Landslide susceptibility mapping using an entropy index-based negative sample selection strategy: A case study of...

Area of Science:

• Geosciences
• Environmental Science
• Geology

Background:

• Landslides pose a significant geological hazard in high-altitude regions like the Himalayas, China.
• Challenging environmental conditions in these areas impede traditional field surveys for landslide assessment.
• Accurate landslide susceptibility mapping is crucial for disaster prevention and land use planning.

Purpose of the Study:

• To develop and validate an improved landslide susceptibility assessment model for Luolong County, Tibet.
• To enhance the accuracy of landslide prediction by optimizing non-landslide sample selection using the IOE model.
• To compare the performance of different machine learning models integrated with the IOE model for landslide susceptibility mapping.

Main Methods:

• Utilized Google Earth satellite imagery to create a landslide database of 2517 debris occurrences.
• Identified twelve conditioning factors including geology, topography, meteorology, hydrology, vegetation, soil, and human activities.
• Integrated the Information of Overlapping Elements (IOE) model with Support Vector Classification (SVC), Multilayer Perceptron (MLP), Linear Discriminant Analysis (LDA), and Logistic Regression (LR) models.

Main Results:

• Optimizing non-landslide samples with the IOE model significantly improved the performance (AUC, accuracy, precision, F1 score) of all coupled machine learning models.
• The IOE-MLP model demonstrated the highest performance, with an AUC increase from 0.8172 to 0.9747.
• Land use, elevation, and slope were identified as the predominant landslide-controlling factors in the study area.

Conclusions:

• The integration of the IOE model with machine learning provides an effective method for landslide susceptibility assessment, particularly in data-scarce, high-altitude regions.
• The IOE-MLP model offers superior predictive accuracy and classification performance for identifying high-risk landslide zones.
• The findings provide valuable data for regional disaster prevention, mitigation strategies, and sustainable land use planning in similar geological settings.