Vous êtes sur la page 1sur 7

Acra Astronaurica Vol. 40, No. IO, pp.

693-699,
0

1998 Published

by

1997
rights reserved
Printed in Great Britain
0094-5765198
$19.00+0.00

ElsevierScienceLtd. All

PII: SOO94-5765(97)00125-2

ON THE RELATIONSHIP BETWEEN ERS-1


SAR/BACKSCATTER
AND SURFACE/SUB-SURFACE
SOIL MOISTURE VARIATIONS IN VERTISOLSt
S. K. SRIVASTAVA,

N. YOGRAJAN,

V. JAYARAMAN,

P. P. NAGESWARA

RAO

and M. G. CHANDRASEKHAR
Indian Space Research Organisation, Antariksh Bhavan, New BEL Road, Bangalore, 560 094, India
(Received 31 January 1997)

Abstract-The
quantitative estimation of soil moisture is usually obtained by employing an empirical
relationship that converts the radar-measured backscattering coefficient (aO)into volumetric soil moisture
content (0,). Owing to the wide range of soil conditions and the complex interactions between microwaves
and matter in the natural environment, existing relationships are usually site specific and have a limited
range of validity. In view of this, the present study aims at realising the potential of ERS-I SAR data
towards estimating the spatial variations of soil moisture in the heavy textured soils of East Godavari
District (A.P.), India. A comparison between ERS-1 SAR measured soil moisture estimates and field
observations indicates that in non-vegetated smooth terrain of vertisols with relatively higher moisture
content (more than or nearly equal to field capacity), the relationship between u0 and 0, is almost
linear up to O-IO cm depth. However, a similar terrain with low moisture content (less than field capacity)
has not been found suitable for soil moisture estimation using ERS-I SAR data. While the results
indicate the sensitivity of SAR data to soil moisture variations, surface moisture estimates which have
been computed from SAR data have been found inadequate to extrapolate further to measure the soil
moisture status beyond 20 cm depth. With regard to profile moisture estimation, it has been observed
that, between the neighbouring soil layers, correlation between SAR-observed moisture content and in
situ

measured soil moisture status deteriorates as the distance between layers increases. 0 1998Published
by Elsevier Science Ltd

1. INTRODUCTION

One of the key application areas of ERS SAR data


is soil moisture estimation and mapping. The
capability of spaceborne active microwave techniques
to sense near-surface soil moisture has been, for
more than a decade, an area of considerable
research interest. A number of field experiments,
airborne measurements and ERS-1 SAR measurements have identified radar sensitivity to nearsurface moisture as a function of surface roughness
and soil texture for various combinations of the
radar sensors parameters of frequency, polarisation
and angle of incidence with respect to nadir [l-4].
These studies have demonstrated that the existing
radar backscattering
coefficients (a) and soil
moisture variations (A&) relationships are usually
site specific and have a limited range of validity.
Hence, an empirical approach towards estimating
soil moisture variations using ERS-1 SAR data is
called for. The present investigation aims at (i)
evaluating ERS SAR capability towards soil moisture
assessment and mapping in vertisols and (ii)

tPaper IAF95B6.04 presented at the 46th International


Astronautical Congress, Oslo, Norway, 2-6 October
1995.

exploring the feasibility of estimating the profile


moisture storage. The study area chosen for the
detailed investigation is non-vegetated terrain with
minimal surface roughness so that the influence
of surface roughness could be avoided to some
extent.

2. STUDY

AREA

The study area, parts of East Godavari district


(A.P.) (Fig. l), is situated between 1620 to 18
north latitude and 8133 to 8235 east longitude
which are covered by ERS-l/SAR (path 811 and
row 205) in descending node with 3 days repetitivity
cycle of the first ice phase. The first test site is
situated between Rajanagaram and Murari (called
the Murari site) and second test site is very
close to Ramchandrapuram (called the Ramchandrapuram site). Both the areas were non-vegetated and
the total area included in the study was around
200 ha each. The soils of test sites have high
clay content ranging from 40 to 60% throughout
the profile. These are self-mulched soils and are
rich in potash, lime and magnesium. Flue-cured
tobacco is cultivated on these soils under rainfed
conditions. Taxonomically, they belong to vertisols
(s).
693

S. K. Srivastava

694
3. METHODOLOC\
3. I. Ground

rf (I/

3.2. SAR data processing

truth collation

soil moisture
samples
and
other
Extensive
ground information
related to soil texture. bulk
density surface roughness, row direction, vegetation
cover related to parameters such as crop height and
density have been collected on the day of ERS-1 SAR
data acquisition
for four scenes. The moisture
contents of the soil at the sampling depths of O--2.
2-5, 5510. IO-20 and 20-40 cm were estimated
mostly between O&30 and I l:30.

Fig. I. Location

The ERS-I SAR scenes were registered to Landsat


TM data of 20th March 1992 which matches with
ERS SAR observation
dates,
This helped
in
identification
of soil sampling locations on radar
imagery. The SAR data were processed for speckle
suppression using an adaptive filter. The SAR digital
values corresponding
to the pixels of sampling
locations were extracted and their average value was
taken as the representative
grey level of the sampling
location.

map of the study areas

Relationship between ERS-I SAR backscatter and surface- sub-surface


3.3. Computation of backscattering values

Backscattering values depend on a number of


parameters including the dielectric behaviour of the
terrain. To calibrate the NRSA-processed ERS-1
SAR data [5], we have developed the following
equation:

pixel spacing of 15 m. eqn (1) is based on


comparative digital count values in the NRSA and
DLR data set (which was found to have rz = 0.97347,
giving an r value of 0.9866). 0 values for study areas
were computed using eqn (1).

rrO= 27.3110807 + 0.7248401(DN,)

4. RESULTS AND DISCUSSION

- O.O0764658(DN,) + 0.0000281(DN;)1

(1)

where u = mean backscattering coefficient in dB for


NRSA data and DN, = mean digital count value for
distributed target in NRSA-processed SAR data
which are in the form of an 8 bit, 4 look product with

Topographic

4.1. Surface soil moisture estimation


A methodology was developed to estimate spatial
variations of soil moisture for both the test sites,
Murari and Ramchandrapuram (Fig. 2). The simplest
method involves direct extraction of from the ERS-1

ERS - 1 SAR
image

map
I
Digitisation
of sampling
locations

Landsat TM
image

I
Sub image
extraction
(reference
image)

V
Sub image
extraction

Map to image

Registration
of landsat TM
image

linkage
1
Transfer of
sampling locations
to SAR image

695

Signature
I
extraction
(SAR DN values)

Soil moisture model

Soil moisture
map

LT-
c

Accuracy
assessment

Field survey

Fig. 2. Methodology for generating soil moisture varaition map.

S. K. Srivastava

696

n,

Test site: Ramchandrapuram


Date: March 3. 1992
t&2
0 = 36.2 0, - I I 8
&.s
0 = 51 z 0, ~- 13.1
&IO
6 = 66.3 0 ~ I6 8
Date: March 9. I992
f&2
O-5
IS-10

6 = 67-.. 3 0 , ~ 15 0
UC= 47.5 0, - 17.5
ri = 67 2 0, - 16 8

0.84
0.79
0.75

0 =

30.10 - 10.3
v = 38.8 I): - 21.6
u0 = 50.7 f?, - 22.9

0.78
0.75
0.69

0 = 64.3 0 ~ 15.X
CT>
= 43.1 0: - IO 7
UC>
= 77.5 0, - IX.3

0 75
0.70
0.69

Test Site: Muran


Date: March 3. 1992
o-2
&5
O-10

0 = 73.6 H. ~ 23.6
o = 76.6 0, - .
4.S
CT= 81.6 III - 25.-

0.70
0 66
0.62

Date: March 9. 1992


o-2
&5
O-10

CT
= 75.3 0, ~ 14.7
61= 79.5 (I1- 20.6
6 = 78.5 H, - 15.6

0.74
0.71
(I.?2

(T= X3.7 0 - x4
o = 72.1 0t - 23.6
0 = 82.501 - 28.5

0 62
0.67
0.61

0 = 647 0, ~ 20 5
0 = 18.9 U - 26.1
0 = 71.1 0: - 29.7

0.72
0.60
0.76

I I.

Date: February
o-2
CL5
&IO
Date: March
LL2
l&5
O-IO

February
&2
&5
C-IO
March
G-2
&5
&IO

I I.

1992

15. 1992

1992

15. 1992

SAR-derived
that

backscattering

coefficient.

Assuming

cTO= mW,,I t c

(2)

rr ul.
the slope of the regression line (dB cm - cm)
(dB % - )
is the intercept of the regression line (dB)

is

Table I presents the results computed using eqn (2).


It is evident from the results that r values are
always higher for the test site at Ramchandrapuram.
Since surface moisture contents in the case of this test
site are always higher than those of test sites at
Murari, higher moisture contents in vertisols presents
better relationship with 0. Furthermore,
in the case
of the test site at Ramchandrapuram,
for days on
which surface moisture content is relatively high its
relationship
with o0 has been found better (e.g. the
moisture content was highest on March 3, 1992 and
the correlation was found to be maximal on that day).
However, in the case of the Murari test site, this trend
is absent. In view of this, it may be concluded that in
vertisols a relatively higher soil moisture
regime
presents a stable relationship
with 0. The poor
response of o- to surface moisture conditions
in
vertisols may be due to the moisture content being
much less than field capacity (FC). Under such
conditions
the number of free water molecules is
lower and the dielectric response approaches that of
dry soil particles.
Hence. ERS-1 SAR sensitivity
towards soil moisture seems to be better above FC
irrespective of textural influence.
Based on the above relationships,
soil moisture
variation maps representing the moisture status from
0 to 5 cm depth have been prepared for all the dates
and for both the test sites. Maps present
the
instantaneous
conditions
of spatial variations
of
surface soil moisture.
Since the test sites were
relatively smooth, the influence of surface roughness
on the sensitivity of backscatter to soil moisture may
not be substantial.
The Rayleigh criterion considers a surface to be
smooth

where,
6,s% is the volumetric water content
(0)and the depth (x)

if

between the soil surface

h<---.-

8 cos 8

Ramachandrapuram

(I 1 February

Volumetric

content

1992)

2.41

Fig. 3. Volumetric

moisture
moisture

content

(cm3km3)
(cmJ cm- )

Relationship between ERS-1 SAR backscatter and surface- sub-surface

697

Murari (11 February 1992)


At ERS-1 SAR C band frequency
of 5.37 GHz with VV polarisation
and 23 incidence angle

1.4.5

Volumetric moisture content (cm3/cm3)


Fig. 4. Volumetric moisture content (cm cm-).

where h is the vertical relief. For C-band ERS-1 SAR


with 6 = 23, h = 0.72 cm which is quite low. It
means that the contribution of surface roughness
towards reducing the sensitivity of Q to surface
moisture cannot be eliminated completely. However,
the present study shows that if the surface moisture
level is high then the sensitivity of & will be more
pronounced towards surface moisture conditions
than towards surface roughness. There are similar
conclusions reported in the literature [6,7]. From this
point of view, the test site at Ramchandrapuram,
where the surface moisture level is relatively high,
may present a better relationship between IYand 6,.

4.2. Towards sub-surface soil moisture estimation


The physical basis of subsurface soil moisture
estimation using ERS-1 SAR data is its penetration
depth capability into the subsurface zone. Moisture
is a key factor affecting the penetration depth. The
penetration depth, LP, is defined as the depth at which
the power of backscattered signals decreases to a
certain value and thus:

LP=

zrr

J;

tan 6

where tan S = ~/c, expressed as a function of the


medium loss tangent and c and t are real and
imaginary parts of dielectric constant respectively.
For C-band, ERS-1 SAR, L, values computed at
varying moisture levels, (6 and 6 values have been
computed using Dobsons model [S]) are presented
graphically in Figs 3 and 4 for both the test sites.
Figures 3 and 4 show the extent to which C-band
radar signals can penetrate into the subsurface
regions of vertisols in field conditions. However,
surface roughness, angle of incidence and state of
polarisation can influence the depth of penetration
quite significantly. In view of this, studies indicate
that, at best, the remotely sensed depth of the soil
may not exceed 5 cm [6,9, lo].

In situ measurements of soil moisture at deeper


layers is very difficult especially in heavy soils such as
vertisols. Several studies indicate the feasibility of
estimating soil moisture variations at deeper layers
from fluctuation of soil moisture in the surface
layer [9-131. One of the earliest studies [lo], in this
regard, suggested to estimate soil moisture at deeper
depth from that at or near the surface soil moisture
estimation was of the form:
S = A + B(d - do) + CS, + D(d - d)

(3)

where S is the soil moisture at depth d and S, that of


the surface or near-surface layer whose depth is 4; A,
B and C are constants in the equation. Hence if d and
4 are known S can be estimated from them.
Towards realising eqn (3) in the study areas, a
correlation
matrix showing the extent of the
relationship between moisture status at the surface
layer to the moisture contents at various depths of the
profile has been derived and is presented in Table 2.
It is quite evident from the matrix that between the
layers the correlation deteriorates as the distance
between layers increases, indicating that a significant
linear relationship between the moisture conditions of
two layers may be expected if the layers are closely
Table 2. Correlation
Depth (cm)
o-2
Ramchandrapuram
1
O-2
2-5
5-10
l&20
2&40
Murari
O-2
2-5
5-10
10-20
2OAa

matrix

Depth (cm)

2-5

>I0

l&20

20-40

0.978
1

0.871
0.941
1

0.482
0.712
0.831
1

0.235
0.368
0.580
0.770
1

0.934

0.721
0.941
1

0.314
0.703
0.712
1

0.127
0.248
0.448
0.612
1

Sample data from all four dates were used to derive the matrix.

S. K. Srivastava et al.

698

Volumetric moisture content (cm/cm3)


0.12
01

0.16
I

0.20
I

0.24
I

0.28
I

0.32
I

0.36
I

0.40
I

0.44
I

Fig. 5. Comparison of field-measured moisture estimates and ERS-1 SAR-derived estimates.

spaced. Since surface moisture status in the case of


the test site at Ramchandrapuram
is relatively high,
interlayer correlations
have improved significantly.
The coefficients A, B and S, of eqn (3) have been
derived for study areas by using linear simultaneous
equations and the equation for study areas is:
S(d) = - 0.0099 + O.O821(d - do) + 0.1222&
- O.O041(d - d)

(4)

Using this equation, the moisture content at different


depths of profile has been computed.
The surface
moisture S,, has been derived from o values. A
comparative
assessment of soil moisture estimation
from field data and SAR cr derived data is presented
in Fig. 5. It is quite evident that at deeper depth there
are more mismatches between SAR estimates of soil
moisture and values observed through field measurements. There is an interesting trend that moisture is
being over-estimated
by SAR data in all the cases,
The results reveal that a lot of improvements
need to
be incorporated
towards more precise measurements
of profile moisture using SAR data.

5.

CONCLUSION

The study on surface and subsurface soil moisture


estimation by using ERS-I SAR data in conjunction
with field observations, laboratory measurements and
models reveals the following inferences:

l A smooth
terrain of vertisols with a relatively
high moisture content has been found suitable to
realise
the potential
of ERS-I
SAR towards
estimating surface soil moisture variations. Similarly,
a terrain with low moisture content presents a weaker
relationship
between
surface moisture
and SAR
backscatter.
l Although
in vertisols the relationship
between
inter- or intra-layer
soil moisture
variations
are
better, ERS-1 SAR capability
towards estimating
subsurface soil moisture variations has been found
effective only up to G-10 cm depth. However, its
capability improves slightly if the surface moisture
status is high.
l While profile moisture
storage can be computed
from surface moisture data derived from SAR data,
inclusion
of more parameters
representing
the
retention
and transmission
characteristics
of soils
may improve the accuracy of measurements.

REFERENCES

Mohan, Shiv, Mehta, N. S., Mehta, R. L., Parul Pate],


Bora, V. H., Sule, K. K., Purohit, S. S., Rai, S. S.,
Singh, A. P. and Ramakrishna, Y., Multi-frequency
microwave measurements for soil moisture. In Proc.
National Symp. on Remote Sensing for Agricultural
Application, IARI, New Delhi, 68 Dec. 1990, pp.

357-363.
Lin, D. S., Wood, E. F., Beven, K. and Saatchi, S.,
Soil moisture estimation over grass covered area

Relationship between ERS-I SAR backscatter and surface- sub-surface


using AIRSAR. Int. J. Remote Sensing, 1994, 15,
2323-2443.
3. Ulaby, F. T., Moore, M. K. and Fung, A. K.,
Microwave

Remote

Sensing

Active

and

Passive,

Vols. 2-3. Artech House, Norwood, MA, 1986.


4. Srivastava, S. K., Jayaraman, V. and Chandrasekhar,
M. G., Evaluation of microwave spectrum for study of
hydrologic processes, IAF-94-B.6.118. 45th International Astronautical Federation (ZAF), 9-14 Oct. 1994,
Jerusalem, Israel.
5. Paudyal, D. R. and Aschbacher, J., ERS-I SAR data
calibration at the Indian NRSA. Asian-Pacific Remote
Sensing J., 1994, 6, 117-120.
6. Dobson, M. C. and Ulaby F. T., Active microwave soil
moisture research, IEEE Trans. Geosci. Remote
Sensing, GE-24(l), 1996, 23-35.
7. Pultz, T. J., Leconte, R., Brown, R. J. and Brisco,

B., Quantitative
soil moisture extraction
from
airborne SAR data. Can. J. Remote Sensing, 1990, 16,
5Ml.
8. Dobson, M. C., Ulbay, F. T., Hallikainen, M. T. and

Rayes, M. A., Microwave dielectric behaviour of wet

699

soil-part
II: Dielectric mixing models, IEEE Trans.
Geosci. Remote Sensing, GE-23(l), 1985, 3546.
9. Rao, P. V. N., Srinivas, K., Ramana, K. V.,
Venkatratnam, L. and Manjunath, A. S., ERS-1
Synthetic Aperture Radar data for soil moisture
estimation. In Proc. National Symp. on Microwave
Remote Sensing, Space Applications Centre (ISRO),
Ahmedabad-380 053, India, on l&11 Jan. 1994, pp.
124-128.
10. Biswas, B. C. and Dasgupta, S. K., Estimation of soil
moisture at deeper depth from surface layer data.
Mausam, 1979, 30, 4, 51 I-516.
11. Arya, L. M., Richter, J. C. and Paris, J. F., Estimating

profile water storage from surface zone soil moisture


measurements under bare field conditions. Wafer
Resources Res., 1983, 19, 403412.
12. Tripathi, R. P. and Ghildyal, B. P., Hydraulic
conductivity and diffusivity relations of mollisols of
Tarai. J. Indian Sot. Soil Sri., 1976, 24, 339-346.
13. Jackson, R. D., On the calculation of hydraulic
conductivity. Soil Sri. Sot. Am. Proc., 1972, 36,
38&382.

Vous aimerez peut-être aussi