test 
n on 
d on 
ice, 
robe 
the 
the 
)AR. 
)uter 
jylute 
the 
nthin 
SNOW 
ct on 
ween 
ce of 
m of 
ation 
railed 
ers. 
; and 
The 
ayers 
iS 22, 
layer 
ature 
S OQ. 
| near 
ıbient 
nal in 
in the 
ırther 
f the 
vered 
n the 
snow and of return from the snow-soil interface, 
with the incidence angle being the most 
important factor affecting the strength of the 
signal. Since the test sites in this study were 
selected after the ground was covered with 
snow, in most cases we did not have uniform 
underlying soil surfaces. In order to obtain all of 
the required SAR data (two different modes and 
polarizations), the aircraft needed four passes 
each during the fall and winter. Table 1 shows 
that the incidence angles for a particular test site 
varied significantly. In addition, the X-band data 
were not collected during the maximum snow 
cover period which forced us to use only the C- 
band data. Another problem was encountered 
with the exact location of the test lines on the 
various SAR imagery. 
Our example, presented in this paper, shows that 
the C-band SAR alone cannot give a good 
indication on the SWE value of snow pack (low 
correlation coefficient). Although the regression 
line has a positive slope, the individual 
observations are widely scattered. This scatter is 
due not only to the variation of the snow 
properties and of the ground surface coverage, 
but also to the inherent ‘speckle’ of the SAR 
even over uniform surfaces. To obtain a more 
precise result, the averaging of more pixels is 
required. However, this would involve a more 
detailed observation of SWE on the ground as it 
also can vary on a short distance. Our general 
observation is that raw DN-s of C-band VV 
polarization provides a better result than the data 
of HH polarization. We also conclude that the 
averaging filter (3 by 3 kernel) provides poorer 
results than the use of raw DN-s. We hope to 
improve on our results with the analysis of the 
multi-temporal SAR data. 
CONCLUSIONS 
At the time of the initiation of the project, very 
limited information was available in the literature 
concerning the prediction of SWE using SAR 
data (NASA, 1981). We received diverging and 
sometimes contradicting advice from experts 
concerning our experimental design. Finally we 
adapted a simple linear regression design for the 
41 
establishment of relationships between SAR and 
SWE data. As we collected a very large amount 
of ground data on snow parameters and of SAR 
data (two incidence angles, two SAR bands, four 
different polarizations), the data analysis 
consumed a lot of time. In addition, we had to 
write our own computer program for the data 
analysis. We also encountered numerous 
technical difficulties, most of them out of our 
control, which made the data analysis a difficult 
task. We are currently in the process of using 
the fall C-band SAR data in combination with 
the winter SAR DN-s. 
Our overall conclusion is that the C-band VV 
polarization provided a better estimate of SWE 
than the HH polarization. The use of a 3 by 3 
averaging filter does not improve the precision 
of SWE estimation but actually decreases it. 
Although our results are not fully conclusive, we 
learned a great deal concerning the pitfalls of 
data collection and analysis. 
Our final conclusion is that the field data 
collection and the SAR data acquisition requires 
extreme precision. We are now in the position 
to conduct further experiments which would 
yield more conclusive results. We are planning 
to use the Canadian RADARSAT and the 
Japanese JERS satellite data to refine our 
procedure of SWE estimation. 
ACKNOWLEDGEMENTS 
We would like to thank the Canada Center for 
Remote Sensing (CCRS) for providing the SAR 
data acquisition under very adverse conditions. 
We acknowledge all the help we received from 
various scientists of CCRS, especially Dr. T. 
Lukowski for working out the absolute 
calibration procedure. Thanks are also extended 
to Mr. W. Smith, Site Service Manager, 
Churchill Falls (L) Co. for the logistic help and 
members of the Water Resources Management 
section for their help in the field data collection. 
The Faculty of Engineering and Applied Science 
of the Memorial University of Newfoundland 
(MUN) and the Canadian Forest Service (CFS, 
Newfoundland and Labrador Region) materially 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996