Comparative application of dispersion curve inversion strategies. Case study of noise arrays in the Euroseistest site, Greece

Abstract

ABSTRACTCurrently, the use of ambient noise arrays has become fairly routine for site characterization applications. Conventionally, the first step in the analysis of ambient noise arrays is the computation of the dispersion curve, defined as the Rayleigh wave phase velocity dependence on frequency. The second more complex step is the inversion of the dispersion curve to obtain a shear‐wave velocity profile. In many engineering applications, where only the time‐averaged shear‐wave velocity, termed Vs30, is needed, a relationship between the Rayleigh phase wave velocity at a given wavelength, VR(λ), and Vs30 is commonly developed for the specific area of study. We compare the results from two recently proposed inversion strategies, the first one is based on the misfit criteria, the second on the Akaike criteria, which we apply to experimental data acquired from the Euroseistest site in Greece. We also show that the two inversion strategies have their limits: for the first strategy, using misfit criteria and constraints, the restriction of describing the dispersion curve within the frequency band defined from the fundamental frequency to ten times this frequency is sometimes difficult to fulfil. On the other hand, the second strategy, using the Akaike criteria, is time‐consuming and requires large data storage. Finally, we conclude that the VR(λ=37m)‐Vs30 relationship is promising but should be confirmed with more data analysis.