APPLICATION OF LOCAL NONLINEAR DIRECT FORECASTING METHOD TO DETECT OCEAN CHAOS AND NOISE
Jiwei Tian · Haiyang yu huzhao · 2000
This paper gives a method for detocting chaos and noise in chaotic time series and in time series with white noise by using a local nonlinear direct forecasting method and the coefficient of correlation between predicted and actual values. The method is applied to Logistic map data. the equator day SST data and data with noise. The results of application show that the method is better than general spectral analysis in detecting chaos and noise. When the data is contaminated by additive noise 34% in proportion to that of the signal, the correlation coefficient is about 0.5. Within this creditable correlation coefficient, the chaotic phenomenon of the time series is detected by the method. When the rates of noise-signal increase, the correlation coefficients decrease and the noise character increases. The correlation dimension and the Kolmogorov entropy can also be determined by the method. Our results showed that the correlation dimension and Kolmogorov entropy of equator day SST data were about 5 and 0.14(1/d), respectively. Use of the above results yielded the main conclusions below. 1) The application of nonlinear direct forecasting to detect the deterministic chaos in natural signals with noise is valid. 2) The time series of the equator day SST is chaotic and contaminated by some additive noises. 3) The higher the noise-signal ratio is, the smaller the corresponding correlation coefficient is.