Development of an AI analysis model that automatically extracts dangerous slopes - Presented at the Integrated Physical Model Technology Research Association 2025 Research Presentation -
PACIFIC CONSULTANTS CO., LTD. (Address: Chiyoda-ku, Tokyo; Representative Director, President Executive Officer: Osamu OMOTO) has developed a technology that uses DEM (Digital Elevation Model) and AI to automatically extract dangerous slopes in order to predict the risk of slope collapse caused by heavy rain disasters, which are becoming more severe and frequent. This technology was presented at the 2025 Research Presentation Meeting (November 19, 2025) hosted by the Integrated Physical Model Technology Research Association.
■ Development background and purpose
Due to the effects of climate change in recent years, there has been an increase in heavy rain disasters, and localized heavy rain caused by linear precipitation bands is becoming more frequent. As a result, the number of landslides is on the rise, and in order to ensure the safety of residents and the continuation of business activities, not only the national and local governments but also private companies are required to quickly identify the risk of slope collapse and take prompt measures. However, conventional visual interpretation methods are time-consuming and labor-intensive, and have the problem of being unable to adequately respond to the ever-increasing disaster risk.
To address this issue, we have developed an AI analysis model based on our long-standing expertise in predicting slope failures. By analyzing topographical data and automatically extracting slopes at high risk of failure, we have made it possible to assess disaster risk more quickly and over a wider area.
■ Overview
This technology replaces the conventional method of visually interpreting slopes created by expert engineers using GIS (geographic information systems) to identify slopes at high risk of surface collapse by modeling them using DEM (digital elevation model) and AI to automatically identify dangerous slopes. The technology was applied to the heavy rains that occurred in northern Kyushu in July 2017 (around Asakura City, Fukuoka Prefecture), and by comparing the actual slope collapses with the predictions made by the AI analysis model, the applicability, challenges, and future prospects of the AI analysis model were examined. The results indicated that the information could be fully utilized to screen the risk of collapses in advance of disaster prevention planning and to determine the installation of disaster prevention facilities.
A diagram showing the correspondence between the collapse determination results (blue frame) based on AI analysis and the actual collapsed locations (red points)
■ Future outlook
Going forward, we will use this technology to assess the risk of landslides across Japan. We will conduct analyses that take into account the topographical characteristics of each region, and work to improve the accuracy of our AI analysis model, contributing to the realization of more practical disaster prevention measures.