Researchers from the University of Cambridge are employing artificial intelligence to expedite landslide detection after major earthquakes and extreme rainfall. This advancement aims to provide crucial time for coordinating relief efforts and minimizing humanitarian impacts.
On April 3, 2024, a magnitude 7.4 earthquake struck Taiwan's eastern coast, marking the strongest quake in the region in 25 years. Although strict building codes protected most structures, landslides severely impacted mountainous and remote villages.
In large and inaccessible disaster areas, responders often rely on satellite images to identify affected regions and prioritize relief efforts. However, mapping landslides manually from satellite imagery is time-consuming. "In the aftermath of a disaster, time really matters," said Lorenzo Nava from Cambridge’s Departments of Earth Sciences and Geography. Using AI technology, Nava identified 7,000 landslides within three hours after acquiring satellite imagery following the Taiwan earthquake.
Nava has been refining his AI method with an international team since the Taiwan incident. They utilize various satellite technologies, including those capable of penetrating clouds and capturing images at night, to improve AI's ability to detect landslides.
Landslides triggered by earthquakes or intense rainfall can be exacerbated by human activities such as deforestation and construction on unstable slopes. In certain conditions, they can lead to additional hazards like debris flows or severe flooding.
Nava’s research is part of a broader initiative at Cambridge focused on understanding how landslides can trigger cascading 'multihazard' events. The CoMHaz group, led by Professor Maximillian Van Wyk de Vries from the Departments of Geography and Earth Sciences, uses satellite imagery, computer modeling, and fieldwork to locate landslides and predict their occurrence.
The researchers are also collaborating with communities to increase awareness about landslides. In Nepal, Nava teamed up with local scientists and the Climate and Disaster Resilience in Nepal consortium to pilot an early warning system for Butwal.
Nava is training AI systems to recognize landslides using optical images of the ground surface and radar data that can penetrate cloud cover. Radar images present challenges due to their greyscale depiction of surface properties but are suitable for AI-assisted analysis that identifies features otherwise unnoticed.
By combining radar's capabilities with optical image fidelity, Nava aims to develop an AI model that detects landslides accurately even under poor weather conditions. His trial after the Taiwan earthquake showed potential by detecting thousands of otherwise unnoticed landslides beneath cloud cover.
However, Nava acknowledges that more work is needed to enhance both accuracy and transparency in AI models. He emphasizes building trust in these models so decision-makers can interpret outputs effectively. "AI can feel like a black box," he noted. "It's important for end users to evaluate AI-generated information quality before making critical decisions."
This effort includes collaboration with organizations like the European Space Agency (ESA), World Meteorological Organization (WMO), International Telecommunication Union’s AI for Good Foundation, and Global Initiative on Resilience to Natural Hazards through AI Solutions.
At a recent ESA meeting in Italy, researchers launched a data-science challenge seeking help from the coding community to improve model functionality while incorporating features explaining its reasoning visually—potentially through maps showing image likelihoods containing landslides—to assist end-user comprehension.
"In high-stakes scenarios like disaster response," Nava stated, "trust in AI-generated results is crucial." The challenge aims at enhancing transparency within model decision-making processes empowering ground-level decision-makers confidently act swiftly."
Reference: Lorenzo Nava et al., 'Brief Communication: AI-driven rapid landslide mapping following the 2024 Hualien City Earthquake in Taiwan,' Natural Hazards and Earth System Sciences (2025). DOI: 10.5194/nhess-25-2371-2025