The Sun emanates a continous flow of charged particles (electrons and
protons) to its surroundings. This flow is called solar wind. The bursts
in the solar wind generate auroras and other space weather phenomena.
The amount of solar wind bursts is known to vary with 11 year periodicity
and to have a close correlation with the sunspot number. Individual bursts
can be associated with large phenomena (coronal holes, flares and mass
ejections) taking place in the solar atmosphere (corona). Thus monitoring
the solar surface has a crucial role in predicting auroras: when e.g. a
coronal mass ejection has been regonized from solar surface observations
its arrival to near-Earth space can be anticipated to happen within 1-2
days. Solar wind observations from satellites residing in the space near
the Sun-Earth line further confirm the predictions.
Observing the solar phenomena alone is not enough to make realiable
auroral predictions. It is namely the solar wind-magnetosphere interaction
processes and certain magnetospheric phenomena which finally dictate
whether the solar wind burst is effective in generating auroras or not.
These phenomena are known in general level and they can be modelled with
computer simulations, but some crucial details still lack of comprehensive
scientific understanding. Progress in this research is naturally of primary
importance for reliable predictions,
but the innovations in the area of artifical intelligence have been useful
also in the space weather research. The long time series of ground based
and satellite observations collected for research purposes can be used
in teaching neural networks which then can be used in different prediction