Forecasting of CME and flare activity using Machine Learning
Abstract
Contrary to the Earth, the Sun does not have a dipolar magnetic field. The Sun is a giant ball of gaseous plasma and it rotates at different velocities in different latitudes. This differential rotation causes the twisting and stretching of its magnetic field. When it reaches a very high magnetic tension, the magnetic lines break and reorganize, releasing all the magnetic tension into the plasma, accelerating, heating and ejecting it from the Sun at very high speeds. The plasma ejected is called a Coronal Mass Ejection (CME) and it can reach the Earth in 4 days.Sometimes the tension is so high that the explosive release accelerates the ions and electrons of the plasma almost to the speed of light. Such intense and energetic explosions are called solar flares and can reach the Earth in less than 10 minutes. CMEs and solar flares can disrupt communications, damage satellite components, create high currents along high-tension lines and irradiate humans flying in space and across the poles of the planet. To prevent accidents and deterioration of assets in space and on Earth, we need to forecast the occurrence of CMEs and solar flares within a window of several days.Physical models have been used to study in detail the underlying causes of CMEs, to understand their build up and release, and to track their path from the Sun to the Earth. But a forecasting tool based on physical models is impossible today: the Sun is too big and too complex, it is impossible to perform a full simulation of the entire star, and it would take months to simulate only a few minutes. We need to rely on different tools for the forecasting of solar activity. Recent advancements in the field of Machine Learning (ML) have shown that complex pattern recognitions can now be made by computers using clever software techniques.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 24, 2019
- Source ID
- FA95501817010
Entities
People
- Giovanni Lapenta
Organizations
- Air Force Office of Scientific Research
- Katholieke Universiteit Leuven
- United States Air Force