NT Py à 
E e o SE 
  
  
shallow drainage lines... Mature stands of A. tortilis and 
Salvadora persica trees with green leaves and a high band 7 (0.8- 
1.lum) reflectance, appeared red in false ‘colour. Some 
qualitative information about the spectral response of vegetation 
and soil was obtained from the colour composite imagery. However, 
a more quantitative knowledge of soil and vegetation response is 
needed .in order to develop an image processing system that will 
successfully discriminate between vegetation types. 
Vegetation boundaries were drawn from 1:50 000 scale 
panchromatic aerial photographs for 1972, using a high 
magnification Stereo Interpretoscope. The interpreted photographs 
were mosaiced together to form a ground data map which was 
classified into six vegetation classes (Table 1) using ground 
data, low altitude vertical colour aerial photographs and 
panchromatic aerial photographs. Canopy cover measurements using 
a 0.5mm” graticule were taken from the panchromatic aerial 
photographs, and estimates for the canopy cover of dwarf shrub 
and grassland were taken from the colour aerial photographs. 
Limited ground data on soil type, vegetation canopy and species 
composition was collected during the summer of 1981 and has been 
used to check the photographic interpretation. In particular, 
observations on the ground and from the air suggest that the soil 
surface within this sedimentary basin has a high but generally 
uniform spectral response. 
Radiance values in bands 5 (0.6-0.7um) and 7 (0.8-1.1um) 
were sampled from a dry-season LANDSAT tape for June 1979, the 
latest then available, using an Interactive Image Analyser (IS). 
For each vegetation class 10 samples were taken, each sample 
being the average for 25 pixels. 
3. RESULTS 
Figure 2 is a scattergram of radiance values with band 7 
values plotted against band 5 values for the six vegetation 
classes (Table 1). Data points tend to cluster in a straight line 
and show that increasing canopy cover is strongly correlated with 
decreasing radiance. 
Points A and B on Figure 2 are taken to represent the 
positions in the feature space of 100 percent vegetation cover 
with a minimum value at A, and bare soil with a maximum value at 
B. The positions of these two extremes cannot be precisely fixed 
as each point is represented by a scatter and not a single value. 
However, if the minimum value is taken to represent 100 percent 
vegetation cover, and the maximum value bare soil, then the line 
joining A to B can be used to divide canopy cover marked along 
the band 5 axis, into six vegetation classes (Table 1). This line 
represents a vegetation cover line similar to that described by 
Graetz et al.(1982), who showed that for semi-arid vegetation in 
Central Australia, vegetation could be represented by a line 
extending from bare. soil with maximum radiance values, to full 
vegetation cover with minimum radiance values. 
This interpretation of the data assumes that changes in 
radiance values are a function of only two components combined in 
different proportions; soil and vegetation. For vegetation 
classes within the test area, the spectral response is dominated 
by band 5, with all data points having a higher band 5 value than 
383 
i 0 0 
aS