ross 
Y 
ed scan 
'Picted to 
;, Onstott 
tween FY 
ximately 
is been 
| (1982b), 
in 
1d 
[ly with 
shows a 
in the 
jand in 
* contrast 
>and being 
his 
7 degraded 
al result, 
Hr to 60° 
ashar, 
sons 
vantage in 
reater 
2 of 
band to 
types 
ce class 
example 
ar both 
  
  
  
incidence angle and frequency through the Rayleigh roughness criterion*. For 
this reason, ice-deformation features (ridging, rafting, hummocking, etc.) are 
generally more prominent at the lower frequencies and always dominate at grazing 
geometries. Radar geometry indeed affects the imagery of ice types and ice 
features as incidence angles vary (Gray et al. (1982a). At X-band a range of 
grey tones appears to prevail in these cold conditions allowing ample class 
separation until Ô = 859 (where we see in the super-shallow geometry that the 
rubble and ridges are white and everything else, black) implying a roughness 
scale of 3 to 4 cm. Schertler (presentation at the SURSAT Workshop in May, 
1980) has, to our knowledge, done the only quantitative study of the behaviour 
of ice radar cross section which spans the full range of incidence angles. This 
work which is more of interest to airborne applications (which use large 
incidence angles to generate wide swaths) showed that although a steady decay of 
absolute signal continued to large incidence angles (0.3 dB/9), contrasts of 
nearly 8 dB remained, until at X-bani, near 9 - 859, the slope ofc? (9) 
steepened. 
3.3 Polarization 
Results given by Onstott et al. (1979) and plotted in Figure 5 show 
that generally VV** responses are higher than HH with the contrast between 
polarizations decreasing with incidence angle. It has been suggested that this 
effect may be related to varying degrees of roughness in the two corresponding 
ground directions. The greater difference with FY over MY with polarization 
seems consistent with the statement. Both polarizations will agree at 
8 2» 0° because the polarizations are indistinguishable. Results from open ocean 
experiments (Jones et al., 1977) also show greater returns at VV than HH 
increasing with incidence angle to about 5 dB contrast at 6 = 60°, 
  
  
Our Ku-band scatterometer results show cross polarization gives 4 to 
7 dB more contrast between FY and MY ice than does like polarization. This 
should be compared to Onstott's results which show even greater advantages in 
contrast with cross polarization ranging from 8 to 10 dB for the conditions he 
studied as shown in Figure 6. 
This advantage in available contrast with cross polarization may not 
however lead to better ice type discrimination since tests show (Guindon, 1982) 
that maximum likelihood classifiers perform equally well using either HH or HV 
data. In addition, probably due to a more suitable match between the dynamic 
range of radar image brightness and photographic printing, like-polarized 
imagery may be preferred (Spedding et al., 1982). For these reasons, future 
operational radars will probably favour a VV or HH configuration rather than pay 
the cost of -10 db in absolute power level for additional, but perhaps 
unnecessary contrast available with cross polarization. 
  
* This may be summarized with the relation 82,/X cos 8«1 for smooth surfaces 
where 2/24 is the ratio of roughness scale to wavelength scale, and 9 is the 
incidence angle. From this relation, as 9 approaches 909, all scenes become 
smooth regardless of frequency. 
** Here and throughout, the polarization, is indicated by letters (either V for 
vertical, or H for horizontal) with the first letter referring to transmit 
polarization and the second, to receive polarization. 
793 
Es "d 
i M E EE E DR NN E IEEE EH Ss LL 
er er re EB EE 
  
PA 
es ulcere. um 
  
  
  
A