Predicting Coronal Mass Ejections transit times to Earth with neural network

TL;DR

This study employs a neural network to predict CME transit times using initial velocity and flare location, revealing drag-like behavior and hemisphere-based differences, with an average prediction error of about 12 hours.

Contribution

It introduces a neural network model for CME transit time prediction using only two parameters, highlighting the drag effect and hemispheric asymmetries in transit times.

Findings

Transit time dependence on velocity shows drag-like pattern.

Acceleration changes to deceleration around 500 km/s.

Shorter transit times for western hemisphere flares.

Abstract

Predicting transit times of Coronal Mass Ejections (CMEs) from their initial parameters is a very important subject, not only from the scientific perspective, but also because CMEs represent a hazard for human technology. We used a neural network to analyse transit times for 153 events with only two input parameters: initial velocity of the CME, $v$, and Central Meridian Distance, CMD, of its associated flare. We found that transit time dependence on $v$ is showing a typical drag-like pattern in the solar wind. The results show that the speed at which acceleration by drag changes to deceleration is $v\approx$500 km s$^{-1}$. Transit times are also found to be shorter for CMEs associated with flares on the western hemisphere than those originating on the eastern side of the Sun. We attribute this difference to the eastward deflection of CMEs on their path to 1 AU. The average error of the NN prediction in comparison to observations is $\approx$12 hours which is comparable to other studies on the same subject.