Abstract: The Singular Spectrum Analysis method,especially on the fundamental property of identifying the oscillatory compo-nents,is described.It is applied to a 30-year surface temperature series of east-southem china with a windows length of 100 days.Two main low-frequency oscillations with period of 35-40day and about 20 days are found to be robust,accounting for about 90% of total variance in the relevant frequency band.

Key words: chaos, time-series, singlar spectrum analysis