814 
Figure 2. One section of the swath of the 36° SIR-B 
data take, over part of Sydney with forest and 
natural waterways to the north west. The swath width 
is approximately 20 km with the incident radiation 
being from the left of the image. 
120 
BAND7 
F = FOREST 
R = RESIDENTIAL 
S = SOIL 
I = INDUSTRIAL 
COMMERCIAL 
120 • 
43 ^ 
80 
40 ■ 
BAND5 
40 80 120 160 
S 
0 
Figure 3. Scattergram of Landsat Bands 7 versus 5 
for various urban surfaces. Mean response values 
of forest (F), mature residential (R), soil (S), 
and intense urban (I) (commercial and industrial), 
are shown with an ellipse representing two standard 
deviations (derived from 20 values). 
Figure 5. 
43° versu 
mature re 
(commerci 
resident! 
ing two s 
BAND 
120 • 
80 
Figure 4. Illustration of the potential changes in 
backscatter and backscatter geometry with incident 
angle, for various idealised urban surfaces; residen 
tial, industrial and cleared development site. 
0 
Figure 6. 
(SIR-B) i 
Forest (F 
urban (I) 
aligned r 
presentin 
due to the inherent difference in the imaged response. 
Neverthless the Landsat data was resampled to 25 m 
pixels using a cubic convolution procedure and regis 
tered to the 36° incident angle data with a standard 
error of the estimate of approximately 60 metres. 
3. COMBINED MULTISPECTRAL RESPONSE 
Landsat data acquired over urban areas typically 
give a similar multispectral response for areas 
recently cleared for development (particularly when 
soil colours are light) and areas that are heavily 
urbanised (typically industrial and commercial land 
use). Classification of these areas as similar cover 
types is a gross error, that leads to log^c.al in 
consistencies when monitoring change over a period 
of time. A scattergram of these two cover types is 
shown in figure 3 (Landsat band 5 versus 7) 
to illustrate the problem. It was considered that 
the respo 
could dis 
differing 
corner re 
in backsc 
in figure 
The scatt 
shown in 
6, for cl 
Similar 
when atte 
areas, th 
mature ve 
a predomi 
and under 
Because t 
a reasons 
an increa 
the trees 
these con 
ponse fro 
scattergr 
used in t 
scale as