Pith. sign in

REVIEW 5 major objections 6 minor 56 references

Extreme Solar Storm Reveals Causal Interactions in Space Weather

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Information-theoretic synergy flux maps the Sun-to-Earth storm chain and powers a more accurate extreme-storm forecaster.

desk verdict Fresh causal-discovery idea, but the forecast evaluation leaks test-period information through the regularizer, so treat the performance claims skeptically. read the letter →

arxiv 2508.06507 v1 pith:I2X3RDMQ submitted 2025-07-27 physics.space-ph astro-ph.EP

Xinan Dai , Haiyang Fu , Zichong Yan , Zitong Wang , Feng Xu , Chi Wang , Yuhong Liu , YaQiu Jin This is my paper
classification physics.space-phastro-ph.EP
keywords spaceweathercausaldiscoverysynergyfluxpartialinformationdecompositiongeomagneticstormforecastingsolar-magnetosphere-ionospherecouplingauroralovaltransfer
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

This paper tries to establish that a statistical quantity called synergy flux, computed from an information-theoretic decomposition of multivariate data, can expose the causal chain that carries a solar storm from the Sun to Earth's ionosphere. Using 273 geomagnetic storms from 1980 to 2024, the authors construct directed causal graphs with stable time delays linking solar activity, solar wind, magnetospheric indices, and vertical electron content. The same synergy-flux measure is then used as a training constraint in SolarAurora, a neural-network forecaster built on spherical Fourier neural operators. The authors report that SolarAurora beats SFNO and Swin-Transformer baselines in predicting F10.7, Dst, Kp, and VTEC during the May and October 2024 extreme storms. If the causal interpretation holds, the framework offers both a discovery tool for hidden magnetosphere-ionosphere couplings and a more reliable way to forecast extreme events.

What carries the argument

The load-bearing object is synergy flux, a quantity extracted from a synergistic-unique-redundant decomposition (a form of partial information decomposition) of the observational variables; it measures information that only appears when a group of source variables is considered together, and the paper uses it to draw directed edges, assign causal strengths and time delays, and build spatial causal maps. That same flux defines the causal-consistency loss that regularizes SolarAurora, whose forecasting backbone is an encoder–SFNO-blocks–decoder network operating on spherical data. The machinery connects discovery and prediction: the causal edges are both the scientific output and the training constraint.

What would settle it

Run SolarAurora and the synergy-flux causal graphs on synthetic data generated from a known coupled dynamical system with prescribed drivers and confounders, and compare recovered edges to ground truth. A second decisive check: retrain SolarAurora with shuffled or randomized causal-consistency constraints; if the forecasting improvement over SFNO and Swin-Transformer survives randomization, the gain is not attributable to causal information.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a synergy-flux measure derived from partial information decomposition recovers a well-defined, hierarchical causal landscape for the coupled solar-magnetosphere-ionosphere system, including causal loops and stable lag times that match known physics. During the May 2024 extreme storm, the method reveals auroral spatial causality patterns in the Arctic that are not visible in standard VTEC maps, and it identifies intensified direct solar-wind control and reciprocal solar wind-magnetosphere coupling (FP↔B and Qi↔B) as features of extreme storms. The authors further claim that a model trained with causal-consistency regularization—SolarAurora—forecasts the May and October 2024 events with lower error than SFNO and Swin-Transformer across multiple variables.

Load-bearing premise

The load-bearing premise is that synergy flux measures physical causation, not mere correlation or confounding, because the same quantity both generates the scientific causal graphs and supplies the training constraint for SolarAurora.

Editorial extensions

If this is right

  • If synergy flux tracks true causation, causal time delays such as the 4.5-hour SPD→Dst lag can be used as fixed early-warning lead times for geomagnetic storm arrival.
  • Auroral spatial causality maps offer a new indicator of the geographic region most at risk of storm impacts, including power-grid stress, beyond what VTEC maps show.
  • Causality-informed training improves forecast accuracy during extreme storms while keeping predictions interpretable.
  • The reciprocal couplings FP↔B and Qi↔B, observed in extreme storms, give testable predictions: future storms with Dst below about -350 nT should show similar reciprocal couplings.

Reading between the lines

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

  • A decisive test of the causal claim would be to run the same synergy-flux analysis on synthetic time series generated from a known coupled dynamical system; if the recovered edges are confounded by shared drivers or smooth trends, the method's physical conclusions would need reinterpretation.
  • The stable time delays could be embedded directly into the model architecture as inductive biases; the paper notes delays are not yet explicitly included, so adding them is a natural next step.
  • If the auroral causality map is a genuine precursor, it could be monitored in near-real time to alert operators of high-risk regions, something the paper does not develop.
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

5 major / 6 minor

Summary. The paper proposes an information-theoretic method, termed 'synergy flux', based on a synergistic-unique-redundant decomposition of multivariate information, to infer spatiotemporal causal graphs among solar wind, magnetospheric, and ionospheric variables during geomagnetic storms. Using data from 1980 to October 2024, it reports stable causal chains and time delays across 273 storms, identifies auroral spatial causality patterns during the May 2024 extreme storm, and introduces SolarAurora, a neural operator model regularized by a causal-consistency loss. The model is evaluated on the May and October 2024 storms and is claimed to outperform SFNO and Swin-Transformer baselines on metrics such as MAPE for Dst, F10.7, and Kp, and RMSE for a 12-variable set.

Significance. If substantiated, the work would be significant for space weather research: it offers an information-decomposition approach to extreme-event causality and a forecasting model that explicitly incorporates causal structure. The paper has several praiseworthy features: it uses public multi-decadal data, provides an explicit data/code availability statement, reports a falsifiable claim about time-delay stability across 273 storms, and benchmarks against two modern deep learning baselines. However, the central method is currently not defined in the main text, the forecast evaluation lacks uncertainty quantification and ablation, and the causal-consistency constraints appear to be derived from a record that includes the test storms. These issues currently prevent the findings from being evaluated or reproduced, so the significance is conditional on addressing them.

major comments (5)
  1. [Figure 2A caption; Figure 4 (causal loss module)] The causal graphs used to define the causal-consistency regularizer are computed from the full record, January 1, 1980 to October 14, 2024, which includes the May and October 2024 storms used for evaluation (Fig. 2A caption). Since SolarAurora is trained with a causal loss derived from these graphs and the training set ends March 29, 2024, the test storms indirectly contribute information to the model through the regularizer, while the SFNO and Swin-Transformer baselines receive no such test-period information. This is a load-bearing leakage because the reported gains, e.g., 13.42% MAPE improvement for Dst, could reflect test-event-tuned inductive bias rather than a general forecasting advantage. Please recompute all causal graphs and causal-consistency constraints using only data up to March 29, 2024 (or otherwise demonstrate that the included 2024 events do not affect the estimated graphs), and rerun the evaluation.
  2. [Results, 'Synergy flux reveals spatio-temporal interaction landscapes' (pages 3-5)] The core quantity, 'synergy flux', is never defined in the main text; the paper refers to a 'synergistic-unique-redundant decomposition method' and to supplementary figures, but no equation, estimator, or parameter choices (lag, binning, significance threshold) are provided. Because both the claimed causal discoveries (Fig. 2) and the SolarAurora causal-loss module (Fig. 4) rest entirely on this quantity, the central method is not reproducible or verifiable as submitted. Please include a precise definition of synergy flux, including the partial information decomposition formulation and the empirical estimator, in the main text or ensure the supplementary material is fully included in the submission.
  3. [Causality-informed extreme space weather prediction; Figure 5] The forecast comparison reports MAPE reductions and radar-chart RMSE values without error bars, confidence intervals, or significance tests, and there is no ablation of the causal-loss component. For example, the claimed Dst gain of 13.42% over SFNO is a single point estimate, and the evaluation covers only three storms, which is too small to assess robustness. Please report uncertainty over ensembles, days, or storms, and compare SolarAurora against an identical architecture trained without the causal-consistency loss to isolate the contribution of the causal constraint.
  4. [Quantitative Evaluation of Causality Learning] The causal interpretation of synergy flux is asserted rather than demonstrated. The same information-theoretic measure is used both to claim physical causal chains and to regularize the forecast model, so any mismatch between statistical synergy and physical causation would propagate into both the discovery and forecasting claims. Please validate the method against established causal-discovery benchmarks (e.g., PCMCI or Granger causality) and provide a concrete test of whether the identified 'matched time delays' predict held-out storm dynamics rather than only describing the same data from which they were estimated.
  5. [Causality-informed extreme space weather prediction (training period)] The manuscript states that the training dataset is from October 19, 2014 to March 29, 2024, but the causal analysis uses 273 storms from January 1980 to October 2024, and the abstract claims '1980-2024 datasets' as a whole. Please clarify exactly which data segments are used for (i) causal graph estimation, (ii) model training, (iii) hyperparameter selection, and (iv) evaluation, and state explicitly whether the 2024 event data are used in any form for model development.
minor comments (6)
  1. [Discussion (paragraph on long-term causal connectivity)] The text 'Comparison of long-term causal connectivity ... with all storm results (Fig. 1A)' should refer to the causal graph in Fig. 2A, not Fig. 1A, which is a schematic overview.
  2. [Causality-informed extreme space weather prediction] The placeholder 'Supplementary Table??' appears in the model evaluation section; the table listing the 12 observational variables is missing and should be included.
  3. [References] Reference 35 is cited as 'Nat. Rev. Earth Environ. 10, 2553 (2023)'; the correct volume is 4, pages 487-505. Please verify all references for accuracy.
  4. [Results, causal loops paragraph] The term 'causal loops' is used for structures such as SPD→Dst→VTEC with SPD→VTEC; these are directed paths, not graph-theoretic loops. Please define the terminology or use a different term such as 'matched direct/indirect pathways'.
  5. [Figure 5c] The radar charts lack axis labels and units; please specify the RMSE metric, the normalization, and the number of evaluation horizons.
  6. [Abstract] The abstract says 'using 1980-2024 datasets', while the forecast model uses only 2014-2024 training data with 2024 test events; please align the wording to distinguish the causal-analysis period from the training period.

Circularity Check

1 steps flagged · score 6.0 of 10

SolarAurora's forecast gain is partially circular because its causal-consistency regularizer is inferred from the same 1980-Oct 2024 record that contains the May/Oct 2024 test storms.

  1. fitted input called prediction [Causality-informed extreme space weather prediction (p.9); Fig. 2 and Fig. 4 captions]
    "The training dataset is from Oct 19, 2014 to March 29, 2024. ... (A) Directed causal graph for Statistical results from 273 geomagnetic storms (January 1, 1980 to October 14, 2024). ... The Causal Module (upper panel) supervises the model by enforcing spatio-temporal causal consistency, using synergy flux derived from information decomposition to construct spatial causal maps (SCM) and temporal causal graphs (TCG)."

    The causal graphs that define the SCM/TCG regularizer are computed from 273 storms spanning Jan 1, 1980 to Oct 14, 2024, which includes the May 2024 and October 2024 events later used to score SolarAurora. The training set ends Mar 29, 2024, so those test-period causal structures enter the model's loss as fitted inputs. Forecasting those same events is then presented as prediction (Fig. 5), while SFNO and Swin-Transformer receive no test-period information. The reported MAPE improvements (e.g., 13.42% for Dst) may therefore reflect test-event-tuned inductive bias rather than an out-of-sample gain. This is a fitted-input-called-prediction pattern: the regularizer is fitted to the target events, then the target events are 'predicted' to demonstrate the regularizer's value.

full rationale

The paper's causal-discovery component (synergy flux) is not circular by itself: it is a defined information-theoretic quantity applied to multivariate solar-terrestrial data, and the resulting causal chains are compared against known physics. The forecast architecture is also a conventional encoder-SFNO-decoder model with a hybrid loss. The central circularity is temporal: the causal-consistency loss appears to be built from synergy-flux graphs spanning the full 1980-Oct 2024 record, including the May and October 2024 storms on which the model is evaluated, even though the supervised training data end in March 2024. This contaminates the evaluation, because the regularizer is fitted to causal structure of the exact test events. The paper also acknowledges that synergy/redundancy quantification remains approximate, which further weakens the causal claims, but this is a limitation rather than circularity. There are no load-bearing self-citations or imported uniqueness theorems. Score 6 reflects a partial reduction: the forecast advantage is not forced by construction (the model can still fail despite these constraints), but the test-period causal fits make the headline May/October 2024 gains non-independent evidence for the method.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claims rest on the assumption that information-theoretic measures from the observed time series capture true physical causality, plus several pragmatic choices (lag, binning, loss weighting, extreme-storm threshold) not reported in the main text. There are no independent physical constraints or formal check on these assumptions.

free parameters (3)
  • causal loss weight
    The relative weight of the causal-consistency term in the hybrid loss is not reported, yet it controls the model's behavior.
  • information estimator parameters (lag, binning)
    The synergy flux computation requires lag, binning, and estimator choices; none are specified in the main text.
  • extreme storm threshold = Dst < -350 nT
    The definition of 'extreme' used to identify common coupling patterns.
assumptions (4)
  • domain assumption Information-theoretic synergy measures computed from time series are valid proxies for causal interactions in the solar-magnetosphere-ionosphere system.
    The paper treats synergy flux as revealing 'causal chains' and uses it as a training target without justification beyond referencing prior PID work.
  • domain assumption The observed variables (12 in total) are sufficient to capture the causal structure with no hidden confounders.
    The causal graph includes only these variables; unmeasured drivers (e.g., solar wind composition, magnetospheric state) could confound the inferred links.
  • domain assumption Stability of causal time delays across 273 storms (1980-2024) implies the process is stationary.
    Used to pool storms and report near-invariant delays.
  • standard math Standard information theory identities (PID, Shannon entropy) hold.
    Background from Williams and Beer and Kolchinsky.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Extreme Solar Storm Reveals Causal Interactions in Space Weather." pith.science (2026). https://pith.science/paper/I2X3RDMQ

@misc{pith2026250806507,
  author       = {Pith},
  title        = {Pith review of: Extreme Solar Storm Reveals Causal Interactions in Space Weather},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I2X3RDMQ}},
  note         = {Machine review of arXiv:2508.06507}
}
read the original abstract

Solar storms perturb Earth's magnetosphere, triggering geomagnetic storms that threaten space-based systems and infrastructure. Despite advances in spaceborne and ground-based observations, the causal chain driving solar-magnetosphere-ionosphere dynamics remains elusive due to multiphysics coupling, nonlinearity, and cross-scale complexity. This study presents an information-theoretic framework to decipher interaction mechanisms in extreme solar geomagnetic storms across intensity levels within space weather causal chains, using 1980-2024 datasets. Unexpectedly, we uncover auroral spatial causality patterns associated with space weather threats in the Arctic during May 2024 extreme storms. By integrating causal consistency constraints into spatiotemporal modeling, SolarAurora outperforms existing frameworks, achieving superior accuracy in forecasting May/October 2024 events. These results advance understanding of space weather dynamics and establish a promising framework for scientific discovery and forecasting extreme space weather events.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

56 extracted references · 50 canonical work pages

  1. [1]

    Minnis, Solar activity and the ionosphere

    C. Minnis, Solar activity and the ionosphere. Nature 181, 543–544 (1958)

    work page 1958

  2. [2]

    Similar patterns are found, with the B→Qi link appearing in the October 2024 storm and the FP↔B and Qi↔B couplings emerging in extreme geomagnetic storms ( 𝐷𝑠𝑡 <−350) (Fig

    Another feature is reciprocal solar wind-magnetosphere coupling FP↔B emerges via newly observed B→FP (May) pathway. Similar patterns are found, with the B→Qi link appearing in the October 2024 storm and the FP↔B and Qi↔B couplings emerging in extreme geomagnetic storms ( 𝐷𝑠𝑡 <−350) (Fig. S1 in supplementary materials). Therefore, we speculate that solar w...

    work page 2024

  3. [3]

    Voosen, NASA seeks to revive lost probe that traced solar storms

    P. Voosen, NASA seeks to revive lost probe that traced solar storms. Science 359, 505 (2018), doi:10.1126/science.359.6375.505

    work page doi:10.1126/science.359.6375.505 2018

  4. [4]

    Normile, Solar Storm Knocks Out Japanese Satellite

    D. Normile, Solar Storm Knocks Out Japanese Satellite. Science 289, 709–709 (2000), doi:10.1126/ science.289.5480.709a

    work page 2000

  5. [5]

    Gou, et al

    T. Gou, et al. , The birth of a coronal mass ejection. Science Advances 5, eaau7004 (2019), doi: 10.1126/sciadv.aau7004

    work page doi:10.1126/sciadv.aau7004 2019

  6. [6]

    Greshko, Extreme solar storm generated aurorae—and ‘surprise’

    M. Greshko, Extreme solar storm generated aurorae—and ‘surprise’. Science (2024), doi:10.1126/ science.zoakd5z

    work page 2024

  7. [7]

    Kusano, et al., A physics-based method that can predict imminent large solar flares

    K. Kusano, et al., A physics-based method that can predict imminent large solar flares. Science 369, 587–591 (2020), doi:10.1126/science.aaz2511

    doi:10.1126/science.aaz2511 2020

  8. [8]

    A. M. Veronig, Can we predict solar flares? Science 369, 504–505 (2020), doi:10.1126/science. abb6150

    doi:10.1126/science 2020

Show all 56 references
  1. [9]

    B. M. Walsh, et al. , Simultaneous Ground- and Space-Based Observations of the Plasmaspheric Plume and Reconnection. Science 343, 1122–1125 (2014), doi:10.1126/science.1247212

    2014 doi

  2. [10]

    Soicher, Response of electrons in ionosphere and plasmasphere to magnetic storms

    H. Soicher, Response of electrons in ionosphere and plasmasphere to magnetic storms. Nature 259, 33–35 (1976)

    1976

  3. [11]

    Similar spatial causality dynamics are noted during the Oct 11, 2024 storm (Dst =∼− 341 nT) (Fig

    The coverage and shape of such auroral spatial causality patterns aligns with previous auroral oval observations (47) (48). Similar spatial causality dynamics are noted during the Oct 11, 2024 storm (Dst =∼− 341 nT) (Fig. S3 in supplementary materials), suggesting a recurrent ...

    2024

  4. [12]

    Keiling, et al

    A. Keiling, et al. , Assessing the global Alfv ´en wave power flow into and out of the auro- ral acceleration region during geomagnetic storms. Science Advances 5, eaav8411 (2019), doi: 10.1126/sciadv.aav8411

    2019 doi

  5. [13]

    S. E. Milan, et al., Overview of Solar Wind–Magnetosphere–Ionosphere–Atmosphere Coupling and the Generation of Magnetospheric Currents. Space Sci. Rev. 206, 547–573 (2017)

    2017

  6. [14]

    J. G. Lyon, The Solar Wind-Magnetosphere-Ionosphere System. Science 288, 1987–1991 (2000), doi:10.1126/science.288.5473.1987. 12

    2000

  7. [15]

    M. C. Kelley, The Earth’s Ionosphere: Plasma Physics and Electrodynamics (2nd ed.) (Academic Press) (2012)

    2012

  8. [16]

    Voosen, GPS satellites yield space weather data

    P. Voosen, GPS satellites yield space weather data. Science 355 (6324), 443 (2017), doi:10.1126/ science.355.6324.443

    2017

  9. [17]

    Gibney, Space-weather forecast to improve with European satellite

    E. Gibney, Space-weather forecast to improve with European satellite. Nature 541, 271 (2017)

    2017

  10. [18]

    Smith, et al., Mapping the ionosphere with millions of phones

    J. Smith, et al., Mapping the ionosphere with millions of phones. Nature 635, 365–369 (2024)

    2024

  11. [19]

    Wang, New chains of space weather monitoring stations in China.Space Weather 8, 08001 (2010)

    C. Wang, New chains of space weather monitoring stations in China.Space Weather 8, 08001 (2010)

    2010

  12. [20]

    Yue, et al., A tristatic phased array radar system in China

    X. Yue, et al., A tristatic phased array radar system in China. Nat. Astron. 8, 673 (2024)

    2024

  13. [21]

    D. N. Baker, How to cope with space weather. Science 297, 1486–1487 (2002)

    2002

  14. [22]

    Hapgood, Prepare for the coming space weather storm

    M. Hapgood, Prepare for the coming space weather storm. Nature 484, 311–313 (2012)

    2012

  15. [23]

    Bolduc, Geomagnetic storms and their impact on power systems.Nat

    L. Bolduc, Geomagnetic storms and their impact on power systems.Nat. Hazards29, 193–211 (2002)

    2002

  16. [24]

    Castelvecchi, Huge SpaceX rocket explosion shredded the upper atmosphere

    D. Castelvecchi, Huge SpaceX rocket explosion shredded the upper atmosphere. Nature (2024)

    2024

  17. [25]

    Chapman, J

    S. Chapman, J. Bartels, Geomagnetism, Volume 1 (Clarendon Press) (1940)

    1940

  18. [26]

    Hasegawa, et al., Transport of solar wind into Earth’s magnetosphere through rolled - up Kelvin - Helmholtz vortices

    H. Hasegawa, et al., Transport of solar wind into Earth’s magnetosphere through rolled - up Kelvin - Helmholtz vortices. Nature 430 (7001), 755–758 (2004)

    2004

  19. [27]

    Zhang, et al

    Q.-H. Zhang, et al. , Direct observations of the evolution of polar cap ionization patches. Science 339 (6127), 1597–1600 (2013), doi:10.1123/science.1231487

    2013 doi

  20. [28]

    Ebihara, T

    Y. Ebihara, T. Tanaka, N. Kamiyoshikawa, New diagnosis for energy flow from solar wind to iono- sphere during substorm: Global MHD simulation. J. Geophys. Res.: Space Phys. 124, 360–378 (2019)

    2019

  21. [29]

    B. T. Tsurutani, G. S. Lakhina, R. Hajra, The physics of space weather/solar-terrestrial physics (STP): what we know now and what the current and future challenges are.Nonlinear Processes in Geophysics 27, 75–119 (2020)

    2020

  22. [30]

    Hayakawa, The Solar and Geomagnetic Storms in May 2024: A Flash Data Report

    H. Hayakawa, The Solar and Geomagnetic Storms in May 2024: A Flash Data Report. Astrophys. J. 979, 49 (2025). 13

    2025

  23. [31]

    C. W. J. Granger, Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37, 424–438 (1969)

    1969

  24. [32]

    Schreiber, Measuring information transfer

    T. Schreiber, Measuring information transfer. Phys. Rev. Lett. 85, 461–464 (2000)

    2000

  25. [33]

    Pearl, Causality: Models, Reasoning, and Inference (Cambridge University Press) (2000)

    J. Pearl, Causality: Models, Reasoning, and Inference (Cambridge University Press) (2000)

    2000

  26. [34]

    X. S. Liang, R. Kleeman, Information transfer between dynamical system components. Phys. Rev. Lett. 95, 244101 (2005)

    2005

  27. [35]

    C. E. Shannon, A mathematical theory of communication. Bell Syst. Tech. J. 27, 379–423 (1948)

    1948

  28. [36]

    Runge, et al., Inferring causation from time series in Earth system sciences

    J. Runge, et al., Inferring causation from time series in Earth system sciences. Nat. Commun. 10, 2553 (2019)

    2019

  29. [37]

    Runge, et al., Causal inference for time series

    J. Runge, et al., Causal inference for time series. Nat. Rev. Earth Environ. 10, 2553 (2023)

    2023

  30. [38]

    Stumpo, et al., Causal inference in space weather by an information theory approach

    M. Stumpo, et al., Causal inference in space weather by an information theory approach. J. Phys. Conf. Ser. 1548, 012019 (2020)

    2020

  31. [39]

    P. L. Williams, R. D. Beer, Nonnegative decomposition of multivariate information (2010), arXiv:1004.2515

    2010 arXiv

  32. [40]

    Kolchinsky, A novel approach to the partial information decomposition

    A. Kolchinsky, A novel approach to the partial information decomposition. Entropy 24, 403 (2022)

    2022

  33. [41]

    T. F. Varley, et al. , Partial entropy decomposition reveals higher - order information structures in human brain activity. Proc. Natl. Acad. Sci. USA 120 (2023)

    2023

  34. [42]

    A. I. Luppi, et al., Information decomposition and the informational architecture of the brain.Trends Cogn. Sci. 28, 352–368 (2024)

    2024

  35. [43]

    Pathak, et al., Fourcastnet: A global data - driven high - resolution weather model using adaptive Fourier neural operators (2022), arXiv preprint arXiv:2202.11214

    J. Pathak, et al., Fourcastnet: A global data - driven high - resolution weather model using adaptive Fourier neural operators (2022), arXiv preprint arXiv:2202.11214

    2022 arXiv

  36. [44]

    Lam, et al., Learning skillful medium - range global weather forecasting.Science 382, 1416–1421 (2023)

    R. Lam, et al., Learning skillful medium - range global weather forecasting.Science 382, 1416–1421 (2023)

    2023

  37. [45]

    Bi, et al., Accurate medium - range global weather forecasting with 3D neural networks

    K. Bi, et al., Accurate medium - range global weather forecasting with 3D neural networks. Nature 619, 533–538 (2023). 14

    2023

  38. [46]

    Kochkov, J

    D. Kochkov, J. Yuval, I. Langmore,et al., Neural general circulation models for weather and climate. Nature (2024)

    2024

  39. [47]

    D. R. Themens, et al., The high latitude ionospheric response to the major May 2024 geomagnetic storm: A synoptic view. Geophys. Res. Lett. 51, e2024GL111677 (2024)

    2024

  40. [48]

    D. K. Karan, et al. , GOLD observations of the merging of the Southern Crest of the equatorial ionization anomaly and aurora during the 10 and 11 May 2024 Mother’s Day super geomagnetic storm. Geophys. Res. Lett. 51, e2024GL110632 (2024)

    2024

  41. [49]

    Q. H. Zhang, et al., A space hurricane over the Earth’s polar ionosphere. Nat. Commun. 12, 1207 (2021)

    2021

  42. [50]

    Zhang, et al., Direct observations of the evolution of polar cap ionization patches.Science 339, 1597–1600 (2013)

    Q.-H. Zhang, et al., Direct observations of the evolution of polar cap ionization patches.Science 339, 1597–1600 (2013)

    2013

  43. [51]

    S. R. Strom, G. Iwanaga, Overview and History of the Defense Meteorological Satellite Program. Crosslink. Aerosp. Corporation Mag. Adv. Aerosp. Technol. 6 (1) (2005)

    2005

  44. [52]

    J. Huba, G. Joyce, J. Krall, Three-dimensional equatorial spread F modeling. Geophys. Res. Lett. 35, L10102 (2008)

    2008

  45. [53]

    C. Wang, G. Branduardi-Raymont, C. P. Escoubet, et al. , Solar Wind Magnetosphere Ionosphere Link Explorer (SMILE): Science and Mission Overview.Space Sci. Rev. 221, 9 (2025), doi:10.1007/ s11214-024-01126-6

    2025

  46. [54]

    Bonev, et al., Spherical Fourier Neural Operators: Learning Stable Dynamics on the Sphere (2023), doi:10.48550/arXiv.2306.03838, arXiv preprint arXiv:2306.03838

    B. Bonev, et al., Spherical Fourier Neural Operators: Learning Stable Dynamics on the Sphere (2023), doi:10.48550/arXiv.2306.03838, arXiv preprint arXiv:2306.03838

    2023 doi arXiv

  47. [55]

    Vaswani, et al., Attention is all you need, in Advances in Neural Information Processing Systems, vol

    A. Vaswani, et al., Attention is all you need, in Advances in Neural Information Processing Systems, vol. 30 (2017), pp. 5998–6008

    2017

  48. [56]

    Liu, et al., Swin Transformer: Hierarchical Vision Transformer using Shifted Windows, in Pro- ceedings of the IEEE International Conference on Computer Vision (ICCV)(2021), pp

    Z. Liu, et al., Swin Transformer: Hierarchical Vision Transformer using Shifted Windows, in Pro- ceedings of the IEEE International Conference on Computer Vision (ICCV)(2021), pp. 10012–10022. 15 Acknowledgments We are grateful to Prof. X. S. Liang for inspiring discussion on ...

    2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.