi Aegean Region, having a normal moisture climate in Turkey are modelled separately by using Hidden Markov Models (HMMs). HMMs are based on Markov Chains (MCs). The most suitable models are decided by comparing Akaike ınformation criterion (AIC), Bayesian ınformation criterion (BIC), mean square error (MSE) and misclassification (Error) rate (MR). It is observed that HMMs give good results for regions that has normal moisture climate compared with the wettest and driest region to estimate the daily precipitation occurrence. On the other hand, they give good results for the wettest region compared with the driest region or with normal moisture climate region to estimate the daily precipitation amount. Also, they successfully predict the most probable states that represents the daily precipitation occurrence by using Viterbi algorithm, when a sequence of observations and the model parameters are known. In this context, by using HMMs which is thought to be more effective than other precipitation models, the precipitation occurrence and precipitation amount are estimated in this thesis study. This work is the first phase to make estimations related to precipitation, providing very fast and less costly computations, and it gives general weather forecast and information about the unknown state of precipitation occurrences.