Spatial assessment of snow cover patterns in the Sutlej River Basin using machine learning approaches and remote sensing data

Abstract

Snow cover information plays a significant role in the hydrology and climate of Himalayan river basins, making it an essential parameter for understanding seasonal flow variations in these regions. This study investigates the spatial variation of snow cover concerning elevation, slope, and aspect ratio across the Sutlej River Basin (SRB) over three seasons, monsoon, winter, and summer, from 2013 to 2021. The study was conducted on the Google Earth Engine (GEE) platform, using two machine–learning algorithms, Random Forest (RF) and Support Vector Machine (SVM), to classify the Landsat satellite data. The study results reveal that the Random Forest classification consistently demonstrated better performance across all three seasons, showing higher overall accuracy and Kappa coefficient values. A decadal increasing trend in Snow Cover Area (SCA) was observed throughout the Sutlej River Basin (SRB). Furthermore, topographic parameters such as elevation, slope, and aspect significantly influenced the spatial distribution of snow cover, showing patterns that contrast with broader climate trends. Specifically, higher elevations particularly those above 4500 m consistently retained substantial snow cover across all seasons. Slopes between 30° and 45°, classified as intermediate gradients, provided an optimal balance between steepness and flatness, promoting maximum snow retention. Regarding aspect, northern and northeastern-facing slopes showed the highest snow accumulation due to reduced solar radiation, which aids in preserving snow during warmer periods. Further, the results highlight the influence of climate variability, with a declining trend in summer snow cover and an increasing trend in monsoon snow cover observed over the past three years (2019–2021).