New possibilities of the PCA-Seq method in the analysis of time series (on the example of solar activity)

Abstract

Abstract When analyzing a 1D time series, it is traditional to represent it as the sum of the trend, cyclical components and noise. The trend is seen as an external influence. However, the impact can be not only additive, but also multiplicative. In this case, not only the level changes, but also the amplitude of the cyclic components. In the PCA-Seq method, a generalization of SSA, it is possible to pre-standardize fragments of a time series to solve this problem. The algorithm is applied to the Anderson series – a sign alternating version of the well-known Wolf series, reflecting the 22-year Hale cycle. The existence of this cycle is not disputed at high solar activity, but there are doubts about the constancy of its period at this time, as well as its existence during the epoch of low solar activity. The processing of the series by the PCA-Seq method revealed clear oscillations fluctuations of almost constant amplitude with an average period of 21.9 years, and it was found that the correlation of these oscillations with the time axis for 300 years does not differ significantly from zero. This confirms the hypothesis of the existence of 22-year oscillations in solar activity even at its minima, like the Maunder minimum.