Using unsupervised machine learning for clustering seismic noise

Abstract

Ground motion induced by non-tectonic sources may hinder recognition of tectonic signals related to microearthquakes or nonvolcanic tremors. Lack of knowledge on seismic noise sources can also negatively affect model training when applying machine learning for weak signal detection. Thus, it is necessary to conduct investigations on seismic noise anatomy, understanding its features and potential sources. In this study, we apply K-means analysis to continuous recordings of a dense seismic array in north China to cluster various types of seismic noise. The analysis indicates that noise field recorded by the array comprises 6 types of signals related to various non-tectonic sources, ranging from road and rail traffic, wind, and nearby power lines.