Systematic assessment of nonlinear soil behavior at KiK-net sites in Japan: Thresholds and controlling site factors

We address two open questions concerning nonlinear soil behavior in a seismic hazard and risk context; (1) which site proxies can be used to predict and map nonlinear site amplification? (2) At which level of ground-motion intensity should such nonlinear models be considered? To answer these questions, we use the KiK-net network in Japan, which includes stations with instruments at both surface and depth, considering events recorded between 1997 and 2024. Using the surface-to-borehole ratio, we systematically capture the empirical effects of nonlinear soil response as the amplitude change and frequency shift between individual events and the linear site response. We then derive station-specific parameters for degree of nonlinearity and threshold for onset of nonlinear behavior. The statistical correlation between nonlinearity and a selection of geotechnical and geological site parameters shows that although parameters characterizing the depth to bedrock and the shallowest part of the soil layer have a promising potential for capturing nonlinear site amplification, the correlation is generally low, suggesting that a single site parameter is not sufficient. As a consistent reference for ground-motion intensity, we empirically calculate PGA emp.rock , as an approximation for PGA recorded on a standard outcropping rock site with V S30 = 760 m / s (average shear-wave velocity of upper 30 m). When analyzing the nonlinear behavior for all recorded events, we define the nonlinear soil behavior as significant when the amplitude change, and frequency shift are greater than 10% for the majority (50%) of the records. We find that in the PGA emp.rock -range 1−3 m / s 2 nonlinear soil behavior is significant only for soft soil stations ( V S30 < 250 m / s) with intermediate sediment thickness (<30 m). While, according to the mean behavior of all sites, regardless of grouping, nonlinearity is significant only at PGA emp.rock > 3 m / s 2 . These results show that for nonlinear site-amplification modeling, the onset of nonlinearity is strongly related to the site conditions.