The loss of glacier mass and the increase in glacial lakes in High Mountain Asia due to climate change led to water-related crises and hazards, such as glacial lake outburst floods (GLOFs). The Sichuan-Tibet Railway (STR), which is crucial for the development of West China, is facing challenges due to these water-related crises. Therefore, there is an urgent need for a high-quality dataset of glacial lakes along the STR corridor, which remains lacking. This study employed a semi-automated lake mapping method, accompanied by rigorous quality assurance and quality control, to create a comprehensive, high-resolution dataset of glacial lakes in approximately 2021, covering the entire STR corridor. The 10-m resolution glacial lake dataset comprises 17,566 glacial lakes, covering a total area of 582.89 ± 31.45 km². This dataset features a larger number of lakes yet exhibits a lower mapping error compared to existing inventories in this region. The high-quality glacial lake dataset has the potential to benefit various applications, including assessing lake mapping accuracy, training deep learning algorithms, evaluating water sources, assessing risks related to glacial lakes, and understanding the interactions between glaciers and lakes.

The rapid expansion of glacial lakes in high-mountain regions, both in number and area, increases the risk of Glacial Lake Outburst Floods (GLOFs) to downstream communities. Optical remote sensing is essential for regional monitoring, as these lakes are frequently widespread and often inaccessible. However, clear image acquisition in mountainous areas is regularly hindered by shadowing, seasonal snow, and cloud cover. In the Himalayas, more frequent imaging is necessary to accurately inventory lakes during the short, snow-free periods. The HJ-2A/B satellites deployed by China now provide daily revisits at 16 m resolution, enabling comprehensive, high-quality mapping of glacial lakes. This study introduces a novel framework that uses a deep-learning U-Net model to detect lakes in HJ-2 imagery automatically and assesses its efficacy for monitoring lake changes in the China—Nepal section of the Central Himalayas (CNCH). Our findings indicate that the frequent imaging capabilities of HJ-2 permit annual lake mapping within a specific 1-mon period, achieving an accuracy and sensitivity to area changes exceeding 99%, with the ability to detect changes as small as 0.004 km². In our case study, we identify 2738 lakes in 2022 and 2739 lakes in 2023, with respective areas of 256.51 ± 14.41 km² and 260.89 ± 14.59 km². Despite certain limitations in geometric and radiometric calibration, this study establishes that HJ-2 imagery is highly effective for consistent, frequent monitoring of glacial lake changes and GLOF risks compared to previous satellite imagery.