DEM terrain height

martinvk

since 10 Aug 2002
How should terrain elevations from TransDEM be interpreted? In other words, what is being measured, the ground or the top of the highest object? Do trees count? Major mountains are obvious but in many other areas, there are undulations that don't appear to be related to the ground.

My recently imported TransDEM formed map has anomalous bumps in various places. Since it is a locally based map, I know from personal knowledge that there are not any actual hills in those areas. What do I do if I model areas I don't know?

When I look at the ground image from other sources like Google Earth, it is possible to see what kind of objects are really there. Is that the best that can be done? Are there any settings in TransDEM to force an average elevation over a defined area?
 
I think it all depends on the quality of the DEM, Martin. Most of our European DEM, at least that from the Viewfinder site, is probably already a bit averaged as rarely better than 50m resolution, often only 70m or 90m. Even the (old) Ordnance Survey 20m DEM data for the UK I got from UKTS is somewhat inaccurate in places. I once considered trying to do a version of the Hong Kong tramway but had to give up. The radar mapping had picked up all the high rise buildings and averaged these into the data, resulting in totally unusable DEM for what, sadly, would have been a fascinating albeit ambitious project.
 
I have found too that the original data is sometimes polluted by nearby objects. On a download of my own hometown, the raised rail grade is completely wiped out because the satellite got the nearby hill on one side, and the rail grade and the factories on the other as a single hillside. The grade then was wiped out totally. I have seen the same with factories out along rivers too. There are odd bumps on the ground where the foundations or buildings are supposed to be. I ended up smoothing them out and then placing my own buildings. In the case of the rail grade, it was nearly impossible trying to smooth that out the way it's supposed to be. Maybe in the future I'll revisit that, but at the time it was an awful lot of work trying to get things to look right.
 
Was afraid of that. Thanks for confirming I was not imagining it. I guess DEM derived maps are OK for rural grass lands but rather out of place for densely built urban areas. Too bad those urban areas tend to be the most interesting. So I'll have to apply some best guess modifications to the terrain.
 
How should terrain elevations from TransDEM be interpreted? In other words, what is being measured, the ground or the top of the highest object? Do trees count? Major mountains are obvious but in many other areas, there are undulations that don't appear to be related to the ground.
As the others have said, it depends. For the older orbital DEMs, SRTM and ASTER, it is very much down to the sensor and the post-processing applied. By nature of the sensor, SRTM does not distinguish between between terrain and buildings/vegetation. Any object that appeared on Radar was accepted. But we are at version 3 now, I think, and some more advanced post-processing may have been applied. There also variants of SRTM, prepared by various organisations, that may have more accurate terrain-only values.

Classic terrestrial DEMs are based on contour lines of topographic maps. So these are terrain-only.

Nowadays, with hi-res LIDAR, we have the Digital Terrain Model (DTM) and the Digital Surface Model (DSM). The DTM is the one we would go for. It is the bare-bones ground, no trees and no buildings - except for embankments, dams and other earthworks, and sometimes including abutments but not the pillars of bridges. But again, this applies to hi-res data only.
 
Back
Top