Towards an update of GMICEs for the Italian territory over a set of ground motion parameters

The rapid determination of major damage location after an earthquake can allow for timely arrangement of first assistance operations by the Civil Protection forces. Traditional macroseismic surveys take time for organization and completion; also, in case of destructive events the most damaged areas might not be immediately accessible for safety reasons. Given the possibility to directly associate the instrumental Ground Motion Parameters (GMPs), which can be estimated in near real-time, to the damages distribution after the occurrence of an earthquake, the most common solution is to use instrumental Macroseismic Intensity (MI) maps for rapid implementation of emergency plans. These maps display instrumental intensity values calculated through Ground Motion to Intensity Conversion Equations (GMICEs) over the struck region. The GMICEs currently in use for Italy were derived by Faenza and Michelini (2010) from a dataset consisting of 87 events (266 records) in the time-span 1972-2004. In this study, we extend the approach proposed by Faenza and Michelini (2010) by carrying out the calculation of GMICEs on eight ground motion parameters:Peak Ground Acceleration, Velocity and Displacement (PGA, PGV, PGD), Arias and Housner Intensities, Pseudo-Spectral Acceleration at 0.3, 1.0 and 3.0 seconds. We first expand a re-elaborated version of the Faenza and Michelini (2010) database with high quality accelerometric data regarding 25 events occurred in the time-span 2002-2016 in Italy and Slovenia (Tiberi et al., 2018). We then derive two sets of empirical conversion equations between each of the selected parameters and MI: the first one using the Linear Least Squares technique and the second one using Orthogonal Distance Regression. Finally, we determine which method provides the best results and which parameters seem to better estimate the observed intensity in terms of R2 of the obtained regression law, and we test the stability of the derived GMICEs using 3-fold cross-validation. Our MI data, expressed in MCS scale and binned into classes at 0.5 intensity intervals, is distributed from 2 to 8, so this is the intensity range in which extracted laws are expected to be valid.