Strong ground shaking has the potential to generate significant dynamic strains in shallow materials such as soils and sediments, thereby inducing nonlinear site response resulting in changes in near-surface materials. The nonlinear behaviour of these materials can be characterized by an increase in wave attenuation and a decrease in the resonant frequency of the soil; these effects are attributed to increased material damping and decreased seismic wave propagation velocity, respectively. This study investigates the 'in-situ' seismic velocity changes and the predominant ground motion frequency evolution during the 2016 Kumamoto earthquake sequence. This sequence includes two foreshocks (M-w 6 and M-w 6.2) followed by a mainshock (M-w 7.2) that occurred 24 hr after the last foreshock. We present the results of the seismic velocity evolution during these earthquakes for seismological records collected by the KiK-net (32 stations) and K-NET (88 stations) networks between 2002 and 2020. We analyse the impulse response and autocorrelation functions to investigate the nonlinear response in near-surface materials. By comparing the results of the impulse response and autocorrelation functions, we observe that a nonlinear response occurs in near-surface materials. We then quantify the velocity reductions that occur before, during, and after the mainshock using both approaches. This allows us to estimate the 'in-situ' shear modulus reduction for different site classes based on V-S30 values (V-S30760 m s(-1)). We also establish the relationships between velocity changes, shear modulus reduction, variations in predominant ground motion frequencies and site characteristics (V-S30). The results of this analysis can be applied to site-specific ground motion modelling, site response analysis and the incorporation of nonlinear site terms into ground motion models.

A theory for modelling the evolution of elastic moduli of grain packs under increasing pressure is combined with a method that accounts for the presence of fine-grained particles to develop a new conceptual framework for computing the seismic velocities of compacting sediments. The resulting formulation is then used to construct a seismic velocity model for California's Central Valley. Specifically, a set of 44 sonic logs from the San Joaquin Valley are combined with soil textural data to derive the 3-D velocity variations in the province. An iterative quasi-Newton minimization algorithm that allows for bounded variables provided estimates of the nine free parameters in the model. The estimates low- and high-pressure exponents that resulted from the fit to the sonic log velocities are close to 1/2 and 1/3, respectively, values that are observed in laboratory experiments. Our results imply that the grain surfaces are sufficiently rough that there is little or no slip between grains. Thus, the deformation may be modelled using a strain energy function or free energy potential. The estimated Central Valley velocity model contains a 27 per cent increase in velocity from the surface to a depth of 700 m. Lateral variations of around 4 per cent occur within the layers of the model, a consequence of the textural heterogeneity within the subsurface.