One basic element for seismic hazard assessment is the empirical definition of ground motion prediction equations (GMPE) to estimate shaking expected for earthquakes of given magnitude and distance. GMPEs are calibrated from data of accelerometric stations, distinguishing among site categories of different lithological type (e.g. hard rocks, more or less stiff soils) expected to cause different levels of ground motion amplification. Such a site classification is commonly based on geological observations and/or geophysical parameters like the mean propagation velocity of seismic waves through subsoil surficial layers. However, doubts have been raised about the effectiveness of results obtained from these conventional methods. Here we propose a methodology of accelerometric site classification relying on peak ground motion observations, exploiting the large amount of such observations available in the Italian National accelerometric database. The method is based on a cluster analysis of differences between observations and predictions provided by GMPEs whose functional form does not comprise site class among the explanatory variables. The new method was applied to the ITalian ACcelerometric Archive (ITACA), extracting a "training" dataset (used to calibrate some GMPEs through regressions) and a "validation" dataset (to select the optimal GMPE form). A cluster analysis was then applied to regression residuals, grouping stations into three categories with increasing value of residual average. Checking the reclassification effectiveness through the examination of differences between independent "validation" observations and predictions of GMPEs adopting the new classification, these proved to be more consistent with site response properties than predictions provided by GMPEs using current classification.
|Titolo:||Site classification of Italian accelerometric stations from cluster analysis of residuals of peak ground motion data regressions|
|Data di pubblicazione:||2019|
|Appare nelle tipologie:||1.1 Articolo in rivista|