pith. sign in

arxiv: 2606.21896 · v1 · pith:J7TYOCGUnew · submitted 2026-06-20 · 🌌 astro-ph.SR · cs.LG

Prediction of Solar Flares Using Photospheric Magnetic Field Parameters with Deep Learning

Yash Chaudhary , Jason T. L. Wang , Chunhui Xu , Yan Xu , Sen Zhang This is my paper

Pith reviewed 2026-06-26 11:45 UTC · model grok-4.3

classification 🌌 astro-ph.SR cs.LG
keywords solar flaresdeep learningphotospheric magnetic fieldexplainable AISHAPpartial dependence plotsspace weather
0
0 comments X

The pith

Deep learning models trained on photospheric magnetic field parameters predict solar flares while SHAP and partial dependence plots reveal which features drive the forecasts.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes training deep learning models directly on photospheric magnetic field parameters to forecast M- and X-class solar flares. It adds SHAP values for global and local feature importance and partial dependence plots to show how individual parameters influence predictions. The goal is to move beyond black-box outputs so that space-weather forecasters can understand and trust the model's reasoning when deciding on mitigation steps for infrastructure risks.

Core claim

A deep learning framework that ingests photospheric magnetic field parameters achieves solar flare prediction while SHAP and partial dependence plots supply both global and local explanations of feature contributions and trends, thereby increasing the interpretability of the otherwise opaque model for operational space-weather use.

What carries the argument

Deep neural network on photospheric magnetic field parameters interpreted by SHAP values for additive feature attributions and partial dependence plots for marginal effect trends.

If this is right

  • SHAP supplies both global rankings and local explanations for individual flare predictions.
  • Partial dependence plots display the marginal trend of each magnetic parameter on flare probability.
  • The combination supports more informed operational decisions in solar physics and space weather applications.
  • XAI integration demonstrates a path for deploying AI models in high-stakes domains where trust is required.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same magnetic-parameter-plus-XAI pipeline could be tested on other solar events such as coronal mass ejections to check whether feature rankings remain consistent.
  • If SHAP rankings align with established flare-triggering physics, the model could guide targeted observations of specific magnetic structures.
  • Operational centers might embed the PDP surfaces into real-time dashboards so forecasters can see how changing a measured parameter would alter the predicted probability.

Load-bearing premise

That the explanations produced by SHAP and partial dependence plots will accurately reflect the physical drivers of flares without separate validation against known solar physics.

What would settle it

A controlled test in which SHAP-identified top features show no statistical association with observed flare occurrence rates in an independent dataset, or in which the XAI-augmented model yields no measurable gain in forecast skill over a non-interpretable baseline.

Figures

Figures reproduced from arXiv: 2606.21896 by Chunhui Xu, Jason T. L. Wang, Sen Zhang, Yan Xu, Yash Chaudhary.

Figure 1
Figure 1. Figure 1: Architecture of the proposed deep learning model for solar flare prediction. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: SHAP feature importance plot: Bar graph ranking [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: SHAP beeswarm plot: Features ranked by impor [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 7
Figure 7. Figure 7: 3D PDP providing enhanced visualization of the [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: 2D PDP showing the interaction between TO [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
read the original abstract

Solar flares, particularly those of the M- and X-class, have a significant impact on human life because of their potential to disrupt critical infrastructure and communication systems on Earth. Accurate prediction of solar flares is crucial for mitigating these risks, but the black-box nature of conventional deep learning models used in flare prediction limits their trustworthiness and interpretability. In this paper, we propose a new approach to solar flare prediction using photospheric magnetic field parameters or features with deep learning. To improve model interpretability, we integrate explainable artificial intelligence (XAI) techniques, including SHapley Additive exPlanations (SHAP) and partial dependence plots (PDPs), into our prediction framework. XAI methods provide transparency by analyzing the importance and interactions of features used by our model. Specifically, SHAP values offer a global and local understanding of the features, while PDPs provide insights into feature-level trends. These techniques demonstrate the potential of XAI in deploying AI-driven solutions in high-impact applications such as solar flare prediction, paving the way for more informed decision-making in solar physics and space weather studies.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes a deep learning approach to solar flare (M- and X-class) prediction that ingests photospheric magnetic field parameters and augments the model with SHAP values and partial dependence plots to improve interpretability and reveal feature importance and interactions.

Significance. If the proposed framework were implemented, validated against standard flare-prediction baselines, and shown to yield both competitive skill scores and physically consistent explanations, it could contribute to trustworthy AI applications in space-weather forecasting. As written, however, the work contains no datasets, architectures, training procedures, performance metrics, or XAI outputs, so no such contribution is demonstrated.

major comments (2)
  1. [Abstract] Abstract: the sentence 'These techniques demonstrate the potential of XAI in deploying AI-driven solutions...' asserts that SHAP and PDPs have already supplied meaningful transparency, yet the manuscript provides neither model outputs, SHAP summary plots, PDP curves, nor any comparison against known physical predictors such as total unsigned flux or Schrijver's R.
  2. No section of the manuscript describes the input feature vector, the deep-learning architecture, the training/validation split, the flare catalog, or any quantitative evaluation; without these elements the central claim that the XAI-augmented model improves both accuracy and interpretability cannot be assessed.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive report. We agree that the submitted manuscript is incomplete and lacks the required methodological details, data descriptions, model specifications, quantitative results, and XAI visualizations needed to support its claims. We will prepare a substantially revised version that addresses these gaps.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the sentence 'These techniques demonstrate the potential of XAI in deploying AI-driven solutions...' asserts that SHAP and PDPs have already supplied meaningful transparency, yet the manuscript provides neither model outputs, SHAP summary plots, PDP curves, nor any comparison against known physical predictors such as total unsigned flux or Schrijver's R.

    Authors: We agree that the abstract overstates what is shown. The sentence will be rewritten to describe the intended use of SHAP and PDPs rather than asserting completed demonstrations. The revised manuscript will include SHAP summary plots, PDP curves, and direct comparisons against established physical predictors including total unsigned flux and Schrijver's R. revision: yes

  2. Referee: [—] No section of the manuscript describes the input feature vector, the deep-learning architecture, the training/validation split, the flare catalog, or any quantitative evaluation; without these elements the central claim that the XAI-augmented model improves both accuracy and interpretability cannot be assessed.

    Authors: We acknowledge that these core elements are absent from the submitted text, which prevents evaluation of the central claims. The revision will add explicit sections detailing the input feature vector (photospheric magnetic field parameters), the deep-learning architecture, training/validation splits, the flare catalog, quantitative skill scores, and the application of XAI methods with supporting outputs. revision: yes

Circularity Check

0 steps flagged

No circularity: application proposal with no derivation chain or fitted predictions

full rationale

The manuscript is framed as an application of existing deep learning and XAI methods (SHAP, PDPs) to photospheric magnetic field data for flare prediction. No equations, derivations, parameter fits, or uniqueness theorems appear in the provided text. The central claim is simply that integrating SHAP/PDPs supplies transparency; this is an unvalidated modeling assumption rather than a reduction of any output to its inputs by construction. No self-citation load-bearing steps, ansatz smuggling, or renaming of known results are present. The work is therefore self-contained against external benchmarks with no circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review is based on abstract only; no free parameters, invented entities, or explicit axioms are stated in the text.

axioms (1)
  • domain assumption Photospheric magnetic field parameters contain information sufficient to predict M- and X-class flares
    The proposal rests on using these parameters as input features.

pith-pipeline@v0.9.1-grok · 5733 in / 1192 out tokens · 25924 ms · 2026-06-26T11:45:34.265753+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

48 extracted references · 14 canonical work pages · 1 internal anchor

  1. [1]

    K.; Liu , H.; Li , Q.; Wang , J

    Abduallah , Y.; Jordanova , V. K.; Liu , H.; Li , Q.; Wang , J. T. L.; and Wang , H. 2022. Predicting solar energetic particles using SDO/HMI vector magnetic data products and a bidirectional LSTM network. The Astrophysical Journal Supplement Series 260(1):16

    2022

  2. [2]

    Abduallah , Y.; Wang , J. T. L.; Wang , H.; and Xu , Y. 2023. Operational prediction of solar flares using a transformer-based framework . Scientific Reports 13:13665

    2023

  3. [3]

    Akiba, T.; Sano, S.; Yanase, T.; Ohta, T.; and Koyama, M. 2019. O ptuna: A next-generation hyperparameter optimization framework. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , 2623--2631

    2019

  4. [4]

    G., and Couvidat , S

    Bobra , M. G., and Couvidat , S. 2015. Solar flare prediction using SDO/HMI vector magnetic field data with a machine-learning algorithm. The Astrophysical Journal 798(2):135

    2015

  5. [5]

    G.; Sun , X.; Hoeksema , J

    Bobra , M. G.; Sun , X.; Hoeksema , J. T.; Turmon , M.; Liu , Y.; Hayashi , K.; Barnes , G.; and Leka , K. D. 2014. The Helioseismic and Magnetic Imager (HMI) vector magnetic field pipeline: SHARPs - Space-Weather HMI Active Region Patches . Solar Physics 289(9):3549--3578

    2014

  6. [6]

    J.; Kim , H.; Bizos , G.; Shin , Y.; Wang , J

    Farooki , H.; Abduallah , Y.; Noh , S. J.; Kim , H.; Bizos , G.; Shin , Y.; Wang , J. T. L.; and Wang , H. 2024. A machine learning approach to understanding the physical properties of magnetic flux ropes in the solar wind at 1 au. The Astrophysical Journal 961(1):81

    2024

  7. [7]

    H.; Bercik , D

    Fisher , G. H.; Bercik , D. J.; Welsch , B. T.; and Hudson , H. S. 2012. Global forces in eruptive solar flares: The Lorentz force acting on the solar atmosphere and the solar interior . Solar Physics 277(1):59--76

    2012

  8. [8]

    Guyon, I., and Elisseeff, A. 2003. An introduction to variable and feature selection. Journal of Machine Learning Research 3:1157--1182

    2003

  9. [9]

    Hassani , Z.; Mohammadpur , D.; and Safari , H. 2025. Solar flare prediction using long short-term memory (LSTM) and decomposition- LSTM with sliding window pattern recognition. The Astrophysical Journal Supplement Series 279(1):27

    2025

  10. [10]

    Li , X.; Li , X.; Zheng , Y.; Li , T.; Yan , P.; Ye , H.; Zhang , S.; Wang , X.; Lv , Y.; and Huang , X. 2025. Prediction of large solar flares based on SHARP and high-energy-density magnetic field parameters. The Astrophysical Journal Supplement Series 276(1):7

    2025

  11. [11]

    Liu , H.; Liu , C.; Wang , J. T. L.; and Wang , H. 2019. Predicting solar flares using a long short-term memory network. The Astrophysical Journal 877(2):121

    2019

  12. [12]

    M., and Lee, S.-I

    Lundberg, S. M., and Lee, S.-I. 2017. A unified approach to interpreting model predictions. In Proceedings of the Annual Conference on Neural Information Processing Systems

    2017

  13. [13]

    Molnar, C. 2025. Interpretable Machine Learning: A Guide for Making Black Box Models Explainable . 3 edition

    2025

  14. [14]

    J.; Christe, S.; Pérez-Suárez, D.; Ireland, J.; Shih, A

    Mumford, S. J.; Christe, S.; Pérez-Suárez, D.; Ireland, J.; Shih, A. Y.; Inglis, A. R.; Liedtke, S.; Hewett, R. J.; Mayer, F.; Hughitt, K.; Freij, N.; Meszaros, T.; Bennett, S. M.; Malocha, M.; Evans, J.; Agrawal, A.; Leonard, A. J.; Robitaille, T. P.; Mampaey, B.; Campos-Rozo, J. I.; and Kirk, M. S. 2015. SunPy—Python for solar physics. Computational Sci...

    2015

  15. [15]

    H.; Sinthong, P.; and Kalagnanam, J

    Nie, Y.; Nguyen, N. H.; Sinthong, P.; and Kalagnanam, J. 2023. A time series is worth 64 words: Long-term forecasting with transformers. In Proceedings of the 11th International Conference on Learning Representations

    2023

  16. [16]

    Peng, H.; Long, F.; and Ding, C. 2005. Feature selection based on mutual information: Criteria of max-dependency, max-relevance, and min-redundancy. IEEE Transactions on Pattern Analysis and Machine Intelligence 27(8):1226--1238

    2005

  17. [17]

    N.; Kaiser, L.; and Polosukhin, I

    Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, L.; and Polosukhin, I. 2017. Attention is all you need. In Proceedings of the Annual Conference on Neural Information Processing Systems , 5998--6008

    2017

  18. [18]

    Zhang , H.; Jing , J.; Wang , J. T. L.; Wang , H.; Abduallah , Y.; Xu , Y.; Alobaid , K. A.; Farooki , H.; and Yurchyshyn , V. 2025. Prediction of halo coronal mass ejections using SDO/HMI vector magnetic data products and a transformer model. The Astrophysical Journal 981(1):37

    2025

  19. [19]

    Zhou, H.; Zhang, S.; Peng, J.; Zhang, S.; Li, J.; Xiong, H.; and Zhang, W. 2021. Informer: Beyond efficient transformer for long sequence time-series forecasting. In Proceedings of the 35th AAAI Conference on Artificial Intelligence , 11106--11115

    2021

  20. [20]

    The Astrophysical Journal , keywords =

    Predicting Solar Flares Using. The Astrophysical Journal , keywords =

  21. [21]

    The Astrophysical Journal , keywords =

    Solar Flare Prediction Using. The Astrophysical Journal , keywords =

  22. [22]

    Solar Flare Prediction Model with Three Machine-learning Algorithms using Ultraviolet Brightening and Vector Magnetograms , journal =

  23. [23]

    Predicting Solar Flares Using a Long Short-term Memory Network , journal =

  24. [24]

    Wheatland, M. S. , title =. Space Weather , volume =. doi:https://doi.org/10.1029/2004SW000131 , eprint =

    work page doi:10.1029/2004sw000131

  25. [25]

    and Wang, Xiantong and Wang, Yu and Sun, Hu and Gombosi, Tamas and Chen, Yang and Hero, Alfred , title =

    Sun, Zeyu and Bobra, Monica G. and Wang, Xiantong and Wang, Yu and Sun, Hu and Gombosi, Tamas and Chen, Yang and Hero, Alfred , title =. The Astrophysical Journal , abstract =. doi:10.3847/1538-4357/ac64a6 , year =

    work page doi:10.3847/1538-4357/ac64a6

  26. [26]

    , booktitle=

    Ma, Ruizhe and Boubrahimi, Soukaina Filali and Hamdi, Shah Muhammad and Angryk, Rafal A. , booktitle=. Solar flare prediction using multivariate time series decision trees , year=

  27. [27]

    and Eler, Danilo M

    Marcílio, Wilson E. and Eler, Danilo M. , booktitle=. From explanations to feature selection: Assessing. 2020 , volume=

    2020

  28. [28]

    URL http://dx.doi.org/10.1145/3447548

    Bento, João and Saleiro, Pedro and Cruz, André F. and Figueiredo, Mário A.T. and Bizarro, Pedro , pages=. doi:10.1145/3447548.3467166 , booktitle=

    work page doi:10.1145/3447548.3467166

  29. [29]

    Exploring SAS

    Interpreting black-box machine learning models using partial dependence and individual conditional expectation plots , author=. Exploring SAS

  30. [30]

    2019 , eprint=

    Visualizing the Effects of Predictor Variables in Black Box Supervised Learning Models , author=. 2019 , eprint=

    2019

  31. [31]

    , Christe , S

    Mumford, Stuart J and Christe, Steven and Pérez-Suárez, David and Ireland, Jack and Shih, Albert Y and Inglis, Andrew R and Liedtke, Simon and Hewett, Russell J and Mayer, Florian and Hughitt, Keith and Freij, Nabil and Meszaros, Tomas and Bennett, Samuel M and Malocha, Michael and Evans, John and Agrawal, Ankit and Leonard, Andrew J and Robitaille, Thoma...

    work page doi:10.1088/1749-4699/8/1/014009

  32. [32]

    Nguyen and Phanwadee Sinthong and Jayant Kalagnanam , title =

    Yuqi Nie and Nam H. Nguyen and Phanwadee Sinthong and Jayant Kalagnanam , title =. Proceedings of the 11th International Conference on Learning Representations , year =

  33. [33]

    , title =

    Hanchuan Peng and Fuhui Long and Ding, C. , title =. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume =. 2005 , doi =

    2005

  34. [34]

    An Introduction to Variable and Feature Selection , journal =

    Isabelle Guyon and Andr. An Introduction to Variable and Feature Selection , journal =

  35. [35]

    Akiba, Takuya and Sano, Shotaro and Yanase, Toshihiko and Ohta, Takeru and Koyama, Masanori , booktitle=

  36. [36]

    Proceedings of the Annual Conference on Neural Information Processing Systems , year=

    A unified approach to interpreting model predictions , author=. Proceedings of the Annual Conference on Neural Information Processing Systems , year=

  37. [37]

    2025 , edition=

    Interpretable Machine Learning: A Guide for Making Black Box Models Explainable , author=. 2025 , edition=

    2025

  38. [38]

    Scientific Reports , year = 2023, month = aug, volume =

    Operational prediction of solar flares using a transformer-based framework. Scientific Reports , year = 2023, month = aug, volume =. doi:10.1038/s41598-023-40884-1 , adsurl =

    work page doi:10.1038/s41598-023-40884-1 2023

  39. [39]

    The Astrophysical Journal Supplement Series , keywords =

    Prediction of Large Solar Flares Based on. The Astrophysical Journal Supplement Series , keywords =. doi:10.3847/1538-4365/ad8b2a , archivePrefix =. 2410.18562 , primaryClass =

    work page doi:10.3847/1538-4365/ad8b2a

  40. [40]

    The Astrophysical Journal Supplement Series , keywords =

    Solar Flare Prediction Using Long Short-term Memory. The Astrophysical Journal Supplement Series , keywords =. doi:10.3847/1538-4365/addc73 , archivePrefix =. 2507.05313 , primaryClass =

    work page doi:10.3847/1538-4365/addc73

  41. [41]

    Gomez and Lukasz Kaiser and Illia Polosukhin , title =

    Ashish Vaswani and Noam Shazeer and Niki Parmar and Jakob Uszkoreit and Llion Jones and Aidan N. Gomez and Lukasz Kaiser and Illia Polosukhin , title =. Proceedings of the Annual Conference on Neural Information Processing Systems , pages =. 2017 , url =

    2017

  42. [42]

    Informer: Beyond efficient transformer for long sequence time-series forecasting,

    Haoyi Zhou and Shanghang Zhang and Jieqi Peng and Shuai Zhang and Jianxin Li and Hui Xiong and Wancai Zhang , title =. Proceedings of the 35th. 2021 , url =. doi:10.1609/AAAI.V35I12.17325 , timestamp =

    work page doi:10.1609/aaai.v35i12.17325 2021

  43. [43]

    , Jing , J

    Prediction of Halo Coronal Mass Ejections Using. The Astrophysical Journal , keywords =. doi:10.3847/1538-4357/adafa0 , archivePrefix =. 2503.03237 , primaryClass =

    work page doi:10.3847/1538-4357/adafa0

  44. [44]

    Predicting Solar Flares Using a Long Short-Term Memory Network

    , keywords =. doi:10.3847/1538-4357/ab1b3c , archivePrefix =. 1905.07095 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4357/ab1b3c 1905

  45. [45]

    The Astrophysical Journal Supplement Series , keywords =

    Predicting Solar Energetic Particles Using. The Astrophysical Journal Supplement Series , keywords =. doi:10.3847/1538-4365/ac5f56 , archivePrefix =. 2203.14393 , primaryClass =

    work page doi:10.3847/1538-4365/ac5f56

  46. [46]

    doi:10.3847/1538-4357/ad0c52 , archivePrefix =

    A Machine Learning Approach to Understanding the Physical Properties of Magnetic Flux Ropes in the Solar Wind at 1 au , journal =. doi:10.3847/1538-4357/ad0c52 , archivePrefix =. 2311.09345 , primaryClass =

    work page doi:10.3847/1538-4357/ad0c52

  47. [47]

    G., Sun, X., Hoeksema, J

    The Helioseismic and Magnetic Imager (HMI) vector magnetic field pipeline: SHARPs - Space-Weather HMI Active Region Patches. Solar Physics , keywords =. doi:10.1007/s11207-014-0529-3 , eprint =

    work page doi:10.1007/s11207-014-0529-3

  48. [48]

    Solar Physics , keywords =

    Global forces in eruptive solar flares: The Lorentz force acting on the solar atmosphere and the solar interior. Solar Physics , keywords =. doi:10.1007/s11207-011-9907-2 , eprint =

    work page doi:10.1007/s11207-011-9907-2