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

[MIS09-P10] A development of noise reduction method for ULF band electromagnetic data using Multi-channel Singular Spectrum Analysis (MSSA) (4)

*Shu Kaneko¹, Toru Mogi², Chie Yoshino², Katsumi Hattori^2,3,4 (1.Graduate School of Science and Engineering, Chiba University, 2.Graduate School of Science, Chiba University, 3.Center for Environmental Remote Sensing, Chiba University, 4.Disaster Medicine Research Institute, Chiba University)

Keywords:MT method, Multi-channel Singular Spectrum Analysis (MSSA), Signal discrimination, Boso peninsula

The electromagnetic phenomena suspected to be related to earthquakes in the ULF band (f<10 Hz) are reported because the skin depth (more than a few km) corresponds to the interior of the crust of the plate boundary. However, the observed data include seismo-electromagnetic signals, natural electromagnetic field variations caused by solar-terrestrial coupling, and artificial noise caused by leak currents from DC-driven railways and power transmission lines. Since the amplitude of seismo-electromagnetic signals is smaller than that of natural magnetic field fluctuations and artificial noise, and especially similar to train noise, analysis has focused on nighttime (e.g., LT 2:30~4:00) when human activities are calm. Therefore, conventional analysis methods may miss seismic EM signals during most of the day. Thus, a method that can discriminate seismo-electromagnetic signals sufficiently, even if the signal-to-noise ratio is poor, is indispensable to extracting seismo-electromagnetic signals buried in other signal components. Therefore, we are developing a signal discrimination method based on Multi-channel Singular Spectrum Analysis (MSSA), which can decompose non-stationary time series.

In the past, a wavelet transform-based method was developed to discriminate the ULF band electromagnetic signal. The method decomposes the time series into wavelet coefficients and estimates the vertical magnetic and horizontal electric fields based on Inter Station Transfer Function (ISTF), then reconstructs by summing the Inverse wavelet transformed estimated wavelet coefficients of target bands. Therefore, once the well-defined ISTF is obtained, the method will be a powerful tool.

In this presentation, we will suggest the appropriate signal discrimination method for noisy data by looking into the performance of relatively clean data (Kakioka (Japan Meteorological Agency) data sets) and noisy data at the Boso peninsula, through the difference of the procedures.