IIT Mandi develops AI-based real-time landslide early warning system

A team of scientists from the Indian Institute of Technology Mandi has developed a real-time landslide monitoring and early warning system aimed at reducing loss of life and infrastructure damage in landslide-prone regions.

The system has been developed in collaboration with the National Mission on Himalayan Studies and other partner agencies. It uses a network of sensors installed at vulnerable sites to transmit real-time data and can detect ground movement of less than one millimetre.

Once even minute movement is detected, the system is automatically activated. Warning mechanisms such as hooters and blinkers are triggered in vulnerable zones, while alerts are simultaneously sent to authorities to enable rapid response.

According to Varun Dutt, Professor , IIT Mandi,, "If there is a movement of even less than one millimetre on a hill, the system gets activated. Once triggered, it sends alerts to 'warning poles' installed in vulnerable areas such as valleys or along roads below the hill – it sets off a hooter and blinker, alerting people about the possibility of a landslide. Additionally, alerts are also sent through SMS, and a web App, an Android-based application to the District Disaster Management Authority (DDMA) control room. From there, the information is also passed on to the police, the NDRF, and other concerned agencies.."

Beyond ground-based sensors, the system also integrates satellite imagery to monitor larger-scale terrain movement over time, improving the accuracy of risk assessment.

He further adds that, "Our system works in real time, but satellite data provides information of a very high level, with a resolution of about 20 meters by 20 meters, while, our system has millimetre-level resolution. So, while satellite data helps us understand whether there was any movement in a particular area over a period of 14 days to two months, our system can detect much finer changes.

The Government of India has now launched the NISAR mission, and its data will gradually become available. As this data becomes accessible, the warnings we generate through satellite inputs will further improve. At present, satellite data is received once every two weeks, and we are integrating that data with our system's data so that we can issue more accurate and timely warnings."

The technology has already been deployed at three landslide-prone locations in Mandi district, where it is providing continuous monitoring and early warning alerts.

Researchers say the model, which uses machine learning and advanced analytics, has achieved over 90 percent accuracy in predicting landslides, marking a significant step forward in disaster preparedness and mitigation in the Himalayan region.