Low frequency ambient noise variation of underwater acoustic data from MACHO hydrophone

Abstract

The 2011 Miyagi earthquake and tsunami in Japan happened right before the Marine Cable Hosted Observatory (MACHO) system was installed in Taiwan's Northeast shore in September, 2011, which reassure the value of MACHO system in observing ocean and earthquake activities in the region. The National Science Council of the Republic of China has also sponsored the project “Marine Observatory in the Northeastern Taiwan (MONET)” to efficiently analyze the large quantity of long-time monitoring raw data from the MACHO system. This paper focuses on establishing an efficient detector to search for special events such as earthquakes, tsunami, cetacean behaviors, and the variations in the ocean environment, which are from the data of hydrophone and conductivity, temperature, depth (CTD) sensors, to provide the immediate and supplementary information for the Central Weather Bureau. The results of the long-time data are massive (about 1T in a week) and valuable, and we need quick and stable procedures to analyze and store it. In general, people who deal with data from the passive sonars have always use experienced human operators to detect special underwater signals, however, operators need to focus on listening to suspicious sounds and observe the spectrogram for long periods of time. The costs of the experiments and instruments are expensive and the detection performances depend largely on operator experience, and their physical and mental condition. Many researchers have been working on how to increase the accuracy rate to back-up the important events more quickly. The detector in this paper includes time-varying ambient noise level estimation via Leq (equivalent continuous sound level, averaged over 30 seconds), and an energy detector with the estimated ambient noise level as threshold. A user friendly UI (user interface) is written to enhance the utilization. Two years of data are analyzed with the detector, and the results are showed the seasonal variation of the local ambient noise level in MACHO system, activities of local environment and the efficiency of analysis of large amount of data.