Solar activity modelled and forecasted: A new approach

A new approach of exploring, predicting and explaining solar activity is introduced. The Lund Solar Activity Model (LSAM) is based on new wavelet methods and a hybrid physics-based neural network. The model uses as input different kinds of indicators of solar activity. Different time scales of solar activity are selected with scalograms and ampligrams. The processes behind the variability are revealed with wavelet time scale spectra. How new solar laws could be discovered with neural networks and how solar theory could be coded into neural networks are then discussed. Finally, forecasts and explanations are described with LSAM.