Ergodic site response model for subduction zone regions

We present an ergodic site response model with regional adjustments for use with subduction zone ground-motion models. The model predicts site amplification of peak ground acceleration, peak ground velocity, and 5% damped pseudo-spectral accelerations of the orientation-independent horizonal component for oscillator periods from 0.01 to 10 s. The model depends on the time-averaged shear-wave velocity in the upper 30 m (V_S30), basin depth, and region and is independent of subduction earthquake type. It has three components: a linear site-amplification term in the form of V_S30-scaling, a nonlinear term that depends on V_S30 and shaking intensity parameterized by peak ground acceleration at the reference-rock velocity condition of 760 m/s, and a basin sediment-depth term for Japan and Cascadia conditioned on the depth to the 2.5 km/s shear-wave velocity isosurface (Z_2.5). A global V_S30-scaling model is provided along with regional adjustments for Japan, Taiwan, South America, Alaska, and Cascadia. The nonlinear model is global, with a functional form that has often been used to fit nonlinear responses inferred from simulations, but here we calibrate it empirically. Relative to a prior model for shallow earthquakes in active tectonic regions, our subduction zone global V_S30-scaling is comparable at short periods (<1.0 s) but weaker at long periods, while the nonlinear site response is generally less pronounced but extends to lower levels of shaking. Basin depth models are conditioned on the difference of the actual Z_2.5 and a V_S30-conditioned mean Z_2.5. Sites with positive differential depths have increased long-period site responses and decreased short-period responses, with the opposite occurring for negative differential depths.