Predicting Solar Cycles from 14C and 10Be Using Deep Neural Networks and Monte Carlo Simulations

Abstract

The Sun is the primary source of energy for the Earth system, and fluctuations in solar activity exert a significant external influence on the planet's climate. Cyclical variations in solar irradiance can be inferred by studying changes in the concentrations of carbon-14 (¹⁴C) and beryllium-10 (¹⁰Be) isotopes. During periods of high solar activity (i.e., more sunspots), the enhanced solar wind and magnetic field shield the Earth from cosmic rays, reducing the production of ¹⁴C and ¹⁰Be in the atmosphere. In this study, various methods to predict future sunspot activity were explored, focusing on deep learning algorithms to develop predictive models. The learning process is based on historical records of ¹⁴C, ¹⁰Be, and sunspot data. The results obtained are both promising and encouraging, highlighting the potential for further advancements in this emerging area of research.