Construction of a VLF/LF band interferometer using a capacitive circular flat-plane antenna and discrimination and identification of observed VLF/LF band signals by machine learning: Preliminary results

Japan is frequently hit by major earthquakes, such as the 2011 off the Pacific coast of Tohoku Earthquake and the 2024 Noto Peninsula Earthquake, which cause enormous human and economic losses. Short-term forecast of earthquakes is effective for mitigating such damage, but this has not been achieved to date. On the other hand, there have been reports of electromagnetic phenomena preceding major earthquakes in various frequency bands, including precursor phenomena in the VLF/LF band (3-300 kHz). In this study, we investigated earthquake-related VLF/LF signals, which has strong electromagnetic emissions due to lightning activity, and it is important to discriminate the VLF/LF signals from those due to lightning activity. In this study, two approaches were attempted: (1) development of a source localization method using VLF/LF broadband interferometry and (2) removal of signals caused by lightning discharges using machine learning.The first approach is expected to spatially discriminate between VLF/LF signals related to earthquakes (which are located near the epicenter and do not move) and signals related to lightning activity (which move with fronts and thunderclouds). The second is to utilize machine learning technology, which has been rapidly developed in recent years, for detection and removal of lightning discharge signals. For example, Wu et al. at Gifu University have succeeded in classifying lightning discharge waveforms in the thunderstorm activity process with an accuracy of approximately 99% using a machine learning technique called Random Forest. In this study, machine learning is expected to efficiently discriminate and eliminate known lightning discharge signals from a large amount of observation data with high accuracy, and analyze the remaining unknown signals to efficiently investigate the relationship between lightning and earthquakes. In this paper, we will describe the specific methods and results of the above two approaches.