The HVSR technique reveals pervasive nonlinear sediment response during the 1994 Northridge earthquake (Mw 6.7)

Abstract

The Northridge earthquake triggered a considerable number of nonlinearsite-response studies. The standard approach involves the standard spectralratio (SSR) method, requiring a suitable reference site (rarely available).Recently, Dimitriu et al. (1999, 2000) proposed the use of the(nonreference) horizontal-to-vertical spectral-ratio (HVSR) technique toassess nonlinear site response. The purpose of this article is to test it onNorthridge-earthquake data.Accelerograms from five sediment sites exhibiting low-frequency resonanceswere used. Hypocentral distances and recorded peak horizontal groundaccelerations (PHA) are between 19.9 km–27.7 km and381 cm/s2–827 cm/s2. For each recording, HVSRs werecomputed in a 6-s moving window, spanning the entire recording and thuscovering a wide range of excitation levels.All five sites are found to exhibit statistically significant negative correlation(r between –0.70 and –0.91) between the site's dominant-resonancefrequency, fres, determined visually from the HVSRs, and the mean(in the 6-s window) horizontal ground acceleration, MHA. Assuming alayer-over-halfspace site structure, it is possible to relate the variation offres to the change in the sediment-layer shear modulus, G,by Gs/Gw = [1-(fress-fresw]/fresw]2, where superscripts `s' and `w' represent the valuescorresponding to `strong' and `weak' motion. Shear-modulusdegradation, Gs/Gw, is a measure of the nonlinear responseof the (softest) layer(s) contributing the most to the resonance; the resultssuggest that these are the upper 30 m of sediments. Substituting thevalues of fres for the `strong' – (MHA > 80 cm/s2) and`weak' – (MHA < 35 cm/s2) motion ranges yielded realisticvalues of Gs/Gw for the sites considered, allowing to rankthem in accordance with the magnitude of their nonlinear response.