Pre-failure topography implementation to predict landslides using a bivariate statistical model

Landslides are natural phenomena that cause significant socioeconomic and environmental impacts in mountainous regions. Statistical models used to predict landslides frequently use Digital Terrain Models (DMTs) to identify scars and to generate thematic maps representing relevant causative factors (e.g., slope, aspect, curvature).The topographical causative factors tell us how some morphometrical parameters control slope stability and the algebraical combination of weighted causative factors (the landslide susceptibility map) displays how the global relationship of the causative factors generates the landslides. However, these DTMs will no longer be representative of the topographical features that triggered landslides when obtained after the occurrence of this events (post-failure DTM) and using archetypal morphometric signatures of past landslides in statistical models will imply relevant conceptual mistakes. A possible solution to this problem is to assume that the pre-rupture topography can be inferred from undisturbed areas adjacent to scars. This work presents a topography reconstruction method using LIDAR elevation points to generate a pre-failure topography DTM from a post-failure topography DTM. The pre-failure topography was used in a bivariate statistical model (Weights of Evidence) to predict landslides in the Quitite and Papagaio basins, in the city of Rio de Janeiro (Brazil). Seven landslide susceptibility models were produced by combining eight conditionally independent causative factors and had their predictive capacity tested by calculating the area under curve (AUC). The final model (AAC = 0.77) highlights the direct topographic and hydrological controls and the indirect lithological and structural controls on the landslides. Landslides are mainly controlled by slopes between 26° and 52°, on North, Northeast and Northwest facing slopes, on concave curvatures with values of a contribution area between 1.8m² e 4.1m². The results take into account the model’s assumptions and provide a synthesized and robust view of the prone landslides areas in an environment of great geodiversity.