Volume 386, Number 1, April IV 2002
|Page(s)||313 - 318|
|Section||Planets and planetary systems|
|Published online||15 April 2002|
Forecasting the solar cycle with genetic algorithms
Instituto Mediterráneo de Estudios Avanzados (UIB–CSIC), Campus UIB, 07071 Palma de Mallorca, Spain e-mail: email@example.com, firstname.lastname@example.org, email@example.com
2 Departament de Física, Universitat de les Illes Balears, 07071 Palma de Mallorca, Spain e-mail: firstname.lastname@example.org
Corresponding author: R. Oliver, email@example.com
Accepted: 25 January 2002
In the past, it has been postulated that the irregular dynamics of the solar cycle may embed a low order chaotic process (Weiss 1988, 1994; Spiegel 1994) which, if true, implies that the future behaviour of solar activity should be predictable. Here, starting from the historical record of Zürich sunspot numbers, we build a dynamical model of the solar cycle which allows us to make a long-term forecast of its behaviour. Firstly, the deterministic part of the time series has been reconstructed using the Singular Spectrum Analysis and then an evolutionary algorithm (Alvarez et al. 2001), based on Darwinian theories of natural selection and survival and ideally suited for non-linear time series, has been applied. Then, the predictive capability of the algorithm has been tested by comparing the behaviour of solar cycles 19–22 with forecasts made with the algorithm, obtaining results which show reasonable agreement with the known behaviour of those cycles. Next, the forecast of the future behaviour of solar cycle 23 has been performed and the results point out that the level of activity during this cycle will be somewhat smaller than in the two previous ones.
Key words: Sun: activity / methods: numerical
© ESO, 2002
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.