The shape of the landscape hidden under thousands of meters of ice in Antarctica remains one of the least known places in the World. It is actually very difficult to measure ice thickness. This is typically done by flying airplanes with a low-frequency radar mounted under the wings: the radar emits an electromagnetic wave that is reflected at the surface of the ice and then at the ice/bedrock interface. We then get the ice thickness below the aircraft by translating the time between both reflections to a distance. While reliable and accurate, it is completely impractical to use this method to map the bed of the Antarctic ice sheet, which is bigger than the US and Mexico combined, at 1 kilometer, the spatial resolution required by ice sheet numerical models.

Despite almost five decades of costly international campaigns of subglacial bed topography measurements, many sectors of the Antarctic Ice Sheet remain completely unmapped. The bed topography is, however, a critical control on the short- and long-term stability of the ice sheet and we cannot provide reliable projections of sea level rise from melting ice sheets if we don't improve our knowledge of the bedrock topography. To overcome this problem, we developed a new method, based on a very simple law of physics: the conservation of mass, to map the bed at an unprecedented resolution and accuracy.

The idea is simple: what comes in, comes out. We combine satellite measurements that tell us where the ice goes, with the rate of snow accumulation and melting at the surface, and constrain the calculation with available measurements of ice thickness from radars, so we can "fill in the blanks" using the principle of mass conservation.

This new description of the bed topography hidden under the ice is called "BedMachine" because it is constantly updated as new data become available. The results reveal the widespread presence of deep submarine ice-covered valleys, with a bed deep enough below sea level to enable strong interactions between ice and oceanic heat. These are regions that are potentially at risk of rapid retreat if warm, salty water from the Antarctic Circumpolar Current makes its way to the continental shelf and then gets in contact with the ice. There is no physical barrier at the bed that could stop this warm water from melting the ice as it retreats inland. We also found some good news: all glaciers flowing across the Transantarctic Mountains and Victoria Land are protected by broad, stabilizing ridges near the region where the ice starts to float. These ridges will act as strong anchor point and these regions cannot lead to a rapid collapse of the ice sheet. 

One of the biggest surprises of this work is that we mapped for the first time, under Denman glacier in East Antarctica, a valley that is more than 3,500 m below sea level, reaching the deepest continental point on Earth. Several attempts had been made to sound the bed topography using radar techniques but the very steep shape of the sidewalls led to side reflections that made it impossible to identify the basal echo. BedMachine provides the first reliable estimate of how deep this 150 km-long valley is. Given the depth of this canyon, this region could be at risk of rapid mass loss if the glacier start to retreat.

These newly unveiled topographic details have a profound impact on the distribution of potential higher-risk zones for rapid sea level rise from Antarctica that call for a re-evaluation by numerical ice sheet models using the new ice-sheet geometry.