Terrain with pronounced relief variations
strongly influences radar backscattering. Ra-
diometric distortions then have to be removed
by value added products such as local inci-
dence angle and local resolution maps, which
are calculated directly from the DEM by recon-
structing the original illumination geometry
from each individual backscatter element. This
information is transformed into the initial
SAR geometry and used to derive the backscat-
tering coefficients c? and y (Haefner et al.,
1994). These preprocessing steps are included
in the ORA scheme (Fig. 1).
Date/Time [UTC] Sensor Type Orbit/Frame
16.01.93/10.10 ERS-1 SAR PRI desc. 7868/2655
16.01.93/21.30 ERS-1 SAR PRI asc. 7875/927
15.02.93/09.30 L-5 TM floating 194/27-28
20.02.93/10.10 ERS-1 SAR PRI desc. 8369/2655
20.02.93/21.30 ERS-1 SAR PRI asc. 8376/927
27.03.93/10.10 ERS-1 SAR PRI desc. 8870/2655
27.03.93/21.30 ERS-1 SAR PRI asc. 8877/927
01.05.93/10.10 ERS-1 SAR PRI desc. 9371/2655
01.05.93/21.30 ERS-1 SAR PRI asc. 9378/927
22.05.93/09.30 L-5 TM floating 194/27-28
05.06.93/10.10 ERS-1 SAR PRI desc. 9872/2655
05.06.93/21.30 ERS-1 SAR PRI asc. 9879/927
10.07.93/10.10 ERS-1 SAR PRI desc. 10373/2655
10.07.93/21.30 ERS-1 SAR PRI asc. 10830/927
21.11.96/05.45 Radarsat SCN W1/We desc. 5469
21.11.96/17.06 Radarsat SCN W1/W2 asc. 5476
16.03.98/05.45 Radarsat SCN W1/W2 desc. 12329
19.03.98/17.19 Radarsat SCN W1/W2 asc. 12379
Table 2: Satellite data sets
4.3 Change Detection
Calculating the ratio between the backscatter-
ing values (dB) of ORA derived synthetic SAR
images, and the y-values of a completely snow-
free or dry-snow reference scene respectively,
the wetness can be mapped and monitored. From
these wetness maps it becomes possible with
the aid of the DEM to deduce the wet snowcover
and to assess the areal extent of the total
snowcover. The ratio of the backscattering
coefficients corresponds to the difference
between the y-values expressed in dbs
(PIESBERGEN et al., 1997). Although the MORA
approach offers good prospects for an area-
covering snowcover determination in open areas
of rugged terrain, several sources of errors
have to be considered, such as a mix-up of wet
snow and bare wet soil, the influence of woody
vegetation, especially forests, and of sea-
Sonal vegetation changes, etc.
5. DATA FUSION
5.1. Basic Concept
To improve the quality of the snowcover moni-
toring, data fusion techniques were applied.
In a first step, the snowcover monitoring is
carried out with SAR-magnitude data. Secondly,
the snowcover is mapped from EO data of ap-
prox. the same acquisition date. Then these
two sensor specific snow maps - geocoded to
the same cartographic reference system - are
used as input in a simple fusion model. The
merging and data processing scheme is illus-
trated in Fig. 2.
5.2. Colour Transformation Method
A simple visual interpretation can be reached
by displaying the data in a RGB-colour cube.
Applying a RGB-YUV-RGB colour transformation
Y 0.299 0.587 0.114
U| = |-0.147 - R* |-0289| - G+ |0.437| : B (Eq. 1)
V 0.615 -0.515 0.100
R 1.140 0
G| 2 Y* -0394|:U * -0,581|: V (Eq. 2)
B 2.028 0
the calculated Y-channel (intensity) is re-
placed by the normalized SAR backscatter val-
ues y. The resulting RGB image then includes
the originally detected snow coverage from the
EO data as well as information on the wetness
conditions, resulting from the SAR-sensor
(Piesbergen & Haefner, 1997).
6. RESULTS
Numerous classification results such as se-
quences of snowcover maps have been produced
and partly published (Haefner et al, 1993,
Piesbergen, Holecz & Haefner, 1997).
Only a few selected examples shall be illus-
trated here to demonstrate the possibilities
of monitoring snow with SAR data, and the data
fusion principles. They all originate from the
medium-size test area GRISONS. In Fig. 3 part
of the original Landsat TM respectively ERS-
SAR images are illustrated.
352 International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
m
isi IC
Fig
Fig
