Polarization Analysis Of Magnetotelluric Time Series Based On Wavelet Transform

Since 1990s, magnetotelluric sounding method has been developed rapidly and been used widely in the research of regional fracture structure, fluid motion, plate tectonics and continental dynamic, etc. It is proved that, magnetotelluric time series obtained around industrial park, downtown and agricultural region are easily interfered by human noises. With the purpose of obtaining stable and reliable impedance estimation, detecting and separating human noises become a significant problem which must be solved in the wide use of modern magnetotelluric method. In respect of spatial attribute, most human noises are local for magnetotelluric field. And the electromagnetic signals produced by these noises are linearly polarized in particular direction. Conversely, magnetotelluric field existed in nature is formed by the mutual influence of solar wind, terrestrial magnetism, the ionospheric current system and lightning. On account of the diversity of field source, the magnetotelluric signals obtained from natural field are nonlinearly polarized. So it is the basis to apply the polarization analysis to magnetotelluric data processing.In this thesis, the author firstly write MATLAB programs based on the geophysical wavelet library, thus the polarization analysis algorithm is achieved. Then this algorithm is used in processiong some kinds of common noises in magnetotelluric signals and the characterization of polarization attributes of these noises is studied. The process of this study is as follows:firstly, we put these common noises in magnetotelluric signals into the measured data which is little interfered and do polarization analysis to these data.Then, several measured data segments contained noises are selected for polarization analysis. Finally, characterization of polarization attributes of the typical noises (impulse interference, rectangular interference, etc) in magnetotelluric signals is obtained.After getting the characterization of polarization attributes, we do the time domain localization to the noise according to their characterization of polarization attributes(depending mainly on polarization ellipticity attribute).And then a lot of measured data are used to validate the reliability of the time domain localization. Ultimately, we process the time-frequency spectrum according to the time domain position of these noises and obtain the signal in which the noises are removed. All these above prove that polarization analysis method applying in noise localization and separation of signal and noise of magnetotelluric measured data is feasible.