Estimation of Site Response Spectra During an Earthquake Based on Kriging Model: A Case Study on the 2010 Darfield Mw 7.0 Earthquake

For estimating site response spectra during an earthquake, a Kriging model-based approach is proposed, which uses typical Ground Motion Prediction Equations (GMPEs) as a baseline and epsilon (ε) as a correction factor. The Kriging model is constructed ro predict epsilons based on five input parameters: rupture distance (R rup ), epicenter distance (E pi ), the clockwise angle (A z ), the averaged shear-wave velocity in the top 30 m (V s30 ) and altitude of the site (H a ). A case study is conducted on the 2010 Darfield Mw 7.0 earthquake. The results demonstrate that, even with limited input data, the proposed method yields lower Mean Squared Error (MSE) values compared to the median estimates derived from standard GMPEs. Meanwhile, the method captures the "bumpy" shape of an individual site response spectrum. Prediction accuracy is high when the target site is close to the sites used for training the Kriging model while deviations increase when the target site is far from the training sites.