AGU Fall Meeting, San Francisco, USA, 5 - 9 December, 2011
Aldo Zollo, Simona Colombelli, Hiroo Kanamori
The mega-thrust, Mw 9.0, 2011 Tohoku earthquake has re-opened the discussion among the scientific community about the effectiveness of Earthquake Early Warning (EEW) systems, when applied to such large events. Many EEW systems are now under-testing or -development worldwide and most of them are based on the real-time measurement of ground motion parameters in a few second window after the P-wave arrival. Currently, we are using the initial Peak Displacement (Pd), and the Predominant Period (τc), among other parameters, to rapidly estimate the earthquake magnitude and damage potential.
A well known problem about the real-time estimation of the magnitude is the parameter saturation. Several authors have shown that the scaling laws between early warning parameters and magnitude are robust and effective up to magnitude 6.5-7; the correlation, however, has not yet been verified for larger events.
The Tohoku earthquake occurred near the East coast of Honshu, Japan, on the subduction boundary between the Pacific and the Okhotsk plates. The high quality Kik- and K- networks provided a large quantity of strong motion records of the mainshock, with a wide azimuthal coverage both along the Japan coast and inland. More than 300 3-component accelerograms have been available, with an epicentral distance ranging from about 100 km up to more than 500 km. This earthquake thus presents an optimal case study for testing the physical bases of early warning and to investigate the feasibility of a real-time estimation of earthquake size and damage potential even for M > 7 earthquakes.
In the present work we used the acceleration waveform data of the main shock for stations along the coast, up to 200 km epicentral distance. We measured the early warning parameters, Pd and τc, within different time windows, starting from 3 seconds, and expanding the testing time window up to 30 seconds. The aim is to verify the correlation of these parameters with Peak Ground Velocity and Magnitude, respectively, as a function of the length of the P-wave window. The entire rupture process of the Tohoku earthquake lasted more than 120 seconds, as shown by the source time functions obtained by several authors. When a 3 second window is used to measure Pd and τc the result is an obvious underestimation of the event size and final PGV. However, as the time window increases up to 27-30 seconds, the measured values of Pd and τc become comparable with those expected for a magnitude M≥8.5 earthquake, according to the τc vs. M and the PGV vs. Pd relationships obtained in a previous work.
Since we did not observe any saturation effect for the predominant period and peak displacement measured within a P-wave, 30-seconds window, we infer that, at least from a theoretical point of view, the estimation of earthquake damage potential through the early warning parameters is still feasible for large events, provided that a longer time window is used for parameter measurement. The off-line analysis of the Tohoku event records shows that reliable estimations of the damage potential could have been obtained 40-50 seconds after the origin time, by updating the measurements of the early warning parameters in progressively enlarged P-wave time windows from 3 to 30 seconds.