James Lea spends a great deal of time looking at satellite imagery, particularly of cold places. The glaciologist from the University of Liverpool has been studying photos of the Greenland ice sheet and its edges in recent years, cataloging melt ponds as they shape, when they drain, how big they expand. Lea and colleague Stephen Brough frequently turn to computers to sort through the volume of images, and it is painful yet necessary work.
But the human eye overcomes the artificial one often. Lea discovered a gem in October 2020 while browsing at some older imagery. “At present, he tweeted to colleagues, “Doing a lot of boring manual tests, but it certainly helps when you come across utter stunners of photos like this one.
The Operational Land Imager (OLI) on the Landsat 8 captured a natural-color picture (top) of glaciers and mountains in eastern Greenland, just inside the coastline, on July 30, 2019. In the early evening sunshine, alpine terrain cast long shadows. The organic Landsat 8 data was draped over a global digital elevation model in the second image to capture the area’s topography. A bit northeastern part of Gunnbjørn Fjeld, the tallest mountain in Greenland, lies the area shown in the photos. In the Watkins Range, Greenland’s tallest mountain range, the area is filled with nunataks, rugged peaks that stick through glaciers and ice sheets.
Using the Moderate Resolution Imaging Spectroradiometer (MODIS) devices on NASA’s Aqua and Terra satellites, Lea and Brough usually view Greenland’s frozen surfaces, providing two views island at a precision of approximately 250 meters each pixel. Using Google Earth Engine’s computing power, they monitor melting pools and freshwater lakes that collect in the warm summer sunlight on top of the ice. The evolution of these ponds is of great concern to scientists because they frequently leak to the bottom of the ice, drain into the ocean (affecting the sea level), and lubricate glaciers’ movement as they flow downhill.
Lea periodically cross-checks his melt-pond exposures against Landsat images, which are acquired less frequently but with a much higher resolution (15 to 30 meters per pixel), because the area observed by MODIS is so large. Here is a MODIS view from the next morning, showing the same region as the Landsat picture for comparison. Both were collected when Greenland was in the middle of a widespread heatwave that caused 90% of the ice sheet to melt.
“We can verify with the Landsat 8 resolution whether lakes that maybe only a few pixels in MODIS are really lakes, or whether they are shadows, ice debris, or nunataks,” said Lea. The low sun angle in the picture above helps us to see a worst-case equivalent of what the regular MODIS image mosaic displays. Shadows occupy an even narrower region in elevated Sun-angle images. But considering that several photographs taken at various Sun angles may be made up of the MODIS composite, this can generate a problem with false positives. One of the validation steps we have undertaken with the data is looking at the low Sun-angle photos.