Abstract

Info

Title:: Magnetotelluric Impedance Recognition Based on K-medoids Clustering Analysis

Author(s):: ZHAO Xuan; YAN Jia-bin; HUANG Xiang-yu; HU Tao; 1. School of Geosciences and Info-physics, Central South University, Changsha 410083, Hunan, China; 2. Key Laboratory of Non-ferrous Resources and Geological Hazard Detection of Hunan Province, Central South University, Changsha 410083, Hunan, China

Keywords:: impedance estimation; real and imaginary components; clustering analysis; K-medoids; euclidean distance for impedance; coherence; compactedness; numerical simulation

Abstract:: In the complicated observation environment, the magnetotelluric signal is disturbed by different types or degrees of noise. Traditional estimation and extraction methods are hard to obtain stable impedance, such as the power-spectrum method and Robust method. Based on the physical properties of the real and imaginary components of magnetotelluric impedance, the impedance of signal disturbed slightly by noise distributes intensively, and the impedance of signal disturbed severely by noise distributes dispersedly. So, the impedance extraction and identification based on K-medoids clustering analysis was introduced, and then the euclidean distance for impedance was proposed to describe the similarity of impedance. The K-medoids clustering analysis can divide the impedance of signal disturbed slightly by noise into the same cluster and other impedance of signal disturbed severely by noise into different clusters based on the similarity. According to the selection criteria of cluster including coherence and compactedness criteria, the optimal cluster of impedance in the disturbed environment was chosen. Based on the simulation experiment and instance analysis, the K-medoids clustering analysis can recognize high quality signal in noisy environment and recovery the real impedance.