YM-N4-04455 J S C 18907 t
.* A Joir:t Program for Agriculture and Resources Inventory Surveys Through Aerospace -
Remote Sensing Yield Wocit:. Development
JANUARY 16, 1984
3P.r.Z 1 L SOYBEAN Y I E U ) COVARIANCE MODEL ( E $ 4 - 10 134) bE,Ai,IL S b Y t E A h Y I E L D C L V A E I M ~ C L hdk-2 1922 H(;XL (Natio:.al cceh1.i~ aria A t a o s p h e r i c A d m l c i s t r a t l o ~ ~ ) 25 p .;L i i 3 i / E k A 0 1 LSLL 02C Llnc id s
3 G O l i . 4
SUSAN L , CALL1 S CLARENCE SAKAMOTO
USDA/NOAA C I A /MODEL RANCH
b ROO 1 200 ? % E ERA BUIG fog CHERAY STREE! OLUMBIA, MO 65201
Lyndon B. Johnson Space Center Houston. Texas 77Q58
https://ntrs.nasa.gov/search.jsp?R=19840013854 2018-07-12T09:03:28+00:00Z
JSC Form 1424 (Rrv N m 75)
'For salo bv the Nat~onal Technical Information Smlce. Spcngfield. Virginia 22161
3. R m p a t ' s Catalog No.
5. Report Date
January 16, 1984 6. f'erformlng Organlratcon Code
8. Performcng 0rganczat:m Rcpat No.
10. Work Unct No.
11. Contract or Grant No.
13. Type of Repon and Percod Covered
14. Sponsoring A- code
1. Report No.
\IM-N4-04455, ;SC 18907
NASA - JSC
2. Gowrnment Accesscon No.
4. Tctle and Subtctle
B r a z i l 5ovbean Y ie l d Covariance Model
7. Author(s1
Susan L. C a l l i s and Clarence Sakamoto
9. Perforrncng Organlation hbme and Address
USDCjNOAA CIAD/Model s Branch Room 200, Federal Bldg, 608 Cherry S t - Columbia, MO 65201
12. Sponsor~ng Agency Name and Address
Na t ions l Aeronautics and Space Admin i s t r a t i on Lyndon B. Johnson Space Center Houston, TX 77058
IS. Supplementary Notes
16. Abmm
A model based on m u l t i p l e regress ion was developed t o es t imate soybean y i e l d s f o r t h e seven soybean-growing s t a t e s o f B r a z i l . The meteoro log ica l data o f these seven s ta tes were "pooled" and t h e years 1975 t o 1980 were used t o model s ince the re was no techno log ica l t rend i n the y i e l d s du r ing these years. P r e d i c t o r va r iab les were der ived from monthly t o t a l p r e c i p i t a t i o n and monthly aversge temperature.
17. Key Words (Suqgsstsd by Authu(s1)
Mu1 t i p 1 e regress i o n ana lys i s P red ic to r va r iab les
'8. D~sn~butcon Statement
19. ~ s c w ~ t y a m l f . (of thk myor*'
Unc lass i f i ed 1 A
21. No. of Papss
I 2 4
20. Secur~ty Classif. (of thls psg)
Unc lass i f i ed
22. Rice'
BRAZIL SOYBEAN Y I E L D COVARIANCE MODEL
Susan L. C a l l i s
and
C l arence Sakamoto
AISC Models Branch
January 16, 1984
I NTRODUCTI ON
The put-pose o f t h i s study was t o se lec t weather va r iab les t h a t could be
used t o estimate the soybean y i e l d s f o r t he country of B raz i l . Soybean produc-
t i o n i n Brazi 1 had i t s incep t ion i n t he s t a t e of R io Grande do Sul where
soybeans were f i r s t p lanted i n r o t a t i o n w i t h wheat. Present ly, soybeans a re
planted i n seven southern states: R io Grande do Sul, Santa Catarina, Parana,
Sao Paulo, Mato Grosso, Goias and Minas Gerais. This means t h a t a v a r i e t y o f
growing condi t ions are covered i n t h i s extensive area. F igure 1 i s a map o f
t he soybean-growi ng areas of Brazi 1.
The northern pa r t o f t h e soybean area, i nc l ud ing t he s ta tes o f Mato Grosso,
Goias, and Minas Gerais, a re character ized by a "wet-arad-dry" c l ima te w i t h a d r y
season o f two or more months. The r e s t of t he area i s subt rop ica l humid w i t h
r e l a t i v e l y abundant r a i n f a l l which i s we1 l - d i s t r i buted throughout t h e year bu t
w i t h s l i g h t l y more r a i n f a l l i n t he warm months. Summers are hot and w in te rs
r e l a t i v e l y mild. Temperatures i n t h e summer are above 400C i n t he p l a i n s o f
R io Grande do Sul when warm a i r masses penetrate t h e plains. Southern Sao
Paulo i s t he northern l i m i t f o r f r o s t occurrence.
The soybean growing season begins w i t h p l a n t i n g i n October and November and
runs through A p r i l and May when harvest s ta r ts .
METHOD
Mu1 t i p l e regression ana lys is o f y i e l d w i t h se lected agroc l imat ic ind ices
was used t c der i ve a su i t ab l e model. The f i r s t approach taken was t o develop a
model f o r each state. However, l a ck o f data f o r several s ta tes posed a problem
and no su i t ab l e models were derived. Therefore, i t was decided t o c rea te a
covariance model whereby a l l t h e ava i l ab l e data a re combined t o ob ta in a model
f o r the country.
ORIGINAL PAGE C9 9F POOR QUALITY
Figure 1. Soybean-proving a r e a of B r a z i l . (Pitcher, 1971)
Since moisture stress was considered a prime determinant o f y ie ld , the
index P-PET (p rec ip i t a t i on m i nus potent ia l evapotranspiration) was used i n the
regressions.
The regression equation i s : A Y = U + BID + B2RDFNi + B3(P-PET)i + E
where A Y = Estimated y ie ld ,
a = Constant,
B j = Coef f ic ients of the variables j = 1 - 3,
D = Dummy var iable t o adjust each state 's y i e l d t o a base y i e l d set by
Rio Grande's do Sul 's y ie ld ,
RDFNi = Deviat ion from normal o f t o t a l p rec ip i t a t i on f o r month i ,
P-PET1 = P rec ip i t a t i on minus PET f o r month i , and
E = Unexplained e r ro r
I n developing the model, various procedures o f the S t a t i s t i c a l Analysis
System (SAS I n s t i t u t e Inc., 1979) were used. The procedures used and the
operations performed w i t h each are summarized i n t he Appendix. The .elected
model had the highest R~ and included variables s i g n i f i c a n t a t (o r close t o )
the 10 per cent leve l and agronomically meaningful.
DATA
The Braz i l crop data f o r 1961 t h r u 1977 were obtained from the Foreign
Agr icu l tura l Service (Sam Ruff, personal communication, 1982). The data were
recorded w i th year o f y i e l d as year o f harvest, so the weather in f luencing
the crop occurred during year-1.
Meteorological data from 1975 through 1980 were used t o model because
there i s no apparent trend i n the y i e l d data during t h i s period. Table 1 1 i s t s
- ORIGINAL PAGE M' OF POOR QUALITY
a Montes Cleros
Figure 2. Location of Meteorological Stations Used to Derive Data Sets for the Brazil Soybean Model.
STATE - Goias
Mato Grosso
Minas Gerais
Parana
METEOROLOGICAL S TAT1 OK
Bras i l i a /Cruze i ro Goiania
Ponta Pora Pres Prudente Foz do Iguacu
Montes Claros Araxa
Londrina Santa Branca Foz do Iguacu Por to Uniao
Rio Grande do Sul Ij u i Veranopolis Farroupi iha Sao B ~ r j a Passo Fundo J u l i o de Cas t i lhas Santa Maria
Santa Ca ta r ina
Sao Paulo
Por to Uniao Sao Joaquim Vacaria
Pindorama Ribei ro Pre to Mococa Bauru Jau Limeira Camp i n a s Monte Alegre do S u l T i e t e Sao Pa7:lo Jundiqr
Table 1. Meteorological S t a t i o n s Used t o Derive Data S e t s f o r the B r a z i l Soybean Model.
t h e s t a t i ons used t o der i ve t he meteorological da ta se t f o r each state.
F igure 2 shows the l oca t i on of each s ta t ion . The seven s ta tes o f R io Grande - 1
do Sul, Santa Catarina, Parana, Sao Paulo, Mato Grosso, Goias, and Minas Gerais i 8
were included i n the covariance model. 1
PROCEDURES ,
. The o r i g i n a l va r iab les used i n t h e regress ion equat ion included "dummy - .
v a r i ables" f o r s i x o f t he seven s ta tes (Rio Grande do Sul was t h e excep t i r k . ;
Thit "durnmy var iab les" ad jus t t h e con t r i bu t i ons t o y i e l d o f each o f t he s ta tes t
t o a base y i e l d which, i n t h i s case, i s Rio Grande do Su l ' s y i e l d . Weather
va r iab les selected r e f l e c t e d ava i l ab l e no i s t u re f o r t h e months December through
March. The var iab les which were s i g n i f i c a n t a t t he 10 per cent l e v e l were t h e
"dummy var iab les" f o r Sao Paulo, Parana, and Santa Catar ina and t h e P-PET fo r
January and February (see the Appendix f o r a d e f i n i t i o n o f P-PET). The coef-
f i c i e n t f o r the Santa Catar ina "dummy var iab le " was negative, i nd i ca t i ng ' i t s
y i e l d was below t h a t o f t h e norm set by R io Grande do Sul; t h e c o e f f i c i e n t s
f o r t he o ther dummy var iab les were pos i t i ve . The c o e f f i c i e n t s f o r January and
February P-PET were pos i t i ve , i n d i c a t i n g a need f o r mois ture du r i ng those
c r i t i c a l months. The s t a t i s t i c s o f t h e selected model a re summarized i n Table 2. I TEST RESULTS !
A j a ckkn i f e t e s t was run on t h e f i n a l model. I n t h i s tes t , a year was
e l im ina ted from the crop data and t he model was used t o p r e d i c t t h a t year 's
y i e l d . This process was done f o r each successive year beginning w i t h 1975.
The t e s t had t o be run separately on each state. The r e s u l t s a re p r i n t ed on
Tables 3 through 9 and p l o t t e d on Figures 3 through 9.
APPEND I X
Def in i t ion o f Variables
P-PET, p rec ip i t a t i on minus potent ia l evapotranspiration, i s an index used
t o measure the amount o f moisture avai lab le f o r p lant growth. Potent ia l
evapotranspiration i s determi ned by the procedure devel oped by Thornthwai t e
(1948) which uses only temperature:
PET = \ A /
where I = Heat index, which i s the sum o f the 12 monthly i ndices i .
T = Monthly temperature i n OC, and
a = An empir ical exponent, 6.75 x 10-713 - 7.71 x 10'512 + 1.79 x 10'21
The durat ion o f day1 i g h t i s used t o adjust po ten t ia l evapotranspiration as a
por t ion o f 12 hours.
S t a t i s t i c a l Analysis System Procedures Used
PROC CORR
PROC PLOT
PROC STEPW I SE
PROC STEPWISE FORWARD
PROC STEPW I SE BACKWARD
Computes cor re la t ion coe f f i c i en ts between variables, inc lud ing Pearson product-moment and weighted proca.ct-moment corre lat ion.
Graphs one var iable against another, producing a p r i n t e r p lot .
Provides f i v e methods f o r stepwi se regression. Stepwi se i s useful when select ing variables t o be included i n a regression model from a co l l ec t i on o f independent variables.
Begins by f ind ing the one-variable model t ha t produces the highest ~ 2 . For each o f the other independent variables, FORWARD calculates F- s t a t i s t i c s r e f l e c t i n g the cont r ibu t ion t o the model i f the var iable were t o be included.
Begins by ca lcu la t ing s t a t i s t i c s fo r a model inc luding a l l the independent variables. The variables are deleted from the model one by one u n t i l a l l the remaining variables produce F -s ta t i s t i cs s ign i f i can t a t the .10 level.
PRO6 STEPW I S E STEPW I SE
PHUC STEPWISE MAXK
PROC PETM
PROC ZINDEX
The stepwise method i s a modi f icat ion of the forward select ion technique, d i f f e r i n g i n t h a t var iables a1 ready i n the model do not necessari ly stay there. Af ter a variable i s added (as i n the forward select ion method) the stepwise method looks a t a11 tne variables alreaay lncluaea i n the moael and oeletes any var iable tha t does not produce an F-statistic s ign i f i can t a t the .10 level. Only a f te r t n i s check i s made and the necessary delet ions accomplished can another var iable be added t o the model.
(Maxlmum K 2 imp ~vement) Unl ike the three techniques above, t h i s method does not s e t t l e on a s ingle method. Instead i t looks f o r the "best' two-variable model, the "best" three var iab le model, and so for th.
Uses l a t i t u d e and mean monthly temperature t o ca lcu la te Thornthwai t e ' s po ten t ia l evapo- t ranspi r a t i o n f o r each month.
Uses monthly PET'S, p rec ip i ta t ion , SS (begi ning moisture i n surface layer), AWCS (avai lab le water capacity i n surface layer) , SU (beginning moisture i n the underlying 1 ayer ) , and AWCU (avai 1 able water capacity i n the underlying layer ) t o ca lculate Palmer's SON moisture budget, drought index Z, ET, and ET.
ORIGINAL PAGL iS OF POOR QUALITY
Orc Nmm r)dOrCaD .Dm.?)'-*
ORIGINAL PA= 1s OF POOR QUALW
0 P
(\ra'(Ca'Om L 0 9VI--+u o A ama-.lou
W C ~U'a2On'IP C - V) S O W u ' l c U I *0* a wb+abf- W J ...... C W O OOCOOC
a CCCCCC 2 0
A X Y W W I a m aw
U C d lL I1 w -0Q z 0 x P 0 II
a 1
. . . CCCCCC
~llcR13iRI.c 9 + b G f Q % a 3 c 5 4 m r t .--lc\th-V I& -cc---c 0 000003 ......
OOOOQO
(U9'-7'mnl 09nmrcru
ul a'LnOS(\rO a -mss(Uc C brYC0JI9+ w L I ( U I U I ~ ~ a 0000e0 ......
OOOCOO
-++a 0u'mlnmaJ (Cd-dQN 9a9mOLlr OVImnJ49 rl'rUch'cr(U . * . * * * OCOCCC
COOOOO
CCCOOC
OnlG!NkL E-.;':;Z 1s OF POOH Q'JkLIn
to +al la' I-' I I I I I I I I I I I I
~ -
-n (r. a 0 J 00
- . . - ~
Q) o a oocooo a' Z
(\Ia'crmu'.- SV,-@+U rCLn9- .04
C @lPCU'hll' V: . C O ~ W . l c l f ' Q b F P Q i * + ......
OCCCCC
r l c r - ~ r ! r n NbW\rcCtUC P-P-9.34nJ COEICdS -'FC(\I-U' -CC+.-C OOOOCO ...... 0003ao
NICrcgrtnnJ a'Q(?w-(U O-SC&D d*s<aS rC-9rn4l- r r R J d d d c r OCOCOO ...... CCOCOO
oeocco COaDaCaotF Q1@6U'Q@ m-cCIcCI
ORIGINAL F:iQT ::j OF POOR QUALlrY
4d60C6 VI +cc.&a'6
M & b l U @ e O C \ r W - (\lf-ma(C* a coccco a o a h a ~ a ~r. mm.qm;n;m 5 C c===c,?
0 I I 1
c c m - l c c 'a-999- +m--r.SQQ PlQ rl-em +r-+Q1- Q Q 4 4 Q 4 ...... OCCCOO
OOCCCO
IL c 0 J
LL: C c VIP >- W -1 )--LC I t-LL: I *e cn u LC* 2LL;c d > W
CCCCCC
--crOP.-?C +C(aCaC.-Q +-3ror-o d(UQQ(7;t emf-mom Ccc-do CCCCCC ...... OCCOOO
e - e c c m ~ ; ~ (CI0CPJIJ.- d r c o m Q d U'ISR-N W C 9 \ C 9 + f l ~ r t c c c c l O 0 0 0 0 0 ...... QOCCOC
PL
OT
O
F Y
IEL
D*Y
IEL
DY
H
PL
OT
,OF
OR
SY
IEL
D*Y
IEL
~~
YR
S
YY
HO
L U
SF
Q
IS
P
SY
MB
OL
US
ED
IS
o
Fig
ure
6.
State of Parana Soybean Model.
NO
VE
ORIWAL PAGE 1s OF POOR QUALITY
- ....*. PC w CeCQOO - o a oooooo 9 z
(L099@- SVI.-.lP-U rtrnDn0Q -V(CD.(FVI S O r - W b l n rcr-r-ar-+ ...... COOOOO
lL 0 0 2
LL' I- u m e * UJ + w o I c w I *+ ;n u to- *Ow nrr U'Ja zul
acr W C J lL II W -30 z C x f u I1 a Q 3
CCCCCC
m & J n A a r OOOOQO ...... OCOOCO
)-UP JJ 1'-u CC I>-)- x m VI U 57 COW >* $UC LTV: - >.a B Y L WWQ !Yu'
xu' a Lric~ + L I I W c -CO X J Z 0 F W Y I C- 0 II -9-
a LL'O r C >J z * U J . a 0-
W m WVI r -s C >c m
la +a 1 0 ' I - I I I
\ I
' . I
\ I
\ I
\ \
I \
I \ I -.
\ I . I
\ 1
I
B 1 6 +IC
0 1 u' / Id
0 I t 0
0 I
0 I
0 I
/ I rl
0 I 9) I 0
0 /
I 0 I 3
#' t I
e 0 cp
I 9) I .n l a c h +(c 0
V) I a' I- 4 I 1 I cn I I
0 a I Q t 1 6 3 1 A 0 I Z ; r * I 0
/ I IP-
2 c7 *(c u.l
1 a' 0 \ I d
\ I Q) U
\ I (d I U
\ I ul
\ I I
\ I h
\ I I Q) I LC
I 1 M
I I z I9 +P- 15 I #-l I I I I I I I I I I I I I I u' *h
1 u' I I. -*---*----+----*----.----+----+----+----+----+ 1
C
ORIGINAL PA= W OF POOR QUALITY -
W r -&903;0 a ooooco a momr-am (L mmrlC7r'I-l 2 ...... o o oooaoo
U 1 1 ) I
Dr-bCb+ ~ ~ ~ L - R J O m I?a(crrcw x
C b@(rCOb - w MQNHQC a ccococ Q: IcrcbCbr- Z ...... 0
OCOQOO U
A X E w w a e m
mu-
ORIGINAL PAGE IS OF POOR QUALITY
ORIGIhlAL PAGE 19 OF POOR QUALITY
If b 0 r(
ecys&ac Nrcu'mm- m 9 ~ 9 9 9 ~ m
.J C IcQD(\Ia'(Db - w m m w a c b u a ocoooo c a 99efi99 z 0 . . a * . . 0 I oooooc 0
wwti - rT=amCllra a m ~um-flb P Q*M.?~ trw o QWWJU- - me-m-c ~ U C J J c \ ruccb-~ a ~c'rca-J: LL ll W W 6mU@&- 2 ................. z c > m l ~ n a e 9 9 u --dd-c
baCFrnrrd -9mu'clc Ln4VI(EJIN Icu-rcbu' m0uw'P'9 C(Hr(CC(c(
COCOCO ...... COOCOC
L' J cut I-IAJ C C la-c mzi V, U x5 '-am *>- &WO VdC 4a.U A Y X WWQ KV) av Q
W O J > I L I I W C '-00 a . J Z 0 >W Y I C- 0 II J>- a a I&'* ") -0
>J z * I r : a 0- W d* a Mu, r -lf C a-c V)
ORIGINAL PAGE 1s OF POOR QUALITY