The variability (sigma) of predicted strong ground motion affects the estimated seismic hazard and the subsequent risk assessment. Hence, assessment of the variability to enable its reduction is important. In this context, residuals computed between observed ground motion levels and those predicted using a ground-motion prediction equation (GMPE) allow assessing the variability due to repeatable source, path, and site effects. Quantifying the residuals by removing ergodic assumption reduces sigma and helps in estimating the seismic hazard more accurately, particularly at low probabilities of exceedance. In our study, we investigate sigma and its components (source, path, and site terms), using a large strong-motion dataset from Japan (9,844 KiK network recordings from 136 shallow earthquakes recorded at 174 sites within 300 km hypocentral distances). We also evaluate the between-event and within-event variability. As reference, we choose a published GMPE developed for shallow crustal earthquakes in Japan (Zhao et al., 2016). We identify event clusters, defined based on focal mechanism similarity, to estimate the location and path terms. Additionally, we apply the region-less approach to define a path term for everyevent-station path; these path terms are then smoothened using a closeness index (Lin et al., 2011). Further, we investigate the azimuthal dependence of ground motion variability due to source effects for strike-slip earthquakes. We find that the overall ground motion variability is reduced significantly, to an extent comparable to previous studies on ground motion datasets. Based on our findings, we propose that our corrections and variance components help to develop a site-specific GMPE.
|Original language||English (US)|
|Journal||Bull. Seis. Soc. Am.|
|State||Published - 2020|