Is nonlinearity needed to predict the Kumamoto Mj 6.5 and Mj 7.3 mainshocks at KUMA site?

Abstract We present the results of 2D numerical simulation carried out using FLAC3D-7.0 for the blind predictions of the shaking induced by the 2016 Kumamoto mainshocks Mj 6.5 and Mj 7.2. The blind exercise was organized by the IASPEI/IAEE Joint Working Group on the Effects of Surface Geology on Seismic Motion (JWG-ESG6), for the 6th International ESG Symposium held in Japan in August 2021. We have drawn our subsoil model using geophysical data provided by the JWG-ESG6 and validated using a weak motion Mj 4.6 event recorded at the target site for predictions (KUMA) and at the reference site (SEVO). Prediction of events Mj 6.5 and Mj 7.2 are performed with linear elastic and nonlinear elasto-plastic simulations, the latter using the Mohr-Coulomb constitutive law to characterize some shallow layers. Their syntheticwaveforms are almost identical and fit the data satisfactorily, most likely due to the mechanical parameters and the constitutive model chosen being not able to induce important nonlinear effects. Also, and more surprisingly, a 1D linear code can reproduce the data. These results provide an important warning against the widespread use of equivalent-linear code for ground response analysis, which, if used for predictions in the Kumamoto case, would have largely underestimated the recorded shaking.