326
Figure 2. Positions of test sites.
the World Reference System was chosen as being repre
sentative of the intensively cultivated Highveld
region which produces a major portion of the country's
grain products and also contains a variety of soils
and crops (see Fig. 1). Nine test sites were selected
to include the different soil types and crop produc
tion patterns. The locations of the test sites are
given in Fig. 2. Each test site contained ca. 100
fields and was ca. 11 km x 7 km in size, representing
about 50% cultivation.
The area has a summer rainfall of 600 - 700 mm per
year, has an altitude between 1400 and 1700 m above
sea level and is fairly flat. The test sites have a
maximum slope of 5% with a mean slope of 3%; thus no
topographic problems were encountered.
2.2 Ground reference data
To facilitate the collection of ground reference data,
the data set was divided into planting data and growth
data. The planting data consisted of soil colour,
planting date, crop type, cultivar, row spacing, row
direction and seed density. These data were recorded
for a sample of the 100 fields within each of the nine
test sites. The growth stages of seven typical crops
were defined by the Department of Agriculture (Malan
and Turner, 1982). Codes of growth stage, average
height, foliage distance, physical condition, weeds,
disease and insect damage were monitored for some
fields every 18 days coincident with the satellite
overpass. The sample of fields comprised 48 maize, 5
sorghum, 4 sunflower, 14 wheat, 1 potato and 1 oats.
Data derived from 1:250 000 scale land type series
of maps as published by the Soils and Irrigation In
stitute of the Department of Agriculture and the sur
face soil colour are tabulated in Table 1. The abbre
viations used are defined on the land cover type maps.
2.3 Satellite data
Multispectral scanner (MSS) data sets were collected
for 10 August 1979, 1 February 1980, 19 February 1980,
13 April 1980, 19 May 1980 and 24 June 1980. These
data sets were recorded on LANDSAT 3 and distributed
in band-sequential format by EROS Data Center, USA.
Eighteen data sets for the period November 1980 to
December 1981 were recorded on LANDSAT 2 satellite and
collected locally at the Satellite Remote Sensing
Centre (SRSC) of the Council for Scientific and In
dustrial Research (CSIR) in CCRS format (as used by
the Canada Center for Remote Sensing).
Table 1. Land characteristics.
Test
area
Pedosystem
Abbre
viation
Colour of surface
soil
BL
Petrus Steyn
Ca
Grey with yellow
DO
Koppies
Dc
Dark brown to black
DU
Petrus Steyn
Ca
Brown to yellow-brown
Blackhill
Ea
Black to brown-black
ED
Mooibult
Bd
Red-brown to brown
Koppies
Dc
Dark brown to black
GR
Clifton
Bd
Red-brown to brown
Blackhill
Ea
Black to black-brown
LI
Senekal
Ca
Brown to yellow-brown
Koppies
Dc
Dark brown to black
RE
Petrus Steyn
Ca
Brown to yellow-brown
VA
Blackhill
Ea
Black to brown-black
Koppies
(Schultzdale)
Dc
Dark brown to black
BR
Petrus Steyn
Ca
Brown to yellow-brown
3 DEVELOPMENT OF THE METHOD AND DATA ANALYSIS
The steps taken to develop the method and to analyse
the data were as follows:
• The ground reference data were reformatted
• field samples were located
• the satellite data were standardized
• suitable vegetation indices were defined
• a Tasseled cap matrix was derived
• field sample means were extracted from satellite
data.
Details are given below.
3.1 Ground reference data
Planting dates for six crops within each test site
were extracted from the survey data (see Table 2).
Crop calendars were generated as a means of display
ing the temporal variability of the growth stages for
the crops for each field.
Table 2. Planting Dates for WRS 182-79 in 1981 (in
day number)
Test site
Maize
Sunflower
Sorghum
Wheat
Potatoes
Beans
Bloukop
279-283
168-253
BRAR
293-298
146
285
Dover
328-358
356-2
342-345
CXjndas
295-310
5
144-202
Edenville
338
336-342
329-339
76
Grootvlei
295-319
3
Reitz
307-318
329
Overall
range
279-358
336-5
329-345
141-253
285
329-3
Planting
period
79 days
34 days
16 days
112 days
39 days
3.2 The extraction of field sample data
The field boundaries within each test site at
1:50 000 scale were digitized and stored in raster
image format. A mask within the monitored field
boundaries was generated. All MSS data sets were
