arly indepen- 
thonormal set 
calculated by 
Jering, 1963) 
's representa- 
egetation and 
s from satel- 
' standardized 
extracted for 
îlues together 
re stored for 
the present 
3 MSS data. 
3 data 
ne test sites 
site temporal 
rghum and sun 
planting date 
ariability in 
ected by local 
ces. 
of MSS data 
compensate for 
tal count and 
difference 
ed from EROS 
nship: 
(1) 
i be converted 
(2) 
2 
Sr. are given 
trome et al., 
1 values of 
min-^'min) 
max-R'min) 
MAIZE 
315 0 50 100 150 200 250 300 350 
I960 1981 
JULIAN OATE 
SORGHUM 
Compensation for station-to-station differences was 
carried out by means of a linear transformation of 
the MSS data using the calculated values of A and B. 
4.2.1.2 Compensation for satellite-to-satellite 
differences 
The method and values prepared by Parris and Rice 
referred to above were used (Rice et al., 1983). 
4.2.2 The Tasseled cap Transformation for the study 
area. 
JULIAN OATE 
The application of the Gram-Schmidt process to the 
data corrected as described in 4.2.1.1 and 3.3.2 
above produces the required set of orthogonal unit 
vectors. 
The matrix of the Tasseled cap transformation for 
the Grootvlei test site was found to be 
SUNFLOWER 
315 0 50 100 150 200 250 300 350 
1980 1981 
JULIAN OATE 
0.30 
-0.36 
-0.26 
(Turner, 1986) and 
in 3.3.3. 
0.38 0.34 
-0.80 0.37 
-0.77 0.58 
SBI, GVI and YVI 
0.69 
0.30 
0.05 
were defined as 
4.3 Vegetation indices 
Figure 5. Temporal Plots of Growth Stages of Field 
Samples within WRS 182-79. 
Table 4. Range of Absolute Radiances (Anon, 1976 and 
Strome, 1975) 
EROS 
CCRS 
R 
max 
R 
min 
R' 
max 
R' 
min 
LANDSAT 
1 
MSS 
4 
2.48 
0 
3.00 
0.0 
MSS 
5 
2.00 
0 
2.00 
0.0 
MSS 
6 
1.76 
0 
1.75 
0.0 
MSS 
7 
not calibrated 
not calibrated 
LANDSAT 
2 
MSS 
4 
2.63 
0.08 
3.00 
0.0 
MSS 
5 
1.76 
0.06 
2.00 
0.0 
MSS 
6 
1.52 
0.06 
1.75 
0.0 
MSS 
7 
3.91 
0.11 
4.00 
0.0 
LANDSAT 
3 
MSS 
4 
2.50 
0.04 
2.50 
0.0 
MSS 
5 
2.00 
0.03 
2.00 
0.0 
MSS 
6 
1.65 
0.03 
1.75 
0.0 
MSS 
7 
4.50 
0.03 
4.00 
0.0 
The calculated values of A and B are given in Table 5. 
Table 5. Coefficients of Linear Transformation for 
SRSC 8-bit data to EROS 7/6 bit data 
A 
B 
LANDSAT 
1 
MSS 4 
0.61 
0.00 
MSS 5 
0.50 
0.00 
MSS 6 
0.25 
0.00 
MSS 7 
not calibrated 
LANDSAT 
2 
MSS 4 
0.59 
-4.02 
MSS 5 
0.57 
-4.36 
MSS 6 
0.60 
-5.26 
MSS 7 
0.26 
-1.85 
LANDSAT 
3 
MSS 4 
0.51 
-2.06 
MSS 5 
0.51 
-2.06 
MSS 6 
0.51 
-2.06 
MSS 7 
0.51 
-2.06 
Sample wheat test sites were formed by concatenation 
of known samples from wheat fields representative of 
WRS 182-79 (Malan and Turner, 1984). A temporal plot 
of the VI and NVI is given in Fig. 6. 
5. DISCUSSION OF THE RESULTS 
The coefficients of the matrix of the Tasseled cap 
transformation obtained in the present work were 
found to differ significantly from those obtained by 
Kauth and Thomas for MSS data representative of the 
US Corn Belt. Although the Tasseled cap transforma 
tion does indeed exploit the structure of the MSS 
data to produce independent indices SBI and GVI, the 
orientations of the distribution of the bispectral 
plots appear to be dependent on geographic location. 
This study has therefore illustrated that even when 
the MSS data is reduced to the same standard condi 
tions applied by Kauth and Thomas for the US Corn 
Belt, the coefficients of the Tasseled cap transfor 
mation are different for US and South African condi 
tions. However, the concepts evolved by Kauth and 
Thomas would still appear to be directly applicable 
to agricultural MSS data collected for the South 
African Highveld. 
The vegetation indices, as shown in Fig. 6, form 
two distinct sets representing the wheat and non 
wheat cover types. Seasonal variations seem to be 
significant and may be correlated with the quantity 
and timing of rain and the growth stage of the 
wheat. The single observation for the 1982 season 
seems anomalous in two respects: The maximum index 
is greater than for the other two seasons and the 
division between wheat and non-wheat is less pro 
nounced. These significant variations may be due to 
higher rainfall during October 1982 (1627 mm) than in 
1981 (316 mm) and 1983 (1266 mm). 
5.1 Future developments 
The generation of the temporal profiles of the spec 
tral and ground reference data sets for each field 
and an analysis of these data in terms of the spec 
tral separability of the crops, the influence of the 
planting date, row direction, etc. on the vegetation 
indices, the spatial variability of vegetation in 
dices for similar crops within a LANDSAT scene, and