An automatic preselection strategy for magnetotelluric single-site data processing based on linearity and polarization direction

Abstract

The magnetotelluric response function can be severely disturbed by cultural electromagnetic noise. The preselection strategy is one of the effective ways to remove the influence of noise when calculating the response function. This study proposed three new parameters (the amplitude ratio predicted amplitude ratio and linear coherence (PLcoh) between the predicted and observed electric fields and the dispersion degree of the magnetic polarization direction (DDpol)) to detect noisy data, making the preselection strategy automatic. The first two were used to evaluate the linearity of binary linear regression to constrain incoherent noise, while the last was used to evaluate the magnetic polarization direction to constrain coherent noise. Finally, the technique is illustrated by applying it to two field datasets and comparing it with the previous studies. The results showed that these parameters can be used to effectively identify contaminated data, and a reliable response function can be obtained by using these parameters to extract high-quality data when intermittent noise contaminates field data.