The interevent and interstation ground-motion variability of the updated Italian strong-motion database (Italian Accelerometric Archive [ITACA]) has been explored through the development of new empirical ground-motion prediction equations (GMPEs) for Italy. The regressions have been performed on 241 three-component waveforms from 27 earthquakes with moment magnitudes ranging from 4.8 to 6.9, recorded by 146 stations at distances up to 200 km. The site classification follows the schemes previously proposed for Italy, in which two soil classes are defined, considering both shear-wave velocity and deposit thickness. The regression analysis uses the values of the explanatory variables (magnitude, fault distance, site class, and style of faulting) recently revised in the framework of a project funded by the Italian Department of Civil Protection. The equations have been derived for peak ground acceleration, peak ground velocity, and 5% damped spectral accelerations at 18 periods from 0.03 to 2 sec.
The residual variance has been decomposed into interevent, interstation, and record-to-record components by applying a random effect regression scheme. The interevent and interstation error distributions have been analyzed as function of periods to detect sites and events for which predicted values significantly deviate from observations. For periods up to 0.35 sec, the interstation is the dominant component of variance, indicating that an improvement in the site classification could lead to a refinement of the GMPEs. For longer periods, the three components of variance provide similar contributions, indicating that a reduction of the uncertainty can be achieved by reducing the epistemic uncertainty affecting the physical model. The interevent error highlights the peculiarity of few earthquakes, suggesting that the evaluation of regional GMPEs can be important when specific scenario studies should be carried out. The interstation variability allows us to detect stations with peculiar site response and to assess the goodness of the considered site classification scheme.