Pith. sign in

REVIEW 3 major objections 5 minor 85 references

Large Model Driven Solar Activity AI Forecaster: A Scalable Dual Data-Model Framework

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A fully automatic solar activity forecaster can match or beat human forecasters on next-day strong-flare prediction while running in minutes.

desk verdict A real end-to-end system with operational potential, but the claim of beating human forecasters rests on an apples-to-oranges evaluation; worth peer review, but the comparison needs fixing. read the letter →

arxiv 2508.06892 v1 pith:AYXSAC72 submitted 2025-08-09 astro-ph.SR physics.space-ph

Jingjing Wang , Pengyu Liang , Tingyu Wang , Ming Li , Yanmei Cui , Siwei Liu , Xin Huang , Xiang Li
show 7 more authors
Minghui Zhang Yunshi Zeng Zhu Cao Jiekang Feng Qinghua Hu Bingxian Luo Bing Cao
This is my paper · ORCID
classification astro-ph.SRphysics.space-ph
keywords solarflareforecastingspaceweathermulti-modaltransformermaskedautoencoderactiveregioncoronalholefilamenthuman-in-the-looplabeling
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that the whole job of a solar-activity forecaster—perceiving the Sun, characterizing active regions, and issuing next-day strong-flare probabilities—can be automated end to end, and that the automated system forecasts at least as well as experienced human forecasters. It reports a dual data-and-model framework with three connected modules: a perception network that fuses multi-wavelength images into daily feature maps, analysis tools that derive physical parameters, and a prediction network that fuses images, flare history, and physics parameters into M-class-or-stronger flare probabilities. The full pipeline runs in a few minutes on one GPU, and on a January 2023–May 2024 test period it reports higher F1 and ROCA scores than the operational human forecasts it was compared with. A sympathetic reader would care because it is a concrete existence proof that operational flare forecasting need not wait on human analysis.

What carries the argument

The engine is FPNet's Efficient Masked Autoencoder for Flare Prediction (EMA-FP), built around Physical Prior-guided Adaptive Masking (PPAM). PPAM uses magnetogram-derived segmentation to keep tokens in active-region patches while masking quieter pixels, so the masked-autoencoder pretraining concentrates representational capacity on flare-relevant structure. A multi-modal transformer then fuses co-aligned EUV 304 Å images, line-of-sight magnetograms, and a 24-hour flare index, while a physics mapper embeds scalar parameters (F10.7 flux, magnetic classification, neutral-line R value) into the same latent space; a learned global token aggregates cross-modal information and feeds a sigmoid head

What would settle it

Re-score the same test period with both the AI and each human forecast at one pre-registered threshold (e.g., 0.5) and on the identical active-region list obtained from a single shared numbering scheme. If the human forecasters' F1 reaches or exceeds 0.502 (or ROCA reaches 0.9065) under those conditions, the paper's central claim of AI superiority is refuted.

Watch

Extended reading notes

Core claim

On its own terms the central discovery is that a fully automatic perception–analysis–prediction pipeline can match or beat human forecasters on 24-hour strong-flare forecasting. The claim is quantitative: for active-region forecasts the independently evaluated model reports F1 0.5020, TSS 0.4392, BSS 0.2409, and ROCA 0.9065; the fully automatic system reports F1 0.4925, TSS 0.4596, and ROCA 0.9040; both are above the corresponding human operational scores. The load-bearing design is that domain knowledge can be injected without breaking automation: magnetogram-derived masks protect active regions during masked pretraining, physical parameters are embedded into the transformer latent space, a

Load-bearing premise

The entire outperformance claim rests on the assumption that the human forecasts were scored fairly—same test days, same thresholds, and the same active-region samples as the AI system; if the threshold choices or sample pools favored the AI, the headline result could be an artifact of evaluation design.

Editorial extensions

If this is right

  • If the reported numbers hold, next-day strong-flare forecasts can be produced automatically in minutes rather than by a human team, with accuracy at least matching current operational forecasts.
  • The perception module's cross-instrument generalization means new telescopes and wavelengths can be added without re-labeling from scratch, lowering the cost of operational adoption.
  • The small performance gap between the fully automatic system and the human-annotated variant indicates the upstream modules do not currently destroy the forecasting signal, supporting end-to-end autonomy.
  • The physical-prior masking strategy gives a concrete way to handle rare-event imbalance in other solar eruption products, since it concentrates learning capacity on active regions without discarding full-disk context.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural extension the authors leave implicit is applying the same perception–analysis–forecast loop to coronal mass ejection arrival times or solar energetic particle events, where the same multi-modal fusion bottleneck appears.
  • A decisive test would be a blind out-of-sample comparison on a later solar cycle with thresholds fixed before the test period; that would separate genuine skill from any test-set threshold tuning in the baselines.
  • The different active-region populations used for the AI and human baselines (5746 vs 3925 samples) make a shared, agency-independent region-numbering standard a prerequisite for unambiguous future machine-versus-human benchmarks.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper presents a three-module 'Solar Activity AI Forecaster' for 24-hour forecasts of M-class and stronger flares: SPNet (multi-modal detection of active regions, coronal holes, filaments), IATools (characterization, tracking, and flare history), and FPNet (a transformer-based flare prediction network with physics-guided masking). The system is evaluated on a held-out period (January 2023-May 2024) against operational forecasts from SWPC and SEPC, CCMC Flare Scoreboard models (NOAA, ASSA), and an SVM baseline. The authors report that FPNet* and the fully automatic SA-AI forecaster achieve F1 around 0.50 and ROCA around 0.90 for active-region forecasts, and state that the model outperforms human forecasters, with the full pipeline running in under 6 minutes.

Significance. If the performance claims are robust, this is a noteworthy advance in operational space weather forecasting: it is one of the first end-to-end, large-model frameworks that attempts to replicate the entire OODA forecasting loop, and the authors provide a substantial public dataset, an independent test window, and cross-instrument generalization checks. The strengths include the modular design, the human-in-the-loop data expansion, and the use of threshold-free metrics (ROCA, BSS) for part of the comparison. However, the central head-to-head claim against human forecasters rests on evaluation-protocol choices that are currently not fully justified.

major comments (3)
  1. [Section 3.6, Table 4] The binary-metric comparison between FPNet*/SA-AI and the operational baselines is not like-for-like. The operational probability forecasts are converted to binary predictions using TSS-optimized thresholds selected on the same test set (SWPC AR Th=0.2, NOAA full-disk Th=0.3, ASSA full-disk Th=0.4), while FPNet* and SA-AI use a fixed 0.5 threshold. Selecting thresholds on the test set can inflate F1/TSS for the baselines and makes the reported margins (e.g., 'surpassing SWPC by 17% in F1') dependent on post-hoc calibration. The threshold-free metrics (ROCA, BSS) in the same table are not affected, and at fixed Th=0.5 the pattern is even more favorable to FPNet*, so the central claim may survive re-analysis; nevertheless, the paper should base its headline comparison on either a common fixed threshold or properly cross-validated threshold selection, rather than presenting test-set-optimiz
  2. [Section 3.2.4 and Table 4] The SA-AI forecaster is evaluated on a different AR population and a different ground-truth event list than SWPC: SWPC contributes 3,925 AR-samples/472 NOAA ARs, whereas SA-AI is scored on 5,746 AR-samples/1,006 SEPC ARs, with flare-event labels derived from IATools' own SEPC-rule identification (770 vs 691 strong flares). F1, TSS, and ROCA are base-rate and sample-dependent; comparing them across these two sets is not a valid head-to-head. Thus the claim that the fully automatic SA-AI forecaster 'outperformed human forecasters (SWPC) by 15% in F1' is not supported by the table as presented. The only clean same-population comparison in Table 4 is FPNet* vs SWPC (both on the SWPC SRS sample). The authors should either restrict the 'outperforms human forecasters' claim to FPNet* and present the SA-AI results separately, or re-score SWPC on the SEPC AR sample using the same flare-event defi
  3. [Section 3.2.4 and Section 3.6] For the SA-AI forecaster, both the feature inputs (AR identification, tracking, Flare Index) and the evaluation labels are products of the same IATools/SEPC pipeline. The flare-event list used as ground truth is derived from SEPC flare-alert rules applied to GOES flux and AIA 193 localizations, and the paper explicitly notes that this yields a different event count (770) than SWPC records (691). This means the end-to-end evaluation may reward a model that is internally consistent with its own label generator rather than one that predicts independently verified events. An independent validation against SWPC/NOAA event reports over the same days, with a common AR-to-flare association rule, would be needed to substantiate the 'outperforms human forecasters' conclusion for the fully automatic SA-AI system.
minor comments (5)
  1. [Section 2.2.3] Typo: 'verision' should be 'version' ('the multi-modal verision (MV1)').
  2. [Table 4] The table as provided in the manuscript text lacks explicit column headers, so the reader cannot map the numeric columns to TP/TN/FP/FN and the metric definitions without external guessing. Please ensure the published table has clear column labels for each metric and threshold row.
  3. [Section 5 (Data Availability)] The data URLs for SDO/HMI, SOHO, and GOES-16/SUVI all point to the same JSOC 'lookdata' URL; GOES-16 SUVI data are not served there. Please provide correct, distinct repository links for each instrument.
  4. [Abstract and Conclusion] The phrase 'outperforms or matches human forecasters' is too broad given the evaluation asymmetries described above. Please qualify the claim to the specific metrics and sample populations for which it is established (e.g., FPNet* on the SWPC-defined AR sample for ROCA/BSS/F1).
  5. [Section 3.5] The statement 'we have calibrated the probabilistic forecasting model so that the binary classification model uses a threshold of 0.5' is not accompanied by a calibration analysis (e.g., reliability diagram). A brief calibration check would help justify the fixed 0.5 threshold used for FPNet*/SA-AI.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: predictions are evaluated on held-out data against external baselines; noted asymmetries are evaluation-fairness issues, not circularity.

full rationale

The paper's claimed derivation chain is an empirical evaluation, not an analytic derivation. FPNet* takes as inputs HMI magnetograms, AIA 304, F10.7, and SWPC-derived flare index/area/magnetic type, and predicts M/X flares in the next 24 h; the target interval is disjoint from the prior-24 h flare index, so the prediction is not the input by construction. The fully automatic SA-AI forecaster replaces SWPC-derived parameters with SPNet/IATools outputs, and still predicts future flare occurrence on a held-out January 2023-May 2024 window not used in training. No fitted parameter is renamed as a prediction: the only fitted values are thresholds, and those are applied to the operational baselines (SWPC Th=0.2, NOAA Th=0.3, ASSA Th=0.4) while FPNet*/SA-AI use a fixed 0.5 threshold, so the threshold fitting can only improve the baselines and cannot manufacture the AI systems' stated advantage. The differences in AR samples (3925 SWPC vs 5746 SEPC/IATools) and flare-event counts (691 SWPC vs 770 SEPC) are genuine threats to the fairness and validity of the head-to-head comparison, but they are not circularity: the AI scores are not defined in terms of the baseline scores, and the outcome labels are not the same as the model inputs. Self-citations (e.g., [56]-[60], [86]) appear as contextual references for precursor identification and SEPC verification conventions, and are not used to justify the central forecast-comparison claim. Thus no step reduces the paper's result to its own inputs.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

No new physical entities are postulated. EMA-FP and PPAM are software components, not invented physical objects; they introduce no falsifiable physical handles outside the system itself. The free parameters are the test-fit thresholds and unstated hyperparameters that the headline comparison depends on.

free parameters (5)
  • SPNet foreground-background weight ratios = 6:1 (AR), 4:1 (CH), 2:1 (FL)
    Selected on the validation set to maximize IoU for each solar feature (Section 3.2.1); these class-weight ratios control the segmented areas and are fitted to data.
  • TSS-optimized thresholds for human baselines = SWPC AR 0.2, NOAA full-disk 0.3, ASSA full-disk 0.4
    Chosen on the 2023-2024 test set itself (Table 4 notes, Section 3.6), so the baselines' binary metrics are fitted to the test window, favoring the AI comparison.
  • FPNet binary decision threshold = 0.5
    Fixed threshold for the AI model, stated as calibrated in Section 3.4, but the calibration data and method are not specified.
  • CH/FL merging distance threshold and small-area filters = not specified
    Predefined thresholds used by IATools to merge boundary-close regions and filter small features (Section 3.2.2); the numerical values are not reported.
  • FPNet architecture hyperparameters (depth N, PPAM sampling probabilities, masking ratios) = not specified
    Described as configurable (Section 2.2) but never given numerical values, preventing reproduction.
assumptions (5)
  • domain assumption The held-out test window (Jan 2023-May 2024) is representative of future operational conditions.
    The evaluation approach assumes the test period is statistically similar enough to deployment; solar cycle phase and instrument availability could differ.
  • domain assumption GOES soft X-ray flux and NOAA/SWPC event reports are accurate ground truth for M/X-class flare occurrence and timing.
    All FI and label construction uses these records (Eq. 1, Section 3.2.3); if event catalogs are wrong or biased, training and evaluation share the same bias.
  • domain assumption Human expert annotations of ARs, CHs, and FLs are an acceptable reference for SPNet accuracy.
    Manual labeling by SEPC forecasters is used both to build Labels V1 and the independent test set (Section 3.1); annotator subjectivity is acknowledged in the paper.
  • domain assumption Transformer and MAE pretraining objectives transfer to solar images once augmented with the PPAM masking scheme.
    The architecture choice (ViT, MAE, MulMAE) is motivated qualitatively in Section 2.2, but no ablation or convergence argument is given.
  • domain assumption No data leakage between the expanded training labels (1996-2022) and the independent test set (Jan 2023-May 2024).
    The paper states the test set was not involved in iterative training (Section 3.1), but the same forecasting team created both training labels and test annotations, so criterion drift is a residual risk.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Large Model Driven Solar Activity AI Forecaster: A Scalable Dual Data-Model Framework." pith.science (2026). https://pith.science/paper/AYXSAC72

@misc{pith2026250806892,
  author       = {Pith},
  title        = {Pith review of: Large Model Driven Solar Activity AI Forecaster: A Scalable Dual Data-Model Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AYXSAC72}},
  note         = {Machine review of arXiv:2508.06892}
}
read the original abstract

Solar activity drives space weather, affecting Earth's magnetosphere and technological infrastructure, which makes accurate solar flare forecasting critical. Current space weather models under-utilize multi-modal solar data, lack iterative enhancement via expert knowledge, and rely heavily on human forecasters under the Observation-Orientation-Decision-Action (OODA) paradigm. Here we present the "Solar Activity AI Forecaster", a scalable dual data-model driven framework built on foundational models, integrating expert knowledge to autonomously replicate human forecasting tasks with quantifiable outputs. It is implemented in the OODA paradigm and comprises three modules: a Situational Perception Module that generates daily solar situation awareness maps by integrating multi-modal observations; In-Depth Analysis Tools that characterize key solar features (active regions, coronal holes, filaments); and a Flare Prediction Module that forecasts strong flares for the full solar disk and active regions. Executed within a few minutes, the model outperforms or matches human forecasters in generalization across multi-source data, forecast accuracy, and operational efficiency. This work establishes a new paradigm for AI-based space weather forecasting, demonstrating AI's potential to enhance forecast accuracy and efficiency, and paving the way for autonomous operational forecasting systems.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

85 extracted references · 74 canonical work pages

  1. [1]

    D., Tsurutani, B

    Gonzalez, W. D., Tsurutani, B. T. & Cl´ ua de Gonzalez, A. L. Interplanetary origin of geomagnetic storms.����� ������� ���������, 529–562 (1999)

    1999

  2. [2]

    Leka, K. D. Brousseau, D. (ed.)������������� ��� ����������� ��� ����� ������� �� ����� �������. (ed.Brousseau, D.)���� ��������� ��������� ������ ������, 100.01 (2018)

    2018

  3. [3]

    P.�� ���The Economic Impact of Space Weather: Where Do We Stand?���� ����������, 206–218 (2017)

    Eastwood, J. P.�� ���The Economic Impact of Space Weather: Where Do We Stand?���� ����������, 206–218 (2017)

    2017

  4. [4]

    Space Weather: The Solar Perspective.������ ������� �� ����� ��������, 2 (2006)

    Schwenn, R. Space Weather: The Solar Perspective.������ ������� �� ����� ��������, 2 (2006)

    2006

  5. [5]

    solar flare effects

    Tsurutani, B. T.�� ���A brief review of “solar flare effects” on the ionosphere. ����� ���������, RS0A17 (2009)

    2009

  6. [6]

    N.�� ���A major solar eruptive event in July 2012: Defining extreme space weather scenarios.����� ���������, 585–591 (2013)

    Baker, D. N.�� ���A major solar eruptive event in July 2012: Defining extreme space weather scenarios.����� ���������, 585–591 (2013)

    work page 2012

  7. [7]

    Marov, M. Y. & Kuznetsov, V. D.����� ������ ��� ������ �� �����, 1–26 (Springer International Publishing, Cham, 2021). URL https://doi.org/10.1007/ 978-3-319-02847-7 1-1. 24

    work page 2021

  8. [8]

    Space weather: the solar perspective: An update to Schwenn (2006)

    Temmer, M. Space weather: the solar perspective: An update to Schwenn (2006). ������ ������� �� ����� ���������, 4 (2021)

    work page 2006

Show all 85 references
  1. [9]

    & Alauzet, F

    Amari, T., Canou, A., Aly, J.-J., Delyon, F. & Alauzet, F. Magnetic cage and rope as the key for solar eruptions.���������, 211–215 (2018)

    2018

  2. [10]

    H.�� ���Low Altitude Solar Magnetic Reconnection, Type III Solar Radio Bursts, and X-ray Emissions.��������� ��������, 1676 (2018)

    Cairns, I. H.�� ���Low Altitude Solar Magnetic Reconnection, Type III Solar Radio Bursts, and X-ray Emissions.��������� ��������, 1676 (2018)

    2018

  3. [11]

    Wang, Y.-M., Hawley, S. H. & Sheeley, N. R., Jr. The Magnetic Nature of Coronal Holes.����������, 464–469 (1996)

    1996

  4. [12]

    D.�� ���Highly structured slow solar wind emerging from an equatorial coronal hole.���������, 237–242 (2019)

    Bale, S. D.�� ���Highly structured slow solar wind emerging from an equatorial coronal hole.���������, 237–242 (2019)

    2019

  5. [13]

    J., Skelton, A., Horne, R

    Oughton, E. J., Skelton, A., Horne, R. B., Thomson, A. W. P. & Gaunt, C. T. Quantifying the daily economic impact of extreme space weather due to failure in electricity transmission infrastructure.����� ���������, 65–83 (2017)

    2017

  6. [14]

    Huang, X.�� ���Short-term solar eruptive activity prediction models based on machine learning approaches: A review.������� ����� ����� ����������, 3727– 3764 (2024)

    2024

  7. [15]

    Gong, L.�� ���Solar Active Regions Detection and Tracking Based on Deep Learning.����� ����������, 121 (2024)

    2024

  8. [16]

    & Lindsey, C

    Hamada, A., Creelman, M., Jain, K. & Lindsey, C. F ArSide Trained Active Region Recognition (F ASTARR): A Machine Learning Approach.��� ������������� ������� ���������� ���������, 53 (2025)

    2025

  9. [17]

    & Napoli, C

    Grycuk, R., Scherer, R., De Magistris, G. & Napoli, C. Real-time detection and classification of active regions from solar images using sector-based hash- ing.������� �� ������������� ���������, 102604 (2025). URL https://www. sciencedirect.com/science/article/pii/S187775032500081X

    2025

  10. [18]

    A.�� ���Improvements on coronal hole detection in SDO/AIA images using supervised classification.������� �� ����� ������� ��� ����� ��������, A23 (2015)

    Reiss, M. A.�� ���Improvements on coronal hole detection in SDO/AIA images using supervised classification.������� �� ����� ������� ��� ����� ��������, A23 (2015)

    2015

  11. [19]

    & Datta, A

    Bandyopadhyay, S., Das, S. & Datta, A. Comparative Study and Development of Two Contour-Based Image Segmentation Techniques for Coronal Hole Detection in Solar Images.����� ����������, 110 (2020)

    2020

  12. [20]

    Jarolim, R.�� ���Multi-channel coronal hole detection with convolutional neural networks.��������� ��� ���������������, A13 (2021)

    2021

  13. [21]

    S., Kempton, D

    Ahmadzadeh, A., Mahajan, S. S., Kempton, D. J., Angryk, R. A. & Ji, S. Toward Filament Segmentation Using Deep Neural Networks.����� �������� 25 arXiv:1912.02743 (2019)

    1912 arXiv

  14. [22]

    A., Banda, J

    Schuh, M. A., Banda, J. M., Bernasconi, P. N., Angryk, R. A. & Martens, P. C. H. A comparative evaluation of automated solar filament detection.����� �������(2014)

    2014

  15. [23]

    & Wang, H

    Liu, D., Song, W., Lin, G. & Wang, H. Solar Filament Segmentation Based on Improved U-Nets.����� ����������, 176 (2021)

    2021

  16. [24]

    Diercke, A.�� ���A universal method for solar filament detection from H�obser- vations using semi-supervised deep learning.��������� � ���������������, A213 (2024)

    2024

  17. [25]

    & Zeng, C

    Zhu, G., Lin, G., Yang, X. & Zeng, C. Flat U-Net: An Efficient Ultralightweight Model for Solar Filament Segmentation in Full-disk H�Images.��� ������������� ����������, 176 (2025)

    2025

  18. [26]

    & Csillaghy, A

    Giger, M. & Csillaghy, A. Unsupervised anomaly detection with varia- tional autoencoders applied to full-disk solar images.����� ���������, e2023SW003516 (2024). URL https://agupubs.onlinelibrary.wiley.com/doi/abs/ 10.1029/2023SW003516. E2023SW003516 2023SW003516

    2024 doi

  19. [27]

    Leka, K. D. & Barnes, G. Photospheric Magnetic Field Properties of Flaring versus Flare-quiet Active Regions. IV. A Statistically Significant Sample.��� ������������� ����������, 1173–1186 (2007)

    2007

  20. [28]

    Schrijver, C. J. A Characteristic Magnetic Field Pattern Associated with All Major Solar Flares and Its Use in Flare Forecasting.��� ������������� ������� ����������, L117–L120 (2007)

    2007

  21. [29]

    G.�� ���The Helioseismic and Magnetic Imager (HMI) Vector Mag- netic Field Pipeline: SHARPs - Space-Weather HMI Active Region Patches.����� ����������, 3549–3578 (2014)

    Bobra, M. G.�� ���The Helioseismic and Magnetic Imager (HMI) Vector Mag- netic Field Pipeline: SHARPs - Space-Weather HMI Active Region Patches.����� ����������, 3549–3578 (2014)

    2014

  22. [30]

    Wang, J.�� ���Parameters Derived from the SDO/HMI Vector Magnetic Field Data: Potential to Improve Machine-learning-based Solar Flare Prediction Models.��� ������������� ����������, 175 (2019)

    2019

  23. [31]

    Abduallah, Y., Wang, J. T. L., Wang, H. & Xu, Y. Operational prediction of solar flares using a transformer-based framework.��������� ���������, 13665 (2023)

    2023

  24. [32]

    Grim, L. F. L. & Gradvohl, A. L. S. Solar Flare Forecasting Based on Mag- netogram Sequences Learning with Multiscale Vision Transformers and Data Augmentation Techniques.����� ����������, 33 (2024)

    2024

  25. [33]

    ����� ���������, e02440 (2020)

    Jiao, Z.�� ���Solar Flare Intensity Prediction With Machine Learning Models. ����� ���������, e02440 (2020). 26

    2020

  26. [34]

    & Reuveni, Y

    Landa, V. & Reuveni, Y. Low-dimensional Convolutional Neural Network for Solar Flares GOES Time-series Classification.��� ������������� ������� ���������� ���������, 12 (2022)

    2022

  27. [36]

    Park, E.�� ���Application of the Deep Convolutional Neural Network to the Forecast of Solar Flare Occurrence Using Full-disk Solar Magnetograms.��� ������������� ����������, 91 (2018)

    2018

  28. [37]

    & Inceoglu, F

    Pelkum Donahue, K. & Inceoglu, F. Forecasting solar flares with a transformer network.��������� �� ��������� ��� ����� ����������, 1298609 (2024)

    2024

  29. [38]

    Sun, P.�� ���Solar Flare Forecast Using 3D Convolutional Neural Networks.��� ������������� ����������, 1 (2022)

    2022

  30. [39]

    Abduallah, Y., Wang, J. T. L., Nie, Y., Liu, C. & Wang, H. DeepSun: machine- learning-as-a-service for solar flare prediction.�������� �� ��������� ��� ��������������, 160 (2021)

    2021

  31. [40]

    & Piana, M

    Guastavino, S., Marchetti, F., Benvenuto, F., Campi, C. & Piana, M. Operational solar flare forecasting via video-based deep learning.��������� �� ��������� ��� ����� ���������, 399 (2023)

    2023

  32. [41]

    Huang, X.�� ���Deep Learning Based Solar Flare Forecasting Model. I. Results for Line-of-sight Magnetograms.��� ������������� ����������, 7 (2018)

    2018

  33. [42]

    Liu, H., Liu, C., Wang, J. T. L. & Wang, H. Predicting Solar Flares Using a Long Short-term Memory Network.��� ������������� ����������, 121 (2019)

    2019

  34. [43]

    & Piana, M

    Guastavino, S., Marchetti, F., Benvenuto, F., Campi, C. & Piana, M. Implementa- tion paradigm for supervised flare forecasting studies: A deep learning application with video data.��������� � ���������������, A105 (2022)

    2022

  35. [44]

    Li, M.�� ���Knowledge-Informed Deep Neural Networks for Solar Flare Forecasting.����� ���������, e2021SW002985 (2022)

    2022

  36. [45]

    & Gradvohl, A

    Ribeiro, F. & Gradvohl, A. L. S. Machine learning techniques applied to solar flares forecasting.��������� ��� �����������, 100468 (2021)

    2021

  37. [46]

    Sun, Z.�� ���Predicting Solar Flares Using CNN and LSTM on Two Solar Cycles of Active Region Data.��� ������������� ����������, 163 (2022)

    2022

  38. [47]

    Tang, R.�� ���Solar Flare Prediction Based on the Fusion of Multiple Deep- learning Models.��� ������������� ������� ���������� ���������, 50 (2021). 27

    2021

  39. [48]

    Han, K.�� ���Research Progress in Solar Flare Prediction Methods.�������� �� ��������� ��� ��������������, 035025 (2025)

    2025

  40. [49]

    Kim, T.�� ���Solar farside magnetograms from deep learning analysis of STEREO/EUVI data.������ ����������, 397–400 (2019)

    2019

  41. [50]

    & Lee, H

    Jeong, H.-J., Moon, Y.-J., Park, E. & Lee, H. Solar Coronal Magnetic Field Extrapolation from Synchronic Data with AI-generated Farside.��� ������������� ������� ����������, L25 (2020)

    2020

  42. [51]

    Park, E.�� ���Generation of Solar UV and EUV Images from SDO/HMI Magnetograms by Deep Learning.��� ������������� ������� ����������, L23 (2019)

    2019

  43. [52]

    Shin, G.�� ���Generation of High-resolution Solar Pseudo-magnetograms from Ca II K Images by Deep Learning.��� ������������� ������� ����������, L16 (2020)

    2020

  44. [53]

    M., P¨ otzi, W

    Jarolim, R., Veronig, A. M., P¨ otzi, W. & Podladchikova, T. Instrument- To-Instrument translation: Instrumental advances drive restoration of solar observation series via deep learning.����� ��������arXiv:2401.08057 (2024)

    2024 arXiv

  45. [54]

    Bobra, M. G. & Couvidat, S. Solar Flare Prediction Using SDO/HMI Vector Mag- netic Field Data with a Machine-learning Algorithm.��� ������������� ������� ���, 135 (2015)

    2015

  46. [55]

    Wang, J.�� ���Solar Flare Predictive Features Derived from Polarity Inver- sion Line Masks in Active Regions Using an Unsupervised Machine Learning Algorithm.��� ������������� ����������, 140 (2020)

    2020

  47. [56]

    & Liu, S

    Wang, J., Luo, B. & Liu, S. Precursor identification for strong flares based on anomaly detection algorithm.��������� �� ��������� ��� ����� ���������, 1037863 (2022)

    2022

  48. [57]

    Liu, S.�� ���A selective up-sampling method applied upon unbalanced data for flare prediction: potential to improve model performance.��������� �� ��������� ��� ����� ����������, 1082694 (2023)

    2023

  49. [58]

    & Wang, X

    Li, M., Cui, Y., Luo, B., Wang, J. & Wang, X. Deep neural networks of solar flare forecasting for complex active regions.��������� �� ��������� ��� ����� ����������, 1177550 (2023)

    2023

  50. [59]

    ��� ������������� ������� ���������� ���������, 43 (2023)

    Deng, H.�� ���Two-stage Hierarchical Framework for Solar Flare Prediction. ��� ������������� ������� ���������� ���������, 43 (2023)

    2023

  51. [60]

    & Zhang, Y

    Wang, J., Luo, B., Liu, S. & Zhang, Y. A Strong-flare Prediction Model Developed Using a Machine-learning Algorithm Based on the Video Data Sets of the Solar 28 Magnetic Field of Active Regions.��� ������������� ������� ���������� ������ ���, 54 (2023)

    2023

  52. [61]

    Reichstein, M.�� ���Deep learning and process understanding for data-driven Earth system science.���������, 195–204 (2019)

    2019

  53. [62]

    Bi, K.�� ���Accurate medium-range global weather forecasting with 3d neural networks.���������, 533–538 (2023)

    2023

  54. [63]

    Lam, R.�� ���Learning skillful medium-range global weather forecasting.������� ���, 1416–1421 (2023)

    2023

  55. [64]

    Chen, K.�� ���FengWu: Pushing the Skillful Global Medium-range Weather Forecast beyond 10 Days Lead.����� ��������arXiv:2304.02948 (2023)

    2023

  56. [65]

    Chen, L.�� ���FuXi: a cascade machine learning forecasting system for 15-day global weather forecast.��� ������� ��� ����������� ��������, 190 (2023)

    2023

  57. [66]

    Wang, X.�� ���XiHe: A Data-Driven Model for Global Ocean Eddy-Resolving Forecasting.����� ��������arXiv:2402.02995 (2024)

    2024 arXiv

  58. [67]

    Zhang, Y., Wei, C., He, Z. & Yu, W. Geogpt: An assistant for understanding and processing geospatial tasks.������������� ������� �� ������� ����� ����������� ��� �����������������, 103976 (2024). URL https://www.sciencedirect.com/ science/article/pii/S1569843224003303

    2024

  59. [68]

    Guastavino, S., Legnaro, E., Massone, A. M. & Piana, M. Artificial Intelligence for the Characterization of the 2024 May Superstorm: Active Region Classification, Flare Forecasting, and Geomagnetic Storm Prediction.��� ������������� ������� ���, 53 (2025)

    2024

  60. [69]

    M.�� ���Real-time solar image classification: Assessing spectral, pixel-based approaches.������� �� ����� ������� ��� ����� ��������, A38 (2019)

    Hughes, J. M.�� ���Real-time solar image classification: Assessing spectral, pixel-based approaches.������� �� ����� ������� ��� ����� ��������, A38 (2019)

    2019

  61. [70]

    & Ishii, M

    Nishizuka, N., Sugiura, K., Kubo, Y., Den, M. & Ishii, M. Deep Flare Net (DeFN) Model for Solar Flare Prediction.��� ������������� ����������, 113 (2018)

    2018

  62. [71]

    Domingo, V., Fleck, J. M. & Poland, A. The solar and heliospheric observatory (soho) mission: An overview.����� ����������, 3–21 (2009)

    2009

  63. [72]

    H.�� ���The Solar Oscillations Investigation - Michelson Doppler Imager.����� ����������, 129–188 (1995)

    Scherrer, P. H.�� ���The Solar Oscillations Investigation - Michelson Doppler Imager.����� ����������, 129–188 (1995)

    1995

  64. [73]

    D., Thompson, B

    Pesnell, W. D., Thompson, B. J. & Chamberlin, P. C. The Solar Dynamics Observatory (SDO).����� ����������, 3–15 (2012). 29

    2012

  65. [74]

    Schou, J.�� ���Design and Ground Calibration of the Helioseismic and Magnetic Imager (HMI) Instrument on the Solar Dynamics Observatory (SDO).����� ����������, 229–259 (2012)

    2012

  66. [75]

    Gan, W.�� ���The Advanced Space-Based Solar Observatory (ASO-S).����� ����������, 68 (2023)

    2023

  67. [76]

    Deng, Y.-Y.�� ���Design of the Full-disk MagnetoGraph (FMG) onboard the ASO-S.�������� �� ��������� ��� ��������������, 157 (2019)

    2019

  68. [77]

    P.�� ���EIT: Extreme-Ultraviolet Imaging Telescope for the SOHO Mission.����� ����������, 291–312 (1995)

    Delaboudini` ere, J. P.�� ���EIT: Extreme-Ultraviolet Imaging Telescope for the SOHO Mission.����� ����������, 291–312 (1995)

    1995

  69. [78]

    R.�� ���The Atmospheric Imaging Assembly (AIA) on the Solar Dynamics Observatory (SDO).����� ����������, 17–40 (2012)

    Lemen, J. R.�� ���The Atmospheric Imaging Assembly (AIA) on the Solar Dynamics Observatory (SDO).����� ����������, 17–40 (2012)

    2012

  70. [79]

    M.�� ���The GOES-R Solar UltraViolet Imager.����� ���������, e2022SW003044 (2022)

    Darnel, J. M.�� ���The GOES-R Solar UltraViolet Imager.����� ���������, e2022SW003044 (2022)

    2022

  71. [80]

    Dosovitskiy, A.�� ���An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale.����� ��������arXiv:2010.11929 (2020)

    2010 arXiv

  72. [81]

    He, K.�� ���Masked Autoencoders Are Scalable Vision Learners.����� �������� arXiv:2111.06377 (2021)

    2021 arXiv

  73. [82]

    & Zamir, A

    Bachmann, R., Mizrahi, D., Atanov, A. & Zamir, A. Avidan, S., Brostow, G., Ciss´ e, M., Farinella, G. M. & Hassner, T. (eds)��������� ����������� ������ ���� ������ ������������. (eds Avidan, S., Brostow, G., Ciss´ e, M., Farinella, G. M. & Hassner, T.)�������� ������ � ���� �...

    2022

  74. [83]

    E., Ellerman, F., Nicholson, S

    Hale, G. E., Ellerman, F., Nicholson, S. B. & Joy, A. H. The Magnetic Polarity of Sun-Spots.��� ������������� ���������, 153 (1919)

    1919

  75. [84]

    McIntosh, P. S. The Classification of Sunspot Groups.����� ����������, 251–267 (1990)

    1990

  76. [85]

    S.����������� ������� �� ��� ����������� ��������(Elsevier Academic Press, Cambridge, 2006)

    Wilks, D. S.����������� ������� �� ��� ����������� ��������(Elsevier Academic Press, Cambridge, 2006)

    2006

  77. [86]

    30 ���� �Schematic diagram of the SA-AI forecaster framework and the visual output of each module

    Cui, Y., Liu, S., Ercha, A.�� ���Verification of SPE Probability Forecasts at the Space Environment Prediction Center (SEPC).������� ����� ����� �������� ��, 1292–1298 (2016). 30 ���� �Schematic diagram of the SA-AI forecaster framework and the visual output of each module. It...

    2016

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.