pith. sign in

arxiv: 2604.18379 · v1 · submitted 2026-04-20 · 💻 cs.LG · eess.SP· physics.geo-ph· physics.space-ph

Forecasting Ionospheric Irregularities on GNSS Lines of Sight Using Dynamic Graphs with Ephemeris Conditioning

Mert Can Turkmen , Eng Leong Tan , Yee Hui Lee This is my paper

Pith reviewed 2026-05-10 05:06 UTC · model grok-4.3

classification 💻 cs.LG eess.SPphysics.geo-phphysics.space-ph
keywords ionospheric forecastingdynamic graphsGNSSephemeris conditioningirregularitiesgraph neural networksROTIprobabilistic classification
0
0 comments X

The pith

Dynamic graphs conditioned on future satellite positions forecast ionospheric irregularities on lines of sight that appear only later.

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

The paper establishes that representing ionospheric observations as nodes in a dynamic graph, with edges shifting according to satellite motion, permits conditioning the forecast model on the known future graph structure. This ephemeris conditioning supports probabilistic predictions of irregularities specifically on lines of sight that become active only during the forecast window. Tested on multi-GNSS data from a Singapore receiver pair over more than two years, the resulting IonoDGNN model reaches a Brier Skill Score of 0.49 and PR-AUC of 0.75, outperforming persistence by 35% and 52% respectively, with larger gains at longer lead times and for newly rising satellites. Ablations confirm that both the graph structure and the future conditioning contribute, and the model retains skill on unobserved nodes through spatial propagation from neighbors.

Core claim

We model the ionosphere as a dynamic graph over ionospheric pierce points (IPPs), with connectivity that evolves as satellite positions change. Because satellite trajectories are predictable, the graph topology over the forecast horizon can be constructed in advance. We exploit this property to condition forecasts on the future graph structure, which we term ephemeris conditioning. This enables prediction on lines of sight that appear only in the forecast horizon. The IonoDGNN model achieves a Brier Skill Score of 0.49 and a PR-AUC of 0.75 on real multi-GNSS data, improving over persistence baselines.

What carries the argument

Dynamic graph over ionospheric pierce points with ephemeris conditioning on the future connectivity derived from satellite positions.

If this is right

  • The model can issue predictions for satellites that rise only during the two-hour forecast window.
  • Under simulated coverage dropout, skill is retained on affected nodes via message passing from observed neighbors.
  • Both graph structure and ephemeris conditioning contribute to gains, with larger improvements at longer lead times.
  • Dynamic graphs preserve the actual time-varying sampling of GNSS observations better than fixed-grid products.

Where Pith is reading between the lines

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

  • The same conditioning technique could apply to other sensor networks whose observation geometry evolves predictably.
  • Operational GNSS services might incorporate such forecasts to anticipate effects on newly acquired satellite links.
  • Message passing on evolving graphs may prove useful for other atmospheric parameters measured along moving lines of sight.

Load-bearing premise

That spatial dependencies among ionospheric pierce points can be captured by message passing on a graph whose edges are set by current and future satellite positions.

What would settle it

A controlled test in which ephemeris conditioning is removed and the receiver operating characteristic AUC for satellites that rise during the forecast window falls from 0.95 to near 0.5.

Figures

Figures reproduced from arXiv: 2604.18379 by Eng Leong Tan, Mert Can Turkmen, Yee Hui Lee.

Figure 1
Figure 1. Figure 1: Examples of cross-station QC outcomes. (a) Confirmed event: both NTUS and SIN1 detect a coincident [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: IonoDGNN architecture. The history module ( [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Threshold-independent evaluation of all model variants. (a) ROC curves. (b) Precision–recall curves; the [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Forecast skill as a function of lead time for all model variants. (a) Brier Skill Score. (b) Precision–recall AUC. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Model predictions (top row) and absolute error (bottom row) at +60 minute lead time for two test-set events. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: BSS and PR-AUC for retained and dropped nodes under simulated coverage loss. Dashed red line indicates [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Predicted event probabilities for samples excluded from training. (a) Unverified segments, where the second [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
read the original abstract

Most data-driven ionospheric forecasting models operate on gridded products, which do not preserve the time-varying sampling structure of satellite-based sensing. We instead model the ionosphere as a dynamic graph over ionospheric pierce points (IPPs), with connectivity that evolves as satellite positions change. Because satellite trajectories are predictable, the graph topology over the forecast horizon can be constructed in advance. We exploit this property to condition forecasts on the future graph structure, which we term ephemeris conditioning. This enables prediction on lines of sight that appear only in the forecast horizon. We evaluate our framework on multi-GNSS (Global Navigation Satellite System) data from a co-located receiver pair in Singapore spanning January 2023 through April 2025. The task is to forecast Rate of TEC Index (ROTI)-defined irregularities at 5-minute cadence up to 2 hours ahead as binary probabilistic classification per node. The resulting model, IonoDGNN, achieves a Brier Skill Score (BSS) of 0.49 and a precision-recall area under the curve (PR-AUC) of 0.75, improving over persistence by 35\% in BSS and 52\% in PR-AUC, with larger gains at longer lead times. Ablations confirm that graph structure and ephemeris conditioning each contribute meaningfully, with conditioning proving essential for satellites that rise during the forecast horizon (receiver operating characteristic AUC: 0.95 vs.\ 0.52 without). Under simulated coverage dropout, the model retains predictive skill on affected nodes through spatial message passing from observed neighbors. These results suggest that dynamic graph forecasting on evolving lines of sight is a viable alternative to grid-based representations for ionospheric irregularity forecasting. The model and evaluation code will be released upon publication.

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

0 major / 4 minor

Summary. The paper introduces IonoDGNN, a dynamic graph neural network framework for forecasting ionospheric irregularities on GNSS lines of sight. It represents the ionosphere as a graph over ionospheric pierce points (IPPs) whose connectivity evolves according to satellite positions; because trajectories are predictable, the model conditions predictions on the future graph topology (ephemeris conditioning). This enables probabilistic forecasts of ROTI-defined irregularities for lines of sight that appear only during the forecast horizon. Evaluated on multi-GNSS data from a co-located receiver pair in Singapore (January 2023–April 2025) at 5-minute cadence up to 2 hours ahead, the model achieves BSS = 0.49 and PR-AUC = 0.75, improving 35 % and 52 % over persistence, with larger gains at longer lead times. Ablations confirm contributions from graph structure and ephemeris conditioning (ROC AUC 0.95 vs. 0.52 without conditioning for rising satellites), and the model retains skill under simulated coverage dropout via spatial message passing.

Significance. If the results hold, the work offers a principled alternative to gridded ionospheric models by preserving the native, time-varying sampling geometry of GNSS observations. Ephemeris conditioning exploits a domain-specific predictability property to solve the otherwise intractable problem of forecasting on unseen lines of sight. The reported ablations and robustness to node dropout provide concrete evidence that dynamic graph message passing captures useful spatial dependencies. Code release upon publication will support reproducibility and adoption in GNSS and space-weather applications.

minor comments (4)
  1. Abstract: the data span (January 2023 through April 2025) should be accompanied by explicit train/validation/test split dates and a statement confirming temporal separation to preclude leakage.
  2. Methods section (graph construction): provide the precise rule used to define edges between IPPs (e.g., distance threshold, visibility criteria) and how the adjacency matrix is updated at each time step.
  3. Experiments: the persistence baseline is appropriate, but its probabilistic formulation (how the forecast distribution is obtained from past observations) should be stated explicitly for reproducibility.
  4. Results: the statement of 'larger gains at longer lead times' would be strengthened by a table or supplementary figure reporting BSS and PR-AUC stratified by lead time (e.g., 15 min, 30 min, 60 min, 120 min).

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and detailed summary of our work, for highlighting the significance of the dynamic graph approach with ephemeris conditioning, and for recommending minor revision. We appreciate the recognition that the framework preserves the native sampling geometry of GNSS observations and provides evidence for the value of spatial message passing.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper describes an empirical dynamic graph neural network (IonoDGNN) trained on held-out multi-GNSS data to forecast ROTI irregularities. Performance metrics (BSS 0.49, PR-AUC 0.75) and ablations (graph structure, ephemeris conditioning) are evaluated on future observations not used for fitting. No derivation step reduces a claimed prediction to a fitted input by construction, no self-definitional equations appear, and no load-bearing self-citations or uniqueness theorems are invoked. The framework remains self-contained against external persistence baselines and dropout simulations.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard domain assumptions from ionospheric physics and graph neural networks; no new physical entities are introduced and no free parameters are explicitly fitted in the abstract description.

axioms (1)
  • domain assumption Ionospheric irregularities can be represented at discrete ionospheric pierce points whose connectivity evolves with satellite geometry.
    This underpins the dynamic graph construction described in the abstract.

pith-pipeline@v0.9.0 · 5639 in / 1380 out tokens · 53215 ms · 2026-05-10T05:06:46.481441+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

51 extracted references · 51 canonical work pages

  1. [1]

    X. Pi, A. J. Mannucci, U. J. Lindqwister, and C. M. Ho. Monitoring of global ionospheric irregularities using the worldwide gps network.Geophysical Research Letters, 24(18):2283–2286, 1997

    work page 1997

  2. [2]

    L. Liu, Y . J. Morton, and Y . Liu. Ml prediction of global ionospheric tec maps.Space Weather, 20(9):e2022SW003135, 2022

    work page 2022

  3. [4]

    Ionobench: Evaluating spatiotemporal models for ionospheric forecasting under solar-balanced and storm-aware conditions.Remote Sensing, 17(15):2557, 2025

    Mert Can Turkmen, Yee Hui Lee, and Eng Leong Tan. Ionobench: Evaluating spatiotemporal models for ionospheric forecasting under solar-balanced and storm-aware conditions.Remote Sensing, 17(15):2557, 2025

    work page 2025

  4. [5]

    Hernandez-Pajares, J

    M. Hernandez-Pajares, J. M. Juan, and J. Sanz. The ionosphere: effects, gps modeling and the benefits for space geodetic techniques.Journal of Geodesy, 85(12):887–907, 2011

    work page 2011

  5. [6]

    Krypiak-Gregorczyk, P

    A. Krypiak-Gregorczyk, P. Wielgosz, and A. Borkowski. Ionosphere model for european region based on multi-gnss data and tps interpolation.Remote Sensing, 9(12):1221, 2017

    work page 2017

  6. [7]

    Martire, S

    L. Martire, S. Krishnamoorthy, P. Vergados, L. J. Romans, B. Szilágyi, X. Meng, J. L. Anderson, A. Komjáthy, and Y . E. Bar-Sever. The guardian system-a gnss upper atmospheric real-time disaster information and alert network.GPS Solutions, 27(1):32, 2023

    work page 2023

  7. [8]

    Accuracy and consistency of different global ionospheric maps released by igs ionosphere associate analysis centers.Advances in Space Research, 65(1):163–174, 2020

    Peng Chen, Hang Liu, Yongchao Ma, and Naiquan Zheng. Accuracy and consistency of different global ionospheric maps released by igs ionosphere associate analysis centers.Advances in Space Research, 65(1):163–174, 2020

    work page 2020

  8. [9]

    Performance and consistency of final global ionospheric maps from different igs analysis centers.Remote Sensing, 15(4):1010, 2023

    Wei Li, Keke Wang, and Kaitian Yuan. Performance and consistency of final global ionospheric maps from different igs analysis centers.Remote Sensing, 15(4):1010, 2023

    work page 2023

  9. [10]

    X. Ren, X. Zhang, W. Xie, K. Zhang, Y . Yuan, and X. Li. Global ionospheric modelling using multi-gnss: Beidou, galileo, glonass and gps.Scientific Reports, 6:33499, 2016

    work page 2016

  10. [11]

    G. O. Jerez, M. Hernández-Pajares, A. Goss, F. S. Prol, D. B. M. Alves, J. F. G. Monico, and M. Schmidt. Two-way assessment of ionospheric maps performance over the brazilian region: Global versus regional products.Space Weather, 21:e2022SW003252, 2023

    work page 2023

  11. [12]

    Bhattacharyya

    A. Bhattacharyya. Equatorial plasma bubbles: A review.Atmosphere, 13(10):1637, 2022

    work page 2022

  12. [13]

    P. M. Kintner, B. M. Ledvina, and E. R. De Paula. Gps and ionospheric scintillations.Space Weather, 5(9):S09003, 2007

    work page 2007

  13. [14]

    C. R. Aguiar, J. F. G. Monico, and A. O. Moraes. Impact of ionospheric scintillations on gnss availability and precise positioning.Space Weather, 23(2):e2024SW004217, February 2025

    work page 2025

  14. [15]

    Geirhos et al

    R. Geirhos et al. Shortcut learning in deep neural networks.Nature Machine Intelligence, 2(11):665–673, 2020

    work page 2020

  15. [16]

    Zhao et al

    X. Zhao et al. A novel short-term prediction model for regional equatorial plasma bubble irregularities in east and southeast asia.Space Weather, 23(2):e2024SW004224, Feb 2025

    work page 2025

  16. [17]

    Atabati et al

    A. Atabati et al. Prediction of ionospheric scintillation with convgru networks using gnss data.Remote Sensing, 16(15):2757, 2024

    work page 2024

  17. [18]

    Liu et al

    H. Liu et al. The short-term prediction of low-latitude ionospheric irregularities leveraging a hybrid ensemble model.IEEE Transactions on Geoscience and Remote Sensing, 62:1–15, 2024

    work page 2024

  18. [19]

    Atabati, M

    A. Atabati, M. Alizadeh, H. Schuh, and L.-C. Tsai. Ionospheric scintillation prediction on s4 and roti parameters using artificial neural network and genetic algorithm.Remote Sensing, 13(11):2092, 2021

    work page 2092

  19. [20]

    Trachuentong, P

    S. Trachuentong, P. Supnithi, L. M. M. Myint, and S. Saito. Ionospheric scintillation prediction using decision tree and rainforest techniques. In2024 10th International Conference on Engineering, Applied Sciences, and Technology (ICEAST), pages 89–92, Luang Prabang, Laos, May 2024. IEEE. 15 Forecasting Ionospheric Irregularities on GNSS Lines of Sight Us...

    work page 2024

  20. [21]

    Nowcasting of amplitude ionospheric scintillation based on machine learning techniques.IEEE Transactions on Aerospace and Electronic Systems, 58(6):4917–4925, 2022

    Otávio Carvalho, Pedro Augusto Araujo da Silva de Almeida Nava Alves, Ricardo Yvan de la Cruz Cueva, and Alex Oliveira Barradas Filho. Nowcasting of amplitude ionospheric scintillation based on machine learning techniques.IEEE Transactions on Aerospace and Electronic Systems, 58(6):4917–4925, 2022

    work page 2022

  21. [22]

    Y . Chen, Y . Liu, K. Yang, L. Li, C. Xiong, and J. Wang. Ionospheric time series prediction method based on spatio-temporal graph neural network.Atmosphere, 16(6):732, 2025

    work page 2025

  22. [23]

    F. Yu, H. Yuan, S. Chen, R. Luo, and H. Luo. Graph-enabled spatio-temporal transformer for ionospheric prediction.GPS Solutions, 28(4):203, 2024

    work page 2024

  23. [24]

    H. S. Kelebek et al. Ioncast: A deep learning framework for forecasting ionospheric dynamics. InProceedings of the Conference, 2025. Preprint available at arXiv:2511.15004

    work page arXiv 2025

  24. [25]

    Kaselimi, N

    M. Kaselimi, N. Doulamis, A. Doulamis, and D. Delikaraoglou. Geometric deep learning for ionospheric tec modeling using a temporal graph convolutional network.Neural Computing and Applications, 37(22):17179– 17192, 2025

    work page 2025

  25. [26]

    Global 4d ionospheric stec prediction based on deeponet for gnss rays.IEEE Transactions on Geoscience and Remote Sensing, 2024

    Dijia Cai, Zenghui Shi, Haiyang Fu, Huan Liu, Hongyi Qian, Yun Sui, Feng Xu, and Ya-Qiu Jin. Global 4d ionospheric stec prediction based on deeponet for gnss rays.IEEE Transactions on Geoscience and Remote Sensing, 2024

    work page 2024

  26. [27]

    Global ionospheric modeling using multi- gnss: A machine learning approach

    Shuyin Mao, Grzegorz Kłopotek, Yuanxin Pan, and Benedikt Soja. Global ionospheric modeling using multi- gnss: A machine learning approach. InProceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pages 5774–5777. IEEE, 2024

    work page 2024

  27. [28]

    Generic dynamic graph convolutional network for traffic flow forecasting.Information Fusion, 100:101946, 2023

    Yi Xu, Liangzhe Han, Tongyu Zhu, Leilei Sun, Bowen Du, and Weifeng Lv. Generic dynamic graph convolutional network for traffic flow forecasting.Information Fusion, 100:101946, 2023

    work page 2023

  28. [29]

    Dahg: A dynamic augmented heterogeneous graph framework for precipitation forecasting with incomplete data.Information, 16(11):946, 2025

    Hao Tang, Hao Yang, and Wei Zhang. Dahg: A dynamic augmented heterogeneous graph framework for precipitation forecasting with incomplete data.Information, 16(11):946, 2025

    work page 2025

  29. [30]

    Gnss combined ultra-rapid solution summary product

    International GNSS Service (IGS). Gnss combined ultra-rapid solution summary product. NASA Crustal Dynamics Data Information System (CDDIS), 2000

    work page 2000

  30. [31]

    B. Lim, S. Ö. Arık, N. Loeff, and T. Pfister. Temporal fusion transformers for interpretable multi-horizon time series forecasting.International Journal of Forecasting, 37(4):1748–1764, 10 2021

    work page 2021

  31. [32]

    Chartier, Cathryn N

    Alex T. Chartier, Cathryn N. Mitchell, and Daniel R. Jackson. New capabilities for prediction of high-latitude ionospheric scintillation: A novel approach with machine learning.Space Weather, 16(11):1817–1836, 2018

    work page 2018

  32. [33]

    B. A. Carter, J. L. Currie, T. Dao, E. Yizengaw, J. Retterer, M. Terkildsen, K. Groves, and R. Caton. On the assessment of daily equatorial plasma bubble occurrence modeling and forecasting.Space Weather, 18(9):e2020SW002555, 2020

    work page 2020

  33. [34]

    Dow, Ruth E

    John M. Dow, Ruth E. Neilan, and Chris Rizos. The international gnss service in a changing landscape of global navigation satellite systems.Journal of Geodesy, 83:191–198, 2009

    work page 2009

  34. [35]

    Spatio-temporal characteristics of ionospheric irregularities in low latitude regions during the peak of solar cycle 25.Advances in Space Research, 76(1):254–268, 2025

    Napat Tongkasem, Pornchai Supnithi, Phimmasone Thammavongsy, Michi Nishioka, Septi Perwitasari, Susumu Saito, Jeff Klenzing, and Lin Min Min Myint. Spatio-temporal characteristics of ionospheric irregularities in low latitude regions during the peak of solar cycle 25.Advances in Space Research, 76(1):254–268, 2025

    work page 2025

  35. [36]

    Multi-gnss products and services at igmas wuhan innovation application center: strategy and evaluation.Satellite Navigation, 3:20, 2022

    Xingxing Li, Qingyun Wang, Jiaqi Wu, Yongqiang Yuan, Yun Xiong, Xuewen Gong, and Zhilu Wu. Multi-gnss products and services at igmas wuhan innovation application center: strategy and evaluation.Satellite Navigation, 3:20, 2022

    work page 2022

  36. [37]

    J. H. King and N. E. Papitashvili. Solar wind spatial scales in and comparisons of hourly wind and ace plasma and magnetic field data.Journal of Geophysical Research: Space Physics, 110(A2):A02104, 2005

    work page 2005

  37. [38]

    W. Li, S. Song, and X. Jin. Ionospheric scintillation monitoring with roti from geodetic receiver: Limitations and performance evaluation.Radio Science, 57(5):e2021RS007420, May 2022

    work page 2022

  38. [39]

    J. T. Emmert, A. D. Richmond, and D. P. Drob. A computationally compact representation of magnetic-apex and quasi-dipole coordinates with smooth base vectors.Journal of Geophysical Research: Space Physics, 115(A8):A08322, 2010

    work page 2010

  39. [40]

    L. Yang, C. Chatelain, and S. Adam. Dynamic graph representation learning with neural networks: A survey. IEEE Access, 12:43460–43484, 2024

    work page 2024

  40. [41]

    Fast graph representation learning with PyTorch Geometric

    Matthias Fey and Jan Eric Lenssen. Fast graph representation learning with PyTorch Geometric. InInternational Conference on Learning Representations (ICLR) Workshop on Representation Learning on Graphs and Manifolds, 2019. 16 Forecasting Ionospheric Irregularities on GNSS Lines of Sight Using Dynamic Graphs with Ephemeris Conditioning

    work page 2019

  41. [42]

    Pytorch geometric temporal: Spatiotemporal signal processing with neural machine learning models

    Benedek Rozemberczki, Paul Scherer, Yixuan He, George Panagopoulos, Alexander Riedel, Maria Astefanoaei, Oliver Kiss, Ferenc Beres, Guzman Lopez, Nicolas Collignon, and Rik Sarkar. Pytorch geometric temporal: Spatiotemporal signal processing with neural machine learning models. InProceedings of the 30th ACM International Conference on Information and Know...

    work page 2021

  42. [43]

    Tgm: a modular and efficient library for machine learning on temporal graphs

    Jacob Chmura, Shenyang Huang, Tran Gia Bao Ngo, Ali Parviz, Farimah Poursafaei, Jure Leskovec, Michael Bronstein, Guillaume Rabusseau, Matthias Fey, and Reihaneh Rabbany. Tgm: a modular and efficient library for machine learning on temporal graphs. InAdvances in Neural Information Processing Systems (NeurIPS) Datasets and Benchmarks Track, 2025

    work page 2025

  43. [44]

    How attentive are graph attention networks? InarXiv, 2021

    Shaked Brody, Uri Alon, and Eran Yahav. How attentive are graph attention networks? InarXiv, 2021

    work page 2021

  44. [45]

    Gomez, Łukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. InProceedings of the 31st International Conference on Neural Information Processing Systems (NeurIPS), pages 6000–6010, 2017

    work page 2017

  45. [46]

    The precision-recall plot is more informative than the roc plot when evaluating binary classifiers on imbalanced datasets.PLoS ONE, 10(3):e0118432, 2015

    Takaya Saito and Marc Rehmsmeier. The precision-recall plot is more informative than the roc plot when evaluating binary classifiers on imbalanced datasets.PLoS ONE, 10(3):e0118432, 2015

    work page 2015

  46. [47]

    Foster and Adrian N

    Matthew P. Foster and Adrian N. Evans. An evaluation of interpolation techniques for reconstructing ionospheric tec maps.IEEE Transactions on Geoscience and Remote Sensing, 46(7):2153–2164, 2008

    work page 2008

  47. [48]

    Autocorrelation method for temporal interpolation and short-term prediction of ionospheric data.Radio Science, 34(2):459–464, 1999

    Plamen Muhtarov and Ivan Kutiev. Autocorrelation method for temporal interpolation and short-term prediction of ionospheric data.Radio Science, 34(2):459–464, 1999

    work page 1999

  48. [49]

    Prayitno Abadi et al. Modeling post-sunset equatorial spread-F occurrence as a function of evening upward plasma drift using logistic regression, deduced from ionosondes in southeast asia.Remote Sensing, 14(8):1896, April 2022

    work page 2022

  49. [50]

    LargeST: A benchmark dataset for large-scale traffic forecasting

    Zezhi Liu, Yuxuan Liang, Ting Hua, Yaguan Li, and Roger Zimmermann. LargeST: A benchmark dataset for large-scale traffic forecasting. InAdvances in Neural Information Processing Systems, volume 36, 2023

    work page 2023

  50. [51]

    Gampii-good: Gnss observations and products downloader

    Zhou, Feng and collaborators. Gampii-good: Gnss observations and products downloader. GitHub repository,

    work page

  51. [52]

    Version 3.1, GPL-3.0 license. 17

    work page