In this article, we investigate, by means of simulations, the reliability of the attenuation parameter k when it is evaluated in the presence of site amplification effects. We consider a single soil layer overlying the bedrock, and we assume that k describes the attenuation due to propagation through the soil layer. We generate synthetic spectra for events with magnitudes from 2 to 6, taking into account different soil-layer properties. In particular, we consider 1D site transfer functions that account for fundamental resonance frequencies varying from 1 to 12 Hz and S-wave velocities ranging from 360 to 1000 m/sec. A least-squares fit to determine the slope of the high-frequency decay is performed over spectral windows having different widths.
The results show that site resonance at both intermediate (4 Hz ≤ f0 < 9 Hz) and high (f0 ≥ 9 Hz) frequencies leads to unreliable k estimates due to the spectral windows selected to perform the least-squares fit not being wide enough to average out local peaks resulting from site amplification effects. This is an important result because, in the analysis of weak motion data, the exploitable frequency band is usually no wider than those adopted in this work. The results relevant to the simulations that consider high fundamental resonance frequencies, close to or inside the spectral windows adopted to perform the fit, indicate that the contribution of k to the high-frequency decay can be completely masked even if an excellent fit is obtained. In these cases, the soil properties estimated from k values may be unreliable, especially when high impendance contrasts exist. For fundamental resonance frequencies well below the frequency range selected for least-squares fitting, fairly good results can be obtained.