10:00 〜 10:15
[SEM12-05] Development of time-domain noise reduction method using MSSA (Multi-channel Singular Spectrum Analysis) for Boso MT data:Verification at Kakioka
キーワード：MT法、MSSA（マルチチャンネル特異スペクトル解析）、ノイズ除去、房総半島
The Boso Peninsula is one of the areas of high crustal activity in Japan due to the triple junction of three plates (the North American plate, the Pacific plate, and the Philippine Sea plate) located near the peninsula. In particular, the source areas of huge earthquakes (The 1703 (Genroku, M8.2) and 1923 (Taisho, M7.9) Kanto earthquakes) are located around the southwest of the peninsula, and the Slow Slip Events occur in the southeast. The magnetotelluric (MT) survey was conducted from 2014 to 2016 to estimate the electrical resistivity distribution in the area. The electrical resistivity reflects temperature, pressure, and the portion of the pore water. Thus, the MT survey is expected to provide a new aspect of the information about the plate interface.
The MT method estimates the electrical resistivity from the ratio of the natural electromagnetic field variation. However, the observed data includes the noise caused by human activity (e.g., the leak current from the DC-driven railways), which is artificial noise. The noise affects the data in the frequency domain because the waveform of the artificial noise appears rectangular-like. Moreover, the noise collapses the assumption of the MT method that is correlated with the magnetic and electric field and has a large amplitude. The estimated MT response using the conventional MT noise reduction method shows the noise response and /or makes the estimation error large. To estimate the realistic MT response, we have developed a new noise reduction method in the time domain based on MSSA (Multi-channel Singular Spectrum Analysis) as a preprocess before the conventional noise reduction method in the frequency domain.
In the MSSA process, first, create Hankel matrixes (window length M × (data length N -M+1)) in each channel and then perform singular value decomposition (SVD) of the covariance matrix. After the SVD, each channel is decomposed to the C channel × M principal components (PC). We can reconstruct the time series by selecting the PC based on its characteristics, such as the correlation and contribution rate.
The abstract of the method follows: 7ch data (target site 5 components (horizontal magnetic field 2ch (Hx and Hy), vertical magnetic field 1ch (Hz), and horizontal electric field 2ch (Ex and Ey)) and reference site 2ch (magnetic field 2ch (Rx and Ry))) are used. First, apply MSSA to 7ch, then eliminate the trend. Second, apply MSSA to 4ch (Hx, Hy, Rx, and Ry) to extract solar-origin magnetic field fluctuation. Finally, apply MSSA to 7ch (extracted magnetic field by step 2, Hz, Ex, and Ey) to obtain the MT signal in Hz, Ex, and Ey.
In practice, by applying this method to the Boso MT data, which contains the rectangular-like noise by the artificial noise, we can reduce the effect of the noise even in the daytime which the observed data is highly polluted. The MT response after the MSSA process shows smoother curves and smaller estimation errors. Therefore, the new method is useful to remove the artificial noise and extract the MT signal.
To verify the capability of the MSSA-based noise reduction method to remove the artificial noise, the MSSA process was performed on the data at the Kakioka geomagnetic observatory (JMA), adding a rectangular pulse and spike. In this presentation, we show mainly the result of the verification.
The MT method estimates the electrical resistivity from the ratio of the natural electromagnetic field variation. However, the observed data includes the noise caused by human activity (e.g., the leak current from the DC-driven railways), which is artificial noise. The noise affects the data in the frequency domain because the waveform of the artificial noise appears rectangular-like. Moreover, the noise collapses the assumption of the MT method that is correlated with the magnetic and electric field and has a large amplitude. The estimated MT response using the conventional MT noise reduction method shows the noise response and /or makes the estimation error large. To estimate the realistic MT response, we have developed a new noise reduction method in the time domain based on MSSA (Multi-channel Singular Spectrum Analysis) as a preprocess before the conventional noise reduction method in the frequency domain.
In the MSSA process, first, create Hankel matrixes (window length M × (data length N -M+1)) in each channel and then perform singular value decomposition (SVD) of the covariance matrix. After the SVD, each channel is decomposed to the C channel × M principal components (PC). We can reconstruct the time series by selecting the PC based on its characteristics, such as the correlation and contribution rate.
The abstract of the method follows: 7ch data (target site 5 components (horizontal magnetic field 2ch (Hx and Hy), vertical magnetic field 1ch (Hz), and horizontal electric field 2ch (Ex and Ey)) and reference site 2ch (magnetic field 2ch (Rx and Ry))) are used. First, apply MSSA to 7ch, then eliminate the trend. Second, apply MSSA to 4ch (Hx, Hy, Rx, and Ry) to extract solar-origin magnetic field fluctuation. Finally, apply MSSA to 7ch (extracted magnetic field by step 2, Hz, Ex, and Ey) to obtain the MT signal in Hz, Ex, and Ey.
In practice, by applying this method to the Boso MT data, which contains the rectangular-like noise by the artificial noise, we can reduce the effect of the noise even in the daytime which the observed data is highly polluted. The MT response after the MSSA process shows smoother curves and smaller estimation errors. Therefore, the new method is useful to remove the artificial noise and extract the MT signal.
To verify the capability of the MSSA-based noise reduction method to remove the artificial noise, the MSSA process was performed on the data at the Kakioka geomagnetic observatory (JMA), adding a rectangular pulse and spike. In this presentation, we show mainly the result of the verification.