<p>Glaciers and glacial lakes are critical indicators of global temperature shifts. The rapid expansion of glacial lakes poses substantial risks of glacial lake outburst floods (GLOFs), leading to devastating impacts on downstream communities and infrastructure. Comprehensive inventories of glacial lakes are essential for understanding the dynamics of glacier lakes and assessing the associated risks. Satellite imagery and deep learning models are used in the present study to overcome the challenges of traditional methods of glacial lake monitoring. The glacial lakes in Himachal Pradesh, located in the Northwestern Indian Himalayas, from 1992 to 2020 were detected and identified through U-net deep learning models to understand the associated disaster risks better using satellite imagery from Landsat 5 and 8 and Digital Elevation Models (DEMs). Water-sensitive bands were used to improve the accuracy of glacial lake mapping. The U-net models were specifically tailored for large-scale glacial lake detection and were validated using high-resolution Google Earth Pro images. The study demonstrated that the U-net models performed exceptionally well, with precision scores ranging from 0.92 and 0.94, recall rates between 0.89 and 0.90, F1 scores from 0.91 and 0.92, and Dice coefficients from 0.75 and 0.76. There is a significant increase in the number of glacial lakes, rising from 572 in 1992 to 835 in 2020, indicating a growth rate of 44% in the study area. The proposed method automates the entire glacial lake mapping process, significantly reducing the need for manual intervention and enabling large-scale efficient monitoring and identification of potentially vulnerable lakes. The findings provide critical insights for climate change research and disaster risk assessments, emphasizing the importance of monitoring glacial lake dynamics to mitigate the risks of GLOFs.</p>