Noise suppression for magnetotelluric using variational mode decomposition and detrended fluctuation analysis

Abstract

Noise suppression is significant for magnetotelluric (MT) data analysis, especially in the face of strong human electromagnetic interference. Variational mode decomposition (VMD) is a novel signal processing method, and the selection of proper decomposition layers plays a key role in its signal processing effects. In view of the fact that detrended fluctuation analysis (DFA) is an evaluation criterion capable of selecting the appropriate number of sub-signals, we propose a noise suppression method for MT by combing VMD with DFA. We firstly estimate the scaling exponent range of MT noisy data and decompose it into a given series of K-order intrinsic mode functions (IMFs). Then, a criterion based on DFA is designed to select the mode number K with the purpose to improve the decomposition accuracy of VMD. The scaling exponent of DFA is used to define the relevant modes for the construction of filtered signals. In this research, the MT data is collected from Qinghai test site and added with strong interference for comprehensive analysis by the remote reference method, VMD method, and the proposed VMD-DFA method. The results have shown that the proposed method could greatly improve the accuracy of signal decomposition and the reliability of the reconstructed signal by adaptively selecting suitable K-order IMFs for different types of interference. Consequently, strong interference is effectively eliminated, and the obtained apparent resistivity-phase curve is stable and smooth. Moreover, the denoised MT data will further strengthen the interpretability of geoelectric information.