IIT Guwahati develops predictive model to identify glacial lake formation in Himalayas

Guwahati, Feb 8: Researchers from the Indian Institute of Technology (IIT) Guwahati have developed a predictive framework to identify locations where glacial lakes are likely to form in the Eastern Himalayan mountains.

The research is intended to provide crucial insights for hazard management and water-resource planning in high-mountain regions.

When glaciers melt, their water accumulates in natural depressions. In the Himalayas, glaciers are rapidly melting due to climate change, leading to the formation of new glacial lakes.

Although the rise in temperature is causing the generation of more meltwater, the formation and location of glacial lakes are influenced by surrounding topography such as bowl-shaped depressions, valleys, flow channels and lakes created by glaciers.

The expansion of glacial lakes increases the risk of their outburst, resulting in floods that can destroy infrastructure and disrupt natural habitats. Past disasters such as the Kedarnath floods in 2013 and the Uttarkashi floods in August 2025 highlight the high stakes involved.

Most existing models developed to address this challenge focus on climate while neglecting landscape features, resulting in a lack of completely reliable forecasts, the research team stated.

To bridge this gap, the IIT Guwahati team created a probabilistic framework considered a smarter way to forecast future glacial lake formation locations.

The team achieved this by using high-resolution Google Earth images and digital elevation models to capture complex landscape features and estimate uncertainty in predictions.

The research team said this makes the forecasts more realistic and reliable, reflecting the natural variability and unpredictability of mountain environments.

The findings have been published in Nature’s Scientific Reports journal in a paper co-authored by professor Ajay Dashora, assistant professor at the Department of Civil Engineering of IIT Guwahati, along with research scholar Anushka Vashistha, and Dr Afroz Ahmad Shah from the Universiti Brunei Darussalam (UBD), Brunei.

During the development process, the research team tested three predictive methods – Logistic Regression (LR), Artificial Neural Network (ANN) and Bayesian Neural Network (BNN).

Among these, the team found the BNN to be the most accurate and showed that certain earth features such as neighbouring lakes, cirques, gentle slopes and retreating glaciers are the strongest predictors of glacial lake formation.

“By pinpointing high-risk areas, the framework can guide early-warning systems for glacial lake outburst floods, help plan safer locations for roads, hydropower projects and settlements, and support long-term water resource management. It offers a practical tool for reducing risks to communities and infrastructure in the Himalayas,” professor Dashora said.

He continued, “Beyond hazard management, the method can help understand how water systems may change as glaciers continue to retreat. Importantly, the framework is adaptable to other glaciated mountain regions around the world, making it a valuable tool for climate-resilient planning and disaster risk reduction globally.”

With the developed framework, the research team identified 492 locations in the Eastern Himalayas where new glacial lakes are likely to form, indicating areas that require careful monitoring and preventive measures.

Professor Dashora said these findings confirm that the shape and structure of the land, often overlooked in previous studies, can play a central role in where and how a glacial lake may appear.

As the next step, the research team plans to integrate moraine development histories, automate data preparation and add field-based validation to the framework to enhance the model’s accuracy and broaden its use for large-scale monitoring of glacial hazards.