Abstract
This paper presents a strategy for estimating latent hazardous areas affected by slope failures with SEM (Structural equation modeling)-based quantitative model with "normative types of training data sets." The quantitative model for producing susceptibility maps of slope failures elucidates the relationship between past occurrences of slope failures (i.e. P_type: past occurrence type of training data sets) and causal factors. The following knowledge-based training data sets (i.e. N_type: normative type) are introduced to the model; i)convex ridge topography, ii)convex plateau-formed topography, iii)concave single hillformed topography, iv)concave many hills-formed topography and v) concave gentle slope topography. As a final outcome, the differences between susceptibility maps produced by using P_type and N_types of training data sets, respectively, are delineated on difference maps which are effective in estimating latent hazardous areas affected by land characteristics unlike those of past occurrences of slope failures.
