Precise Monitoring of Non-volcanic Low-frequency Tremors using Vertical Seismic Array: The case of Tokai Area, Southwest Japan

Non-volcanic low frequency tremor (LFTs) and short-term slow slip events have been found in various subduction zones and the strike-slip San Andreas Fault during the last decade. Previous studies suggest that these slow events occur within the quasi-stable frictional regime downdip of the shallower seismogenic-locked zone. Therefore, a detailed real-time monitoring of these phenomena is one of the useful ways for forecasting the next great earthquakes. In 2007, Geological Survey of Japan, AIST has started an integrated borehole observation in southwest Japan for forecasting the anticipated Tokai, Tonankai and Nankai megathrust earthquakes. Each observatory has three boreholes with different depths (about 30 m, 200 m and 600 m), in which we installed high-sensitivity seismometers at the bottom of every borehole. On the basis of a semblance analysis using this vertical seismic array data, we developed a monitoring system of the LFTs and showed a dramatic improvement of the LFTs detection (Takeda et al., 2009). This study focuses on the analysis of LFTs in Tokai area using vertical seismic array data. The major LFTs episode in the area last for days to week and occur repeatedly every six months. We have one observatory above the middle of the Tokai LFTs zone, where we started the vertical array observation in June 2008. We calculated a semblance for a range of values in apparent velocity space using one-minute long moving windows. It is easy to discriminate seismic signals associated with the LFTs from cultural noise by both the sign and value of the best apparent velocity for that particular window. We calculated total duration of the LFTs activity for each hour by counting the time that the semblance value of the best apparent velocity exceeds a specific threshold. The two year’s time-duration plot suggests that our semblance method detected about ten times in duration than that by the envelope correlation method (ECM). During each major episode, the time-duration plot by ECM shows that there are many gaps in LFTs activity. However, our semblance method revealed that the LFTs signals were almost continuously observed for every hour and the activity gaps seen in the ECM result are artifacts due to the insufficient detection capability. We also found that between major episodes the semblance method detected a lot of short duration episodes continuing from a few minutes to hours, which are unclear in the case of ECM result. By using an empirical relationship between the seismic moment and duration by Ide et al. (2007), we infer that the moment magnitudes of these episodes range approximately from 4 to 5. These short duration episodes would also accumulate stress on the shallower locked portion and influence the occurrence of earthquakes.