Extreme Value Modeling of Maximum Temperature in Upper Northeastern of Thailand

Abstract

The purpose of this research is to find the extreme value modeling of maximum temperature in the upper northeastern of Thailand from 9 meteorological stations using the generalized extreme value distribution. The parameter estimations are obtained from maximum likelihood estimation method under 6 different stationary and non-stationary process settings. The model selections are obtained by 3 criterions: Akaike’s information criterion (AIC), Bayesian information criterion (BIC) and likelihood ratio test (LRT). The return levels of annual maximum temperature are calculated. Moreover, profile likelihood method is applied to calculate the confidence interval of return levels. The results of this study indicate that the 5 meteorological stations were appropriated for stationary and the 4 meteorological stations were appropriated for non-stationary when the scale parameter changed with the fitted trend. When the return levels are considered, Udon Thani meteorological station had a higher return level than other stations for each return period. Hence, this station should be more emphasized. Keywords : generalized extreme value distribution, maximum likelihood estimation, return level,     profile likelihood method