Global Landslide Susceptibility Ensemble at 36-km Resolution

2500 ensemble members of global landslide susceptibility maps were created using blocked random cross-validation and predictor variable perturbation. Each member is derived from a logistic regression equation with 5 predictor variables, which vary between members. The dataset, provided by Anne Felsberg of KU Leuven, is based on research published in Natural Hazards and Earth System Sciences in 2022.

Use Cases

• Assessing global landslide risk based on ensemble susceptibility maps.
• Quantifying prediction uncertainty in geospatial models based on the provided standard deviation maps.
• Comparing logistic regression model performance based on the ensemble of 2500 members with varying predictor variables.
• Studying the impact of variable selection on landslide susceptibility estimates based on the described data-driven predictor selection methodology.

Strengths

• Contains 2500 ensemble members, allowing for robust uncertainty quantification.
• Provides both ensemble average and standard deviation maps for analysis.
• Based on a peer-reviewed methodology published in a scientific journal.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Spatial resolution is 36 km, which may be too coarse for local-scale analysis.
• Last update date is unknown; freshness unverified.

Provenance

Source

Felsberg, A., Poesen, J., Bechtold, M., Vanmaercke, M., and De Lannoy, G. J. M. (2022): Estimating global landslide susceptibility and its uncertainty through ensemble modelling, Natural Hazards and Earth System Sciences.

Collection Method

Combination of blocked random cross validation and predictor variable perturbation.

Geography

Global