Effective denoising of magnetotelluric (MT) data using a combined wavelet method

Noise interference, especially from human noise, seriously affects the quality of magnetotelluric (MT) data. Strong human noise distorts the apparent resistivity curve, known as the near-source effect, causing poor reliability of MT data inversion. Based on analyzing the frequency characteristics of human noise resulting from the surrounding environment, a new waveletbased denoising method is proposed for both synthetic and real MT data in this paper. The new technique combines multiresolution analysis with a wavelet threshold algorithm based on Bayes estimation and has a remarkable effect on denoising at all band frequencies. The multi-resolution analysis method was employed to reduce long-period noise, and a wavelet threshold algorithm was used to eliminate strong high-frequency noise. In this research, the improved algorithm was assessed via simulated experiments and field measurements with regard to the reduction in human noises. This study demonstrates that the new denoising technique can increase the signal-to-noise ratio by at least 112% and provides an extensive analysis method for mineral resource exploration.

---

Opracowanie rekordu ze środków MNiSW, umowa Nr 461252 w ramach programu "Społeczna odpowiedzialność nauki" - moduł: Popularyzacja nauki i promocja sportu (2020).