We describe an Italian database of strong ground motion recordings and databanks delineating conditions at the instrument sites and characteristics of the seismic sources. The strong motion database consists of 236 corrected recordings from 86 earthquakes and 101 recording stations. Uncorrected recordings were drawn from public web sites and were processed on a record-by-record basis using a procedure utilized in the Next-Generation Attenuation (NGA) project to remove instrument resonances, minimize noise effects through low- and high-pass filtering, and baseline correction. The number of available uncorrected recordings was reduced by 52% (mostly because of s-triggers) to arrive at the 236 recordings in the database. The site databank includes for every recording site the surface geology, a measurement or estimate of average shear wave velocity in the upper 30 m (Vs30), and information on instrument housing. Of the 86 sites, 39 have on-site velocity measurements (17 of which were performed as part of this study using SASW techniques). For remaining sites, we estimate Vs30 based on measurements on similar geologic conditions where available. Where no local velocity measurements are available, correlations with surface geology are used. Source parameters are drawn from databases maintained (and recently updated) by Istituto Nazionale di Geofisica e Vulcanologia and include hypocenter location and magnitude for small events (M < 5.5) and finite source parameters for larger events.

Ground motion prediction equations (GMPEs) have recently been developed in NGA project that are intended for application to shallow crustal earthquakes in tectonically active regions. We investigate the compatibility of those models with respect to magnitude-, distance-, and site-scaling implied by Italian strong motion data. This is of interest because (1) the Italian data is principally from earthquakes in extensional regions that are poorly represented in the NGA dataset and (2) past practice in Italy has been to use local GMPEs based on limited data sets which cannot resolve many source, path, and site effects known to be significant. We find that the magnitude scaling implied by the Italian data is compatible with the NGA relations. However, the Italian data attenuate faster than implied by the NGA GMPEs at short periods, and the differences are statistically significant for three of four relations. Three regression coefficients are re-evaluated for the three affected NGA GMPEs to reflect the faster attenuation; a constant term, a geometric spreading and anelastic attenuation term, and a source depth term. The scaling of ground motion with respect to site shear wave velocity is consistent between the NGA models and Italian data. Moreover, the presence of nonlinearity in the Italian data is confirmed and found to be generally compatible with what is provided by NGA site terms. The scatter of Italian data is much higher than in the NGA models, although only the intra-event error is sufficiently well established by the data to justify modification of NGA models. On the basis of these findings, we recommend that NGA relations, with the aforementioned minor modifications, be used to evaluate median ground motions for seismic hazard analysis in Italy.