Modeling annual extreme precipitation in China using the generalized extreme value distribution

Extreme precipitation events are the major causes of severe floods in China. In this study, four time series of daily, 2-day, 5-day, and 10-day annual maximum precipitation from 1951 to 2000 at 651 weather stations in China were analyzed. The generalized extreme value (GEV) distribution was used, to model the annual extreme precipitation events at each station. The GEV distribution was also modified to explore the linear temporal trends in the extreme events. The results showed that more than 12% of the stations have significant (p-value < 0.10) linear trends. Decreasing trends are mainly observed in northern China, and increasing trends are observed in the Yangtze River basin and northwestern China. The return periods of extreme precipitation have changed for stations with significant temporal trends. The 50-year event observed in parts of the Yangtze River basin, and northwestern China during 195160, has become a more frequent 25-year event in the 1990s. The spatial distribution of the return levels of the 651 stations are closely related to the climatic mean precipitation, and are influenced by the East Asian summer monsoon system (return levels are measures of extremity-for example, a 10 year return level is the value that can be expected to be exceeded on average once in every 10 years). The return period of extreme precipitation, that caused the 1998 severe floods in the Yangtze River basin, was also evaluated from a probabilistic perspective.