State fusion entropy for real-time and site-specific analysis of landslide stability changing regularities

Stability analysis is of great significance to landslide hazard prevention, especially the real-time stability. However, many existing stability analysis methods are difficult to analyse the real-time landslide stability and its changing regularities in a uniform criterion due to the unique landslide geological conditions. Based on the relationship between displacement monitoring data, deformation states and landslide stability, a state fusion entropy method is herein proposed to derive landslide instability through a comprehensive multi-attribute entropy analysis of deformation states which are defined by a proposed joint clustering method combining K-means and cloud model. Taking Xintan landslide as the detailed case study, cumulative state fusion entropy presents an obvious increasing trend after the landslide entered accelerative deformation stage and historical maxima match highly with the evolutionary stages of Xintan landslide in key time nodes. Reasonable results are also obtained in its application to several other landslides in the Three Gorges Reservoir in China. Combined with field survey, state fusion entropy may serve as a novel index for judging landslide stability changing regularities and for landslide early warning.