To test this possibility, areas of similar terrain (i.e., forest and mountain) and 
incidence angle were compared on Pass 1 and Pass 3 to see if look direction was a factor 
in detectability. Although it was not a comparison of the exact same terrain it was the 
same general region and terrain conditions. Specifically, the segments of each pass 
containing only mountains and forests were re-interpreted. The results are presented in 
Table III. Recall that the accuracies for the entire length of the Passes were: Pass 
1--62.7 per cent and Pass 3 -76.4 per cent. Note, however, when only the mountain/forest 
environs were studied the accuracy figures were reversed (i.e., Pass 1--77.5 per cent and 
Pass 3 - 62.3 per cent). From these data it is suggested that the overall higher accuracy 
for Pass 3 was due to the fact it passed through more plain and open terrain whereas Pass 
1 imaged more forest and mountain area producing lower detection numbers and accuracy 
figures. These conclusions are supported by a similar observation made by Lo (1986) using 
SIR-A data of the United States. In theory, Pass 1 should have lower detection figures 
since the higher incidence angle (52.3 degrees versus 40.9 degrees for Pass 3) would 
produce more radar shadows and grazing areas than the steeper incidence angle of Pass 3. 
If this observation is a logical assumption, then a major reason for the higher per cent 
accuracy and detection figures for Pass 1 must be the change in look direction. 
Also, the number and per cent commission errors were lower for Pass 1 than Pass 3. 
It is hypothesized that surface roughness may be more dominant than topography as an 
influence on signal return for Pass 1 reducing the possibility for detection error. However, 
the difference in the number of errors is too small to warrant a precise cause-effect 
relationship without additional supportive data. Note also that in the attempt to identify 
settlements in mountain/forest areas in Pass 1, twice as many commission errors were 
produced over a smaller area (36 versus 18) than in the original interpretation of the entire 
Pass. Undoubtedly, the factors of interpretation consistency, constant visibility, and 
identification confidence need to be examined in this regard. 
SUMMARY 
This study has examined some of the relationships among settlement detectability, 
the environment (target-background contrast), incidence angle, and look direction. Three 
L-band HH polarized SIR-B imagery data takes over south and central Germany and 
northeastern France provided a study area of over 20,000 sq. km. The observations below 
are presented as the result of this investigation. 
1) Steep (low) incidence angles of less than 20 degrees appear to be of minimum 
utility for settlement detection in northern boreal forest environments. 
2) The greatest number of commission errors was generated at the lowest incidence 
angle. In such cases, topography was a major factor along with signal penetration of the 
surface canopy in generating specular reflectors. Vegetation covered wetlands, cultivated 
field crops at surface water capacity, terrain slopes, terraces, and edges of mature forest 
stands all generated high returns that were mistaken, at times, for settlements.