The british geological survey has developed a new semi-automated method that combines artificial intelligence and satellite data to detect actively moving slopes that could pose a landslide risk across great britain. Working with the university of florence in italy, researchers built a workflow that applies a type of artificial intelligence called machine learning alongside clustering tools to mine large volumes of interferometric synthetic aperture radar, or insar, data. The approach is designed to overcome the difficulty and cost of traditional landslide monitoring at national scale, offering a way to flag unstable slopes that warrant closer investigation.
Previously, the british geological survey relied on insar for landslide monitoring, benefiting from its extensive data coverage but facing major analytical challenges due to data volume and complexity. In the new study, the semi-automated analysis highlighted around 3000 slopes that showed consistent movement of over 2.5 mm per year between 2018 and 2022. These actively moving slopes affect approximately 14 000 km of road and 360 km of railway – 2.4 per cent and 1 per cent of the entire national network, respectively. The tool has helped to classify more than 300 000 slopes around the uk and has highlighted 3000 slopes that have moved in a four-year period, providing landslide specialists with a systematic starting point for further work.
Researchers stress that the unstable slopes detected by the method are not all confirmed landslides but represent priority areas for detailed mapping, field surveys and assessment of impacts on infrastructure such as roads, railways and buildings. By automatically identifying the most critical areas for ground motion, the method is presented as a practical disaster management tool that supports smarter prioritisation of maintenance and mitigation, reducing costs and improving safety. Next steps will focus on refining the analysis using more detailed topographical data so that it can move from identifying unstable slopes to automatically mapping individual landslides, including their types, extents and likely triggers. The results are intended to be shared with local authorities, infrastructure owners and the natural hazards partnership, and the full research is described in the paper titled “machine learning and clustering for supporting the identification of active landslides at national scale.”
