An improved algorithm for identifying shallow and deep-seated landslides in dense tropical forest from airborne laser scanning data

Landslides are natural disasters that cause environmental and infrastructure damage worldwide. They are difficult to be recognized, particularly in densely vegetated regions of the tropical forest areas. Consequently, an accurate inventory map is required to analyze landslides susceptibility, hazard, and risk. Several studies were done to differentiate between different types of landslide (i.e. shallow and deep-seated); however, none of them utilized any feature selection techniques. Thus, in this study, three feature selection techniques were used (i.e. correlation-based feature selection (CFS), random forest (RF), and ant colony optimization (ACO)). A fuzzy based segmentation parameter (FbSP optimizer) was used to optimize the segmentation parameters. Random forest (RF) was used to evaluate the performance of each feature selection algorithms. The overall accuracies of the RF classifier revealed that CFS algorithm exhibited higher ranks in differentiation landslide types. Moreover, the results of the transferability showed that this method is easy, accurate, and highly suitable for differentiating between types of landslides (shallow and deep-seated). In summary, the study recommends that the outlined approaches are significant to improve in distinguishing between shallow and deep-seated landslide in the tropical areas, such as; Malaysia.