Investigating the Kinematics of Coronal Mass Ejections with the Automated CORIMP Catalog

TL;DR

This paper introduces the CORIMP catalog, an automated method for detecting and analyzing coronal mass ejections in solar coronagraph data, improving speed and accuracy over manual methods and other automated systems.

Contribution

The paper presents a new automated CME detection catalog, CORIMP, utilizing advanced image processing and machine learning techniques to enhance CME analysis.

Findings

CORIMP effectively detects and tracks CMEs in LASCO data.

CORIMP shows comparable or improved performance relative to existing catalogs.

Unsupervised machine learning helps distinguish multiple CMEs occurring close together.

Abstract

Studying coronal mass ejections (CMEs) in coronagraph data can be challenging due to their diffuse structure and transient nature, compounded by the variations in their dynamics, morphology, and frequency of occurrence. The large amounts of data available from missions like the Solar and Heliospheric Observatory (SOHO) make manual cataloging of CMEs tedious and prone to human error, and so a robust method of detection and analysis is required and often preferred. A new coronal image processing catalog called CORIMP has been developed in an effort to achieve this, through the implementation of a dynamic background separation technique and multiscale edge detection. These algorithms together isolate and characterise CME structure in the field-of-view of the Large Angle Spectrometric Coronagraph (LASCO) onboard SOHO. CORIMP also applies a Savitzky-Golay filter, along with quadratic and linear fits, to the height-time measurements for better revealing the true CME speed and acceleration profiles across the plane-of-sky. Here we present a sample of new results from the CORIMP CME catalog, and directly compare them with the other automated catalogs of Computer Aided CME Tracking (CACTus) and Solar Eruptive Events Detection System (SEEDS), as well as the manual CME catalog at the Coordinated Data Analysis Workshop (CDAW) Data Center and a previously published study of the sample events. We further investigate a form of unsupervised machine learning by using a k-means clustering algorithm to distinguish detections of multiple CMEs that occur close together in space and time. While challenges still exist, this investigation and comparison of results demonstrates the reliability and robustness of the CORIMP catalog, proving its effectiveness at detecting and tracking CMEs throughout the LASCO dataset.