ing and
certain
t. After
er areas
1e flood
/ers and
lenburg
lenburg
erbruch
ankfurt
bildkarte
nporaler
) der
Yon aus
-Aufnahmen
urt/Oder
04 2004 A
EI
orer
0107
4 vam 20 08 96
7
om 230797
om 30.07.97
u& Kruger
The determination of the accuracy of the extracted water areas has
not yet been finished and will be performed on the basis of a
comparison with the aerial imagery taken during the flood event.
The first results confirm, that in general the error of the
determination of the water-land boundary is nearly of the same
magnitude as the geocorrection rms error and depends mainly on
image resolution. Misclassification are in the frame of about 3 to
7.5 % and can be connected with vegetation standing in the water
of flooded areas and with wind effects on the water surface.
2.1.4 Land cover determination. The investigation of the
land cover was carried out on the basis of multitemporal Landsat-
TM images. The data used were acquired in 1995, 1996 and 1997
at different vegetation stages. For the determination of the classes
to take into account for the classification an experimental land
cover class nomenclature of the Bundesanstalt für Gewässerkunde
was applied, which was especially designed for hydrological
investigations and includes 13 relevant classes:
Dense built-up
Loose built-up
Decidouos forest
Coniferous forest
Mixed forest
6. Agricultural fields
7. Grassland
8. Orchards, Winery and tree nursery
9. Fens and heath
10. Water
11. Bare soil
12. Dumps. open cast mines
13. Rocks, glaciers
Nd 2 PD =
In the study area 11 out of these 13 classes could be found, rocks
and glaciers (class 13) as well as orchards and winery (class 8) did
not exist in the given region.
The data processing was carried out in the well known multitem-
poral land cover classification steps including the determination of
vegetation indices like the NDVI, the correlation assessment and
reduction with PCA and the hierarchical supervised classification
based on an maximum likelihood criterion. The classification
results after each hierarchical step were evaluated using ground
truth information from different sources (topographic maps,
CORINE data. airborne images. field data). This comparison led if
necessary to modifications in the supervising reference class
signatures. The classification of urban areas, which is especially
difficult with remote sensing data, was supported by an interactive
correction of the settlements based on the PCA transformation of
the Landsat- TM data.
2.2 Photogrammetric analysis
The photogrammetric data processing was based on the
following sources:
e aerial color images (scale 1:23000) of the area Guben-
Hohensaaten
e airborne scanner images, taken by DPA, with a ground
resolution 0.27 m of the area Hohensaaten-Stützkow
2.2.1 Photogrammetric processing of the aerial color
images. In the preprocessing phase the images were scanned
with an resolution of 15um in 3 bands. For a better data
handling only the red band was taken into digital photogram-
metric processing. The necessary number of ground control
points have to be digitized from topographic maps, because
there were no field measurings during the flood-period. There-
fore the product data precision is limited to the acurassy of the
used 1:10000 scale topographic maps. In addition to that the
- ground control point definition was difficult because of the
water cover and cloud shadows.
As a result there were found a sufficient number of ground
control and tie points for bundle adjustment, DEM generation
and stereo measuring. The DEMs will be used for
orthophotorectification. On its base we document the water-
land boundary at the moment of data acqusition in three
dimensions. This helps to calibrate flood models for future
flood warning and management systems.
2.2.2 Processing of digital airborne scanner data. During
the photo flight missions the new digital photogrammetric
equipment DPA was used for data aquisition too. With its 7
channels it allows to store panchromatic stereo images for
measuring aims and multispektral nadir images for thematic
work simultanously.
During the photo flight data from inertial systems were
recorded synchronously to the image data. This allows to
adjust image geometry which was disturbed by flight motions.
The resulting image becomes geocorrected by integration of
DGPS data and, where available, by ground control points.
The storage of inertial data allows to take data along a nonli-
near track, for instance the river axis. The resulting line
position error can be eliminated in the preprocessing phase.
This technology was tested on the Oder river in the area of
Frankfurt/Oder to Küstrin and in the area from Hohensaaten to
Gartz. In the preprocessing phase there were some problems
with the INS data intergration into image data. In the vicinity
of Hohensaaten the Oder river turns about 80? to the right. For
INS data integration this causes a change of the x- and y-axis
directions. Therefore the image had to be split into parts
before and after the turn and process them separately.
Additionally, during such a flight the plane motion is much
more dynamical, so a greater number of image lines had to be
considered in a single preprocessing step. As the operational
memory for this procedure was limited a lower resolution than
the original data had to be selected to process them.
The resulting rms error in geocoding of the whole image is
1.8 m thus being smaller than the cartographic precision of the
ground control point source. The data product is planned to
use for documetation of the flood and for calibration of flood
models, too.
3. DEVELOPMENT OF AN INTEGRATED SPACE
BASED FLOOD RISK INFORMATION AND
MANAGEMENT SYSTEM (FRIMS)
As mentioned above, one of the experiences of the Oder river
flood was the lack of an efficient and practicable information
system for assessing and forecasting the flood situation in the
total Oder catchment area. Besides the need to improving
hydrodynamic models (precipitation outlet) we have also to
consider problems concerning the international data
availability and the possibilities of the data exchange and an
insufficient consideration of Earth Observation data within the
decision making processes.
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 187
