Solar flare forecasting through data assimilation with avalanche models

Christian Thibeault ( Université de Montréal )

Sandpile models are lattice-based computational frameworks offering a powerful approach to the modelling of solar flares, encompassing a range of scales too wide to be accommodated by more conventional methods. In this talk, I will show that despite inherently stochastic components in these models, the presence of long-range temporal correlations implies that prediction of future flare behavior is possible given information on past flare events. This property, in conjunction with data assimilation, can be used to design forecasting schemes that significantly outperform climatological forecasts. I will end by presenting an application of this method to the forecast of large flares, working off a set of GOES X-Ray flux time series.