Learning to Add and Subtract


Scambos and his colleagues had created the Mosaic of Antarctica map by layering hundreds of digital images of Antarctica taken by Aqua and Terra MODIS between late 2003 and early 2004. To help Fricker find the source of the elevation drop, he would use a similar technique. But instead of “adding” the layers of MODIS images together to improve the amount of detail in the map, he would “subtract” images captured at different times to figure out what had changed.

Scambos had been working with Robert Bindschadler of NASA’s Goddard Space Flight Center for the past two years, each scientist using data from different NASA sensors to explore the idea of “image differencing.” Fricker’s discovery was a perfect test case for them, and they jumped at the chance to apply their techniques to her find. Photo-like images can’t provide actual elevation change in meters or feet, but if they are captured under similar enough lighting conditions, differences in shadows or brightness between two images can verify that the slope of the surface, and therefore its elevation, has changed.

  Page 1Page 3
  Differenced MODIS images

“With image differencing, we had to be even more constrained than MOA,” Scambos explains. A change in the Sun’s position at the time of an image would give a false impression of change on the continent’s surface. To keep the shadows consistent, the researchers only used images acquired at a certain time of day and a certain time of year (between late November and early December). Scambos then collected as many images fitting those criteria as possible, cleaning up any flaws created by blowing snow or clouds. Once he “subtracted” the images, he could see actual changes in the surface of the snow and ice.

“Near the grounding line, there were lots of changes that occurred, lots of crevasses [cracks in the ice] that you could see because the moving crevasse ridges and troughs had shifted between images,” explains Scambos. The MODIS images confirmed that the elevation inland of the ice shelf grounding line had changed over an oval-shaped area about 15 by 30 kilometers. The area had clearly dropped in elevation, consistent with Fricker’s ICESat interpretation.


Subtraction of satellite images captured at roughly the same time of day on different dates revealed changes in ice sheet elevation. Differences in shadows or brightness show where the slope of the ice surface changed. An image of part of the Whillans Ice Stream from 2002 (top) subtracted from an image from 2005 (middle) revealed that a section of the ice stream sank (bottom). Scientists inferred that water had drained from a subglacial lake, which they named Subglacial Lake Engelhardt. (Images courtesy Ted Scambos, National Snow and Ice Data Center.)

  Aerial photo downstream of Lake Helen

“We knew immediately we were onto something big,” Scambos says. The elevation change at the surface was only the visible sign of a change that took place beneath the thick layer of fast-moving ice that covered the area. Buried beneath the ice stream, the cause of the change couldn’t be directly observed. It had to be deduced.

Could it be a sediment shift? Elsewhere in Antarctica, scientists had previously documented that substantial amounts of sediment could move under the ice, but Fricker and her colleagues didn’t think that sediment was responsible here. Under the ice, the thick mud would likely have had an uneven surface. The area Fricker and her colleagues studied had a smooth surface, and it changed more quickly than sediment could likely move. One of geology’s oldest principles is that when a liquid fills a basin, its top surface is smooth and parallel to the horizon. “It took us about a day to convince ourselves it was water,” Scambos recalls.

“It’s been known for a very long time that there are lakes under Antarctica’s ice,” Fricker explains. While the surface of the stream is frigid, the underbelly is warmer, with geothermal heat and the friction of the ice’s movement producing meltwater. “There are 145 documented subglacial lakes, and people are discovering more and more.” The lake she and her collaborators found was similar in area to Lake Tahoe (although not nearly as deep). The lake sat in the vicinity of Siple Coast, under the Ross Ice Shelf.


This aerial photo was taken downstream from Subglacial Lake Engelhardt (nicknamed Lake Helen), looking toward the Ross Ice Shelf. Ridges on the ice surface are crevasses, or cracks in the ice. Shifting and deepening of crevasses that are visible in satellite images are indicators of ice movement. (Photo courtesy Christina Hulbe, Portland State University.)

Antarctica map showing subglacial lakes


They knew they had found a subglacial lake, and because the elevation had dropped, they knew the lake had drained. But where did the water go, and why did it move?

“After we found that first lake, we went ahead and mapped all of the ice streams around Siple Coast,” Fricker recounts. When they did, they found 14 areas under the ice where elevation rose, fell, or oscillated between February 2003 and June 2006. Drawing from nearby features , they proposed names for the four biggest areas: Subglacial Lake Engelhardt (the largest one), Subglacial Lake Conway, Subglacial Lake Mercer, and Subglacial Lake Whillans.


The lakes discovered by Fricker and her colleagues (white dots) add more information to a large body of data about Antarctica’s subglacial lakes (black dots), now numbering well over 100. (NASA map by Robert Simmon, based on data from the Radarsat Antarctic Mapping Project, Ted Scambos, Chris Shuman, and Martin J. Siegert.)

ICESat elevation change graph ICESat elevation change graph

The simplest explanation for simultaneous elevation changes was that water was moving between these lakes. Glaciologists had previously documented that movements of meltwater beneath the ice can change the ice sheet’s surface elevation. Any subglacial water is subject to tremendous force from the weight of the ice overhead. As the ice stream above the lakes shifts, pressure increases in one area, and the water squishes to another area. The water flow into the new lake increases the pressure there, and eventually, that lake drains into another. As the lakes fill and drain, the elevation of the ice sheet above them rises and falls.


Between October 23, 2003, and June 2, 2006, average elevation of the ice surface along this Geoscience Laser Altimeter System (GLAS) track dropped from roughly 53 meters to about 44 meters. After March 3, 2006, the elevation drop was negligible, indicating that the event that caused the sinking had ended. (NASA image by Robert Simmon, based on GLAS data courtesy Helen Amanda Fricker, Scripps Institution of Oceanography.)

  Pressure field, lake map

Based on elevation data, image differencing, and ice thickness measurements that suggested where the ice would be leaning most heavily on the water below, Fricker and her colleagues were able to deduce that a complicated network of waterways underlies this region of Antarctic ice. They were also able to describe how some of the subglacial lakes are connected to one another based on where the overlying ice exerted pressure.

“It’s fascinating to watch the water drain in one place and appear in another,” Fricker says. “It’s also incredible to think that you can actually get an idea of what’s going on in that subglacial environment just by looking at the surface. This is under a kilometer of ice we’re talking. We’re actually seeing what’s going on.”


This image incorporates ice pressure data and measurements from the Geoscience Laser Altimeter System (GLAS) sensor on NASA’s ICESat satellite. Rainbow colors show the range of elevation changes (either up or down) observed between 2003 and 2006, with red indicating the greatest change and purple indicating the smallest. Relative pressure exerted by the ice sheet appears in grayscale, with white indicating the greatest pressure. Pools of water are likeliest to form in areas of low pressure. The yellow bands indicate possible pressure “ridges” separating different pressure fields and, consequently, separate water basins. (Image courtesy Ted Scambos, National Snow and Ice Data Center.)