904 
Grassland type 2 - mean & standard deviations 
of dns for TM bands 3 and 4 
Bare soil - mean & standard deviations of dns 
for TM bands 3 and 4 
Parcel L1 
Parcel L2 
Parcel B1 
Sample 
Spec. 
Sample 
Spec. 
size 
band(TM) 
size 
band(TM) 
5 
4 
mean 
30.18 
31.12 
5 
4 
mean 
24.87 
3 
17.54 
17.55 
3 
21.42 
4 
s.dev 
1.30 
2.04 
4 
s.dev 
1.66 
3 
0.22 
0.30 
3 
0.29 
10 
4 
m 
29.98 
31.17 
10 
4 
m 
24.14 
3 
17.43 
17.58 
3 
20.98 
4 
sd 
2.17 
2.59 
4 
sd 
1.75 
3 
0.28 
0.34 
3 
0.48 
15 
4 
m 
30.97 
30.35 
15 
4 
m 
25.03 
3 
18.00 
17.11 
3 
21.22 
4 
sd 
2.28 
3.04 
4 
sd 
1.94 
3 
0.31 
0.43 
3 
0.54 
20 
4 
m 
31.52 
30.86 
20 
4 
m 
23.57 
3 
18.32 
17.40 
3 
20.65 
4 
sd 
2.68 
3.31 
4 
sd 
2.11 
be incori 
25 
3 
0.38 
0.49 
3 
0.65 
detectior 
4 
m 
32.17 
31.53 
25 
4 
m 
24.21 
3 
18.70 
17.78 
3 
20.76 
are pred 
4 
sd 
2.92 
3.72 
4 
sd 
2.56 
strategie 
3 
0.43 
0.59 
3 
0.72 
30 
4 
m 
31.23 
31.76 
30 
4 
m 
22.92 
3 
18.15 
17.91 
3 
19.14 
4 
sd 
3.24 
4.23 
4 
sd 
2.83 
3 
0.48 
0.66 
3 
0.89 
Source: data assembled by Mr Jin-King Liu, April 1985 
detect cover types with less than five per cent of the total 
cover. It would seem that only one pixel per ten hectare 
plot would be sufficient to provide adequate 
discriminative power representing a reduction in data of 
one hundred times. Such .economies are very attractive 
both to those designing data transmission systems as 
well as to those handling and archiving data. 
It should be emphasised that the precision with which the 
estimate can be made on the basis of the sample is not 
because the discriminative power of the sensor is special 
in spectral terms. The confidence with which different 
cover can be detected is because each parcel has been 
managed to a relatively homogeneous crop cover. It is 
an awareness of this behavioural feature which should 
The abo\ 
strategie 
processii 
5 CONS 
RURALl 
If spatiall 
hundredt 
to be as | 
pose the 
such eco 
on-board 
spatially! 
reduction 
and a hal 
desirable 
the spect 
spectral c 
good exa 
that a sat 
of the spe 
record, 
whether r 
natural re 
appncatio 
survey of 
applicatio 
which col 
the applic 
essential 
managerr 
govemme 
On the ba 
it has bee 
effective i 
cover in n 
therefore, 
construct