Eruptive events of the Sun are often accompanied by the acceleration of energetic particles which can have significant impacts on Earth’s environment. However, the mechanism responsible for such phenomena is not sufficiently well understood to be able to predict their occurrence.Satellites and ground-based observatories probe the Sun’s photosphere, chromosphere and corona and are key in studying solar activity. Numerical models have attempted to bridge the gap between the physics of the solar interior and such observations. The next step for realistic simulations would be to forecast the short term evolution of the Sun’s photosphere.
The following work explores how notions borrowed from meteorology (i.e. data assimilation) and artificial intelligence (i.e. neural networks) could be used to forecast short term solar activity for space-weather modeling purposes. More specifically, I’ll describe our attempt at implementing data assimilation in a radiative MHD model of the Quiet Sun to forecast its evolution over a short period of time. However, not all physical quantities in the model are directly observable. For example, direct measurements of plasma motions at the photosphere are limited to the line-of-sight component. Multiple algorithms have been consequently developed to reconstruct the transverse component from observed continuum images or magnetograms. I’ll discuss how neural networks can be used in conjunction with numerical models of the Sun to recover the transverse velocity field (and potentially other unobservable quantities) in solar photospheric plasma from observational data.