P-wave based extended source models for Earthquake Early Warning applications

Standard approaches for the peak motion prediction in Earthquake Early Warning (EEW) methods are typically based on the point-source approximation and on 1D empirical attenuation relationships, depending on magnitude and hypocentral distance. For large events (M>6) such a simplified representation is inadequate and may result in unreliable predictions of the expected shaking, thus reducing the effectiveness of the EEW systems. Here we use a methodology to improve the accuracy of real time ground motion prediction, through the fast determination of magnitude and fault plane geometry. The approach is based on the real-time, progressive measurement of the P-wave maximum amplitude along the recorded seismograms (LPXT curves, X is D, V or A for displacement, velocity or acceleration) at different azimuths and distances. The LPXT curves are used to estimate the seismic moment and the source duration, which is, in turn a measure of the source length, assuming a near-triangular moment rate function. At the same time, the azimuthal variation of peak amplitudes provides a first-order estimation of the best fault plane solution mechanism and, possibly, the dominant rupture direction. These estimates are used to build simplified kinematic source models. The fault geometry is constrained combining the classical scaling laws and the mechanism early solutions. The rupture speed and the rise time are selected accounting for the medium elastic properties and the event magnitude. A single patch slip distribution is imposed: its extension and position with respect to the nucleation are controlled by the size estimates from LPXT curves and by preliminary directivity estimates, respectively. The convolution of these models with pre-computed Green’s functions provides complete wavefield synthetic seismograms and thus early estimates of the expected intensity measures (PGA/PGV) at the EEW target sites. The alert decision scheme is thus defined upon the exceedance of a user-compliant PGA/PGV threshold by the predicted synthetic values. Preliminary off-line tests on the Mw 6.5 Norcia earthquake case-study have shown the feasibility of this approach through the consistency between the fast predicted PGVs from synthetics and the observed values, due to the improved modeling of the up-dip and along strike directivity effects