Abstract:

We find ourselves in an ever-evolving environment, where the past fifty years have marked a pivotal moment in scientific thinking. This shift is particularly evident in the scientific approach to analyzing natural-induced hazards. Geohazards annually contribute significantly to loss of life and property, with mass movements standing out as widespread occurrences globally. The study of extreme events and their repercussions on landscape stability is a critical area in environmental research. In this paper, we showcase the advancements in the integration of Artificial Intelligence (AI) and Remote Sensing (RS) for improved landslide assessments, leveraging developments in Earth Observation (EO) data analysis. We highlight the application of Object-Based Image Analysis (OBIA), which have not traditionally been tailored for landslide studies but have proven effective in this context. The framework enables the translation of complex real-world landslide scenarios into analyzable objects through segmentation algorithms, applying subsequent classifications via rule-based or advanced Machine Learning (ML) algorithms. We demonstrate how ML has the potential to revolutionize geoscience data analysis and address major societal concerns presented by landslide hazards by tapping into the vast reserves of geoscience data. ML algorithms, particularly Random Forest (RF), integrated into an Object-Based Image Analysis (OBIA) workflow, demonstrated adaptability for sub-zone landslide mapping on a local scale. Given the increasing frequency of extreme meteorological events driven by climate change, the integration of UAV datasets, Structure from Motion (SfM), and advancements in OBIA and AI can respond effectively by enabling precise and accurate analysis of landslide and rockfall failures. Our results affirm that rotational landslides and their thematic sub-zones were adequately recognized and mapped through the ML procedure.