STUDY OF THE SPECTRAL RESPONSE OF SOYBEANS
Getulio T.Batista
Bernardo F.T.Rudorff
Instituto de Pesquisas Espaciais-INPE
Ministerio da Ciencia e Tecnologia-MCT
Av.dos Astronautas, 1758 - C.Postal 515
12201 - S~o Jose dos Campos, SP, Brazil
Commission Number VIr
ABSTRACT
The spectral responsesof two soybeans varieties planted monthly
(in a period of threemonthst and their relationship to percent
ground cover, biomass and yield were investigated .. Bidirectional
reflectance factor was measured throughout the growing period
(five campaigns at 20~day interval, approximately) of six
experimental plots of 5 x 10 m'size (three plots of each variety
planted at one-month intervall using a 17 band (visible and NIR)
portable radiometer. In each campaign, six measurements per plot
were taken in order to obtain a representative estimate of the
plot reflectance factor .. Plant height and development stage were
estimated concomitant to the radiometric measurements~ biomass
estimates were sampled three times per plot during specific
development stages (flowering, grain filling, and maturation)
and total plot production was measured at-harvest. The percent
ground cover was estimated through photographs taken simultaneously to the radiometric measurements. Results indicated that
differences in reflectance between the two varieties correspond
to differences in their biomasses. However, in the visible
peaked at 550nm there is a slight difference between the two varieties .. Preliminary analysis of spectral trajectory plots and
vegetation indices indicated that the reflectance measures are
related to the studied agronomic variables.
1. INTRODUCTION
Soybean CUltivation in Brazil experienced a great expansion
through the last decade when the national production jumped from
1.5 (1970) to 15 million tons (1980) (ElllBRAPA 1981) and for the
crop year of 1987/88, 18 million tons are expected "to be produced
according to current official Brazilian government estimates.
Although the majority of the area planted to soybeans (10 million
hectares, approximately) is in the southern and central part of
the country, there is a great potential to bring other lands into
production specially in the north and northeast of Brazil(EMBRAPA
1981).
Among several factors that determined this tremendous increase in
the soybean production in Brazil such as market,double cropping
with wheat, government credit (production and commercialization) ,
and the establishment of cooperative organizations, is the governmental support for research and technical assistance to the
farmers .. A good example is the establishment of the National
Center for Soybean Research (CNPSo), located in Londrina,parana
State as an E!:lBRAPA (Empresa Brasileira de Pesquisa Agropecuaria) unit ..
As a result from this research effort, new varieties became a vailable' specially to be planted in non-traditional soy})ean
areas. This is the case of the two varieties considered in this
study, which are being tested in the Paraiba Valley, which is a
rice producing (15,000 ha of irrigated rice - Miyasaka and Medina,
1981) and a livestock area.
These varieties were introduced in this region to provide alternatives for cattle feeding during the dry winter season and also for
green manure to be incorporated into the soil asa rotating crop
for the rice production areas keeping them free of weeds during the
winter time. These varieties are also being tested for grain
production at the Experimental Station of Pindamonhangaba, SP,from
the S~o Paulo government. They were selected among several other
varieties since 1956 because they are tolerant to photoperiodism,
low temperature, soil acidity, and excessive soil moisture, typical
of the Paraiba Valley soils (Miyasaka and Medina, 1981).
Several papers have reported conclusively that spectral data are
significantly correlated with photosynthetically active portion
of the crop canopy (e~g.Boutton and Tieszen, 1983; Holben et al.,
1980; Daughtry et al., 1984; Kollenkark et ale, 1981; Blad et ale,
1985).
Crist (1982) analyzing very large data sets from field spectroscopy studies has established the normal spectral profile for
soybean spectral response. However~ several factors, may determine deviations from that normal profile, one of them being
the planting date (early or late planting) .Therefore, it remains
important to investigate the spectral behavior of soybeans planted
during the winter growing season in a tropical environment.
This paper reports on an experiment in which spectral information
obtained throughout the growing cycle of two varieties of soybeans
planted at different dates during the winter time was investigated
and derived vegetation indices were correlated with total fresh
biomass, plant height, percent soil cover, and grain yield.
2. EXPERIMENTAL PROCEDURE
This experiment was conducted at the Hydro-agricultural Research
Center - Eng. A. G. Borba from the DAEE (Departamento de Aguas e
Energia ~letrica de S~o Paulo), located at Pindamonhangaba, S~o
Paulo (22954' Sand 45925' N).
Six plots corresponding to three planting dates (one month interval) of two varieties (IAC~9 and Santa Maria - 8M) were investi gated. The plots were 10 x 5 m with 0.5 m of row spacing, 25
seeds/m (manually planted) with 10 rows per plot on a flat clayey
soil (BT) which was fertilized before planting. Water, weeding ,
and preventive pest control were supplied as needed throughout
the growing cycle and all biotic and abiotic events were regis tered ..
Six spectral measurements (three centered on the row and three
between rows) per plot were obtained five times with 20 day -
VII ... 546
intervals, approximately, throughout the growing cycle with a
hand-held nadir pointing portable radiometer with 17
spectral
bands from 400 nm to 1050 nm. Measurements were taken between
10:40 and 13:50 in clear sky conditions. Only occasionally,
smoke from a nearby small plant appeared on the sky but care was
taken to avoid its interference. The radiometer with an IFOV of
10 0 was held at 3 m above the crop canopy. A reference reading
was taken from a BaS04 panel prior and after each on the crop
spectral measurement. The procedure used for the spectral data
collecting followed recommendations from Biehl and Robinson
(1983) and procedure used by Holben et ale (1980).
Simultaneously with the spectral measurement a 35 mm photograph
was taken for percent soil cover estimate as suggested by Cihlar
et ale (1987). Plant height was obtained close to the radiometric
measurements based on the mean of 10 plants per plot, per date.
Total fresh biomass was taken three times per plot during
beginning of flowering, grain filling, and maturation,and
therefore not necessarily coincident with the radiometric
measurements. Two meters of one row (1 m2 ) were randomly selected,
cut and immediately weighted per plot, per biomass estimate. The
grain yield was estimated by total plot harvesting (47 m2 ) .
3. RESULTS AND DISCUSSION
Before making the crop reflectance analyses, the performance of
the portable radiometer was checked. It was hypothesized firstly
that there was no change on the reflectance property of the
reference panel throughout the measurement campaigns, and secondly
that the sky was constantly clear (no atmospheric interference
changes), therefore any change on the radiance readings from the
reference panel was due to solar radiation and or due to
instability of the radiometer. Since the solar radiation changes
are quite predictable, the radiometer performance could be
assessed under the assumptions made. It can be noticed on the
figure 1 that the panel response is quite smooth throughout the
measuring period. In fact, the largest variation occurred on
September 15, when the difference between the largest radiance
reading (at 10:45) and the lowest (at 13:15) was around 10%
for the 550 nm and 8% for 850 nm wavelengths following the
variation on the sun-target geometry. The largest spurious
variation were around 2% for the two wavelengths analyzed. Therefore, for those clear sky conditions we could have spaced more
the measurements of the reference panels without compromising
data quality.
Figure 2 shows the seasonal variation on the spectral response
of the two varieties planted at three different dates (May 15,
June 19 and July 15). The spectral curves follow very closely the
pattern presented by Bowker et ale (1985).
VU-547
C.N.
3
340
320
3CO
11:00
11:30
Fig.l - Radiance measurements of the BaS04 reference panel for
the two extreme dates.
Figure 3 shows the seasonal spectral profile of the six analyzed
plots based on the ratio vegetation index (ratio of the near
infrared band centered at 850 nm and the chlorophyll absorption
band centered at 650 nm). It can be noticed that the Santa Maria
variety senesces earlier than the IAC-9 (plot 1 vs 2 and 3 VB 4).
Despite of only few measurements taken throughout the growing
cycle, the asymmetry of the soybean variability about the mean
described by Crist (1982) is noticeable on the profile of plots
1, 2 and 4 (Figure 3).
Figure 4 shows in a comparative way the spectral trajectory of
the soybean varieties studied. Plots 5 and 6, planted later,
presented the greatest variation on both visible and near infrared reflectance factors.
Figure 5 shows that spectrally the two varieties studied are very
similar. However, in the visible band, at the peak of green
reflectance (550 nm), the Santa Maria variety has a slightly
greater reflectance than the IAC-9, except for the September 15
observation on plots 5 and 6, which were planted on July 15 and
had lower than 20 percent soil cover.
Figure 6 shows that ratio vegetation index is quite well correlated with all agronomic variables studied. The correlation with
fresh biomass was exceptionally high, especially considering that
the spectral measurements were not taken at the same date of the
biomass estimates. This is very encouraging considering that these
varieties are planted with the primarily objective of biomass
production.
In spite of the fact that these varieties were planted during the
winter time, the results were quite comparable to other previous
findings (e.g. results reported by Perry and Lautenschlager (1984) ,
Holben et alo (1980); Daughtry et ale (1984); Kollenkark
et ale
(1981)) .
1... 548
501
PLOT5-SM
AWT6
AUG 25
~
!...
///
...
P~~_~ 1NOV 03
OCT 08
f/--PW~'~:IA~
,-----
IAC
I
/'
-~,
_",,~... , /....... J
1
50-1
SEPT 15
AUG 25
l-----
P
7
, ...
,
I tII!'I
PLOT3-IAC
~
// PLOT 4-S
j
PLOT4-SM
OCT 08
'
~/
PLOT3j
lAC --- NOV 03
/---- PlOT 3
~
,,/"PLOT
'
/
PLOT 4-SM
SM
PLOT3-IAC
,/
,/
.-,
NOV 30
,-,
,/-----"
~4
SM
lAC
/
/
//
-l
0
Q
...... 50
CD
~~;;6
1
0
a:
NOV 30
~LOT5
SM
lAC
J----';
/,i
PLO~;lAC1
SEPT 15
PLOT5-SM
400 550
j
AUG 08
O!!
400 550
PLDT,~-IA~
,
850
SEPT 15
PLOTl-~AC j
OCT 08
PLOTl-IA~
1NOV 03
?
850
1050
PLOTI-IAC,/
,F'"
"
-... ...
,
PLOT 2- SM
I
1050 400 550
850
1050 400 550
850
1050 400 550
850
1050
WAVELENGTH (N M)
Fig. 2 - Average spectral reflectance of the six plots, on five dates of two soybean varieties
(IAC-9 and Santa Maria-SM)
e
RVI
15
15
PLOT 5
SM
15
PLOT3
PLOT6
lAC
10
10
10
5
PLOT 4 5
SM
lAC
1
5
PLOT2
SM
to
M
C\J
0
~
r0 "'"
If)tO
C\J
r:-J
0
I"/')
.;;t If) to
0
I
I
>-Z-lCDCLI->UZ Ill
OZQ., LI.
>-z..J"a...I->UZ£D
C;:);:):')LlJUOWCW
;=e...,...,ccnOZQ.,LI.
C~:')~WUOI&ICLLl
2.,.,C"
Fig. 3 - Seasonal change in the ratio vegetation index (RVI) of
soybeans (1= planting, 2= flowering, 3= grain filling,
4= maturation, 5= harvesting of SM variety, 6= harvesting of IAC-9 variety).
B.R.F. TRAJECTORY60
;=e 50
z
o 40
an
~30
3
PLOT I-lAC
4
2
-'7
~ 20
LI..
0:
CD
5
3~-S'40
2
1
4
4
60
60
NM VS. 650NM
5
PLOT4-SM 50
5
30
5pLOT~O
4 "J.IAC
2
1 30
20
20
10
10
048
o
4
8
12
o
4
8
12 16 20
650 N
Fig. 4 - Seasonal spectral trajectory of two soybean varieties
(1- Aug 8, 2- Sept 15, 3- Oct 8, 4- Nov 3, 5- Nov 3D,
1987) .
Vll-.. . . . "'-'.~
-=-__
PL_OT 5- S M
60
50
40
30
2100
_lI"0
a:::
o
~-~-~~=-~--==~~~~~
-_1__ _ . . ____~
s~
-- .. - -- - -i"lc
~--~~~--~~.-~~~--~-,~~~
550
850 )..{nm)
NOV30
550 850
AUG25
o
I<.)
<t:
u.
~
60
~
50
40
t;
PLOT3-IAC
_--0--- ---_-(')
30
~
20
10 f=== ____
..SM
O~__~~~__~~.-~~~__~-,_-~~A~C~
550 650 550 850 [550 650
550 850)" (nm)
AUG25
SEPT15
OCT 08
NOV 30
a:::
..J
<t:
Z
o
i=
<.)
w
a:::
o
iii
Fig. 5 - Seasonal variation
on soybean reflectance
factor at two spectral
bands, centered at 550 nm
and 850 nm.
PLOT4-SM
~
*_
*_
....
60
50
PLOTT-IAC
40
~
PLOT 2-SM
30
20
10 I=~_=_=-_~~=~
O~--~~~--~~~--.-~--~-.-,--~
)\.(nm)
550
650
AUG 25
NOV 30
(a)
(b)
IS
2I-
;e
!!..
.- -co':,
:x:
~
..
!AI
.•
:x: 50
... ....
: .....:,C
. ..
·1
100
1()()1
•
••
•
a:::
>
u
---.
•••
50
e.
..J
II
II
0
r " 0,71
..
(I)
411
n .. 20
1
.. 4.220
t
STAT
0
co-
-•
!AI
0
.....
0
r "0.77
n ,. 56
f
"8.966
STAT
SC" 0.86+ 16.49 RVX-oSS(Rvrl
R2 ... 79
(e)
...
(d)
....
r
e
r :0.76
=0.88
16CO
n =18
..... co
2ro,L-~~~~~~~~~
o
10
20
RATIO
n=6
tTAT= 2.337
kAT=7.477
mro.L-__________________
0
10
20
VEGETATION INDEX(RVI)
Fig. 6 - Relationship between the ratio vegetation index (RVI)
and agronomic variables (a- plant height, b- percent
soil cover, c- total fresh biomass, d- grain yield) .
VII ... 551
4. CONCLUSIONS
In spite of the studied varieties of soybeans to have been
selected for biomass production during the winter growing season,
their spectral response in the visible and near infrared are very
similar to the pattern of soybean reflectance reported on previous
researches.
The ratio vegetation index (RVI) (near infrared/red) was highly
positively correlated with all agronomic variables studied (plant
height, percent soil cover, total fresh biomass, and grain
yield). The RVI explained 78 percent of the variation in biomass.
This result is particularly relevant considering that these two
measurements were not taken at the same date. Conventional
biolllass estimates at the field level can be very time-consuming,
therefore, spectral measurements might be an alternative method
for reliable biomass estimate of soybeans for the Paraiba Valley
in Brazil.
5. ACKNOWLEDGEMENTS
This experiment was conducted in cooperation with the "Instituto
Florestal de Sao Paulo" that provided the portable radiometer and
with the "Campo de Pesquisas Hidroagricolas - Eng. A.G.Borba"
from the Departamento de Aguas e Energia El~trica de Sao Paulo
(DAEE)" that provided the facilities and support for the field
experiment. Special thanks are due to Mr. Norberto Leite and
Mr. Hideyo Aoki, senior researchers from the DAEE experimental
station and from the "Instituto Florestal de Sao Paulo", respectively. The authors wish also to express their gratitude to Dr.
Raimundo A. Filho and Mr. Carlos A. Steffen for their help on the
field spectral data measurements.
6. REFERENCES
BIEHL, L.La; ROBINSON, B.F. Data acquisition and preprocessing
techniques for remote sensing field research. SPIE 356:143-149,
Society of Photo-Optical Instrumentation Engineers, Bellingham,
WA, 1983.
BOUTTON, T.W.; TIESZEN, L.L.
Estimation of plant biomass by
spectral reflectance in an East African grassland. J. of Range
Management 36(2}:213-216, 1983.
BOWKER, D.E.; DAVIS, R.E .. ; MYRICK, DeL .. ; STACY, K .. ; JONES, W.T.
Spectral reflectances of natural targets for use in remote
sensing studies. NASA Reference Publication 1139, Scientific
and Technical Information Branch, NASA, College Park, MD, 1985.
CIHLAR, J.; DOBSON, M.C.; SCHMUGGE, To; HOOGEBOOM, Pe; JANSE,
A.ReP.; BARET, Fe; GUYOT, Go; LeTOAN, To; PAMPALONI, P.
Procedures for description of agricultural crops and soil in
optical and microwave remote sensing studies. International J.
of Remote Sensing, 8(3) :427-439, 1987.
VIl . . 552
CRIST, E.P. Analysis of the profile characteristics of corn and
soybeans using field reflectance data. AgRISTARS IT-EZ-04130
NAS9-16538. NASA, Scientific and Technical Information
Facility, College Park, MD, 1982.
DAUGHTRY, C.S.T.; GALLO, K.P.; BIEHL, L.L.i KANEMASU, E.T.i
ASRAR, Go; BLAD, B.L.; NORMAN, J.M.; GARDNER, B.R. Spectral
estimates of agronomic characteristics of crops. Machine
Processing of Remotely Sensed Data Symposium, p348-355,
Purdue University, W. Lafayette, IN, 1984.
EMBRAPA-Empresa Brasileira de Pesquisa Agropecuaria. Programa nacional de pesquisa de soja. EMBRAPA - DID, 115p, Brasilia, DF,
1981.
HOLBEN, B.N.; TUCKER, C.J.; FAN, C.J. Spectral assessment of
soybean leaf area and leaf biomass. Photogrammetric Eng. and
Remote Sensing 46(5):651-656, 1980.
KOLLENKARK, J.C.; DAUGHTRY, C.S.T.; BAUER, M.E. Soybean canopy
reflectance as influenced by cultural practices. AgRISTARS
SR-PI-04038 NAS9- 5466, Purdue University, W. Lafayette, IN,
1981 ..
MIYASAKA, So; MEDINA, J.C.; ed. A soja no Brasil, Instituto de
Tecnologia de Alimentos, ITAL, 1981.
PERRY, C.R.; LAUTENSCHLAGER, L.F
Functional equivalence of
spectral vegetation indices. Remote Sensing of Environment
14:169-182, 1984.
VII ... 553