Glacier Dynamics and Water Security in Central Asia: Insights from Earth Observation Satellites and Machine Learning

Monitoring glacier dynamics over the long term is crucial for calibrating hydrological models and validating future runoff projections in catchments. As sensitive indicators of climate change, glaciers are particularly critical in Central Asia, one of the driest continental regions in the northern hemisphere, where their retreat exacerbates water scarcity, ecological disruption, and geopolitical tensions. These glaciers are primarily concentrated in Kyrgyzstan and Tajikistan, serving as vital water sources for agriculture in downstream countries. The Zarafshan and Amu Darya rivers extensively support agricultural activities in Uzbekistan, Tajikistan, and Turkmenistan, while the Syr Darya river sustains agriculture in Kazakhstan, Tajikistan, and Uzbekistan. Any changes in water availability could accelerate water conflicts in the region. In the Ili Basin, glaciers located in China contribute to the flow of water feeding Lake Balkhash in Kazakhstan. A reduction in this flow could threaten the lake's future, raising concerns about a potential repeat of the Aral Sea crisis in the region. Consequently, the future of glaciers emerges as a key driver of transboundary water issues in Central Asia, highlighting the urgent need for cooperative management and sustainable strategies to mitigate these challenges. This study integrates Earth Observation Satellite (EOS) data and assimilated products to analyze glacier area changes across the Amu Darya, Syr Darya, and Ili River basins from 1970 to 2024, employing machine learning methods. A time-series analysis framework was adopted to enhance glacier detection, leveraging ERA-5 monthly snow cover dataset. Seasonal snow cover often obscures glacier detection and leads to overestimations. ERA-5 data were filtered for warm-season imagery and aggregated into annual averages to address this. Snow cover persisting through warm seasons at elevations above 2500 meters was identified as glacier associated. This methodology was validated through comparisons with high-resolution EOS imagery, confirming its accuracy in delineating glacier extents. Trend analysis using Mann-Kendall and Pettitt test revealed a gentle decreasing trend in glacier area, beginning in the late 1990s to early 2000s. During the study period, glacier area declined by 13% in the Amu Darya basin, 21% in the Ili basin, and 29% in the Syr Darya basin. Following this, air temperature records displayed a statistically significant increase, with mean temperature rising by 0.75 ⁰C from 1996-1997. A negative correlation was observed between air temperature and glacier area, with coefficients of -0.44, -0.60 and -0.62 for the Amu Darya, Syr Darya, and Ili basins, respectively. Cross-correlation and R-squared analysis indicated that air temperature variability explained 31% of glacier area changes over short time horizons (e.g., one-year lag).