Identifying Anomalies in Seismic Velocity and Scattering Property Changes at Active Volcanoes Based on Seismic Interferometry and the Local Outlier Probability Method

The identification of anomalies in seismic wave interferometry data is important in the prediction of imminent volcanic eruptions. Herein, we propose using the local outlier probability (LoOP) method to evaluate the degree of anomaly in seismic wave velocities and scattering properties, estimated via seismic wave interferometry. LoOP is the likelihood that an observation is anomalous and is always in the range of 0-1 (0-100%). We quantitatively evaluated the degree of anomaly in seismic wave velocities and scattering properties before and after the eruption of Mt. Aso, Japan, in October 2016 and Mt. Shinmoedake, which lies within the Mt. Kirishima cluster of volcanoes, Japan, in 2017 and 2018. We found that LoOP exceeded 70% 2 to 3 days before Mt. Aso erupted on October 8, 2016, and it exceeded 70% 1 to 5 days before Mt. Shinmoedake erupted on October 11, 2017 and March 6, 2018. Adjusting the reference and quiet periods for the estimation of changes in seismic velocity/scattering property and LoOP calculations can allow the tracking of repeated, significant LoOP increases during times of high volcanic activity. The quantitative evaluation of temporal anomalies in seismic datasets will improve the precision of predictions of imminent volcanic eruptions.