Detector for fast wave trains in the solar radio emission

TL;DR

This paper develops an automatic neural network-based detector to identify fast wave trains in solar radio data, enhancing diagnostics of solar energetic events.

Contribution

It introduces a machine learning method to automatically detect quasi-periodic fast propagating wave trains in radio spectra, expanding diagnostic capabilities.

Findings

Detected 50 QFP candidate events in radio data.

Found 13 candidates associated with global EUV waves.

Demonstrated effectiveness of neural networks in solar radio event detection.

Abstract

Quasi-periodic fast propagating (QFP) wave trains observed in the solar corona after some energetic events (solar flares, coronal mass ejections, jets) open possibilities for diagnostics of spatial and temporal scales of the impulsive energy release processes, that are absent in the standard model of a solar flare. Besides, the dynamics of the wave trains and their characteristic spatial and temporal signatures allow to localize the initial energy release volume magenta and to perform fine diagnostics of the transverse structures of plasma inhomogeneities in the solar corona. However, the small number of such events registered significantly limits their promising diagnostic potential. The aim of this paper is to perform an automatic search for fast wave trains in radio data. We apply classifying neural network/machine learning methods. Dynamic radio spectra obtained by HiRAS radio spectrographs within the 20 MHz -- 2.5 GHz frequency band during 2011 were used. We consider 50 global coronal EUV waves as marker events for more a targeted search in HiRAS data. Our automatic detector revealed 50 independent QFP-candidates events with the temporal signatures similar to that of the fast wave trains, with 13 candidates connected with the global waves.