Prediction of Land Subsidence in Chiba Prefecture Caused by Liquefaction Based on Machine Learning

Abstract

In this study, the authors developed a machine learning model to predict land subsidence caused by liquefaction resulting from an earthquake, with the aim of improving liquefaction hazard maps. A numerical model for predicting land subsidence due to liquefaction using XGBoost (eXtreme Gradient Boosting), which is one of the ensemble machine learning methods, was developed based on the results estimated by the Nankai Trough Earthquake Model Examination Committee. Additionally, the numerical model was applied to estimate the land subsidence in Chiba Prefecture, where liquefaction was extensively observed after the 2011 Great East Japan Earthquake. The results showed a geotechnically valid map of expected land subsidence because significant subsidence was observed in the areas where liquefaction conditions were met.