9-11 Nov. 1999
s, summer first (sf), summer
> selected stands. The height
idth) is 1 m. The histograms
ally and relative frequency
er impression of the close
elative) number of reflected
curves show some typical
no off-terrain reflections in
T œ © m
2.2 & 2
ov s
S> 3 3 3
/
r a mature beech stand.
ai
[entender
|
~ co © —
aq «S E
= o~ oN o
e as
s for a spruce stand.
2604.
?608 -
2606
96001
International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 3W14, La Jolla, CA, 9-11 Nov. 1999
winter last.
e Coniferous trees show very similar patterns in winter and
summer.
e For both deciduous and coniferous trees, there is only a
slight difference between summer last and summer first
pulse.
The last observation shows that it does not make much sense to
go to the extra effort of an additional first pulse flight. The
TopoSys scanner does not allow to store more than one reflected
signal at a time, so this can save costs.
The penetration rate can be estimated at different heights from the
cumulative frequency curves. Results from 50 different stands
showed that any penetration rate lower than 100 percent for
deciduous trees in the winter last data can be explained by the
presence of coniferous trees. The separation between deciduous
and coniferous trees is thus very simple with both a winter and a
summer flight. Since beech and fir are the dominant tree species
in the project area, it was not possible to distinguish between
different deciduous or coniferous tree species.
The penetration rate strongly depends on the number of reflected
pulses. This is not a constant value. Rather, it depends on the
intensity of the reflected rays; signals too low may not be
recognized by the laser scanner. Figure 6 shows the number of
reflected laser dots. The position of the flight stripes can be
clearly recognized as well as different vegetation types. For
coniferous trees, generally much fewer laser points are recorded
than with deciduous trees. This is quite reasonable, since the laser
scanner uses an infrared signal; it corresponds well to the
intensity patterns observable in infrared images.
In the winter flight the difference between the two main tree
species is much lower which can be partly explained by climatic
conditions: The summer flight was undertaken after a long period
of hot dry weather, while the winter flight followed a quite humid
period. Obviously dry and eventually dusty needles are poor
reflectors of the laser signal.
Figure 6 Number of reflected laser dots per square meter as gray coded images. The values range from 0 (black) to more than 50 dots
Proportion of crown coverage of coniferous trees
The 50 stands were used to estimate the proportion of conifers.
For reference, the data of the forest management plan were used,
checked, and eventually corrected with aerial photographs. The
range is from O to 100 % in units of 10 %. The overall accuracy is
unknown but is assumed to be less than +10 %. The results of the
stepwise variable selection regression for the crown coverage
proportion of coniferous trees, Cer, is as follows:
Cer = 83.0 - 0.989 py, — 0.0202 py; > + 0.00733 pyr - pr (1)
Figure 7 Crown coverage proportion estimated from laser
data plotted against reference data which are in 10 % steps
with a regression coefficient R = 0.95 and a standard error of the
estimates of +13.0 %. Again, compared to the accuracy of the
reference data, the result is very good (Figure 7). There is a trend
to underestimate the proportion of the conifers, which is partly
due to the fact that holes are not considered in the method and
thus their area is grouped with the deciduous trees. Taking into
account the accuracy of the ground samples the results are very
good (+13% is scarcely less than the accuracy of the reference
data). Note that equation (1) allows for negative values or values
higher than 100% which certainly needs to be corrected. Yet, the
method is simple and reliable.
(white). Left: Winter last pulse; Center: Summer first pulse; Right: Digital ortho image.
