Uncertainty of class membership in spatial prediction modeling: follow-up study to an application to complex landslides

A previously studied spatial database for representing landslide susceptibility is reanalyzed to bring out aspects of relative quality and uncertainty of the resulting prediction pattern. Using the empirical likelihood ratio function, a processing strategy, based on cross-validation, target pattern and prediction rates, is used to asses the suitability of the database for prediction modeling. Sharing of databases is seen as fundamental for discussing and understanding prediction modeling.