Seismic Risk Assessment of Multiple Cities: Biases in the Vulnerability Derivation Methods for Urban Areas With Different Hazard Levels

Previously, it has been shown that probabilistic seismic risk assessments (PSRAs) at urban scale present important discrepancies when compared with analyses conducted using methodologies from regional or national PSRA. However, conducting site-specific urban-scale analyses for a considerable number of cities may not be feasible due to limitations in time, resources, and in some cases availability of information, and thus more general models or methodologies are used. This brings into the picture the importance of identifying and quantifying the possible biases, discrepancies, and uncertainties when using different methods, both in the hazard and vulnerability components. Regarding the latter, several sources of uncertainty and biases have been identified in 1) the selection of ground motion records, either by using a general pool of records such as the ones from FEMA P695 or by performing a site- or hazard-specific analysis that requires a significant effort, especially in areas with a poor history of seismic instrumentation and even more in regions with no evidence of previous PSRA at all; and 2) the fragility or vulnerability derivation nonlinear dynamic methods: incremental dynamic analysis [IDA], cloud analysis [CA], and multi-stripe Analysis [MSA], among others. Focusing on these sources of uncertainty and bias, and with the challenge to bring solutions for places with scarce information, in this study, we aim to explore the use of different vulnerability derivation assumptions for the three principal cities of Colombia: Bogota, Medellin, and Cali, where most of the economic growth is concentrated. This considers the different seismic hazard levels and tectonic environment contributions in each city. Afterward, a comparison between the results of the analysis without the hazard-specific record selection and the site-specific one for each city is performed to establish the cases in which the former is applicable without adding more biases or uncertainties in the process.