Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Magnetogram-to-Magnetogram: Generative Forecasting of Solar Evolution

Francesco Pio Ramunno, Hyun-Jin Jeong, Stefan Hackstein, André Csillaghy, Svyatoslav Voloshynovskiy, Manolis K. Georgoulis

TL;DR

The study tackles forecasting solar magnetic field evolution by forecasting a full-disc LoS magnetogram 24 hours ahead from its 24 hours-prior state using image-to-image translation with Denoising Diffusion Probabilistic Models. It adopts a hybrid evaluation framework that combines traditional image-quality metrics with physics-based metrics for magnetic flux, active-region size, PIL length, and morphology, quantifying uncertainty through multiple realizations. Results indicate that diffusion-based forecasting preserves dynamic range and many physical features better than a differential-rotation persistence baseline, though some image-quality metrics show mixed results; the work also demonstrates uncertainty localization near complex regions such as polarity inversion lines. The findings support the potential of diffusion-driven tools as interactive, physics-aware visualization and analysis aids for solar telescopes and space-weather forecasting, with clear avenues for incorporating multi-channel data and explainable AI in future work.

Abstract

Investigating the solar magnetic field is crucial to understand the physical processes in the solar interior as well as their effects on the interplanetary environment. We introduce a novel method to predict the evolution of the solar line-of-sight (LoS) magnetogram using image-to-image translation with Denoising Diffusion Probabilistic Models (DDPMs). Our approach combines "computer science metrics" for image quality and "physics metrics" for physical accuracy to evaluate model performance. The results indicate that DDPMs are effective in maintaining the structural integrity, the dynamic range of solar magnetic fields, the magnetic flux and other physical features such as the size of the active regions, surpassing traditional persistence models, also in flaring situation. We aim to use deep learning not only for visualisation but as an integrative and interactive tool for telescopes, enhancing our understanding of unexpected physical events like solar flares. Future studies will aim to integrate more diverse solar data to refine the accuracy and applicability of our generative model.

Magnetogram-to-Magnetogram: Generative Forecasting of Solar Evolution

TL;DR

The study tackles forecasting solar magnetic field evolution by forecasting a full-disc LoS magnetogram 24 hours ahead from its 24 hours-prior state using image-to-image translation with Denoising Diffusion Probabilistic Models. It adopts a hybrid evaluation framework that combines traditional image-quality metrics with physics-based metrics for magnetic flux, active-region size, PIL length, and morphology, quantifying uncertainty through multiple realizations. Results indicate that diffusion-based forecasting preserves dynamic range and many physical features better than a differential-rotation persistence baseline, though some image-quality metrics show mixed results; the work also demonstrates uncertainty localization near complex regions such as polarity inversion lines. The findings support the potential of diffusion-driven tools as interactive, physics-aware visualization and analysis aids for solar telescopes and space-weather forecasting, with clear avenues for incorporating multi-channel data and explainable AI in future work.

Abstract

Investigating the solar magnetic field is crucial to understand the physical processes in the solar interior as well as their effects on the interplanetary environment. We introduce a novel method to predict the evolution of the solar line-of-sight (LoS) magnetogram using image-to-image translation with Denoising Diffusion Probabilistic Models (DDPMs). Our approach combines "computer science metrics" for image quality and "physics metrics" for physical accuracy to evaluate model performance. The results indicate that DDPMs are effective in maintaining the structural integrity, the dynamic range of solar magnetic fields, the magnetic flux and other physical features such as the size of the active regions, surpassing traditional persistence models, also in flaring situation. We aim to use deep learning not only for visualisation but as an integrative and interactive tool for telescopes, enhancing our understanding of unexpected physical events like solar flares. Future studies will aim to integrate more diverse solar data to refine the accuracy and applicability of our generative model.
Paper Structure (3 sections, 2 figures, 1 table)

This paper contains 3 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Implementation and Results
  3. Discussion

Figures (2)

  • Figure 1: Example of input, target, and a comparison of the model's prediction with the baseline persistence model. Input image from 25 October 2014 at 17:08 UTC.
  • Figure 2: The image shows a standard deviation map derived from 8 repeated (semi-stochastic) model predictions. As expected, the variance shows that the model finds it most difficult to predict the PIL of the AR 12192 (indicated by the arrow), which physically undergoes the most significant changes during a flare. Input image from 24 October 2014 at 17:08 UTC. The standard deviation has been computed on images normalised between -1 and 1, thus the unit of measure is in pixel value.