Summary
International Symposium on Nonlinear Theory and its Applications
2009
Session Number:A1L-A
Session:
Number:A1L-A2
Singular Spectrum Analysis of Equatorial Precipitation Data
Naoki Itoh, Jurgen Kurths,
pp.-
Publication Date:2009/10/18
Online ISSN:2188-5079
Summary:
Singular spectrum analysis (SSA) is adopted to the time series of the monthly equatorial precipitation data observed at three stations, Nakuru [1904-1991], Naivasha [1950-1985], and Narok [1913-1991], in Kenya, to explain how climate works in tropical East Africa. By singular value decomposition (SVD) method in the SSA technique, basically, the bivariate precipitation data is mathematically decomposed and then a similarity between the two kinds of time series is discussed. A comparison of the data is performed by a heterogeneous correlation and an expansion coefficient. Annual structures obtained by estimates of this correlation show a similar form in each pair of the stations. 1-st expansion coefficient which explains the most dominant feature in the precipitation, shows relatively high values in rainy season of spring. In the next study some sub time series with background features can be represented by applying caterpillar-SSA to each single time series. A harmonic curve of 12 month cycle resulted in the 1-st mode by using this method can be interpreted as the most dominant characteristics in the time series. Such a result can be obtained at all the stations, i.e. there exists such a common seasonal cycle in Kenya. And results of the shorter cycles in a seasonal sense obtained from the other sub time series, can be interpreted as cycles of a rainy season. On the other hand, 15 month cycle is shown in the higher modes of the sub time series in Nakuru, which may be interpreted as an irregular period in the sense of annual cycle.