Russian Journal of Earth Sciences Russian Journal of Earth Sciences Russian Journal of Earth Sciences 1681-1208 90657 10.2205/2025ES000996 budhdy Спецвыпуск: "Наука о данных, геоинформатика и системный анализ в изучении Земли" Special Issue: “Data Science, Geoinformatics and Systems Analysis in Geosciences” Спецвыпуск: "Наука о данных, геоинформатика и системный анализ в изучении Земли" Artificial Neural Network for Downward Continuation of Anomalous Magnetic Fields Artificial Neural Network for Downward Continuation of Anomalous Magnetic Fields https://orcid.org/0000-0002-7963-3673 Рытов Руслан Алексеевич Rytov Ruslan Alekseevich ruslan.rytov2017@yandex.ru

кандидат физико-математических наук;

candidate of physical and mathematical sciences;

 Институт земного магнетизма, ионосферы и распространения радиоволн им. Н.В.Пушкова РАН (ИЗМИРАН), г. Москва, г. Троицк, Россия Москва Россия Pushkov Institute of Terrestrial Magnetism, Ionosphere and Radio Wave Propagation of the Russian Academy of Sciences (IZMIRAN), Moscow, Troitsk, Russia Moscow Russian Federation 23 05 2025 23 05 2025 25 2 1 5 15 11 2024 15 04 2025 https://rjes.ru/en/nauka/article/90657/view

The downward continuation of an anomalous magnetic field is used for many applications in geophysics. However, such a problem is ill-posed, so it does not have a unique and stable solution. In this paper, we propose an artificial neural network architecture for the downward continuation of the vertical component of an anomalous geomagnetic field measured in a plane at a given height. The inverse problem is solved here by a direct method: the neural network is trained to reconstruct such a distribution of the magnetic field Bdown, which after a stable upward continuation corresponds to the measured field Bup. The performance of the neural network was demonstrated using the example of an anomalous geomagnetic field obtained using the Enhanced Magnetic Model.

The downward continuation of an anomalous magnetic field is used for many applications in geophysics. However, such a problem is ill-posed, so it does not have a unique and stable solution. In this paper, we propose an artificial neural network architecture for the downward continuation of the vertical component of an anomalous geomagnetic field measured in a plane at a given height. The inverse problem is solved here by a direct method: the neural network is trained to reconstruct such a distribution of the magnetic field Bdown, which after a stable upward continuation corresponds to the measured field Bup. The performance of the neural network was demonstrated using the example of an anomalous geomagnetic field obtained using the Enhanced Magnetic Model.

 stray magnetic field magnetic anomaly untrained neural networks inverse modeling downward continuation stray magnetic field magnetic anomaly untrained neural networks inverse modeling downward continuation The study was supported by the Russian Science Foundation grant No. 24-27-00250. The study was supported by the Russian Science Foundation grant No. 24-27-00250.

Abadi M., Agarwal A., Barham P., et al. TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems. — 2016. — DOI: 10.48550/arXiv.1603.04467. Abadi M., Agarwal A., Barham P., et al. TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems. — 2016. — DOI: 10.48550/arXiv.1603.04467. Blakely R. J. Potential Theory in Gravity and Magnetic Applications. — Cambridge University Press, 1995. — DOI: 10.1017/cbo9780511549816. Blakely R. J. Potential Theory in Gravity and Magnetic Applications. — Cambridge University Press, 1995. — DOI: 10.1017/cbo9780511549816. Buchanan A., Finn C. A., Love J. J. Geomagnetic referencing the real-time compass for directional drillers // Oilfield Review. — 2013. — Vol. 25, no. 3. — P. 32–47. Buchanan A., Finn C. A., Love J. J. Geomagnetic referencing the real-time compass for directional drillers // Oilfield Review. — 2013. — Vol. 25, no. 3. — P. 32–47. Coskun U. H., Sel B., Plaster B. Magnetic field mapping of inaccessible regions using physics-informed neural networks // Scientific Reports. — 2022. — Vol. 12, no. 1. — DOI: 10.1038/s41598-022-15777-4. Coskun U. H., Sel B., Plaster B. Magnetic field mapping of inaccessible regions using physics-informed neural networks // Scientific Reports. — 2022. — Vol. 12, no. 1. — DOI: 10.1038/s41598-022-15777-4. Dubois A. E. E., Broadway D. A., Stark A., et al. Untrained Physically Informed Neural Network for Image Reconstruction of Magnetic Field Sources // Physical Review Applied. — 2022. — Vol. 18, no. 6. — DOI: 10.1103/physrevapplied.18.064076. Dubois A. E. E., Broadway D. A., Stark A., et al. Untrained Physically Informed Neural Network for Image Reconstruction of Magnetic Field Sources // Physical Review Applied. — 2022. — Vol. 18, no. 6. — DOI: 10.1103/physrevapplied.18.064076. Kaji C. V., Hoover R. C., Ragi S. Underwater Navigation using Geomagnetic Field Variations // 2019 IEEE International Conference on Electro Information Technology (EIT). — IEEE, 2019. — P. 1–6. — DOI: 10.1109/eit.2019.8834192. Kaji C. V., Hoover R. C., Ragi S. Underwater Navigation using Geomagnetic Field Variations // 2019 IEEE International Conference on Electro Information Technology (EIT). — IEEE, 2019. — P. 1–6. — DOI: 10.1109/eit.2019.8834192. Kingma D. P., Ba J. A. Adam: A Method for Stochastic Optimization. — 2014. — DOI: 10.48550/ARXIV.1412.6980. Kingma D. P., Ba J. A. Adam: A Method for Stochastic Optimization. — 2014. — DOI: 10.48550/ARXIV.1412.6980. Nielsen M. A. Neural networks and deep learning. — San Francisco, CA, USA : Determination press, 2015. — P. 15–24. Nielsen M. A. Neural networks and deep learning. — San Francisco, CA, USA : Determination press, 2015. — P. 15–24. Pollok S., Bjørk R., Jørgensen P. S. Inverse Design of Magnetic Fields Using Deep Learning // IEEE Transactions on Magnetics. — 2021. — Vol. 57, no. 7. — P. 1–4. — DOI: 10.1109/tmag.2021.3082431. Pollok S., Bjørk R., Jørgensen P. S. Inverse Design of Magnetic Fields Using Deep Learning // IEEE Transactions on Magnetics. — 2021. — Vol. 57, no. 7. — P. 1–4. — DOI: 10.1109/tmag.2021.3082431. Pollok S., Olden-Jørgensen N., Jørgensen P. S., et al. Magnetic field prediction using generative adversarial networks // Journal of Magnetism and Magnetic Materials. — 2023. — Vol. 571. — DOI: 10.1016/j.jmmm.2023.170556. Pollok S., Olden-Jørgensen N., Jørgensen P. S., et al. Magnetic field prediction using generative adversarial networks // Journal of Magnetism and Magnetic Materials. — 2023. — Vol. 571. — DOI: 10.1016/j.jmmm.2023.170556. The National Centers for Environmental Information. — 2018. — URL: https://www.ngdc.noaa.gov/geomag/geomag. shtml (visited on 10/14/2024). The National Centers for Environmental Information. — 2018. — URL: https://www.ngdc.noaa.gov/geomag/geomag. shtml (visited on 10/14/2024). Tikhonov A. N., Arsenin V. Y. Solutions of ill-posed problems. — New York : Winston, 1977. Tikhonov A. N., Arsenin V. Y. Solutions of ill-posed problems. — New York : Winston, 1977.