Canopy Height Models (CHM) are how many of us perceive LiDAR data and outputs in GIS systems in 2D, but sometimes it is good to understand how they actually represent the underlying LiDAR point cloud data. A client recently asked about the difference in height between a CHM model he was using and actual tree heights he was measuring in the field. In this case he was using a provided 1x1m resolution CHM raster surface model.
When the LiDAR data is dense enough it is also possible to increase the resolution of the CHM and in this case, we have also produced a 0.3×0.3m resolution CHM raster surface model. We have put together a range of 2D and 3D views of a CHM and the corresponding point cloud to help users understand more about CHM raster surface models and how they relate to the underlying LiDAR point cloud data.
Figures 1 through 5 shows a CHM both in 2D and 3D at different levels of zoom.
Figure 1 shows a 2D view of a 1m resolution CHM
Figure 2 shows a 2D view of a 0.3m resolution CHM
Figure 3 shows a 3D view of a LiDAR point cloud from which the CHM was defined.
Figure 4 shows a 3D view of the 1m resolution CHM (Left) and 0.3m resolution CHM (Right)
Figure 5 shows a 3D view of the 3D LiDAR point cloud for the area zoomed to in Figure 4.
Now if we take a profile of a single row of tree just to look at how the CHM compared to the actual point cloud. In this case these are young plantation trees from the same data as above. I have extracted out a 1m wide profile 26m long to show only a single tree row (Figure 6).
Figure 6 shows the 3D profile of the LiDAR point cloud as extracted from the dataset.
Now looking at the side profile (Figure 7) and 3D view (Figure 8) with the 1x1m resolution CHM raster applied to the point cloud.
Figure 7 side profile of 1m wide tree row and 1x1m raster surface CHM.
Figure 8 3D profile of 1m wide tree row and 1x1m raster surface CHM.
Here is the same view with the 0.3×0.3m resolution CHM applied to the same point cloud (Figure 9). The side profile does not do the 0.3 resolution CHM full justice as viewed in 3D you can see the more refined definition of the tree peaks across the 1m wide profile (Figure 10).
Figure 9 side profile of 1m wide tree row and 0.3×0.3m raster surface CHM.
Figure 10 3D profile of 1m wide tree row and 0.3×0.3m raster surface CHM.
And both in Figure 11.
Figure 11 Side profile showing the 1m and 0.3m CHM raster surfaces.
As you can see the CHM resolution will greatly dictate how accurate the information is when compared LiDAR point cloud. Both are valid and depends on your need and use of the resulting datasets. A 1m resolution CHM is often small in filesize, quick to view and therefore will be fit for purpose for most intended visual uses, especially when viewed at scales of 1:5000 or more. However if you need tree detail, it is best to use a high resolution CHM raster (or ultimately the actual point cloud data itself).
Many are also still using CHM’s for the detection of tree tops, so is also important to understand these effects when using tree detection algorthims using CHM rasters as their primary input layer.
All of these examples are using CHM’s derived using RapidLasso LAStools and a pit free algorithm for deriving the CHM.
If you would like to know more about Canopy Height Models and how the apply them feel free to contact our LiDAR team.