Extreme Events, Long-range Correlation

The extreme events show characteristics such as high devastating and hard to predict. Extreme climate events is a hot topic in the research work of climate change. Under the background of global warming, the extreme events of China shows characteristics such as high intensity and hard to predict. It is always believe that there are no relationship between the extreme events. But it has proved not correct in recent research. In this paper, relationship and prediction of the extreme events is mostly research, we try to study the rules of extreme events and return intervals from a brand new angle, and we use the observation data and it reveals that the similar law exists in the fact meteorological factors.In the first place, the paper takes the typical chaotic system-Lorenz system as our subjects. Then we use the detrended fluctuation analysis method to study the system's long-range correlation with different initial value and parameters. It turns out that the system's long range correlation is related with its initial value's phase space position. When the initial value is close to the unstable equilibrium points, the system's long rang correlation is strengthened, and the scaling exponentsαare bigger,the predictability of the system is better;While the system is in complete chaotic state, its long range correlation becomes weak with the parameters increased, and the scaling exponentsαare descended, the predictability of the system is weaker. It reveals the relationship between system's long range correlation and predictability. We disturb the system equation with random noise and find that the system's long range correlation decreased with the random noise's intensity augmentation. This outcome further turns out that Lorenz system's long range correlation is a quite effective physical as the predictability's token. The paper investigates the height and temperature fields reanalysis data of NCEP/NCAR in the means of detrended fluctuation analysis method, and gives their scaling exponent distribution of East Asia region. The empirical investigation indicates that both of height field and temperature filed have the property of long range correlation, and that match with the traits of space distribution as a whole. As for the grid data of the same layer, the long range correlation varies obviously when the latitude changes. The low-latitude areas near the equator have a bigger scaling exponent and a better long range correlation, whereas the mid-latitude and high latitude areas have a smaller scaling exponent and a worse long range correlation. The average scaling exponent increases with the layer growth of height field data, and the scaling exponent varies obviously when the latitude changes. The scaling exponent of Qinghai-Tibet Plateau is obvious greater than other areas with the same latitude. For the data of temperature, the average scaling exponent decreases with the layer growth then increases, and the average scaling exponent have a definite latitude trait. All in all, the scaling exponent of height field's long range correlation is greater than that of temperature field. The investigation for different seasons display that the long range correlation exists in height field and temperature field as well. And we find that the scaling exponent of winter is greater than that of other seasons. This may provide a theory for the summer flood season forecast which bases on the winter information.The paper has researched the long range correlation of extreme events of the Lorenz system used the method of fixed threshold. It turns out that all of the extreme events with different threshold have long range correlation. And the scaling exponents are similar just smaller than the original series. The long range correlation of extreme events is less effected when the initial value changes, but it is decreased distinctly when the parameters increased. The long range correlation of Lorenz system's extreme events series has the traits of memory when compared with Gauss white noise, and the memory is closed with the threshold. Finally, we use the maximal day air temperature data of 194 stations between 1957 and 2004 ,which from the National Climate Center of China Meteorological Administration, and it reveals that the similar law exists in the fact meteorological factors. In the view of return intervals, we have studied the long range correlation of return intervals when the extreme events were happened. And we found that the constructed extreme event time series also have the characteristic of long range correlation when the original series ones have, meanwhile exponential scale index of the two time series are very close. This is very different from the stochastic return intervals series. The probability density function of time series with long range correlation have essential difference with the stochastic one, and the extreme time series from the former take on accumulative phenomena. So we defined the cluster of the return intervals and found the long range correlation of the time series is the essential reason for their extreme event cluster, that is to say the correlation of inner units of the system cannot but lead to the cluster of the extreme events. We defined the index of extreme events'cluster and investigated the index's stability when the threshold value varied as well. Finally, we use the maximal day air temperature data of 194 stations between 1957 and 2004 which from the National Climate Center of China Meteorological Administration and it reveals that the similar law exists in return intervals of the fact meteorological factors. And the scaling exponential are existed a characteristic with obvious district distribution.