Lactation
Curves and Prediction of Daily and Accumulated Milk Yields in
a
Multibreed Dairy Herd in Thailand Using All Daily Records[1]/
Skorn Koonawootrittriron*,
Mauricio A. Elzo2/, Sornthep Tumwasorn, and Wirot
Sintala
3/
Department of Animal Science, Faculty of Agriculture,
Kasetsart University,
Bangkok 10900, Thailand.
* Correspondence: Present address 3/ Sakon Nakhon Agricultural Research and
Training Center, P.O. Box 3,
Pungkone, Sakon Nakhon 47160, Thailand. E-mail: skornk@hotmail.com
[1]/ This research was supported by the Florida
Agricultural Experiment Station and a grant from
the
Thailand Research Fund under the Royal Golden Jubilee Project,
and
approved for publication as Journal Series No. R-08042.
2/ Department of Animal Sciences, University of
Florida, Gainesville, FL 32611-0910, USA
Abstract
Daily milk
records of 88 cows in a multibreed dairy herd were
used to describe lactation curves and test the ability of seven models to
predict daily milk yields, and accumulated 100-d and 305-d milk yields. The seven mathematical models were a gamma
function, a mixed log second-degree polynomial model, and five polynomial
regression models. Cows were of three
breed groups: Holstein Friesian (HF), 1/2HF-1/2RS (Red Sindhi), and
3/4HF-1/4RS. Lactation number was
classified as first, second, third, and fourth-and-later lactations. Calving age was defined as age less than 30
months, and equal to or greater than 30 months for the first lactation; age
less than 44 months, and equal to or greater than 44 months for the second
lactation; less than 60 months, and equal to or greater than 60 months for the
third lactation; and all ages for the fourth-and-later lactations. Seasons of calving were defined as winter
(November to February), summer (March to June), and rainy (July to
October). Four general types of
lactation curves were found: convex, slightly convex, two-peaked, and
flat. Types of lactation
curves varied across breed group x lactation x calving season and breed group x
lactation x calving age subclasses. The second-degree polynomial model was the
best in terms of predicted minus actual daily and 305-d milk yields, and
computational requirements. For 100-d
milk yield the best model was the sixth degree polynomial model. The application of these results is limited
to HF x RS multibreed herds in the Northeastern region of Thailand. To further help national genetic evaluation
efforts, this study needs to be repeated with a data set that is representative
of the national Thai dairy cattle population.
Key
words: dairy cattle,
lactation curve, milk yield, prediction, multibreed
Introduction
Knowledge of lactation curves and
yields in dairy cattle is important for decisions on herd management and
selection. Different mathematical
models have been evaluated for their ability to describe lactation patterns of
milk yield as well as the ability to predict cumulative milk yields from
partial records (Schaeffer et al., 1977, Batra, 1986, Vargas et al.,
2000). A popular model that has been
widely used to describe lactation curves and predicted lactation yields is the
gamma function (Wood, 1967). However, a linear regression model of yields on
days in lactation (linear and quadratic) and on log of 305 divided by days in
lactation (linear and quadratic) was reported to be better than Wood’s gamma
function for predicting lactation yields (Ali and Schaeffer, 1987). Recently, more complicated nonlinear models
(Grossman et al., 1986, 1999, Morant and Gnanasakthy, 1989) to describe
lactation curves have been proposed, but they require additional mathematical
techniques and may have more computational problems because of the larger
number of parameters that need to be estimated and the greater amount of data
is required to estimate those parameters with a reasonable degree of accuracy.
In Thailand,
although Holstein Friesian is the main breed used for crossbreeding in dairy
herds, most farmers simultaneously raise many types of crossbred dairy cattle
in their farms. Each one of these breed
groups of cattle can potentially have lactation curves of different shapes
depending on calving age and calving season.
Because interactions among these effects can be important, if these
animals are to participate in a genetic evaluation, then lactation curves
within breed group x lactation number x calving age subclasses and breed group
x lactation number x calving season subclasses need to be compared. In addition, it is also important to assess
the ability of various mathematical models to estimate daily milk yields using
all available daily records.
Thus, the objectives
of this study were: 1) to compute actual lactation curves based on means of
daily milk records, 2) to compare the ability of seven mathematical models
including a gamma function, a regression model and five polynomial regression
models to describe lactation curves of individual cows, and 3) to compare the
ability of these seven models to predict cumulative 100-d and 305-d milk yields
of individual cows, within breed group x lactation number x calving age and
breed group x lactation number x calving season subclasses, using all daily
milk yields in a multibreed dairy herd in Thailand.
Materials and
Methods
Animals,
Management, and Records
Analyses were
performed on data provided by the Sakon Nakhon Agricultural Research and
Training Center (SARTC) and collected in an experimental multibreed dairy herd
in northeast Thailand between November 1, 1997 and December 31, 1999. Animals consisted of purebred Holstein
Friesian (HF) and crossbred Holstein Friesian x Red Sindhi (HF-RS), which were
under the cattle breeding project of Rajamangala Institute of Technology (RIT).
A total of
28,452 daily milk lactation records (days 5 to 305) from 106 lactations of 88
straightbred HF and crossbred HF x RS cows were used here (Table 1). There were three breed groups of cows: HF,
1/2HF-1/2RS, and 3/4HF-1/4RS. Lactation
records were from first, second, third, and fourth-and-later lactations.
Calving seasons
were classified as winter (November to February), summer (March to June), and
rainy (July to October). Because of the
overlapping of calving ages and lactation numbers, calving age x lactation
number subclasses were defined. The
resulting seven calving age x lactation subclasses were: 1) calving age less
than 30 months x lactation 1, 2) calving age equal to
or
greater than 30 months x lactation 1, 3) calving age less than 44 months x
lactation 2, 4) calving age equal to or greater than
44 months x lactation 2, 5) calving age less than 60 months x lactation 3, 6)
calving age equal to or greater than
60 months for the third lactation, and 7) calving age greater than 60 months x
lactation 4 and greater.
Animals of all
breed compositions were raised under the same nutritional and management
conditions. All cows were milked twice
a day, once in the morning (0400 hrs.) and once in afternoon (1400 hrs.). Irrigated Napier (Penisetum pupureum)
and Guinea (Panicum maximum) grasses were fresh cut and carried to feed
all animals. However, the quantity and
quality of these green chops was variable and changed from season to season
depending on the weather in a particular year.
During the dry season, milking cows were maintained on Ruzi (Brachiaria
ruziziensis) pastures after the morning milking. Mineral blocks were provided as supplement ad libitum. Animals were fed 8 kg of a concentrate made
at the research station using local grains, crop-residues, modified-rice straw,
molasses, vitamins, and minerals.
All animals in
the herd were treated against internal (IVOMECâ,
ABENDAZOLEâ) and external
parasites (BARICATEâ) every six
months. Breeding age and younger
heifers were vaccinated against viral (e.g. Foot and Mouth Disease,
Hemorrhagic Disease) and bacterial (e.g. Hemophilus sp, Leptospirosis sp)
diseases annually beginning at six months of age. The parasite control and vaccination program used in the HF-RS
multibreed herd followed the guidelines given by the Department of Livestock
Development of Thailand (SARTC, 1999).
Cows were mated
throughout the year to maintain a minimum level of total amount of milk
produced per day. Qualified personnel
observed signs of estrous behavior daily. Animals were inseminated up to three
times. Cows were palpated by a
veterinarian 60 d after the last insemination.
Open cows were placed with a Holstein or a Red Sindhi clean up bull
according to the mating plan. There
were 2 Holstein Friesian and 2 Red Sindhi clean up bulls available every
year. Pregnant cows were dried off two
months before calving.
Table
1.
General description of the data set
|
|
HF |
1/2HF-1/2RS |
3/4HF-1/4RS |
Total |
|
Cows |
75 |
8 |
5 |
88 |
|
Lactations |
88 |
13 |
5 |
106 |
|
First lactations |
18 |
4 |
5 |
27 |
|
Second lactations |
43 |
6 |
- |
49 |
|
Third lactations |
9 |
3 |
- |
12 |
|
Fourth and more lactations |
18 |
- |
- |
18 |
|
Number of observation (daily yields) |
23,986 |
2,972 |
1,494 |
28,452 |
Models and Data
Analysis
Genetic
evaluation models require the construction of contemporary groups where
comparisons among predicted genetic values of animals for traits of economic
importance can be made in a fair manner.
For this Thai data set, one such contemporary group is the subclass
formed by cows from the same breed group x calving age group x lactation number
x calving season subclass.
Unfortunately, preliminary analyses showed that it was unfeasible to use
these subclasses because 56% of them were either empty or had a single
lactation. Consequently, two larger
subclasses were defined: 1) breed group x lactation x calving season, and 2)
breed group x lactation x calving age.
The data set yielded 20 breed group x lactation x calving season
subclasses, and 13 breed group x lactation x calving age subclasses. Analyses were conducted separately for each
one of the breed group x lactation x calving season and the breed group x
lactation x calving age subclasses.
Lactation curves were constructed using means of individual
cow daily milk yields (5 to 305d) within each subclass. Means of daily milk yields within breed
group x lactation x calving season and breed group x lactation x calving age
subclass were plotted using Microsoft Excel 2000
(Dodge and Stinson, 1999).
Daily milk
yields (5 to 305 d) for individual cows were predicted using each of the seven
equations used to model lactation curves within each subclass. The predictive ability of these seven
equations was tested using deviations of predicted minus actual daily milk
yields within lactations. Least squares
means of daily deviations of predicted minus actual daily productions were
computed for each breed group x lactation x calving season and breed group x
lactation x calving age subclass. To
evaluate the predictive ability of the seven models, least squares means of
differences between predicted and actual daily milk yields were tested (t-test)
for their difference with respect to zero, and for differences between
lactation models. The statistical model
used was:
dijkl =
+ subclassi + modelj + cowk
+ eijkl
where
dijkl = difference
between predicted and actual milk production in lactation day j of cow k within
subclass i and model j,
= overall mean,
subclassi = ith
breed
group x lactation x calving season or breed group x lactation x calving age,
modelj = jth
prediction model,
cowk = kth cow,
eijkl = residual.
All effects in
the model were assumed to be fixed, except for the residual term that was
assumed to be independent, identically distributed with mean zero and a common
variance. Because a cow will be in each
subclass type, separate computations were done for breed group x lactation x
calving season and breed group x lactation x calving age subclasses.
Individual cow
actual and predicted 100-d and 305-d milk yields were compared to evaluate the
performance of the seven models to predict milk yield for two traits of
economic importance used in genetic evaluation of dairy cattle. Least squares means of 100-d and 305-d
differences between predicted and actual milk yields were also computed for
each breed group x lactation x calving season and breed group x lactation x
calving age subclass for comparison purposes. These subclass deviations were
used to assess (t-tests) the ability of the seven models to predict 100-d and
305-d milk yields. The statistical
model used was:
dijk
=
+ subclassi + modelj + eijk
where
dijk = difference
between predicted and actual milk production of cow k at 100-d or at 305-d of
lactation, within subclass i and model j,
= overall mean,
subclassi = ith
breed
group x lactation x calving season or breed group x lactation x calving age,
modelj = jth
prediction model,
eijkl = residual.
All effects in
the model were assumed to be fixed, except for the residual term that was
assumed to be independent, identically distributed with mean zero and a common
variance. Separate computations were
done for breed group x lactation x calving season and breed group x lactation x
calving age subclasses.
Models that had smaller differences
between predicted and actual milk yields within breed group x lactation x
calving season and breed group x lactation x calving age subclasses were
considered to have better predictive ability.
The seven
mathematical models used here were as follows.
Model 1: Wood gamma function (Wood,
1967),
[1]
where
yt is the milk
yield on day t in each subclass, a is the initial
yield of lactation, b represents the increasing slope, and c
represents the decreasing slope.
Because nonlinear regression does not guarantee convergence (SAS, 1990),
natural logarithms were taken on both sides of equation [1] giving
where
is the residual. The
predicted yield on day t (yt) was computed as yt
= exp (ln yt).
Model 2: mixed log
second-degree polynomial model of milk yield on day in lactation and log of day
in lactation (linear and quadratic), of Ali and Schaeffer (1987),
[2]
where
yt is the milk yield on day
t in each subclass,
= t /305,
= ln(305/ t), t = days since calving or days in milk, b0, b1,
b2, b3, and b4 are the regression coefficients
where b0 is associated with peak yield, b3 and b4 the increasing slope of the curve and
b1 and b2 are associated with the decreasing slope; and et
is the residual.
Models 3, 4, 5, 6 and 7: second, third,
fourth, fifth, and sixth polynomial regression models, respectively,
Model 3:
[4]
Model 4:
[5]
Model 5:
[6]
Model 6:
[7]
Model 7:
[8]
where yt is the milk yield on day t in each subclass,
t = days since calving or days in milk, b0, b1, b2,
b3, b4, b5 and b6 are coefficients,
and et is the residual.
Estimates of
regression coefficients for models 1 through 7 were obtained using PROC REG of
the SAS program (SAS, 1990). This
program was also used to compute predicted daily milk yields (5 to 305 d) for
all models.
Individual
cow actual milk yields at 100-d and 305-d, predicted minus actual milk yield
daily deviations, and predicted 100-d, and 305-d milk yield deviations were
computed using the general SAS program (SAS, 1990). Least squares means of milk yield deviations per subclass were
obtained using LSMEANS statement of PROC GLM (SAS, 1990).
Predicted
means of daily milk yields per breed group x lactation x calving season and breed group x
lactation x calving age subclass using the
seven models above were drawn using Microsoft Excel 2000 (Dodge and Stinson,
1999). These lines were compared with
the corresponding lines of daily milk yields within subclasses.
Results and
Discussion
Actual Lactation Curves
Plots of actual
lactation curves showed that the shape of actual lactation curves varied across
breed group x lactation x calving season and breed group x lactation x calving
age subclasses. Lactation curves were
classified into four types according to the pattern of milk production from day
5 to 305. Type 1: milk production
increased continuously from the beginning of lactation, it reached a peak
between 50 to 70 d of lactation, then it decreased until the end of the
lactation (convex line); type 2: milk
production peaked before 30 d of lactation, then it steadily decreases to the
end of lactation (slightly convex line); type 3: milk production peaked twice, one between 30 to 70 d of
lactation, and another between 180 and 210 d of lactation (double peak line);
and type 4: milk production remained
fairly constant throughout the lactation (flat line).
Breed group subclasses. First lactation
curves of HF cows were of type 2, their milk production increased after calving
to reach a small peak around 30 d of lactation, and then it decreased until the
end of their lactations. Crossbred 1/2HF-1/2RS
cows had flat lactation curves (type 4), and 3/4HF-1/4RS cows had type 1
lactation curves, with a peak in milk production at about 75 d of lactation.

Figure 1. Means of first lactation
curves of HF, 1/2HF-1/2RS, and 3/4HF-1/4RS cows that calved at ages equal to or
greater than 30 mo
Figure 1 shows means
of first lactation curves of HF, 1/2HF-1/2RS, and 3/4HF-1/4RS cows that calved
at ages equal to or greater than 30 months.
The mean peak milk yield of HF cows (14.2 kg at 37 d in milk) was lower
than that of 3/4HF-1/4RS cows (17.3 kg at 65 d in milk), but higher than the
one from 1/2HF1/2RS cows (13.2 kg at 5 d in milk). The shape of the mean of lactation curves of 1/2HF1/2RS cows was
flatter than those of HF and 3/4HF-1/4RS in this calving age subclass.
Tekerli et al.
(2000) indicated that Holstein cows with flat lactation curves would be
expected to show higher persistency and higher milk yields per lactation. However, when comparing cows across breed
groups this may not be the case. Here
(Figure 1), the pattern mean lactation curves of 1/2HF-1/2RS cows that calved
at ages equal to or greater than 30 months was flatter than that of HF and
3/4HF-1/4RS cows, however, they had lower total 305-d milk yields. The means of
305-d milk yields of first lactations of HF, 1/2HF-1/2RS, and 3/4HF-1/4RS cows
that calved at ages equal to or greater than 30 months were 2,723.6, 2,014.5
and 3,579.0 kg, respectively. Milk
production in these three groups was clearly related to the fraction of Holstein
genes. Higher fractions (HF, and 3/4HF-1/4RS)
produced more milk than 1/2HF-1/2RS.
The 3/4HF-1/4RS group of cows produced more milk than purebred Holstein
cows probably due to the inability of Holstein cows to cope with the
environmental conditions (heat, humidity, internal and external parasite load)
at the SARTC farm in Thailand.
The results obtained
here reconfirmed those obtained in previous studies conducted at SARTC. SARTC (1999) reported that HF raised under
Northeastern conditions of Thailand needed more veterinary care than HF x RS crossbred
cows because of environmental stresses and tropical diseases. The F1 (1/2HF-1/2RS) cows were healthier,
but had lower milk production than HF and 3/4HF-1/4RS. Crossbred 3/4HF-1/4RS and higher HF fraction
(up to 87.25 %HF) cows produced more milk than 1/2HF-1/2RS and HF cows (SARTC,
1999).
Later lactations in
HF cows had similar patterns of lactation curves to the first lactation, except
that they had higher peak yields. On
the other hand, the shape patterns of later lactation curves of 1/2HF-1/2RS
cows were different from those of their first lactations. Later lactations of
1/2HF-1/2RS cows were of type 2 (slightly convex) instead of type 4 (flat). In
all breed groups initial and peak milk yield increased from lactation 1 to 3.

Figure 2. Means of lactation curves of
various lactation numbers in HF cows that calved in summer (a), and in
1/2HF-1/2RS cows that calved in winter (b)
Figures 2a and 2b
contain examples of the types of lactation curves found in the HF-RS herd of
this study. Figure 2a shows that the
means of first lactation curves of HF cows that calved in summer had lower
initial (12.5 kg) milk productions than those of second (17.1 kg), third (20.8
kg), and fourth-and-later (15.5 kg) lactations. All these lactations were of type 2; they had a short initial
increase in milk production, reached a peak yield within the first month of
lactation, and then decreased at various rates depending on the lactation
number until 305 d. The mean peak yield
of first, second, third, and fourth-and-later lactations cows were 14.6 kg (at
10 days in milk), 18.96 kg (at 10 days in milk), 24.2 kg (at 23 days in milk),
and 16.6 kg (at 8 days in milk), respectively.
The means of first
lactation curves of 1/2HF-1/2RS cows that calved in winter (Figure 2b) were
similar to those of HF cows. They had
an initial milk yield (12.2 kg) that was lower than that observed for second
(12.7 kg) and third lactations (16.9 kg).
The mean peak yield in the first lactation of 1/2HF-1/2RS cows (12.5 kg
at 7 days in milk) was also lower than that in the second (16.0 kg at 17 days
in milk) and in the third lactation (20.4 kg at 18 days in milk). Comparison of first and later lactations
could not be done with 3/4HF-1/4RS cows because they had only first lactation
records.
Mean peak yields
were lower and were reached earlier in the first lactation because cows needed
to devote a substantial portion of their nutritional intake to their growth and
development.
Because 1/2HF-1/2RS mature earlier than HF cows (4 years vs.
5 years; SARTC, 1999), nutritional requirements allocated to growth and
development during the second lactation were probably lower for 1/2HF-1/2RS
than for HF cows. Thus, 1/2HF-1/2RS
probably utilized more nutrients for milk production than HF cows, resulting in
longer peak yield times for 1/2HF-1/2RS animals. The degree of maturity of 1/2HF-1/2RS and HF cows was probably
similar at the third lactation, thus mean peak yields were achieved at similar
days. Mean peak yield would decrease as
cows get older and less productive (SARTC, 1999), as found here for
fourth-and-later-lactation HF cows.
Calving age subclasses. The patterns of
shape of lactation curves of older calving ages within lactation number were
similar to those of the younger calving ages within lactations. Older cows of all breed groups had type 2
lactation curves; milk production increased after calving for a short time,
reached a peak in less than a month, and then steadily decreased until 305 d.
Plots of means of
first lactation curves in Figure 3 show that HF cows that calved at less than
30 months of age (12.3 kg at 51 days in milk) had peak yields slightly lower
than those of cows calved at ages equal to or greater than 30 months (13.4 kg
at 25 days in milk). The difference in
peak yield between younger and older cows in the first lactation (1.1 kg) was
smaller than that in the second lactation (4.4 kg), and in the third lactation
(5.3 kg). In addition, younger cows had
flatter lactation curves than older cows in all lactations. In contrast, 3/4HF-1/4RS cows calving at
less than 30 months of age had peak yields (17.0 kg at 75 days in milk) higher
than those of cows that calved at ages equal to or greater than 30 months (16.1
kg at 65 days in milk). Differences in
daily milk production and peak yields translated in an advantage of
approximately 200 kg milk for later lactations over first lactations in HF
cows, and of about 100 kg in 3/4HF 1/4RS cows.
These small differences suggest that it might be advantageous to breed
cows to calve at younger ages within lactations, provided adequate feed is
available. An additional advantage
would be a potential longer herd life of younger calving cows.

Figure 3. Means of first lactation curves of HF (a)
and 3/4HF-1/4RS (b) cows that calved at less than 30 months, and equal to or
greater than 30 months of age
Calving season subclasses. Cows calving in
different seasons had different types of lactation curves. The HF cows that
calved in winter had lactation curves of type 2 (convex lactation curve), while
cows that calved in summer were of type 3 (two peaks per lactation), and cows
that calved in the rainy season had lactations of type 4 (flat lactation
curve). Examples of the types of
lactation curves by calving season are shown in Figure 4. The initial yield of lactation of HF cows
calved in the rainy season (7.5 kg) was lower than that of HF cows calved in
winter (10.6 kg), and summer (12.5kg). The peak yield of the first lactation of
HF cows that calved in the rainy season (11.5 kg) was lower than that of cows
that calved in winter (14.2 kg), and summer (14.6 kg).

Figure 4. Means of lactation curves of
first lactation HF (a) and second lactation 1/2HF-1/2RS (b) cows that calved in
winter, summer, and in the rainy season
Holstein Friesian cows that calved in the rainy season had type 4 (flat) lactation curves with lower initial production and peak yields than cows that calved in winter and summer. The lower production of HF cows in the rainy season was likely due to the incidence of mastitis,