Ground-motion Prediction Equations For Shallow Crustal And Upper Mantle Earthquakes In Subduction Zone

Establishing a set of ground-motion prediction equations(GMPEs)for Japan or other subduction zone requires earthquake source categories in the dataset.Earthqukes in subduction zone are typically divided into three groups: shllow crustal events that occur in the Earth's crust,subduction interface events that occur at the interface between the crust or mantle and the subducting plate,and the subduction slab events that occur in the subducting plate.In the present study,the hypocentral locations published in the catalogues of the International Seismological Centre(ISC-EHB),the Japan Meteorological Agency(JMA)and the National Earthquke Information Centre(NEIC)were assessed and compared.The hypocentral locations for the same earthquakes vary significantly from one catalogue to another.Four classification schemes using locations from these three catalogues were designed.A random effects model was to the strong-motion dataset from these earthquakes to assess the merits of the classification schemes.The results showed that using ISC-EHB locations for events before 2005,and then using the preference order of catalogues as: 1)JMA locations with high precision levels,2)ISC-EHB,and 3)NEIC(excluding the events with a fixed depth)for events since 2005,together with some conditions for subduction interface events,produced the best GMPEs in terms of the maximum log-likelihood.It was also found that having a separate group for the earthquakes above the subduction interface,but with a depth over 25 km,improved the goodness-of-fit of the GMPEs.GMPEs for shallow crusal and upper mantle(SC&UM)derived from strong-motion records in Japan are presented.A large dataset from earthquakes with a moment magnitude(MW)over 4.9 and a reliable earthquake category,up to the end of 2012 was assembled.The GMPEs were based on a set of simple geometric attenuation functions.The models included nonlinear site terms,most of which did mot vary from one model to another.The site terms were site classes based on site period.The model has a bi-linear magnitude function hinged at MW = 7.1.The magnitude coefficient for large events is much smaller than that for the events with MW < 7.1.The effect of volcanoes on the attenuation of seismic waves was modelled.A volcanic path defined as the horizontal portion of a straight line distance(the closest distance between the fault plane and the recording station)that passes through the assumed low-Q zones around active volcanoes.An anelastic attenuation rate that applies to the volcanic path was derived.Most strong-motion records in the dataset used in the present study are from stations with a measured shear-wave profile down to bedrock.A small number of strongmotion records are from strong-motion stations with inferred site classes using H/V response spectral ratio or geological description of the surface soil layers.The effect of site information quality was tested,by examining the results from a dataset containing the strong-motion records from sites with an inferred site class and the results from a dataset without these records.It was found that the overall goodness-of-fit of the GMPEs improved significantly after excluding the records.The within-event residuals were separated into within-site and between-site approximately and the corresponding standard deviations were calculated using a random effects model with all model coefficients presented in this manuscript.The separation of within-event residuals into within-site and between-site components allows for a possibility of adopting different standard deviations for different types of earthquakes and for different site classes in a probabilistic seismic hazard analysis if desired.It was found that the models had a magnitude-distance oversaturation,i.e.,the response spectrum at source(zero distance)decreased with increasing magnitude.This effect cannot be justified physically yet,and so to eliminate it we adopted a distance constant term.The effect of site information quality was tested.GMPEs derived from the strong-motion records in Japan by the Zhao et al.(2006)(referred hereafter Zhao 2006)study,Zhao et al.(2016a,2016b)studies and the present study(referred hereafter Zhao 2016 studies)were also compared.The Zhao 2006 model used strongmotion records up to the end of 2003 and the site class was inferred for most recording statios.The Zhao 2016 models used records between 1996 and the end of 2012 and all recording stations have measured shear-wave velocity profile down to engineering bedrock.The data quality and site conditions in the Zhao 2016 studies are much better than those in the Zhao 2006 study.The Zhao 2016 study developed three separate models,one for shallow crustal and upper mantle earthquakes,one for subduction interface events,and one for subduction slab events.We carried out statistical tests in this manuscript to justify the use of three separate models because the between-event and within-event standard deviations among the three models differ statistically between each other at many spectral periods.Statistical tests were used to justify the use of separate site terms for each model because the site terms among the three models differ statistically at many spectral periods.The predicted spectra from the Zhao 2006 model and the Zhao 2016 model were compared.The Zhao 2006 spectra in a number of magnitude and distance ranges are comparable with those from the Zhao 2016 model but the differences are large in some magnitude and distance ranges.The depth boundary of 25 km in the shallow crustal and upper mantle model and the depth boundary of 25 km differentiating the shallow and deep interface model lead to abrupt change in the predicted spectra at the depth boundary.The change can be considerable at some magnitude and distance ranges and the abrupt change can be averaged out using the depth scaling function recommended by Zhao et al.(2016a).