Publications

Xiao, XX; Liang, S (2024). Assessment of snow cover mapping algorithms from Landsat surface reflectance data and application to automated snowline delineation. REMOTE SENSING OF ENVIRONMENT, 307, 114163.

Abstract
The abundance of remote sensing imagery available has been extensively used for mapping snow cover extent in mountainous regions. However, previous studies have paid little attention to quantifying the uncertainties inherent in snow cover mapping algorithms when using Landsat observations, particularly in the context of delineating the snowline-a pivotal parameter for understanding the spatiotemporal dynamics of snow cover. Additionally, there is an urgent need for an automated processing approach capable of monitoring alpine snowline across expansive mountainous terrains. This study squarely addresses these gaps by primarily focusing on the precise delineation of snowline and the quantification of disparities in determining snowline elevation using eight snow cover mapping algorithms. Our approach is twofold: initially, we comprehensively assessed eight snow cover mapping algorithms using Landsat 8/9 data, contrasting their performance against high-spatialresolution (3 m) snow observations. Subsequently, we introduced a novel snowline delineation method, termed Automated Snowline Delineation on Binary Snow Cover (ASLD-BSC). This method is designed to determine snowline on binary snow cover maps generated by these eight algorithms and was rigorously assessed across 15 catchment basins in America. The comparative analysis of the eight snow cover algorithms revealed a hierarchy of performance, with three algorithms employing multi-band decision trees exhibiting the highest proficiency in snow cover mapping. They were succeeded by four NDSI-based algorithms, with the Blue Snow Threshold algorithm ranking the lowest in terms of performance. Furthermore, our assessment demonstrated that the proposed snowline delineation method, ASLD-BSC, successfully mitigated approximately 1/3 of misclassification pixels and effectively created robust snowline patterns for each binary snow map. When scrutinizing snowline elevation, we observed striking variations in elevation differences among the eight snow cover mapping algorithms relative to the reference snowline elevation (average snowline elevation: 121 m - 258 m; bottom 10% snowline elevation: 253 m - 512 m; top 10% snowline elevation: 206 m - 344 m). These findings underscore the pivotal role that the quality of binary snow maps plays in determining the accuracy of snowline and snowline elevation. Importantly, this study provides a comprehensive guide for selecting appropriate snow cover mapping algorithms, facilitating effective monitoring of Landsat-based snow cover in mountainous areas.

DOI:
10.1016/j.rse.2024.114163

ISSN:
1879-0704