he 
ral 
(v 
Dr 
—r 
Dr 
INTRODUCTION 
The hydro-electric power plant at Churchill Falls, 
Labrador (Canada) is one of the largest power 
generating facility in the world with a drainage 
area of 69 267 km“ containing five reservoirs. 
Monitoring the inflow from such a large 
watershed area constitutes an important part of 
the water management. This monitoring at the 
present is conducted by field observations in 
snow courses during the winter months. The 
collection of field data is very expensive as most 
of the snow courses can be reached only by a 
helicopter and the results do not provide 
accurate results. The field observations could be 
supplemented by remotely sensed data 
collection to improve the run-off modelling and 
could make the procedure less expensive and 
more accurate. 
Researchers suggested that snow-water- 
equivalent can be related to multi-channel 
RADAR returns from the snow fields in open 
areas (NASA , 1981 and Shi et.al., 1990). Since 
the Canada Centre for Remote Sensing (CCRS) 
introduced a RADAR development program 
with the use of an air borne X- (A=3.24 cm) and 
C- (A=5.66 cm) band SAR we could designed an 
experiment at Churchill Falls (in situ) to 
investigate the usefulness of this system to 
derive SWE. The RADAR development 
program was the fore runner for the application 
of a new Canadian RADARSAT satellite which 
was launched in late 1995. 
The experiment was designed such that SAR and 
simultaneous ground data were collected in pre- 
determined sites at several time intervals (with 
and without snow cover). Statistical analysis of 
the ground and RADAR data should provide an 
estimate of SWE that could be built in a 
hydrological model to predict spring run-off. 
EXPERIMENTAL DESIGN 
The Earth Observation System approach 
(NASA, 1981) uses L-, C-, and X- bands SAR 
simultaneously over a snow field. This provides 
three simultaneous equations from which the 
37 
SWE can be calculated. Since we had only two 
bands (X- and C-) we planned to obtain SAR 
data without snow cover after the freeze-up and 
during the maximum snow cover. This would 
replace the use of the L- band. 
Bernier and Fortin (1991) evaluated the potential 
of C- and X- band SAR to monitor dry and wet 
snow cover. More recent applications of C- 
band SAR for mapping melting snow (Donald 
et.al. 1993) and mapping of discontinuous 
permafrost terrain (Granberg, 1994) were 
reported in the Canadian Journal of Remote 
Sensing. 
The determination of SWE requires the use of 
calibrated SAR data. The incidence angle is one 
of the most determining factor in analysis, 
therefore, we chose two incidence angles (30^ 
and 60^) for our data acquisition. 
Since the ground data collection had to follow 
the SAR acquisition immediately, we could use 
only a limited number of test sites. These sites 
were chosen in open areas near to access roads. 
Each target area consisted of a 60 to 120 m long 
base line. Snow parameter data were to be 
obtained at 6 m intervals along the base lines to 
coincide with the CCRS SAR ground resolution 
element. 
On the advice of CCRS scientist we choose a 
regression analysis design such that the ground 
data (SWE) are correlated with single band and 
multiple band multi-temporal SAR data. 
Corner reflectors were manufactured for the 
absolute calibration of SAR data. Two sets of 
two reflectors (one for the X- and one for the C- 
band) were placed in locations corresponding to 
30? and 60" incidence angles. 
DATA ACQUISITION 
Air Borne SAR. 
The snow free data were collected on October 
30th, 1990 after the freeze-up with minimal 
snow cover (less than 10 cm). This flight 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996