IIT Mandi's AI Early Warning System Could Predict Landslides Hours Before

At the heart of this innovation is a low-cost, AI-powered early warning system designed to detect landslide risks before they occur.

Developed by a team of professors and students, the system combines sensor networks with machine learning models to monitor soil conditions in real time and predict potential slope failures hours in advance.

"Climate change is clearly altering rainfall patterns," said Kala Venkata Uday, Associate Professor of Geotechnical Engineering at IIT Mandi.

"Earlier, rainfall used to be light and spread over a longer duration, allowing water to seep into the soil gradually. Now, we are witnessing intense rainfall over short periods. The water doesn't get enough time to infiltrate, increasing the likelihood of landslides," he said.

The system--known as the Flume Test Bed Setup--uses sensors to track parameters such as soil moisture, temperature, and micro-movements in the earth.

These inputs are processed through AI algorithms trained on historical and real-time data, enabling predictions with over 90 per cent accuracy. Crucially, it can issue alerts up to three hours before a potential landslide.

"We have designed a model that analyses both current activity and past events," explained Varun Dutt, Dean at IIT Mandi. "It can forecast whether any ground movement is likely in the next 10, 20, or 30 minutes--or even one to three hours. Essentially, it is an embedded system powered by IoT and AI."

What sets this innovation apart is not just its accuracy, but its affordability. While conventional early warning systems often come with high deployment costs, IIT Mandi's model is designed to be both scalable and cost-effective--making it viable for widespread use in vulnerable regions.

Currently, the system has been deployed across more than 60 high-risk locations in Himachal Pradesh, where it continuously collects data and sends real-time alerts to authorities and local communities.

Students involved in the project say the system goes beyond immediate alerts. "We have used AI as a tool to simulate rainfall scenarios based on past data," said Tanvi, a student researcher.

"This helps us predict future events and even map landslide risk zones for 2050, 2070, and beyond, taking into account changes in land use and land cover," Tanvi said.

Her colleague, PhD scholar Shubham Kumar, emphasised the importance of ground-level analysis.

"Once we identify a landslide-prone area, we prioritise it for detailed study. We collect soil samples, analyse their characteristics, and study movement patterns. Based on this, we decide where to install the system for maximum effectiveness," Kumar said.

In a complementary effort, some students are also developing virtual reality tools that simulate disaster scenarios such as floods. These immersive experiences aim to prepare communities for emergencies, enabling quicker and more informed decision-making.

For residents of mountainous regions, where minutes can mean the difference between safety and disaster, such innovations could prove transformative. By bridging cutting-edge research with real-world application, IIT Mandi's initiative reflects a broader shift--where technology is not just advancing knowledge, but actively safeguarding lives.

- ANI