GDEM V3The announcement is quite promising:
Compared to Version 2, Version 3 has a decrease in elevation void area due to the increase of ASTER stereo image data and improved process, and a decrease in water area anomaly data due to using new global water body data to be described later.So overall: If you use more scenes, it should improve the quality, shouldn’t it? this dataset has a ground resolution of approx. 30m and looks good at first sight. I will proof the contrary. Yet, I have to admit: If you need elevation data for the non-covered region of SRTM this might be a first good startimg point to look at.
water bodiesLet’s quick check the water-body statement first: Yes, I can see a perfectly flat surface at the lake Ogii Nuur.
If you look closely on the lower left corner of this scene: there is an area that should be flooded, right? nearly 15m below the surface of the water body. And if you check the shore, you will see some “burn effect” for the water body. The GDEM values plunge from 13xxm to the water level at 1321m on a very short distance (blue profiles for GDEM, red ones for SRTM): As you can see here, and here, the topography is not that slopy at the shores of the lake. The level of Ogii Nuur is approx. 10m lower compared to the level in the SRTM scene and Wikipedia mentions the surface elevation with 1337m, which is quite similar to the SRTM value with 1330m.
nasty artefactsNow let’s move along the Orkhon river towards the south. The GDEM DEM is a DEM processed with images instead of radar or other nasty stuff. And the number of images heavily affects the scene. In the next image you see the number of scenes used for the DEM processing and the slope values of the scene. There are quite some footprints or stripes visible:
The Orkhon Valley is covered with thickdeposits of Quaternary unconsolidated sediments. Well-rounded basal-ticfluvial deposits that include rocks from a phase of middle to earlyPleistocene volcanic activity and alluvial deposits alternate with loessand glacial loams rich in calcium carbonatepaper So we should assume a leveled topography. But contrary we some artefacts here as well: . If we recalculate the black-white, stretch the differences are very easy to see: