Previous: Introduction Next: QBO and surface air temperature response to solar activity Up: Ext. Abst.

Quasibiennial oscillations in sunspot numbers and stratospheric winds

Different authors have found “similar” but not exactly equal periodicities in both solar and geophysical parameters [4-10]. This differences may be due to the fact that time series with different discretizartion and lengths have been used, or that the processes are independent. To compare correctly the periodicities in solar activity and 30 hPa QBO [11], in the present study the time series have been reduced to equal lengths - 552 monthly values, from January 1953 to December 1998, and the same filtering procedure has been applied. The quasibiennial oscillation in solar activity is very weak compared to the much more powerful 11-year, 5-7 year and other cycles, and an appropriate filtering procedure must be applied in order to identify it. To this end, we have designed a digital bandpass filter with finite-duration response using Fourier approximation of the desired transfer function and a Kaiser window [12]. As the period studied is relatively short, some compromise had to be found between the quality of the filter and the inevitable reduction of the time series. The input parameters we finally chose were low cut frequency f_L = 0.02 month^-1, high cut frequency f_H = 0.05 month^-1, ripple d = 0.05 and transition width D f = 0.02 month^-1. Thus we obtained a 63-point filter, and the time series length decreased at each side by 32 points. The quality of the filter was tested on modeled time series with predefined periodicities of 21.3 and 26.5 months with equal amplitudes, 66 months with a two times higher amplitude, and 132 months with a five times higher amplitude. The Fourier analysis of the filtered time series yields peaks at 21.3 and 27.2 months and a much lower broad peak at 70.0 months - Fig.1.

In Fig.2 the raw QBO data are compared to the filtered values. The amplitudes of the filtered values are slightly shifted with respect to the raw ones, however no periodicities are lost and none are added.

The Fourier analysis of the QBO data gives a single peak at 28.44 months, and in sunspot numbers periodicities at 28.44 months, 23.27 months and 39.40 months are found - Fig.3. If these different peaks do not correspond to independent oscillations, and taking into account the discrete character of the oscillation, their existence indicates that the oscillation has a variable duration [13]. Besides, it has been shown that the period of QBO is not constant either. Therefore, a moving normal window with a width of 128 months (corresponding approximately to one 11-year solar cycle) has been applied with a step of 64 months, looking for periods between 20 and 50 months. In Table 1 the periods of the peaks are presented together with the periods corresponding to 0.7 of the amplitude of the respective peak (in brackets).

Table 1

In some time intervals (April 1966-November 1976, April 1982 - November 1992) the peaks coincide, in others (December 1960 - July 1971, December 1976 - July 1987, August 1987 –May 1996) they partly overlap, and in still others (August 1955 - March 1966, August 1971 - March 1982) stratospheric winds exhibits clear quasibiennial oscillation, while no such periodicity is obvious in sunspot numbers. Therefore, it does not seem probable that the sunspot activity quasibiennial oscillations are an indicator for QBO in stratospheric winds.