GIS-based landslide susceptibility mapping in the Longmen Mountain area (China) using three different machine...

Area of Science:

• Geosciences
• Environmental Science
• Data Science

Background:

• Landslides pose significant global risks to safety and socio-economic stability.
• Accurate landslide modeling and prediction are crucial for effective disaster management and prevention.

Purpose of the Study:

• To assess landslide susceptibility and map landslide risk in the Longmen Mountain area using advanced machine learning techniques.
• To compare the performance of Random Forest (RF), Support Vector Machine (SVM), and Decision Tree (DT) algorithms in landslide prediction.

Main Methods:

• Applied the frequency ratio (FR) method in conjunction with RF, SVM, and DT regression algorithms.
• Utilized 7774 historical landslide and non-landslide points, balanced for training and testing.
• Analyzed influencing factors through multicollinearity analysis and FR method, normalizing environmental factors for improved model performance.

Main Results:

• The Random Forest (RF) model achieved the highest predictive performance with an Area Under the Curve (AUC) of 0.82, followed by SVM (AUC = 0.8) and DT (AUC = 0.69).
• Generated a landslide susceptibility map for the Longmen Mountain region, identifying high-risk areas.
• Validated the effectiveness of machine learning models combined with the FR method for accurate landslide susceptibility assessment.

Conclusions:

• Machine learning models, particularly RF, significantly enhance the accuracy and performance of landslide susceptibility assessment.
• The developed predictive maps offer critical support for disaster prevention, safeguarding lives and property.
• The FR-based machine learning approach is a robust methodology applicable to landslide research and other scientific fields.