Flare Prediction Using Photospheric and Coronal Image Data

摘要: The precise physical process that triggers solar flares is not currently understood. Here we attempt to capture the signature of this mechanism in solar-image data various wavelengths and use these signatures predict flaring activity. We do by developing an algorithm i) automatically generates features 5.5 TB image taken Solar Dynamics Observatory photosphere, chromosphere, transition region, corona during time period between May 2010 2014, ii) combines with other based on history a understanding putative processes, iii) classifies whether active region will flare within $T$ hours, where $T = 2 \mbox{ }24$ . Such approach may be useful since, at present time, there are no models available for real-time prediction. find when optimizing True Skill Score (TSS), photospheric vector-magnetic-field combined yields best performance, area under precision–recall curve, all helpful. Our model performance TSS $0.84 \pm0.03$ $0.81 2$ - 24-hour cases, respectively, value $0.13 \pm0.07$ $0.43 \pm0.08$ curve $T=2$ respectively. These relatively high scores competitive previous attempts prediction, but our different methodology extreme care task design experimental setup provide independent confirmation results. Given similar values across types reported literature, conclude can expect certain baseline predictive capacity using data. believe first vector-magnetic field as well multiple from corona, it points way towards greater integration diverse sources future work.