Land subsidence due to groundwater pumping: hazard probability assessment through the combination of Bayesian model and fuzzy set theory

Abstract

Land subsidence caused by groundwater overpumping threatens the sustainable development in Beijing. Hazard assessments of land subsidence can provide early warning information to improve prevention measures. However, uncertainty and fuzziness are the major issues during hazard assessments of land subsidence. We propose a method that integrates fuzzy set theory and weighted Bayesian model (FWBM) to evaluate the hazard probability of land subsidence measured by Interferometric Synthetic Aperture Radar (InSAR) technology. The model is structured as a directed acyclic graph. The hazard probability distribution of each factor triggering land subsidence is determined using Bayes' theorem. Fuzzification of the factor significance reduces the ambiguity of the relationship between the factors and subsidence. The probability of land subsidence hazard under multiple factors is then calculated with the FWBM. The subsidence time series obtained by InSAR is used to infer the updated posterior probability. The upper and middle parts of the Chaobai River alluvial fan are taken as a case-study site, which locates the first large-scale emergency groundwater resource region in the Beijing plain. The results show that rates of groundwater level decrease more than 1 m yr−1 in the confined and unconfined aquifers, with cumulative thicknesses of the compressible sediments between 160 and 170 m and Quaternary thicknesses between 400 and 500 m, yielding maximum hazard probabilities of 0.65, 0.68, 0.32, and 0.35, respectively. The overall hazard probability of land subsidence in the study area decreased from 51.3 % to 28.3 % between 2003 and 2017 due to lower rates of groundwater level decrease. This study provides useful insights for decision makers to select different approaches for land subsidence prevention.