Guwahati: Researchers at the Indian Institute of Technology (IIT) Guwahati have developed a new predictive framework to assess glacial hazards in the Eastern Himalayas, identifying hundreds of locations where glacial lakes are likely to emerge in the coming years.
Using high-resolution Google Earth imagery along with digital elevation models, the research team mapped 492 potential sites where new glacial lakes could form as glaciers continue to retreat.
The study offers important inputs for disaster-risk reduction, infrastructure planning and long-term water-resource management in fragile high-mountain regions.
The framework is designed to capture complex terrain characteristics while also accounting for uncertainty in predictions, making its forecasts more realistic and operationally useful.
By flagging high-risk zones in advance, the model can support early-warning systems for Glacial Lake Outburst Floods (GLOFs) and help authorities make safer decisions on the placement of roads, hydropower projects and human settlements.
According to Prof. Ajay Dashora of the Department of Civil Engineering at IIT Guwahati, the framework provides a practical tool to reduce risks to Himalayan communities and critical infrastructure.
He noted that the approach can also help researchers understand how mountain water systems may evolve as climate change accelerates glacier retreat.
Beyond the Himalayas, the method has wider relevance.
The researchers said the framework can be adapted for use in other glaciated mountain regions across the world, contributing to climate-resilient planning and global disaster-risk reduction efforts.
The findings, published in Scientific Reports, highlight the critical role of landform characteristics in glacial lake development.
The study confirms that terrain features such as nearby existing lakes, cirques, gentle slopes and retreating glaciers strongly influence where new lakes are likely to form — factors that earlier studies often underestimated.
During the research, the team evaluated three predictive approaches — Logistic Regression, Artificial Neural Networks and Bayesian Neural Networks.
Among these, the Bayesian Neural Network emerged as the most accurate in forecasting potential glacial lake formation.
The researchers plan to further strengthen the framework by incorporating moraine development histories, automating data preparation processes and adding field-based validation.
These enhancements are expected to improve predictive accuracy and expand the model’s applicability for large-scale monitoring of glacial hazards in high-altitude regions.
