[M-IS09] Interdisciplinary studies on pre-earthquake processes

[MIS09-P05] A novel LF/VLF signal discrimination using random forest and interferometer approaches; identification of lightning discharge and possible seismo-EM signals

太田 悠一朗¹、*服部 克巳^2,3,4、三浦 健伸⁵、吉野 千恵²、今住 則之⁶ (1.千葉大学大学院融合理工学府、2.千葉大学大学院理学研究院、3.千葉大学環境リモートセンシング研究センター、4.千葉大学災害治療学研究所、5.千葉大学理学部、6.日本技術士会千葉県支部)

キーワード：LF/VLF信号の弁別、ランダムフォレスト、干渉計、雷放電素過程、地震電磁気信号

Electromagnetic phenomena preceding large earthquakes have been reported in various frequency bands. As an example, Yamada and Oike 1996 reported an increase in the number of pulses in the LF band before the 1995 M7.2 Hyogo-ken Nanbu earthquake. On the other hand, Izutsu 2007 pointed out that some of the increased pulses were likely due to lightning activity. Therefore, for the identification of VLF/LF signals related to earthquakes, the existence of electromagnetic emissions unrelated to earthquakes is a problem, especially the discrimination of electromagnetic emissions from lightning activity. Therefore, we have conducted the following two approaches to construct a system that can properly identify and remove electromagnetic emissions caused by lightning activity, which can be noise. The first approach is to use machine learning to identify and remove known lightning pulses from data obtained at a single observation point. In this study, a random forest, a type of machine learning, was constructed and applied to observed data. The second is to develop a VLF/LF broadband interferometer with antenna systems installed at several stations to construct a system to estimate electromagnetic radiation sources. In this study, observation tests were conducted at three stations: Chiba University, Asahi, and Miho. We compared estimated location of lightning by LIDEN with estimated location of lightning by our VLF/LF broadband interferometer. The details on the results of conducting these two approaches will be reported in the presentation.