Dynamic landslide susceptibility mapping based on the PS-InSAR deformation intensity

Area of Science:

• Geosciences
• Remote Sensing
• Machine Learning

Background:

• Refined landslide risk management requires advanced dynamic susceptibility modeling.
• The Wanzhou channel in the Three Gorges Reservoir Area, a vital transport link, faces significant landslide risks.
• Existing models often overlook crucial dynamic factors influencing landslide occurrence.

Purpose of the Study:

• To explore an extended correlation framework for dynamic landslide susceptibility modeling.
• To develop and validate a dynamic landslide susceptibility mapping (DLSM) approach.
• To assess the impact of incorporating dynamic deformation factors on predictive accuracy.

Main Methods:

• Utilized five machine learning models: logistic regression (LR), multilayer perceptron neural network (MLPNN), support vector machine (SVM), random forest (RF), and decision tree (DT).
• Employed PS-InSAR technology to derive deformation intensity as a dynamic factor.
• Integrated dynamic factors with static factors (topography, geology, hydrology, human activities) for DLSM.

Main Results:

• The decision tree model achieved the highest AUC value of 93.1%, a 2% increase when dynamic factors were included.
• Incorporating dynamic factors improved the predictive accuracy of several machine learning models.
• DLSM results demonstrated better alignment with the actual spatial distribution of landslides.

Conclusions:

• The proposed DLSM framework offers a significant improvement over static models for landslide susceptibility assessment.
• Integrating dynamic deformation intensity enhances the reliability and accuracy of landslide risk management.
• This research provides a valuable reference for dynamic landslide disaster management and prevention strategies.