Glacial lakes mapping using satellite images and deep learning algorithms in Northwestern Indian Himalayas

Global climate change's influence on the Himalayan glaciers has initiated glacier retreats, resulting in the development and evolution of glacial lakes, alterations in river flow dynamics, and changes in glacier boundaries. These changes have led to a rise in the occurrence of glacial lake outburst floods (GLOFs), causing substantial socio-economic damages. Therefore, monitoring and studying these glacial lakes is imperative to comprehend the impacts of climate change on the cryosphere. Automatic and Semi-automatic methods for mapping glacial lakes from remote sensing satellite data have been developed and extensively used in mapping and monitoring glacial lakes. Artificial intelligence (AI) based machine learning and deep learning algorithms have been successfully used for feature extraction and image segmentation in recent years. Fully convolutional neural (FCN) networks-based U-Net architecture, which involves a gradual integration of superficial visual characteristics and semantic information extracted from images to segment small objects effectively, is used in the present study to extract glacial lakes pixel-by-pixel from satellite data, providing a more efficient and accurate method for identifying and mapping these lakes. The Landsat and Indian Remote Sensing imagery were utilized to train, test and map glacial lakes in the Chandra-Bhaga basin of Himachal Pradesh in North Western Indian Himalaya. A total of 134 glacial lakes were mapped in the basin for 2021, covering approximately 4 km2 area, yielding an aggregate accuracy score of 0.90 with a recall of 0.95, an F1-score of 0.96, and an intersection over union (IoU) value of 0.94. The digital elevation model (DEM) was used to remove the mountain shadows identified and extracted as false glacial lakes during the post-processing phase. The mapped glacial lakes were verified and validated using high-resolution satellite imagery from Google Earth Pro. The fully automatic Deep learning-based glacial lake extraction is effective and efficient for developing glacial lake inventory and continuous monitoring to assess the associated glacial lake outburst hazard.