The Flare Likelihood and Region Eruption Forecasting (FLARECAST) Project: Flare forecasting in the big data & machine learning era

• DOI: 10.48550/arxiv.2105.05993
• Publication Date: 2021-01-01
• Authors: Manolis K. Georgo...; D. Shaun Bloomfield; Michele Piana; Anna Maria Massone; Marco Soldati; P. T. Gallagher; É. Pariat; Nicole Vilmer; É. Buchlin; F. Baudin; A. Csillaghy; Hanna Sathiapal; D. R. Jackson; P. Alingery; Federico Benvenuto; Cristina Campi; Kostas Florios; C. Gontikakis; C. Guennou; Jordan A. Guerra; I. Kontogiannis; Vittorio Latorre; Sophie A. Murray; Sung‐Hong Park; S. von Stachelski; A. Torbica; D. Vischi; Mark Worsfold
• Keywords: Flare; Space Weather; Solar flare

The EU funded the FLARECAST project, that ran from Jan 2015 until Feb 2018. had a R2O focus, and introduced several innovations into discipline of solar flare forecasting. were: first, treatment hundreds physical properties viewed as promising predictors on equal footing, extending multiple previous works; second, use fourteen (14) different ML techniques, also to optimize immense Big Data parameter space created by these many predictors; third, establishment robust, three-pronged communication effort oriented toward policy makers, space-weather stakeholders wider public. pledged make all its data, codes infrastructure openly available worldwide. combined 170+ (a total 209 are now available) in algorithms, some which were designed exclusively for gave rise changing sets best-performing forecasting flaring levels. At same time, reaffirmed importance rigorous training testing practices avoid overly optimistic pre-operational prediction performance. In addition, project has (a) tested new revisited physically intuitive (b) provided meaningful clues transition flares eruptive flares, namely, events associated with coronal mass ejections (CMEs). These leads, along algorithms infrastructure, could help facilitate integrated efforts take steps duplication. spite being one most intensive systematic to-date, not managed convincingly lift barrier stochasticity occurrence forecasting: thus remains inherently probabilistic.