Site-specific comparisons of random vibration theory-based and traditional seismic site response analysis

Seismic site response analysis is used to assess the effect of site-specific soil conditions on expected earthquake ground motions. For this purpose, input earthquake rock motions are propagated through a site-specific soil column to obtain motions at the ground surface. Due to the high variability in the expected input rock motions, a suite of input motions must be used and the number of motions should be relatively large to obtain a statistically stable estimate of the surface response spectrum. Unfortunately, the number of input motions used in practice in a site response evaluation is generally less than five, which results in uncertain predictions of the surface ground motion. An alternative to conventional site response analysis is a Random Vibration Theory (RVT) approach. Due to its stochastic nature, RVT analysis can provide an estimate of the site response without the need to choose any input time histories for analysis. Therefore, RVT is a potentially powerful tool for site response analysis that can provide fast and accurate estimates of the surface ground motion at a site. This study applies RVT to equivalent-linear site response analysis and provides site-specific comparisons between RVT site response analysis and traditional site response analysis, which uses time-domain input rock motions. Comparisons are made for various site conditions and earthquake scenarios. The site amplification predicted by RVT was similar to that predicted from the time domain simulations for most periods. However, at the fundamental site period, RVT-based analysis tended to overpredict amplification. A study was performed in which the time series parameters incorporated in RVT were computed from recorded ground motions and compared with those from RVT theoretical predictions. These comparisons were generally favorable, given that the duration of the forward directivity motions was appropriately characterized. RVT site response analysis for forward-directivity motions was performed by characterizing the near-fault, forward-directivity input motions via modified acceleration response spectra with enhanced long period spectral values. Time history site response analyses were performed with forward-directivity and non-directivity input rock motions. In general, the RVT analysis did not accurately predict the site response for the forward directivity motions, revealing that a more thorough characterization of the forwarddirectivity motions is required for input into RVT site response.