Sustained Decline of Annual Snow Cover Area in the Sikkim Himalaya (1987–2025): Multi-Sensor Remote Sensing on Google Earth Engine, Machine Learning, and Projections to 2100

Prakash Pradhan 

Snow cover is a critical component of the Himalayan cryosphere, providing freshwater to millions of people across South Asia and regulating regional climate through albedo feedbacks. Sikkim, a small but ecologically significant state in the Eastern Himalaya, has experienced accelerating glacial retreat and mounting hydrological stress in recent decades. This study presents a comprehensive, multi-decadal analysis of annual snow cover area (SCA) in Sikkim from 1987 to 2025, utilising cloud-masked, cloud-free median composites derived from Landsat 5, Landsat 8, and Sentinel-2 satellite imagery processed on the Google Earth Engine (GEE) cloud computing platform. The Normalised Difference Snow Index (NDSI) threshold of 0.1 was applied uniformly across all sensors and years to map annual SCA at 30-metre spatial resolution. After quality-controlled exclusion of five years with zero valid satellite acquisitions (1984–1986, 2002, and 2012), a final time series of 37 annual observations was analysed. Results reveal a statistically significant and robust long-term decline in SCA (Mann–Kendall z = −4.49, p < 0.001; Sen’s slope = −17.62 km²/year), representing an estimated cumulative loss of approximately 669 km² over the study period. Decadal mean SCA declined monotonically from 1537.75 ± 342.55 km² in the 1990s to 923.22 ± 166.44 km² in the 2020s, a 40% reduction. Six machine learning and statistical models were evaluated via Leave-One-Out Cross-Validation (LOOCV): Support Vector Regression (SVR) achieved the best performance (RMSE = 237.19 km², R² = 39.4%). An inverse-RMSE weighted ensemble of LM, GAM, and SVR projects a continued decline to approximately −256 km² by 2100 under the central scenario (95% PI: −1194 to +937 km²), implying functional disappearance of annual snow cover before the end of the century under business-as-usual warming. These findings provide a quantitative, long-term baseline for cryospheric change in the Eastern Himalaya, with direct relevance to water resource planning, hydrological impact assessment, and climate adaptation policy for the Teesta Basin and downstream communities.