Segmentation and Tracking of Eruptive Solar Phenomena with Convolutional Neural Networks

2025

Read Segmentation and Tracking of Eruptive Solar Phenomena with Convolutional Neural Networks (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-05623.txt.

Introduction

Eruptions on the Sun are cut out and tracked by neural networks trained on masks from an older algorithm, in a 2025 solar physics paper by Oleg Stepanyuk and Kamen Kozarev. Coronal mass ejections, shock waves, flares, and filament eruptions unfold together. Decades of data now demand automated handling.

The earlier Wavetrack method, built on wavelet splitting and brightness ranking, had already described eruptive features across several data types. Here deep networks learn segmentation from Wavetrack's masks, easing the training set logjam that limited data-driven tracking. Pre-trained models are shown on eruption events from a space imaging instrument. The authors present the step as a natural next step for their pipeline.

Catalog record

YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-05623

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Segmentation and Tracking of Eruptive Solar Phenomena with Convolutional Neural Networks Abstract Solar eruptive events are complex phenomena, which most often include coronal mass ejections (CME), CME-driven compressive and shock waves, flares, and filament eruptions. CMEs are large eruptions of magnetized plasma from the Sun’s outer atmosphere or corona, that propagate outward into the interplanetary space. Over the last several decades a large amount of remote solar eruption observational data has become available from ground-based and space-borne instruments. This has recently required the development of software approaches for automated characterisation of eruptive features. Most solar feature detection and tracking algorithms currently in use have restricted applicability and complicated processing chains, while complexity in engineering machine learning (ML) training sets limit the use of data-driven approaches for tracking or solar eruptive related phenomena. Recently, we introduced Wavetrack - a general algorithmic method for smart characterization and tracking of solar eruptive features. The method, based on a-trous wavelet decomposition, intensity rankings and a set of filtering techniques, allows to simplify and automate image processing and feature tracking. Previously, we applied the method successfully to several types of remote solar observations. Here we present the natural evolution of this approach. We discuss various aspects of applying Machine Learning (ML) techniques towards segmentation of high-dynamic range heliophysics observations. We trained Convolutional Neural Network (CNN) image segmentation models using feature masks obtained from the Wavetrack code. We present results from pre-trained models for segmentation of solar eruptive features and demonstrate their performance on a set of CME events based on SDO/AIA instrument data. † † journal: JGR: Machine Learning and Computation † † authors: Oleg Stepanyuk, Kamen Kozarev † † corresponding: Kamen Kozarev, [email protected] Institute of Astronomy and National Astronomical Observatory, Bulgarian Academy of Sciences, Sofia 1784, Bulgaria keypoints We present our general approach to efficient data-driven solar eruption image segmentation of eruptive solar phenomena. We performed a hyperparameter search for the optimal network architecture and trained U-Net-based models for segmentation and tracking. We demonstrate the performance of our models on coronal bright front imagery observed in extreme UV data from the SDO/AIA telescope suite. Our approach has the potential to significantly advance the automated analysis of high-dynamic range solar data. Plain Language Summary Our study focuses on enhancing detection and analysis of imaging observations of solar eruptive events, which is important for predicting space weather impacts on satellites and communication systems. Segmentation and tracking of solar eruptive events based on ground based and remote instrument data is essential in pursuit of better answers about the physical nature of these phenomena. A data-driven approach appears as an obvious solution; nevertheless, engineering of training sets is a challenging task as obtaining labeled ground truth masks requires significant amount of manual work. Existing algorithmic tracking software does not allow to automate the process significantly, as each event or set of the images usually requires separate re-configuration of the software. We used our previously developed wavelet-based algorithmic method, which to a certain extent simplifies the segmentation of solar images, together with synthetic data for engineering of training sets. We describe several image segmentation models and demonstrate their performance, and generally discuss various aspects of high-dynamic range images segmentation in heliophysics domain. 1 Introduction Coronal Mass Ejections (CMEs) are large expulsions of plasma and magnetic field from the solar corona into the heliosphere. CMEs travel outward from the Sun at speeds ranging from slower than 250 km/s to more than 3000 km/s ( Yashiro et al. (2004) ) . Due to their ability to generate large Solar Energetic Particle (SEP) events and severe geomagnetic storms, CMEs are regarded as a significant natural hazard that can cause disruptions in the interplanetary environment, with potential impacts on spacecraft, satellites, and communication systems. Since their discovery, CMEs have been associated with a host of solar activity processes ( Hansen et al. (1971) ; Tousey et al. (1973) ) . They are known to be formed by explosive reconfiguration of solar magnetic fields through the process of magnetic reconnection, but broad theoretical discussion aiming to shed light on the exact formation mechanism still continues ( Shibata & Magara (2011) ; P. Chen (2011) ) . To accurately identify and measure the size of the erupting bubble and the subsequent flux rope, high-cadence, high-sensitivity off-limb extreme ultraviolet (EUV) and white light observations have proven very useful ( Carley et al. (2020) ; Jiang et al. (2021) ) . Due to their increasingly large volume, these observations must be accurately processed which sets demands for development automated pipelines and more advanced interpretation techniques. The launch of the Solar Dynamics Observatory ( Pesnell et al. (2012) , SDO) , with its high-resolution, high-cadence EUV telescope Atmospheric Imaging Assembly ( Lemen et al. (2012) , AIA) , has enabled even more detailed studies of CMEs ( K.A. Kozarev et al. (2015) ) in the lower corona, as well as extensions to modeling conditions for coronal acceleration of solar energetic particles in compressive fronts ( K.A. Kozarev & Schwadron (2016) ; K.A. Kozarev et al. (2017) ; K.A. Kozarev et al. (2019) ) . However, the complexity of CMEs and related phenomena still presents challenges for precise instrument design and data interpretation ( Wuelser et al. (2004) ; Barnes et al. (2019) ) . One of these phenomena are EUV waves, also known as coronal bright fronts (CBF) - relatively dim, large-scale traveling disturbances, observed both on disk and off limb. These are thought to correspond to compressive and shock waves in the low and middle solar corona, and have been previously observed in EUV wavelengths ( K. Kozarev et al. (2010) ; Veronig et al. (2010) ) . Automatic algorithmic or data-driven differentiation between eruptive events in the solar corona can be challenging due to the complex interplay of phenomena in the Sun’s atmosphere. When visually examining observations, other contextual information, such as the presence of a solar flare or coronal mass ejection, can provide clues about the likelihood of observing one of these phenomena. Additionally, working with observations at multiple wavelengths and using time-lapse imagery to observe the propagation and evolution of these structures can further assist in analysis ( Stepanyuk & Kozarev (2024) ) . In recent years, a few groups independently proposed a number of algorithmic, non-data driven approaches for EUV wave detection of varying complexity ( Podladchikova & Berghmans (2005) ; Verbeeck et al. (2014) ; Long et al. (2014) ; Ireland et al. (2019) ; Stepanyuk et al. (2022) ) . Of these perhaps the most advanced are the CorPITA ( Long et al. (2014) ) and AWARE ( Ireland et al. (2019) ) . The CorPITA algorithm employs percentage base difference (BD) images to fit multi-Gaussian shapes to flare-source-centered, sector-averaged portions on the wave along the solar disk. The AWARE algorithm uses more advanced pre-processing in persistence running difference images, to characterize the wave shapes along similar flare-centered sectors, and a random sample consensus (RANSAC) algorithm to select the features. However, using running difference images introduces spurious features and is not recommended for discovering the true (projected) wave shape ( Long et al. (2011) ) . In addition, the persistence imaging approach tends to amplify the noise in the output, requiring

Read the complete text