Deriving spatially varying thresholds for real-time snowmelt detection from space-borne passive microwave observations

Abstract

Previously, detecting surface snowmelt from passive microwave remote sensing data based on thresholding used arbitrarily selected cut-off values. To avoid the problems of the arbitrary threshold, this letter aims to find optimal threshold values through automated calibration. By using a wavelet-transform-based method and an interactive visualization and analysis software tool, 6-year snowmelt data were derived as reference for the calibration. The optimal threshold value is determined by minimizing the melt index error. This study shows that the previously used threshold value tends to systematically underestimate the surface melt. The spatial and temporal variability of the threshold value is examined. Although the optimal threshold value varies from region to region in the Antarctic Ice Sheet, it is relatively consistent from year to year, which implies that the optimally determined threshold value for each region can be extended to other years.