In accordance with the request of the Department of Earth Sciences of the Russian Academy of Sciences (No. 13000/2312-351 dated 11/06/2024 "On providing information on the most important scientific results for 2024"), each of the laboratories of RS RAS has prepared formulations of its most important results obtained in 2024 as a part of the State assignment. On December 13, 2024, a regular meeting of the RS RAS Scientific Council was held, where the laboratory heads presented the results for the colleagues consideration. The presented results were discussed and, following the results of the voting, the results of the Laboratories of Deep Magnetotelluric Studies and the Laboratory of Complex Studies of Geodynamic Processes in Geophysical Fields were selected as the most important.
Result №1
When studying the relationship between electromagnetic and deformation processes on the territory of Bishkek Geodynamic Proving Ground, according to results of magnetotelluric monitoring, electromagnetic pulses associated with seismic events were detected. It was found that the Uchturfan earthquake (China, 01/22/2024 18:09:05 UT, M=7.0-7.27, at a depth of ~13 km) with numerous aftershocks (M=4.9–6.9), the foci of which are located at distances of ~450 km from the registration points, is reflected in all recorded parameters. At the same time a weaker earthquake (Kyrgyzstan, 03/04/2024 06:22:04.4, M=5.3, at a depth of ~3 km), but located closer, does not manifest itself in one of the horizontal components of the electromagnetic field. The reality of appearance of electromagnetic earthquake precursors and coseismic signals observed in the first tens of seconds or minutes after earthquake is shown.
Possible areas of practical application of the result: The research results can be used in the development of monitoring seismic activity methods in potentially dangerous regions.
Drawing. Recording of the response from the January 22, 2024 Uchturfan earthquake (arrow) and its aftershocks (dotted lines) by geophysical equipment (MTU-5 Phoenix Geophysics). The X–axis shows the observation time in minutes, and the Y-axis shows 5 components of the electromagnetic field.
The result was obtained within the framework of the National Assembly of the Russian Academy of Sciences State Assignment: the topic "Studying the deep structure and modern geodynamics of the Tien Shan lithosphere and surrounding areas using a set of geophysical methods" (the head of the topic is Director of the National Assembly of the Russian Academy of Sciences, Doctor of Physico–Mathematical Sciences Rybin Anatoly Kuzmich).
Authors: Bataleva E.A., Matyukov V.E., Nepeina K.S.
Publications. Research results have been reported at conferences and accepted for publication:
1) Bataleva E.A., Matyukov V.E., Nepeina K.S. Earthquake response in electromagnetic field components (Northern Tien Shan) // In the book: Problems of geodynamics and geoecology of intracontinental orogens. Abstracts of the IX International Symposium. Bishkek, 2024. pp. 266-269.
2) Bataleva E.A., Nepeina K.S., Matyukov V.E. Electromagnetic fields generated by the earthquakes of 01/22/2024 Mw 6.9 (China) and 03/04/2024 Mw 5.4 (Kyrgyzstan) at stationary points in the Northern Tien Shan // Problems of Geocosmos—2024, Springer Proceedings in Earth and Environmental Sciences, 2025. (in print)
3) Bataleva E.A., Matyukov V.E., Nepeina K.S. Anomalies in the behavior of components of the Earth's electromagnetic field and their relationship to earthquakes according to magnetotelluric monitoring data // Geodynamics and Tectonophysics, 2025. (in print).
Result №2
The CatalinNet artificial neural network has been developed to detect anomalies in variations of the geomagnetic field. The neural network is based on classical autoencoder architecture. The training sample consisted of daily changes in geomagnetic field magnitude on calm days for 2020, 2021 and 2022 at Ak-Suu (Northern Tien Shan) base station of the geomagnetic monitoring network of RS RAS. The neural network has 5 hidden layers with a total number of trainable parameters equal to 3.5×106. The trained neural network reproduces typical signs of normal data well, whereas in the case of observations containing various anomalies, it demonstrates a deterioration in the quality of recovery. This property of the model was used to divide the data into two classes: norm and anomaly. The recovery error in the form of mean absolute error (MAE) is used as a measure of data anomaly. Verification of the model based on test data for Ak-Suu station for 2017, 2018 and 2019. She showed that the neural network can confidently identify geomagnetic anomalies, in particular, the proportion of correctly identified anomalies was higher than 97% for all cases.
Possible areas of practical application of the result: the development of methodological approaches for detecting anomalies in geomagnetic data, for a comprehensive study of geophysical processes associated with the preparation of seismic events.
Figure. The general scheme of the CatalinNet neural network architecture. Input and output data – daily variations in the magnitude of the geomagnetic field The result was obtained within the framework of the State Assignment of RS RAS the topic "Study of geophysical fields and processes as the basis for earthquake forecasting based on monitoring and modeling of inelastic processes in seismogenerating media" (head of the topic – Leading Researcher, Head of the Laboratory of Complex Research of Geodynamic processes in Geophysical Fields of the Research Station of Russian Academy of Sciences, Ph.D. Imashev Sanzhar)
Author: Imashev S.A.
Publications:
1) Imashev S.A. Method for detecting anomalies in geomagnetic field variations based on artificial neural network // Geosystems of Transition Zones. 2024. vol. 8, No. 4. pp. 343–356. https://doi.org/10.30730/gtrz.2024.8.4.161-173
2) Imashev S.A. A program for detecting anomalies in variations in the magnitude of the geomagnetic field based on the CatalinNet neural network // Computer program registration certificate 2024686523, 11/11/2024
