4.3 GIS-Analysis
Based on the existing sociological studies of macrophytes
and the own experiences conditions for the incidence and
growth of macrophytes were defined. They depends
mainly on sediment, water depth and related conditions
like aspect and slope. This ancillary data were used
together with the images for the classification.
Another GIS application is to overlay the data sets for
change detection. For this macrophyte coverages
obtained by the classification of the aerial photographs
from 1992 and 1995 were overlayed (Fig. 2).
Macrophytes, May 1992
EM Macrophytes, May 1995
i Land
Fig. 2.: The change of distribution of the macrophytes
between 1992 and 1995
5. RESULTS
The processing of the data of different sensors illustrates
that the acquisition time of the images is very important
for the quality of image interpretation and classification. A
high amount of chlorophyll and suspended material in the
water falsifies the spectral reflectance of the macro-
phytes. In addition they decrease the visibility of the
ground. In the Landsat TM scene from July the macro-
phytes differ from the background maximally up to 3.5 m
water depth. The visible macrophytes could be separated
only in 2 classes. The low resolution of the scene does
not permit a sharp demarcation of macrophytes in the
Greifswalder Bodden. The population areas especially
near the coast are too small for the rough resolution. That
means a lot of mixed pixels.
The Spot images represent the boundary of macrophytes
better, that's why Spot scenes can be used as a filter
mask for macrophytes. A classification and separation of
the vegetation under water was not possible with the Spot
scenes.
Photo flights offer great possibilities to get high quality
photographs in accordance to the requirements of the
application; scale of the photos, film material, filter and
others can be considered. A further advantage of photo
flights is the flexibility in flight planning. Thus, the optimal
time can be chosen.
144
The spectral classification of the aerial photographs did
not bring satisfying results in spite of carefully selected
training areas. Up to 2 m water depth the classes are be
very good separable. If the water depth increases the
classification accuracy decreases. Deeper than 4.5 m the
variance of the spectral class values of the macrophytes
and the background are higher then the distance of their
class centers. Therefore a definite delimitation of macro-
phytes is impossible in this depth only by spectral classi-
fication. The combination of ancillary data together with
remote sensing data within a GIS supports the classifica-
tion process. The use of ancillary data improved the
classification results on an average of 9 percent.
The evaluation of the vegetation changes (Fig. 2) indicate
seemingly a high decreasing of macrophytes from 1992
to 1995. A check of some test sites in summer of 1995
did not supported this result. In the time of the photo
flights the vegetation was not enough developed because
of cold and stormy weather conditions before.
6. CONCLUSION
All tested sensors can be used in the digital mapping of
macrophytes. The use of a special sensor depends on
the application. The best results come from aerial photo-
graphs taken by a photogrammetric camera in the scale
of 1:10000. This photos represent an acceptable area of
ground. That facilitate the view over the area, the rectifi-
cation and the georeferencing. The processing of images
is more effective than the processing of the 'Low-Cost'-
Images taken by an amateur camera in the scale of
1:5000. The Low-Cost-Images should be used only for
small areas close to the shore because the rectification
and geocoding is very difficult. A lot of images could not
be georeferenced without a high number of put out con-
trol points in the water area. On the other side the classi-
fication results are finer than from the aerial photographs.
The self-made photoflights with an amateur camera are
very flexible and their costs do not exceed one fourth of
the photoflight with a photogrammetric camera.
The use of satellite imagery in the Greifswalder Bodden
for mapping macrophytes is limited because of the low
resolution (Landsat TM) and/ or the absence of multi-
spectral bands (Spot).
The integration of remote sensing data within a GIS is
necessary to combine the advantages of several sensors
and to compensate their disadvantages, which are
increased by special properties of under water research.
Moreover the integration improves the quality of the
classification results and approves the comparability of
data at different times.
7. ACKNOWLEDGMENT
This work was sponsored by German Science Foundation
(DFG) and supported by the Deutsche Gesellschaft fuer
Luft- und Raumfahrt (DLR) and the research group
Unterwassertechnik.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
