Abstract
Three methods are well known for estimation of spectral densities of descrete time series; Blackman-Tukey, Cooley-Tukey (FFT) and auto-regressive model (AR) methods. The fundamental properties of FFT and AR methods are reviewed for the methods of spectral analyses of turbulent winds since Blackman-Tukey method is equivalent to Cooley-Tukey method in the viewpoint of statistical errors of estimates. In this paper it is shown that the choices of record length and sampling frequencies by using the scale of turbulence and the nondimensional frequencies are very useful. Comparisons of these two methods are illustrated by numerical examples on observed data. FFT method has a few demerits which are not sufficiently refined by averaging and/or window operations, while AR method is preferable for estimating spectral densities of turbulent winds though these statistical errors are not defined.
