Snow Depth Modelling using Hybrid Optical - SAR Approach

This study investigates a hybrid Optical-SAR approach for modelling snow depth in North Sikkim, India, a region characterized by complex terrain and frequent cloud cover. By integrating Sentinel-2 multispectral imagery with Sentinel-1 SAR data, the research aims to improve the accuracy and reliability of snow depth estimations. The Gradient Boosting Regression (GBR) model was employed, utilizing key variables such as the Snow Index, SAR backscatter values, and NDVI. The model demonstrated strong predictive performance, with an $\mathrm{R}^{2}$ of 0.922 and an RMSE of 0.26. The snow depth maps generated are essential for hydrological planning, disaster management, and climate studies, offering valuable insights into snow dynamics in challenging environments. This approach highlights the potential of combining optical and radar data with machine learning techniques for enhanced environmental monitoring.