Determining the threshold of extreme events with detrended fluctuation analysis

Abstract

When the dynamic system is in the extreme states， or it is affected by an intensive disturbance， we will get an extreme event. The detrended fluctuation analysis (DFA) measures the long-term correlation of a system， so the extreme events has no effect on systems long-term correlation. Based on this idea， we determined the threshold of extreme events with detrended fluctuation analysis and the results testify its effectiveness. By this method， we analyzed the exreme events in Beijing for the temperature and rain. The results show that before the metaphase of the 1970s， the frequency and inensity of the extreme temperature events is mitigatory， and from 1970s to 1990s， there were very few extreme events. But from the metaphase of the 1990s to now， the extreme temperature events are frequent. The extreme low temperature event has a downtrend from the 1950s to now and never happened after 1978s. The frequency and intensity of extreme precipitations are periodic. The frequency and intensity were high in the metaphase of 1950s and 1960s, and were low in the remaining period. The transition occurred in the 1970s.