The Research Of Magnetotelluric Time-domain Waveform Denoising

Magneto telluric is a electromagnetic sounding method based on electromagnetic induction. The method used regional or global distributed natural electromagnetic field observed on the surface to study the electrical properties and distributed characteristics of subsurface information. This method has many advantages such as high efficiency, low cost, deep exploration depth, good vertical and horizontal resolution capability, small topographic effects and weak shielding effect of high resistivity layer. With the Ministry of Land and Resources to carry out the Sinoprobe project, magnetotelluric as a highly effective techniques probing the deep earth, plays an irreplaceable role in Sinoprobe.In the actual measurement, however, magnetotelluric is susceptible to electromagnetic noise interference, especially in the mining area. Large number of magnetotelluric sites collected in these areas are disturbed by strong noise. We can often found many class square waveforms, class triangle waveforms and other man-made noise waveforms in the time series of those sites. This article is based on an assumption that the waveforms such as class square waveforms is not from natural field, but from mining area as a result of a variety of noise. On the basis of this assumption, we first add simulation waveforms to the time series of100Ω-m uniform half-space model to study the influence of magnetotelluric sounding curve contaminated by noise waveforms.Aiming at the characteristics of time domain noise waveform, we propose a method called top-hat transform for noise separation, and summarize the scope of application and limitations of this method. To improving the accuracy of waveform denoising, we study a method about noise waveform recognition and waveform fitting for signal to noise separation algorithm, then we build a program base on MATLAB which can remove the noise waveforms from time series. We use this program to remove simulation waveforms and measured magnetotelluric site, and compare the processing result with which result from morphological filtering. We analyzed the treatment effect of these two methods at the same time. Finally we summarizes the significance of time domain denoising in improve the quality of magnetotelluric sounding curves by contrasting different sounding curves of de-noised magnetotelluric sites.