How robust are in situ observations for validating satellite-derived albedo over the dark zone of the Greenland Ice Sheet?

Calibration and validation of satellite-derived ice sheet albedo data require high-quality, in situ measurements commonly acquired by up and down facing pyranometers mounted on automated weather stations (AWS). However, direct comparison between ground and satellite-derived albedo can only be justified when the measured surface is homogeneous at the length-scale of both satellite pixel and in situ footprint. Here we use digital imagery acquired by an unmanned aerial vehicle to evaluate point-to-pixel albedo comparisons across the western, ablating margin of the Greenland Ice Sheet. Our results reveal that in situ measurements overestimate albedo by up to 0.10 at the end of the melt season because the ground footprints of AWS-mounted pyranometers are insufficient to capture the spatial heterogeneity of the ice surface as it progressively ablates and darkens. Statistical analysis of 21 AWS across the entire Greenland Ice Sheet reveals that almost half suffer from this bias, including some AWS located within the wet snow zone. Plain Language Summary Ground measurements of reflectivity, such as those made by automated weather stations, are often used to determine the accuracy of satellite measurements. But the footprints of the instruments mounted on automated weather stations are usually much smaller than the pixel of the satellite image, meaning that comparison between the two is only justified when the surface is relatively uniform. We use high resolution imagery collected by a UAV to demonstrate that the surface of the Greenland Ice Sheet is often not uniform at the scale of both the weather station and satellite pixel due to the presence of impurities, surface water and crevasses. This means that a point measurement of reflectivity might not capture the full variability of the surface, resulting in discrepancies when compared to satellite image pixels. Furthermore, weather stations are usually located on safe areas of flat, bare ice or snow, so they usually overestimate reflectivity in comparison to the satellite pixel. We argue that if unrepresentative ground measurements are removed from satellite comparison exercises then the uncertainty in satellite products could be reduced. Hence, the long-term decline in Greenland Ice Sheet reflectivity between 2000 and 2012 might be more significant than previously thought.