Author Archives: Pseudogley


1. Identifying the vegetation height by subtracting the DSM (Digital Surface Model) from the DEM (Digital Elevation Model). The result of this process is a vegetation height model.

2. Identifying the peaks of single trees by inverting the vegetation height model and calculating the flow accumulation.

3. Extracting points of flow accumulation.

4. Re-inverting the points of flow accumulation.

5. Calculating the variance of the tree peaks.

6. Classifying trees into single-story and multi-story areas with a defined threshold.


How has the DEM been derived from the LiDAR dataset?
The DEM has been derived from the LiDAR dataset by using ground returns and a linear interpolation technique (0.2 m cell resolution). The ground returns were identified with help of a ground-finding algorithm (Zimble et al. 2003, p. 174-175).

How has the tree height/DSM information been retrieved from the LiDAR dataset?
The DSM has been retrieved from the LiDAR dataset by subtracting the ground DEM from the first return of the LiDAR dataset. Trees and their heights have been identified operating a complicated model (Zimble et al. 2003, p. 174-175).

How has the vertical structure been classified into single- and multi-story areas?
The vertical structure has been classified into single-story areas if the tree height variance of the 30.0-m cell has been below 1.54 m. It has been classified into multi-story areas if the tree height variance has been more than 1.54 m (Zimble et al. 2003, p. 176).

What is the principle difference between the single-/multi-storage data set and the characterization of the forest areas using the coefficient of variation approach (section 4)?
The coefficient of variation provides a unitless ratio between the standard deviation and mean of the tree heights. This causes independence of sample sizes and comparability of samples of unequal sizes (Zimble et al. 2003, p. 179-180).

Which errors of the approach can be attributed to post spacing?
Errors of the approach can be attributed to large post spacing (2.0 m or more). This gives the opportunity of missing peaks of the trees and receiving too low heights or detecting the sides of trees as lower peaks. Other possible errors are multiple tree counts caused by crowns with multiple peaks (Zimble et al. 2003, p. 177).

Zimble, D. A., Evans, D. L., Carlson, G. C., Parker, R. C., Grado, S. C. & P. D. Gerard (2003): Characterizing vertical forest structure using small-footprint airborne LiDAR. – In: Remote Sensing of Environment 87. – 171-182.