Estimating subsurface velocity and resonance frequency changes using two decades of KiK-net data in the Iwate Prefecture, Japan

SUMMARY Earthquake ground motion is strongly influenced by near-surface geology, which governs its amplification, duration and spatial variability. Under intense shaking and depending on the material strength, sediments often exhibit nonlinear behaviour, producing large deformations that reduce shear-wave velocity, shift resonance frequencies and increase damping. We analyse over two decades of borehole–surface recordings from 28 stations in Iwate Prefecture, Japan, collected by the Kiban Kyoshin network (KiK-net), to quantify these effects. Frequency-domain analysis (stacked Stockwell power spectral density) and time-domain interferometric methods (multitaper deconvolution and phase cross-correlation) provide consistent results, revealing systematic decreases in both resonance frequency and seismic velocity with increasing peak ground acceleration (PGA). Frequency shifts inferred from the surface data mainly reflect the shallowest layers, whereas velocity changes estimated with borehole-referenced methods capture a path-averaged perturbation between the surface and borehole sensors and therefore depend on borehole depth. The data set is divided into seven PGA bins based on surface recordings, with the 1–5 $\mathrm{cm\ s^{-2}}$ bin serving as a baseline for comparison, representing linear site conditions. Across all stations, relative velocity reductions average $\sim$ 12 per cent in the 200–400 $\mathrm{cm\ s^{-2}}$ PGA range, corresponding to a shear modulus reduction ($\mu /\mu _0$) of about 23 per cent. Nonlinear effects are most pronounced at sites with thicker sedimentary deposits, which are mainly found in the central valley and northern foothills of Iwate Prefecture. In contrast, $V_{\rm S30}$ shows no clear correlation with the observed nonlinearity, as its averaging effect masks thin low-velocity layers near the surface that are prone to nonlinear response during strong shaking. These results underline that nonlinear site response is highly site-specific, and that large observational data sets are crucial for robust characterization across a seismic network.