Evaluation of Satellite Remote Sensing Albedo Retrievals over the Ablation Area of the Southwestern Greenland Ice Sheet

摘要: Abstract MODerate resolution Imaging Spectroradiometer (MODIS) albedo products have been validated over spatially uniform, snow-covered areas of the Greenland ice sheet (GrIS) using so-called single ‘point-to-pixel’ method. This study expands on this methodology by applying a ‘multiple-point-to-pixel’ method and examination spatial autocorrelation (here semivariogram analysis) in situ observations, high-resolution WorldView-2 (WV-2) surface reflectances, MODIS Collection V006 daily blue-sky heterogeneous surfaces lower ablation zone southwest Greenland. Our results 232 ground-based samples within two pixels, one being more than other, show little difference accuracy among narrow broad band albedos (except for Band 2). Within homogenous pixel area, were very close (error varied from − 4% to + 7%) range ASD standard errors. The analysis revealed that minimum observational footprint needed representative sample is 30 m. In contrast, pixel, size was not quantifiable due autocorrelation, far exceeds effective retrievals. Over high heterogeneity ground measurements 4–7%, partly known undersampling darker often are impassable foot (e.g., meltwater features shadowing effects crevasses). Despite sampling issue, our errors stated general product 5%. Thus, suggests performs well heterogeneous, low-albedo, area zone. Furthermore, we demonstrate methods alone insufficient characterizing validating variation displayed area. true because distribution data deviations substantial range, with average values 10th 90th percentiles − 0.30 0.43 across all bands. if only point taken validation, randomly selected either tails, error would appear be considerable. Given need multiple in-situ points, concurrent derived existing AWSs, (low-flying vehicles (airborne or unmanned) imagery (WV-2)) resolve sub-pixel variability zone, thus, further improve characterization Greenland's albedo.