On signal-noise decomposition of timeseries using the continuous wavelet transform: Application to sunspot index

摘要

We show that the continuous wavelet transform can provide a unique decomposition of a timeseries in to 'signal-like' and 'noise-like' components: From the overall wavelet spectrum two mutually independent skeleton spectra can be extracted, allowing the separate detection and monitoring in even non-stationary timeseries of the evolution of (a) both stable but also transient, evolving periodicities, such as the output of low dimensional dynamical systems and (b) scale-invariant structures, such as discontinuities, self-similar structures or noise. An indicative application to the monthly-averaged sunspot index reveals, apart from the well-known 11-year periodicity, 3 of its harmonics, the 2-year periodicity (quasi-biennial oscillation, QBO) and several more (some of which detected previously in various solar, earth-solar connection and climate indices), here proposed being just harmonics of the QBO, in all supporting the double-cycle solar magnetic dynamo model (Benevolenskaya, 1998, 2000). The scale maximal spectrum reveals the presence of 1/f fluctuations with timescales up to 1 year in the sunspot number, indicating that the solar magnetic configurations involved in the transient solar activity phenomena with those characteristic timescales are in a self-organized-critical state (SOC), as previously proposed for the solar flare occurence (Lu and Hamilton, 1991).