ERS-1 and JERS-1 SAR Data Analysis for Soil Moisture and Land Cover Studies
Y.S. Rao, P.V.N. IRao",L. Venkataratnam"and K.S. Rao
Centre of Studiies in Resources Engineering
Indian Institute of Technology, Powai, Bombay-400 076, India
Fax: +91-22-57834180,Email: k,[email protected]
* National Remote Sensing Agency, Agriculture and Soil Group
Balanagar, Hyderabsd-500 037, India
STUDY AREA AND DATA SOURCES
Abstract:- ERS-1 and JERS-1 S A R dlata acquired wiith one
day difference were compared to each other to study the effect
of frequency and incidence angle on soil moisture and various
land cover features. Optical data from IRS-1 satellite were also
used for comparison with microwave data. Temporal ERS-1
SAR data acquired in three months (3 scenes) were also
included in the analysis.
The study area for the analysis of three sensor data is located
around Rajahmundry town with Latitude 17".1 ' and Longitude
81".47', Andhra Pradesh, India. The area contains wide variety
of cover types such as reserved forest, plantation, croplands
(paddy, sugarcane, pulses), barren land, urban area, hilly
terrain and water bodies. Some part of the study area is
inundated with water in rainy season and later used for
The response of ERS-1 SAR data for soil moisture is better plantation of commercial crops such as tobilcco, chilly and
than that of JERS-1 SAR data. Effelct of roughness and
cotton in the winter.
vegetation on soil moisture was observed in both the SAR data
The details on data acquisition from different sensors are
sets. The temporal changes in soil moisture using three ERS-1
S A R scenes were clearly seen. Classification accuracy of given in Table I.
various land cover features are very poor in microwave (data as
compared to optical data.
GROUND TRUTH DATA
INTRODUCTION
There has been growing interest in remote sensing community to use microwave SAR data of JERS-1, ERS-1,2 and
Radarsat satellites due to its availability irrespective of cloud
cover and sun illumination. Recent studies [1]-[5] using S A R
data have been shown that ERS-1, JERS-1 S A R and SIR-C
systems are useful for observing soil moisture and dynamics of
agriculture crops as well as providing land cover mapping and
area estimation.
The objective of this research is to studly the ability of ERS-I
and JERS-l sAR data for 'Oil moisture and land 'Over
classification. For this, we have used the available data of
JERS-I, ERS-1 S A R and optical data of Indian Remote Sensing
(IRS-1) satellite. In the subsequent sections, study area, ,ground
truth, data sources and results and discussions are given.
In January, many fields were harvested paddy fields. Some
harvested fields were used for growing second crops such as
pulses (green and black gram) and sunhamp. The crops were
grown and fully covered the soil surface. Some harvested area
were again planted with paddy and it was grown about 30 cm
height. Many sugarcane fields were closed to harvesting stage
and some fields were harvested. Low-lying water logged areas
in rainy season were used for growing commercial crops such
as tobacco, chilly, tomato etc. in winter season. The height of
the tobacco was 1.2 m with big leaves and flowers. Water
content in a single leaf was 42 grams. Each plant had for about
13
Therefore, the tobacco plant with leaves and stalk
(diameter 2 cm) may have highest water content as compared to
other crops. The area also
mango, cashew plantation and
reserved scrub and highly grown thick forest. The soil
moisture in many crops vary from 5% to 18% of weight in
J;anuary 1993.
0-7803-3068-4/96$5.00@1996 IEEE
163
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on July 3, 2009 at 07:21 from IEEE Xplore. Restrictions apply.
Table I. Different sensors and their data on different dates converted to backscattering values using available calibration
used for the analysis.
equations [ 6 ] , [ 7 ]It. can be seen from Figs. 2 and 3 that points
related to tobacco and rough fields were deviated from fitted
Sensor
Band
Resolution
Acas. Date
line obtained with smooth bare, medium rough and sparce
vegetation fields. Even though the soil moisture in tobacco
18 m
Jan. 22, 1993
ERS-1 S A R L-band
fields were low, its oovalues were high due to high vegetation
30 m
Jan. 21, 1993
ERS-1 S A R C-band
water content in tobacco. High correlation is observed between
(SARI
Dec. 17, 1992
ERS-1 0" and soil moisture. However, the sensitivity is low as
compared to JERS-1 S A R . Large scattering in JERS-1 Plot may
Feb. 25, 1993
be attributed to the restrictions on the use of JERS-1 calibration
IRS-1
2,3,4 bands
36.25 m
Jan. 2, 1993
equation.
All the images were classified individually and combinely
using supervised maximum-likelihood classifier and the
DATA PROCESSING
classification accuracy is given in Table 11. The classification
ERS-1 and ERS-1 S A R data were resampled to IRS-1 optical accuracy using single S A R data is about 30%, whereas optical
data (36.25 m) and co-registered all the images. Different color data gives 67%. Combinations of optical and microwave data
combinations using ERS-1, JERS-1 S A R and IRS-1 data were gives a small increment (10%) in ClaSSifiCatlOn accuracy.
made for better identifications of various features. Principal
component and IHS analysis were also done. Ground truth soil
moisture locations were identified in January S A R images and
CONCLUSIONS
average digital numbers (DN) of 3x3 window were taken.
JERS-1 and ERS-1 S A R data were analysed for soil moisture
and land cover classification. Large differences in the tone of
image were observed between ERS-1 and ERS-1 for paddy and
RESULTS AND DISCUSSIONS
pulses. Correlation between oo and soil moisture is higher
for
~ i 1 shows
~ . 42 km by 36 km area of JERS-1, ERS-1 sm ERS-1 as compared to JERS-1. The relations are to be verified
images on different dates. F~~~ comparison of ERS-1 and Using more ground truth data points. Temporal Variations ln
~ ~ s s-m1 images in J
~ it is observed
~
~that various
~
ERS-1
~
S A, R data using three scenes show that the sensor is
features can easily be identified in JERS-1 sm imagery as suitable for soil moisture monitoring. As the classification
poor using sAR data
is
compared to that of ERS-1. Paddy fields are seen as dark tone accuraCY
(extreme lower right area) in ERS-1 imagery, whereas it is maximum-likelihood classifier, new techniques based on
networks can be
bright in the ERS-1. This can be attributed to the larger texture and
incidence angle (35") of JERS-1 as compared to that of ERS-1
(23"). Just above paddy area, pulse crop is seen dark tone in
ACKNOWLEDGMENTS
JERS-1. The same area is bright in ERS-1. The centre of the
image is mango-cashew plantation which can be seen clearly in
The authors are thankful to Pr0f.V.S. Chandrasekaran, Head,
E R S - 1.
CSRE for his constant encouragement and support for this
Temporal ERS-1 S A R scenes show variations in brightness. work. We are grateful to DST for the financial support under
Out of three ERS-1 scenes, Dec. 17 scene is very bright due to the grant ES/23/13 1/91.
cyclone and rainfall in this month. Maximum tonal variations
is observed mainly in agricultural areas as compared to highly
grown plantation and forest areas. In February, overall imagery
REFERENCES
is dark due to dry soil conditions in this month.
[ 11 M.C. Dobson, K. Sarabandi and F.T.Ulaby, "Preliminary
Soil Moisture Analysis
analysis of ERS-1 S A R for forest ecosystem studies," IEEE
Ground truth soil moisture values in January 22 of bare
Trans Geosci. and Remote sensing, Vol 30, No 4, pp. 203smooth, medium rough, rough, and crop fields were plotted
211, 1992.
against backscattering coefficient (0') in Figs. 2 and 3 for
ERS-1 and JERS-1 S A R respectively. DN values were
164
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on July 3, 2009 at 07:21 from IEEE Xplore. Restrictions apply.
[2] S. Mohan et al., “Estimation of soil moisture using EIRS-1
SAE7 data, Proc. second ERS-I symposium, Hamburg, Table 11. Classification accuracy of different features using
various combinations of sensor systems.
Germany, Oct. 11-14, pp. 241, 1993.
I
I
I
I
1
131 K. Musiake et al., “Extraction of soil moisture informat’ion
Kogakuin University, Tokyo, Japan, pp. 3 11-326.
[4]T. Nishidai, ‘‘Early results of Fuyo-1, Japanese Earth
Resources Satellite (JERS-l).” Int. J. Remote Sensing, Vol
14, NO 9, pp 1823-1833, 1993.
[ 5 ] P.C. Dubios, J. van Zyk abd T. Engman, “Measuring soil
moisture with imaging radars,” IEEE Trans. Geosci. and
Remote Sensing, Vol. 23, N0.4, DD. 915-!)26. 1995.
[6]
guide;o ~JASD’A’~
SGproducts,’ EOC, NASDA,
Japan, March 10, 1993.
171 D.R. Paudyal and J.Aschacher, “ERS-1 S A R data calibration at the Indian National Remote Sensing Agency,” AsianPacific Remote Sensing Journal, Vol. 6, N0.2, PP. 117-1 19,
1994.
JERS-1 SAR
Settlement 59.8 88.6 92.8
Tobacco 51.4 71.4 71.4
hlses
49.4 57.1 63.3
Paddv
0.0 5.0 79.3
Fallow
0.0 2.8 44.4
R.forest 0.0 3.8 3.9
Mango 52.1 2.0 47.9
Tapioka 47.7 0.0 25.0
Water 27.2 14.9 43.9
Sugarcane15.6 62.2 55.6
Sand
64,2 o,O o,o
93.8
97.1
59.2
91.4
38.9
9.1
85.1
66.7
1,2
Overall 33.4 30,0 48.0
51.2
0.0
20.8
99.0
97.1
65.3
67.9
68.1
18.2
45.8
2.0
86.0
77.8
4,9
57,5
55.7
100.0
83.7
99.3
97.9 100.0 92
100.0 100.0 35
’75.5 87.8 49
72.9 100.0 140
100.0 100.0 100.0 72
63.6 68.8 84.4 77
41.’7 43.7 37.5 48
27.3 43.2 20.5 44
17.5 ‘97.4 57.9 114
48.9 88.9 77.8 45
7,4
81
67.1. i,2,3 78,7
Temp. - Temporal; B-bands; E-ERS- 1; J-JE:RS- 1
ERS-1 SAR
ERS-1 SAR
ERS-1 SAR
Fig. 1. JERS-1 and ERS-1 S A R images of an area 42 km by 36 km acquired on different dates.
-7-
(0)
ERS-1
.
SkR
SWI
(b) JERS-1
0
-lot
-1 1
/
’
SIope =43.2 8
r= 0.7 403
A
A
A
-1 3
*
-1 4
* Bare8rmedium
A
D
A
rough
Tobocco
Sunhamp
Rcugh Fields
Soil Moisture (W Weight)
Soil Moisture (sg Weight)
Fig. 2 . Backscattering coefficient
in January.
*
(GO
) versus soil moisture for (a) ERS-1 SAR and (b) JERS-1 SAR data
165
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on July 3, 2009 at 07:21 from IEEE Xplore. Restrictions apply.