The integration of slow-moving landslide features, MT-InSAR data, damage survey results and deep learning algorithms for building vulnerability zoning and forecast

The Three Gorges Reservoir area is characterized by complex natural and geological conditions, leading to frequent landslides along the Yangtze River's main stream and its tributaries. Regular human activities, including the regulation of the Yangtze River water levels, urban development, and infrastructure expansion, combined with heavy rainfall, dynamically alter the state of existing slow-moving landslides and provoke new slope failures. Despite the significant risk, there is a notable lack of vulnerability zonation and prediction methods at the township level. This study introduces a comprehensive approach aimed at assessing landslide vulnerability in Dazhou Town, located in the northeastern region of Wanzhou District. The approach integrates information on landslide features with multi-temporal interferometric SAR (MT-InSAR) displacement monitoring and deep learning techniques to assess and predict building vulnerability. Key challenges addressed include the integration of Long Short-Term Memory (LSTM) networks optimized through Bayesian methods to enhance predictive accuracy, and the application of the Rolling Time Series Forecast Distribution Analysis (RTSFDA) method to quantify prediction uncertainties. The methodology focuses on slope units for vulnerability zoning, which is crucial for local disaster management and effective identification of high-risk areas, especially along reservoir banks. Scenario-based forecasting, incorporating optimistic, moderate, and pessimistic scenarios, provides a thorough assessment of potential future risks, facilitating long-term planning and emergency preparedness. This research demonstrates the effectiveness of combining background ancillary data with earth observation technologies and deep learning methods, offering valuable insights and tools for enhancing landslide risk management for future scenarios.