photogrammetry which drastically improves the penetration
rate.
e the high point density of laser scanner systems, at least
partly compensates for low penetration rates, thus even in
dense forests there may be enough information available to
create good ground elevation models.
However, there are still some major problems with laser scanners
that must be overcome or are inherently unanvoidable;
e the penetration rate may be nearly zero in areas with
extremely dense vegetation (young deciduous trees);
e the huge point number and the totally arbitrary distribution
along with the massive shift towards off-terrain elevations
needs new approaches to interpolate DEMs. This is
comparable to the unqualified point distribution of image
matching techniques which does not always lead to correct
DEMs.
Laser scanning in its practical application is a very new
technology. It has the potential to provide information far beyond
the mere creation of DEMs, particularly in fields outside the
photogrammetric community. Thus the technology needs to be
promoted to those areas, and algorithms have to be provided that
will allow other disciplines to benefit from the information
available with laser data. One already well-known field of
application is the creation of digital 3D city models, mainly
driven by the mobile telephone companies. In this paper, the
focus is on the application of laser scanner data in forestry.
2 LASER SCANNING IN FORESTRY
As mentioned, laser scanning has great potential to provide
information about vegetation, therefore it may well be used in
forestry. In contrast to photogrammetry, or more general, remote
sensing — which focuses on the thematic aspects in a geo-coded
context — the main application of laser data lies in their geometric
aspects. With photogrammetric methods, the crown surface can
be measured; the lack of ground information, however, makes
estimation of tree heights very inaccurate. The great advantage of
laser scanning in this regard is its ability to penetrate vegetation.
Therefore the two techniques complement one another.
Forestry usually needs tree information on a stand basis. A stand
in the context of a forest enterprise is an area that is dealt with as
one unit in planning and working. The borders of these units
frequently follow roads, ridges, valleys, or arbitrary — often
simple straight — lines. There may or may not be a connection to
the terrain. The size of forest stands depends on several criteria,
such as silviculture-methods, steepness, tree species, infra
structure, or even legal aspects. The vegetation within a forest
stand may change due to differing conditions (valley — ridge; soil
changes; etc.) or operative actions within the forest enterprise.
In contrast to the "economic unit", laser scanning simply provides
geometric information. Thus laser data can be used in two ways:
Either one can take the stand borders as provided by the forest
inventory and analyze the laser data within these stands; or one
can comprise areas of similar geometric characteristics, i. e.
International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 3W14, La Jolla, CA, 9-11 Nov. 1999
similar tree height, density, and species, which yields
"geometrically homogenous stands".
It is not yet clear which method better fits the needs of forest
enterprises. This certainly depends on several aspects and up to
now there is little experience available. It may well be that laser
scanning will change traditional techniques of forestry, especially
along with legal actions. An example; in mountainous countries
like Austria there is a growing awareness of the dangers of large
clear-cuts on steep hillsides. Thus there is a tendency towards
uneven aged stands, cutting only particular trees from time to
time. It is very difficult to estimate timber volume of such stands
from terrestrial or photogrammetric measurements.
For the following research, "geometric stands" are delineated
from laser data along with aerial photographs in order to
determine about the characteristics of different tree species,
heights, density etc. This information can later be used to find out
about the real, often fairly mixed stands.
3 RESEARCH AREA AND DATA
The research area chosen encompassed the eastern part of the
Vienna University of Agricultural Sciences research forest, 60 km
south of Vienna (Figure 1). The area of interest covers about 10
square kilometers. The terrain is hilly with elevations between
300 to 700 m above sea level. Average hillslope is 34 percent,
with extremes up to 100 percent. The following data are available
for the region:
e A precise forest inventory;
e Tree and stand information (height, species, breast height
diameters, vitality, leaf area index measurements, etc.) for
over 1000 trees;
e three laser scanner flights, two summer flights with first and
last reflected pulse recorded respectively, and a winter
flight (leafless period) with last pulse recorded. The system
used is the TopoSys scanner with ground resolution of 10-
15cm in flight direction and 1.5 meters across. Approx. 340
million laser dots were collected;
e photogrammetric false color infrared imagery from both
summer and winter flights (scale 1:10,000);
e terrestrial measurements from a survey during the winter of
1999 with approx. 2000 ground points and data from more
than 1000 trees.
The research forest reflects vegetation typical for Central Europe:
Mixed and pure stands of coniferous and deciduous trees with
mainly spruce and beech. In between, but less important, fir, oak,
larch, pine, and an insignificant number of other deciduous trees.
Laser scanners as used in this research provide only geometric
information. For each recorded pulse three coordinates, X, y, Z are
calculated. No radiance information is collected. Therefore only
geometric characteristics may be extracted; however, there is à
relation between several (thematic) stand parameters and
geometric aspects of the laser data.
Internation:
4 CHARAC
4.1 Laser patterns of
The main "geometric c
dealt with are
e mean stand heigh
® tree species, grou
e proportion of bas
SD
Figure 2 Perspecti
show the lines of
h
