Abstract Glaciers in Tajikistan serve as a crucial water source in Central Asia. However, existing glacier data products suffer from several limitations, including outdated temporal coverage, coarse spatial resolution, and inadequate mapping of debris-covered glaciers. To address these issues, we have developed the Tajikistan Glacier Inventory (TGI), a high-resolution national-scale glacier dataset derived from 3 m resolution PlanetScope imagery acquired in 2024. Using a deep-learning model enhanced with an attention mechanism, we achieved precise glacier mapping, with validation confirming high overall accuracy (F1 = 92.1%, IoU = 85.4%). The model performed robustly in classifying both clean ice (F1 = 93.3%) and debris-covered ice (F1 = 90.8%). The TGI identifies 12,325 glaciers covering an area of 7,720.26 km 2 . It provides a systematic, glacier-level distinction between clean and debris-covered ice nationwide. This new inventory indicates an approximate 11% reduction in glacierized area relative to the Randolph Glacier Inventory, which aligns with the significant glacier loss observed over the past two decades. Released as an open-access resource, the TGI establishes a timely and reliable baseline for assessing regional water security, glacier-related hazards, and climate-change impacts across Central Asia.
Li et al. (Sat,) studied this question.