Data Processing of Magnetotelluric Survey Data Using Maximum Entropy Method and IIR Filter

Abstract

Data processing of magnetotelluric (MT) survey has been applied frequency by frequency based on the Fast Fourier Transform (FFT), since the FFT gives us the response functions (RFs) of the earth in the frequency domain directly. However, applying FFT processing to MT data may not be optimum. It is because the MT data is in general non-stationary since the source of MT is the transient fluctuation of electric current in the ionosphere. As well known, the FFT assumes time series to be a stationary so that we develop the data processing without FFT. We focus on an IIR filter called ‘pole on pedestal’ that extracts the signal at a specific frequency. Combining this IIR filter and the Hilbert transform, the RFs are calculated in time domain. In addition, it is important to remove the time segments contaminated by noise out of the whole recorded time series before calculating the RFs. Several coherences (for example, partial or multiple coherences) have usually been used to remove the segments contaminated by noise. However, it is known that these coherences are insufficient to select the segments. We apply the maximum entropy method (MEM) to the selection of the segments contaminated by large noise. The MEM searches and removes these contaminated segments easily. As a result we developed the time domain processing of MT data using MEM and IIR filter, and applied this processing to the real data acquired at the Nankai trough. Comparing the conventional and novel data processing, our novel data processing gives us more optimum RFs than the conventional processing.