The Applicability Research Of Approximate Entropy In Abrupt Climate Change Detection

Approximate entropy is a nonlinear dynamics index of quantitatively measuring the complex of the time series and it has been used for abrupt dynamic detection of time series, but the existing researches are only for the performance tests of ideal model time series and the applicability in the actual observations has not been systematically studied. In view of this, we develop the applicability of approximate entropy and test the performance of approximate entropy in the detection of abrupt climate change. Various trends exist in many observation data, such as periodical trends caused by seasonal variation, linear trends and polynomial trends brought about by global warming. In addition, observational data often contains noises, disturbances and some other false information. In order to research the applicability of abrupt climatic change detection of the approximate entropy, in present paper, firstly the effects of various trends and different noises on moving cut data-Approximate Entropy are investigated. Secondly, we use weather generator to simulate a large number of long time series of temperature and precipitation data and study the applicability of moving cut data-Approximate Entropy in abrupt changes detection in dynamics structures of main meteorological elements, such as temperature, precipitation, etc. At last, the abrupt climatic changes of the precipitation of China’s northwestern region and the PDO index are detected by moving cut data-Approximate Entropy. We test the abrupt detection performance of moving cut data-Approximate Entropy method in the actual observation data. The main results are as follows:(1) The detection results of moving cut data-Approximate Entropy are little affected by various trends. By doing a large number of numerical tests on model time series, we test the degree of effects of periodical trends, linear trends, second order polynomial trends and higher order polynomial trends in nonlinear ideal time series on abrupt change detection results of moving cut data-Approximate Entropy. It is indicated that the detection results of moving cut data-Approximate Entropy are little affected by these signals of stationary and non-stationary trends. (2) Moving cut data-Approximate Entropy has strong anti-interference ability for noises. By analyzing the effects of random spikes and Gaussian white noise on abrupt change detection results of moving cut data-Approximate Entropy, we find that the abrupt change detection results of moving cut data-Approximate Entropy are little affected by the number of random spikes accounted for the length of the original time series and the size of random spikes. It is indicated that moving cut data-Approximate Entropy has strong anti-noise ability for random spikes. When Gaussian white noise is added to the nonlinear ideal time series IS1of which the length is2000, for the different lengths of sliding subsequence, the critical value of SNR at which the abrupt change points can be detected is about SNR=22dB, that may has something to do with the size of the sample. It is showed that moving cut data-Approximate Entropy also has strong anti-interference ability for Gaussian white noise.(3) Whether for a large number of simulation data, or for the actual observation data, moving cut data-Approximate Entropy method can detect climate mutation effectively.The applicability of moving cut data-Approximate Entropy in abrupt changes detection in dynamics structures of different meteorological elements is studied. We use weather generator to generate three kinds of meteorological elements, which are daily maximum temperature, daily minimum temperature and daily precipitation, and build1000time series of three kinds of meteorological elements separately, which have abrupt changes in dynamics structures.1000time series of various meteorological elements are detected by moving cut data-Approximate Entropy separately. Then the ApEn series detected by moving cut data-Approximate Entropy are diagnostic analyzed by MTT method. The results show that vast majority of abrupt change points of the1000series of the three meteorological elements can be accurately detected by moving cut data-Approximate Entropy method, indicating that the moving cut data-Approximate Entropy method applies to abrupt changes detection in dynamics structures of three meteorological elements, such as daily maximum temperature, daily minimum temperature, daily precipitation, etc. Meanwhile, the abrupt changes detection results by moving cut data-Approximate Entropy method of daily precipitation data of the Yuzhong, Linxia meteorological observation station in Northwest China from1960to2006and monthly PDO index data from1960to2006are with the previous studies the same conclusion, which will further confirm the reliability of the moving cut data-Approximate Entropy to detect abrupt changes and lay a solid foundation for the wide applications of the present method in observational data.