In: Paparoditis N., Pierrot-Deseilligny M.. Mallet C. Tournaire O. (Eds). IAPRS. Vol. XXXVIII. Part 3A - Saint-Mandé, France. September 1-3. 2010 
Figure 4. Results of the Otscher test site, (a) orthophoto adapted 
from © www.bing.com/maps; (b) a z -image; (c) nDSM derived 
from the combined DSM minus the DTM; (d) difference model 
(DSM^ - DSM ni i s ). The spatial resolution of the images is 
0.50 m. 
As the ALS data acquisition was carried out under leaf-off 
conditions and, consequently, several first echoes were reflected 
from stems or branches within the deciduous tree crowns, the 
moving planes interpolation leads to a significant 
underestimation of the surface heights. 
For the test site Vorarlberg the available ALS data have low 
point density, which increases the probability to smooth the 
DSM nl i s surface. This can clearly be seen in the shadings of the 
DSM niax and the DSM m i s in Fig. 5b,c and in the building and 
vegetation profiles shown in Fig. 5e. Considerable smoothing 
effects of the DSM^ occur along the building border (cp. the 
building in the right part of the figure) and within the forest. 
Using the DSM niax (Fig. 5b) instead of the DSM^ (Fig. 5c) for 
rough surfaces improves the accuracy of the final DSM and 
decreases these smoothing effects. However, with decreasing 
point densities the amount of nodata values of the calculated 
DSM niax increases (cp. Fig. 5b). To fill these data gaps the 
DSM^ was used. Due to the irregular point distribution of the 
used laser scanner (saw tooth pattern), the DSM niax appears to 
be rather rough even for smooth and sloped open areas (see 
Fig. 5b,c). 
/•1.-Ç.-T P1 
• I* - m • 
¿s- ; 
: T. .> 
Meters 
0 50 100 200 nDSM 
1 1 1 1 1 1 1 1 1 Efl >30m 
• DSM^ • DSM„„ 
• ALS points | 00m 
P2 
* .V 
e) 
fc. 
» .A ^ **' 
Figure 5. Illustration of DSM calculations based on (b) the 
highest point within a defined raster cell (DSM niax ) and (c) 
moving planes interpolation (DSM m | s ). In (b) the nodata cells of 
the DSM ]nax are highlighted in red. In (d) an nDSM is shown. 
The profiles in (e) illustrate the smoothing effects of the moving 
planes interpolation for a forest (PI) and for a building (P2). 
The spatial resolution of the CIR orthophoto (a) is 0.25 m, those 
of the DSMs and the nDSM 1.0 m. 
For such irregular point distributions it is suggested to search 
the k-nn points, which are used for the moving planes 
interpolation, separately in the four quadrants around the (x.v) 
interpolation location in order to avoid extrapolation effects. 
The combination of both DSMs depending on the cr z -values 
minimizes this effect as far as possible. The derived nDSM is 
shown in Fig. 5d. 
For the test site Neusiedler See the difference between the 
DSM^ and the DSM ir) i s is substantial for reed areas (cp. 
Fig. 6). Reed is one of the dominant plants in this region and it 
is of great importance for their ecosystems. Therefore, the 
underestimation of the reed surface in the order of 1.0 to 1.5 m, 
as shown in Fig. 6d, would constrain e.g. biodiversity analyses 
(i.e. determination of the horizontal and vertical distribution of 
reed). Based on the applied combined DSM calculation even 
small geometric structures in reed (i.e. linear structure) can be 
detected, as shown in Fig. 6b, which was not possible using the 
DSM m | S alone. 
225