CME Transit Time Prediction Based on Coronagraph Observations and Machine Learning Techniques

A coronal mass ejection (CME) is a significant release of plasma and accompanying magnetic field from the solar corona, and is one important source of severe space weather events. With the accumulation of CME observations by coronagraphs and the excellent performance of convolutional neural network (CNN) in image classification, fast and accurate prediction of the transit (arrival) time of CMEs became possible. In this study, we present a new prediction method utilizing both a deep learning framework and the physical characteristic of CMEs based on remote-sensing observations. In total, the initial and arrival data of 168 geo-effective CME events from 2000-2020 are collected for the study. A convolutional neural network model is trained with the coronagraph images of the events observed by SOHO/LASCO. The output of the trained CNN is further combined with the initial CME speed to carry out a linear fitting process. The comparison with the prediction results merely based on a CNN or a linear fitting by CME speed indicates that the hybrid model can improve the accuracy of CME arrival time prediction.