Abstract: The Multi-Taper method (MTM) of spectral analysis provides a means for spectral estimation with a low variance and high resolution.It is particularly well-suited for the diagnosis analysis of weakly signals with a time-depended am plitude and frequency ag ainst a highnoise background.We briefly introduced basic outline of MTM and apply this approach to global,Northern and Southern Hemisphere meansurface temperature time series.Results are as follows.(1) There is an significant low-frequency=40-70 year timescale oscilation superim pose on a linear warming trend in global and Northern Hemisphere annual meantemperature time series,which appears mainly contribution to fluctuation of warming process.(2) Decadal signal is considerably significant in Northern Hemisphere annual mean temperature time series,while interannual quasiperiods signals (periods of 5 year shorter) are signif icant in Southern Hemisphere meant emperature series.Reconstructed ENSO band and quasibiennial components of above time series each exhibit modulation both amplitude and frequency.(3) Compared to the Black-Tukey method and Maximum entropy method,the spectral estimations based on MTM are of high resolution and strong stability.