Milk production of Holstein-Friesian, Jersey and crossbred cows milked once-a-day or twice-a-day in New Zealand

ABSTRACT

The objective of this study was to compare Holstein-Friesian (F), Jersey (J) and crossbred (F × J) cows milked once-a-day (OAD) or twice-a-day (TAD) in New Zealand for milk, fat and protein yield, lactation persistency and average somatic cell score (SCS). Data consisted of 223,149 herd-test records (89,297 and 133,852 OAD and TAD, respectively) from 11,848 F; 11,677 J and 27,720 F × J spring-calving cows between 2008 and 2012. Compared to TAD systems, cows milked OAD yielded 722, 28.0 and 22.2 kg less milk, fat and protein, respectively. Crossbred and J cows were less affected than F cows by OAD milking with a reduction in milk yield traits of ≤ 19.0%, while in F cows the reduction ranged between 19%–25%. Cows milked OAD had higher SCS than cows milked TAD (6.20 vs. 6.08). The greatest SCS difference in OAD and TAD systems was in first-lactation cows (6.40 vs. 6.02). Persistency of production traits was greater in F and F × J cows milked OAD than in F and F × J cows milked TAD, but J cows had similar milk and protein persistency in both systems. Overall, cows milked OAD had greater persistency than cows milked TAD (79%–90% vs. 76%–85%). These results show that, in commercial herds, the relative difference in production traits between OAD and TAD systems was smaller than those reported in experimental trials.

KEYWORDS:

• Milking frequency
• dairy cattle
• New Zealand
• milk yield traits
• lactation persistency

Introduction

Dairy production in New Zealand is typically a spring-calving, pasture-based system. This seasonal system is largely determined by the pattern of pasture growth, where calving of cows precedes a period of rapid pasture growth and cows are dried-off in late autumn, when the pasture growth is slowing (Holmes et al. 2002). In this context, dairy farming has been mostly undertaken using twice-a-day (TAD) milking. However, milking once-a-day (OAD) for the entire season has been adopted by some farmers for herd management and lifestyle benefits (Davis et al. 1999). The popularity of this system increased during the early 2000s (Bewsell et al. 2008), but the number of herds milked OAD has been stable in recent seasons. Currently, up to 5% of herds in New Zealand are milked OAD for the whole season (Stachowicz et al. 2014).

In New Zealand, three controlled research studies have compared cows milked OAD and TAD during the entire lactation (Holmes et al. 1992 [one season]; Cooper 2000 [one season]; Clark et al. 2006 [four seasons]). In these studies the breed groups were: Holstein-Friesian (F), Jersey (J) and their crosses Holstein-Friesian × Jersey (F × J) (Holmes et al. 1992), F × J (Cooper 2000) and purebred F and J (Clark et al. 2006). In general, F and F × J cows milked OAD were more affected than J cows because those breeds had decreased milk production (milk yield and milk solids yield) per cow and per hectare at a greater magnitude than J cows. Jersey cows may not have been affected as much because they can accumulate milk in their udders for longer periods than F cows (Davis et al. 1999).

In the literature, definitions of lactation persistency show a great inconsistency (Grossman et al. 1999). Swalve & Gengler (1999) defined persistency of lactation yield as the ability of the cow to produce milk yield constantly over the lactation period. This trait is positively associated with health and fertility (Dekkers et al. 1998), but has not been directly selected for in the dairy population in New Zealand (Morris et al. 2008), and is not currently included in international evaluations (Cole & Null 2009).

Although there are several definitions of lactation persistency (Grossman et al. 1999; Lopez-Villalobos et al. 2005; Togashi & Lin 2006), there is no a general agreement about the most suitable function to calculate it (Swalve & Gengler 1999). Swalve & Gengler (1999) indicated that an appropriate measure of persistency should be independent of total yields. They indicated that ratios of partial yields (splitting the lactation period in 90- or 100-day intervals) are more suitable and have been used widely since they are easy to compute and compare ascending and descending phases of lactation.

Under OAD systems, Harding et al. (2002) indicated that persistency is a major concern because generally cows dry-off early in the season and for this reason selection for persistency is the main option to reduce production losses in OAD systems (Clark et al. 2006; Hickson et al. 2006). Hickson et al. (2006) is the only study comparing lactation persistency in OAD and TAD systems. They used data from the study of Clark et al. (2006) and calculated three different measures of persistency (criteria based on ratios, variation of test-day yields and applications of random regression test-day models). In general, cows from TAD farms had shown better persistency than cows from OAD farms.

Quantifying milk production, lactation persistency and SCS under both milking systems is desirable because breed comparison in both OAD and TAD systems has only been evaluated previously in research facilities. The objective of this study was to compare, using data from commercial farms, milk production traits (milk, fat and protein), lactation persistency and somatic cell score between spring-calving F, J and F × J cows milked either OAD or TAD for an entire lactation in New Zealand.

Materials and methods

Data

A total of 16.8 million herd-test records of milk yield (MY), fat yield (FY), protein yield (PY) and somatic cell count (SCC) recorded from 2008 to 2012, and pedigree information were provided by Livestock Improvement Corporation (LIC), New Zealand. Only records from F, J and their crosses were considered, discarding animals missing breed identification.

Initial data were restricted as follows. Firstly, herd-test records were sorted based on milking frequency code provided by LIC (cows milked either OAD or TAD). In the present study, OAD herds are those in which 100% of the cows were milked OAD for the whole lactation. Using the GPS Visualizer (Schneider 2012), TAD herds were selected in a radius of 30 km from OAD herds. In some cases, in a given single map co-ordinate, an OAD herd was surrounded by several TAD herds. In such cases all TAD herds were selected. Also, in other cases, in a given single map co-ordinate, there were several OAD herds and only one TAD was found and selected. Secondly, herds with less than 40 cows were deleted from the analysis. Only records from spring-calving cows in their first five lactations were considered. Lastly, lactations where the period from calving to the first herd-test record was longer than 45 days were removed from the analysis.

Cows were considered pure-bred when their breed composition was ≥ 93.75% (15/16) from a particular breed (F or J), and F × J cows were those cows not meeting the criteria of purebred cow (i.e. ≥ 3/16 breed F or J and ≤  14/16 breed J or F, respectively). Coefficient of heterosis in a cow was calculated using the following equation (Dickerson 1973):

where and are proportion of F or J in the sire and and are proportion of J and F in the dam, respectively. Somatic cell count (SCC) was transformed to somatic cell score (SCS) calculated as SCS = log₂ (SCC).

The final data set included 89,297 herd-test records from 278 herds milked OAD and 133,852 herd-test records from 248 herds milked TAD. A summary of the population studied is presented in Table 1 by breed and lactation number.

Table 1. Number of herd-test records, number of lactations and cows per milking frequency and breed group.

	Once-a-day				Twice-a-day
	Holstein-Friesian	F × J^1	Jersey	Total	Holstein-Friesian	F × J^1	Jersey	Total
Herd-test records	11,525	44,005	33,767	89,297	38,081	72,531	23,240	133,852
Lactations	4116	14,813	9881	28,810	11,632	22,215	6298	40,145
Cows	3234	11,795	7348	22,377	8614	15,925	4329	28,868

¹Holstein-Friesian × Jersey crossbred cows.

Lactation curves

Lactation curves were modelled using a fifth order Legendre polynomial (Kirkpatrick et al. 1990). Considering as the level of production of a trait measured on day (t) of the lactation from calving. A polynomial of fifth order, which described a single observation, was defined as:

where are the regression coefficients to estimate, are the functions normalised to which is standardised to unit of time described by: where,  = 5 d and  = 270 d, which means that herd-test records between 5 to 270 days were converted respectively to the interval −1 to +1 (Schaeffer 2004).

According to Spiegel (1971), first fifth Legendre polynomial's functions of standardised unit of time and coefficients were calculated as:, , , , , .

The polynomial's equation presented includes the fixed regression coefficient for the population and random regression coefficients as deviation from the fixed population for each combination cow-lactation. For the cow-lactation ‘i' in days in milk ‘t' the equation is rewritten as follow:

where are the fixed regression coefficients of the population, are the random regression coefficients for each combination cow-lactation and is the random error associated with each observation of day t and cow-lactation i. The ASREML 3.0 software package (Gilmour et al. 2009) was used to model the lactation curves for MY, FY PY and SCS.

Pearson's correlation (r) between actual and estimated daily yields and SCS was calculated as a measurement of goodness of fit. Correlation summarises the discrepancy between observed values and the values predicted under the model used.

In addition to the fifth-order Legendre polynomial, polynomials of order 2, 3 and 4 were also tested. The decision to use a fifth-order polynomial was based on the smallest ‘Akaike's Information Criterion’ (AIC) value (Akaike 1973), which was achieved with polynomial of order 5. Therefore, this was selected to predict daily MY, FY, PY, lactation persistency and average SCS.

Persistency

In this study, lactation persistency of each combination cow-lactation was calculated as the area under the lactation curve from day 121 to 180 divided by the area under the curve from day 1 to 60, expressed as a percentage (Lopez-Villalobos et al. 2005). Compared to other measures of persistency, Lopez-Villalobos et al. (2005) found that the selected measure in this study was less dependent of total yields, as recommended by Swalve & Gengler (1999).

Statistical analysis

Accumulated yields of milk, fat and protein for each cow were estimated using the polynomial equation as the sum from day 5 to actual lactation length through 270 days. With respect to average SCS, it was calculated as the mean of the estimated SCS using the orthogonal polynomial's function. Then the following statistical linear model was used to obtain least square means and standard errors for each parameter of the lactation curve, lactation persistency, predicted accumulated milk, fat and protein yields and average SCS using PROC MIXED of SAS version 9.3 (SAS Institute Inc.). The model was:where is the MY, FY, PY or average SCS of herd-season i, milking frequency j and lactation number k; is population mean; is random effect of contemporary group i defined as cows calving in the same herd and production season (i = 1,2 … 1568); is the fixed effect of the milking frequency (j = 1 and 2); is fixed effect of the lactation number k (k = 1,2 … 5); is the interaction between milking frequency j and lactation number k; is the regression coefficient associated with the linear effect of proportion of F in milking frequency 1; is the regression coefficient associated with the linear effect of proportion of F in milking frequency 2; is the regression coefficient associated with the linear effect of coefficient of heterosis in milking frequency 1; is the regression coefficient associated with the linear effect of coefficient of heterosis in milking frequency 2; is the regression coefficient associated with linear effect of deviation days from the median calving date of the herd per a given season and is the residual random error associated to the observation .

Results

Tables 2–5 detail the estimate of regression coefficients describing the lactation curves of MY, FY, PY and SCS by lactation number for cows of the different breed groups under the two milking frequencies.

Table 2. Least square means and standard errors of regression coefficients of the lactation curve for milk yield modelled with a fifth-order Legendre polynomial fitted to Holstein-Friesian (F), Jersey (J) and first F × J crossbred cows of different lactation number and milking frequency (MF).

MF^1	Breed	Lactation number	α0	α1	α2	α3	α4	α5
OAD			11.16^a ± 0.09	−5.08^a ± 0.05	−0.55^a ± 0.04	0.57^a ± 0.04	−0.24^a ± 0.04	−0.03^a ± 0.04
TAD			13.65^b ± 0.08	−6.32^b ± 0.05	−0.10^b ± 0.04	0.50^a ± 0.04	−0.58^b ± 0.04	0.45^b ± 0.04
OAD	F		13.37^a ± 0.16	−5.91^a ± 0.09	−0.47^a ± 0.08	0.66^a ± 0.06	−0.45^a ± 0.06	0.13^a ± 0.06
TAD	F		17.69^b ± 0.11	−8.17^b ± 0.06	0.03^b ± 0.05	0.70^a ± 0.04	−0.80^b ± 0.04	0.83^b ± 0.04
OAD	F × J		12.61^a ± 0.14	−5.61^a ± 0.10	−0.62^a ± 0.08	0.57^a ± 0.07	−0.26^a ± 0.07	0.22^a ± 0.08
TAD	F × J		15.65^b ± 0.13	−7.13^b ± 0.10	−0.22^b ± 0.08	0.67^a ± 0.07	−0.60^b ± 0.07	0.56^b ± 0.08
OAD	J		11.13^a ± 0.08	−5.10^a ± 0.05	−0.55^a ± 0.04	0.55^a ± 0.04	−0.21^a ± 0.04	−0.03^a ± 0.04
TAD	J		13.51^b ± 0.09	−6.14^b ± 0.06	−0.33^b ± 0.05	0.51^a ± 0.05	−0.48^b ± 0.05	0.39^b ± 0.05
OAD		1	8.25^a ± 0.09	−3.51^a ± 0.06	−0.78^a ± 0.05	0.39^a ± 0.04	0.01^a ± 0.04	−0.43^a ± 0.04
TAD		1	10.45^b ± 0.09	−4.72^b ± 0.06	−0.22^b ± 0.05	0.33^a ± 0.04	−0.43^b ± 0.04	0.11^b ± 0.04
OAD		2	10.70^a ± 0.09	−4.81^a ± 0.06	−0.58^a ± 0.04	0.53^a ± 0.04	−0.23^a ± 0.04	−0.05^a ± 0.04
TAD		2	12.96^b ± 0.09	−5.91^b ± 0.05	−0.11^b ± 0.04	0.37^a ± 0.04	−0.52^b ± 0.04	0.44^b ± 0.04
OAD		3	11.83^a ± 0.09	−5.47^a ± 0.06	−0.47^a ± 0.05	0.60^a ± 0.04	−0.29^a ± 0.04	0.08^a ± 0.04
TAD		3	14.44^b ± 0.09	−6.68^b ± 0.06	−0.03^b ± 0.05	0.52^a ± 0.04	−0.66^b ± 0.04	0.49^b ± 0.04
OAD		4	12.39^a ± 0.09	−5.73^a ± 0.06	−0.45^a ± 0.05	0.64^a ± 0.04	−0.33^a ± 0.04	0.13^a ± 0.04
TAD		4	15.08^b ± 0.09	−7.08^b ± 0.06	−0.06^b ± 0.05	0.65^a ± 0.04	−0.66^b ± 0.04	0.59^b ± 0.04
OAD		5	12.63^a ± 0.09	−5.87^a ± 0.06	−0.48^a ± 0.05	0.71^a ± 0.04	−0.33^a ± 0.04	0.14^a ± 0.04
TAD		5	15.29^b ± 0.09	−7.20^b ± 0.06	−0.07^b ± 0.05	0.61^a ± 0.04	−0.62^a ± 0.04	0.64^b ± 0.04

¹MF OAD, milking once-daily; MF TAD, milking twice-daily.

^a,bLeast square means with different superscripts are significantly different between milking frequencies with breed-type or within lactation number (P < 0.05).

Table 3. Least square means and standard errors (x10²) of regression coefficients of the lactation curve for fat yield modelled with a fifth-order Legendre polynomial fitted to Holstein-Friesian (F), Jersey (J) and first F × J crossbred cows of different lactation number and milking frequency (MF).

MF^1	Breed	Lactation number	α0	α1	α2	α3	α4	α5
OAD			65.13^a ± 0.40	−16.83^a ± 0.24	−2.97^a ± 0.19	1.76^a ± 0.17	−2.01^a ± 0.15	−0.47^a ± 0.16
TAD			76.76^b ± 0.39	−22.09^b ± 0.25	−1.24^b ± 0.20	1.59^a ± 0.17	−3.66^b ± 0.16	1.23^b ± 0.17
OAD	F		65.16^a ± 0.67	−17.43^a ± 0.40	−1.87^a ± 0.31	1.57^a ± 0.27	−2.26^a ± 0.27	−0.24^a ± 0.23
TAD	F		79.74^b ± 0.45	−24.97^b ± 0.26	0.82^b ± 0.20	0.68^a ± 0.17	−3.69^b ± 0.17	3.63^b ± 0.14
OAD	F × J		67.34^a ± 0.68	−18.07ª ± 0.48	−2.55^a ± 0.39	1.23^a ± 0.34	−2.08^a ± 0.33	0.36ª ± 0.36
TAD	F × J		77.99^b ± 0.68	−22.79^b ± 0.48	−0.89^a ± 0.40	1.21^a ± 0.35	−3.20^a ± 0.34	1.46^a ± 0.37
OAD	J		65.34^a ± 0.43	−17.06^a ± 0.28	−3.06^a ± 0.23	1.74^a ± 0.20	−1.96^a ± 0.19	−0.42^a ± 0.19
TAD	J		77.31^b ± 0.50	−21.76^b ± 0.34	−2.02^a ± 0.27	1.58^a ± 0.24	−3.44^b ± 0.23	1.19^b ± 0.23
OAD		1	50.82^a ± 0.42	−11.26^a ± 0.26	−2.86^a ± 0.21	1.13^a ± 0.18	−1.12^a ± 0.17	−1.98^a ± 0.18
TAD		1	62.08^b ± 0.40	−16.69^b ± 0.26	−0.60^b ± 0.21	0.94^a ± 0.19	−3.12^b ± 0.18	0.27^b ± 0.19
OAD		2	62.97^a ± 0.41	−15.75^a ± 0.25	−3.02^a ± 0.21	1.60^a ± 0.18	−2.02^a ± 0.17	−0.63^a ± 0.18
TAD		2	73.69^b ± 0.40	−20.28^b ± 0.26	−1.37^b ± 0.21	1.15^a ± 0.18	−3.45^b ± 0.17	1.05^b ± 0.18
OAD		3	68.72^a ± 0.42	−18.40^a ± 0.26	−2.86^a ± 0.21	1.91^a ± 0.18	−2.37^a ± 0.17	−0.04^a ± 0.18
TAD		3	80.55^b ± 0.40	−23.31^b ± 0.26	−1.23^b ± 0.21	1.71^a ± 0.18	−4.05^b ± 0.17	1.32^b ± 0.18
OAD		4	71.01^a ± 0.43	−19.13^a ± 0.27	−3.05^a ± 0.22	2.03^a ± 0.19	−2.47^a ± 0.18	0.08^a ± 0.19
TAD		4	83.32^b ± 0.40	−24.98^b ± 0.26	−1.38^b ± 0.22	2.19^a ± 0.10	−3.94^b ± 0.18	1.57^b ± 0.19
OAD		5	72.12^a ± 0.43	−19.61^a ± 0.28	−3.09^a ± 0.23	2.11^a ± 0.20	−2.52^a ± 0.19	0.26^a ± 0.20
TAD		5	84.18^b ± 0.41	−25.22^b ± 0.27	−1.62^b ± 0.22	1.94^a ± 0.19	−3.75^b ± 0.18	1.95^b ± 0.20

¹MF OAD, milking once-daily; MF TAD, milking twice-daily.

^a,bLeast square means with different superscripts are significantly different between milking frequencies with breed-type or within lactation number (P < 0.05).

Table 4. Least square means and standard errors (x10²) of regression coefficients of the lactation curve for protein yield modelled with a fifth-order Legendre polynomial fitted to Holstein-Friesian (F), Jersey (J) and first F × J crossbred cows of different lactation number and milking frequency (MF).

MF^1	Breed	Lactation number	α0	α1	α2	α3	α4	α5
OAD			47.71^a ± 0.33	−15.19^a ± 0.19	−1.03^a ± 0.16	0.57^a ± 0.14	−0.77^a ± 0.14	−2.68^a ± 0.14
TAD			55.95^b ± 0.31	−19.44^b ± 0.19	0.91^b ± 0.16	0.29^a ± 0.15	−2.03^b ± 0.14	−1.28^b ± 0.14
OAD	F		51.81^a ± 0.58	−16.61^a ± 0.33	−0.45^a ± 0.29	0.55^a ± 0.25	−1.32^a ± 0.22	−2.51^a ± 0.23
TAD	F		64.50^b ± 0.40	−24.01^b ± 0.22	2.21^b ± 0.17	0.23^a ± 0.16	−2.57^b ± 0.14	0.16^b ± 0.15
OAD	F × J		51.24^a ± 0.52	−16.34^a ± 0.36	−1.11^a ± 0.32	0.30^a ± 0.27	−0.79^a ± 0.29	−1.89^a ± 0.31
TAD	F × J		60.54^b ± 0.51	−21.56^b ± 0.36	0.85^b ± 0.32	0.53^a ± 0.28	−1.94^a ± 0.29	−0.90^a ± 0.31
OAD	J		47.88^a ± 0.33	−15.35^a ± 0.21	−1.09^a ± 0.18	0.53^a ± 0.16	−0.66^a ± 0.16	−2.64^a ± 0.16
TAD	J		56.13^b ± 0.38	−19.00^b ± 0.25	−0.02^b ± 0.21	0.40^a ± 0.19	−1.73^b ± 0.19	−1.46^b ± 0.19
OAD		1	36.02^a ± 0.34	−9.89^a ± 0.21	−1.58^a ± 0.17	0.09^a ± 0.15	0.15^a ± 0.15	−3.97^a ± 0.15
TAD		1	43.83^b ± 0.32	−13.79^b ± 0.20	0.41^b ± 0.17	−0.15^a ± 0.15	−1.39^b ± 0.15	−2.33^b ± 0.15
OAD		2	45.99^a ± 0.33	−14.27^a ± 0.20	−1.15^a ± 0.17	0.43^a ± 0.15	−0.75^a ± 0.15	−2.72^a ± 0.15
TAD		2	53.66^b ± 0.32	−18.08^b ± 0.20	0.86^b ± 0.17	−0.13^a ± 0.15	−1.85^b ± 0.15	−1.27^b ± 0.15
OAD		3	50.60^a ± 0.34	−16.60^a ± 0.20	−0.84^a ± 0.18	0.65^a ± 0.15	−0.97^a ± 0.15	−2.31^a ± 0.15
TAD		3	59.09^b ± 0.32	−20.81^b ± 0.20	1.20^b ± 0.17	0.36^a ± 0.15	−2.36^b ± 0.15	−1.16^b ± 0.15
OAD		4	52.57^a ± 0.34	−17.41^a ± 0.21	−0.73^a ± 0.18	0.77^a ± 0.16	−1.14^a ± 0.15	−2.24^a ± 0.16
TAD		4	61.24^b ± 0.32	−22.10^b ± 0.20	1.06^b ± 0.18	0.78^a ± 0.16	−2.38^b ± 0.15	−0.88^b ± 0.16
OAD		5	53.38^a ± 0.35	−17.81^a ± 0.22	−0.83^a ± 0.19	0.92^a ± 0.17	−1.15^a ± 0.16	−2.15^a ± 0.16
TAD		5	61.90^b ± 0.33	−22.41^b ± 0.21	1.03^b ± 0 .19	0.60^a ± 0.16	−2.17^b ± 0.16	−0.76^b ± 0.16

¹MF OAD, milking once-daily; MF TAD, milking twice-daily.

^a,bLeast square means with different superscripts are significantly different between milking frequencies with breed-type or within lactation number (P < 0.05).

Table 5. Least square means and standard errors of regression coefficients of the somatic cell curve modelled with a fifth-order Legendre polynomial fitted to Holstein-Friesian (F), Jersey (J) and F × J crossbred cows of different lactation number and milking frequency (MF).

MF^1	Breed	Lactation number	α0	α1	α2	α3	α4	α5
OAD			6.29^a ± 0.02	0.88^a ± 0.01	0.38^a ± 0.01	−0.16^a ± 0.01	0.08^a ± 0.01	−0.16^a ± 0.01
TAD			6.26^a ± 0.02	0.82^a ± 0.01	0.39^a ± 0.01	−0.16^a ± 0.01	0.09^a ± 0.01	−0.13^a ± 0.01
OAD	F		6.50^a ± 0.02	0.97^a ± 0.01	0.27^a ± 0.01	−0.21^a ± 0.01	0.10^a ± 0.01	−0.14^a ± 0.01
TAD	F		6.32^b ± 0.02	0.88^b ± 0.01	0.34^b ± 0.01	−0.19^a ± 0.01	0.12^a ± 0.01	−0.12^a ± 0.01
OAD	F × J		6.45^a ± 0.04	0.91^a ± 0.03	0.36^a ± 0.02	−0.24^a ± 0.02	0.09^a ± 0.02	−0.10^a ± 0.02
TAD	F × J		6.28^a ± 0.05	0.88^a ± 0.03	0.36^a ± 0.02	−0.17^a ± 0.02	0.09^a ± 0.02	−0.16^a ± 0.02
OAD	J		6.29^a ± 0.02	0.87^a ± 0.01	0.39^a ± 0.01	−0.15^a ± 0.01	0.07^a ± 0.01	−0.16^a ± 0.01
TAD	J		6.24^a ± 0.03	0.78^b ± 0.02	0.43^a ± 0.01	−0.15^a ± 0.01	0.08^a ± 0.01	−0.13^a ± 0.01
OAD		1	6.44^a ± 0.02	0.55^a ± 0.01	0.47^a ± 0.01	−0.07^a ± 0.01	0.05^a ± 0.01	−0.13^a ± 0.01
TAD		1	6.18^b ± 0.02	0.58^a ± 0.01	0.52^a ± 0.01	−0.11^a ± 0.01	0.06^a ± 0.01	−0.09^a ± 0.01
OAD		2	6.08^a ± 0.02	0.86^a ± 0.01	0.43^a ± 0.01	−0.13^a ± 0.01	0.07^a ± 0.01	−0.17^a ± 0.01
TAD		2	6.06^a ± 0.02	0.82^a ± 0.01	0.43^a ± 0.01	−0.15^a ± 0.01	0.09^a ± 0.01	−0.14^a ± 0.01
OAD		3	6.18^a ± 0.02	0.97^a ± 0.01	0.36^a ± 0.01	−0.18^a ± 0.01	0.09^a ± 0.01	−0.17^a ± 0.01
TAD		3	6.19^a ± 0.02	0.88^b ± 0.01	0.36^a ± 0.01	−0.17^a ± 0.01	0.10^a ± 0.01	−0.14^a ± 0.01
OAD		4	6.32^a ± 0.02	1.00^a ± 0.01	0.33^a ± 0.01	−0.19^a ± 0.01	0.08^a ± 0.01	−0.16^a ± 0.01
TAD		4	6.34^a ± 0.02	0.89^b ± 0.01	0.34^a ± 0.01	−0.17^a ± 0.01	0.10^a ± 0.01	−0.15^a ± 0.01
OAD		5	6.44^a ± 0.02	1.02^a ± 0.02	0.32^a ± 0.01	−0.23^a ± 0.01	0.08^a ± 0.01	−0.15^a ± 0.01
TAD		5	6.52^a ± 0.02	0.90^b ± 0.02	0.32^a ± 0.01	−0.19^a ± 0.01	0.09^a ± 0.01	−0.13^a ± 0.01

¹MF OAD, milking once-daily; MF TAD, milking twice-daily.

^a,bLeast square means with different superscripts are significantly different between milking frequencies with breed-type or within lactation number (P < 0.05).

At the beginning of the lactation (intercept), cows milked TAD had higher values for yields of milk fat and protein than cows milked OAD. For SCS, however, the intercept was only different in F, F × J and first-lactation cows, where SCS was higher in cows under OAD systems.

In general, the predictive ability of fifth-order Legendre polynomial for the four traits studied was high for all the traits (r ≥ 0.98) (data not shown), and it showed a more flexible lactation shape than the other Legendre polynomials tested.

Across milking frequency, typical milk lactation curves (rapid increase in milk yield up to about 30–45 days post calving, and then a gradual decline to the end of the lactation; Grossman et al. 1999) were found in mature cows (second to fifth lactation) (Figure 1) and in the three breeds (Figure 2), although a less well-defined peak, in particular in J cows, is observed for lactation curves in OAD systems.

Figure 1. Predicted milk yield from calving to 270 days of lactation across lactation (L₁–L₅). A, Cows milked once-a-day; B, cows milked twice-a-day.

Abbreviations: L₁, first lactation; L₂, second lactation; L₃, third lactation; L₄, fourth lactation; L₅, fifth lactation.

Figure 2. Predicted milk yield from calving to 270 days of lactation of Holstein-Friesian, Jersey and F × J crossbred cows. A, Milked once-a-day; B, milked twice-a-day.

Importantly, for each breed and milking frequency, the predicted peaks occurred before 30 days of lactation. However, in first-lactation cows, for both milking frequencies, the same lactation curve did not exist. The predicted lactation curve for first-lactation cows milked OAD had a predicted peak of production at the beginning of the lactation, which gradually declined to a period of stability between day 30 and day 100 of lactation followed by a constant decline to the last day modelled. In first-lactation cows milked TAD there was a predicted peak at the beginning, a gradual decline to day 240 of lactation with a clear increase in the last 30 days of lactation.

Least square means for predicted MY, FY, PY and average SCS and persistency for the lactation are presented in Table 6. The results show that cows milked OAD yielded 722 kg less milk (21.1%), 28.0 kg less fat (16.9%), 22.2 kg less protein (17.2%) and had 0.12 greater SCS (1.9%) than their equivalent TAD cows.

Table 6. Predicted least square means with standard errors of total yields and lactation persistency of milk, fat and protein, and average somatic cell score (SCS) in New Zealand dairy cattle, by milking frequency (MF), breed group and lactation number.

MF^1	Breed^2	Lactation number	Milk yield		Fat yield		Protein yield		Average SCS
			Mean (kg/cow)	Persistency (%)	Mean (kg/cow)	Persistency (%)	Mean (kg/cow)	Persistency (%)	Mean
OAD			2698^a ± 30	78.7^a ± 0.4	138.1^a ± 1.5	89.8^a ± 0.4	106.7^a ± 1.2	84.6^a ± 0.4	6.20^a ± 0.02
TAD			3420^b ± 27	75.9^b ± 0.4	166.1^b ± 1.4	85.1^b ± 0.4	128.9^b ± 1.1	79.4^b ± 0.4	6.08^b ± 0.02
OAD	F		2879^a ± 54	79.4^a ± 0.7	135.5^a ± 2.4	88.8^a ± 0.8	108.7^a ± 1.9	84.3^a ± 0.7	6.30^a ± 0.04
TAD	F		3824^b ± 37	75.8^b ± 0.5	167.1^b ± 1.7	83.5^b ± 0.5	136.6^b ± 1.4	78.1^b ± 0.5	6.13^b ± 0.02
OAD	F × J		2787^a ± 40	80.5^a ± 0.6	143.3^a ± 2.1	90.5^a ± 0.7	110.4^a ± 1.6	86.0^a ± 0.6	6.28^a ± 0.04
TAD	F × J		3446^b ± 36	76.9^b ± 0.5	168.8^b ± 1.9	86.4^b ± 0.6	130.9^b ± 1.4	80.6^b ± 0.6	6.06^b ± 0.03
OAD	J		2427^a ± 30	78.3^a ± 0.5	137.1^a ± 1.8	90.8^a ± 0.6	101.4^a ± 1.3	84.7^a ± 0.6	6.11^a ± 0.03
TAD	J		2929^b ± 34	77.4^a ± 0.6	162.2^b ± 2.0	88.1^b ± 0.7	118.8^b ± 1.5	82.4^a ± 0.6	6.05^a ± 0.03
OAD		1	2006^a ± 30	82.8^a ± 0.4	104.2^a ± 1.5	92.4^a ± 0.5	79.0^a ± 1.9	88.4^a ± 0.5	6.40^a ± 0.02
TAD		1	2675^b ± 27	77.2^b ± 0.4	131.9^b ± 1.4	85.1^b ± 0.4	100.5^b ± 1.1	81.2^b ± 0.4	6.02^b ± 0.02
OAD		2	2603^a ± 30	78.8^a ± 0.4	133.8^a ± 1.5	90.1^a ± 0.5	103.3^a ± 1.2	84.9^a ± 0.4	5.97^a ± 0.02
TAD		2	3277^b ± 27	76.3^b ± 0.4	159.7^b ± 1.4	85.8^b ± 0.4	124.2^b ± 1.1	79.9^b ± 0.4	5.85^b ± 0.02
OAD		3	2868^a ± 30	77.4^a ± 0.4	147.1^a ± 1.5	88.6^a ± 0.5	113.9^a ± 1.2	83.3^a ± 0.5	6.06^a ± 0.02
TAD		3	3611^b ± 27	75.3^b ± 0.4	175.3^b ± 1.4	84.9^b ± 0.4	136.5^b ± 1.1	78.7^b ± 0.4	5.99^b ± 0.02
OAD		4	2985^a ± 31	77.5^a ± 0.4	151.7^a ± 1.5	89.0^a ± 0.5	118.0^a ± 1.2	83.2^a ± 0.5	6.21^a ± 0.02
TAD		4	3750^b ± 28	75.2^b ± 0.4	181.3^b ± 1.4	84.7^b ± 0.4	141.2^b ± 1.1	78.5^b ± 0.4	6.15^a ± 0.02
OAD		5	3032^a ± 31	77.2^a ± 0.5	153.9^a ± 1.5	88.8^a ± 0.5	119.5^a ± 1.2	83.0^a ± 0.5	6.34^a ± 0.03
TAD		5	3785^b ± 28	75.4^b ± 0.4	182.3^b ± 1.4	85.0^b ± 0.4	142.2^b ± 1.1	78.9^b ± 0.4	6.36^a ± 0.02

¹MF OAD, milking once-daily; MF TAD, milking twice-daily.

²F,  Holstein-Friesian; J, Jersey; F × J, crossbred cows.

^a,bLeast square means with different superscripts are significantly different between milking frequencies with breed-type or within lactation number (P < 0.05).

Comparison across breed groups showed that F cows were more affected than the other breed groups by OAD milking. Holstein-Friesian cows milked OAD yielded 24.7%, 18.9% and 20.4% less milk, fat and protein, respectively, than F cows milked TAD. Compared to J cows under TAD systems, J cows milked OAD yielded 17.1%, 15.5% and 14.6% less milk, fat and protein, respectively.

The reduction in milk production traits from TAD to OAD was not constant across lactation numbers. First-lactation cows milked OAD had 21%–25% less milk, fat and protein yield compared to their counterpart cows milked TAD. In mature cows, this difference was 16%–21%, reflecting an interaction between milking frequency and lactation number.

The difference of SCS across milking frequency was greatest in first-lactation cows (6.3%) and decreased in successive lactations, indicating a significant interaction between milking frequency and lactation number. In fourth- and fifth-lactation cows, no difference was found across milking frequency. The results also indicate an interaction between breed groups and milking frequency, since F and F × J cows milked OAD had higher values of SCS than their counterpart cows milked TAD, meanwhile no difference in the average SCS was observed in J cows milked either OAD or TAD.

The results presented in Table 6 indicate that, in general, cows milked OAD had higher persistency of milk, fat and protein yield (79%–90% vs. 76%–85% in OAD and TAD, respectively). Within breeds, J cows milked OAD and TAD had similar milk and protein persistency; in contrast, F and F × J cows milked OAD had slightly greater lactation persistency for the three production traits than their counterparts milked TAD.

Discussion

The three breed groups milked OAD yielded 15%–25% less milk, fat and protein than the respective breed groups milked TAD. However the reduction in milk yield was, in relative values, greater in F cows than in F × J and J cows, which is in agreement with Cooper (2000) and Clark et al. (2006) and confirms that a breed × milking frequency interaction exists.

The reduction in milk yield in J cows milked OAD (15%–17%) was smaller than the reduction of 20%–25% estimated in a farmlet study carried out in New Zealand (Clark et al. 2006). For the other breed groups, the relative differences were considerably smaller than the studies of Cooper (2000) in F × J, and Clark et al. (2006) in F cows. Both studies noted a difference ranging between 29%–32% in milk, fat and protein yield. Similar to the present study, the latter authors also observed that, compared to F cows, J cows were less affected by OAD milking.

The results of this study are important as the data came from commercial herds instead of experimental herds, and the reductions in milk and milk solids yield were not as large on commercial farms as has been reported in experiments. The latter suggest that commercial farms milking OAD have increased their stocking rate (SR) as a way to compensate the decrease in milk production (Stockdale 2006), and it appears that the increased SR may have been lower than the 17% suggested by Cooper & Clark (2001), because cows under OAD systems have similar (or slightly less) dry matter intake to cows under TAD (Stelwagen et al. 2013).

The reduction in the milk yield traits found in this study compared to experiments may be due to low input farms operating TAD, because the average yield of milk solids presented in Table 6 (295 kg/cow) is lower than the national milk solids average for the period 2008–2012 (330 kg/cow; LIC & Dairy NZ 2014), although this cannot be confirmed.

No data were available showing how long each farm had been operating under OAD milking systems; however, it is assumed that farmers have systematically selected and culled cows unsuited to less frequent milking, as theorised by Woolford et al. (1982). Tong et al. (2003), using the first two seasons of the study from Clark et al. (2006), reported a reduction between lactation length in TAD and OAD (from 26 to 8 days) and in milk solids (kg/cow) among J cows (from 55 to 42 kg), supporting the possibility of selection for OAD milking.

The greater reduction in first-lactation cows milked OAD compared to the same group milked TAD has been found in previous studies (Clark et al. 2006). In contrast, Cooper (2000) found that reduction in the milk, fat and protein yield was not affected by lactation number, and they observed that mature cows milked OAD tended to have greater losses than first-lactation cows milked OAD. The greater reduction in first-lactation cows found in this study could be attributed to cows in their first lactation having a relatively immature cisternal development in comparison to later lactations (Knight & Wilde 1993). Clark et al. (2006) theorised that those cows might not be capable of accommodating the storage of high amounts of milk. In this study, the reduction of milk solids between OAD and TAD in first-lactation cows was 17%. A reduction of 24% and 38% in J and F cows, respectively, was reported in the pastoral study of Clark et al. (2006). The findings might indicate that first-lactation cows of this study are daughters of cows more adapted to OAD; consequently, the reduction observed is lower than the reduction reported in experiments.

The results showed increased SCS in cows milked OAD compared with cows milked TAD, which is consistent with experimental studies (Holmes et al. 1992; Cooper 2000; Clark et al. 2006). Kamote et al. (1994) had suggested that higher levels of SCS could be caused by a ‘concentration effect’ rather than higher somatic cell count. However, several famers milking OAD have reported lower SCC (and therefore SCS), after culling high SCC cows and other cows deemed unsuitable for OAD. Dalley et al. (2007) attributed this occurrence to better management practices and possibly to less exposure to pathogens during the milking process because cows are only milked once daily. The variation in SCS observed in the initial three lactations may be the result of high SCS cows being culled prior to fourth and fifth lactations in OAD milking systems.

The lactation persistencies observed in the present study for the milk yield traits were slightly greater in OAD compared with to TAD (79%–90% vs. 76%–85%). These results were unexpected because there is evidence of an accelerated mammary involution with lower milking frequency (Knight & Dewhurst 1994; Capuco et al. 2003; Bernier-Dodier et al. 2010). Further, Hickson et al. (2006) had shown that F and J cows milked OAD had lower persistency than cows milked TAD. Those results were obtained in two of the three calculations of persistency, and neither of these calculations was used in this study. In a third measure of persistency (based on accumulated yields from quadratic polynomial curves; Kamidi 2005), Hickson et al. (2006) found that J cows milked OAD and TAD had better lactation persistency of milk, protein and lactose than F cows milked OAD and TAD. Comparison of the results presented in this study with those in the study of Hickson et al. (2006) should be viewed with caution due to criteria differences in the measures of lactation persistency (Grossman et al. 1999).

The findings of this study might also support the conclusions from the study of Woolford et al. (1982). They indicated that there may be an important genetic component to the decrease of milk yield when cows went from TAD to OAD and, therefore, some cows are more tolerant and adapt better to the OAD regime in successive lactations than other cows. Those cows which are genetically predisposed to adapt to OAD systems show greater lactation persistency when they are milked OAD (Clark et al. 2006; Hickson et al. 2006).

Conclusion

The three main breed groups in New Zealand produced significantly greater milk, fat and protein yield when milked TAD; however, the relative difference to OAD found in this study is smaller than reported in studies on research farms. In addition, greater lactation persistency under OAD systems in F and F × J may be an indication that cows have adapted to OAD systems. With the exception of J cows, cows milked OAD tend to show higher SCS, although these values varied with increasing lactation number.

Acknowledgements

The principal author acknowledges the support from Programa Formación de Capital Humano Avanzado, Becas Chile, CONICYT doctoral scholarship, and information provided by Livestock Improvement Corporation (Hamilton, New Zealand). Comments and discussion with NW Sneddon were extremely helpful.

Disclosure statement

No potential conflict of interest was reported by the authors.