Guwahati: Researchers at the Indian Institute of Technology (IIT) Guwahati have developed a model to predict where new glacial lakes may form in the Eastern Himalayas, aiming to help authorities manage risks and plan water resources in mountainous regions.
As glaciers retreat, water collects in natural depressions, forming lakes that did not exist previously. Rising temperatures caused by climate change are accelerating this process, increasing the risk of sudden floods that can damage settlements, infrastructure, and ecosystems. Past incidents, such as the 2013 Kedarnath floods and the 2025 Uttarkashi floods, illustrate the potential impact.
The formation and growth of glacial lakes are influenced not only by melting but also by local terrain, including slopes, valleys, and natural hollows. Earlier prediction methods largely relied on climate data and often overlooked terrain factors, limiting their accuracy.
To address this, the IIT Guwahati team developed a probabilistic approach that integrates topographic data with high-resolution satellite imagery and elevation maps. This method accounts for the natural variability of mountainous landscapes, improving prediction reliability.
The research, published in Scientific Reports by Nature, is co-authored by Professor Ajay Dashora (IIT Guwahati), research scholar Anushka Vashistha, and Afroz Ahmad Shah (Universiti Brunei Darussalam).
The team tested three techniques, Logistic Regression (LR), Artificial Neural Networks (ANN), and Bayesian Neural Networks (BNN), with BNN emerging as the most accurate. Key factors influencing future lake formation included nearby existing lakes, glacial cirques, gentle slopes, and retreating glaciers.
Professor Dashora said the model can support early-warning systems, guide safer road and hydropower project planning, and improve long-term water management. He added that the framework is adaptable to other glacier-rich regions worldwide.
Using the model, researchers identified 492 potential sites in the Eastern Himalayas where glacial lakes may form, indicating areas that require monitoring and preventive action.
The team plans to further refine the system by including historical moraine data, automating the workflow, and validating predictions in the field to enhance accuracy for large-scale monitoring.