The EM Noise Signals In Highly Contaminated Areas And Study Of New Methods To Process MT Data

To estimate more precise magnetotelluric(MT)impedance tensor in high correlated noise area is much more difficult.The magnetotelluric data processing methods consist of preprocessing the time series,estimating the spectrum,calculating the power spectrum,approximate MT impedance tensor and computing the apparent resistivity and phase.The dissertation is focused on the new methods of detecting and removing the noise before the traditional MT data processing methods were studied and noise was analysised.In the time domain,the independent component analysis method was used to process the MT time series.After calculating the independent components,the components of noise are removed.The new time series are constructed using the left components.The Robust method can be used to process the new time series.The generalized S transform was studied to improve the spectrum quality and I have used polarization ellipse to detect the cultural noise based on the continuous wavelet spectrum.The MT time series are preprocessed by the generalized S transform.The instantaneous stectra is filtered using the data truncation method,the relative covariance method and the median method.The impedance tensor was estimated using Berdichevsky formula with instantaneous stectra.The synthetic data was made to test the new methods.The natural MT is expected to be non-polarized due to its wide variety of sources and most cultural noise is linearly polarized in a certain direction.With the characteristic,the polarization ellipse is used to detect the cultural noise in the MT time series.The signal-noise separation method is used to improve MT impedance tensor in time-frequency domain.Using the remote site which is lack of correlated noise and the local site,the separation tensor is obtained so that the observed data in the local site can be separated into MT signal and noise.The MT impedance tensor and the noise 'impedance tensor' can be estimated simultaneously.I use the signal-noise separation method to process the synthetic data and observed MT data.The results showed that the apparent resistivity and phase curve was much more smooth than that obtained from the Robust method and remote reference method.The signal-noise separation method has advantage to process the observed MT data which is highly contaminated by the correlated noise.In order to process the MT data,we should observe the time series in advance and distinguish the types of noise.The types of noise can be detected by using polarization ellipse so that the location of noise can be found.The filters are designed to filter the MT data with different methods.