Susceptibility Model for Sinkholes Caused by Damaged Sewer Pipes Based on Logistic Regression

Abstract

The occurrence of anthropogenic sinkholes in urban area can cause serious social losses. A damaged and aged sewer pipes beneath the road contribute to occur such a phenomenon. This study used the best subsets regression method to develop a logistic regression model that calculate the susceptibility for sinkholes induced by damaged sewer pipes. The model was developed by analyzing both the sewer pipe network and cases of sinkholes in Seoul. Among numerous sewer pipe characteristics were analyzed as explanatory variables, the length, age, elevation, burial depth, size, slope, and materials of the sewer pipe were found to influence the occurrence of sinkhole. The proposed model reasonably estimated the sinkhole susceptibility in the area studied, with an area value under the receiver operating characteristics curve of 0.753. The proposed methodology will serve as a useful tool that can help local governments choose a cavity inspection regime and prevent sinkholes induced by damaged sewer pipes.