Persistency
oflactation
yields: a
review
N.
Gengler
IJB.
&
btechnie,Frctltt
Untvcrstlatrc iles $icnccs Agrononiquca 8-5030 GenblorsAbrtnct
FcdEcocy
d
yiclds dudlg thc laaation iss!
importedldadm
q|'vc Ftatnc&r.It
can bc dcfncd as tDcabilig to tDaintair lhr lerrcl dproduaion during thc ldation- Rlal pcrsisaascf, for a constant le\Gl of productioq c8tr
be disiaguisbcd &,om
apannt
Frsins|gy.
P.rsi$cocy wasdcdr.d
i!
Eany difrct€ot xay5.Thtlc
Eajot malhcodlical d.f,nitionc arc bascd: a) on mathdtrstical lacarionorw
nodcls: b)o!
ratioc of panial totrl or othcr yiclds rhuing thc laaation aod c) on tbc variation of thosc yiclds. Dcfinition of pcrsisrcocy can bc occndcdo
ftt
andprucin yiclds. Heritsbility valu6 for pcrsist€nsy arc found beturcco 0.01
ad
> 0.30, wilh ooct rnlucs surnd 0.10. lvioc rralucs rcportcd for mcasrcs bascd oo variatioo and on longcr pcriods wcrc bigb6 rhen tbo6a for ratio ncasucs.Ilcritebility for
ht
ard €sp.ciatly Fotcin pcrsistcncics erc lowertia!
thosa rqortcd for milk pcrsilrency. Inportancaof pcrsistcDcy is double. It aficcts eccl|racy of yield evaluations bascd on hcompletr laclation rccords (c.&, tcslday nodcls)
ad
it
hasar
cconomic importancc on its own; asit
rcduc€s fccd and othcr managcm€nt cods. Genclicaaluations for persistcncy can bc concicved as scpcfatc aahutioq combined (multiple-{rait) with or deduccd Arom
yicld araluation.
l.
Introduction
The main income
for
most dairy farmers is based onmillg
fat and protein yieldsof
theircows. Especially since the
introduaion
ofmilk
quotas
in
the
European
Unioq
maximumimprovement
of
yields
is
not
necessarily economically optimum. As a matterof
fact,if
profits
are afunstion
of
returns minus costs,reduction
of
costs must
be
consideredto
improve proEts when increases
in
retums are limited.A
wayto
reduce costs isto
distributethe
sametotal yield
more equally
over
thewhole lactation. The distribution
of
lactationyield
is known
under the nameof
persistencyof
lactation
yields,
often
simply
called persistency. The precise definition is often notwell
worked
out, but in
generalwe
say thatthe lactation
ofa
cow is more persistentif,
for
the
sametotal
yield, the
animal peaks lowerand
the
lactation curve
is
flatter.
Better persistenryis
considered advantageousfor
scertain number
of
reasons.The
two
mostimportant are the better use of cheap roughage (e.g., S6lkner and
Fuchs, 1987) and the reduction
of
stress dueto
high peak production (e.g.,
Zimmermann and Sommer, 1975).The concept
of
persistencyis
also often related to the mathematical descriptionofthe
laaation curve. Several different measures are classically used
to
describe persistency. Wewill
describethe
most important
persistency mea$rresand
we
will
deal
with the
genetic aspectsof
persistency andits
introduction
incurrent and
future
selection schemesfor
vieldtraits.
2.
Delinition
of persistency2.
L
Persistency of milkyields
Different
approachesexist
to
define persistency and there is a certain ambiguity.In
general persistency
can
be
defined
as
theability
to
maintaina
more
or
less
constantyield during the lactation.
If
we
consider thatthe
shapeof
the
curveis diferent
&om
onediffercncc
otists
even8t a
constant level
of
produclion (Grossman et aL,
t986).
Thereforewe
can define pcrsistcncy as afunction of
the flatncssof
the
lacation
qrrve.
For
an animal this means thatit
is more persistent as an othcrifthe
lacrationorve
has aflatter
shape.In
theliterahrc
this
conccptis
calld
persistencyof
the lactatioq
or
p€rsist€ocyof
the
lactationcurve
or
persistency
of
milk
yields
or
persistency.It
is
clcar that
the
shapeof
thc lactation curve depords also on thctotal
yiel4
represcntedby the
sr&cc
below
ThereforeGengler
(1995)
disinguished
bawccn
ryparcnt
or
obscnred pcrsistency
and
realpersistency.
Thc 6rst
is
defined
without
cxonsidering total yields constant
2.2. Persistenqt
offat
and proteinyields
It
is
obviousrhat
thc
same conceptsrs
for
milk
yields can be generalizedfor fat
andprotei4 wen
if
such studies have been ratherseldom
in
the
literature (e.g., Kandzi
andGlodeh
1990;
Bouloc
andBoichard,
l99l;
Swalve, 1994).
2.
Lrctetion currcs
The milk
production
of
a
cow
can besubdMded
into
three parts.
An
ascending phase between calving and peakyield, a
nearconstant production around
peak
and
a descendingpart after
peak
(e.g.,
Gengler,1990). Lactations
failing
to
show
the
first
phase,
or
showing steadily
increasingproduction
are
called atypical
lactations. Shankset
d.
(1981)
summarized proportions reportedin diferent
papers.The literature
isnot
very
consistent conceming
the
relativeimportance
of
atypical curves as proportionsfrom
15%for
Fenis et
d.
(1985)to
45Vofor
Schneeberger
(1978)
are
given.
Thesediferences
are dueto
different
definitionsof
atypical,
differencesin
populations
studies,management
of
animals
and
recording procedures (e.g., Belgium).But
the
existenceof
atypical lactations is a fundamental problemmathematical
models used
must deal
with.Iactation
curves are atso important in the fieldof incomplete lactation octension (e.g., and Van
Vlech
1973; Schaeffer and1976) and
of test{ay
models (e.g.,Ptak
Schaefer, 1993).
3.
Mrthcmeticd
lectetion
cunc
modelc3.I.Inffirctiot
Difcrcnt
approach*
have been usedto
6nd a
function
to
ft
obs€rveddaily
yidds.Difercnt
objectives
of
thesc
functions
alsoorist. Thc
threc most important
are:
l)
the adjustmentfor
individualcows
to
describc agivan curve and
wentually
the persistency, 2)the adjustment
of
a curvefor
groupsof
cows,which
is often
donefor
management reasons and3)
the estimationof
305 day yields usingthe adjusted
ctwe,
thetotd
yield
being only the surface below the curve.Several
difierent models exists
(e.g.,Wood
1967;
Grossmanand Koops,
1988;Cobby and
Le
Du,
1978) andwe
will
try to
group themto
make the comprehensionofthe
models easier.3. 2. Eryonential
functions
The use
of
exponentialfunctions
goes back to Gaines (1927).But
Wood (e.g., 1967) popularized this approach.It
was then used byseveral
other
authors
(e.9., Kellogg
a
al.,1977).
tfis
basic
function
is
often
calledWood's
incompletegamma
function.
If
we define the daily yield at dayt
by y1, the modelts:
lt
=
atbe'd
(l)
where
4
b
and
c
are
parameters,a
being linkedto
peak yield, bto
the increasing phaseslope
and
c to
the
decreasing
phase.Parameters can be estimated using the natural logarithmic transformation (e.g., Shanks et al.,
l98l;
Grossmanet
al.,
1986;
Batra
et
al.,1987;
Congleton
and Everett, 1980a
and1980b):
This
trandormation yields
a
multipleregression
of
In(y1)
on
ln(l)
and
t
with
paramAcrs ln(a), b and c that can be estimdedby
ordinary least
squares.This
firnction
hasthe
weaknessthat
scasonaleffects have
astrong influence
on
ys
and thereforc
thefollowing
modiEcation
was
proposed
(e.g., Grossman etal.,
1986; Batrs et al., 1987):yt
=
ubed(l+usin(r)+vcodx))
(3)
where
u
and
v
are
additional
par.meters associatedwith
seasonalvariation
other thanseason
of
calving andr
is the dayof
the year expressed as radians.Wth
certain assumptionswe can
also
usc the
logarithmic transformation:(yJ
=ld")
+b
ldl)
-
cr+udn(r)+vo(r)
(a)
The problem
with
this transformation is that the number ofparameters is augrnented sotherefore the
fit
should be better.but
we lose residual degreesof
freedorn, orfrom
the morepractical
point of
vieq
we
needto
know
atleas
6 test-days.S chneeberger
(1981)
proposed anothermodification
of
Wood's
gammafunction
to
take
the time
of
initial
production
(re)
intoaccount:
y,
={1-10)b"{r'to)
(s)
Schaeffer
et
d.
(19?7)
used
a
morecomplex
exponentialfunctio4
calleda
one-compartment open model.3.3.
Multiphesic
mdels
As the lactation can be easily subdivided
in
different phases, the idea was developedto
use multiphasic models.An
exampleis
given by Grossman and Koops (1988):r,
=i{",u,[r-t""n'z(u1(r-
c1))]]
(6)where y1 is the production at day
q
a1,b;
andci
are parameters associatedwith
phase i; tanhis the
hlpcrbolic
tangent and n the numberof
phases.This
method
works
well
if
2
or
3 phases are used (Grossman andKoops,
1988),but 6 or 9 parametcrs must be estimated. This
fast limits
thc
use
of
this
approach
to
situations
wherc enough
obscnations
are known.This
methodhas
a
certain
numberof
advantlge
but
has
bccn
scldom used.
Thcmain
reason
is
that
despite
its
theoraical
rd*t"g.C
it
csn
not bc
linearizd
so
thatparameter cstim.tion is
dificult.
3.4.
Polynonial
od
iwvrse
polyonial
ndels
Ali
and Schaetrer (1987) used an inverse quadratic polynomial modelfirst
describd
by Nelder (1966) and used by Yadav et al.(1977)
for
dairy cattle. This modet is written as:ylr
=Fr+
p6t-t
+p2t
(7) where the parameterp
represents peak yield,p9
is associatedwith
increasing andp2
with
decreasing phase
and
ylt
is the
inverscof
yield
at
day
t.
The
model
is
written as
amultiple regression
ofy[l
onI
andt-1. It
hasthe advantage that the number of parameters is
limited
to
3.3. 5. Multip le rc gre ssion
Another interesting way
to
describe thelactation curve
is
the
multiple
regression approachfirst
describedby
Ali
and Schaeffer (1987). The regression model has the form:yt
=
po+pr,'t
+p2yl
*yrr+parl
(8)where: 71
=t|3oS
and
,=tn13g57r1.
The
parametersare
again
linked
to
the lactationcuwe:
pq with thc peak yield, p1 andp2
\,ith
decreasingand
p3
and
pa
with
increasing phase.
This
model permitsa
good descriptionof
the
cune
accordingto
results presentedby
Ali
and Schaeffer (1987),but
it
F
obsenrations
to
cstimatc
thesc
parameten.This model was uscd
in thc
modelingof
test-day
CID) genaic
waluation
as corrcctionfor
days inmilk
attcstday
by Ptak and Schacffer(1993).
In
this contod
corrections are donefor
all
TD
records
in
lactations
of
cowsgrouped
in
givcn
agc.seasonclasscs
andthercfore
the
mrmber
of
parametersis
not important.A
new
dwelopment
is
the
randomregression
model
(c.g.,
Jamrozik
andSchacfrer, 1995) whcre
the
lactationqrrve
isdescribd
phenotlpicaly
and geneticaly usingAli
and Schaeffer(1987) multiple
regression.Fixcd regrcssion coef6cients are extimated
for
give,n region-age'season classesand
random(genaic)
regression coefficientsfor
an animalusing
the
gcnetic
covariances
amongcoefficients
and
the
numerator
relationshipmatrix among rnimals.
3.6. Other
mdels
Several
other
models canbe
used andhave been
described
in
the
literature.
An
example
is
the
vibration model reported by
Hayashi
et
at
(1986). Cobby and
Le
Du
(1978)
described
a
certain
number
of
alternative
models,
essentially
modifiedexponential
function bascd
on
Wood's tunction (e.g., 1967).4.
Pcnistency mcrsurcs
1,1.
Measres
fuyd
on
Iacwion
atrvv mdels
Thc most oornmon mathernatical model is based on
lVood
(1967). This model has thcpartiailuity
thatit
is rather easyto
dwelop
asimple formula
of
parameteis associatedwith
penistency. This simple formula is:
P =
c-G+D
Another way persistency can be deduced
out of such an mathematical model is
g"en
by Grossman and Koops (1988). They associatedwith their
multiphasicmodel,
persistencyto
the duration of their second phase:P=2bi
(10)For
multiple
regression
modelspersistenry
can
be
linked
to
slope
of
the decrcasing phasc of lactation.Random
regression
models
(e.g.,Jamrozik
and
Schaeffer,1995) can
provide breeding valuesfor
individual
day yields andfor
partial,
total
or
other ladarion ields.
Persistency measures can than be developpedout ofthese values.
Oher
formulas
exist
to
describepersistency
out
of
lactation curve
models (Rowlandset
al.,
1982),but their
importancefor
the
description
of
persistencyis
rather limited. Another problem is that one should beable
to
describe
all
types
of
lactations, including atypical ones.4.2. Measures based
on
ralios
between total,partial, mcimum,
or
otheryields
The
ideato
useratios
as
measuresof
persistenry
is
old,
it
goes
at
least back to
(e)
Sandcrs (1930).
Hc
dcdncd pcrsistcncyrs
theratio
benvecn mean and pcakyicld.
This ideawrs
uscdby
S6lkner and Fuchs (1987). Theydefned
two
pcrsistency measunes,Prourxz
utd
PTgttA)G,
trs:Ploulxz
=
Max.
yield6.pr.ro
r
100(l
l)
(12)
Mean yield 3r,
r.ru
Higher
values
for
Ff€ntAJA
andrycM$o,
arre
associued
with
lowerpersistency. This not very logical fact must bc
remembered
when
comparing thernto
othermahods.
Keown
et al.
(1986) used
anmodification of this mahod:
r=
MT;{tjldo-''''.r*
(13)Y I el O.r d.y 3OO
This measure is also linked negatively
to
persistenry
and
sharestherefore
the
sameintuitive
weakness
as
PfUt4ArO
andPrs,tcx3
Others suoh as fohansson and Hansson
(1940)
introduced
ratios
berween
partial yields. These methods were popular and manyvariants
of
their
two
measures, P2.g and P3. ,exist:
Dandl (1982) also modifed P2,,
but
with
thc
intention
of
crcating
a
meanrrerssociucd
with
asc,cndingand
descending plrasasof
laciation called
Pl
and P2.
She defined:D--
-.---
-
Max'
Yiclda:nr-rs
,r*
rrordAx3 =
ffit;e--d.,'tln,
-
Partialyieldar.16l_26'2r=lE;;@;
^
Partialyield*,rouro
",r=
lfrti'yidl;_-,,=Pffig',*
,,,
= Partialytcld,r-
rzr-zro.
roo
-'
Psrtialyielddry,3l.l,
(17)
(18)
All
thcr*io
mahods
basedon
P21 andP3
sssociste
higher values
to
bctterpersistency.
Other
ratio
methods exist,but
all
sharethe
samc basicdefinitions
of
ratios
betweenccrtain partial,
peah
daily,
totd
or
otheryields.
1.3.
Measres
based onvwiation of
yields
The
definition
of
persistency as flatnessof
the
ctrve
can
be
easily
described
bymeasures of dispersion. This rather simple idea
seems
to
have been seldom usedin
the
past.Only
sinceis
useby
S6lkner
and Fuchs in1987,
it
has been
reconsideredby
other scientists (Gengler, 1990; Swalve, 1994).Sdlkner and Fuchs
(1987)
definedtwo
persistency mea$lres based
on the
standard deviations of tes-day yields:Some
scientist
(e.g., Ericson
et
at.,1988) used the same ratios but expressed them
as perc€ntages.
Others modified
therq
as the persistency measure defined by Mahadwan ( I 95 I ) :-
Pfitial idddD. r.re-Partial yiddd$ r.Dt=@rro,
Psoz
=
(le)
where
TD;
is the ith test-dayfeld,
pp
is themean test-day yield and
p
is the numberofthe
last test-day before 200 days in milk.Pso:
=
(20)where
TD;
is the ith test-dayyield,
pp
is themean test-day yield and p is the number
ofthe
last
test-daybefore
305
days
in
milk.
This measure has a weak point, the lactation needsto
be finished,or
at least considered finished.Both
methods are affectedby
the fact that
a (14)(15)
1i(-,-r*)'
gircn
tcstday
do€s not neccssary take place atthe samc stage of tactation.
An
Lnportantrdrnntagc
of
thesemeasres
isthat their
distribution
can
be
cmpiricallyconsidered
ncarcr
to
a
normal one tlran
foi
ratio measurcs (S6lkner and Fuchs,l9g7). But
at
the
same
time
all
pcrsistency
measuresbased
on
v?riation
mahods
sharc
thedisadvutsge that their
vatues become nearerto zero
with
higher persisency._
But
according
to
Gengler
(1990)
andSwalvc
(1994), the
mcasuresbased
on
thcvalation
of
tcst{ays
arc in0uenced by erraticvariations
that
rrc
obscrvedfrom
test-davto
testday.
Gengleret al.,
1995 used thereforc the persistency measure calledpyv
measured as yield variation and defined as:*=1{,S.(H'.(g-,*'f
(2t)
where
M
is the partialmilk yield from
dayI
to
day
100,M2
is the partialmilk
yield fromday
l0l
to
day
200,
M3
is
the partial milk
yield from day 201
to
305 andM1
is the totalmilk yield from day
I
to
day
305.
Gengler(1995)
modified slighrly
the
definitionreplacing
the 0.5
by
0.25.
He
used
a suggestion made earlier (Gengler, 1990) and expressed persistency on a relative scale intra-lactation.5.
Influencc of
milk
yield
on pcrsistencyThe
problem
of
links
betweenpersistency
of
lactation and
milk
yields
hasinterested scientists
for
many years
(e.g., Gaines, 1927; Mahadevaql95l).
There are at leasttwo
reasonsfor
this
link. Total
vield
isthe
areabelow the cuwe,
and
the
leld
atevery moment
of
the lactation is a functionof
the
qlrve.
On the other handit
is clear that ananimal
with
very high production at peak havelikely
a steeper slope than another that islow
producing (Genglea 1990). Therefore we can
assume
that the milk yield
hasan
important inlluence on persistency.Many
scientists
have
andlzcd
rclationships berween persistency, measured through
diferent
param€tcrs, andinitial,
partia[
total,
ac.
milk
yields.
The
rcslltiii
found
does
not
show
a
consistcnt
pattern.Early
studiesfound
a
negative
relationship bctween persistency andtotal yield
(Gaheq
1927).
l-atu
lvlahadevan
(1951) found
apositivc relationship. This result is
verified
bySchnecberger (1981).
Now
it
is
clear that therdationship
betcreen
total
lcld
andpcrsistenq
depecrds
asscntially
on
the pcrsisrencymea!
re used(solkner
andFuc\
1987). Persistency measures basedon
ratiosshowed
a
positive
rclationstrip,
but
themethods based
on
variation
a
negative
.1relationship. Gengler
(1990)
showedthat
this later relationship is closeto
a linear one.If
we
,
takethe definition
givenby
Grossmanet
al.
:
(1986),
persistencyshould
be independentty I
measured,
or
oorrectedafter,
for milk
yields. Therefore agood
persistency measure shouldbe independent
from
yields,or
correctedfor
the influenceofyields.
As
a
conclusion
we
can
state
thatpersistency
is
dependenton
yields,
especiallytotal
yields,
but
the
direction
of
therelationship depends
on the
measures used.The
ratio
meaiures
show
a
positive
one,whereas
the
variation
measures
show
anegative relationship.
The
reason
for
thiscould be
that the
first
arehigNy
atrected bythe level
of
production
and the
second areinlluenced by variation in production,
with
thisvariation
more
importantfor
high
producing cows.The
influence
of
total
yields
has two
components,
a
genetic
and
a
non
genetic, thereforeif
phenotypic relationships are forcedto
zero
through
the
definition
of
real
t
persistency, genetic relationships exists. Table
l.
gives results obtainedby
Gengler
(1995)
:using
a
persistency measure
based
on
a
i
functionofthe
variationofyields
6.
Environmental
inlluences onpcnistenqr
6.1. Parity and age at
caling
t
I
,:i
The
effect
of
rgc
8nd
parity
onp€rsittency
hrs
been
*udied
by
severalscicntists.
Certain
scientists separste
thesceffects
or
worked
with
separate lactations(e.g.,
Sdlkner and Fuclrst 1987;
Gengler, 1990),while
others considered them together(Suders,
1930).All
authors ageedtlut
thefirst
lactationis more persistent than thc others (e.g., Shanls ct al.,
l98l;
Dancll,
1982; Keown et al., 1985;S6lkncr and Fuchsi 1987). The most conunon
ocplanation
is
based
on
I
lowcr
level
of
devclopment
of
the
mammary glandfor
first
puity
cows (e.g., S6lkner and Fuchs, 1987).The
results concerningthe
influenceof
age within parity
are
not so
similar.Mahadwan (1951)
for
orample did not find an influence,Danell (1982)
and Grossmanet
al.(1986) found
a
very
mdl
influencethat
ismore important
for
younger cows. Grossmanet al. (1986) did
not find
it
sigrificant.
Others such as Smith and Legates (1962) described a reductionin
persistenc,with
ageduring
thefirst lactation and an augmentation during later
ones. Gengler
(1990) found
results
that supportedthe
later
hypothesis.All
authorsagre€
that
it
would be
important
to
studyeffects
on
persistency
of
parity and
ageseparately.
6.2. Season oJ
calving
Most
research
that
has
been
doneconfrms the influence
of
seasonof
calving on persistency (e.g., Mahadevar\l95l;
Cady andMc Dowell,
1980; Danell, 1982;Fenis
et al.,1985;
Grossmanet
at.,
1986; S6lkner
andFuchs, 1987;
Gengler,
1990).
OnlySchneeberger (1981) did not
confrm
this.But
the different
studiesdid not find
the
sameinfluence. Gengler
(1990)
explained
that obviously such differences are dueto
climaticand
management
differences
among
thepopulation studied. Some papers
described studies done inNorth
America (e.g.,Fenis
etal., 1985;
Keown
etd.,
1986) who found thatthe
most
persistent lactations beginsat
endsunrmer.
But
European studies as Mahadevan (1951), Danell (1982) and S6lkner and Fuchs(1987)
found
that
the
most
persistent lactations beginin
fall-winter. Gengler (1990)crplained
that
these resultsare also afhcted
by
the length
of
the
measuredpoiod.
As
amatter of fact the way persiitency is measured also atrecrs the time of the apparent modmum
persistcncy
(Solkner
and
Fuchs,
1987). Longer intervals tendto
shift this mo<imumto
eadier dates in the year.6,3, InJluence
of
gesaion
The new c8lf
thst
a
cow
is
carrying during the lastpart
of its
lactation is
ableto
influencc
the
potential
milk
production
of
acow. Thc influence on daily yields during this
period
is
direct and
creates
an
indirect influenceon
persistency.Sweral
ways
havebeen
described
to
measure
this
influence. Some scientists used the days open approach(e.g.,
Schneeberger,l98l;
Danell,
1982;Grossman
et al.,
1986; Sdlkner and
Fuchs,1987), others used days
carried
calf
(e.g.,Keown
et
d.,
1986). Gengler
(1990)
took
calving
interval
as
the
parameter.
Animalswere
often
grouped
in
classes
(e.g., Schneeberger, 198l;
Danell,
1982;Kmwn
etal.,
1986; Gengler, 1990). Sdlkner and Fuchs(1987) used a linear and quadratic regression on days open.
The
results found
were
not
similar. Some people found no influence (e.g., Danell, 1982; Grossmanet al.,
1986), others such asSdlkner
and
Fuchs
(1987),
Schneeberger (1981), and Smith and Legates (1962) showedthat gestation influenced persistency. S6lkner
and
Fuchs (1987)
observed
a
linearrelationship.
They also
showed
that
the persistency measurescovering
a
long
period aremore
afec-ted thanthose covering
shortperiods. Gengler (1990)
found a
relationshipof
gestationwith
persistencyonly in the first
lactation and
it
had a quadratic component.7.
Seritrbility
of persistencyThe
literature
on
persistency measures showsthat the
different
measures have alsodiflerent levels
of
heritability.This
additional sourceof
variationfor the
heritability
makesdifficult.
Genglcr (1990) gave a reviewof
theUtcrature.
Results
werc
bctween
undcr
0.01(Shanls et
al.,
l98l)
and over 0.30 (Smith andI*gatesi
^1962).
Another
importani point
isthat
holding
thc milk yicld
constant reducedthe.heritability @anell, l9B2).
According
to
Sdlkner and Fuchs
(1987),
heritabilitiesiere
rather constsnt
fiom
lactationto
lactatio4
butrathcr diferent
betwecn
thc
pcrsistency measuresthcy
snrdied. Ttrcy found that longer mcalnrresincluding
thc
whole
lactation
andrrariation
mahods
had the
highestheritabilitieg
p5p3
hada
heritability of
-O.Zt,0.22 and 0.22
in
the three first i""t"tions.
Gcnglcr (1995) found that vatues
for
apparentTg
r""l
persistency arc closeif
persijcncy
isdefined
as
variuion
of
partiat
yieias. Heritab.rlityof
reat persistencyfor milk
yields*"t yt
0.14
higher than
fat
(O.O6j andprotein (0.04)
(table
I
and Tabte2).
Similarlfuls
were reported
by
Swalve(199a)
andKandzi and Glodek (1990).
t.
Rcpcetability
of
pcrsistencyAnother important question
is
what
isthe repeatability
of
persistencyfrom
lactationto
lactation.
Most
researchhas been
donel11iol
by
tactation (S6lkner and
Fuchs,1987; Gengler, 1990),
but
it
is
interestingto
assess how similar aretwo
persistencyrec;rds
are
for
the same animal. Resuttsfrom
Gengler(1995)
showthat
real pcnistenciesfor milk
fat
and protein yields
had
slightly
higher repeatabilirics(Iable
2)
comparedto
apparent persistencies Cfablel).
Thc ratherlow
iesultsfor
repeatability
indicate
that
at
least persistencyfor first
and later lactations can beconsidered
diferent
traits.9.
Corrclrtions
bctwccn
mi[q fat
andprotein
persistcncicsVery
few
results are known showing thecorrelations between
mil(
fat
and
proteinpersistencies.
Table
I
andTable
2
show the values obtained by Gengler (1995). They showthat
genetic correlations
are
higher
thanphenotypic correlations. Correlations are
higher betwecn
protein
and
persistency between milk and fat or protein.10. Economic
importrncc
ofpcrsistcncy
_-
Economic
importanc€of
persistencylinked
to
the
reductionof
costs obtainedbater
pcnistcncy.
Two tlpes
of
odsts. Frst bctter
persistency reduces costs 8s shovmby
Sdlkner and Fuchs(l
and Genglcr (1995)
who
linked persistencyto
the replaccmcntof
conc€Nrtratcsby
rough4gc.Gengla
(1995) found that this reduction givespersistency
a
relativc weight
compared
to
leld
of
around
3
/o.
B*ter
persistency reducesalso health
and
reproductive
costs.Here he found
around
7
Yoof
the
relative economic value of yield.ll.
ConclusionIn
the
literature
persistencyis
mosttydefined as lactation
or milk
yield
persistency, and persistenciesof fat
andprotein yield
areseldom
considered.
Different
persistency measureswere
describedin
the
literature. Three greattpes
of
measures can be defined:l)
me^cures basedon ratios
between total,partial
or
otherfelds; 2)
measures based onvariation
of
testday
lelds
and
3)
measures based on mathematical modets. Environmentalinfuences on persistency are
well
described inthe literaturg
evenif
there are
diferencesbetween persistenry measures and population studied.
The
most important effects are milkyields and seasonal effects. The heritabilities
of
persistency measures reported in the literature
range from under 0.05
to
over 0.30.Acknowledgemcnt
The author is
Chargi
de Rccherchesof
the
FondsNational
Belge de
la
RechercheRcfcrcnces
Ali,
T.E. atrd Schacfrcr,L.R,
19t7. Accounting forcovarianccs among tes day
nilt
yicld in dairy cows. Can J. Anim. Sci.. 67: 637{44.Bara,
T.R,
Li4
C.Y., McAlliscr,A.t.,
l€,
4.J.,Roy, G.L., Vcsely,
J.A,
Wauthl
J.M. andWintcs,
K.A.,
1987. Multitrait estimation ofgenetic paramctcrs
of
lactation curvcs in Holncis hcifcr& t. DaLy Sci., 70: 2105-21Il.
Bouloc, N. aod Boichar( D., 1991. Classification of
lactation
crnrls
itr
g@t.J.
Dairy
Sci., 74(Suppt
l):
230(Ahra
r).C5dy,
RA"
ad
Mc DwcU,RE., l9to,
Persist ncyand othcr lactation paramclers
in
cms$rcd dairy cadc. J. Daily Sci., 63 (Suppl.l):
lll
(Abdraci).CobW, J.lvL and
L D\
Y.L.P., 197t. On fittingc1rwcs to laclation data Anim. Prcd., 26: 127-133.
Conglclon,
w.R
and
Ewr€G
RW.,
l9E0a. Ap'plication of thc incompletc gamma funaionto prEdict omulativc
nilk
production. J, DaitySci., 63: 109-119.
ConSJcron, W.R
ttd
Evtrltl, RW., 1980b. Enor and biash
using tbc immplets gamma fi$ction to dcscribc lactation crl'\rcs. J. Dairy Sci., 63:l0l'
l0t.
Dancll.
8.,
1982.SMics
on
laclationidd
ad
indivi&El
tcs{ay
yisldsof
SqEdish dairycowa
ltr.
Pcrsistcocyof mi&
yield and itsconclation
with
lacadonyield
Acta Agtic.Scand., 32: 93-101.
Ericson, trC,
Demll,
B.
and
Rrnd.l,
t
,
1988.Croe$rccding rficsts bctrEeD
two
S$&disb dairy brceds for produaion raits. LivrsL Pmd-Sci., 20: 175-192.Fcnis
T.A,
fy{aoI.L.
and Aadcron,C.R,
f 9E5. Scloctiog for laclalion ctwcad
hilk
yicld Lldairy canlc. ,. Dairy Sci., 68: 143&l'14t.
Cnincs, W.L., f927. Pcrsisrensy of lactation in dairy cows. Univ.
Ilioois,
Agdc. Exp. Sto- Bu[.,228i 355424.
Gcnglcr,
N.,
1990.Endc dc
la
persislamc desproduaions laiti&rs Pendalt la laclation cbcz
la
vachc Pic-Noirc. Travail defin
d'Chtdc'Facult4
Unill
sitsirs dcs
ScienccsAgronomique
&
Gemblou:t, Gemblouv',Bcl6un\ 9Epp.
Gcnglcr, N., 1995. Usc of mixcd modcls ro apprcciatc thc pcnisrency of yiclds during rhc laclation of
milk cows. Ph.D. Thesis. Factltd Univcrsitairc dcs Scicnccs Agronomiqucs
dc
Gemblou+Gemblourg Bclgiulo, 23 lPP.
Gcoglcr, N. , Kco*n, J.F. atrd Vatr Vleclq L.D., 1995.
Various persistcncy mcasures and relationsbipe . with total, partial and pcak milk yields. Brazil.
J. Genct., lE: 237-243.
Grossmar\
M.
and Koops, N.J.,l9tt.
Mttltipbasicamlysis of lactation crwes
in
dairy catllc. J.Dairy Sci.,
7l:
l59t-1608.Grossnaq M., Kuclq A. L. and Norto4 ILW., f 9E6.
lactation curvcs
of
purcbrcd and crossbred dairy catlle. J. Dairy Sci-, 69:195-203-Ha)'ashi, T., Nagaminc, Y. and Nishida A., 19E6. A
vibration modcl to ddcribe lhe lactation curl€
of a dairy cow. Jpn.
I.
Zootcch. Sci., 57: 471-47E.Jamrczilq
J.
and Schaefier,L.R,
1995. Genctic parameters forflst
lactatation milk,frt
and proteinields
of
Canadian Holsleinswith
arandom regression model.
J.
Dairy Sci., 78 (Suppl. l):249.Joharssoq 1., and Hanssoq
A.,
1940. Causcs of variation in milk ard butterfat yieldin
datycows. Kungl.
ladrbr.
Akab. Tidskt., 79:l-rz7.
Kandzi, A. and Glodelq P., 1990. Die Eftzienz der Selektion aufgrund der Einsatzleis$ngcn von Ersrkatbskohen.
II.
Mineilun8:
die Bezicbungen nvischen Einsatdeistungen uDd dcr Persistenz der l:ktation. Ziichtungskunde, 62: I lt-128.Kc[ogg, D.W., Urquhart N.S. rod
frcge"
A"J.,t9n.
Fgirn'ring
ltrolScin laaation cuwrswith
agrnna
curve, J. Dairy Sci., 60: 1308-1315.Kcs/tr, LF. and
Vat
Vlcdq L.D., 1973. Extcrdinglaaadon
rords
il
progrcssro
305day cqui\latcotJ.Ihiry
Sci., 56: 1070-1079.UaUacrran, P,, 1951, Thc cficct of cm,imnmcnt and
hcrdity
o!
laaadon"tr.
Pcrsisrcncyof
lecrarion J.
furie
Sci., 4t: t9-93.Ncldlr,
J.A.,
1966. Iwcssc polynomials,a
uscftlgtoup
d
multi-bcror
rcsponsc firnaions. BioDdric{Zl: 126,l|l.
Rab
E
aodsch*fEr, L.R,
193.
Usc of tcst day yiclds forgoic
aalrntion of dairy sircs andcmts. Livec
Pnd
&i.,
34: 23-34.noc'lards, G.J., Luc.y, S. rnd
&$s.ll,
A.ld., f 9t2. Aconparisoo of difr.rcot Eodels of thc lactation
cnwr in dairy ca0lc. Aaim Prod, 35: 135-t,14.
Satrdcc,
ILG.,
1930. Thc aaalysisof
thc lactationq|'vc
into rn ximum yield and Frsigcncy. ,.Agric. Sci., 20: 145-1E5.
Scbacfrcr,
L.R
ad
Burnsidc, E.B., 1976. Estimaring the sbapc of thc ladation quvc. Can. J. Anim.Sci., 56: 157-170.
Schacficr,
L.R,
Mindcr,CJ.,
Mclvlilla4
I.
andBunsidc, 8.B., 1977. Nonlincar tcchniqucs for
prcdiaing 305day
lacration produaionof
Holstcils and Jcrsefs. J. Dairy Sci., 60: 1636-1641.
Scbnccbcrgcr,
lvt,
197t.Ttc
sbaFof
thc lactation c-urvc and thc cgimation of milt yicld in BrownSwiss cattlc.
Ph.
D.
Thcsis,Eidg.
Tcch. Ilochscbulc, Zlricb, S\yitr€rlaod.Schnccbcrgcr,
lvt,
1981. Inhcritancc of lactatioDin Swiss Brc$d cattle. J. DaLy Sci., 64: 483.
Shanfs,
RD.,
Bcrgcr, P.J., Frccmaq AE.Dic&inson
F.N.,
l9tl.
Getrctic aspc.is lactatiolorws.
J. Dairy Sci., 64: 1852-1860.Snir\
J.W. aDd lrga&sr J.E., 1962. Facrorspcrsigcocy eDd
ia
importarccin
305day laaadon produaion" J. Dairy Sci., 45: 676{77(Abdra.r).
i
Soll(ncr, J. and Frcbs,
W.,
l9E?.A
compadsoo dificcnt mcaoucs of pcniscncy withspcid
rlsp.c
!o
uriaion of
ldday
nilk
yidds.'Li\cst Prod Sci.. 16: 305-319.
Swalvc,
ILII,
194.
Egimationof
brccdinglaluc
n.ingtrdry
nodcls" Prcscotcdat
th.
Opco Scssion of thclDtcttul
Annual Mcctiag5{
Augusr, 1994,
ftaua,
Ontario, Canada Woo4 P.D.P., 1967. Algcbraic modcl of thc lacationorrvc in cattlc. Naore,
2l6i
164-165,Yadav, M.C., Katpatal, B.C. and Kausbih S.N., 1977.
Componcnts
of
gamma-typciltrction
of
a lactation c1rnrcatd
faclors aficcting thcn itrIlariam
andits
F iGiatr cro6s€s. Indiatr J.Anin. Sci.. 47: 520 - 525.
Zimmcrnunrl
E.
and
Sommcr,Il,
1973. Zuml:ktationnrcrlauf
von K0hen
inHochlcistungshcrdcn und dcsscn B€cir lussrsg durch nicbtcrtlichc Faktorcn. Znchtungslondc 45: 75{8.
Tablc
I
Gcnaic corrclalions (aborr), hcritabilitics and rEFatabilities (on rcpcatabilitics b€twcen brcakels) and phenotypic correlations (bclow thc diagonal) among apparcnt milh fat and pmrein Frsincncies
Traits
Milk
Dcrsistcncr
Fatocrsistencr
Protcin DcrsistcncvMilk pcrsistency 0.t4
(0.24)
0.E3Fatpc6istcncy
0.52 Protcinpersistency
0-52 0.06 (0.14) 0.70 0.E9 0.8E 0.05 (0. r0) Tablc 2Gcnctic correlations (above), beritabilitics and npcatabilitiG (oq repealabilitics betrvccn breakcts) and phenogpic corrclatiors (below lhc diagonal) among rEal milh faa and protcin pcrsistcncics
Mi Milk p€rsistcncy Fat Frsistens, Protein persistcnsy 0. t4