AI Now Predicts Landslides Before They Strike

When mathematician Antoinette Tordesillas pulled up satellite images of Kimtang village in central Nepal, the red warning on her screen told a frightening story: hundreds of villagers were living directly on top of land ready to slide.

The cracks in concrete steps and trees growing at odd angles had been subtle clues. But artificial intelligence analyzing radar satellite data revealed what human eyes couldn't see: the ground was slowly shifting beneath their feet, possibly headed toward a devastating collapse.

Landslides kill thousands of people worldwide each year and cause billions in damage. They've long seemed impossible to predict, striking suddenly and without warning. But new AI systems are changing that equation entirely.

Scientists like Tordesillas are teaching machine learning algorithms to spot the tiny warning signs that appear days, weeks, or even years before a slope gives way. Radar satellites capture images showing ground movement as small as millimeters per year. The particles of earth begin separating from each other like "dancers following some unwritten choreography," as Tordesillas describes it.

The technology is already making a difference. In the UK, researchers used AI to analyze 300,000 slopes across Great Britain and found 3,000 that are actively moving. The work would have taken humans months or years to complete manually, but AI finished it in hours.

In Nepal, the timing proved crucial. Kimtang's residents had already been relocated once in 2019 after a nearby landslide. The AI revealed their new home sat on the most unstable ground in the entire region. Armed with this information, Tordesillas and her team worked with villagers to develop ground-level monitoring systems and plan evacuation routes.

The analysis even identified safe zones. The village high school sits on one of the most stable spots, making it an ideal emergency gathering point.

The Ripple Effect

This breakthrough extends far beyond individual villages. Climate change and human activities like construction and mining are making landslides more common. In the US alone, they kill up to 50 people annually and cost billions in damage.

The AI systems can monitor vast regions continuously, something impossible for human analysts. British researchers found that slow ground movements are affecting 8,700 miles of roads and 220 miles of railway across their country. Early warnings could prevent infrastructure damage and save lives.

Alessandro Novellino, a geoscientist at the British Geological Survey, calls it "enabling new science that we couldn't do in the past." The technology works because scientists guide the AI using their deep knowledge of how slopes fail, making it a specialized tool rather than a general algorithm prone to errors.

As of January 2025, no landslide has struck Kimtang. The villagers now have something precious: time to prepare, knowledge of the risk, and a plan for staying safe.