Denoising of magnetotelluric data by polarization analysis in the discrete wavelet transform domain

The development of denoising techniques of magnetotelluric (MT) data affected by cultural noise is currently one of the most important objective to make magnetotellurics reliably in urban or industrialized areas. In this work, a new denoising technique of MT data affected by temporally localized noise is proposed. It is based on the polarization analysis of the MT field in the time-frequency domain achieved through a discrete wavelet transform. This transform, thanks to the possibility to operate in both time and frequency domains, allows the automatic detection of transient components within the MT signal possibly due to disturbances of anthropic nature. Unlike the continuous wavelet transform, it permits to reconstruct the denoised signal in the time domain in order to test the effectiveness of the filter. Applications to both synthetic and field MT data have shown the ability of the implemented filter to detect and remove effectively the cultural noise.