A.М. Appalonov1, Yu.S. Maslennikova2, O.N. Sherstyukov3

1-3 Kazan Federal University (Kazan, Russia)

1 artem309_97@mail.ru; 2 jsmaslennikova@kpfu.ru; 3 Oleg.Sherstyukov@kpfu.ru 

Problem Statement. Radio communications, navigation systems, and telecommunications are highly dependent on the state of the ionosphere. One of the key parameters of the ionosphere is its complete electronic content. Changes in the concentration of free electrons affect the passage of radio waves through the ionosphere, therefore, global maps and the analysis of two-dimensional maps of this parameter make it possible to predict the conditions of signal propagation and optimize the operation of various radio electronic systems.

Recently, algorithms based on neural network approaches have become widely used to study various physical processes, as they have a greater generalizing ability than classical linear algorithms, which makes it possible to capture more complex relationships in the source data.

Goal Development of a new method for decomposing global maps of the complete electronic content of the ionosphere using neural networks, additionally comparing the results obtained with data from previous studies.

Results. A new method has been developed for decomposing the spatiotemporal components of ionospheric dynamics using global maps of the full electronic content for the 23rd, 24th, and 25th cycles of solar activity, produced using an autoencoder type neural network. The possibility of compressing the source maps to 10 main components is shown, while preserving 93% of the initial variance. The presence of an equatorial anomaly on the main spatial component of decomposition and the presence of a cyclic component of 25~27 days in the main temporal component is confirmed, which is consistent with the results of previous studies. Additionally, the features of the learning process of a neural network of this type in the context of the task are discussed.

Practical significance. The presented results make it possible to use a hidden representation of an autoencoder type neural network to reduce the data dimension of global maps of the electronic content of the ionosphere, which makes it easier to analyze, as well as use the new representation as data to build a model for predicting the global dynamics of the total electron concentration of the ionosphere.

Appalonov A.M., Maslennokova Yu.S., Sherstyukov O.N. Application of deep learning neural networks for the analysis of spatial and temporal components of the decomposition of the total electronic content of the ionosphere. Radiotekhnika. 2025. V. 89. № 1.
P. 172−179. DOI: https://doi.org/10.18127/j00338486-202501-16 (In Russian)

1. Brjunelli B.E., Namgaladze A.A. Fizika ionosfery. M.: Nauka. 1988. 527 s. (in Russian).
2. Mendillo M. Storms in the ionosphere: Patterns and processes for total electron content. Reviews of Geophysics. 2006. V. 44. № 4. DOI: 10.1029/2005RG000193.
3. Fitzgerald T.J. Observations of total electron content perturbation of GPS signals caused by a ground level explosion. J. Atoms. S. Terr. Phys. 1997. V. 59. P. 829-834.
4. Maksimov D.S., Kogogin D.A., Nasyrov I.A., Zagretdinov R.V. Vlijanie solnechnyh vspyshek 5-12 sentjabrja 2017 g. na regional'nuju vozmushhennost' ionosfery Zemli po dannym GNSS-stancij, raspolozhennyh v Privolzhskom federal'nom okruge Rossijskoj Federacii. Solnechno-zemnaja fizika. 2023. T. 9. № 2. S. 52-59 (in Russian).
5. Ivanov D.V., Ivanov V.A., Rjabova M.I., Ovchinnikov V.V. Jeffekty dispersii v ionosfernyh radiokanalah. Ch. 2. Vidy chastotno-vremennoj dispersii. Radiotehnika. 2024. T. 88. № 12. S.63−77. DOI: https://doi.org/10.18127/j00338486-202411-10 (in Russian).
6. Talaat E.R., Zhu X. Spatial and temporal variation of total electron content as revealed by principal component analysis. Ann. Geophys. 2016. № 34. Р. 1109–1117. DOI: 10.5194/angeo-34-1109-2016.
7. Ahmad Muhammad, Fatih Külahcı. A semi-supervised total electron content anomaly detection method using LSTM-auto-encoder. Journal of Atmospheric and Solar-Terrestrial Physics. 2022. V. 241. Р. 1364-6826. DOI: 10.1016/j.jastp.2022.105979.
8. Khristoforov S.V., Bochkarev V.V. Model of the global distribution of the total electron content based on deep dense convolutional autoencoder. Journal of Physics: Conference Series. 2018. V. 1141 012067. DOI: 10.1088/1742-6596/1141/1/012067.
9. Shhiryj A.O. Ispol'zovanie nejronnyh setej dlja dal'nejshego razvitija programmnoj chasti apparatno-programmnyh kompleksov radiozondirovanija ionosfery. Jelektromagnitnye volny i jelektronnye sistemy. 2024. T. 29. № 5. S. 55−60. DOI: https://doi.org/10.18127/j15604128-202405-08 (in Russian).
10. Appalonov A.M., Maslennikova Ju.S. Nejrosetevoe prognozirovanie dinamiki jekvatorial'noj anomalii po dannym polnogo jelektronnogo soderzhanija ionosfery. Tehnika radiosvjazi. 2021. № 3(50). S. 29–42. DOI: 10.33286/2075-8693-2021-50-29-42 (in Russian).
11. Maslennikova Yu.S., Bochkarev V.V. Principal Component Analysis of Global Maps of the Total Electronic Content. Geomagnetism and Aeronomy. 2014. V. 54(2). Р. 216–223. DOI: 10.1134/s0016793214020133.
12. Appalonov A.M., Maslennikova Ju.S. Analiz global'noj dinamiki polnogo jelektronnogo soderzhanija za 23-j i 24-j cikly solnechnoj aktivnosti s ispol'zovaniem metoda glavnyh komponent. Radiotehnika. 2023. T. 87. № 12. S.46−55. DOI: 10.18127/j00338486-202312-06 (in Russian).
13. Liou C.-Y., Cheng C.-W., Liou J.-W., Liou D.-R. Autoencoder for words. Neurocomputing. 2014. V. 139. Р. 84-96. DOI: 10.1016/j.neucom.2013.09.055.
14. LeCun Yann, Bengio Yoshua, Hinton Geoffrey. Deep learning. Nature. 2015. V. 521. P. 436-444. DOI: 10.1038/nature14539.
15. JPL. URL: https://www.jpl.nasa.gov (дата обращения: 15.11.2024).
16. Iwasawa Kenkichi. On some types of topological groups. Annals of Mathematics. 1949. V. 50. № 3. Р. 507–558. DOI: 10.2307/1969548. JSTOR 1969548.