On two implicit issues in prediction modeling of landslide susceptibility

Abstract—this contribution argues the interpretation of prediction
rate curves of landslide susceptibility and the corresponding 
prediction-pattern uncertainties. The focus is on two main issues in 
any kind of spatial prediction modeling: (1) isolating the meaningful 
parts of prediction-rate curves from a cost-benefit point of view and 
(2) comparing the qualities of prediction patterns obtained by 
different mathematical models and/or dissimilar spatial evidences. 
Mathematical models, methods, and databases generated prediction 
maps (we prefer the term prediction patterns) hard to evaluate due 
to inevitable relativity of measures, representations and confidence. 
This is a major problem, with modelling assumptions and 
justifications mostly ignored or poorly discussed. We consider 
prediction-rate curves, obtained by cross-validation, as standardized 
procedure. The curves are the result of cross-validating prediction 
patterns with the distribution of occurrences more recent than the 
ones used to generate the patterns. The two issues concern all types 
of modeling independently of algorithmic complexity or database 
formats. We examine an application example and its analytical 
strategy to point at resolving problems of pattern evaluation, 
comparison and uncertainty measure. The need becomes evident of 
collaborative efforts towards solutions analyzing a common multi
format database.