Abstract. The aim of the study is to confirm the importance of discriminate different types of slope movements for a better landslide susceptibility evaluation. The study was applied to the sample area of Calhandriz (11.3 km2) in the area North of Lisbon. Sixty shallow translational slides, 23 deeper translational movements and 19 rotational movements were selected for statistical analysis. Landslide susceptibility assessment was achieved using a data-driven approach: the Information Value Method (Yin and Yan, 1988). The method was applied both to the total set of considered landslides and to each type of slope movement, and the obtained success rates for the highest susceptibility classes are higher in the latter case. The different types of landslides are not equally conditioned by the considered instability factors. Information scores are higher for lithology, concordance between slope aspect and dip of the strata, and slope angle, respectively, for rotational movements, translational movements and shallow translational slides. The information value of the variables "presence of artificial cut (roads)" and "presence of fluvial channel" is systematically high for the three types of slope movement, pointing out the importance of both anthropogenic influence and bank erosion on slope instability in the study area. Different types of landslides have neither the same magnitude nor equal damaging potential. Furthermore, technical strategies to mitigate landsliding also depend on landslide typology. These are additional reasons to discriminate between different types of slope movements when assessing landslide susceptibility and hazard.
Using the rotation and random forest models of ensemble learning to predict landslide susceptibility
