The long-term variations of the occurrence frequency of the extreme rainfall events at five Taiwan stations are analyzed. The linear trend is determined by binomial logistic regression method. The change-point of the abrupt change during the period of analysis is determined by the nonparametric Mann-Whitney-Pittitt test and the parametric Bayesian analysis approach based on Poisson distribution. Significant increasing trend of the occurrence frequency of the extreme rainfall events, with the durations of 1, 3, 6, 12, 24, 48, and 72 hours, is detected in the data of Taipei station. At Hualien, a station located at eastern Taiwan, the increasing trend of the extreme rainfall events with short durations (≤6hr) is detected. No significant trend is detected at Taichung, Tainan and Taitung. A change-point significant at the 5% level is detected in Taipei's yearly frequency data of the extreme rainfall events with 7 different durations. Although the change-points of the events with different durations appear in different years, the change point of short-duration events occurred during 1982-84 and long-duration events occurred during 1976-77. The change-points of 10-year accumulative frequencies of the extreme 1-hr events are also detected. We find that 1983-1994 is a period with the most frequent extreme rainfall events in Taipei, Hualian and Taitung stations, while 1988-2003 is a period with the least frequent events in Tainan and Taichung. In other words, the rainfall intensity has decreased over south and western Taiwan, but increased over north and eastern Taiwan after 1979. The findings in this study suggest the importance of understanding the relationship between the decadal-scale variations in large-scale atmosphere and sea surface temperature and its influence on Taiwan local climate. The complex terrain effect may significantly modify the influence of large-scale circulation and result in distinct contrast in the local climate responses.
The long-term variations of the occurrence frequency of the extreme rainfall events at five Taiwan stations are analyzed. The linear trend is determined by binomial logistic regression method. The change-point of the abrupt change during the period of analysis is determined by the nonparametric Mann-Whitney-Pittitt test and the parametric Bayesian analysis approach based on Poisson distribution. Significant increasing trend of the occurrence frequency of the extreme rainfall events, with the durations of 1, 3, 6, 12, 24, 48, and 72 hours, is detected in the data of Taipei station. At Hualien, a station located at eastern Taiwan, the increasing trend of the extreme rainfall events with short durations (≤6hr) is detected. No significant trend is detected at Taichung, Tainan and Taitung. A change-point significant at the 5% level is detected in Taipei's yearly frequency data of the extreme rainfall events with 7 different durations. Although the change-points of the events with different durations appear in different years, the change point of short-duration events occurred during 1982-84 and long-duration events occurred during 1976-77. The change-points of 10-year accumulative frequencies of the extreme 1-hr events are also detected. We find that 1983-1994 is a period with the most frequent extreme rainfall events in Taipei, Hualian and Taitung stations, while 1988-2003 is a period with the least frequent events in Tainan and Taichung. In other words, the rainfall intensity has decreased over south and western Taiwan, but increased over north and eastern Taiwan after 1979. The findings in this study suggest the importance of understanding the relationship between the decadal-scale variations in large-scale atmosphere and sea surface temperature and its influence on Taiwan local climate. The complex terrain effect may significantly modify the influence of large-scale circulation and result in distinct contrast in the local climate responses.