SAFE-RISCCS — Nepal's Early-Stage AI Landslide Early-Warning Pilot

Nepal is exceptionally exposed to landslides: government estimates put about 59% of the country's land as landslide-prone, and more than 80% of it lies on a slope. The 2015 Gorkha earthquake alone triggered roughly 45,000 separate landslides. Nepal's National Disaster Risk Reduction and Management Authority (NDRRMA) has been building a national Landslide Early Warning System with support from Australia's Department of Foreign Affairs and Trade.

Within that effort, researchers from the University of Melbourne (led by Professor Antoinette Tordesillas), Tribhuvan University, and partners in Britain and Italy are testing SAFE-RISCCS ("Spatiotemporal Analytics, Forecasting and Estimation of Risks from Climate Change Systems"), an AI system designed to fuse rainfall gauges, ground-movement sensors, local resident observations and satellite imagery from NASA, the European Space Agency and Japan's JAXA to flag landslide risk up to weeks in advance. It is being rolled out at two high-risk pilot sites: Kimtang in Nuwakot district and Jyotinagar in Dhading district.

As of mid-2025, sensors at each site had been collecting data for only about two months, and the project team says a dataset spanning a year or two is needed before the AI component can reliably generate automated graphical alerts from rainfall forecasts. No accuracy figures, false-alarm rates, lead-time validation or lives-saved estimates have been published for SAFE-RISCCS, and no independent evaluation of the system yet exists.

This is, honestly, an early pilot rather than a proven early-warning tool: its case for inclusion rests on a credible government-anchored institutional setup (NDRRMA as implementing partner, bilateral Australian government funding, an international university consortium) rather than on demonstrated forecasting performance. It is included here as a transparent, cautionary example of AI adoption in progress — worth tracking as the data record lengthens, not yet worth relying on.

Read the full analysis: https://phys.org/news/2025-08-landslide-prone-nepal-ai-powered.html