Machine learning for reconstruction of polarity inversion lines from solar filaments - Abstract and method overview

Aug 4, 2026

Upload by Oliver

Solar filaments serve as tracers of polarity inversion lines, separating opposite magnetic polarities on the solar photosphere. Since filament observations predate systematic magnetic-field measurements, historical filament catalogs enable reconstruction of polarity maps when direct magnetograms are unavailable. The reconstruction is often ambiguous and manually performed, so the work presents an automatic machine-learning model that outputs multiple magnetically consistent polarity maps, validated using the McIntosh catalog. User-guided polarity reference inputs reduce prior-knowledge gaps and enable uncertainty estimation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Introduction

Motivation from historical chromospheric observations

Prior work on polarity inversion line reconstruction

Algorithmic formulation using filament fragments

Deep learning and uncertainty considerations

arXiv :2405 .06293v1 [ cs .LG] 10 May 2024 Machine learning for reconstruction of polarity inversion lines from solar filaments Vaclovas Kisielius 1 · Egor Illarionov 1,2 © The author(s)•••• Abstract Solar filaments are well-known tracers of polarity inversion lines that separate two opposite magnetic polarities on the solar photosphere. Because observations of filaments began long before the systematic observations of solar magnetic fields, historical filament catalogs can facilitate the reconstruction of magnetic polarity maps at times when direct magnetic observations were not yet available. In practice, this reconstruction is often ambiguous and typically performed manually. We propose an automatic approach based on a machinelearning model that generates a variety of magnetic polarity maps consistent with filament observations. To evaluate the model and discuss the results we use the catalog of solar filaments and polarity maps compiled by McIntosh. We realize that the process of manual compilation of polarity maps includes not only information on filaments, but also a large amount of prior information, which is difficult to formalize. In order to compensate for the lack of prior knowledge for the machine-learning model, we provide it with polarity information at several reference points. We demonstrate that this process, which can be considered asthe user-guided reconstruction or super-resolution, leads to polarity maps that are reasonably close to hand-drawn ones, and additionally allows for uncertainty estimation. Keywords: Prominences, Magnetic fields, Machine learning 1. Introduction About 50 years ago, deriving magnetic-field information from chromospheric observations was the main way to compensate for the lack of magnetic-field V. Kisielius [waclove@yandex.ru](waclove@yandex.ru) [E.A. Illarionov](E.A. Illarionov) [egor.illarionov@math.msu.ru](egor.illarionov@math.msu.ru) 1 Moscow State University, Moscow, Russia 2 Institute of Continuous Media Mechanics, Perm, Russia SOLA: main.tex; 13 May 2024; 0:58; p . 1 maps required for practical purposes and research. Patrick McIntosh, one of the founders of the space weather field, mentioned in McIntosh (1972) up toten advantages of this approach that were actual for those days. Most of these advantages are less relevant today due to the availability of full-disk solar magnetographs with high spatial and temporal resolution. However, at least one of these advantages remains actual – the existence of chromospheric observations for more than 120 years (at the time, McIntosh said 70 years; for an extended time period, see, e.g. , Tlatova, Vasil’eva, and Tlatov, 2017; Mazumder et al. , 2021; Chatzistergos et al. , 2023) . This motivates continuing attempts to use these data to reconstruct magnetic-field maps. The approach described in McIntosh (1976) focuses on the reconstruction of magnetic polarity only. He considers the five basic chromospheric structures (filaments, filament channels, fibrils, arch-filament systems, and plage corridors) derived from Hα observations and uses their positions to infer the polarity inversion lines, which in turn define the magnetic polarity map. It should be realized that the approach is not an exact algorithm, but an analysis, combined with the author’s experience and understanding of what can and cannot be. For this reason, the dataset created by McIntosh (1964), which spans about 45 years, remains a unique resource in the field of solar physics. The downside of this unique approach is that it can unlikely be consistently extrapolated forward or backward in time. Such attempts were made in Makarov, Fatianov, and Sivaraman (1983) and Makarov and Sivaraman (1983) and cover the time period from 1904 to 1981 based on Kodaikanal, Meudon, and later Hα observations. However, for many reasons, this Archive should also be considered as a unique and non-reproducible approach. This motivates us to investigate an algori

FAQs

Why are solar filament catalogs useful for polarity map reconstruction?

They record polarity inversion line tracers long before systematic solar magnetic-field observations, enabling reconstruction at times when direct magnetograms were not yet available.

What problem does the proposed machine-learning approach address?

Manual reconstruction is typically ambiguous; the model automatically generates multiple polarity maps consistent with filament observations and supports uncertainty estimation.

How does the method compensate for missing prior knowledge?

It provides polarity information at several reference points, which can be viewed as user-guided reconstruction or super-resolution, bringing results closer to hand-drawn maps.

Aug 4, 2026

Upload by Oliver