Application de l'analyse spectrale à la description séquentielle des paramètres thermo-halins dans une région d'upwelling

Continuous sampling is now widely used in oceanography and requires new tools for data processing. The theory of stochastic processes and especially spectral analysis offers powerful ones. We are dealing in this paper with bivariate time series : temperature and salinity were continuously measured during the CINECA-CHARCOT II cruise, in the upwelling area of North-West Africa. The record has been splitted into three parts corresponding to straight lines. Those time series have been prewhitened using the first finite difference filter, spectra and cross-spectra were smoothed using the Parzen window. Power spectra reveals obvious peaks. Coherencies are higher in low frequencies. The basic phenomena, mixing due to the upwelling (transfer functions can be related to the slopes of T-S diagramms), is disturbed by a noise stronger in high frequencies. [NOT CONTROLLED OCR]