the soil
capacity
lation of
of field
(2)
of field
as been
M 1/3bar 7
ty and at
for this
udes the
optimum
used can
y relate
ntitative
alibrated
reason
alibrated
a) (3)
he local
ed a 32-
. These
)cessing
; carried
. Owing
IAPRS & SIS, Vol.34, Part 7, "Resource and Environmental Monitoring", Hyderabad, India, 2002
to the large disk space requirement (about 2 GB) and data
handling, the 32 bit real 0° images were converted to 8 bit
unsigned images. A user defined linear relation was used for
this purpose, so that once 8 bit DN values were extracted from
image, later on it can be inverted to get backscattering
coefficient values in order to carry out quantitative analysis.
5.6 Image registration
Adjacent path SAT images of IRS L-III were mosaiced
sidewise. The RADARSAT EL1 image was geo-referenced and
IRS LISS-III image was co-registered with respect to ELI
image keeping RADARSA-1 ELI image as reference image
nearest neighborhood method of resampling (Duggin and
Robinore, 1990).
5.7 Signature extraction
After registering both the images sampling locations were
identified on the images. Backscatter coefficient values for
sampling locations were extracted from SAR image. IRS L-III
data was used to classify crop covered soil and bare soil in
order to avoid the inclusion of crop covered fields in the
analysis.
5.8 Model development
Variation in SAR signature due to soil moisture was studied for
agricultural areas at lower incidence angle (16°) using the
backscatter values extracted from RADARSAT-1 Extended
Low-1 beam mode SAR image. Soil moisture retrieval models
had been done for the following combinations using linear
regression analysis.
- Backscattering coefficient with gravimetric soil
moisture.
- Backscattering coefficient with volumetric soil
moisture.
- . Backscattering coefficient with soil moisture
represented in terms of percentage of field capacity.
- . Backscattering coefficient with soil moisture
represented in terms of percentage of available water.
After the development of soil moisture retrieval models with
the different values of soil moisture a comparative evaluation of
different models has been performed.
6. RESULTS AND DISCUSSION
In order to express the soil moisture in a more realistic way so
that it should be less sensitive to soil texture and more useful
for agricultural applications, four separate models have been
developed using the soil moisture values in terms of
gravimetric soil moisture, volumetric soil moisture, soil
moisture in percentage of field capacity and soil moisture in
percentage of available water. Since objective of the present
study was to incorporate the effect of soil texture in the soil
moisture retrieval model, enough care has been taken to avoid
the effects of other noise parameters like crop cover and surface
roughness. Absence of other noise parameters would enhance
the effect of soil texture on the radar backscatter. For this
purpose only bare fields have been included in the analysis in
order to avoid the 2-way attenuation in radar backscatter
coefficient. Moreover out of all the bare fields only those fields
have been included in the analysis, which were smooth or
having small values of surface roughness. Results of all these
models are given in Table 2. Table 2 indicated that there is a
723
significant increase in R? for model developed with the values
of soil moisture in terms of percentage of available water. R?
increased from 0.86 to 0.93 by represent ting the soil moisture
in percentage of available water instead of gravimetric soil
moisture. A marginal increase in R? was also observed from
0.86 to 0.88 in case of replacing gravimetric soil moisture with
volumetric soil moisture. R? has further increased from 0.88 to
0.90, when volumetric soil moisture has been replaced by soil
moisture in percentage of field capacity. Similar results have
been observed by Ulaby et. al., (1986b) by expressing the soil
moisture in terms of percentage of field capacity. Study
indicated that gravimetric soil moisture and volumetric soil
moisture are less sensitive way to represent soil moisture while
using microwave remote sensing technique. This is due to the
fact that gravimetric soil moisture and volumetric soil moisture
does not take into account the relative proportions of various-
sized particles in a given soil medium which is one of the main
factors governing the relative percentage of bound water and
free water in the mixture of soil and water (wet soil). Since
field capacity and wilting point of a particular soil are directly
linked with the soil texture and microwave are sensitive to
proportion of free water and bound water in a given soil hence
the last two models were found to be better than the first two
models. In particular the forth model developed with the values
of soil moisture in percentage of available water was found to
be the best with the highest value of R?.
MODEL USED Coefficients
Soil Moisture = A + B * 0° | A B R°
Soil Moisture (weight) 46.2 3.6 0.86
=A+B*o°
where Soil-Moisture(weight)
represents gravimetric soil
moisture
Soil Moisture (volume) 67.9 5.4 0.88
=A+B*o°
where Soil-Moisture(volume)
represents volumetric soil
moisture
Soil Moisture (% r.c., 168.6 13.4 | 0.90
=A+B*o°
where Soil-Moisture(ær.c.)
represents soil moisture in
percentage field capacity
Soil Moisture (% A.w.) 224.8 19.7 | 0.93
=A+B*o°
where Soil Moisture, ow.)
represents soil moisture in
percentage of available
water.
Table 2. Results of various models developed with different
units of soil moisture
7. CONCLUSIONS
Study indicated that percentage of available water is a more
realistic way to represent soil moisture as it directly describes
the parameter affecting microwave interaction in the soil
medium. Thus it is possible to reduce the effect of soil texture
in the sensitivity of microwaves towards the soil moisture. The
concept of available water in a soil medium is based upon the
assumption that the water present in the soil medium at 15-bar
pressure, which is unavailable to plants, is actually the bound
