One-dimensional site response analysis (1D SRA) remains the world state of practice for assessing site-specific site response in engineering projects. The 1D SRA numerical approach condenses the complexities of the 3D wave propagation phenomena into a simple horizontally polarized wave vertically traveling through a soil column, thus leading to errors in site response predictions. This dissertation proposes two approaches: (1) an approach to capture the effect of shear-wave velocity (V_S) spatial variability on site response using 1D SRAs, and (2) an approach for conducting 1D SRAs to account for the effect of unmodeled features affecting site response (e.g., inclined waves) and the potential for higher site amplifications. These approaches and the findings learned during their development are herein presented to provide practical recommendations expected to improve site response predictions using 1D SRAs.

A numerical investigation using 2D and 1D SRAs is conducted to develop an approach for capturing 2D V_S spatial variability effects on site response using 1D SRAs with randomized V_S profiles. The limitations of 1D SRAs with V_S randomization are mainly due to (1) the excessive randomization and the assumption that the resulting mean site response is representative, and (2) the intrinsic inability of 1D SRAs to capture wave propagation effects (e.g., constructive interference). Results from this investigation indicate that the 84th seismic response estimated from 1D SRAs conducted with fifty randomized V_S profiles generated using the Toro model (1995) with V_S standard deviation, σ_lnVs = 0.25 approximates well the median 2D site response at the site’s fundamental frequency, regardless of what the site-specific σ_lnVs is. Comparisons with data from four borehole sites classified as Group A (Tao and Rathje, 2020) support this observation.

An approach for conducting 1D SRAs is developed based on comparisons between 1D SRA predictions and borehole ground-motion data, with two objectives: (1) to improve site response predictions, and (2) to account for the 1D-SRA bias and the potential for underpredicting the estimated seismic response. The first objective is achieved by using randomized V_S profiles with σ_lnVs and damping multipliers (D_mul) that reduce intrinsic errors carried in 1D SRAs, such as the overpredictions at the site’s resonant frequency. Results from this work indicate that the σ_lnVs-D_mul pair leading to the lowest root mean square error between the observed and 1D SRA-based transfer functions and amplification factors is σ_lnVs = 0.25 and D_mul = 3. The second objective is achieved by acknowledging the 1D-SRA bias (c_3D^SRA ) and the potential for under- and overpredictions due to modeling errors carried by 1D SRAs, quantified as the standard deviation of the site-specific bias-corrected mean residuals (ϕ_S2S^SRA ). The c_3D^SRA is used to bias-correct results from 1D SRAs with V_S randomization, thus obtaining the best estimate site response. The potential for higher site amplifications is subsequently accounted for by computing the 95th site response percentile using ϕ_S2S^SRA.