Previous: Introduction Next: Conclusions Up: Ext. Abst.

Results and Discussion

The time series analyzed were the annual averages, the 12 months running average (12m), 36 months running averages (36m) and 12m-36m. Spectral analyse were made using the Multiple Taper Method (Thomson, 1982), the Maximum Entropy Spectral Analysis ­ MESA (Ulrich and Bishop, 1975), and a Iterative Regression Model (Wolberg, 1967) to calculate amplitudes and phase of  periodicities.

Figure 1 shows the spectrum obtained with the MESA applied to the annual average time series.  It is observed that the more significant periodicities are 9.7 years (solar cycle), 2.36 years (QBO) and 4 years for Natal, and 13 and 7.3 years (solar cycle) and 2.4 (QBO) for   C. Paulista.

Fig.1 Spectrum obtained with the MESA method applied to yearly average data for Natal (right panel ) and C. Paulista (left panel).

Similar results were obtained  using Multiple Taper and Iterative Regression Methods. The solar cycle and the QBO cycle have been identified in the ozone time series, and the periodicities are: for QBO, 2.4 years for both stations; and for the solar cycle, near 10 and 8 years at Natal and C. Paulista.

The ozone average for the period of data analysed in this paper is 265.4 ± 4.3 DU for Natal and 269.3± 6.8 DU for C. Paulista.  The relative amplitude of the solar cycle signal were  1.0% of the ozone average for 9 years at Natal  and 1.5% for 7.8 years at C. Paulista. The relative amplitude of the QBO signal was 1.2 % for 2.4 years at Natal and 1.4% at 2.4 years at C. Paulista.

Filtering the data with a given smoothing technique could result in a QBO or solar signal more clearly seen. Table 2 shows the correlation coefficients calculated between total ozone monthly averages, 12m, 36m and 12m-36m time series and the sunspot number and zonal wind index time series.

Table 2 ­ Correlation coefficients

It is observed fromTable 2 that higher ozone and sunspot number correlations are observed at Natal than at C. Paulista, which  is in agreement with observations that the solar cycle is more easily seen at the equator (Zerefos et al., 1997). The 36m  time series has the higher correlation, because smoothing ozone data with these elliminates the  lowest periods. Ozone and QBO correlations are also higher for Natal, and at Natal the correlation is positive (r=+0.70, 12m-36m), and C. Paulista is negative (r=-0.58, 12m-36m), which is the pattern observed in equatorial and extratropical latitudes (Zerefos et al., 1992; Hollandsworth  et al., 1995; Kane et al., 1998).

In Figure 2 the 36m total ozone time serie and the sunspot number are shown for Natal and C. Paulista. It is observed that ozone and sunspot number have a variation very similar at Natal, and out of phase in C. Paulista. Cross-correlation spectral analysis have resulted in a lag of 0 years between ozone and sunspot number at Natal, and a lag of 2 years at C. Paulista (ozone leads sunspot number by 2 years).

Fig.2  Ozone 36 months running averages  time series and sunspont number for Natal (upper panel) and C. Paulista (lower panel).

In Figure 3 the zonal wind index at 30 hPa and the ozone anomaly time series  (12m-36m) are shown for Natal and C. Paulista. It is observed that at Natal ozone and zonal wind vary nearly in phase, and at C. Paulista a large out-of-phase variation is observed. Cross-correlation spectral analysis results in a lag between total ozone and zonal wind index of +2 months at Natal, and ­14 months at C. Paulista. So at Natal the ozone maximum is on average 2 months later than the zonal wind maximum, and at C. Paulista it has a maximum 14 months earlier than the zonal wind. These results are in agreement with those obtained by Hollandsworth et al. (1995), Kane et al. (1998), among others, using satellite data in latitudinal ranges.

Fig.3  Ozone 36 months-12 months running averages  time series and zonal wind index  at 30 hPa for Natal (upper panel) and C. Paulista (lower panel).