Terrain elevation

Terrain elevation addresses the question of how high the ground or ice surface is at a location, to what vertical accuracy, at what grid spacing and relative to which vertical reference. Results are digital elevation models (DEMs). A digital surface model (DSM) represents the top reflective surface including vegetation and buildings, while a digital terrain model (DTM) represents bare ground ^[1]. Ice-sheet and glacier surface elevation is included; vertical ground motion, the gravity field and ice mass are separate problems.

Copernicus DEM GLO-30 (1 arc second) and GLO-90 (3 arc seconds) are free global DSMs built on TanDEM-X data acquired between about 2011 and 2015, with older DEMs filling gaps ^[1]. The Copernicus collection description states absolute vertical accuracy below 4 m at 90 percent linear error (LE90) ^[1]. The German Aerospace Center (DLR) releases a TanDEM-X 90 m DEM for scientific use, a non-edited DSM with possible voids and artefacts ^[2][3]. DLR states a 12 m product goal of absolute vertical LE90 below 10 m ^[3] and reports an achieved absolute height error of about 1 m ^[2]; each figure refers to its own product and definition. The Shuttle Radar Topography Mission (SRTM) flew in February 2000 and covers about 80 percent of land ^[4]. Its 1 arc second product spans 60 degrees north to 56 degrees south and is void-filled from other DEMs ^[5]. JAXA states that AW3D30, a 1 arc second dataset, is free for commercial and non-commercial use ^[6].

The Polar Geospatial Center publishes ArcticDEM and the Reference Elevation Model of Antarctica (REMA) as mosaics at 2 m, 10 m and 32 m ^[7][8]. For REMA, it states absolute uncertainty below 1 m over most of the area ^[8]. In the United States, the 3D Elevation Program (3DEP) defines airborne lidar quality levels QL0 to QL3 with vertical root mean square errors of 5, 10, 10 and 20 cm ^[9].

Spaceborne lidar and radar altimetry supply height samples rather than wall-to-wall grids. ICESat-2 ATL06 land-ice height uses 40 m segments spaced 20 m between 88 degrees north and south, with a 91-day repeat ^[10]. The National Snow and Ice Data Center (NSIDC) describes centimetre-scale precision on flat ice, degraded by roughness, steep slopes and crevasses ^[10]. The Global Ecosystem Dynamics Investigation (GEDI) records ground elevation per 25 m footprint up to 51.6 degrees latitude ^[11]. Against airborne lidar over terrain in the Netherlands, Switzerland and New Zealand, one study found root mean square errors of 0.93 m for ICESat-2 and 2.96 m for GEDI ^[12]. CryoSat-2, carrying a radar altimeter, produced Greenland and Antarctic ice-sheet DEMs and elevation-change maps for 2011 to 2014 ^[13][14].

Vertical references differ: the DLR 90 m DEM, ICESat-2 ATL06 and ArcticDEM strips use WGS84 ellipsoid heights, while NASADEM, an SRTM reprocessing, converts to a geoid reference, so mixed products need datum reconciliation ^{[3][10][7][15]}. Acquisition epoch matters on surfaces that have since changed: Copernicus DEM reflects 2011 to 2015 and SRTM a February 2000 snapshot ^[1][4].