This paper has two primary purposes. First, we fit the annual maximum daily rainfall data for 6 rainfall stations, both with stationary and non-stationary generalized extreme value (GEV) distributions for the periods 1911-2010 and 1960-2010 in Taiwan, and detect the changes between the two phases for extreme rainfall. The non-stationary model means that the location parameter in the GEV distribution is a linear function of time to detect temporal trends in maximum rainfall. Second, we compute the future behavior of stationary models for the return levels of 10, 20, 50 and 100-years based on the period 1960-2010. In addition, the 95% confidence intervals of the return levels are provided. This is the first investigation to use generalized extreme value distributions to model extreme rainfall in Taiwan.
# 13-004/III (2012-01-07)
- Lan-Fen Chu, National Science and Technology Center for Disaster Reduction (NCDR); Michael McAleer, Erasmus University Rotterdam, Kyoto University Japan, Complutense University of Madrid; Szu-Hua Wang, Chinese Cultural University, Taiwan
- Generalized extreme value, Extreme rainfall, Return level, Statistical modelling
- JEL codes:
- Q54, Q51, Q57