This study presents effective probabilistic procedures for evaluating ground-motion hazard at the free-field surface of a nonlinear soil deposit located at a specific site. Ground motion at the surface, or at any depth of interest within the soil formation (e.g., at the structure foundation level), is defined here in terms either of a suite of oscillator-frequency-dependent hazard curves for spectral acceleration, S-a(s)(f), or of one or more spectral acceleration uniform-hazard spectra, each associated with a given mean return period. It is presumed that similar information is available for the rock-outcrop input. The effects of uncertainty in soil properties are directly included. This methodology incorporates the amplification of the local soil deposit into the framework of probabilistic seismic hazard analysis (PSHA). The soil amplification is characterized by a frequency-dependent amplification function, AF(f), where f is a generic oscillator frequency. AF(f) is defined as the ratio of S-a(s)(f) to the spectral acceleration at the bedrock level, S-a(s)(f). The estimates of the statistics of the amplia fication function are obtained by a limited number of nonlinear dynamic analyses of the soil column with uncertain properties, as discussed in a companion article in this issue (Bazzurro and Cornell, 2004). The hazard at the soil surface (or at any desired depth) is computed by convolving the site-specific hazard curve at the bedrock level with the probability distribution of the amplification function. The approach presented here provides more precise surface ground-motion-hazard estimates than those found by means of standard attenuation laws for generic soil conditions. The use of generic ground-motion predictive equations may in fact lead to inaccurate results especially for soft-clay-soil sites, where considerable amplification is expected at long periods, and for saturated sandy sites, where high-intensity ground shaking may cause loss of shear strength owing to liquefaction or to cyclic mobility. Both such cases are considered in this article. In addition to the proposed procedure, two alternative, easier-to-implement but approximate techniques for obtaining hazard estimates at the soil surface are also briefly discussed. One is based on running a conventional PSHA with a rock-attenuation relationship modified to include the soil response, whereas the other consists of using a simple, analytical, closed-form solution that appropriately modifies the hazard results at the rock level.
