x
the distance to the nearest GPS point,
x
the GPS orthometric height, and
x
the height difference.
For all 15 GPS control points, a total of 342 LIDAR points fell within the 5 m
radius limits. The average difference in orthometric heights between the LIDAR and
GPS points was -0.37 m, with a standard deviation of 0.25 m. Because the majority of
the GPS validation points were on extensive flat surfaces within city parks and sports
fields, one would expect the height differences between the locations of individual
LIDAR hits and the nearest GPS points to be nearly constant within the 5 m radius.
Figure 6 shows a typical site in a city park near the waterfront with an extensive, flat,
grass field. In this case, the observed height difference based on the LIDAR elevations
was not constant, and the spatial distribution of differences showed a pattern. Because
we recorded the GPS time tag, we were able to classify the LIDAR points based on the
GPS time to distinguish points from different flight lines.
Taking the site shown in Figure 6 as an example, Figure 7 shows the LIDAR
ground points coded by flight line. The LIDAR points within 5 m of the GPS point are
represented by larger symbols and consist of two flight lines. Figure 8 shows the same
LIDAR points within 5 m of the GPS location, in this case coded by height difference.
From these figures, it appears that the height difference is related to the LIDAR flight
line. When the height differences are plotted against GPS time, the systematic height
difference is more apparent (Figure 9). One set of points flown at GPS time 47,600 s, is
designated line 1; another at time 62,010 s is termed line 2. The height differences for
line 1 range from -0.26 m to +0.03 m, with a mean difference of -0.12 m and a standard
deviation of 0.08 m. The height differences for line 2 range from -0.46 m to -0.20 m,
with a mean difference of -0.33 m and a standard deviation of 0.07 m. Thus the height
differences are distinct for each flight line although they have a similar variance range
and overlap slightly. It is clear that if adjustments in the form of a vertical offset are
applied to the data, these should preferably be specific to flight lines. Examining the
rest of the LIDAR points within 5 m of each of the GPS survey validation points
reveals a similar pattern, with distinctive height differences for different flight lines
(Figure 10). Another way to look at this relationship is to plot the GPS-LIDAR height
differences against GPS time (Figure 11). It is clear from this plot that, while the 0.9 m
adjustment to the LIDAR elevations was appropriate in an aggregate sense, the
variation in offsets between flight lines add an extra error term to the DEM elevations.
In many cases this difference is reduced because the mean elevation for a DEM grid
cell is computed from all points lying within the cell.
Another problem with the LIDAR data collected in 2000 involved large
variations in the density of returns. A reduction in the laser power resulted in flying the
Charlottetown LIDAR survey at lower altitude than planned (Webster et al., 2004).
This also resulted in a lack of LIDAR returns from near-infrared targets such as black
asphalt pavement and rooftops (Figure 5). This resulted in the clear delineation of the
167
Airborne Laser Altimetry
the distance to the nearest GPS point,
x
the GPS orthometric height, and
x
the height difference.
For all 15 GPS control points, a total of 342 LIDAR points fell within the 5 m
radius limits. The average difference in orthometric heights between the LIDAR and
GPS points was -0.37 m, with a standard deviation of 0.25 m. Because the majority of
the GPS validation points were on extensive flat surfaces within city parks and sports
fields, one would expect the height differences between the locations of individual
LIDAR hits and the nearest GPS points to be nearly constant within the 5 m radius.
Figure 6 shows a typical site in a city park near the waterfront with an extensive, flat,
grass field. In this case, the observed height difference based on the LIDAR elevations
was not constant, and the spatial distribution of differences showed a pattern. Because
we recorded the GPS time tag, we were able to classify the LIDAR points based on the
GPS time to distinguish points from different flight lines.
Taking the site shown in Figure 6 as an example, Figure 7 shows the LIDAR
ground points coded by flight line. The LIDAR points within 5 m of the GPS point are
represented by larger symbols and consist of two flight lines. Figure 8 shows the same
LIDAR points within 5 m of the GPS location, in this case coded by height difference.
From these figures, it appears that the height difference is related to the LIDAR flight
line. When the height differences are plotted against GPS time, the systematic height
difference is more apparent (Figure 9). One set of points flown at GPS time 47,600 s, is
designated line 1; another at time 62,010 s is termed line 2. The height differences for
line 1 range from -0.26 m to +0.03 m, with a mean difference of -0.12 m and a standard
deviation of 0.08 m. The height differences for line 2 range from -0.46 m to -0.20 m,
with a mean difference of -0.33 m and a standard deviation of 0.07 m. Thus the height
differences are distinct for each flight line although they have a similar variance range
and overlap slightly. It is clear that if adjustments in the form of a vertical offset are
applied to the data, these should preferably be specific to flight lines. Examining the
rest of the LIDAR points within 5 m of each of the GPS survey validation points
reveals a similar pattern, with distinctive height differences for different flight lines
(Figure 10). Another way to look at this relationship is to plot the GPS-LIDAR height
differences against GPS time (Figure 11). It is clear from this plot that, while the 0.9 m
adjustment to the LIDAR elevations was appropriate in an aggregate sense, the
variation in offsets between flight lines add an extra error term to the DEM elevations.
In many cases this difference is reduced because the mean elevation for a DEM grid
cell is computed from all points lying within the cell.
Another problem with the LIDAR data collected in 2000 involved large
variations in the density of returns. A reduction in the laser power resulted in flying the
Charlottetown LIDAR survey at lower altitude than planned (Webster et al., 2004).
This also resulted in a lack of LIDAR returns from near-infrared targets such as black
asphalt pavement and rooftops (Figure 5). This resulted in the clear delineation of the
167
Airborne Laser Altimetry