Mining Solar Big Data with the Flare Detective

• Activity Number: 433
• Type: Invited
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Henry "Trae" Winter III*+
• Companies: Harvard
• Keywords: Automated Tools ; Statistics ; Science ; Astronomy ; Sun ; Big Data

Abstract

The largest solar flares can release more than 10 million times the energy of every nuclear weapon created. Despite decades of study, it is still difficult to determine when a solar flare will occur. In the past, the solar telescopes had a very limited field of view and missed the majority of flares. Therefore, a small number flares were studied in detail to discern key features of individual flares. The Atmospheric Imaging Array (AIA) instrument launched onboard the Solar Dynamics Observatory (SDO) satellite drastically changed the nature of flare observations. The full Sun is now observed in amazing detail 24/7, providing ~3 terabytes of uncompressed data a day. To handle this huge influx of data, automated tools have been developed to detect and analyze events on the Sun. The "Flare Detective" records about 65 flare events a day or over 23,000 events over the mission lifetime so far. This has created an immense data set that has only barely been mined. In this talk, I will describe the Flare Detective and its use. I will also show how the relationships between large flares and "mini-flares" were discovered using the Flare Detective and simple statistical techniques.