Exploring the Gap between AI-Derived Landslide Risk, Risk Preparedness, and Risk Communication: A Comparative Study from the Indian Himalayas

The Geospatial Artificial Intelligence (Geo AI) drastically enhances the capacity to estimate landslide risk. Within the Geo AI platform, numerous machine learning algorithms are being developed to achieve optimal prediction accuracy. These techniques allow for a top-to-bottom approach, wherein scientific perspectives are reflected in risk mapping. However, existing studies rarely explore how community preparedness aligns with this scientific risk mapping. Therefore, this study aims to compare AI-derived risk estimation with community preparedness. Additionally, it investigates the gaps in disaster risk communication. The study is conducted through a case study in the Eastern Himalaya of India, which is one of the most landslide-prone regions in the world. Climate change, earthquakes, rapid population growth, and unplanned infrastructure development are cumulatively generating the risk of fatal disasters in the region. Overall, the methodological framework consists of three steps. First, we develop a landslide risk model using GIS and machine learning algorithms such as deep learning. In the second step, community preparedness is assessed. Risk preparedness data is collected through a GPS-based field survey involving 200 households. The next step compares risk estimation with risk preparedness. Furthermore, the study analyzes the effectiveness of disaster risk communication in addressing the identified gaps. The findings of the study indicate that risk preparedness is complex among communities and varies at the local level, even though the degree of risk is almost the same. The study recommends improving risk communication to bridge the gap between AI-defined predicted risk and community preparedness.