Workshop on Fracture Dynamics: Theory and application to earthquakes, Madrid, Spain, September 26-28, 2005
V. Convertito, R. De Matteis, M. Lancieri, A. Zollo and G. Iannaccone
Humans life loss reduction planning in the event of large earthquakes occurrence requires that maps, known as shake maps (Wald et al., 1999) of instrumental ground-motion being generated and communicated to civil protection for emergency response operations. In order to have a realistic representation of the potential damages it is important to set up a robust and reliable tool to estimate strong ground-motion parameters such as, peak ground acceleration, peak ground velocity or displacement to correlate with fragility curves of building environment. Classically, this goal is accomplished using the attenuation laws that, given the magnitude and the source-to-site distance, allow to estimate the strong ground-motion parameters both for rock or soft-soils. However, classical attenuation laws have some limitations in so far as, do not allow to take into account for the details and complexity of the causative fault. As consequence, the shaking maps might not represent the real pattern of damages. Many efforts have been made in recent years in order to introduce both empirically and theoretically (e.g., Abrahamson et al. 1997, Sommerville et al. 1997, Spudich et al. 1997) seismic source parameters in the classical attenuation laws. However, these approaches suffer for some inherent limitations particularly due to lack of data that, for large earthquakes recorded in near-fault range generally are saturated or not available at all. On the other hand, a non secondary problem concerns the fact that rarely the data on which the attenuation laws are retrieved are collected in the same region in which they will be applied giving rise to underestimations or overestimations of the predicted ground motion (host-target problem). In the present work, we propose an application of "near-real time" peak ground motion regression analysis for the Campania region (southern Italy) in which an advanced dense network, equipped with high dynamic instruments, is going to be installed for early warning applications. During the earthquake an increasing number of peak ground motion values are available so that it should be useful to perform "near-real time" data regression. This approach is based on the updating of the regression model using further information, about source mechanism and geometry, available with increasing time after event occurrence. As a consequence, a more realistic shaking map should be obtained. One of the main problems of this approach concerns the selection of the regression model, i.e. the equation and its associated coefficients, that is then used to predict the peak ground motion in all the sites where no records are available. However, this difficulty can be partly overcame by putting some constraints inferred from an existing data-base relative to the source region under study. This is true even if the data-base covers a range of magnitudes lower than that of interest for the early warning, alarm system. In fact, from the data-base it is possible to retrieve scaling laws that can be extrapolated for the larger magnitudes and average values of the anelastic attenuation that do not depend on the magnitude range. Thus, regional attenuation laws that provide the required constraints on the "near-real time" regression model can be retrieved using synthetic accelerograms for the magnitude range of interest. We analyzed a data-base of waveforms recorded from 1988-2003 by the Istituto Nazionale di Geofisica e Vulcanologia (INGV) network in the region of interest for a magnitude range Md (1.5, 5.0). We inverted P-wave spectra by assuming an -2 spectral model (Brune, 1970) in order to estimate seismic moment, corner frequency and Qp values. We used the stochastic approach proposed by Boore (1983) to generate peak ground accelerations from which a regional attenuation law has been obtained. In order to test the method and to compare the classical shaking maps with those obtained in ?near-real time?, we show a test case for an Mw 6.0 earthquake for which synthetic recordings at the early warning network are available.