AGU Fall Meeting, San Francisco, USA, 5 - 9 December, 2011
A.Emolo, A.Zollo, L.Elia, C.Martino, S.Colombelli, G.Festa, G.Iannaccone
We present an approach to Earthquake Early Warning for Southern Italy that integrates regional and on-site systems. The regional approach is based on the PRESTo (PRobabilistic and Evolutionary early warning SysTem) software platform. PRESTo processes the 3C acceleration data streams from the stations of the Irpinia Seismic Network (ISNet) for P-waves arrival detection and, once an event has been detected, provides location and magnitude estimations, and peak ground motion predictions at target sites. The earthquake location is obtained through an evolutionary, real-time probabilistic approach based on an equal differential time formulation profiting of information from both triggered and not-yet-triggered stations. The magnitude is estimated by an empirical relation that correlates it with the peak displacement measured on the first 2-4s of signal, starting at the detected P-wave. The peak ground motion parameters at a site of interest can be finally estimated by specific ground motion prediction equations once that the location and magnitude are known. All estimates are provided as a probability density functions with an uncertainty that typically decreases with time. Alarm messages can thus reach vulnerable structures before the destructive waves. PRESTo has been under continuous real-time testing during the past two years using data streaming from ISNet and has produced a bulletin of about a hundred low-magnitude events. Meanwhile, PRESTo has been tested off-line playing back in the system both real and synthetic seismograms for moderate-to-large events. The tests performed have shown that a stable solution is generally reached within 10s from the origin time when a dense seismic network is deployed in the source area. The on-site system is based on the realtime measurement, at near-source stations located at increasing distances from the earthquake epicenter, of the peak displacement and P-wave dominant period, measured on a 3s window after the first P arrival. These values are compared to thresholds, set for a minimum magnitude 6 and instrumental intensity VII, based on the empirical regression analyses of strong motion data. Thus, each site independently provides an alert level according to the measured parameters and a decision table, and then the method identifies the extent of the potential damaged zone as inferred from continuously updated mapping of the alert levels at the near-source stations. Due to the unavailability of large magnitude events recorded at ISNet, the feasibility and robustness of such an approach have been tested off-line using data recorded in Italy and Japan by dense seismic networks, providing a very consistent matching between the rapidly predicted and observed damage zones. While in the regional approach the ground-motion parameters are computed only using the earthquake source parameters, the integrated approach is likely to provide a more robust prediction of the potential earthquake damaging effects. At the same time, the combined use of both near-source and distant stations, allows for a faster alert notification with respect to a classical on-site system with sensors deployed near the target.