CLUSTERING-BASED ANALYSIS OF GROUND MOTION DIRECTIONALITY: PRELIMINARY INTENSITY MEASURES AND HYPOCENTRAL CORRELATIONS IN THE ITALO-BALKAN PENINSULA

Earthquake ground motion parameters exhibit significant variability when considering the simultaneous interaction of horizontal components and their orientation within the horizontal plane. Traditionally, these effects have been analyzed by evaluating each direction independently and establishing probabilistic correlations. This study proposes an alternative approach that unifies the Cartesian components of ground motion into a single framework, exploding an extensive dataset of recorded seismic events from the Italian and Balkan peninsulas. By representing all response ordinates in polar coordinates, this method effectively captures the variability across all horizontal orientations as a function of the rotation angle relative to the azimuth of maximum response. A convex hull approach is employed to determine the shape of the grouped ground motion, using the newly proposed maximum rotation-independent principal ground acceleration as a reference. These results were then used to approximate a simplified ellipsoidal envelope model for correlating the maximum rotation-independent principal ground acceleration with its orthogonal component, denoted b value. Some seismological classifications and Intensity Measures AI-driven clustering were performed to highlight underlying parameters and propose a generalized expression for the location under study. By incorporating these directional effects, which are often overlooked in probabilistic seismic hazard analyses and in the majority of building codes, the findings provide a refined framework for preliminarily estimating ground motions directionality. These findings enhance seismic design and retrofitting by correlating structural orientation and earthquake directionality, improving resilience and performance.