Abstract: The time series of hydrometeorology has lost its consistency, because of the influences of climate change and human activities. Based on the daily temperature data during the period of 1960-2013 in 89 meteorological stations over Hunan Province, the annual mean maximum temperature (AMMT) and peak over threshold (POT) are selected to detect the inconsistent characteristics of the extreme high temperature indices. The results showed that there were significant inconsistencies in the AMMT series of 59 sites (66.3%) and the POT series of 23 sites (25.8%) out of 89 stations. The parameters of five extreme distributions GLO, GEV, GPA, GNO and PE3 are estimated by using the L-moments method. From the results of Cramer-von Mises (C-M) test, we found that the GNO distribution function can better fit the series of extreme high temperature indices in Hunan Province. The non-stationary series is modified by the reduction pathway. Based on the GNO model, we calculated the estimated return levels for both the modified and non-modified extreme series at different return periods and assessed the changes of the extreme series at three different return periods. Under the conditions of climate change, the AMMT series showed an enhanced trend of intensity and a shortened trend of the return period in north Hunan, central Hunan and southeast Hunan. However, the POT series showed similar frequency characteristics variation only in northern Hunan and southeastern Hunan.