Interevent Variability in Ground-Motion Prediction Models: The Role of the Source Parameter Selection

We investigate the dependence of interevent residuals on the choice of source parameters used in ground-motion prediction models calibrated for peak ground acceleration, peak ground velocity, and peak ground displacement. Using a dataset of 877,566 recordings from 1586 earthquakes with magnitudes ranging from 1 to 6.5 in central-southern Italy, we perform multiple mixed-effects regressions, exploring different approaches for the source scaling component. We compare the interevent standard deviation of models based on various source parameters, including moment magnitude, local magnitude, radiated energy, source spectral value at 3 Hz, moment magnitude with stress drop, and moment magnitude with apparent stress. Our results show that combining moment magnitude with either stress drop or apparent stress yields the lowest variability across all peak parameters, as expected. In addition, using local magnitude effectively captures the stress-drop-related component of variability. For the analyzed magnitude range, the source spectral amplitude at 3 Hz performs similarly to local magnitude in this regard, without saturating for large magnitudes. These findings suggest that source parameter choices complementary or alternative to moment magnitude can help reduce interevent variability. However, the suitability of models based on parameters other than moment magnitude depends on the specific application.