The Choice The Temperature Threshold Value Based On The Generalized Pareto Distribution

According to recent statistics, the economic losses caused by global climate change and related extreme weather and climate events have increased by an average of10times in the past40years. As far as China is concerned, due to extreme weather events and weather disasters caused by natural disasters accounted for70%of the above. The results show that the global warming caused by human factors almost the same as natural factors, even more than the amplitude of natural climate system. The frequency and intensity of the extreme weather event are changing. However, for a long time, there is a little research on extreme happened so that the field is as a weak link to the climate change research.As a series of the extreme climates happened, people try to make the accurate judgments and projections to the extreme events. To determine threshold value u is critical. It is a prerequisite to estimate parameters for further study. The article based on the generalized Pareto distribution theory to choice the threshold value. However, in the GPD model, how to select the appropriate threshold value is a problem. If threshold value is too small, the amount of data being analyzed will increase, estimator will be biased estimate. On the other hand, the threshold value is too large, the amount of data for analyzing will reduce, the bias of analysis will reduce, the data is too little to increase the variance of estimation. How to select the appropriate threshold, it is a trouble has been plagued with extreme researchers. At first, this article mainly introduced the three limit distributions of extreme value theorem and the correlation contents, then introduced modeling threshold excesses, based on the Generalized Pareto Distribution theory. Then, we introduce the parameter estimation method of penalty functions, estimated the parameters of density function of the generalized Pareto distribution. With the average remaining life figure and automatic selection method to choice the threshold value, compared the advantages and disadvantages of both methods, and applying to actual temperature data, get the desired results. Currently, the international climate change research in the extreme is the most common value of5%threshold as an extreme value. At the end of the article, we compared the threshold value by the average remaining life figure and automatic selection method with the selected threshold value of5%. The result is that the threshold value by the average remaining life figure and automatic selection method selected is better than the selected threshold value of5%.