Recipient organisationUniversity of ExeterSource-published name: University of Exeter
Funding£192K
PeriodFeb 2025 — Jan 2027
In plain English
AI plain-English summary
Landslides, sinkholes, and ground collapses kill thousands of people each year, yet current satellite and sensor data is too fragmented and slow to give reliable early warnings. This project tackles a core problem: existing AI models for spotting geological hazards are brittle. They work well on the data they were trained on but fail when faced with new terrain or different sensor types. The researchers will fuse multiple types of remote-sensing data—radar, optical images, ground-motion monitors—into a single high-precision picture. They then plan to inject expert geological knowledge directly into the AI, so the model understands that a crack near a road behaves differently from a crack on a bare hillside. Finally, they will build a system that tracks how a disaster evolves over time, flagging subtle precursor signals before a catastrophic failure. If successful, the work could transform how infrastructure agencies and emergency responders monitor unstable slopes, mining areas, and coastal cliffs. Instead of reacting after a collapse, they could receive automated, interpretable warnings days or weeks in advance. The approach is designed to generalise across different landscapes, meaning it could be deployed globally without retraining from scratch.
View original technical description
Geological disasters have become one of the most significant environmental threats, creating risks to human life and economic development. It is therefore imperative to conduct fine-grained precise disaster monitoring and early warning to guide disaster prevention and emergency rescue strategies, which can greatly contribute to saving lives, reducing economic losses, and safeguarding communities. However, several critical challenges, including inefficient multi-dimensional geological environment remote sensing (GERS) data fusion, low generalization of data-driven AI models, and inconclusive disaster evolution patterns, remarkably limit the precise identification and warning of geological disasters from GERS data. To address these challenges, this project will comprehensively integrate multi-modal GERS data and cutting-edge AI technologies to create an innovative knowledge-driven disaster identification and evolution associative analysis methodology with high accuracy, interpretability and reliability. Specifically, 1) a novel multi-modal GERS data fusion method will be developed to effectively ensemble remote sensing images and monitoring data to obtain high-precision fused data; 2) An original knowledge-driven interpretation scheme will be designed to combine the spatial relationships and domain expert knowledge of geological environment in order to improve the precision and generalization of disaster and geological context identification; 3) A disaster evolution analysis mechanism will be developed to recognize the evolution process of typical geological disasters and warn early disaster signs; 4) A creative knowledge-driven method with high accuracy, interpretability and reliability will be developed for disaster identification and evolution analysis. This project will revolutionize the way disasters are predicted and monitored, which empowers more swift responses to the geological threats, significantly reducing the devastation caused by geological disasters.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know