You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Remote sensing for resources development and environmental management
Damen, M. C. J.

s present
MSS data,
ating of this
of high
Mapper (TM) and
1 supply data
the other hand
the handling
ed by most
d national
te sensing
the Indonesian
such a
f remote
1 use of remote
fits of a
he capability
that multi
d. Such a
om imagery of
realising these
studies are
findings and
such a multi
ted equatorial
oring of
of Indonesia,
oducts Tanaman
mesian Case
in Indonesia, a
md its
.ng for Crop
ig for
îe remote
Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986
Assessment of soil properties from spectral data
G.Venkatachalam & V.K.R.Jeyasingh
Indian Institute of Technology, Bombay
ABSTRACT: Application of computer aided analysis of Landsat data to the finer levels of identifying soil character
istics, such as grain size, is beset with difficulties because of the poor resolution of the Landsat data. The difficulty
is essentially due the low correlation between grain sizes of soils and their reflectance values which precludes predic
tion of grain sizes from reflectance to a satisfactory degree. Hence, an attempt has been made in this study to evolve
mathematical models which give better correlations between the gradation of soils and their spectral reflectance chara
The first model is in the form of a linear relationship between the grain sizes and the third Principal Components
of the reflectance data. A second model with a higher predictive ability has also been suggested based on the concept
of Primary, Secondary and Tertiary Principal Components. Further, using non-linear optimization techniques, a third
model having a very high correlation has also been evolved. These models were based on a group of soil samples, colle
cted from different parts of India.
The model obtained through optimization was found to have the best predictive ability. Therefore, the optimization
technique has been extended to a larger sample group of fourteen different soils. The accuracy of the model so evolved
has been tested statistically and the model applied successfully to predict grain sizes of a few soils collected from
certain other parts of the country.
This study will help in the application of Landsat data for estimating grain size characteristics of soils, if their
responses in the four bandpasses are known.
Spectral reflectance characteristics of soils vary broadly
due to many factors such as colour, mineral content,
covertype, moisture content, texture or grain size,
type of parent rock and land-forms. Multispectral sca
nner (MSS) data collected from space-borne platforms
have been successfully used in exploration for dema
rcation of land-forms, soils and rock types on a broad
basis. But, studies related to soil characteristics such
as grain size are limited because of the poor resolution
of the MSS data.
A good amount of research work has been reported
in estimating soil texture in terms of sand, silt and
clay fractions using microwave remote sensing. Lab
oratory studies by Gerberman (1979) showed the exis
tence of a definite linear relationship between sand
levels and their reflectances in the visible and near
infra-red regions, in a mixture of sand and clay. MSS
data in the visible and near infra-red regions could
not be effectvely used to study the variations in soil
texture and other characteristics because of their poor
correlations. Since a number of orbiting platforms
are routinely collecting information in the visible and
near infra-red regions, it is worthwhile to evolve a
method to use this information to study the soil charac
teristics. Hence the present investigation has been
taken up with an objective of evolving a relation- ship
between the grain size characteristics of naturally
available surface soils and their reflectance values
measured in the laboratory in the visible and near
infra-red regions of the electromagnetic spectrum (EMR)
by siutable transformation techniques.
their relfectance values corresponding to the four bands
of the Landsat data.
3. Analysis of the data to evolve a reliable model
by suitable transformation techniques to predict the
grain size characteristics of soils.
4. Testing the model to find the accuracy of predic
2.1 Experimental setup
A Cema sungun with an operating temperature of 3 50 0
degrees Kelvin was used as a light source.
An Exotech model 100-A Landsat ground truth radio
meter was used as a sensor. It observes the reflectance
in four bands corresponding to Landsat except for band-
7, where the range is 0.8 to 1.0 micron.
A barium sulphate coated plate was used as a standard
reference surface. It was assumed to have 100 percent
reflectance and specular in nature.
2.2 Experimental investigations
Initially, five types of soils from two different states
in India, with ten samples for each type, have been
used for reflectance measurements in the laboratory
under oven dry conditons. The average reflectance
values of these samples are given in Table 1. For these
five soils the grain size characteristics were also deter
mined. These data have been used for arriving at the
three models stated above. In the second phase, similar
experiments have been carried out on a larger group
of 14 soils representative of a variety of Indian soils.
The present investigation has been carried out in the
following four stages:
1. Formulation of an experimental setup in the lab
oratory where observations could be taken under contr
olled conditions.
2. Collection of soil samples and measurement of
2.3 Statistical analysis
The primary objective of the statistical analysis was
to transform the original reflectance data with a view
to improve the correlation between the grain sizes
and their transformed reflectance values. The well
known Principal Component transformation has been
used. The four band data have been transformed