D-A-CH TAGUNG 2011 - SGEB

uildings representing a real building stock was constructed based on typical buildings for theTaiwan area.The results of the modelling show that the proper consideration of (a) the betweenearthquakecorrelation and (b) the parameters of spatial correlation (correlation function() ) is very important when estimating seismic losses for distributed portfolios. Both typesof ground motion correlation affect the probability of joint damage and the characteristics ofloss distribution; the peculiarities of the influence were analysed and are described in literature.In our work, we showed that the relative influence of each type of correlation is not equal. Forthe considered case of a scenario earthquake, the full-scale variations (i.e., changes withinconsidered limits) in the correlation distance (spatial correlation) lead to larger variations in thecharacteristics of damage and loss than the full-scale variations of between-earthquakecorrelation. The difference is especially pronounced when considering the median values ofloss distribution or the probability that at least one critical element of a lifeline will bedamaged. When the spatial correlation of ground motion is used for such estimations, thepossible variations in the between-earthquake correlation may be neglected.The shape of the site-to-site correlation function, i.e., the rate of decrease of the coefficientof spatial correlation with separation distance, seems to also be important for the lossassessments when modelling spatially correlated ground motion fields. As expected, a high rateof decrease of the coefficient of spatial correlation with separation distance would increase theinfluence of within-earthquake variability and, in turn, would increase the possibility of largelosses at particular locations but decrease the variability of total loss.If information on correlation models is not available, it is necessary either to obtain upperand lower bound estimates by assuming the extreme characteristics of correlations or to makesome assumptions about correlations based on reported models. In this situation, it is necessaryto bear in mind that the spatial correlation of ground motion, in principle, depends on thechosen ground motion model. On the other hand, the correlation structure depends on the localgeology as well on peculiarities of the propagation path (azimuth-dependent attenuation) (e.g.,Sokolov et al. [29]). Thus, a single generalised model of spatial correlation may not beadequate for large areas.We realise, of course, that there are several limitations of our study. Our findings relate to asingle earthquake and to a relatively small area, the dimensions of which are comparable to themaximum value of correlation distance in the spatial correlation model. A comprehensiveanalysis requires considering territories of various sizes and various building densities, as wellas considering the influence from multiple earthquakes. The damage analysis was based onlyon peak ground acceleration, and it included some arbitrary assumptions about the parametersof fragility curves and replacement costs. Obviously, for practical calculations requiring theestimation of the absolute values of loss, these simplified assumptions have to be reconsidered.ACKNOWLEDGEMENTSThe authors would like to thank Julian J. Bommer for his thoughtful review comments andsuggestions, which significantly helped to improve the earliest version of the article. Theconstructive comments from anonymous reviewers are gratefully acknowledged. We wouldalso like to thank Chin-Hsun Yeh for providing necessary data used in this study, valuablecomments and suggestions. This work was sponsored by Deutsche Forschungsgemeinschaft(DFG), Germany, Project WE 1394/14-1.19173