Contrasting short-term dynamics of supraglacial ponds along the Hindu Kush-Himalaya revealed by PlanetScope imagery and deep learning

An increasing number of supraglacial ponds have formed and expanded on the surface of debris-covered glaciers across the Hindu Kush-Himalaya (HKH) mountain range in the last decades. Despite the pronounced spatio-temporal variability observed in supraglacial ponds at annual and decadal scales, investigations of their seasonal changes are limited over large spatial scales. These investigations are critical for evaluating their impacts on glacier ablation and dynamics and predicting water resource availability. Here, we produced detailed seasonal maps of supraglacial ponds at five sites of the HKH for the years 2017 to 2022 using a deep-learning-based mapping method applied to PlanetScope imagery. Using these maps, we investigate pond seasonality and interannual variability. We found that (1) the average pond number and percentage ponded area over the debris-cover area were higher in the Central Himalaya (417, 1.55%) and Eastern Himalaya (481, 1.93%) compared to those in the Hindu Kush (142, 0.20%) and Western Himalaya (153, 0.19%); (2) pond percentage area over debris-cover area showed an increase in the Karakoram (+0.2% in an absolute sense), Central Himalaya (+0.6%) between 2017 and 2020, and Eastern Himalaya (+0.9%) between 2018 to 2021; (3) supraglacial ponds reached their peak at the onset of the ablation season (May-June) in the Karakoram and the Hindu Kush, during the pre-monsoon season in the Western and Central Himalaya, and during the monsoon or post-monsoon period in the Eastern Himalaya; (4) the Central Himalaya displayed a highest occurrence of persistent ponds (17.2%), while only 4.3% of supraglacial ponds in the Karakoram were persistent. Our results provide a spatially diverse and temporally detailed dataset that serves to advance the understanding of supraglacial pond dynamics across the Hindu Kush-Himalaya.