Stochastic characteristics of seismic excitations at a non-uniform (rock and soil) site

Abstract

This study analyzes data recorded at the dense array in the Parkway Valley, Wainuiomata, New Zealand, a small alluvial valley surrounded by graywacke outcrops. The array consisted of stations on both the sediment basin and the surrounding soft rock, with station separation distances pertinent for earthquake engineering applications. The array's configuration renders itself uniquely for the study of the spatial variation of seismic ground motions at sites with irregular topography for bridge response evaluations: the locations of the soft-rock stations surrounding the valley may be viewed as the locations of the bridge's abutments, and the locations of the stations in the basin as those of the bridge's intermediate piers. A further-away station, north-east of the valley, provided information on firm-rock data. The analysis suggested that the motions in the valley were dominated by surface waves and considered two time windows for the evaluation of the stochastic characteristics (power spectral densities and lagged coherencies) of the data. During the first time window, at the onset of the excitation, the motions in the valley exhibited both low and high frequency energy that propagated down the valley. During the second window, high-amplitude, low-frequency, surface-wave energy controlled the horizontal motions in the valley. Very significant variabilities were observed in the amplitudes of the power spectra of the motions throughout the array during both windows indicating a complex wave propagation pattern affecting not only the valley but also the soft rock that surrounds the valley. Significant correlations in the motions appeared in the form of “hills” in the lagged coherency estimates when the energy of the power spectra of the station pairs peaked at similar or close-by frequency ranges, irrespective of whether the station pairs were located on the same or different ground types. The valley data indicated significant correlations in the dominant frequency ranges of the surface waves that control the motions during the two windows analyzed. Interestingly, motions at the soft-rock and soil stations, as well as those at the soft-rock stations and the firm-rock station, are significantly correlated at the onset of the excitation. This result may or may not depend on the event analyzed, as the spectra of its motions at the firm-rock, soft-rock and basin stations during this window peak at similar frequency ranges, and, in addition, the motions at the soft-rock and basin stations appear to be controlled by similar surface-wave energy at higher frequencies. On the other hand, during the later windows, there is tremendous amplification of the motions in the valley, whereas the spectral amplitudes of the soft-rock data are, comparatively, very low. The differences result from the fact that the seismic excitations within the valley have a significantly longer duration than the seismic ground motions recorded at the rock stations due to the formation of additional, low-frequency surface waves. This observation may have significant implications for the design of bridges supported at variable site conditions. Whereas the motions at rock (or excitations at the bridge abutments) have essentially ceased, the motions in the valley (or excitations at the bridge piers) undergo the severe part of the earthquake excitation. This scenario of seismic excitations is not currently considered in bridge response evaluations, but can adversely affect the structure's response. The analysis provides insight into the stochastic characteristics of seismic excitations at sites with irregular subsurface topography.