Similar to the GDEM1 validation, the GDEM2 was evaluated over the CONUS against the “GPS on Bench Marks” data set (Gesch et al., 2011) as an absolute geodetic reference . GDEM2 postings were bi-linearly interpolated over the benchmarks for the comparison
The GDEM2 was also evaluated against other 1 arcsecond elevation grids over the CONUS: the National Elevation Dataset (NED), the Shuttle Radar Topography Mission (SRTM) DEM, and the original GDEM1. The NED data set has a published vertical accuracy of 2-3 meter root-mean-squared error (RMSE), and thus was used as the primary reference for pixel-by-pixel grid comparisons. As with the Japan study, the CONUS assessments were segmented by land cover classes to look for relationships between accuracy and cover type
The land cover dataset used is the 2006 update of the National Land Cover Database (NLCD), which includes 19 classes 5
derived from 30-meter Landsat data. Like the GDEM1 validation study, all 934 1×1 degree GDEM2 tiles covering the CONUS were
included in this validation effort. The US effort focused on vertical accuracies and land cover effects