44th COSPAR Scientific Assembly. Held 16-24 July, 2022

2022.

Vorobev D., Blumenau M., Fridman M., Хабарова О. В., Obridko V.

We show a possibility of automated detection of solar magnetic tornadoes, using the classic computer vision and deep learning methods. We define magnetic tornadoes, independently of their origin, as magneto-plasma objects in the solar corona in which a magnetic field is twisted. Typically, a whole magnetic tornado rotates resembling tornadoes in the terrestrial atmosphere. Meanwhile, there are also tornadoes in which only plasma flows upward along the magnetic field spiral but the whole structure just shakes. The lack of identified structures is one of many problems that prevent studying the physics of magnetic tornadoes and the processes associated with them. In particular, the filamentary rotating structures are well detectable only at the limb, while one can only make suppositions about their presence at the solar disk. Our method relies on analyzing SDO/AIA images at wavelengths 171 Å, 193 Å, 211 Å, 304 Å, to which several different algorithms are applied, namely, convolutional and recurrent neural networks and optical flow calculations. The new technique combines several approaches that are established in various fields of data analysis. Such an approach allows detecting the structures with sufficient accuracy and recall. For training objects, we used magnetic tornadoes previously described in the literature [e.g., Wedemeyer et al. 2013, ApJ; Mghebrishvili et al. 2015 ApJ] as well as newfound ones. Our method made it possible to detect those structures, as well as to reveal previously unknown magnetic tornadoes.

Keywords: Neural networks, solar physics, solar magnetic tornadoes