engleski

Fourier vs. wavelet analysis: the case of the Adriatic Sea surface temperature

In this paper we compare two approaches (Fourier and wavelet) to spectral analysis of time series, using five-day averaged maps of remotely sensed Adriatic Sea surface temperature (SST) as an example. Both analyses have been performed in conjunction with empirical orthogonal function (EOF) analysis of the surface temperature fields. The Fourier power spectra of the first EOF mode (the dominant part of the overall signal, used to assess the difference in the approaches) were calculated using the Welch averaged periodogram. The Morlet mother wavelet was used in calculating the continuous wavelet transform (CWT). By using the CWT we were able to decompose 1-dimensional time series of the Adriatic SST variability into 2-dimensional time-frequency space, discerning the series` dominant modes of variability as well as changes in those modes over a 14-years period (1985-1998). The Fourier analysis singled out important, but by definition global, periodicities uniformly spanning the whole analysed period. In contrast, the wavelet analysis, besides annual and semi-annual harmonic, offered plenitude of other periods, with modal intensities changing in time.

Fourier analysis; wavelet transform; Adriatic; sea surface temperature

nije evidentirano