Correlation matrices as a method of source function isolation and classification in ambient noise

Abstract

At low frequencies (1–100 Hz), the deep-ocean ambient noise field is a mixture of many source types, including seismic activity, marine life, ship traffic, seismic airgun surveys, and nonlinear surface wave interactions. The location of the sensors used to monitor these types of noise typically result in an extremely dynamic recorded ambient noise field. This study proposes the use of correlation matrices as a tool for identifying the predominant noise sources present in samples of ambient noise. These correlation matrices work by identifying frequency ranges in which the noise levels tend to change at the same time, thus identifying frequency ranges where noise is driven by one particular source mechanism. The effectiveness of this strategy will be demonstrated using low-frequency deep-ocean ambient noise recorded in the Pacific, Atlantic, and Indian Oceans by the Comprehensive Nuclear-Test Ban Treaty Organization (CTBTO) hydroacoustic monitoring system.