Geomagnetic field modeling with dual-architecture neural networks

Abstract

Traditional geomagnetic field modeling requires balancing computational efficiency with the resolution of complex dynamics. A dual-architecture deep learning framework is introduced as a high-efficiency surrogate for global modeling. A sinusoidal representation network (SIREN) captures the continuous, quasi-static internal field, and a transformer with an ‘adaptive recursion’ mechanism models the dynamic external field. The recursive mechanism adjusts computational depth according to geomagnetic activity, optimizing efficiency and accuracy. Trained on CHAOS-8.4 data, the framework reproduces the global field with high fidelity. Under identical CPU hardware conditions, the SIREN-based surrogate model achieved an inference speed-up of approximately 1100 times compared to traditional implementations. Native graphics processing unit (GPU) acceleration further enables large scale parallel computations without additional engineering effort. This study validated the feasibility of deep learning as a robust real-time surrogate for complex geophysical simulations.

Keywords

Geomagnetic field modeling / Deep learning / Sinusoidal representation network / Recursive transformer / Surrogate modeling

References

[1] Bergen KJ, Johnson PA, de Hoop MV, Beroza GC. Machine learning for data-driven discovery in solid Earth geoscience. Science, 2019, 363(6433 eaau0323

[2] British Geological Survey (2025) World Data Centre for Geomagnetism, Edinburgh. Retrieved from https://wdc-dataportal.bgs.ac.uk/

[3] Camporeale E. The challenge of machine learning in space weather: nowcasting and forecasting. Space Weather, 2019, 17(8): 1166-1207

[4] Chen Z, Tian H, Li HM, Tang RX, Ouyang ZH, Deng XH. The study on the global evolution of energetic electron precipitation during geomagnetic storm based on deep learning algorithm. J Geophys Res-Space Phys, 2023, 128(3 e2022JA030974

[5] Conde D, Castillo FL, Escobar C, García C, García JE, Sanz V, et al.. Forecasting geomagnetic storm disturbances and their uncertainties using deep learning. Space Weather, 2023, 21(11 e2023SW003474

[6] Devlin J, Chang MW, Lee K, Toutanova K (2019) BERT: pre-training of deep bidirectional transformers for language understanding. In: Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). Association for Computational Linguistics, pp 4171‒4186. https://aclanthology.org/N19-1423/

[7] Dungey JW. Interplanetary magnetic field and the auroral zones. Phys Rev Lett, 1961, 6(2): 47-48

[8] Finlay CC, Olsen N, Tøffner-Clausen L. DTU candidate field models for IGRF-12 and the CHAOS-5 geomagnetic field model. Earth Planets Space, 2015, 67(1): 114

[9] Graves A (2016) Adaptive computation time for recurrent neural networks. Preprint at arXiv:1603.08983

[10] Hulot G, Balogh A, Christensen UR, Constable CG, Mandea M, Olsen N. The Earth’s magnetic field in the space age: an introduction to terrestrial magnetism. Space Sci Rev, 2010, 155(1‒4): 1-7

[11] King JH, Papitashvili NE. Solar wind spatial scales in and comparisons of hourly wind and ACE plasma and magnetic field data. J Geophys Res-Space Phys, 2005, 110(A2 A02104

[12] Kloss C (2024) ChaosMagPy: Python package for the CHAOS geomagnetic field model. Zenodo. https://doi.org/10.5281/zenodo.16091680

[13] Kloss C, Beggan C (2025) The 14th generation of the international geomagnetic reference field. In: EGU General Assembly Conference Abstracts. EGU25-3566. https://doi.org/10.5194/egusphere-egu25-3566

[14] Kloss C, Finlay CC, Olsen N, Tøffner-Clausen L, Gillet N, Grayver A. The CHAOS-8 geomagnetic field model. Earth Planets Space, 2026, 78 21

[15] Krizhevsky A, Sutskever I, Hinton GE (2012) ImageNet classification with deep convolutional neural networks. In: Advances in Neural Information Processing Systems 25 (NIPS 2012). NIPS, pp 1‒9. https://papers.nips.cc/paper_files/paper/2012/hash/c399862d3b9d6b76c8436e924a68c45b-Abstract.html

[16] Kuslits L, Horváth A, Wesztergom V, Beggan C, Rubóczki T, Prácser E, et al.. An alternate representation of the geomagnetic core field obtained using machine learning. Earth Planets Space, 2024, 76(1 77

[17] Laperre B, Amaya J, Lapenta G (2018) Prediction of the Dst-index using a long short-term memory network. https://www.iihe.ac.be/~dldecker/Doc/Abstracts/AstroGeoPlasma/Oral/abstract_Brecht_Laperre.pdf

[18] Li G, Zhou XH, Chen CJ, Xu LA, Zhou F, Shi FS, et al.. Multitype geomagnetic noise removal via an improved U-Net deep learning network. IEEE Trans Geosci Remote Sens, 2023, 61 5916512

[19] Li J, Chen B. The secular variation prediction method of geomagnetic field in Chinese mainland based on long short-term memory neural network. Acta Seismol Sin, 2025, 47(3): 390-409

[20] Liu H, Liu Z, Liu S, Liu YH, Bin JC, Shi F, et al.. A nonlinear regression application via machine learning techniques for geomagnetic data reconstruction processing. IEEE Trans Geosci Remote Sens, 2019, 57(1): 128-140

[21] Merrill RT, McElhinny MW, McFadden PL. The magnetic field of the earth: paleomagnetism, the core, and the deep mantle,, 1998Academic Press: 63

[22] Otzen M (2022) Geostatistical simulation and deep learning in geomagnetism. PhD thesis, Technical University of Denmark, Kgs. Lyngby

[23] Pelkum Donahue K, Inceoglu F. Forecasting solar flares with a transformer network. Front Astron Space Sci, 2024, 10 1298609

[24] Ren XD, Yang PX, Mei DK, Liu H, Xu GZ, Dong Y. Global ionospheric TEC forecasting for geomagnetic storm time using a deep learning‐based multi‐model ensemble method. Space Weather, 2023, 21(3 e2022SW003231

[25] Sakpal S (2024) Prediction of space weather events through analysis of active region magnetograms using convolutional neural network. Preprint at arXiv:2405.02545

[26] Sitzmann V, Martel JNP, Bergman AW, Lindell DB, Wetzstein G (2020) Implicit neural representations with periodic activation functions. In: 34th Conference on Neural Information Processing Systems (NeurIPS). NIPS, pp 7462‒7473

[27] Tapping KF. The 10.7 cm solar radio flux (F_10.7). Space Weather, 2013, 11(7): 394-406

[28] Tasistro‐Hart A, Grayver A, Kuvshinov A. Probabilistic geomagnetic storm forecasting via deep learning. J Geophys Res-Space Phys, 2021, 126(1 e2020JA028228

[29] Tóth G, Sokolov IV, Gombosi TI, Chesney DR, Clauer CR, De Zeeuw DL, et al.. Space weather modeling framework: a new tool for the space science community. J Geophys Res-Space Physics, 2005, 110(A12 A12226

[30] Upendran V, Tigas P, Ferdousi B, Bloch T, Cheung MC, Ganju S, et al.. Global geomagnetic perturbation forecasting using deep learning. Space Weather, 2022, 20(6 e2022SW003045

[31] Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN et al (2017) Attention is all you need. In: 31st Conference on Neural Information Processing Systems (NIPS 2017). NIPS, pp 1‒11. https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf

[32] Wintoft P, Wik M. Exploring three recurrent neural network architectures for geomagnetic predictions. Front Astron Space Sci, 2021, 8 664483

[33] Yang J (2024) Research on geomagnetic signal denoising in geomagnetic navigation based on deep learning. Master thesis, Chang'an University, Xi'an

[34] Zhang JY, Feng Y, Zhang JX, Li YJ. The short time prediction of the Dst index based on the long-short time memory and empirical mode decomposition–long-short time memory models. Appl Sci, 2023, 13(21 11824

Funding

Laoshan Laboratory(LSKJ202501100)

Key Laboratory of Polar Geology and Marine Mineral Resources (China University of Geosciences, Beijing), Ministry of Education(PGMR-2025-101)

National Natural Science Foundation of China(42076224)

RIGHTS & PERMISSIONS

The Author(s)