DEPARTAMENTO DE ~TODO
S
QUANTI
T
ATIVOS
AGR0
-
CLIMATIC MODELING AS AN AGRI
C
U
LTUR
A
L
RESEARC
H
TOOL
Fernando
L.
Garagorry Everaldo R. Porto(Trabalho apresentado no First Inter-American Syrnposium on Agro-Clirnatic Modeling and Inforrnation Systerns, Caracas, Venezuela, de 20 a 26 de setembro de 1981)
70333 Brasília, DF Brasil
Garagorry, Fernando L
Agro-climatic modeling as an agricultural research tool, por Fernando L. Garagorry e Everaldo R. Porto. Brasília, EMBRAPA-DMQ,
1981.
13p (EMBRAPA-DMQ/A/52)
1. Climatologia agrícola - Modelo. 2. Mo delos matemáticos. 3. Pesquisa Operacional. I. Porto, Everaldo R., colab. 11. Empresa Bra sileira de Pesquisa Agropecuária. Departamen to de Métodos Quantitativos, Brasília, D.F~ 111. Título IV. Série.
FERNANDO
L. GARAGORRY
AND
EVERALDO
R.
PORTO
EMBRAPA,
BRAZIL
S U M M A R Y
The paper
illustrates
the use of an agro-climatic
model
so as to suggest
agricultural
research
priorities.
More generally,
some of the main
aspects
of the modeling
activity,
as it
lSbeing
conducted
at EMBRAPA
(the Brazilian
Public
Corporation
for Agricultural
Research)
will
be
discussed.
INTRODUCTION
In this paper
we will
try to illustrate
the use of
an agro-climatic
model
in connection
with
more
traditional
techniques,
so as to suggest
agricultural
research
priorities.
More
generally,
the paper
will
consider
some
aspects
of the
modeling
activity
as it is being
conducted
at EMBRAPA
(the
Brazilian
Public
Corporation
for Agricultural
Research).
The
central
idea with
regard
to modeling
and simulation
at EMBRAPA
is that
several
feed-backs,
acting
upon
the researchers
and
the research
managers,
can be obtained
through
these
activities.
This will
help
them
to gain understanding
about
complex
systems
and set reasonable
research
priorities.
THE USE OF SYSTEMS
APPROACH
AT EMBRAPA
The
modeling
activity
15 done within
the general
fram~
work of
the
systems
approach
applied
to
agricultural
research
(ALVES
1975,DENT
&
ANDERSON
1971,DALTON
1975,SPEDDING
1979).In
other words,
it
is seen
as
an
essential
component
to the
use
of that approach.
So,
in order
to
place
the
modeling
activity
1n its appropriate
dimensions,
we must
review
briefly
some
aspects
of
the
systems
approach.
Firstly,
we must
identify
some
systems
that
should
be
studied.
Fig.
1gives
a short
list
of
labels
that
can be
attached
to some
of
the systems
that
are
of
interest
to
EMBRAPA.
Of course,
these
labels
are not
very
descriptive;
for instance,
"PESTS"
may refer
to
an insect-plant-water
system
whose
descrip-tion requiresa
lengthy
paper.
The
important
point
to be stressed,
is that we take
farms
("farming
systems")
as
the
reference
systems
to be studied.
AlI
other
systems,
which
may
range
from
the whole
country
to a single
animal,
should
be studied
in relationship
to
the recommendations
that must
be given
to the
farmers.
For
instance,
the study
of an agro-climatic
system,
of
the
weather-soil-plant
type,
for the semi-arid
region
of Brazil,
was
considered
essential
in order
to
develop
technologies
that
can be
acceptable
to
the farmers.
Secondly,
we must
characterize
some
activities
that
should
be carried
out
in order
to
use
the systems
approach.
They
are
necessary,
although
by
no means
sufficient,
if
we want
to
claim
that
the systems
approach
is being
used.
On one
side,
we
have
the traditional
activities
that can
be
roughly
grouped
into
experimentation
and sampling.
They
are
essentially
analytical,
and have
dominated
agricultural
research
up to present
days.
On the other
side,
some new
activities
should
be developed
in
parallel
to the
ones
that have
characterized
the
"experiment
station".
Here we include
the
use of mathematical
AGRO-CLIMA T IC
SOCIO- ECONOMIC
ECOLOGICAL
LESS
DETAIL
•
•
•
•
•
FARMS ~~---
FOCAL
SYSTEMS
(REFERENCE LEVEL AT E~RAPA)
•
•
•
CROP PRODUCTION
ANIMAL
PRODUCTION
FORESTRY
•
MORE
DETAIL
•
•
SOIL - PLAN - WATER
PESTS
ANIMAL
NUTRITION
•
•
m
e
nt o
f large
i
n
f
o
rmati
o
n
sys
t
e
ms
(for in
s
tanc
e,
a clim
a
ti
c
d
a
t
a
base).
T
h
ese
a
cti
vi
t
ies
are es
sentiall
y
sy
nth
e
ti
c
,
at l
ea
st
l
n
the obv
io
us
s
e
ns
e
th
a
t
th
e
y try to consider
s
i
multa
n
eously
a
lar
g
e nu
m
b
er
of
va
ria
bl
es
to
ge
ther
with
th
ei
r
r
e
lationsh
ips
.
B
ut
,
mor
e
t
han
that,
i
n
connec
t
i
o
n
with
comput
er
mod
els
a
n
d
case
s
tud
ies,
a conscious
effort
is ma
d
e
to stud
y
n
o
t onl
y
h
ow b
ut
a
l
so
why the system
works
the way
it does
.
This
is
a ce
ntra
l
aspect
of
t
he systems
approach,
which
requires
sy
nt
he
s
is
ra
th
e
r
th
a
n
a
n
alysis
(ACKOFF
1979a,
1979b)
.
In th
i
s
c
ont
ex
t,
s
y
nth
es
i
s
m
ea
n
s
not
o
n
ly
t
h
e com
b
i
n
a
t
ion
of
th
e
d
i
ff
e
r
e
nt
compon
e
nts
of
a s
ystem
,
but
also
the considera
t
ion
o
f
l
a
r
ge
r
sys
t
e
m
s
,
th
a
t
cont
a
in
th
e
on
e we are interested
in
.
For i
n
s
t
a
n
ce,
t
o stud
y
f
a
rmin
g
sy
st
em
s
in the semi
-
arid
tropic
of Brazil
, it
was co
n
si
d
ere
d
neces
s
a
ry
to
stu
dy an agro
-
clima
t
ic
system
at
th
e
c
ou
n
ty
leveI;
a
ls
o,
t
he
study of some other
systems
,
con
t
ai
n
i
n
g
eco
nom
ic,
m
a
rket
i
n
g
a
n
d
employment
aspects
will
be require
d
.
F
o
r exa
mp
le
,
i
f
a
n
agr
o-c
l
imatic
or a farm management
mod
e
l
indi
cate
s
th
a
t
th
e
b
es
t
th
i
n
g
is to plant
mellons
and most
farme
r
s
in
a regio
n
follow
th
i
s
recommendation,
the result
may be an econom
i
c
disas
te
r
.
THE ROLES
OF MODELING
AND
SIMULATION
In this section,
we wi
ll
d
i
s
cuss
th
e
m
al
n
f
unction
s
th
a
t
we
associate
to the modeling
an
d
si
mul
a
t
i
on
ac
tiv
i
t
ie
s
,
w
ith
i
n
th
e
research
environment
of EMBRAPA
.
A
t th
is
s
t
age,
i
t is
a
ssum
e
d
th
at a group
of researche~has
ide
nti
f
i
ed
a
n
impo
rt
a
nt
s
y
st
em,
a
n
d that
they are willing
to alloca
t
e
a s
u
b
stant
ial
part
o
f
th
eir
time
t
o i
t
s study
.
W
e ca
n t
a
k
e a
s
a
n
exa
mpl
e
a water
-
soil
-
plant
system,
th
a
t
can b
e
commo
n
t
o
m
a
n
y
f
a
r
mers
in a given
region.
Tipically
,
the
rese
a
rch
e
r
s
h
ave
som
e
kno
w
led
ge
a
nd biolo
gi
cal
d
a
t
a
rel
e
v
a
nt
to
the stu
dy
.
T
h
e
d
a
t
a were
ob
t
ai
n
e
d
throu
g
h
th
e
u
se
of tradition
a
l
techniqu
e
s
,
appl
ied
w
i
t
hi
n
t
he wo
rk
o
f
s
pecific
d
is
ciplin
e
s
(
pl
a
n
t
physiolo
gy
,
soi
l ph
y
s
i
cs
,
wa
t
er
ma
n
age
m
e
nt
,
et
c
.
)
.
W
e ar
e
not
referrin
g
h
ere
to
th
e c
l
i
m
atic
da
t
a
th
a
t h
av
e
been
recorded
ov
e
r
the
yea
r
s
.
Later
on we will
co
mm
e
nt
on th
e
av
a
ilability
of
Suppose
th
a
t it
is
found
that
a mathematical(computer)
model would
b
e
useful
to und
er
stand
how and why
the system
works
the way it does,
and how
it could
be changed
or controlled.
First,
we note th
a
t
the model
i
s
not
an end
ln itself,
but
a means
to
gain unde
r
st
a
nd
i
n
g
.
S
e
c
o
nd
,
a
s soon as w
e
start
the job of
building
th
e
m
ode
l
,
mos
t
pro
b
a
bl
y
w
ith
th
e
help
o
f a modelin
g
expert,
a f
i
rst
fee
d-b
ack,
ac
t
i
n
g
upon
th
e
researchers,
appears
ver
y c
le
arly
.
This
is indica
ted
by
(1)
i
n Fig.
2
.
In fact,
it
is found
very
s
o
o
n
th
a
t
th
e av
ail
a
ble
knowledge
a
nd data
are not
enough
to bu
i
ld
th
e
mo
de
lo
T
he r
e
s
e
archers
must
perform
sever
a
l
activiti
es,
l
i
ke
r
eviewi
n
g
t
h
e
literatur
e
,
formulating
new
hypothes
es
and holdi
n
g
discu
s
sio
ns
w
ith
oth
e
r
scientists.
B
esides
that
,
th
e original
multidiscip
l
i
n
a
r
y
t
eam
st
a
rts
to work
in an
int
e
rdis
ciplinary
way,
50as to
comb
i
n
e
th
e
points
of view
and
the tech
niques
of several
discip
l
i
n
es
(
BI
R
N
BAUM
19
7
9
,
PAYNE
19
7
9).
T
h
en,
sup
p
ose
that, wi
th
the av
a
ilable
information,
we
are abl
e
t
o formula
t
e
a ma
th
emat
i
ca
l
mod
e
lo
At
this
stage,
a
se
cond
feed-back
(in
d
icated
b
y
(2
)
i
n Fi
g
.
2) can be activated
throu
g
h
the use of simulatio
n.
I
n this p
a
p
e
r,
the word
s
i
mulation
i
s
u
sed
in the sense
of numeri
c
a
l
ex
p
e
riment
a
tion.
It includes
pa
r
a
met
erization
and sensitivi
t
y
s
tud
ie
s,
p
e
rformed
with
the modelo
The simulation
exerci se
c
a
n
i
ndic
a
te
some points
where
more
re
s
earc
h
is necessa
r
y
,
t
oge
ther
with
some
oth
e
rs
where
the
availabl
e
k
n
owledge
sho
ul
d
b
e
c
o
nsidered
s
a
tisfactory.
Besides
that, it
may point
to seve
r
a
l
we
aknesses
in th
e
model,
50that
some corre
c
t
i
on
s
s
ho
uld
b
e
m
a
d
e
.
T
h
e two feed
-b
acks
desc
ribed
abo
v
e
are the most
important
,
at th
e presen
t
s
t
age
of
t
he
m
o
d
e
ling
act
i
vity
at EMBRAPA.
Dependin
g
upon the system
und
e
r
study
and the available
model,
s
o
me other
feed-backs
can be activated.
For instance,
we can show
the model
outputs
to external
exp
e
rts,
extension
agents
or farmers,
and ask
them to critici
ze
the results.
But,
in
g
eneral,
a
t the present
stage,
our mod
els
a
re not
int
e
nd
e
d
to
a
d
v
ise
the farmers
directly
;
RESEARC
HERS
®
CD
SIMULATION
MODELING
DATA
I---.!BASE
Fig.
2-
Two main feed-backs produced by modeling and simulation.MATHEMATICAL
(COMPU TING)
OUTLINE OF AN AGRO~CLIMATIC MODEL
In this section we will present an OVerVleW of an agr~
climatic model for the semi-arid tropic of Brazil. A detailed
description will be published elsewhere. Here, only a summary
of the main aspects, illustrated in Fig. 3, will be given.
The system is of the water-soil-plant type. For each
planting period and for each year a water balance is carried
out and the productivity is estimated.
Geographic boundary. A county with meteorologic station.
Crops. Only annual crops are considered.
Climatic data. The model requires daily rainfall data for a
series of years (tipically, between 20 and 50 years are avail
-able) and mean monthly temperature and relative humidity. In
the present version, solar radiation and potential evapotransp~
ration are calculated with formulae proposed by Hargreaves, that
make use of the temperature and relative humidity data and take
into account the latitude of the meteorologic station. The
com-puter program, written with small modules, is very flexible, 50
that different routines may be used, in the future, to estimate
potential evapotranspiration.
Soil. Water holding capacity is estimated for the most frequent
type of agricultural soil in the county.
Planto Root development is estimated for the crops of interest
(mainly corn, beans and sorghum), for the chosen type of soil.
Evapotranspiration coefficients are estimated or, else, taken
from the literature.
Simulation step. The user specifies the length of the planting
period (tipically, 5 days). If one chooses 5 days, the year is
divided into 73 periods. Then, the daily rainfall and potential
evapotranspiration values are grouped into five days values.
Productivity. For each planting period and each year, the model
estimates the productivity (defined as the ratio between actual
RAINFALL
EVAPOTRANSPIRATION
ATMOSPHERE
DATA
• TEMPERATURE
BASE
·HUMIDITY
• SOLAR RADIATION
SOIL
PLANT
• WATER HOLDING
~-fI •ROOT DEVELOPMENT
CAPACITY
·CANOPY DEVELOPMENT
USER'S
DATA
FOR
EACH
RUN
MANAGEMENT
·PLANT
I--~·SOIL
·WATER
PRODUCTIVITY
VALIDATION
(FIELD EXPERlENCE)Fig.3 - Basic components of an agro-climatic model for the semi - arid tropic of Brazil (annual crops).
ent response functions can be easi1y tested. The user must glve two va1ues between O and 1, so as to c1assify the estimated pro-ductivity into "good", "fair" ar "bad". For instance, if PROD is the estimated productivity and the values 0.8 and 0.5 are given, then the result is considered to be good if PROD ~0.8, fair if 0.8~ PROD 2 0.5 and bad if PROD <0.5.
Outputs. The computer gives two kinds of outputs: numeric and graphic. For each planting period, the program gives the rela -tive frequency of good, fair and bad productivity results (also the relative frequency of "acceptable" results, defined as the sum of good and fair ones, is printed); mean runoff and mean water deficit are also printed. These same values are plotted. Validation. The model results corresponding to each of the last
ten years with available weather data are compared to field ex -perience, with the cooperation of farmers and extension agents.
USE OF THE MODEL
Example 1. We will illustrate the use of the model with a simula tion that takes the form of a traditional
2x2
factorial experimento Probably, most people would say that we run four simulations. We prefer to say that we run one simulation that consisted of four trials. So that the term simulation carresponds to a hypotheti -cal field experiment, and each trial is equivalent to one experi -mental plot.In this first example, we consider the county of Irecê, in the State of Bahia. This is the main beans producing county ln Brazil; 50, beans is the chosen crop.
In arder to run the model, the user must glve a set of evapotranspiration caeficients (k ) and a set of yield response
c
factors (k). Now, we have twa sources for these sets af values: y
we can take them from the literature ar we can use the values estimated by our researchers. So, we have four cambinations that take the form of a
2x2
factorial experimentev
al
u
es
gi
ve
n by DOORENBOS
&
KASSAM
(
1
9
7
9)
,
a
nd with
EMB
R
APA
th
e
values esti
ma
te
d
by
E
MBRAP
A
's
rese
ar
che
rs
a
t th
e
C
e
nt
er
for
Agricultur
a
l
Re
s
ea
rc
h
af
t
he Semi
-
A
rid
Trap
i
c
(CPATSA)
.
Table
1
g
i
ves
a
summary
a
f
some
af th
e
results
prin
te
d
b
y
th
e campu
t
er
.
TABLE
1.BE
S
T
P
L
AN
TI
NG
I
N
T
E
RVAL
AN
D PE
R
CE
NTA
G
E
OF YEA
R
S
WITH
ACCEP
T
AB
L
E
YI
E
L
D
(Caun
t
y:
Irecê
.
C
r
op
:
Be a
ns)
CROP EV
APOTR
AN
SPIRAT
I
O
N
CO
E
FFIC
I
EN
TS
(
k
)
c
YI
E
LD
R
E
SPO
NSE
FACTORS (k )
YF A
O
E
M
B
RAPA
F
A
Oc
t
23/0ct
27
O
ct
1
3/
a
v
6
O
sz
%
62%
E
MB
R
Nav
7/Nav
11N
av
7/N
av
16
A3
5
%
35%
P
A
C
an
c
l
usians.
Several
ca
n
c
l
u
s
ia
ns
ca
n b
e
d
e
duced
fram
th
e
study
af
th
e c
a
mp
u
ter
autput
We
l
ist h
ere
a
f
ew af
th
em.
(1)
The best plan
t
i
n
g
s
e
as
an
li
e
s
in l
a
t
e
Oc
tab
er
a
n
d earl
y
Navember
,
in agr
eement
wi
t
h
field
ex
pe
r
ie
n
ce
.
(2)
A c
h
a
n
ce
af
3
5
%
a
f y
e
a
r
s
w
ith
acce
pt
a
ble
yield,
w
hi
c
h
mea
n
s
ane aut af three y
e
a
rs,
is much
claser
to field
e
x
pe
r
ience
than
the 6
2%
val
u
e
.
(4)
The m
a
del
is very
r
ab
u
s
t w
i
th
re
g
ard
t
o kc val
u
es,
f
or
a
nyane
af the tw
a
sets
a
f k
values.
y
Th
e madel
is very
sensit
i
ve
with
rega
rd
t
o
k
val
u
es,
fo
r
yanyan
e
a
f
th
e
t
wa se
t
s
af k
values
.
cRecommenda
ti
ons.
S
ev
eral
re
c
omm
e
ndat
i
ons
can be submitted
to the
researche
rs,
50that the
y ca
n be
i
n a better
position
to s
e
t
reasonabl
e pri
or
i
ti
e
s
f
o
r t
he
i
r work.
Of course,
the ultim
a
te
decision
may
d
epe
n
d
o
n s
e
vera
l
f
actors
not
included
into
th
e
model o
r c
o
mplete
l
y
inde
p
enden
t
of the system
under
study
(for
instanc
e, the avai
l
a
b
i
l
i
t
y
of p
e
rsonnel
or equipment).
We
list
h
e
re so
me of the rec
om
mendations
s
ubmitted
to the researchers;
thes
e recommendations
are vali
d
f
or the county
of Irecê.
(1)
(2)
( 3)
Do
not plan
a
ny
"bes
t pl
a
nting
period"
field
experiment,
s
ince the mode
l
a
lr
eady
give
s
re
a
sonable
and robust
results.
G
ive high pri
o
ri
t
y
t
o
t
h
e st
ud
y
of different
water
stress
re
sponse
functions
.
P
lan
s
om
e
e
x
p
e
riment
where
the plots
are submitted
to d
i
fferent
leve
ls
of water
stress
at
dif-ferent stages
of the plan
t
cy
cle.
M
e
a
nwhile,
in other
s
imulations,
use the k
value
s
estimated
at CPATS
A
,
since
y
the results
they give are c
lose
to field experlence
.
Give low priority
to
a
n
y
new estim
a
tion
of the k
values;
c
the values
in the
l
i
t
e
r
at
u
re
should
be considered
s
atis-factory
for the t
ime b
e
in
g
.
E
x
ampl
e
2.
T
h
e c
o
u
nty
of Santan
a
do Ip
a
nema,
in the Stat
e
of
Alagoas,
is also i
n
t
h
e
s
e
mi-arid
tropic
of Brazil,
but
differs
quite c
learly
fr
o
m
I
rec
ê with
re
g
ard
to dominant
so
i
ls
and
rain-falI d
i
st
ributi
o
n
.
A
s
i
mulation
a
nalogous
to the one in the
previous
example wa
s performed
for the county
of Santana
do
Ipa-nema,
where
57 y
e
ar
s
of rainf
a
ll
data were
available.
Besides
the r
ai
n
fall
data,
the other
difference
with
the first example
lies in the root development
data;
in fact,
the model
requires
this information
in order
to relate
the type of soil with
the
planto
Table
2
gives a summary
of the results
for this e
x
emple.
TABLE
2.
BEST
PLANTING
INTE
R
VAL
AND
PERCENTAGE
OF YEARS
WITH
ACCEPTABLE
YIELD
(County:
Santana
do Ipanema.
Crop:
Beans)
CROP EVAPOTRA
N
SPIRATIO
N
COEFFICIENTS
(kc)
YIELD
RESPONS
E
FACTOR
S(k )
Y
F
A
O
EMBRAPA
F
A
May
2
6/
Ma
y
30May
21/May
30O
86
%
86%
E
M
B
R
May
26
/M
ay
30May
26/May
30A
81
%
79
%
P
A
From
the e
xa
min
a
tion
o
f
the computer
output,
s
umm
ar
i
ze
d
in Tab
1e
2
, w
e
c
onc1ude
that
the resu1ts
are very
robust
wi
t
h
regar
d
to d
i
fferent
sets
of k
a
nd k
va1ues.
Th
e
re-c y
f
ore
, for
th
e t
im
e
bein
g
,
w
e
do not s
e
e
any urgency
to repeat
the
f
ie1d e
x
p
e
r
i
m
e
nts
in order
to improv
e
our estimates
of the k
and
c
k
coeffici
e
nts.
In other words,
the mode1
suggests
that this
y
s
ort of
fi
e1d experiments
has
low priority
in that
county.
Examp1e
3.
The irrigation
specia
1
ists
at CPAT
S
A
have
deve10ped
or
adapted
some water
m
anagement
techn
o
1
o
gies,
th
at
c
o
u1d be
of great
h
e
1p
for the sma11
farmers
.
These
techno1ogies
are based
on the
storage
of runoff water
to be used
in shortage
periods.
They
un
d
eE
stand
that there
a1ready
exist
some
techno1ogies
that
cou1d
g
i
v
e
a
high yie1d,
if the farmers
cou1d
store
at 1east
100millim
e
ters
of
runoff
water.
The model
shows
that,
in alI the counties
already
studied, there
are several
periods
with
runoff
well
above
the 100
wllimete~
mark,
in a
large
percentage
of years.
On
the one hand,
this
confirms
the hypothesis
that
the main
problem
is not
the
lack of water,
but the uneven
distribution
of rainfall.
On the
other hand,
these results
show
that EMBRAPA
can disseminate
those technologies
among
the farmers,
and expect
a high
chance
af
successs;
but, before
doing
that,
the upmost
priority
should
be
given to
the feasibility
analysis
of the proposed
technologies
at
the farm leveI.
So,
if we consider
the
individual
disciplines
in-volved,
the priority
shifts
from water
management
(which developed
or adapted the technology)
to farm management
and economics
(to
see
if the technology
can be acceptable
to the farmers).
CONCLUDI
N
G
REMARKS
The central
idea of this paper
was
to present
the use of
an agro-climatic
model
as an illustration
of the general
approach
followed at EMBRAPA
with
regard
to modeling
and simulation.
A
similar account
could
have
been
illustrated
by means
of other
models
(for instance,
pest
control
or cattle
management
models).
In any case,
the work
on agro-climatic
modeling
is probably
the
best
example
to give
a thorough
picture
of the present
situation
with regard
to modeling
and
information
systems
at
EMBRAPA.
On the positive
side, we can
list
the following
aspects:
(1)
A
reasonable
know-how
has been
gained
with
regard
to
the de
-velopment
and maintenance
of computer
models.
Once
an
initial characterization
of the system
to be studied
has been
given,
a first version
of
the mathematical
model
and
the
cor-responding
computer
program
will
be
ready
in
a
few
weeks.
Since
both the model
and the
computer
program
are subject
to
frequent
changes,
the main
concern
shifts
from
formulation
to
maintenance.
Therefore,
a central
importance
is being
given
to modern
programming
techniques
(mainly
top-down,
modular
and structured
programming),
50as to simplify
the maintenance
(2)
There
is a large
amoun
t
of climatic
da
t
a
.
I
n
fact,
many
stations
have
coll
e
ct
ed
these
data
for m
o
re
t
h
a
n
sixty
years,
which
places
Bra
z
il
i
n
a very
good positi
o
n
to undertake
clim
a
tological
s
tud
ies.
(3)
EMBRAPA's
research
e
r
s
are very
interested
ln
u
si
n
g
computer
models.
The exp
erience
h
as made
it clea
r
th
a
t
c
o
mputer
technology
offers
,
perhaps,
t
h
e only
pos
sib
ili
t
y
to study
many comple
x
syst
e
m
s
.
(4)
EMBRAPA's
expandi
ng
computer
installation
is among
the bests
in Latin Americ
a
in th
e field
of scientific
d
a
t
a
processing
.
On th
e
negative
side,
we
can mention
two
p
r
ob
lems
that will
certainl
y
r
e
c
e
i
v
e
muc
h
attention
in the near
fut
u
re:
(1)
A
n on-lin
e
c
l
i
m
a
t
ic
data
base
is not
availa
b
le.
For the
most
pa
rt
,
th
e weather
data
are still
in sheets
and books,
scattered
t
h
r
ou
ghout
the country.
So that,
although
alI
kinds
o
f clima
t
ic
models
can be quickly
developed,
they can
no
t be
put
i
n
to operation
because
the data
are n
o
t
readily
available
for computer
processing.
In other
words,
the
m
o
deling
know
-h
ow
is far ahead
of the availability
of the
information
systems
required
to use
the models.
(2
)
M
o
st
re
s
earc
h
s
t
ati
o
ns
do not have
as yet
rema
t
e
terminaIs,
l
i
nk
e
d
to the c
en
t
ral
computer,
due
to telec
ommun
ication
pro
ble
ms.
This
is
t
he case,
for instance,
of the Center
fo
r Agr
i
c
ultur
a
l
Research
of the Semi
-
Arid
Tropic.
Therefore,
com
pu
ter
process
ing
is dane
at headquarters
a
n
d
th
e outputs
are
ma
iled
to th
e researchers
.
As
it can be ex
pect
ed
,
this
c
reates
conside
rab
l
e
difficulties
t
o
t
he
use of com
pu
t
er
mod
el
in
g
.
Once
these
problems
a
re sol
v
ed,
th
e use of computer
m
o
deling
and inform
a
tion
systems
in connectio
n
to
agr
o-clim
a
t
i
c
s
tudies
will beco
m
e
a matter
of rout
i
n
e
.
But
.
in the mea
n
time.
even with
th
e
dif
f
i
c
ulties
mentione
d
above
.
models
are already
being
used
at EMBRAPA
to set some
r
es
earch
priorities.
For instance.
the
model presented
in thi
s paper
had
a significant
impact
on project
af the Cent
e
r
f
o
r Agric
u
ltural
Research
af the Semi
-
Arid
Trapic.
Several field exp
er
i
ments,
with
different
craps,
will
include
th
e measurement
a
f
t
he parameters
re~uire
d
by
th
e m
a
del
(aman
g
the
se, raat and c
a
na
py
develapment
,
l3ngth af the variaus
phase
s
of th
e plant
cycle, pl
ant
respanse
t
o water
s
t
r
e
ss
,
etc
.
),
even
tho
ugh these me
a
sur
ements
were
na
t
amang
t
hei
r
m
ain
abjectives.
T
hu
s
,
several
res
earchers
have
clearly
understoad
that
they can
R
E F E R E N
C
E
S
1.
ACKOFF,
R.L.
(19
7
9
a
).
"
T
h
e F
utu
re
o
f Opera
t
i
o
nal
Research
is Past".
J.of the Op
era
t
i
on
a
l
R
e
s.
S
oc
.,
30
:
93
-
104.
2.
(
1
979b).
"Resurrecting
the Fut
u
re
of Opera
-tional R
e
s
ea
r
c
h"
.
J. of the Opera
t
iona
l
R
es
.
Soc
.
,30:189-19
9
.
3.
ALVES, E.R.
de A. (197
5
).
"
O Enfoque
de S
i
s
t
emas
na EMBRAPA
".
EMBRAPA
,
Brasili
a
.
4.
BIRNBAUM,
P
.
H.(
1979
)
.
"Acade
m
ic
I
nt
erd
i
s
c
iplinary
R
esearch:
Probl
e
m
s and Practice".
R
&
D Managemen
t
,
1
0: 17-22
.
5.
DALTO
N, G. E.
(Ed.)
(1975)
.
S
t
udy
of Agr
i
cultural
Systems.
Applied Sc
ience
Pu
b
lishe
rs,
Lond
on
.
6.
DOOREN
B
O
S
,
J.
&
K
AS
SAM
,
A.H.
(1
979
).
Yie
l
d
Response
to Wa
t
e
r
.
FA
O
,
Ro
me
.
7.