Phenomenological Model of Solar Cyclicity

This work presents a phenomenological model of solar cyclicity that integrates the secular, 11-year, and quasi-biennial variations of solar activity. Using a two-stage modication of the Singular Spectrum Analysis (SSA) method with window lengths of 132 months and 28 months, we performed a decomposition of the monthly mean total sunspot number (SSN) time series. An amplitude modulation of the 11-year and quasi-biennial variations by the secular cycle has been identifed, which allowed us to construct an additive-multiplicative model of the interrelations between the components of the series. A correlation has been found between the duration of the 11-year cycle and the rate of change of the secular variation 4.5 years prior to its onset. By applying an LSTM neural network for the secular cycle, approximating the 11-year cycle with a superposition of lognormal functions, and tting the quasi-biennial variations with a superposition of Airy functions, a forecast for the 25th solar cycle was produced. The forecast correlates with observational data with a determination coeffcient of R2 = 0.79. The results confirm the effectiveness of the proposed approach for modeling solar cyclicity.