Spatial forecast of landslides in three gorges based on spatial data mining.

Area of Science:

• Geosciences
• Remote Sensing
• Artificial Intelligence

Background:

• The Three Gorges region faces significant landslide risks due to its complex terrain and dense population.
• Frequent landslide disasters pose a tremendous potential threat to the area.

Purpose of the Study:

• To investigate spatial landslide forecasting in the Three Gorges.
• To establish criteria for landslide prediction using multiple factors.
• To perform intelligent spatial landslide forecasts for Guojiaba Town.

Main Methods:

• Utilized China-Brazil Earth Resources Satellite (Cbers) images.
• Established 20 forecast factors including spectra, texture, vegetation coverage, and hydrological data.
• Employed the C4.5 decision tree algorithm for landslide criteria mining and forecasting.
• Compared the proposed method with seven other landslide prediction models.

Main Results:

• The developed method successfully identified dangerous and unstable regions for landslides.
• The spatial landslide forecasts for Guojiaba Town were accurate.
• The proposed method demonstrated significantly higher forecast precision compared to IsoData, K-Means, Mahalanobis Distance, Maximum Likelihood, Minimum Distance, Parallelepiped, and Information Content Model.

Conclusions:

• The C4.5 decision tree approach, integrating Cbers imagery and multiple factors, is effective for spatial landslide forecasting.
• The method provides a valuable tool for risk assessment and mitigation in landslide-prone areas like the Three Gorges.
• The study highlights the superiority of the proposed intelligent forecasting method over traditional techniques.