I am trying to analyse elevation data for a coastal (almost flat) city with the highest elevation of about 30m. I have DEMs from ASTER, SRTM, and a DSM from ALOS W3D-30. I have realized that the values are too high: for ASTER, 0-250, for SRTM, 0-101, for ALOS, 0-92. Are the values not in metres (m) or is there something I need to do to have the correct (approximately) elevation values?
It's impossible to pinpoint what is going on without looking at the data per se, but a wild guess would be due to the DEM being a DSM and not a DTM. I don't know how well you know these terms, but on a general note:
- A DSM (Digital Surface Model) means a model getting all elevations above the surface, meaning tree canopy, top of buildings, etc. It's the raw elevation data on most platforms.
- A DTM (Digital Terrain Model) is a model getting all elevations on the surface. It is normally derived from a DSM, by means of manual and automatic processing to remove surface objects. A big part of photogrammetry nowadays is algorithms to more precisely remove the objects without the input of a human technitian (which is a very laborious process, requiring 3D glasses, powerful machines and hours of work).
- A DEM (Digital Elevation Model) is an umbrella term that encompasses either of the above.
SRTM, ALOS and ASTER are all DSMs, meaning they don't have any processing to get to terrain levels (unless you acquired them from a company that did so). Since all of them reveal bigger heights than the stated 30m, I'd guess you have tall buildings somewhere in your city, that are affecting the model.
Notice, also, that all of these have a measuring error imbued in them. SRTM error varies from place to place, but you can expect something around 4-9m; according to this study, ALOS seems to be about the same. ASTER, on the other hand, has a very high variance. While SRTM and ALOS are radar data, ASTER is an image satellite whose DEM was generated through stereo pairs. Meaning, where there is cloud, there will be a spike. Regardless, all three models rendered heights around or above 100m, so a problem with the data itself is highly unlikely.
Side from all these, you may want to question how reliable that 30m figure is. Did it come from a high precision GNSS or topographic survey? Does it coincide with the peaks seen in your DEMs? If you really do need to correct your DEM for any reason, there are ways to do so, but they are not easy and require specialized software. Especially if what you want is to convert from DSM to DTM.