Rapid prediction of landslide dam stability using the logistic regression method

Abstract

The accurate and rapid prediction of landslide dam stability is of great significance for emergency response planning. However, current rapid prediction methods for the landslide dam cannot quantitatively consider the influence of landslide debris grain size distribution. A database was established based on 1434 documented historical landslide dams, including formed-unstable and formed-stable cases from around the world. The logistic regression method was utilized to develop new methods for the rapid prediction of landslide dam stability, which can consider the morphological characteristics and particle composition of the landslide dams as well as the hydrodynamic conditions of the upstream dammed lake. According to the available information on landslide debris particle composition, the newly proposed rapid prediction methods were classified as either detailed or simplified based on 27 and 150 cases, respectively. Based on the database, several typical methods for the rapid prediction of landslide dam stability were chosen to compare with the newly proposed methods. The performances of each method testify to the rationality of the new methods.