Development of Strong-Motion Database for the Sumatra-Java Region

Abstract

"This study focuses on the development of an Indonesian strong ground motion database. This research will be an important contribution to Indonesian earthquake hazard assessment. This work will be perform in three (3) stages: (1) construction of earthquake-related parameters catalogue; (2) compilation of site condition information using various techniques; and (3) processing recorded ground-motions using well-known procedures to produce ground motion parameters commonly used in seismology and engineering applications. The construction of the database in this study produced 3090 records from about 249 earthquakes. It found among 249 earthquakes in this study 63 earthquakes categorize as interface/megathrust, 161 earthquakes as intraslab, 4 earthquakes as crustal, and 21 earthquakes are of unknown type. Then, the database will be used for investigating which published Ground Motion Prediction Equations (GMPE) are appropriate to use in Indonesian earthquake hazard assessment. The relevant GMPEs compared in this study are based on the resemblance of geologic and tectonic conditions of the regions where the GMPEs were developed to the study area. Twelve GMPEs have been considered in this study, consisting of nine GMPEs derived for subduction-zone event types (intraslab and interface regimes) and three GMPEs derived for the crustal regime. The analysis of GMPEs in this study was done using the graphical analysis of residuals between the observed ground motion value and the corresponding values predicted by each GMPE. The visual analysis of the statistical graphs presented in this study indicates four GMPEs (Youngs (1997), Zhao (2006), Kanno (2006) and Lin-Lee (2008) match with the recorded data reasonably well, while the others have poor fit with the data. In this study, we also rank the GMPEs using the quantitative method proposed by Scherbaum et.al (2004). The Scherbaum e.al (2004) scheme shows that comparison of PGA/PSA with threshold value 0.0005 m/s2 gives a better output than using all data. In this study, we found that among all models, only the Youngs (1997) and Zhao (2006) provide predictions consistent with the data and are classified as C class (the lowest capability class)."