DTM/DSM and Other Lidar Grids

By today's standards, the first DEM's were crude and generalized.  They had large spacing between postings.  It was also unclear how they accounted for vegetation and buildings, so many like SRTM and ASTER are really DSMs.  Those created photogrammetrically were probably DTMs.

With the advent of lidar (lasers collecting many thousands of elevations per second), the resolution and detail increased substantially.  We can now create multiple surfaces, each with a specific name and properties.

DTM creation


Five grids for Fort Monroe, Virginia

  • DTM from USGS.  This has had light edits to fill and place the entire moat at one elevation. The castle walls remain.
  • Drop in the bucket 1 m grids from MICRODEM.  These take the largest or smallest lidar return in the pixel.
    • DTM from just the ground points in the lidar point cloud.  The buildings, vegetion, and the moat are voids
    • NVS.  A 2 m grid removes the spikes from tree trunks, in cells where no returns penetrated to the ground.
    • DSM.  Only the SW corner of the moat has elevations.  Lidar has problems getting returns from still water.
    • CHM

Grid Synonyms Comments MICRODEM produces from point cloud, with drop in the bucket algorithm
DEM--digital elevation model   Generic term, without implying what is shown.  There is not widespread agreement about these terms, and in particular USGS now uses DEM for what is here called a DTM.  This usage follows the Maune book, and most European usage.  
DSM—digital surface model First return
  • Easiest to create
  • Only need to identify and remove obvious anomalies like bird strikes.
  • Not to be confused with digital soil map
DTM—digital terrain model Bare earth
Last return
  • Requires identification (classification) of ground points, buildings, and power lines. 
  • Challenges with tree stumps and large buildings.
  • Building ghosts look unrealistic. 
  • TIN gridding can create anomalies at data voids.
Yes, if point cloud has ground classification
DTM creation
CHM—canopy height model HAG, height above ground
nDSM, normalized digital surface model
  • Difference between DSM and DTM (or DSM and NVS), which is the height of vegetation.
NVS—Non-vegetated surface  

  • Easier to compute than DTM, and buildings affect most users of the grids, so leaving them in may create a more useful data set.

DoD: DEM of difference   This should be called a grid, and not a DEM.  


DSM--First return DEM.  This includes trees and buildings. DTM--Bare earth DEM CHM Difference grid, shown as Elevation map (red is tall).

DTM/DSM for the same area in Portsmouth, UK, from the UKOS.

The DTM is very generalized, and does not provide a good visualization of what is there.  It also has clear ghosts from the building removal process.

For the  DSM, each point in the point cloud will be placed at the nearest grid node, and the highest value retained. 

This is elevation coloring; the holes will make reflectance maps problematical.  It might be a good idea to use a larger grid spacing, or try to interpolate across the voids.
For the NVS, which might approximate a DTM.  each point in the point cloud will be placed at the nearest grid node, and the lowest value retained.   You need a grid size that is large enough that there are multiple returns in each grid cell, and one of them must have gotten to the ground.


Two lidar surveys a month apart (April 2010 and July 2010) in Pennsylvania, from OpenTopography.org.

Note the spikes corresponding to tree trunks in the leaf on red DTM, which are missing in the green leaf off DTM.

The black leaf on DSM is generally a little higher than the blue leaf off DSM.  Viewed in a reflectance map the two DSMs are very hard to differentiate.

Lidar interpretation.

Urban lidar displays

Last revision 1/16/2023