Risk Assessment of Groundwater Depletion Induced Land Subsidence: A Case Study in Taiyuan Basin, China

Abstract

Groundwater depletion induced land subsidence affects the safety of local communities. In this research, a data-driven approach is applied to predict and assess the risk level of land subsidence in Taiyuan Basin, Shanxi Province, China based on Geographic Information Systems data and field investigation data. First, the relevance vector machine is introduced to model and classify the risky/non-risky ground subsidence cases. Next, an ensemble classifier using multiple relevance vector machines and elastic net is proposed in this research. Multiple land subsidence locations in Taiyuan Basin have been investigated in this study. Five benchmarking machine learning classification algorithms including decision tree, random forest, multi-layer perceptron, support vector machine and classical relevance vector machine have been compared in this study. Four evaluation metrics including accuracy, sensitivity, specificity and area under the receiver operating characteristic curve have been introduced to assess the classification performance of all the algorithms tested. Computational results demonstrated the outperformance of the proposed approach in classifying risky land subsidence cases compared with other benchmarking algorithms.