A novel approach for the separation of frozen water bodies from snow to enhance snow mapping accuracy in high-altitude wetlands

Abstract

Precise estimation of snow cover is of great importance for hydrological, environmental, and climate change impact studies due to the significant role it plays in runoff generation and climate regulation in high altitudes, where frozen / glacier lakes pose a potential threat from outbursts of floods. Although many remote sensing-based indices are employed to delineate snow and to separate frozen water bodies from snow, it is difficult, especially in high-altitude wetlands, due to their spectral similarity, which may result in an overestimation of snow. Moreover, most of the existing snow indices rely on thresholding techniques to distinguish snow from waterbodies, but this approach has limitations due to its high spatio temporal variability, leading to reduced accuracy in snow mapping. In this context, the present study developed a new constraint based snow mapping approach (CBSMA) for the separation of frozen waterbodies /glacial lakes from spectrally similar snow-covered areas without the need for thresholding and demonstrates its potential for precise snow mapping compared to other existing spectral indices using Sentinel 2 imagery. The evaluation of CBSMA based on qualitative and quantitative analyses indicated overall agreement values of above 95 % and F1 score values of above 0.95 across all study areas and periods. The poor correlation of CBSMA with the existing indices ensures a reduction in spectral confusion and could improve the precision in the separation of frozen and non-frozen water bodies from snow cover areas, even without the need for thresholding or masking. Hence, CBSMA can be used as a promising alternative to the existing snow mapping methods.