Discrepancy between observation and simulation of ground motions for the 2016 Gyeongju earthquake, Korea

Most of ground-motion prediction equations (GMPEs) of the Korean Peninsula have been developed by using the stochastic simulation method due to not enough strong ground-motion data. It has been controversial whether these GMPEs are proper models for predicting a large earthquake which might occur in near future because the GMPEs have not been verified with observed data yet. ML 5.8 Gyeongju earthquake, the great earthquake ever occurred in the Korean Peninsula after modern instrumental earthquake monitoring was initiated in 1978, occurred on September 12, 2016 provides an opportunity for verifying the GMPEs of Korea. We compared the observed ground-motions recorded on 201 seismic stations operated by Korea Meteorological Administration and Korea Institute of Geoscience and Mineral Resources to the predictions estimated from the GMPEs and the stochastic simulations. Most of simulated-PGAs or predictions of GMPEs appeared to underestimate the observed ground motions about 8 times in average. These discrepancies seem to be caused by the assumption that all stations located in rock-site in GMPE development or ground-motion simulation. We applied the FIR filter method to evaluate proper site response functions and corrected the discrepancies between the observations and the predictions.