Threshold Model And Its Application In Extreme Low Temperature

In recent ten years, frequent extreme weather events around the world has caused numerous catastrophic consequences to the world. At present, the extreme climate has become one of the focus that each government, the public and the scientific community pay attention to. Early at the end of the20th century, many international scholars ever called for "the strengthening of long-term observation of extreme weather and climate events and variability in prediction of future variability", which also aroused the attention of global climate scientists. Around the world, due to extreme weather events caused by natural disasters, human’s life has been influenced and the impact of production rise, while economic losses are even more alarming. For example, there are all over the world drought, flood, severe cold, high temperature and other extreme weather events. Their frequent occur always causes very serious natural disasters, which results in the rising number of deaths. According to recent statistics, in the past three to four decades, losses caused by global climate change,extreme weather and related events have rose to an average of10times than before. As for China, natural disasters caused by extreme weather and meteorological disasters accounted for75%of the whole damage. The study also shows that global warming caused by human activity is more than the amplitude of natural climate system. Especially in recent years, the frequency and intensity of extreme weather events change a lot. However, few studies on extreme weather existed so far. Thus the research on the climate change field is a weak link. Especially in the last three years, the global frequency of extreme weather events and extreme weather have become a hot research issue. Generally people used to apply extreme value theory to dealing with such problems.At present, we mainly have two kinds of models in the extreme value analysis. One kind is extreme value theorem model (EVT), which is mainly used to calculate in the maximum value modeling. The extreme value theorem has guaranteed that the limit distribution is belonged to one of the three extreme value distributions which are termed as the Gumbel,Frechet and Weibull distribution, or may use their general form-generalized extreme value (GEV) family of distributions; Another kind is the generalized Pareto distribution model, the model is used when data exceed some high threshold. At first, this paper introduced the three limit distributions of extreme value theorem and the correlation contents, and introduced modeling threshold excesses, minimum value is derived theoretical model of the generalized Pareto distribution by the generalized Pareto distribution of maximum value; and then introduced some related theories of extreme value index estimate as well as several methods of threshold value selection:the method of mean residual life plot; the method of the stability of parameter estimator; the method of Hill estimate; the method of subsample bootstrap. The threshold in this paper is selected by the mean residual life plot method combined with the parameter estimation method to find the stability threshold. Finally, the actual temperature data with the theoretical threshold model is given in extreme low temperatures, obtained very good results.