SOM-based Visualization for Classifying Large-scale Sensing Data of Moonquakes

Large-scale seismic data were obtained from seismometers emplaced on the Moon by the NASA Apollo missions conducted during 1969-1977. Previous analyses of lunar seismic data revealed that deep moonquakes occur periodically from identical sources at depths of about 700-1200 km. The deep moonquakes occurring from identical sources have high similarities among waveforms. This similarity is important to classify their sources and to investigate the moonquake generation mechanism. Therefore, classification of moonquakes has been undertaken. For that purpose, we developed a web system for visualizing moonquakes considering their waveform similarity to advance the study of moonquake classification. Our system maps moonquake data to two-dimensional output space using a Self-Organizing Map (SOM). We embed Hadoop in the back-end system to apply SOM to large amounts of moonquake data. In this paper, to select features for SOM, we evaluate several features based on classified data. Using selected features, we apply SOM to moonquake data and discuss the results.