50
play an important role for the
classification.
For the results under discussion the
UN-ECE forest damage classification
scheme was used as an European
standard, though structural elements
are not yet considered in it.
yellowing
leaves,needles
loss
11-25%
26-60%
61-100%
< 10 %
0
1
2
11-25 %
1
2
2
26-60 %
2
3
3
> 60 %
3
3
3
lead tree
4
Table 1 UN-ECE forest damage
classification scheme for trees.
Class definition 0: healthy, 1: low
damage, 2: medium damage, 3: heavy
damage, 4: dead
The principal differences in the
evaluation of visual features for
damage classification: leaves or
needles loss, discolouration on one
side, structural elements in addition
in the other side indicate a general
problem in photo interpretation.
1 Problem definition
The analysis of multispectral scanner
data which was the main goal of the
BMFT project, it is distinct from the
photo interpretation for the follo
wing reasons:
The spatial resolution of scanners is
defined by the instantenous field of
view (IFOV) or the image element. Its
size is determined by the sensor's
field of view (FOV) and the flight
altitude.
For the image element the integral
spectral radiance is recorded. The
integral reflected spectral radiance
may be produced from forest cover of
one species, generating so-called
"pure pixels", or surface elements,
which contain also portions of other
plant communities, road network,
gravel etc. , resulting in so-called
"mixed pixels", with their conse
quences on spectral analysis. A
comparison of CIR-film and multi
spectral data performance for forest
damage evaluation is made in table 2.
Multispectral scanner data cover a
wider spectral range than CIR-photos.
For interpretation purposes the
spectrum of vegetation can be devided
into three rather distinct regions
Johnson, 1969; Sinclair et. al.,
1971; Walter et al., 1981; Rock et
al.; 1986. In the visible region
(400-700 nm) the main part of the
radiation energy is absorbed by plant
pigments of the upper leave or needle
layer. Reflection and transmission is
rather low. In the near infrared
(NIR, 700-1300 nm) the reflection of
green vegetation is high, depending
on the species and number of layers.
Reflectance is influenced by cell
structure background and water
content. In the short wave infrared
(SWIR, 1300-4000 nm) reflectance can
be attributed to water content, cell
structure and background. Leaf area
index in relation to canopy
reflectance is more important in the
infrared than in the visible.
Spectral reflectance depends on plant
species, age, form and orientation of
leaves or needles, branches, stem,
background and health status, Koch,
1987; Hermann et al., 1988, Kirchhof
et al. 1988, Hoffmann et al. 1989.
On the ground are soil cover, surface
roughness, relief, mineral supply,
humidity and heavy metal content main
parameters, which influence the
optical behaviour, Collins, 1983;
Kronberg, 1985, The saisonal
variation of spectral signature of
plants is described by Hildebrandt,
1976, and Tanner et al., 1981.
2 Objective of spectral measurements
In support of the cooperative
research project of the BMFT
additional spectroradiometer
measurements became necessary.
Spectral analyses of forest stands in
the test site Stadtwald Frankfurt
revealed the importance of primary
and/or secondary effects of damage on
the change of spectral signatures,
Guttmann et al., 1987. More detailed
information could not be derived from
multispectral scanner data of pixels
sizes 5 x 5 m 2 to 10 x 10 m 2 .
As primary effects of damage were
identified
- change of the spectral signature of
tree components (branches, leaves,
needles, barks, lichens)
- discolouration of leaves, needles
- loss of biomass, leaves, needles
- orientation of branches, leaves,
needles, roll of leaves
- change of crown structure and
texture, anomalous ramification
The course of the spectral signature
will be changed by above mentioned
primary effects. Loss of biomass,
change of crown structure and
anomalous branching produce an
increase of optical transparency, of
shadow portions and as a consequence
an augmentation of background
radiation as secondary effects.
For the improvement of our understan
ding of spectral signature changes by
tree damage, a measurement program of
tree components was developed. The
central theme is the understanding of
reflectance behaviour of tree com
ponents in different compositions
(layers), and its application for the
selection of spectral bands and image
processing algorithms for multispec
tral classification to optimize da
mage identification and separation,