Earth observation techniques for seismic building vulnerability assessment

Abstract

The impact of natural disasters such as earthquakes on mankind has increased dramatically over the last decades. Global urbanization processes and increasing spatial concentration of exposed elements such as people, buildings, infrastructure, and economic values in earthquake prone regions induce seismic risk at a uniquely high level. The increase of urban population has occurred during a time period that is comparatively short with respect to the return time of severe earthquakes. Earthquakes that had little impact in the past, when they hit sparsely populated and spatially fragmented settlement areas (i.e., small towns and villages), will nowadays shake urban agglomerations with a considerable of people. This situation, when left unmitigated, is expected to cause unprecedented death tolls, enormous economic and ecological losses, critical infrastructure and service failures, and poses a significant threat for civil security, and a sustainable development in the future. To mitigate those perils requires detailed knowledge about seismic risks. As an important constituent element of seismic risk, the seismic vulnerability of the built environment has to be assessed. In particular, it is crucial to have information about the building inventory and its behavior with respect to a certain level of ground shaking. In this manuscript, we how to derive sets of valuable features from remote sensing data and combine them with in situ collected seismic vulnerability information of buildings. Statistical learning methods are deployed to estimate the seismic vulnerability level of the building inventory spatially complete and consistent. Experimental analyses are carried out for the city of Padang (Indonesia). Accuracy measures show promising empirical evidence (mean absolute percentage error of 10.6% and a linear correlation (R) of 0.59) and underline the potential of the approach for an area-wide seismic vulnerability assessment of buildings based on remote sensing.