Statistical Modelling of Seismmic Vulnerability of Buildings for South Iceland considering the Spatial Correlation of Ground Motion Intenisty

Abstract

An Mw6.30 earthquake occurred in south Iceland in May 2008. The epicentre and fault rupture occurred close to small villages and farms, affecting over 5000 residential buildings. Despite significant damage, no residential buildings collapsed. It is desirable to know the ground motion intensity at various locations in order to develop an empirical vulnerability model; however, ground-motion observations are only available for a limited range of sites. Groun motion Prediction Equations (GMPEs) are commonly used to predict desired ground motion intensity measures (IM) at a given site. There are several interpolation methods available to improve the predictions if local ground motion data for the study event is available. Since IMs or their logarithms are normally distributed, spatially correlated, and correlated with each other at a given location, the conditional multivariate normal (MVN) distribution can be used for this purpose. This paper uses the MVN-based approach to perform PGA interpolation using local GMPE. We specifically present: 1- spatially correlated PGAs using MVN formulation and 2- an advance empirical vulnerability model based on zeroinflated beta regression calibrated for five building typologies in south Iceland.