EUV Wave Detection and Characterization Using Deep Learning

Abstract

Coronal Mass Ejections (CMEs) are the most violent solar bursts. They cause severe disturbances in the solar–terrestrial space and affect human activities in many aspects, especially causing damage to high-tech infrastructure. It usually takes few hours for a CME to arrive at the Earth after eruption. Therefore, many efforts have been devoted to CME arrival time prediction, so that we have enough time to take action before a CME arrives at the Earth. For predicting CME arrival time, it is vital to detect the CME origin, arrival and departure speed in a coronagraph. It has been widely accepted that Extreme Ultraviolet (EUV) waves are associated with CMEs, so EUV waves are the signatures of CMEs as CMEs originate and traverse the solar disk, specifically for front-side CMEs. In this paper, two deep neural networks are developed to first detect EUV waves and then outline their wavefronts, giving early signatures of CMEs. Usually, CMEs are recorded by coronagraphs as they transit the corona, so our proposed method can obtain a certain time ahead compared with conventional CME forecasting. In addition, the parameters for describing EUV waves can be more easily deduced, benefiting the subsequent statistical analysis of CMEs. The experimental results demonstrate the effectiveness of the proposed model for detecting EUV waves and generating their outlines.