325
Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986
56 sugar-beet ml „
57 turnip
42 sugar-beet min
' , R . ;
>ns of spec-
L objects
oational
Remote
obinoon, G.F.
jricultu ral
space, aerial
at ional
Remote
3. 1983.
bee a
ues daus les
nes -
res spec-
n.
83. Issledo-
sociolisti-
3: 120 -
4. Metod
zenki ugodij
ledovanie
LANDSAT temporal-spectral profiles of crops
on the South African Highveld
B.Turner
Optical Sciences Division, Nat. Physical Research Lab., Pretoria, South Africa
ABSTRACT :
LANDSAT MSS data were standardized for variations in digital count coding between different receiving stations,
satellite sensor response and solar illumination of the scene. The Kauth and Thomas approach has been used to
obtain the independent variables of soil brightness, greenness and yellowness. The coefficients of the rota
tion matrix were modified to suit local conditions.
Three vegetation indices, the vegetation ratio, the normalized vegetation index and the greenness index have
been determined on a field-by-field basis using data recorded on sixteen overpasses within thirteen months. An
intercomparison of the spectral-temporal profiles for a variety of crops is in progress. A correlation between
the spectral profiles and a ground reference set consisting of detailed information on planting, growth and
environmental conditions is to be undertaken.
The spectral separability of the crops in terms of the vegetation indices and the field conditions is being
evaluated.
1 INTRODUCTION
The need to develop objective, repeatable and time-
efficient methods for crop identification has been
recognized by researchers particularly in the United
States of America since the 1970's. The identifica
tion of cultivated crops from LANDSAT imagery has
traditionally been accomplished by the recognition of
certain crop characteristics which change throughout
the growing season (i.e. temporal changes)(MacDonald
and Hall, 1980). Success depends upon the correct
estimation of the time of occurrence of a particular
crop feature which is stable from year to year and
which does not overlap with similar characteristics of
other crops growing in the same geographic region. In
the few cases where such features can be identified, a
unique signature for each crop can be accurately and
consistently defined (Malan and Turner, 1984).
An alternative approach was introduced in 1976 when
Kauth and Thomas (Kauth and Thomas, 1976) showed that
a fixed two-dimensional subspace of the LANDSAT four-
dimensional measurement space contained most of the
spectral information for a very large range of agri
cultural, seasonal and meteorological conditions
(Hall, 1982). Furthermore, they proposed that two of
the transformed axes of the subspace related to
changes in scene brightness and changes in the leafy
matter in the vegetative canopies.
Other researchers (Jordan, 1969; Pearson, 1972;
Rouse et al., 1973; Colwell, 1974; Maxwell, 1976;
1978; Richardson and Wiegand 1977) have suggested cer
tain mathematical combinations or vegetation indices
in an attempt to establish a relationship between
LANDSAT multispectral scanner (MSS) data and the
ground cover canopy.
It is also important to realize that there are a
number of variables which significantly influence the
data collected by earth orbiting satellites. For
quantitative studies, methods of data analysis must be
selected or developed to take the effects of these
variables into account.
In the present study, techniques which have been
successful elsewhere have been applied to agricultural
MSS data collected for the Highveld region of RSA.
Control data concerning the characteristics of the
crop canopy have been obtained from an agricultural
survey conducted simultaneously with LANDSAT overpas
ses by officers of the Highveld Region of the Depart
Figure 1. Location of test area.
ment of Agriculture (Malan and Turner, 1982).
A correlation has been sought between time series
of spectral reflectance properties of the vegetative
cover as measured by LANDSAT MSS data and known
features of the actual crops.
2 STUDY AREA DESCRIPTION AND DATA SOURCES
2.1 Choice of test area
The LANDSAT-scene identified by a code WRS 182-79 in
