Development of an AI-based landslide prediction model and nationwide early warning framework

As climate change intensifies, the frequency and magnitude of extreme rainfall events are increasing, leading to more frequent landslides and debris flows. In South Korea, where approximately 63% of the land area is mountainous, landslide damage occurs repeatedly because of complex topographic conditions. This highlights the need for spatially detailed and timely landslide early warning information to support climate change adaptation. However, existing landslide early warning systems primarily rely on meteorological data and provide warnings at the administrative-unit scale, limiting their applicability to local-level prevention and preparedness. This technical note presents the conceptual design and implementation of a nationwide landslide early warning framework developed as a non-structural measure for proactive landslide prevention and preparedness. The proposed framework integrates meteorological observations and forecasts from the Korea Meteorological Administration (KMA), national geospatial datasets, and time-series optical satellite imagery. Landslide susceptibility is calculated by combining an AI-based, data-driven landslide susceptibility model with a physically based infinite slope stability model, which are independently constructed and subsequently linked using rule-based flow directions. In addition, an anthropogenic activity map derived from satellite-based time-series land-cover change information is incorporated to reflect changes in surface conditions during the early warning process. The framework is designed to automatically generate landslide early warning information across the entire country at a spatial resolution of 100 m × 100 m and a temporal resolution of 3 hours. The resulting outputs are regularly disseminated through a web-based service. The proposed framework provides a practical example of a climate change adaptation technology that supports landslide prevention and preparedness under increasingly frequent extreme rainfall conditions.