|
|
|
|
I
I
rologi-*
oplica-
tellite
relief
oristic
differ-
scale,
ization
i moni-
jes and
> suit-
sary to
goal of
1 tech-
of the
to the
ile in
à fully
pproach
sforma-
nd then
sources
A). The
onitor-
l prin-
possi-
ig dur-
'ess of
ins or
ibility
ly. But
e fre-
system
ational
re spe-
reduce
equency
allow a
extent
well as
It is
the purpose of this paper to demonstrate the
advantages and disadvantages of the use of EO
and SAR sensors, and of the synergism of these
systems, using data fusion techniques. In ad-
dition, characteristic problems have to be
solved when working on different scales, and
over a longer time period.
2. TEST SITES AND DATA SETS
To investigate the effects of scale on the
classification procedure of SAR data and the
accuracy to be achieved, as well as to evalu-
ate the possibilities and advantages of the
data fusion techniques, three different test
sites were selected in the Swiss Alps (Tab.
1). They reach from a very small, local area,
asking for very high precision, to a rather
large area covering almost the entire alpine
region of Switzerland. Digital Elevation Mod-
els (DEM) are available of all sites. Loca-
tion, size, elevation range and characteris-
tics of the DEM's are listed in Table 1. The
available satellite data sets are compiled in
Table 2.
In addition, snow parameters (e.g. tempera-
ture, depth, density, water equivalent) were
measured in the Davos test area during several
overflights. Meteorological records are avail-
able from various stations within or close to
the test sites.
3. CLASSIFICATION OF EO-IMAGES
Snowcover mapping and monitoring with Landsat-
or SPOT-data is operational and has been dis-
cussed repeatedly (Ehrler, 1998; Piesbergen &
Haefner, 1997; Brüsch, 1996; Haefner, 1996;
Steinmeier, 1995). :
After a careful geometric and radiometric cor-
rection of the TM data (Piesbergen & Haefner,
1997), a simple RGB composite was used as ba-
sis for a snowcover mapping for this study.
The channels considered were No. 4 (nir) as
red, No. 3 (red) as green and No. 5 (mir) as
blue, making use of the significantly differ-
ent reflectance properties of various snow
types in channels 4 and 5. A separation of
snow from clouds is reached with channel 5.
Reflectance values in the visible range
(channel 3) allow a clear separation of the
snowcover from the snowfree parts, due to the
relatively high contrast between snow and bare
Soil, pasture or forest.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
Testsite Size Elevation DEM
sq. km range msl grid size m
height re-
solution m
DHM Davos
DAVOS 10 1500-2850 | 10 x 10
0.1
DHM25
GRISONS 1260 600-3000 | 25 x 25
0.1
SWISS ALPS | -16'000 500-4500 | DHM50
50 x 50
1.0
Table 1: Test sites and its characteristics
Copyright DMH25: O Bundesamt für
Landestopographie (3017a)
Courtesy DHM50: TYDAC
4. SNOWCOVER MONITORING WITH SAR-DATA
4.1. General Approach
The application of SAR data in high mountain
,areas requires a new method to reduce the re-
lief induced distortions. It is based on the
radiometric and geometric correction of the
images, and on the generation of a new syn-
thetic image, using the optimal-resolution
approach (ORA). These resulting images, con-
taining the fully interpretable parts of
crossing orbits, are almost free of non-
interpretable parts caused by layover and ra-
dar-shadows, and have a better local spatial
resolution (Haefner et al. 1994). For monitor-
ing purposes an extension of ORA to the mul-
titemporal optimal resolution approach (MORA),
combining several ascending and descending
image pairs was achieved (Piesbergen et al.,
1997, Piesbegen and Haefner, 1997). The MORA-
concept is outlined in Fig. 1.
4.2. Preprocessing
The removal of relief induced radiometric dis-
tortions requires a high precision geocoding.
This geometric correction has to consider the
sensor and the processor characteristics.
Therefore, it has to be based on a rigorous
range-Doppler approach (MEIER et al., 1993).
It is obvious that the geocoding has to be
accomplished in the same geodetic-cartographic
reference system as for the Landsat data, and
which is also valid for the DEMs. For our pur-
poses, it certainly is the official Swiss Ref-
erence System.
351
