The aim of this work is to complement a previous article by Vila (1998) presenting long-term observations of the behavior of seismic noise. We start by computing the Fourier transforms (FFTs) of time series (30-min length, one series per day) of the last 6 years of operation in an attempt to confirm whether the observations follow the pattern observed in the previous analysis. The analysis of the microseismic peak using 6 years of data confirms a long time (seasonal) variation of the amplitude, showing in addition the same variation in the frequency at which the maximum occurs. This points out the necessity of clarifying some of the observations in Vila (1998), and some modification in the conclusions is also necessary. A detailed analysis of the behavior of several spectral components of the seismic noise reveals that the seasonal component is presented in the frequency range 0.05-2.0 Hz. The amplitude of the seasonal component has a clear frequency dependence, being higher for frequencies close to the microseismic peak. The main differences between the spectral components are in accordance with its own amplitude level. No significant differences between the behavior of noise of the three components have been reported. Although for the geographical location of the station, the seasonal component is present in the atmospheric pressure, all attempts of correlation with seasonal components for diverse frequencies of the noise spectra have failed. This indicates that the pressure seasonal component cannot be related directly to the same component of the very long term noise spectral evolution.