Principal component and spectral analyses of palaeo-climate time series

摘要

Mathematical modelling and time series analysis techniques are important tools for extracting information from complex geotime series. These techniques also facilitate a fair degree of prediction, which is one of the prime goals of science. The data analysis strategy for such a purpose mainly involves spectral analysis and pattern classification. The aim of pattern classification and frequency analysis is to assign observations or patterns into semantic categories. Traditional statistical methods generally applied during the past years fail to recognize patterns from high dimensional georecords. Principal component analysis (PCA) is a powerful tool in identifying patterns in such records and provides useful means for reducing the number of dimensions without loss of much information. Here we have carried out spectral analysis and PCA of a climate record for approximately 28,000 yrs spanning from 1.15 to 29.78 kyr, off central Japan in the northwest Pacific. Our analysis reveals a dominant oscillation corresponding to the well known 'Heinrich Cycle'. The physical significance of the results has been discussed and the observed cyclic pattern corresponding to the global 'Heinrich Cycle' originating from the North Atlantic and Greenland ice rafting fluctuations has been linked to the Pacific phenomenon and Asian monsoon system.