A Novel Rule-based Approach In Mapping Landslide Susceptibility

Area of Science:

• Geosciences
• Environmental Science
• Geographic Information Systems

Background:

• Existing landslide susceptibility mapping (LSM) methods lack transparency and exhibit subjectivity.
• Understanding landslide-conditioning factors at the local scale is crucial but often hindered by opaque models.
• High subjectivity in re-classifying susceptibility scores reduces the quality of LSM products.

Purpose of the Study:

• To develop and validate a novel, transparent rule-based system for landslide susceptibility mapping (LSM).
• To improve the understanding of landslide-conditioning factors and reduce subjectivity in susceptibility classification.
• To assess the reliability and performance of the proposed LSM method at the pixel-level.

Main Methods:

• A rule-based system was developed, relating landslide-conditioning factors without logical operators.
• Shannon entropy was used to determine factor priorities and rule uncertainties.
• Rule-level uncertainties were mapped for pixel-level reliability assessment, followed by If-Then rule application for classification.

Main Results:

• The proposed rule-based LSM method demonstrated high performance with an Area Under the Curve (AUC) of 0.934 in a case study.
• The approach provided transparency in identifying factor priorities and states/values for each pixel.
• Uncertainties associated with susceptibility rules were accessible, interpretable, and replicable.

Conclusions:

• The novel rule-based system offers a transparent and less subjective alternative for landslide susceptibility mapping.
• The method enhances insights into LSM processes by clearly defining rules and uncertainties.
• This approach is beneficial for deriving interpretable and reliable landslide susceptibility information.