Extraction of landslide information and analysis of driving factors using multi-source remote sensing data and Bayesian algorithms

Landslides, as a typical type of geological hazard, are characterized by strong suddenness and high destructiveness, posing a serious threat to human life and property safety. In this study, multi-source remote sensing data were integrated with advanced algorithms to analyze the temporal and spatial evolution of landslides as well as their triggering factors. The results indicate that the Bayesian algorithm accurately identified three key landslide events (2008, 2014, and 2019) between 2000 and 2022 without any human intervention, corresponding respectively to the initial deformation stage, accelerated sliding stage, and seasonal activity stage of the landslides. Subsequently, through NDVI difference analysis based on Landsat and Sentinel-2 data, the expansion areas of the three landslides were quantitatively extracted—32,000 m² in 2008, 83,000 m² in 2014, and 132,000 m² in 2019. The results show an accelerating expansion trend, with the annual expansion rate increasing from 8,500 m²/a during 2008–2014 to 9,800 m²/a during 2014–2019. Further analysis using the Geographical Detector method revealed that soil type (q = 0.06), slope (q = 0.05), and vegetation coverage (q = 0.04) are the dominant factors affecting landslide stability.