Nonlinear Seismic Ground Response Analysis: Code Usage Protocols and Verification against Vertical Array Data

One-dimensional seismic ground response analyses are often performed using equivalent-linear procedures, which require few, generally well-known parameters. Nonlinear analyses have the potential to more accurately simulate soil behavior, but their implementation in practice has been limited because of poorly documented and unclear parameter selection and code usage protocols as well as inadequate documentation of the benefits of nonlinear modeling relative to equivalent linear modeling. Regarding code usage/parameter selection protocols, the following are described: (1) when input motions are from ground surface recordings, we show that the full outcropping motion should be used without converting to a "within" condition; (2) Rayleigh damping should be specified using at least two matching frequencies with a target level equal to the small strain soil damping; (3) the "target" soil backbone curves used in analysis can be parameterized to capture either the soil's dynamic shear strength when large-strain soil response is expected (strains approaching 1%), relatively small-strain response (i.e., γ < 0.3%) as inferred from cyclic laboratory tests, or a hybrid of the two; (4) models used in nonlinear codes inevitably represent a compromise between the optimal fitting of the shapes of backbone and hysteretic damping curves, and we present two alternatives for model parameterization. The parameter selection and code usage protocols are tested by comparing predictions to data from vertical arrays. We find site amplification to be generally underpredicted at high frequencies and overpredicted at the elastic site period, where a strong local resonance occurs that is not seen in the data. We speculate that this bias results from over-damping.