INTRODUCTION
An understanding of the consistencies and changes in settlement detection ability
related to radar system and environmental parameters is of potential interest to diverse
segments of the scientific community. On one hand such data can supply useful input to
environmental monitoring efforts and population assessments at many levels--particularly
where such information is dated or absent. Settlement detection and data inferred from
such data (e.g. population estimates, urban growth models) are important factors and
desired by many resource management users--especially those in developing nations. On
the other hand the correlations between urban morphology (structure characteristics) and
changes in radar signal response are of interest to scientists attempting to define and
model the theoretical nature of system-target relationships and specular reflection
parameters relevant to semi-automated interpretation techniques.
In this experiment SIR-B imagery of central Europe was examined to explore the
influence of incidence angle, look direction, and environment-
terrain conditions on settlement detectability. Although previous efforts in this regard
have contributed to an eclectic collection of data, few have been able to employ multiple
images of the same area. Using airborne K-band imagery of five New England study areas
Simpson (1969) found that 100 percent detection was possible for cities over 7,000
population. Henderson and Anuta (1980) examined K-band and X-band airborne imagery of
diverse areas in the United States as well as one Seasat-SAR (L-band) image of central
Pennsylvania. With the former imagery a scale of 1:200,000 was determined necessary to
attain consistent detection of settlements with a population of at least 1,000. However,
only a 68.3 per cent accuracy for similar size settlements was attainable with the
low-incidence (ca 21-23 degrees) angle Seasat Image. Drake and Patton (1980) stated it
was difficult to identify towns on Seasat imagery of the Delmarva peninsula as the signal
return was similar to that from agriculture and forest vegetation. The specular returns
from settlements were found to be similar to and confused with such structures as poultry
barns.
Dowman and Morris (1982) were able to detect around 16 per cent of isolated rural
buildings, 37 per cent of houses, and 70 per cent (HH polarization) and 87 per cent (HV
polarization) of high rise buildings using X-band imagery of an area west of Ottawa,
Ontario. Plotting of more recent residential developments was more successful than of
older areas. Seasat SAR L-band imagery of four U.S. urban areas (some with two
different look directions) was used by Hardaway, et.al. (1982) to explore cardinal effects
of urban structures. Streets/houses at angles of 20 degrees or less from perpendicular to
the radar look direction were found to accentuate radar echoes, but there was no apparent
relationship at greater angles.
