® open canopy closure decreases the relationships
between ground parameters and spectral signatures
of stands. Hence, if relationships are diminished
regionally, then this information should be taken into
account under supervised classification.
Aerial colour infrared photographs are the second source
of reference information on stand characteristics. These
photographs were taken for large areas of the
Karkonosze-Izerskie Range test area in 1984, utilizing
Kodak Aerochome Infrared Film. Detailed stereoscopic
interpretation of the photographs resulted in classification
of four levels of damage of spruce stands, hardwoods and
mixed stands, as -well as various phenomena
accompanying forest decline: 1) openings, 2) windbreaks
and windfalls, 3) loose canopy closure, 4) afforestations
and 5) regenerations.
The aerial photographs significantly improved the analysis
of satellite data. They made possible the determination of
spatial extent of the different factors characterised in the
field for the classification of Landsat satellite images.
3. KARKONOSZE - IZERSKIE RANGES TEST
AREA
Aerial colour infrared photographs taken in July 1984
were the supplementary source of information on the state
of forests within the study area. A map of forest health
prepared from these photographs, was used as a the
reference for making appropriate selection of training and
test areas, used further under supervised classification of
satellite data. The classification was done, utilizing Polish
and foreign experience in the field of analyzing forest
areas on satellite images (Horler et al.,1986, Kadro, 1988,
Rock et al, 1986, Wiliams et al, 1986, Zawila-
Niedzwiecki, 1994).
As a result of classifications three maps of the
test area were prepared:
- based on 1984 TM image
- based on 1990 TM image
- based on 1976 MSS image.
After thorough analysis of TM data 9 forest
categories were distinguished:
. Healthy / slightly impaired spruce stands
. Moderately impaired spruce stands
. Heavily impaired spruce stands
. Clear-cuts
. Openings
. Vegetated openings
. Deciduous stands
. Mixed stands
. Dwarf mountain pine.
« o 400. »0N-—
Classification image based on 1990 satellite data reveals
serious destruction of forests at western part of the study
area. Extensive areas were totally deforestated in this
region. Three levels of deforestation were presented at
the classification image:
e clear-cuts - deforestations with bare soil, i.e. the
areas, where dead trees were recently removed:
e openings - deforestations with some decayed trees
and sparse vegetation;
e vegetated openings - old deforestations already
covered with grass, shrubs and regeneration.
It should be mentioned, that deforestations also cover
large areas outside Poland (in Czech Republic). It can be
it
e
easily observed in he form of bright patches on the map.
?
As far as other forest classes are concerned, quite healthy
and slightly damaged forests cover only small valley
areas. Dying and dead stands can be still seen on the
classification image, but these areas will be also
deforestated in the near future. Distinct stratification of the
study area can be observed: eastern part is characterised
by larger contribution of slightly and moderately impaired
stands, while western part is heavily deforestated.
In order to assess dynamics of forest destruction 1976
Landsat MSS image was analyzed. Despite of coarser
resolution of that image, compared to TM imagery, three
levels of damage to spruce stands were discriminated, as
well as openings and non-coniferous categories. Then
1976 and 1990 classification images were compared, in
order to assess extent and character of forests changes in
this region.
4. DISCUSSION OF RESULTS
Comprehensive analysis of Landsat TM satellite images
for 1976 and 1990 shows that spectral differentiation of
spruce stands in this region permits discrimination of
three classes of forest quality. Class No 1 includes
healthy and slightly impaired stands, with loss of needles
up to 30%. Initial phases of spruce damages cannot be
separated from healthy stands with the use satellite data.
Hence, training fields defining stands with loss of needles
0 - 30% should be treated as representatives of one forest
category.
Accuracy of discriminating the succeeding classes of
spruce damage (> 30%) is dependent on local
conditions: canopy closure, soil cover, understorey
vegetation, etc. For instance, if stands with open canopy
closure contain openings with vegetation cover, the
spectral signature for these stands changes towards a
better class of quality. Generally, it is possible to
distinguish heavily impaired stands with loss of needles
30 - 60%, as well as dying and dead stands, which lost
over 60% of needles.
It is worth to mention, that young coniferous stands were
delineated well on satellite image from 1984. This class
includes mainly spruce stands (and partly pine stands) at
the age 7-25 years, covering 9% of the Karkonosze and
Izerskie Ranges test area. The other classes, which are
marginally discerned on satellite images, i.e. grasslands
and afforestations, were well discriminated. As
afforestation areas cover about 796 of the total classified
area, this class significantly characterises structure of the
forest areas.
The main forest categories were classified with high
accuracy, exceeding 80%. Taking into account complex
natural phenomena, existing in the region we conclude,
that the analysis of both structure and health conditions of
forest at the Sudety Mountains has acceptable accuracy.
Comparative analysis of two classified images -
Karkonosze and Izerskie Ranges 1984 and 1990
enabled an evaluation of qualitative changes in the forest.
802 International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
in £f» f^» ss — 14
na
€Y ON à mA
