Developing Ground-Motion Prediction Equations Based On The Strong-Motion Records Obtained In The Southwestern Part Of China

A ground-motion prediction equation(GMPE)is an important component of earthquake engineering and plays an important role in seismic hazard analysis and the generation of seismic hazard maps.Usually,the horizontal ground motions have larger amplitudes and are considered to have more influence on the structural safety than the vertical component,and therefore many GMPEs for the horizontal component have been developed.However,in recent years,many studies show that the vertical ground motions,such as the response spectra,can often be greater than those of the horizontal ground motions.The vertical ground motions are not only important for a long-span bridge or a building with a large spacings between columns or shear walls but are also important for many types of structures near the earthquake fault.There are three kinds of methods to develop a GMPE for a given region.The first one is to develop a GMPE from an abundant strong ground-motion records from the target region.The second one is to modify a selected reference model using a small dataset from the target region and the third one is to develop a model based on the ground motions from stochastic ground motion models.The southwestern part of China is one of the most seismically active regions in China where a GMPE is particularly important for this region;and this study aims at building GMPEs for both horizontal and the vertical components for this region.This study assembled a strong-ground motion database from this region first,then various factors affecting ground motion characteristics were explored,and finally,methods of describing these factors in the ground motion model were introduced.The unbalanced distributions of records with respect to earthquake parameters make it difficult to develop a robust GMPE from this dataset only and selection of reference models,one for the horizontal and one for the vertical components,respectively,was carried out first.This study selected five GMPEs developed using a larger number of strong-motion records for the horizontal and five GMPEs for the vertical components of the ground motions and then the following analysis was carried out,(1)Evaluated the overall performances of those GMPEs for the recorded ground motions;(2)Assessed the mismatch in the source,path,and site effects between the reference models and those of the recorded ground motions using correction functions;(3)Calculated the leftover standard deviation of residuals after correction functions were applied.Based on the comparison of the model parameters in the last 3 steps,this study developed an evaluation method for selecting the reference model for the horizontal and vertical components,respectively.Based on the functional form and some model coefficients from the reference model,this study developed two GMPEs,one for the horizontal and one for the vertical components for the southwestern part of China.This study then compared the model coefficients,standard deviations,and predicted ground motions predicted by the GMPEs from this study with those of the reference models.The following specific research results were obtained:(1)Selected horizontal GMPEs include those by Abrahamson and Silva(2008),Abrahamson et al.(2014),Chiou and Youngs(2008),Chiou and Youngs(2014),and Zhao et al.(2016a).The Abrahamson and Silva(2008)and Zhao et al.(2016a)for the shallow crustal and upper mantle earthquakes were found to overestimate the horizontal ground motions whereas the other GMPEs under-predict the ground motions of the dataset in this study at most periods and a correction term terms were applied to the model residual.The corrected residuals were then decomposed into between-event,between-site,and within-site parts,the corresponding standard deviations of the model from Abrahamson and Silva(2008)and Zhao et al.(2016a)are smaller than those of the other three models.The coefficient of the constant correction term and the corresponding standard deviations were used to establish an the first overall goodness-of-fit parameters,and the GMPE by Zhao et al.(2016a)has the highest scores among the five GMPEs for the horizontal component,the GMPE by Abrahamson and Silva(2008)has the second high scores.Coefficients of the correction functions based on magnitude,path,and site effect parameters were used as the second GMPE evaluation index,and the model from Zhao et al.(2016a)has the highest score.The standard deviations of the leftover residuals after all correction functions were applied,the third model evaluation index based on the leftover standard deviations,and the GMPE by Zhao et al.(2016a)has the highest score.Overall,GMPE by Zhao et al.(2016a)has the best scores and this GMPE was selected as the reference horizontal model for the southwestern part of China;(2)For the vertical ground motions,a similar procedure for the horizontal component was used for selecting the reference GMPE for the vertical component and the GMPEs for the vertical ground motions are from Bozorgnia and Campbell(2016),?a?nan et al.(2017),Gülerce et al.(2017),Stewart et al.(2016),and Zhao et al.(2017),respectively.The model from Zhao et al.(2017)model has the highest score for all three model evaluation indexes and this GMPE was selected as the reference model in this study for the vertical ground motions;(3)For the horizontal component,the GMPE developed for the southwestern part of China in this study used a similar model functional form and the same values for some of the model coefficients from Zhao et al.(2016a).The model coefficients were derived by a random effect analysis and the magnitude scaling rates for events with M_W>7.1 were derived by the records from the large overseas events within a distance of 80 km and records from the M_W7.9 Wenchuan earthquake.The between-event standard deviations from this study are smaller than those of the reference model at most periods,the within-event standard deviations from this study are larger than those in the reference model at periods up to 1.2 s.The predicted PGAs of an M_W 6.0 event by the GMPE for the horizontal component from this study are smaller than those from the reference model,especially at sites within 30 km.The predicted spectra by GMPE from this study attenuates slower than those for the reference model at a distance within 30 km,and two models have similar attenuates rates at a longer distance.(4)For the vertical ground motions,a GMPE for the vertical ground motions for the southwestern part of China was developed by using a similar method to that of the GMPE for the horizontal components,and the reference GMPE is the model from Zhao et al.(2017)for the vertical ground motions from the shallow crustal and upper mantle earthquakes in Japan.The between-event standard deviations from the GMPE for the vertical ground motions of this study are smaller than those of the reference model at spectral periods up to 0.2 s;the within-event standard deviations from the GMPE developed by this study are larger than those in the reference model at almost all periods.For the vertical ground motions from an M_W 6.0 or 7.0event at a distance of 20 km,the predicted spectra by the GMPE of this study are smaller than those from the reference model at periods up to 0.2 s,and the predicted spectra from two models are similar at moderate and long periods.