REMA Strips

Strip DEM files correspond to the overlapping area of the input stereoscopic imagery pair strips as they are collected by DigitalGlobe’s constellation of polar-orbiting satellites. Strip DEM dimensions will vary according to the satellite sensor that acquired the images and the off-nadir angle of collection. Most strips are between 13 km and 17 km in width, and 110 km and 120 km in length.

Strip DEM files are provided at 2-meter or 8-meter spatial resolution in 32-bit GeoTIFF format. Elevation units are meters and are referenced to the WGS84 ellipsoid. No ground control or altimetry registration has been applied to the strips.

REMA Mosaic

MOSAIC TILES

In addition to the individual time-stamped strips, REMA includes a mosaic for 100×100 km tiles. The mosaic tiles are generated at 8 meter posting.

In addition to the complete 8 meter dataset, reduced-resolution, resampled versions are availbled at 100-meter, 200-meter, and 1-km resolutions. The reduced-resolution datasets have an alternate filled version.

Methodology

Hillshade representation images were generated for each DEM strip segment and these were visually inspected and classified based on visual quality (i.e. lack of erroneous surfaces due to clouds, shadows, etc). Images were either accepted, manually edited to mask erroneous surfaces, or rejected. All strips were registered to altimetry point clouds obtained from Cryosat-2 radar and ICESat GLAS laser campaign 2D (25 Nov. to 17 Dec., 2008). For Cryosat-2 registration, only vertical bias corrections with a 1-sigma uncertainty of less than 0.1 m and residuals with a standard deviation of less than 1 m were applied. For ICESat, only corrections with a vertical residual of less that 0.35 m were applied.

Quality-controlled strip DEMs are mosaicked into 100 km by 100 km tiles with a 1-km wide buffer on each side to enable coregistration and feathering between tiles. For each tile, strips with altimetry registration are added first, in order of ascending vertical error, with a linear distance-weighted edge feather applied to the strip boundaries. The error value at each pixel is the registration error and the date stamp is the day of DEM acquisition. In areas where edges of strips have been feathered, the error and date stamp are averaged with the same weighting as the elevation. Once all registered strips have been added, unregistered strips are added to fill gaps and are coregistered to the existing, registered data in the mosaic. Each quality-controlled, unregistered strip overlapping a gap is tested for quality of coregistration, with the strip with the smallest coregistration residual selected to fill the the gap. Again, a distance weighted feathering is applied to smooth strip edges.

If Cryosat-2 registered data are available within a tile, those data are used and any ICESat registrations are ignored. If neither Cryosat-2 or ICESat registered data are available, the quality-controlled strip with the most coverage of the tile is added first and serves as a relative reference. Unregistered strips are then coregistered to the mosaic and added as described above. Figure 2 shows the distribution of tiles registered to Cryosat-2, ICESat or alignment to neighbors. Tiles around the edge of the ice sheet, within the CryoSat-2 SARin mode zone, are mostly registered to contemporaneous Cryosat-2 altimetry with the exception of coast tiles with too little land surface or extensive crevassing that prevent successful altimetry registration. Most of the interior tiles are registered to ICESat and therefore have a nominal date stamp of late December 2008, although little or no secular surface elevation change is expected in these regions on sub decadal time scales. Some tiles missing registration, and thus registered through alignment, are found around the pole hole and along a narrow zone in to the northeast of the pole hole. Lack of registration in these cases was due to missing or lower quality DEM data, resulting in registration residuals larger than the threshold. Once the tile is complete, the tile is then registered to surrounding, registered tiles using the overlap provided by the edge buffer. Tiles edges are then feathered to smooth any offsets between tiles edges and then buffers are cropped.

Finally, we apply a coastline mask using the British Antarctic Survey (BAS) land/ice classification polygons from the Antarctic Digital Database. Since this coastline is of a lower resolution and does not precisely match REMA in several areas, we buffered the coastline by 800 m and masked as all surfaces within the buffer that are less than 2 m from the local mean sea level.