Monitoring Snow Water Equivalent (SWE) Using MODIS Time-Series Data and Machine Learning in Turnagain Arm, Alaska

Abstract. Global warming and water scarcity have made snowfall an essential area of study. Rapid melting of glaciers and snowfields transforms ecosystems and water availability, emphasizing the need to measure snow water equivalent (SWE). This study employs SWE, the Normalized Difference Snow Index (NDSI), and Snow Depth from MODIS time-series images to monitor and analyze snow-related changes effectively. MODIS images, with their consistent temporal resolution, are ideal for tracking seasonal and annual snow variations. This research examines MODIS data spanning 2007 to 2023 during the winter months (December to March) to evaluate changes in snow-covered areas and their water equivalents. The study focuses on Turnagain Arm, a prominent waterway in the north-western Gulf of Alaska. Machine Learning methods were applied to model SWE variations, using NDSI and Snow Depth as predictors. A test-train split approach was implemented to ensure robust and reliable results. Data from six USDA monitoring stations around Turnagain Arm supported the model's accuracy and relevance. The findings reveal significant trends in snow coverage and water storage over time, providing valuable insights for understanding snowmelt dynamics and informing strategies for water resource management in the region. This comprehensive approach demonstrates the potential of integrating remote sensing data and Machine Learning techniques to monitor environmental changes caused by global warming.