813 
Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986 
Evaluation of combined multiple incident angle SIR-B digital data 
and Landsat MSS data over an urban complex 
B.C.Forster 
Centre for Remote Sensing, University of New South Wales, Sydney, Australia 
ABSTRACT: As part of the NASA sponsored SIR-B experiment, digital data with incident angles of 17°, 36° and 
43° were recorded over Sydney, Australia, and havebeen used in a study of radar imagery for urban purposes. 
The effect on radar backscatter of the multifaceted and oriented features found in urban regions has been 
examined as a means for improving urban discrimination when combined with Landsat multispectral data. Imagery 
at different incidence angles were registered to each other and to Landsat data and analysed using an image 
analysis computer system. While systematic interpretation of the radar imagery is complicated by the high 
response from urban features aligned at right angles to the incident radiation, the combined radar and Land 
sat images are shown to give good discrimination between sites cleared for development and those heavily 
urbanised. These areas show a similar Landsat response but are markedly different in radar. Moreover in older 
residential areas, with significant tree cover, the Landsat response is dominated by the vegetation signature, 
while radar is shown to provide an increased response from the underlying buildings. 
1. INTRODUCTION 
A number of previous studies have examined urban 
areas using radar imagery, these include Bryan (1979) 
and, Hardaway and Gustafson (1982). In general these 
studies have concentrated on the so called 'cardinal 
effect'. Here the intersection of roads and build 
ings tend to act as corner reflectors when they are 
aligned at right angles to the incident radiation. 
They showed that backscatter was very sensitive to 
street alignment, and reduced from a maximum to a 
constant value as the street orientation approached 
a threshold value of 20° to 25° away from the right 
angle relationship. Both these studies used single 
incident angle synthetic aperture radar data from 
earlier NASA missions, Seasat and SIR-A (Shuttle 
Imaging Radar-A), while the present study had the 
advantage of multiple incident angle data in a digi 
tal form compatible with digital Landsat MSS data. 
Data over Sydney was acquired as part of the NASA 
SIR-B (Shuttle Imaging Radar-B) experiment launched 
on the 17th flight of space shuttle in October of 
1984. This 8-day mission collected microwave data of 
many parts of the Earth's surface. Table 1 compares 
the characteristics of Seasat, SIR-A and SIR-B. 
Table 1. Characteristics of NASA Spaceborne Synthetic 
Aperture Radars. 
Seasat 
SIR-A 
SIR-B 
Spacecraft' 
800 km 
260 km 
225 km 
Altitude 
Wavelength 
23.5 cm 
23.5 cm 
23.5 cm 
Polarisation 
HH 
HH 
HH 
Look Angle 
20° 
47° 
15°-57° 
Swath Width 
100 km 
50 km 
14-44 km 
Azimuth 
25 m 
40 m 
25 m 
Resolution 
Range Resolution 
25 m 
40 m 
58-17 m 
Data at three incident angles were acquired over 
the Sydney region (43°, 36° and 17°), as shown in 
figure 1, imaged from a path to the northeast of the 
city. Part of the swath of the 36° data take is shown 
in imaee form in figure 2. 
2. IMAGE REGISTRATION AND PREPROCESSING 
The SIR-B data was available in a digital form at 
12.5 m pixel centres. Because this was substantially 
less than the inherent resolution of the data, and 
to reduce speckle effects, the data was resampled to 
a pixel size of 25 m using a cubic convolution re 
sampling procedure. Following this a total of 50 
control points were selected in an overlapping area 
of the 43 and 36° incident angle images (with over 
lap dimensions of approximately 10 by 40 km). These 
were naturally occurring points (road and stream 
intersections, centres of small parks, water/land 
features, etc) arrayed in banks of five across 
track. Their location and number were designed to 
give maximum registration accuracy, particularly 
across track where the maximum errors due to geo 
metry would occur. Using a 2nd order polynomial 
transformation and holding the 36° data fixed, re 
gistration errors of 2.2 pixels across track and 
0.9 pixels along track (standard error of the esti 
mate) were determined. A nearest neighbourhood re 
sampling procedure was used to register the 43° inci 
dent angle data to the 36° data. 
Data from Landsat bands 7 (near infrared) and 5 
(visible red) were viewed simultaneously with the 
radar data and a total of 12 common points were 
determined. Selection of common points was difficult 
Figure. 1. Coverage of SIR-B radar swaths at 
17°, 36° and 43° incident angles.