Continuously monitoring and mapping glacial lake variation is of great importance for determining changes in water resources and potential hazards in alpine cryospheric regions. The semi-automated glacial lake mapping methods used currently are hampered by inherent subjectivity and inefficiency. This study used improved YOLOv5 strategies to extract glacial lake boundaries from Sentinel-2 imagery. These strategies include using the space-to-depth technique to identify small glacial lakes, and adopting the coordinate attention and the convolution block attention modules to improve mapping performance and adaptability. In terms of glacial lake extraction, the improved YOLOv5-seg network achieved values of 0.95, 0.93, 0.96, and 0.94 for precision (P), recall (R), mAP_0.5, and the F1 score, respectively, indicating an overall improvement in performance of 12% compared to that of the newest YOLOv8 networks. In High Mountain Asia (HMA), 23,108 glacial lakes with a total area of 1847.5 km² were identified in imagery from 2022 using the proposed method. Compared with the use of manual interpretation for lake boundary extraction in test sites of HMA, the proposed method achieved values of 0.89, 0.87, and 0.86 for P, R, and the F1 score, respectively. Our proposed deep learning method has improved accuracy in glacial lake extraction because it can address the challenge represented by frozen or high-turbidity glacial lakes in HMA.

Light-absorbing particles, which are vital components of aerosols, can cause significant snow albedo darkening and accelerate melting. However, restricted by the poor quality of remote sensing-based aerosol products in High Mountain Asia (HMA), previous studies have seldom reported the long-term pattern of aerosols. In this study, we analyzed the spatial and temporal distribution characteristics of AOD in HMA and surrounding areas using Moderate Resolution Imaging Spectroradiometer and Ozone Monitoring Instrument data from 2004 to 2023. The Mann-Kendall test was applied to analyze the temporal trend and abrupt changes in AOD, while Rotated Empirical Orthogonal Function was used to identify subregions and investigate spatiotemporal variations. Moreover, random forest and XGBoost-Shap models were employed to quantify the contributions of the aerosols to changes in snow albedo and melting. The results indicate that the annual (monthly) average highest and lowest AOD occurred in 2021 (April) and 2022 (September) between 2004 and 2023, respectively. The AOD first increased and then decreased during our study period and an abrupt decline was detected in 2013. The REOF model revealed three regions in HMA (northern, southwestern, and southeastern parts) with strong variations in AOD load, which are strongly correlated with atmospheric circulation and monsoon driving. Specifically, REOF1, REOF2, and REOF3 are primarily associated with frequent dust events during springtime atmospheric circulation and anthropogenic emission transport during the monsoon season. Aerosol types were divided into four types, BC aerosol, DUST aerosol, MIX aerosol, and clean conditions, whose proportions were 16.7%, 16.1%, 6.6%, and 60.6%, respectively. The clean conditions constituted the main aerosol type in the region. The AOD notably decreased snow albedo (17.8%) and increased snowmelt (9.0%); we highlight the contribution of AOD to the intensification of snowmelt. These results could provide guidance for further studies on the relationship between snowmelt and AOD.