Assessment of Site Amplification Factors in Southern Lima, Peru Based on Microtremor H/V Spectral Ratios and Deep Neural Network

Abstract

Evaluation of site amplification factors (SAFs) of seismic waves has been one of the important issues for evaluating seismic hazards. The authors have proposed a deep neural network (DNN) model in order to cost-effectively and accurately estimate SAF from microtremor horizontal-to-vertical spectral ratio (MHVR). In this study, we assessed the SAFs in southern Lima, Peru by estimating from MHVRs and DNN. First, we validated the applicability of the DNN model to Lima by estimating the SAFs from the MHVRs observed at seismic stations in Lima. From the comparison with the observed SAFs derived from spectral inversion technique, we confirmed that the SAFs in Lima were accurately estimated by the DNN model. The SAFs in the southern Lima including Chorrillos and Villa El Salvador districts were evaluated by applying the DNN model to the observed MHVRs at approximately 250 sites. We found that large amplifications at low frequency around 1 Hz were expected in the southeastern coastal areas formed by eolian sands whereas smaller amplification were estimated in the northwestern areas mainly located on alluvial deposits.