A reference site has to be free of amplification or deamplification effects, namely with no troughs and peaks in its Fourier amplitude spectrum. At the Cavola landslide we show that this spectrum is dependent on the direction of propagation of wavefronts for incidence angles in the range 30°–90°. Our study is based on comparison of spectral ratios from observations and 2D numerical simulations. We have modeled propagation in a 2D profile for SH and SV waves with several incidence angles in a range from 0° to ±90°, where 0° and 90° are, respectively, vertical and horizontal incidence, except that ±90° denotes Rayleigh waves in the P-SV. We discuss in detail the result for angles of incidence of 0°, ±20°, ±60°, ±90°. We have obtained observed horizontal-to-horizontal spectral ratios using three reference sites. Two of these have matching receivers in the model, located at the opposite ends of the 2D profile. Overall observations are matched best when the reference site is located on the same side of the landslide as the incoming wavefront. We also find general agreement of the observed H/H spectral ratios from earthquakes with H/V and H/H spectral ratios from noise, and the match between H/V values from noise and synthetic spectral ratios using an absolute, flat half-space reference is very good. On the other hand, 1D modeling performs poorly in comparison with 2D modeling in our case, for which the shape ratio h/D=0.2 is intermediate between primarily 1D and strongly 2D wave propagation according to the classification of Bard and Bouchon (1985).

You do not currently have access to this article.