Time Series Theory Mutation Detection Method And Its Application

In the study of global climate change, abrupt change points detecting climate time series is an important content in the research of the climatic time series, nonlinear, non-stationary, makes the climate time series point mutation detection requirements of high. Especially in recent years, our country climate extreme events appear frequency increasing trend, therefore, the climatic time series with the detection and research of point mutations appears especially urgent and necessary.This paper introduces several commonly used time sequence mutation detection method; introduces the latest foreign proposed by RJMCMC method. And its application to Wuhan China nearly50years of temperature time series of mutation research, will be an annual average daily temperature exceeds a given threshold value of the total number of sequence is defined as a sequence of annual extreme high temperature days. The site of Wuhan summer annual extreme high temperature days sequence and winter annual extreme low temperature days sequence mutation points of research and analysis, the results showed:summer air temperature of Wuhan is gone from a low to high, again from high to low process from1951to2004. two point mutations are1958and1968,1958and1968be made the boundary, from1951to1958and1969to2004, the number of days appear extremely high relative to the1959-1968years to less,1959-1968days to the extremely high summer changes to a larger extent, belongs to the variant climate; The other two period belong to smooth line climate; And the extreme temperature of winter in Wuhan experience the process of "a high to low and low to high" from1951to2004,it has three point mutations, respectively, for the year of1958,1964and1970, with the three points to the boundary mutation.1951-1958and1965-1970in the frequency of extreme low temperature relatively more,1959-1964and1971-2004and the number of days in extreme low temperature is opposite less,1951-1958winter days and more extreme low partial moderate, belongs to the cold climate type;1959-1964days to the extreme winter scarce, belongs to the warm climate type; The other two extreme low temperature in winter days period have greater variation, belongs to the variant climate. Historical data indicate that:the1906-2000of temperature changes with the obvious stages and mutation characteristics in Wuhan, the temperature in the20th century, experienced "a low, high, low and high"4stages, and now at variable periods of high temperature. On the cyclical temperature changes, the Wuhan temperature change cycle in about12years, the temperature changes of Northern Hemisphere common cycle close to13years. Therefore, the results which obtained with the historical data are basically consistent.In the above research results, makes a simple prediction about the number of the winter extreme low temperature days in Wuhan, compare the predicted to the direct prediction methods, it shows that:RJMCMC algorithm-based prediction method is more effective than the direct prediction method.