Detecting Slow Slip Signals in Southwest Japan Based on Machine Learning Trained by Real GNSS Time Series

With the increase in Global Navigation Satellite System (GNSS) observations, the requirement for objective and automated detection of slow slip event (SSE) signals hidden in displacement time series is increasing. However, machine learning for GNSS time series has rarely been attempted. Especially, the physical meanings of the spatio-temporal noise variations and their effects on the detection performance have been not so deeply discussed. In this study, we conducted a single-site SSE detection based on machine learning trained by real GNSS observations of southwest Japan to directly consider the complicated spatiotemporal characteristics of observational noise. Based on a catalog of 284 short-term SSEs, approximately 26,000 time series containing SSE signals or noises were extracted as training data. The signal data predominantly had an amplitude of 1.5-2.0 mm. The model architecture following the Generalized Phase Detection, which was originally proposed for seismic wave detection, was then adopted. We obtained an accuracy of 75% for the test data. As expected, the detectability were mainly controlled by the signal amplitude, and false positive appears to be caused primarily by the temporally correlated noise that resemble the onset or termination of the SSE signal. We examined the correlation between detection performance and noise properties at each site, such as standard deviation and slope of power spectrum. The analysis of this study is expected to facilitate a straightforward evaluation of the influence of noise characteristics on the detection performance, and clarify the crucial topics to improve detection precision.