On the Discrepancy of Recent European Ground-Motion Observations and Predictions from Empirical Models: Analysis of KiK-net Accelerometric Data and Point-Sources Stochastic Simulations

Titre de la revue : Bulletin of the Seismological Society of America
Volume : 98
N° : 5
Pagination : 2244-2261
Date de publication : 01/10/2008

Ground-motion amplitudes from recent European earthquakes of moderate magnitudes have been observed to be systematically smaller than the values expected from a number of popular empirical ground-motion prediction equations used for seismic hazard analysis. It has been suggested that these discrepancies are caused by regional variations in seismotectonic character and as a consequence that seismic hazard estimates using these ground-motion models would be too high. In the present article, we explore the hypothesis that the discrepancy can simply be explained by the fact that these models adopt magnitude-independent functional forms (letting Y designate the response spectral acceleration M designate the magnitude, then d log Y=dM is assumed independent of both magnitude and distance). The data collected by the KiK-net array (see the Data and Resources section) provides a unique opportunity to test this hypothesis and to analyze some of the pitfalls of deriving magnitude-independent functional form models and applying them for predictions of ground motion from smaller events (and vice versa). Borehole rock KiK-net ground-motion data (337 events, 3894 records) have been used to derive empirical ground-motion models with magnitude-independent functional forms for various magnitude ranges. By using these new ground-motion models and stochastic simulations, we discuss the ground-motion distance decays and magnitude effects for ground-motion models obtained with different magnitude range datasets. This analysis clearly indicates that response spectral amplitudes of ground motions from large earthquakes decay slower with distance than those from small earthquakes and confirms that the magnitude scaling of ground motion decreases as earthquake magnitude increases. Using stochastic simulations, we demonstrate that the observed decay in scaling could be a mixture of geometrical decay from extended sources and the fact that response spectral values instead of Fourier spectral values are considered. New ground-motion models (with functional forms including coefficients to model the observed magnitude-dependent scaling and decay rate) have finally been calculated for both surface and borehole site conditions of the analyzed dataset. These models show similar decays for intermediate period or moderate magnitude earthquakes. Our site classifications remove most of the statistical trend of the site effect and suggest that source and path effects could dominate the aleatory variability.