High spatial-resolution loss estimation using dense array strong-motion near-fault records. Case study for Hveragerði and the Mw 6.3 Ölfus earthquake, South Iceland

Abstract

The most recent and costliest damaging earthquake in Iceland is the 𝑀w6.3 29-May-2008 Ölfus earthquake to date. In particular, Hveragerði town located in the extreme near-fault region, suffered intense horizontal peak ground accelerations (PGA) of ∼40–90%g and large amplitude and long-period near-fault pulses, recorded on a dense urban strong-motion array in the town. In this study we collated a high-spatial resolution exposure database (building-by-building) complete with actual reported losses and classified the buildings by building materials and construction year according to the code design requirements in place at the time. We took advantage of the array data and evaluated a set of well-known ground motion intensity measures (IM), including PGA, pseudo-acceleration response spectra at short-to-long periods, Arias Intensity and Cumulative Absolute Velocity. We applied empirical Bayesian kriging geostatistical analyses to generate high-resolution shakemaps and provide IM estimates for each building. The shakemaps showed significant and systematic variation of the IMs across the small study area, with the lowest ground motions observed centrally and highest values in the outskirts. Furthermore, correlation analysis was carried out for the damage ratio and the exposure data IMs, but only low-to-moderate correlations were observed. A key reason is the incurred losses were primarily due to damage to non-structural components, to which the code design requirements do not apply. We carried out a seismic loss assessment in Hveragerði for the earthquake scenario of the Ölfus earthquake both on building-by-building, and municipality levels of spatial resolution. We applied both local and global fragility models for associated with detail building typologies identified based on the SERA taxonomy scheme. The results show that the global fragility functions severely underestimate the seismic performance of the building stock, except for one-story reinforced concrete buildings, while overall the masonry buildings were associated with the most predicted and observed losses. On the other hand, the local models predicted losses that conformed well with the observed damages to timber and concrete buildings. The high-spatial resolution predictions of losses gave results that better correlated with the observed losses in most typologies.