Farm systems research at Ruakura – a 60-year legacy underpinning profitable and sustainable pasture-based dairy systems

ABSTRACT

The Number 2 Dairy (No. 2 Dairy) unit at Ruakura, Hamilton, New Zealand was established as a dairy research farm in 1943 and became recognised globally as one of the leading institutes for farm systems science. The research undertaken was fundamental to the ‘systemisation’ of component research into replicable, efficient, competitive, and sustainable dairy systems. From McMeekan’s seminal research in the 1950s on rotational grazing to the experiments of Carter, Campbell, Bryant, Penno and others, who brought greater resolution to the key principles influencing successful outcomes on-farm, the more than 60 years of experimental research led to ‘good practice’ rules that were to transcend geographical, climatic, and system-level interactions. Much of the research can only be found in Annual Reports or in summary form in conference proceedings and not in an electronic form. Our objective was to collate and discuss the research undertaken, thereby ensuring that the legacy of this great work and the wisdom amassed are not lost. To this end, we have compiled a compendium of the research undertaken at No. 2 Dairy, with the hope that it be used to inform the scientific rationale for system design and to stimulate future hypotheses for system improvement.

KEYWORDS:

• No. 2 Dairy
• grazing systems
• rotational grazing
• set stocked
• stocking rate

Introduction

Globally, the traditional New Zealand system of dairy production is regarded as the archetypal grazing system (Roche et al. 2017a, 2017b) and it is a prototype for a circular bioeconomy (Muscat et al. 2021), wherein outputs are provided for, primarily, by natural inputs, with little, if any, need for imported inputs beyond superphosphate and some replacement potassium fertiliser. It truly reflects the capture of solar energy in the most nutritionally complete food for human consumption (Park 2009; Gorska-Warsewicz et al. 2019). The system has been imitated by a few (e.g. Ireland, Australia, the UK, South Africa, the UK; Roche et al. 2017a, 2017b), but New Zealand’s latitudinal situation as well as the maritime influence result in natural and repeatable climatic statistics (Roche et al. 2009a) that provide its farmers with advantages over competitors. For example:

• cows can graze for 365 days of the year, with no requirement for housing infrastructure or supplementary feeding;

• pasture supply and animal demand can be managed to ensure that animals are appropriately fed;

• milk production/cow is less than in housed systems, but milk production/ha is equivalent to those systems utilising crops in the form of a total mixed ration; and

• operating expenses/kg milk are the lowest globally (Dillon et al. 2005), maximising margin/kg milk sold.

Roche et al. (2017a) reviewed the century of grazing research undertaken between 1917 and 2017 and acknowledged that the farm systems research undertaken at Ruakura was fundamental in the ‘systemisation’ of component research into a replicable, efficient, competitive, and sustainable dairy system, known by most as the New Zealand system of dairy farming. Unfortunately, however, much of this research can only be found in New Zealand Department of Agriculture Annual Reports or in summary form at the annual Ruakura Farmers’ Conference, both of which are difficult to access in the new era of digital archiving and search engines.

Our objective was to collate the research undertaken (Table 1) into themes and discuss implications in the context of pasture-based dairy farming today, thereby ensuring the legacy of this great work and the information amassed are not lost. To this end, we have compiled a compendium of the results from farm system research projects undertaken at Number 2 Dairy research farm (No. 2 Dairy) Ruakura, Hamilton, New Zealand during the ∼60 years from the appointment of Campbell (Mac) McMeekan as the Research Station Superintendent in 1943, to the closure of No. 2 Dairy in 2004. Within this period, there were 20 full year or multiple year farmlet experiments, plus numerous component experiments that were designed to either complement the main experiment in progress or further investigate aspects of pasture and cow management. We have organised the experiments into logical themes, which allows both the presentation of the results and, also, a longitudinal assessment of progress in the different thematic areas. The outline of this review should allow the reader to read individual themes of interest without the need to read the manuscript in its entirety.

Table 1. Timing and brief description of, and references for major farm systems experiments at No. 2 Dairy between 1944 and 2004.

Period	Experiment #	Type of Experiment^a	Theme	Experiment name	References
1944–1957	1	F	1 & 3	Nutrition - Lifetime project	McMeekan and Walshe (1963)
1945–1957	2	F	1	Effects of grazing method – ‘controlled versus uncontrolled’ grazing	McMeekan (1947, 1954, 1956, 1960)
1957–1961	3	F	1	Interaction of grazing method and stocking rate	McMeekan and Walshe (1963)
1958–1961	4	C	1	Grazed pasture parameters	Campbell (1966a, 1966b, 1966c)
1961–1965	5	F	7	Genetic quality, stocking rate and pasture management	Carter (1964)
1965–1966	6	F	2	Redistribution of available feed	Campbell and Clayton (1966)
1966–1967	7	F	6	Meal feeding of dairy cows	Annual report (1966–67)
1967–1968	8	F	6	Feeding and management of dairy cows	Annual report (1967–68)
1968–1974	9	F	8	Integration of dairying and beef raising	Campbell et al. (1974)
1972–1976	10	F	7	Comparison of Jersey and Friesian-Jersey crossbreds	Campbell (1977)
1977–1979	11	F	6	Pasture and maize rotation in dairying	Campbell et al. (1978)
1979–1982	12	F	10	Effects on production of milksolids of using nitrogen to increase pasture growth	Bryant et al. (1981), Bryant (1983)
1982–1986	13	F	2	Autumn-winter management and dairy production	Bryant (1984, 1990), Bryant and L'Huillier (1986)
1986	14	C	7	Comparative performance of Jersey and Friesians in mid-lactation	L’Huiller et al. (1988)
1986–1987	15	C	2	Management of the dairy herd during a dry summer	Bryant and Macdonald (1987)
1986–1987	16	C^1	11	The effect of bovine somatotropin in a sustained release preparation (Somidobove) on milk production of cows at pasture in New Zealand	Davis et al. (1999)
1986–1988	17	F	4	New ryegrasses and Matua Prairie grass	Thom et al. (1990, 1992)
1989–1990	18	F	2	Early spring management	Bryant (1990)
1990–1993	19	F	7	Friesian vs Jersey comparison	Ahlborn and Bryant (1992), Penno (1998), Spaans et al. (2018)
1991–1993	20	F	4	Pasture cultivar comparison	Thom and Prestidge (1996), Thom and Bryant (1996)
1993–1998	21	F	6 & 10	1.75 t MS/ha	Penno et al. (1996); Macdonald (1999); Macdonald et al. (2017)
1996–1997	22	C	6	Balancing pasture and maize silage diets for dairy cows using urea, soybean meal or fishmeal	Macdonald et al. (1998)
1997–1998	23	C	2	Mowing pasture to improve milk production	Kolver et al. (1999)
1998–2001	24	F	5 & 11	Whole farm efficiency experiment	Macdonald et al. (2001, 2008a, 2010, 2011), Roche et al. (2016)
1998–2001	25	F	9	Effects of time on calving on dairy production	Auldist et al. (2002), Spaans et al. (2019)
2001–2004	26	F	7	Holstein-Friesian Strain experiment	Macdonald et al. (2005, 2007, 2008b), Lucy et al. (2009)

^a Signifies type of experiment. F = Farmlet, C = Component, C¹ = Component imposed on a Farmlet experiment.

Pioneering system level research

From the arrival of McMeekan as Superintendent of Ruakura in 1943 to its closure as a research farm in 2004, the research undertaken at No. 2 Dairy helped to shape a small rural nation dependent on agriculture for employment and agricultural exports for national wealth in a post-war world (McCloy 2014). Despite the distance from its principal export market (the United Kingdom), New Zealand became the largest exporter of dairy products in the world. This success was dependent on the nation’s ability to produce milk for butter, cheese, and, later, milk powder, at a price and quality standard that no other country could match. It was the requirement for low production costs that fuelled 60 years of scientific research.

In 1886 the Government purchased 137 acres (55 ha) and it was vested to the Waikato County Council in 1888 under the Waikato Agricultural College Model Farm Act, with the proviso that the land be used as an agricultural college and model farm. The Ruakura Experimental Station was established in 1901 with an additional 690 acres (279 ha) being purchased. In 1912, the station became the first Farm School of the Dominion and continued, as such, with short courses for ex-servicemen from World War 1, and then as a farm school for youth.

As with the proverb, ‘it’s an ill wind that blows nobody any good’, it was the severe outbreak of Facial Eczema in New Zealand in 1938 that was to set in motion a chain of events that would result in the development of No. 2 Dairy as the powerhouse of farm systems research that it became. Because of the extent of the disease in the North Island in 1937–38, the Government awarded a grant of £17,000 (∼$2.3 million today) to study the disease and, so, was born the Ruakura Research Station (Scott 1997), although the campus had been used for agricultural education for more than 20 years previously. The Dairies at Ruakura had been part of the farm school, but with McMeekan arriving in 1943, No. 2 Dairy was established as a research farm. Most experiments at No. 2 Dairy were undertaken over several years, but only two were spatially replicated (i.e. replication of individual farmlet treatments). This decision was made because the spatial replication was deemed cost prohibitive with the land area and number of cows used. McMeekan believed that the lack of spatial replication was overcome, to a large degree, by undertaking the experiment over several years; furthermore, he believed that it was more important to consider treatments over multiple years because of the large effect that weather could have on experimental outcomes. In his book ‘Grass to milk’, McMeekan (1960) stated that, in his opinion,

replication of complete ‘farm type’ experiments is impractical on many counts despite its scientific desirability, but by carrying all such studies on for several seasons, this weakness is overcome to a considerable degree by ‘replication in years’. In point of fact, it is my general opinion and experience that where a choice between the two lines of attack has to be made, it is more valuable to repeat any grassland experiment involving livestock over several years than to have several replications running in the one year.

He supported this by also stating that, ‘Pasture production is influenced far more in its total and seasonal supply by yearly and seasonal variations in climate, so that replication in years is vital’. The exposure of treatments to weather variability in undertaking experiments over multiple years also helped in interpreting results for farmers in differing locations and situations.

McMeekan’s view that the purity of treatment preservation should be maintained over several years was, however, also a limitation. This discipline resulted in a failure to adjust the experimental methods in some studies, even when it was evident that a treatment was vastly inferior to ones against which it was being compared. For example, the experiment investigating the effects of ‘Controlled’ versus ‘Uncontrolled’ grazing strategies (Theme 1 & 3) on the performance of rising 1 and 2 yr olds, demonstrated that the Uncontrolled grazing strategy resulted in excessive parasite burdens. Yet, despite the experiment progressing for 12 yrs, no investment was made in a management strategy to overcome this limitation for the ‘Uncontrolled’ grazing treatment. If an intervention had been investigated, the very large positive effect of the ‘Controlled’ grazing strategy on young stock growth and development may have been much less and farmers slower to transition to full rotational grazing better served by the research undertaken at No. 2 Dairy. Candler (1962) emphasised this point and in later yrs (e.g. in the ‘1.75 tonne Milksolids’ experiment), the researchers changed treatments after a couple of years to better represent the questions requiring an answer.

From the outset, No. 2 Dairy dealt with current issues facing farmers, demonstrated ‘best practice’, and ensured that farmers were engaged in the ongoing studies. Campbell Percy McMeekan was the first farm systems scientist engaged at Ruakura and his policy was that ‘research was useless unless it was applied’ (Scott 1997). This tradition of connectedness to the farmer continued in all of the grazing research undertaken at No. 2 Dairy, with a strong focus on applied research questions and experiment design, the testing of component-level changes in multi-year farm systems experiments, demonstration and farmer-partner farms, and the dissemination of experimental results directly by scientists. As a result, there were frequent visits to No. 2 Dairy by discussion groups. Furthermore, farmers attended the Ruakura Farmers’ Conference Field Days, which started in 1949, with two conference days and a third day that involved visits to the research farms. At their height, these field days attracted up to 8000 visitors (Scott 1997) and allowed interaction between researchers and farmers in the presentation of the latest experiment results.

In our presentation of the significant contribution of the research undertaken at No. 2 Dairy, we do not mean to undervalue or belittle the tremendous contributions of others. We, unreservedly, acknowledge that excellent research investigating soil science and fertiliser application, agronomy and grazing management, and animal genetics and nutrition has been undertaken in their component disciplines in many universities and research stations and in a multitude of countries; much of this work has been reported on by Roche et al. (2017a). However, the concept of multi-year, system-level research experiments was pioneered at No. 2 Dairy and has been synonymous with the development of the archetypal seasonal, spring-calving grazing dairy production system; so much so, in fact, that the system developed at No. 2 Dairy has become known globally as the New Zealand grazing system and, even, domestically, as the No. 2 or Ruakura system of farming (Roche et al. 2017a, 2017b).

The research farm

The No. 2 Dairy Research farm at Ruakura was located on the eastern outskirts of Hamilton city, New Zealand (37°47′ S, 175°19′E and approximately 40 m above sea level). The area was, at the time of establishment, considered as being of marginal quality for dairy production, as the land was on the edge of a peat swamp. In the book Grass to Milk, McMeekan (1960) describes No. 2 Dairy as being of poor natural fertility, with approximately 70% being reclaimed consolidated peat. The pastures were generally considered to be inferior, with a high proportion of Yorkshire Fog (Holcus lanatus), browntop (Agrostis capillaris), and weeds. The land was considered, by competent field observers, as typical of those producing 100 lb butter fat/acre (∼200kg MS/ha/yr; McMeekan 1960).

Depending on the experiment, between 30 and 65 ha were devoted to permanent grassland and divided into small paddocks (i.e. defined grazing area), each of which was contained by electrified galvanised wire and serviced by laneways (i.e. cow tracks), which facilitated cow and machinery movement, and water for drinking troughs (Roche et al. 2017b). Soil was a Te Rapa peaty silt loam, which is a Humic Aquic Haplorthod in soil taxonomy or a Humose Groundwater-Gley Podzol in the New Zealand classification.

Fertiliser was applied annually:

• up to 1964, 250 kg/ha of serpentine superphosphate/year (N:P:K:S 0:6.8:0:8.4);

• from 1964 to the mid 1980s, 500 kg/ha of 15% potassic superphosphate/year (N:P:K:S 0:7.7:7.5:9.4);

• from 1985, 600 kg/ha of superphosphate (N:P:K:S 0:9:0:11) and 100 kg/ha of muriate of potash (N:P:K:S 0:0:50:0) were applied annually;

• Nitrogenous (N) fertilisers were not applied until 1979.

Milksolids (MS; milk fat plus protein) production/ha without feed importation increased from approximately 580 kg (average for 1945–1955: McMeekan 1956), with young stock reared on the farm, to about 925 kg 35 years later (Bryant et al. 1981), with young stock reared off farm, and, consistently, to just over 1100 kg/ha two decades later, with the incorporation of 200 kg/ha of N from urea fertiliser (N:P:K:S 47:0:0:0; Macdonald et al. 2001, 2008a). With the use of supplementary feeds, MS production/ha was increased to 1750 kg MS/ha (Macdonald et al. 2017).

New Zealand’s original dairy cows were derived from Shorthorn cattle, used for draught work in the forests. Jersey cows arrived in the country in the 1860s. The original herd at No. 2 Dairy was predominantly Jersey and was one of the first to use the New Zealand Dairy Board's artificial breeding service. From 1974, the herd was inseminated with Holstein-Friesian semen and gradually transitioned from brown to black and white. A general description of Materials and Methods used over the 60 yr are described in Appendix 1.

No. 2 Dairy – proving the value of change

At the 1976 Annual General Meeting of the Waikato Section of the New Zealand Institute of Agricultural Science, the then under-secretary for Agriculture, the Right Hon. J. B. Bolger, questioned researchers and advisors on why there was a failure in the uptake of research results by farmers, as a major gap had opened between milk fat (milk fat) production/ha at No. 2 Dairy and the average Waikato dairy farm. His prompt led to an analysis by Archie Campbell that attempted to identify the major reasons for, and economic consequences of, the discrepancy between No. 2 Dairy production and national average yields (Campbell et al. 1977). They concluded that the difference between 2.1 cows/ha on commercial farms and 3.25 cows/ha on No. 2 Dairy accounted for 66% of the difference in net farm income; a further 25% was explained by a combination of cow genetic merit and management expertise. The authors acknowledged that the effects of the greater stocking rate (SR; cows/ha) were not independent of the ‘fine-tuning’ of management; they were, in fact, dependent on:

• having an early and concentrated calving;

• ‘drying off’ cows on body condition and culling early if required to ensure cows were in adequate body condition at the next calving;

• high genetic merit (quality) cows and use of artificial breeding;

• good heat detection and recording;

• grazing management;

• disease prevention.

Campbell et al. (1977) concluded that the South Auckland farmer did not utilise the feed that was grown because they did not have sufficient stock of good genetic merit to consume it at the right time and turn it into saleable products. Designing a ‘simple’ system to improve New Zealand dairy productivity became the goal of No. 2 Dairy’s research team.

Six decades of research

In Table 1, we present a catalogue of the experiments undertaken at No. 2 Dairy between the early 1940s and its closure in the early 2000s and we present these in a series of themes. During this time there were, 20 multi-year, system-level comparisons of:

• strategic grazing management policies (e.g. rotational grazing vs set-stocking), and how they interacted with SR and cow genetics;

• tactical grazing management policies to ensure high quality pasture was available for cows in early lactation, despite slow growth rates in winter, and to manage dry summers;

• replacement stock rearing strategies to maximise productivity;

• evaluating new ryegrass varieties and the role of prairie grass in rotational grazing systems;

• the role of SR in farm profitability and environmental sustainability;

• the practicality and economics of supplementing dairy cows with non-pasture feeds (i.e. supplementary feeding);

• the appropriateness of different cow breeds and genetic selection lines for grazing systems;

• integrating dairy and beef production;

• appropriateness of seasonal calving;

• consequences of N use in grazing dairy systems;

• some component experiments and value of well-kept records.

We have grouped experiments into these logical themes and discussed the implications of results for modern New Zealand production scenarios. In the final theme (Theme 11) we present research that did not fit into the other themes: the use of No. 2 Dairy in the development of a ‘Digital Twin’ of a pasture-based dairy system, an evaluation of a computer model used to optimise weekly management, some of the component studies that were superimposed on farmlet experiments, and a series of papers that were generated in the early 2000s that were enabled by the meticulous records kept from the experiments undertaken at No. 2 Dairy.

Theme 1: strategic grazing management convincing the world of the benefits of rotational grazing

In 1927, Stapledon reported that increasing the period of ‘rest’ between defoliations resulted in increased dry matter (DM) production (Stapledon 1927c) and these findings were, subsequently, confirmed by others (Woodman and Norman 1932). At the same time, however, Woodman et al. (1928, 1929, 1931) and Woodman and Norman (1932) reported that more frequent defoliations resulted in higher quality feed; they concluded, nevertheless, that pasture DM production was positively associated with digestible organic matter (OM) production, suggesting that there may be a suitable ‘rest’ period that accommodated the increased DM production and, yet, maintained feed quality. Further research in plant morphology, physiology, and functional genomics over successive decades was to provide reasons for the observed effect of ‘resting' pastures (summarised in Roche et al. 2017a).

At a farm system level, recognition of this ‘pasture resting’ effect on pasture OM production led to the hypothesis that animal production would be greater from intermittently grazed swards (i.e. ‘controlled’ or rotational grazing) than continuously grazed swards (i.e. ‘uncontrolled’ or set stocking). The ability to ‘rest’ pastures was facilitated by the development of an electric fence in the early 1930s and subsequent improvements to facilitate longer fences and more consistent voltage by Doug Phillips at Ruakura (Jones 1988). Farmers were already taking action, as subdivision of larger farm areas into smaller ‘paddocks’ (i.e. a grazing area) and utilisation of rotational grazing were reported as widespread in New Zealand in the 1930s (Holford 1937). But there were no research results to quantify the ‘size of the prize’ that could be attributed to the capital investment or to suggest in what way the system would be best managed. This was to change between 1945 and 1963, with two decades of research to understand the advantages and disadvantages of both approaches and how they interacted with cow genetics and SR.

The experiments – (Experiment 1, 2, 3). Lifetime project, grazing method, SR and genetics

Three experiments were undertaken in succession by McMeekan and Carter between 1945 and 1964 to compare farm system-level effects of ‘Rotational Grazing’ with ‘Set Stocking’ and the system-level factors that potentially interact with grazing strategy on the animal and farm responses. It was these experiments that were to lay the groundwork for what was to become known as the New Zealand system of dairy farming. A fourth experiment (Campbell 1966a, 1966b, 1966c) provided the measurements that explained the reasons for the increase in productivity from increasing SR and employing a controlled (rotational) grazing strategy.

Experimental methods

The first was a 12-year long experiment (1945–1957) entitled The Lifetime Project, (McMeekan 1956, 1960: McMeekan and Walshe 1963) although it is probably better known as the ‘Controlled versus Uncontrolled Grazing Experiment’; in this, McMeekan established two comparable farmlets, one for each of the grazing strategies. The comparison began with replacement heifer calves at weaning (∼9 weeks of age), wherein Jersey calves were managed in either a rotational grazing system, with surplus pasture conserved as hay for subsequent feeding during periods of deficit, or a set stocked grazing system, with minimal conservation (discussed in Theme 3). At the time of first calving, animals in each grazing strategy were randomly assigned to either a Rotational Grazing (Controlled) or Set-Stocking (Uncontrolled) treatment group (2.25 cows/ha) in a 2 × 2 factorial arrangement (two grazing strategies during rearing × two grazing strategies as lactating cows).

The second experiment was undertaken over four years and involved a 2 × 2 factorial arrangement, with Jersey cows either rotationally grazed or set stocked at either a low (2.35 cows/ha) or high (2.95 cows/ha) SR (McMeekan and Walshe 1963). A third experiment in this systems theme was undertaken over three full lactations and involved a 2 × 2 × 2 factorial arrangement of treatments. In this experiment, the interaction between grazing strategy (Rotational Grazing vs Set Stocking), SR (2.35 vs. 2.95 cows/ha), and cow genetics (High vs. Low Genetic Merit) were evaluated (Carter 1964). This experiment is discussed further in Theme 7. Finally, in an attempt to explain the results of the McMeekan ‘Controlled versus Uncontrolled’ grazing management experiment (as reported by McMeekan and Walshe 1963), Archie Campbell set up a component experiment that was a replica of the original experiment.

Results and discussion

Milk production and the interaction between grazing strategy and SR

Early comparisons of the effects of rotational grazing on milk production were positive, with McMeekan (1947) concluding that rotational grazing resulted in a 26% per year advantage in milk fat yield in lactating dairy cows when compared with set stocking. This early support for rotational grazing was challenged, however, by the long-term experiment investigating the interaction between grazing strategy for young stock and lactating animals (The Lifetime Project; McMeekan 1956, 1960); a 12-year experiment demonstrated that the effects of grazing strategy on production per cow and per ha were small: cows managed under rotational grazing produced, on average, only 13% more milk than their peers under the set-stocking management strategy: 160 and 141 kg milk fat/cow per yr for the rotational grazing and set-stocked cows, respectively; Table 2: equivalent of ∼291 and 256 kg MS, respectively. This result questioned whether the capital investment in fencing and tracks to facilitate rotational grazing was justified.

Table 2. Summary results from a series of experiments investigating the effect of grazing strategy (Rotational Grazing: RG; Set Stocking: SS) on milk and pasture production, and how these effects were modified by interacting system-level factors: stocking rate (SR) and cow genetic merit (GM)^a.

Experiment^b	Comparison	Year		Grazing management		Stocking rate		Genetic merit			Main effects P-value			Interaction P-value
				RG	SS	Low	High	High	Low	SED^c	MGT	SR	GM	MGT-SR	MGT-GM	SR-GM	MGT-SR-GM
1	Grazing strategy	1948–1957	FCM^c, kg/cow
			Fat, kg/cow	160.0	141.4					2.30	<0.001
2	Grazing strategy × SR	1957–1961	FCM^c, kg/cow	4085	3728	4142	3671			84.2	<0.01	<0.001		0.07
			Fat, kg/cow	186.4	169.4	188.6	167.2			3.96	<0.01	<0.001		<0.05
			Fat, kg/ha	494	449	443	493
			Lwt^c, kg	366	380	376	371			3.6	<0.01	0.22		0.17
			Pasture, kg/ha	12,738	11,124	11,637	12,226			1358	0.28	0.68		0.87
3	Grazing strategy × SR × GM	1961–1965	FCM^c, kg/cow
				173.4	150.6	167.5	156.5	175.4	148.6	2.08	<0.001	<0.001	<0.001	<0.01	0.2	0.79	0.38
			Fat, kg/ha	460	399	394	462

^a LSR – Low stocking rate = 2.35 cows/ha; HSR – High Stocking rate = 2.95 cows/ha; HGM – High Genetic merit; LGM – Low Genetic merit;

^b Experiment 1 compared the two grazing management strategies (RG versus SS) over a 12-year period. Experiment 2 compared cows in either RG or SS at 2 stocking rates (Low = 2.35 & High = 2.95 cows/ha). Exp 3 again compared grazing management & SR and included cows of High Genetic Merit, from the Ruakura herd or Low Genetic Merit cows sourced from commercial farms.

^c SED = Standard Error of the Difference; FCM = fat corrected milk; Lwt = cow live weight.

McMeekan (1960) hypothesised that the smaller than expected differences were, primarily, because more pasture was conserved as hay on the ‘Controlled farmlet’, a consequence of greater pasture production; however, as there was large wastage associated with this, at least some of the potential advantages of ‘Controlled grazing’ were lost. This lack of appreciable difference in milk production led McMeekan (1956) to conclude that greater SRs were central to capturing the benefit of rotational grazing and was ‘a potent weapon in obtaining high production of animal products from high-producing grassland’.

To test this hypothesis, a further 4 years were dedicated to investigating the interaction between grazing strategy and SR. In essence, the debate on the superiority of rotational (controlled) grazing over set stocking was settled by the publication of this work (McMeekan and Walshe 1963), which was the first of many joint New Zealand-Irish collaborations in applied dairy production science.

McMeekan and Walshe (1963) concluded that rotational grazing increased milk fat production per hectare by 7%, 6%, 22%, and 29% (in years 1–4, respectively) at the high SR. The low-stocked rotational grazing system produced 10% to 12% more milk fat/ha than the set-stocked system in years 3 and 4 of the experiment, a similar amount reported from the 12-year experiment (McMeekan 1956). The results confirmed McMeekan’s hypothesis on the interaction between grazing strategy and SR, with the most likely reason for the greater milk production being an increase in pasture growth and utilisation; Importantly pasture growth in the last 2 years of the experiment was recorded as 13.7 and 10.2 t DM/ha for the Controlled and Uncontrolled farmlets, respectively (Campbell 1966a). The Ruakura research was consistent with the results of Bryant et al. (1961a, 1961b) in Blacksburg, Virginia, who reported that rotational grazing increased the stock-carrying capacity by between 20% and 30% and increased milk production per hectare by 18%. Similarly, O’Sullivan (1984) reported a similar interaction between grazing strategy and SR in beef cattle. In his study, rotational grazing increased production/ha by 33% at the high SR, but by only 7.6% at the low SR.

The very high milk fat production/ha attained in the high stocked-rotational grazing treatment (McMeekan and Walshe 1963) is also noteworthy. In this treatment, cows produced >560 kg milk fat/ha, which is approximately the equivalent of 950 kg MS/ha. That is the equivalent of >13,000 kg of 4% fat-correct milk/ha and it was being produced in the late 1950s without nitrogen (N) fertiliser or purchased supplements. By comparison:

• Lewthwaite (1964) compared dairy production in the Waikato with Wisconsin. Extrapolating from the statistics he presented, Waikato herds in 1960 produced approximately 6500 kg 4% fat-corrected milk, 260–280 kg milk fat, and 440–480 kg MS/ha annually, or one-half of the average production at No. 2 Dairy;

• Wickham et al. (1978), in presenting a comparison of 1014 herds of cows by the New Zealand Dairy Board (unpublished), reported average milk fat production of 252–264 kg milk fat/ha annually; based on other publications, this would be equivalent to 5600–5900 kg 4% fat-corrected milk and 430–450 kg MS/ha annually.

The more than 100% greater milk fat production/ha at No. 2 Dairy in the 1950s was testimony to the successful system of dairy production developed by McMeekan over the previous 20 years. In fact, the production achieved by McMeekan and Walshe (1963) is only marginally inferior to what Macdonald et al. (2017) reported from No. 2 Dairy during the mid-1990s without use of N fertiliser: 14,000 kg 4% fat-corrected milk, 605 kg milk fat, and 1044 kg MS produced/ha; and what Waikato farmers produced in 2019/20 (Dairy Statistics): ∼13,750 kg of 4% fat-corrected milk, 598 kg milk fat, and 1063 kg MS/ha, while using 128 kg N/ha per year (Waikato Regional Council, unpublished). One hundred and twenty-eight kg N fertiliser would be expected to increase milk production by 1329, 47, and 99 kg of 4% fat-corrected milk, milk fat, and MS, respectively, suggesting that average milk production on Waikato dairy farms in 2019/20 has not yet exceeded that achieved at No. 2 dairy in the late 1950s. This was despite half a century of genetic progress and the animal husbandry and grazing management improvements that have been developed in the interim.

The interaction with genetics

Although the evidence in favour of both (1) the superiority of rotational grazing over set stocking and (2) the production advantages of increasing SR appeared beyond dispute, it was recognised, however, that ‘the animals involved could hardly be considered representative of the average dairy cow in the industry’ (Carter 1964); it was postulated that if animals of average productive ability had been used in the high SR system they ‘would have been dry by Christmas’ under the severe treatments imposed (McMeekan 1960).

The resulting experiment was to be the final nail in the coffin of set stocking in New Zealand dairy farming, with a 2 × 2 × 2 factorial arrangement established to test the interaction between grazing strategy (rotational grazing vs. set stocking), SR (2.35 cows/ha or 2.95 cows/ha), and cow genetic merit (Low or High). Half of the cows in the experiment were sourced from commercial dairy herds that used milk recording but had not used artificial insemination (AI) (i.e. low-genetic-merit cows; LGM). These cows were compared with cows from the research station, where cows had been bred by AI to ‘the best merit sires available’ (Carter 1964; high-genetic-merit cows; HGM) since the initial techniques were developed by James and co-workers at Ruakura (i.e. for ∼15 years).

Confirming the previous study (McMeekan and Walshe 1963), the negative effect of SR on milk fat yield/cow was less, and the positive effect on milk fat yield/hectare was greater, in the rotational grazing system compared with the set-stocked treatment, irrespective of cow genetic merit (Table 2).

Explaining the reasons for the increased productivity

Although pasture measurement techniques had been reported in NZ (e.g. Sears 1951), for the first time that we are aware, Archie Campbell reported on a new method for measuring pasture. The new pasture assessment method measured the DM yield of pasture both before and after grazing, providing an estimate of net pasture production (gains through new growth minus losses from all sources): this was termed ‘herbage accumulation’. It was argued that this parameter, rather than pasture DM production, provided a better representation of changes in feed DM supply to the grazing animal and, therein, could provide a better explanation for the treatment effects.

In a series of three papers on ‘Grazed pasture parameters’ published in the Journal of Agricultural Science in 1966 (Campbell 1966a, 1966b, 1966c), Campbell reported that ‘Controlled’ grazing at the High SR resulted in greater net pasture production in autumn and winter than other treatments; in comparison, Low SRs resulted in greater net pasture production in spring and summer, although the latter effects were not statistically significant. The annual increase in pasture growth in favour of the Rotational grazed system was 1610 kg DM/ha/yr (Campbell 1966a), an increase of approximately 15%, which would correspond to a 25–30% increase in production (i.e., gains above maintenance). Despite the greater pasture production under low SRs in spring and summer, the high SR in combination with ‘rotational’ grazing resulted in better pasture control (utilisation) in spring and summer than when any other treatment combination was imposed.

Campbell concluded that per hectare milk fat production was positively related to the percentage utilisation of available pasture DM, but negatively related to yield of available DM, highlighting that feed quality and, by connection, milk production declines with pasture DM yield. He also highlighted the importance of nutrition in early lactation on total milk production/cow; per cow milk fat yield was positively correlated with yield of available DM in the two months after calving (August and September). This is consistent with subsequent research by Bryant and Trigg (1979), Roche (2007), and Kay et al. (2013), where the negative effect of a feed restriction in early lactation was reported to last for several months after the restriction ended and, quantitatively, was two to three-times greater than what the anticipated response to supplements would be.

Campbell (1966a) highlighted the importance of lactation length on milk production per cow when he reported that yield of available DM in the penultimate month of lactation (April) was important to per cow milk fat yield. There was no relationship between available pasture DM yield and milk fat production in other months, probably because pasture growth/ha exceeded herd demand in these months (Roche et al. 2009b). Campbell (1966a) also noted that there was no consistent difference between so-called ‘Day’ (i.e. reserved for grazing during the day) and ‘Night’ (i.e. reserved for grazing during the night) paddocks in net pasture production under the ‘uncontrolled’ management system, with the conclusion that management system was as important as SR in increasing the utilisation of pasture and animal production from pasture.

Summary and implications

Although farmers had begun rotational grazing during the 1930s, there was disagreement among grazing enthusiasts as to the advantage of the strategy, other than for labour management, and there was little quantitative evidence to inform the discussion. The 20 years of ground-breaking experiments at Ruakura were to end the debate.

Under low SRs, there were only small differences between grazing strategies; certainly, too small to justify the necessary capital investment. But, with higher SRs, the advantage of rotational grazing increased because of increased net pasture accumulation (i.e. pasture growth and utilisation), greater heifer growth rates and reduced mortality from intestinal parasites (Discussed in Theme 3), and a consistent 20% to 30% improvement in milk fat production/ha, with only small changes in operating expenses, all of which justified the capital expense. These conclusions, coupled with the lack of interaction with cow type, as confirmed by similar effects in beef cattle, subsequently, paved the way for rotational grazing to become commonplace in New Zealand and throughout the world.

The approach to farm systems experimentation, with treatments replicated in time rather than spatially, defined the experimental method for systems research for the remainder of the twentieth century. In grazing research, arguably, no systems-level modification has come close to equalling the effects of rotational grazing under higher SRs for increasing the production and utilisation of pasture/ha and the conversion of that feed into milk. To quote McMeekan (1960): ‘No greater force exists, for good or evil, than the control of SR in grassland farming’.

Lessons learned in experimental design

The interaction between grazing strategy, SR and time identified by McMeekan and Walshe (1963) was a very important discovery; in their study, the superiority of controlled rotational grazing increased each year of the experiment and, in fact, only became material in Year 3 of the experiment. This result, which was confirmed in subsequent experiments, supported McMeekan’s premise that replication in time was more important than spatial replication when evaluating significant system-level changes (McMeekan 1960). Unfortunately, failure to recognise this has created a difficulty for publishing such studies in scientific journals.

The accepted statistical approach is for the herd to be the unit of statistical replication and not the animal (Bello et al. 2016). Although technically correct that animals within a herd are not truly independent of each other, especially in a competitive grazing environment, replication of herds within a grazing system and, in particular, in multi-year experiments is both impractical and resource prohibitive. Furthermore, a half century of farm systems research, which has defined the efficient system of farming in operation today, would have been lost to the annals of scientific discovery if such an edict had been enforced in the latter half of the twentieth century.

We agree with the lead author of the aforementioned scientific opinion (Bello et al. 2016) in her subsequent review (Bello and Renter 2018) when advocating ‘for a responsible practice of statistics in the animal sciences’. It is our position, however, that instead of being non-negotiable in the definition of independent replicates, reviewers of scientific manuscripts for farm system experiments should consider whether non-treatment variability has been minimised through appropriate experimental protocols, whether the results are reproducible and sensible, and whether the literature will be improved by the contribution made by the research work undertaken.

Theme 2: tactical grazing management: optimising rotational grazing

Background

In Theme 1, we reviewed the research undertaken at No. 2 that detailed the benefits of rotational grazing as a strategy for annual management of pastures; but, little research had been undertaken to determine how best to manage pasture seasonally (i.e. tactical management), particularly between the autumn-winter ‘lull’ and peak growth in spring. Grazing systems that maximise pasture harvested directly by the cow and minimise the dependency on non-pasture feeds (i.e. supplementary feeds) are designed to best synchronise pasture supply with herd demand profiles (Roche et al. 2017b).

Herd demand is regulated by physiological state (i.e. growth, pregnancy, lactation), while pasture supply is controlled by temperature, light, soil nutrient and pH status, and available moisture (Roche et al. 2009a, 2009c). In temperate regions and with temperate grass and legume species (i.e. ‘cool season’ or C3 species), pasture growth is minimum during winter, peaks during mid-spring and declines slowly through summer (moisture dependent) and autumn to the nadir in winter again (Figure 1). This supply is best matched by cows calving in mid-winter (Theme 7; Spaans et al. 2018), when the requirements of non-lactating cows coincide with low pasture growth and peak nutrient demands of lactation and reproduction occur during peak pasture growth and digestibility in spring (Roche et al. 2009b, 2009d).

Figure 1. Daily pasture growth rate (vertical bars; kg DM/ha/d) and per ha DMI of dairy cows at 2 stocking rates, 2.2 (

), and 4.3 (

) cows/ha. PSC = Planned start of calving, PSM = planned start of mating.

Although the synergy between temperate pasture supply and cow demand appears utopian, the profiles do not exactly match (Figure 1):

• winter pasture growth rates are not adequate to meet requirements, when the herd is at the SRs necessary to utilise the spring pasture flush and maximise the advantage of rotational grazing (Theme 1; Campbell 1966a);

• spring growth can exceed demand, particularly in regions with extended periods of low temperatures during winter (i.e. low pasture growth rates in winter), exacerbating the difference between pasture growth in winter and spring, and creating surpluses and, potentially, greater grazing residuals that must be corrected before summer; and

• summer and early autumn rainfall and, therefore, pasture growth is unpredictable.

Approaches adopted to overcome the differences in feed supply and demand profiles invariably involved deferring a block of ‘surplus’ pasture grown in autumn into winter and offering spring-conserved hay during April, May, and June. One such system entailed ‘setting aside’ 15% to 30% of the farm from the time that ‘autumn rains’ (McMeekan 1960) stimulated an increase in pasture growth post-summer (generally mid-March at No. 2 Dairy), deferring the grazing of these pastures until winter, and managing cows on the smaller area with hay conserved during the spring surplus. Deferring pasture ensured that an adequate mass of pasture was available at the start of calving, but the sward was, then, over 100-day regrowth: pasture pre-grazing mass was >5000 kg DM/ha (calculated from Hutton and Parker 1973). As a result, feed quality for early lactation cows was poor, with a large amount of dead material, described by McMeekan as ‘rotten bottom’ pasture (McMeekan 1960). Such feed had been identified as one of the four key factors limiting milk production by Professor Robert Boutflour of Harper-Adams College in 1928 (Boutflour 1928).

An alternative approach, and colloquially known as the Wallace system (Wallace 1958), after Lindsay Wallace, Ruakura’s first animal nutrition scientist, involved ‘setting aside’ two-thirds of the farm for half the time in the latter half of the ‘autumn flush’ (i.e. for 50–60 days), with the cows maintained on the other one third of the farm, using hay conserved during the spring surplus. By using such a large area, the ‘Wallace system’ reduced the length of time pasture was closed, increasing feed quality for early-lactating cows, but reducing the DM yield available/ha. It also involved considerable feeding pressure on the remaining one-third of the farm during late autumn-early winter. Further variations included closing a smaller part of the farm – McMeekan (1960) stated ‘that closure of 15% of the farm was quite satisfactory, as long as it was closed in the early stage of the autumn flush’ – or slowing the rotation during the late autumn-winter.

In all cases, autumn management was facilitated by using hay conserved during spring surplus. But, this was a fine balance; McMeekan and Walshe (1963) highlighted the importance of high SRs; but, high SRs increase the size of the winter feed deficit, while coincidentally reducing the amount of pasture available in spring for conservation. For example, it was common for 1/3 acre/cow (0.12 ha) to be reserved in spring to make 18–20 bales of hay/cow (∼350 kg DM/cow), which then facilitated 1/3 of an acre/cow to be deferred in autumn for winter pasture. This strategy, however, created the potential for a level of underfeeding of cows during peak lactation.

Various tactical strategies can be used to overcome any one of these potential limitations in the seasonal grazing system, but they are inter-dependent to varying degrees and must, therefore, be considered together. Furthermore, it is important to quantify advantages and disadvantages of each approach. To address this important question, 5 experiments were undertaken at No. 2 Dairy between 1965 and 1990, which were to underpin the grazing practices that, alongside high SRs and rotational grazing of high genetic merit cows, were to become synonymous with the New Zealand system of dairy farming. These are considered under the sub-themes:

1. Deferral of the spring surplus to the autumn deficit as hay (1965–66);

2. Designing an autumn, winter, spring management strategy to maximise production from grazed pasture (1982–1990);

3. Pasture topping to increase cow DM intake (DMI) and improve mid-season pasture quality (1997–98);

4. Managing through a dry summer (1987).

Conservation of spring surplus as hay to feed in the autumn feed supply deficit – The Experiment

The first experiment (Experiment 6, Table 1) was undertaken in 1965–66 and aimed to quantify the effect of transferring feed from the spring to the autumn-winter on the farm’s ability to transcend the winter feed deficit and the associated effects on spring milk production (Campbell and Clayton 1966).

On 1st June, 2 farmlets, each of 9.3 ha, were stocked with 40 cows that had been sourced from the Grazing strategy × SR × Genetics experiment (described in Theme 7). Cows were randomly allocated into either a HIGH or LOW feed allowance farmlet: each farmlet received 2.8 kg DM/cow/day of autumn-saved pasture for ∼60 to 70 days, with cows on one treatment farmlet receiving an additional 4 kg DM/cow/day of pasture hay (i.e. total DMI of 6.8 kg DM/cow/day; HIGH) and cows on the other treatment farmlet receiving 2 kg DM/cow/day (i.e. total DMI of 4.8 kg DM/cow/day; LOW). This equated to approximately 1.9% and 1.4% of their live weight (Lwt), respectively. From calving, both groups were grazed together on 18.5 ha and the cows were fully fed on pasture.

Results and discussion

The LOW cows were 27 kg Lwt lighter at calving than their contemporaries in the HIGH treatment and produced 196 kg milk and 12 kg fat less during that lactation. Although a formal way of assessing cow ‘condition’ was not yet established anywhere in the world (Roche et al. 2009e), these effects are consistent with much more recent experiments on the effect of calving body condition score (BCS) on early lactation milk production; Berry et al. (2006) reported that 27 kg Lwt ∼ 1 BCS unit (10-point scale, where 1 is emaciated and 10 obese) and, with Roche et al. (2007b), reported a 12 kg milk fat difference between cows calving at a BCS 4.0 compared with those calving at 5.0 (same 10-point scale). The results are also consistent with several other studies (summarised by Roche et al. 2009e).

The experiment highlighted the importance of feeding levels in the autumn or, probably, the importance of BCS at calving, but also identified that it was possible to manage the cows through autumn-winter with less supplements than previously thought. Campbell and Clayton (1966) described how, as SR is increased, a conflict arises between available grazing in spring and the need to make conservation. They concluded that cows’ intake in spring should not be restricted with the aim of making conservation, as they believed that with increasing SRs, there would need to be greater direct use of pasture by lactating cows.

Designing an autumn, winter, spring management strategy to maximise production from grazed pasture

Experimental methods

Arnold Bryant established a 4-year experiment (Experiment 13, Table 1) in 1982 to define the optimum pasture cover at calving and pasture allocation during winter and spring. One hundred and ninety-two cows and 52 ha were randomly allocated to eight equal farmlets (i.e. 24 cows on 6.5 ha; SR = 3.7 cows/ha). All herds had access to 150 kg DM silage/cow during the autumn-winter (∼3% of annual feed DM requirements), but no supplements were fed to lactating cows. There were 2 phases to this Experiment, Phase 1; 1982–1984 and Phase 2; 1984–1986, with the cows being re-randomised between the 2 Phases. Deciding that more data were needed to define exactly how and when the rotation should be changed post-calving, a one season experiment (Experiment 18) was established in 1989.

During the first two years (Phase 1; 1982–1984), different autumn pasture management strategies on each farmlet were established to achieve differing amounts of pasture on the farms at the Planned Start of Calving (PSC; 15th July). A range of autumn rotation lengths (Table 3) were established to determine the effect of autumn rotation length on pasture availability and the effect of winter rotation length on both the long-term and short-term pasture supply/cow. Winter-spring herd feed demand was intensified by having calving confined to 7 weeks.

Table 3. No. 2 dairy experiment design for 1982/83 & 1983/84 for autumn-winter (April to July), (Experiment 13, Phase 1: Adapted from Bryant and L'Huillier 1986).

	Nominal rotation* (days)		Actual average rotation (days)
Farmlet	April–May	June–July	April–July	Aug–Nov
1	128	96	99	27
2		64	84	26
3	96	96	84	26
4		64	72	25
5	64	96	67	24
6		64	55	25
7	32	96	60	24
8		64	46	25

*A rotation of 100 days means that 1/100th of the farm was grazed each day.

In Phase 1, at one extreme, the rotation during most of the autumn was 128 days in autumn and 96 days in winter (Farmlet 1); at the other extreme, the farmlet was managed using a 32-d rotation during autumn and a 64-d rotation in winter (Farmlet 8). Once calving had commenced, the rotations on all farmlets were gradually shortened so that, by the end of August, all farmlets were on a 16-d rotation. This 16-d rotation was maintained through September and early October and increased to 32-d by the end of October (Table 3).

During the subsequent two years (Phase 2; 1984–1986), 3 spring-rotation strategies were imposed on each of the three autumn management protocols, allocating available feed at different rates: from very short rotations (11 days) to ensure cows were well fed in very early lactation, to long rotations (36 days), in which pasture was rationed to allow adequate recovery time between grazing events (Table 4). The winter rotations were gradually shortened post-calving so that each farmlet attained its nominated spring rotation by mid-August. The nominated (16- or 8-d rotations) were maintained until October when they were all gradually increased to a 32-d rotation by the end of November.

Table 4. Experimental design at No. 2 dairy for 1984/85 and 1985/86. (Experiment 13, Phase 2: adapted from Bryant and L'Huillier 1986).

	Nominal rotation* (days)		Actual average rotation (days)
Farmlet	April–July	August–November	April–July	August–November
1	128	16	107	23
2		32	107	36
3	96	16	77	22
4		32	77	34
5		8	77	11
6	32/64	16	46	21
7		32	46	34
8		8	46	11

*A rotation of 100 days means that 1/100th of the farm was grazed each day.

The last of these experiments (Experiment 18; 1989/90) involved 4 early lactation grazing strategy treatments replicated twice (i.e. 8 farmlets). Each farmlet was randomly allocated 6.5 ha and 21 Holstein-Friesian cows; (i.e. SR = 3.7 cows/ha; Table 5). The farmlets were managed similarly pre-calving, with the four treatments imposed post-calving. Grazing strategies involved varying rotation length between 14 and 56 days immediately post-calving (i.e. farmlets were grazed between 1.0 and 4.0 times in the 56 days from the PSC; Table 5). The aim was to further investigate the interaction between pasture cover at calving and early lactation feeding levels on subsequent production and pasture growth.

Table 5. Effect of differing rotation lengths on farm cover (kg DM/ha) and subsequent milksolids production (Experiment 18; 1989/90).

Treatment	1	2	3	4
Period	*Rotation length (Days)
21/7 – 3/8	100	58	41	33
4/8 – 17/8	60	37	26	20
18/8 – 31/8	52	35	21	14
1/9 – 14/9	42	28	14	7
No. of rotations (21/7 – 14/9, 56 days)	1.0	1.5	2.5	4.1
Farm cover at 18 July kg DM/ha	2000	1900	1900	1900
Farm cover at 22 Sept kg DM/ha	1600	1500	1200	1100
MS/cow to 1 October	78	88	83	81
MS/cow to 9 February	248	267	257	246

*A rotation of 100 days means that 1/100th of the farm was grazed each day.

Results and discussion

In Phase 1 of Experiment 13, the proportion of the farm grazed between 1st April and 30th June was negatively associated with the amount of pasture for grazing on the farm in July (Figure 2). In other words, autumn rotation length (i.e. the number of days between the autumn and winter grazing events) was positively associated with the amount of pasture on the farm in July. Importantly, however, the effect was greater than the difference in amount of pasture removed in autumn (i.e. for every 1 t DM pasture removed in autumn, there was more than a 1 t DM difference in farm cover in July). This is consistent with what we know about the profile of growth of temperate pasture species.

Figure 2. Relationship between proportion of farm grazed during April, May and June, and the amount of pasture on the farm in July (Experiment 13, Phase 1).

Note: A rotation of 100 days means that 1/100th of the farm was grazed each day.

The sigmoidal nature of regrowth in temperate pasture species results in a positive relationship between pasture DM yield and growth rates, at least until light interception in the sward is maximised (Brougham 1957; Voisin 1959; Fulkerson and Donaghy 2001). This means that the greater the pasture mass, the faster it is growing, until maximum light interception. Some of the additional pasture cover at calving in the longer rotation farmlets was because the average daily pasture growth rate during winter was greater on these farmlets because of the greater average pasture cover (Figure 2); for every 100 kg DM increase in pasture cover, growth rate increased 7 kg DM/day. These results are similar to those reported in Ireland by Carton et al. (1988) and Roche et al. (1996), the latter team concluding that every 1 kg DM removed during autumn grazing resulted in a ∼1.25 kg DM reduction in available pasture in early spring due to a lower tillering rate and a lower leaf area index in the sward that was grazed later in autumn.

The practical consequences of utilising the sigmoidal nature of temperate pasture growth are significant. An increase in pasture cover at PSC from, for example, 2100–2200 kg DM/ha, provides enough additional feed for a dry cow or more than half the requirements of a milking cow during the first month in milk. This ‘grass grows grass’ phenomenon, as it was to become colloquially known, was an important ‘discovery’ in the development of management strategies to facilitate higher SRs: ‘the slower the rotation during winter, and the earlier in autumn this is established, the more will be the amount of feed on the farm at the start of calving’ (Bryant 1990).

Importantly the concept of ‘grass grows grass’ refers to the whole farm, not individual paddocks. In the late 1990s the benefits of leaving higher post-grazing residuals in winter on subsequent pasture growth rates was tested in early and late July at No. 2 Dairy (Macdonald 1998), with pastures that were grazed for 2, 4, 8 or 24 h and, after these times, were protected from further grazing using cages. The regrowth resulting from different post-grazing residuals was measured over the next 52 days and ranged from 864 to 1773 kg DM/ha. Even though the areas grazed for 24 h were grazed much harder than those grazed for 2 h, the accumulated growth after 52 days was 38% greater. The area that had the least severe grazing had the lowest subsequent regrowth, although it maintained the highest pasture cover at the next grazing. Leaves have a limited lifespan and if left ungrazed they die before the next grazing. In more laxly grazed pastures net DM accumulation is reduced because the amount of pasture lost to decay is greater.

More extensive work in 2006 (Lee et al. 2008) identified that a very severe grazing (<3 cm) reduces pasture regrowth and can reduce plant persistence due to removal of plant energy reserves, while lax grazing (>6 cm) reduces the herbage quality of the subsequent pasture on offer. So, while regrowth may not be affected in the short term, repeated lax grazing events will result in a less productive and less persistent pasture (Lee et al. 2008).

The additional feed produced through longer autumn rotations increased the farm’s early season milk production. A rotation of 80–120 days between April/May and July resulted in 10% to 15% more milk fat production before Christmas, when compared with rotation lengths of less than 50 days. Although the fact that longer rest periods result in greater DM accumulation in pastures had been well established more than half a century prior to the No. 2 experiment (Stapledon 1927a, 1927b, 1927c), this was the first time that the biophysical relationship between inter-grazing duration and pasture DM yield was quantified in dairy cow performance in a farm system experiment.

In Phase 2 of Experiment 13, the interaction between pasture cover at calving (i.e. autumn grazing management) and spring grazing management was quantified (Figure 3). Using a fast rotation immediately after calving (average 11-d from August to November) increased the amount of feed available for the cows during very early lactation, but the resultant decline in pasture cover resulted in reduced pasture growth rates, on average, across the farm. When compared with the farmlet that maintained a slow rotation (average 34-d from August to November) through winter, the difference in pasture cover created during winter remained until November. Importantly, the farmlet with the fast rotation during winter produced 650 kg DM/cow less pasture during this period.

Figure 3. The effect of rotation length in a winter-spring feed deficit on the long-term pasture cover (i.e. feed availability; Experiment 13 Phase 2). Treatments were: a slow rotation prior to calving and post-planned start of calving (

), a fast rotation prior to calving and slow post-calving (

), fast rotation prior to calving and post-calving (

).

The experiment was key in understanding the importance of controlled rotation lengths during winter-spring, with a fast rotation in winter increasing the amount of feed available/cow in very early lactation, but reducing feed available/cow for the next 120 days. Typically, rotations that are too fast immediately after calving result in large September feed deficits, which persist through October (Figure 3) and long into the seasonal breeding season. An additional and important point identified in these experiments were that short rotations were more detrimental to future pasture growth rates than grazing intensity (i.e. post-grazing residual) in early spring.

The development of decision support systems to optimise pasture management

The results of these experiments highlighted the interconnectedness of autumn, winter, and spring pasture management and the implications of this on feed availability for the first 150 days in milk and, therefore, the lactation:

• Campbell reported on the importance of feeding cows in autumn and ‘cow condition’ on early lactation milk production, although this was before the development of an official BCS scale;

• Bryant and L'Huillier (1986) were to complicate this salient discovery, however, when they published that long rotations in autumn (i.e. a low pasture allowance/cow) increased pasture cover at calving and feed available/cow in early lactation, highlighting the interaction between autumn and spring requirements; and

• more feed was available/cow during spring if available pasture was rationed during winter and early spring, increasing milk production per cow.

The results highlighted the need for a ‘balance’ between pasture allocation for the cow in autumn to ensure the cow was of appropriate BCS at calving, pasture cover at calving to ensure sufficient amounts of available feed for a high stocking rate and compact calving, and pasture allocation to the cow in the spring to ensure total lactation yield is optimised and not just early lactation yield. To achieve this ‘balance’, the experiments were used to develop two Decision Support Systems (DSS) that were to revolutionise autumn-winter-spring grazing management through their simplicity.

The first DSS was colloquially known as the Spring Rotation Planner (SRP) and became, arguably, the most important DSS in the grazing farmer’s arsenal. Bryant’s experimental results highlighted the importance of rationing available pasture once the cows calve, irrespective of actual farm pasture cover, until spring weather facilitates pasture growth equivalent to herd demand. Grazing more quickly will reduce pasture growth and increase the likelihood and size of a feed deficit and the duration of low feed availability. It was, therefore, important not to allow too fast a rotation in winter, as any feed deficits were magnified.

The SRP (Figure 4) utilises the fact that feed demand of a pregnant non-lactating cow is less than a calved lactating cow. As a result, the DSS is designed to ration the amount of feed offered each day from the start of calving and through early lactation until pasture growth rate has increased and is equivalent to herd demand (approximately 60 days after planned start of calving; PSC). A small area of the farm is offered initially, as the entire herd is non-lactating and demand is low. This ensures pasture cover remains high and pasture growth rate through winter is maximised. As the season advances, cows transition from non-lactating to lactating, herd feed demand increases, and the SRP offers a greater area of the farm each day. The increase in area allocated provides more feed, but controls the allocation to ensure pasture cover is maintained in an optimum range for pasture growth.

Figure 4. The spring rotation planner dictates how much area should be allocated each day from calving to 'balance day' (i.e., when pasture growth is equal to herd demand). The farmer must ration this area between dry and lactating cows. The area allocated increases with time, matching the increasing number of cows calved and the greater dry matter intake of lactating cows (Developed from Experiment 13 & 15). The provided example is a 100 ha farm with a planned start of calving in early July. Rotation length (

), Area grazed/day (

).

Following the development of the SRP, the autumn rotation planner (ARP; Figure 5) was introduced for use in early autumn to ensure pasture cover at calving is optimised and pasture during autumn is optimally and simply rationed to the herd. The ARP is, in effect, the reverse of the SRP and was designed to gradually increase rotation length (reducing the area grazed on each successive week) from a particular week in autumn to approximately 4 weeks before PSC (early-winter). This ensured that the area being grazed daily declined gradually, and, by the inverse association, the length of the inter-grazing period (i.e. the rotation) increased. In reverse to the SRP, the ARP reduces feed allocation/day as the herd transitions from lactating to non-lactating and demand for feed declines. By increasing rotation length, average pasture cover increases and, with it, average growth rates.

Figure 5. The autumn rotation planner developed from autumn-winter grazing management experiment that dictates how much area can be allocated each day. This area is rationed between dry and lactating cows to ensure dry cows are allocated enough for condition score gain, while the milkers are allocated enough for maintenance and milk production. The provided example is for a 100 ha farm, with a planned start of calving (PSC) in early July. Rotation length (

), Area grazed/day (

).

Both the SRP and ARP are extraordinary in their simplicity. Proper use maximises autumn-winter growth rates, irrespective of the prevailing weather, so that non-lactating cows gain BCS, thereby avoiding the negative effects of an autumn restriction reported by Campbell and Clayton (1966), and rations feed during winter to ensure pasture available for cows in spring is maximised. The system evolved to maximise pasture production and utilisation through discipline in grazing management; in effect, by utilising the ARP and the SRP, the cows could ‘have their pasture, and eat it too’. Because of their operational simplicity and, yet the unconscious discipline in optimal pasture management, these two DSS were instrumental in sustainable management of high SR systems through limited feed supply in winter.

Managing a dry summer

Background

In a nutrient and moisture unrestricted environment, the natural profile of pasture growth in temperate ‘cool-season’ pastures, like perennial ryegrass-white clover (Lolium perenne L- Trifolium repens L. i.e. C3 pasture species), is to grow slowly through winter, peak in mid-spring (i.e. October-November in the Southern Hemisphere), and decline through summer and autumn (Figure 1; Macdonald et al. 2007; Roche et al. 2009b, 2009c). This natural pattern of growth is, however, affected by temperature and moisture deficits. An analysis of 40 years of data from DairyNZ research farms indicated that variation in weekly pasture DM yield is twice as great during summer-autumn as it is during winter-spring (i.e. coefficient of variation of ∼20% vs 10%; unpublished data).

Because the majority of dairy production in New Zealand is from what is called ‘dry land’ or ‘unirrigated’ farming, pasture growth and mid- and late-season milk production are dependent on summer rainfall, making it less predictable. Although conserved spring pasture or purchased supplements can be used to maintain milk production in such scenarios, these feeds are more expensive than pasture and may be more useful when pasture growth slows in autumn-winter.

With the recommendation emanating from the early research at No. 2 Dairy to use rotational grazing and increase SR, the summer period became a time of greater risk for feed supply. It was generally considered appropriate to maintain the late-spring rotation (at No. 2 Dairy, this was generally ∼30 d); importantly, no research had been undertaken to determine what the most appropriate rotation length was for summer or what was the most appropriate course of action with the arrival of the ‘autumn rains’.

The 1986/87 dairying season was notable for the dry summer. Milksolids processed in South Auckland for December and January were 9% and 22% less than the same months of the previous season. Rainfall at Ruakura during November and December was 60% less than the average for the previous 10 years, with only 90 mm falling between 28 October and 19 January. There was uncertainty about the management that most quickly improved levels of feeding and milk yield after good rainfall and there was no information to help decide what rotation to adopt and the extent to which this is affected by whether supplements are used. A rainfall event at Ruakura of 122 mm during mid- to late January provided the opportunity to test some hypotheses at No. 2 Dairy (Experiment 15).

Experimental methods

This was a short-term (8-wk) component experiment established after a dry late spring-early summer, which was followed by 122 mm rain in mid-January. Four farmlets were randomly allocated 8.1 ha and 28 cows (SR = 3.46 cows/ha), and a 2 × 2 factorial arrangement of treatments was established: farmlets were randomly assigned to one of two rotation lengths: 10 or 40 days, with or without 4 kg DM/cow/day pasture silage as a supplement during the first 4-wk. The aim was to establish the most appropriate rotation length and the value of supplementation, following an extended dry summer.

Results and discussion

At the start of the experiment, the average farm cover for all farmlets was 2200 kg DM/ha and at the end of the 8 weeks it was 2500 and 2250 kg DM/ha for the 40- and 10-d rotation farmlets, respectively. The cows on the 40-d rotation produced less milk during the first four weeks of the experiment, as they had only one quarter of the ‘daily’ pasture area of the short-rotation herd. However, they produced more milk during the subsequent four weeks. By week 8, the herd on the 40-d rotation was producing 0.12 kg milk fat/cow/day (∼0.17 kg MS) more than the herd that had completed 4 rotations during the same period (i.e. 10-d rotation). Although the total (8 wk) MS production was similar, 26.5 & 26.7 kg MS for the 10-d & 40-d rotation groups, respectively, the better performance at the end of the experiment, together with the greater amounts of feed on the farm demonstrated the value of establishing a longer rotation following the drought-breaking rains.

Use of available supplementary feed following the rain resulted in greater post-grazing pasture residual height and mass (3.08 vs 2.75 cm and 2000 vs 1860 kg DM/ha) during the four weeks of supplementation, irrespective of rotation length. There was no difference, however, in the amount of feed on the farm at the end of the 8-wk experiment. Supplemented herds produced more milk (19g milk fat/kg DM supplement during the supplementary feeding period and 6g milk fat/kg DM supplement once supplementary feeding ceased; 25 g milk fat/kg DM supplement or ~35 g MS/kg DM supplement). Supplemented cows were also heavier at the end of the 4-wk feeding period.

In conclusion, the experiment provided evidence that the grazing rotation should be lengthened rather than shortened when rain falls following a dry summer, irrespective of whether supplement is available. Supplement used after significant rain events in summer (i.e. those that will increase growth rates) increased pasture residuals, but did not affect post-grazing pasture regrowth in this experiment. A financial analysis of the results concluded that, with the milk and supplement price at the time, the total response to supplementary feeding was insufficient to warrant their use. This conclusion, however, is sensitive to prevailing milk and supplement prices.

Optimising pasture quality (Experiment 23)

Background

In the 1990s, the focus on improving milk production/cow led to an increasing interest in decreasing SR. Furthermore, Penno (1998) reported that the SR required to optimise farm profitability (i.e. economic farm surplus; EFS) was generally lower than that required to maximise milk production/ha. But, reducing SR would be expected to reduce pasture utilisation/ha, particularly in spring, which was recognised as important for maintaining high pasture quality.

Milking fewer cows, while maintaining or improving profitability, would require cows to consume more pasture every day. Earlier work had identified that intake in spring may be limited by the low DM content of high-quality pasture (John and Ulyatt 1987) and by the time and energy required to harvest pasture by grazing (Parsons et al. 1994). It was considered that while mowing residual pasture after grazing by cows that have been offered a high allowance can maintain pasture quality through the late spring (Holmes and Hoogendoorn 1983; Hughes 1983), mowing and wilting pasture before grazing in spring and early summer might overcome the low DM content and harvesting efficiency factors limiting intake and improve production at the time and, subsequent pasture quality and production.

Experimental methods

The experimental aim was to investigate the effects of ‘normal’ pasture management compared with either mowing pasture pre- or post-grazing during spring and summer. A short-term component experiment (Experiment 23) during spring 1997 and summer 1997/98 (165-d) was established by Eric Kolver that had areas of pasture subjected to three treatments: (i) a ‘Control’ that was not mown; (ii) pasture mown approximately 24 h before each grazing; and (iii) pasture ‘topped’ (mown) approximately 24 h after each grazing.

The area used for the experiment were non-contiguous and were balanced for initial average herbage mass, soil type, and pasture growth potential (18 t DM/ha/year). Treatment areas were grazed by three Holstein-Friesian herds (stocking rate: 3.0 cows/ha; 72 kg Lwt/t DM grown). Cows (n = 18) were blocked by age, breeding worth (BW), milk production, and Lwt, and were randomly allocated to treatments according to a crossover design where cows went from one treatment to another after a washout period. The pastures were grazed at intervals of 28 days during spring (September, October, November 1997) and summer (December 1997, January, and February 1998), resulting in six experimental periods of 14 days.

Results and discussion

Either mowing pasture before grazing or topping pasture after grazing in spring, reduced pasture production by >30% (71 vs 54 kg DM/ha/day) during the subsequent growth period, when compared with an untreated ‘Control’. Pasture growth in summer was increased by 20% (42 vs 48 kg DM/ha/day) in the mechanically treated pastures; but, overall, mowing and topping reduced pasture DM growth/ha by 4% and 10% for mowing and topped pasture, respectively. Spring pasture quality was not affected by treatment, but summer pasture quality was greater in the topped and pre-mowing treatment pastures; metabolisable energy (ME) content of summer pastures that had been mown before grazing (+0.2 MJ ME/kg DM) or topped after grazing (+0.6 MJ ME/kg DM) were greater than the untreated ‘Control’ pastures.

During spring, mowing or topping had inconsistent effects on Lwt change: change in Lwt was +0.30, −0.47 and +0.65 kg/cow/day for the ‘Control’, Topped and Mowed treatment cows, respectively. But, in summer, cow Lwt responded similarly to MS production, as both mowing and topping treatments resulted in greater Lwt gain (0.76 kg/cow/day) when compared with the ‘Control’.

Pre-grazing mowing increased average DM content of the wilted pasture by 7.5% units when compared with the ‘Control’. However, DMI was reduced by 2.4 kg DM/cow/d in November. As a result, MS production/cow was 11% (0.11 kg MS/cow/d; total kg/cow) less in both mechanical intervention treatments during October. However, the effect of treatment was reversed in summer, with MS production/cow 12% greater (0.13 kg MS/cow/d; total kg/cow), presumably due to the improvement in pasture quality following mechanical correction of post-grazing residuals. However, although both mowing and topping increased the quality of pasture and the yield of MS per cow in summer, the overall benefits for MS production/ha were small or negative. The topping treatment did increase total MS/cow produced during the six 14-d experimental periods (by 4.4 kg MS/cow, or 4.6%), but both pre-mowing and topping resulted in a decline in pasture DM production and, as a result, the number of cows the farm could sustainably feed. An additional 864 kg of DM/ha was removed from the ‘Control’ pastures by extra cows; this is the equivalent of >60 kg MS/ha, if utilised by a higher SR (or via conservation).

The results were conclusive; pre-mowing of pasture or post-grazing topping increased pasture quality in subsequent grazing events and associated milk production/cow, but reduced the amount of pasture produced and utilised/ha, and, therefore, reduced the potential milk production/ha. These results were consistent with previous research by Bryant (1982) and Stakelum and Dillon (1991) in Ireland. An experiment in Tasmania (Irvine et al. 2010) confirmed lower MS production/ha on pre-mowed pastures and identified a further challenge, that mowing in undulating areas resulted in pasture being contaminated with soil and refused by the cow.

Despite clear and consistent results in New Zealand and internationally, there continued to be interest in the perceived benefits of pre-grazing mowing, and it came to be advocated by farmers and advisors, particularly at the Lincoln University Demonstration farm in Canterbury, in New Zealand’s South Island. Further research in the South Island (Lincoln University Research Farm; Kay et al. 2018) identified that mowing prior to grazing reduced pasture density and pasture harvested, including the silage conserved, and required a greater amount of pasture silage to be fed. Importantly, mowing before grazing did not improve MS production, or BCS.

Summary and implications

The recommendation to increase SR to optimise farm productivity created a risk of exacerbating shortfalls in available feed, especially during winter-spring, but also during summer. At the same time, the ability to ‘rest’ pastures in between grazing events, under rotational grazing systems, created an ability to ‘ration’ pasture that would have been more difficult under set stocking.

The ‘tactical pasture management’ research undertaken at No. 2 Dairy over a 20-year period established the need for cows to be optimally conditioned at calving, even before the establishment of recognised BCS scoring systems (Roche et al. 2004), but also the competing need to reduce the daily supply of pasture/cow in autumn, so that an adequate farm cover was established at the beginning of winter. This was to create a feature of farming systems unique to New Zealand: using lactation length to manage feed demand. Recognition that cows needed to be at a BCS 5.0 (although not defined as such in the 1960s), while creating a pasture cover of 2100–2300 kg DM/ha (depending on location) resulted in a system in which thin cows were dried off early (to provide time to get to the optimum BCS, while reducing per ha feed demand) and culled cows that were surplus to need in the next lactation to reduce feed demand/ha. Combined, these management strategies shortened the average lactation length and milk yield of New Zealand cows relative to their international peers, but they were critical to successfully establishing the next lactation. The ARP facilitated this by creating a discipline around how much farm area and pasture/day were available, stimulating farmers to consider which cows should be culled, dried, or allowed to continue milking.

‘Rationing’ this pasture through winter and spring may appear logical, but quantifying the value of this, even if the farm pasture cover was lower than ideal, was critical to the system success: a fast rotation through winter on a low cover, reduced pasture growth/cow by 650 kg DM before Christmas when compared with a slow rotation. This knowledge and the SRP created a winter-spring discipline and surety of management success that enabled farmers to ‘see the light at the end of the tunnel’, even in the wettest and coldest of July-August days. Critics of the system labelled it ‘Controlled Starvation’, referring to the limit-feeding of cows during the first two months of lactation, when most dairy systems internationally are trying to stimulate greater energy and protein intake. This criticism fails to recognise:

1. That the SRP is a pasture management DSS and not a DSS for feeding cows. Diligence to the SRP will ensure maximum pasture growth during winter-early spring and that any pasture restriction is ‘Controlled’. The SRP does not preclude the use of supplementary feeds, if necessary and appropriately costed;

2. That production animals adjust production to match their intake of nutrients, particularly in early lactation. This means that if one cow consumes less than another, all other things being equal, it will produce less milk, but it will not lose more BCS (Roche 2007; Roche et al. 2007a, 2009e); the cow’s metabolism adjusts to its intake of nutrients. Limit-feeding, therefore, should not be considered ‘starvation’;

3. That the response to supplementary feeds is least in spring in pasture-based system (see; Stockdale 2000; Roche 2017) and that the use of supplementary feeds is most efficient and profitable if employed in the autumn.

With hindsight, these results appear common-sense. However, the ability to balance these competing system needs was a limitation to increasing SR and maximising land productivity. Optimising this management required a knowledge of agronomy and animal nutrition and physiology. But the need to understand these scientific disciplines was removed in the development of the SRP and APP DSS, which systemised the results of the No. 2 Dairy experiments. The fact that they are still used by most farmers today, despite system-level changes to include supplementary feeding, is testimony to their effectiveness and simplicity. Their modification and use in other pasture-based systems is testimony to their brilliance and the genius of Arnold Bryant and co-workers; after all, ‘imitation is the greatest form of flattery’.

Theme 3: rearing replacement stock – quantifying the cost benefit of management for greater growth rate in replacement heifers

Background

The pasture-based system is designed to synergise the natural patterns of pasture growth with the profiles of nutrient requirements of production animals. The demand for feed is dependent on both animal numbers (i.e. SR) and animal physiological state (i.e. growing, pregnant, lactating, non-lactating, gaining or losing condition, etc.). Growth, BCS, and lactation profiles, therefore, can vary more than would be expected in housed systems (Roche et al. 2007a, 2017b), as the pasture system is more prone to variation in feed supply, while recognising the sentient nature of the animals under the farmer’s care. This can be particularly important in the growth of replacement stock, wherein the lack of a daily production signal to help ‘pivot’ the system to provide supplementary feed is not as obvious or, often, as logistically manageable.

Dairy heifer growth rate and Lwt at first calving have been regarded as important benchmarks in farm management for at least a century in science publications (McCandlish 1922) and, certainly, prior to the establishment of No. 2 Dairy. McCandlish (1922) reported that ‘stunted and generally undernourished’ heifers ‘cannot be expected to do their best work when they come into the producing herd’. In fact, he said, ‘the improper raising of heifers is a factor of great influence in causing the low average production of milk and butterfat in many of the dairy herds of today’. Several studies followed to investigate the role of management (e.g. weaning age) and nutritional factors in heifer development, reproduction, and, subsequent, milk production (Salmon and Eaton 1925; Bender and Bartlett 1929). However, these studies either investigated the supplementation of grain-based diets or investigated practical heifer-rearing strategies using locally sourced feeds and, as such, were of little use in evaluating the importance of heifer growth rates under pasture-based systems. Considering the reported reduction in productivity of poorly grown heifers, it was important to evaluate the role of early life nutrition in pasture-based grazing systems (Experiment 1) and it was to this challenge, McMeekan and, more recently, Macdonald and co-workers, put their considerable efforts.

Experimental methods

As previously outlined in Theme 1, McMeekan (1956, 1960) established a 12-year long experiment entitled ‘The Lifetime Project’ in 1945, (McMeekan 1956, 1960), in which two comparable farmlets were established (Experiment 1). The initial phase of this project involved a comparison of Rotational Grazing vs. Uncontrolled Grazing (i.e. set stocking) strategies and the effects on growth rate and performance of replacement heifer calves from weaning (∼9 weeks of age) to first calving. Jersey calves were managed in either a rotational grazing system, with surplus pasture conserved as hay for subsequent feeding during periods of deficit, or a ‘Set-stocked’ grazing system, which allowed minimal conservation. At calving, half of the heifers from each treatment group were randomly assigned to either ‘Rotational Grazing’ or ‘Set stocking’ as lactating cows in a 2x2 factorial arrangement of treatments. In this way, the long-term effects of grazing strategy (i.e. early life nutrition and heifer growth rate) could be investigated without the confounding effects of differences in lactation management strategy. To our knowledge, this was the first reported experiment investigating the effects of nutritional rearing strategies on the performance of heifers in a pasture-based production system.

A second experiment of note was established in the ‘Grazing Unit’ contiguous to the No. 2 Dairy platform to compare the long-term implications of prepubertal and post-pubertal feeding strategies (Macdonald et al. 2005). Briefly, an experiment with a 2 × 3 × 2 factorial arrangement was established (2 breeds × 3 pre-pubertal growth rates × 2 post-pubertal growth rates): Holstein-Friesian and Jersey heifers were managed to grow at three different rates of average daily gain (ADG) from weaning until the approximate time of puberty; within each one of those six treatments, heifers were randomly assigned to management strategies that achieved two different growth rates between first breeding (∼15 months of age) and first calving.

Over a four-year period, 689 heifer calves (2 breeds: 259 Holstein-Friesian and 430 Jersey) were acquired from 10-commercial dairy farms at 4-d of age and reared at the ‘Grazing Unit’ until approximately 22 m of age when they were returned to their farm of origin. Physical measurements and milk production data were recorded for the first 3 lactations. Although the experiment was not undertaken on the lactating dairy platform at No.2, it was designed and managed by No. 2 research staff on the contiguous Grazing Unit research farm and is material to the Theme being presented.

Results and discussion

Growth and lactation (Experiment 1)

The grazing strategy affected the growth and survival of heifers. Heifer calves that were rotationally grazed from weaning, grew more quickly (Figure 6), were 27 kg heavier entering their first winter (∼9–12 months of age), and by the time of their first calving, were 63 kg heavier (∼20%) than their set-stocked peers (Figure 6). In addition, animals in the rotational grazing system had a substantially reduced mortality rate during winter (McMeekan 1954). McMeekan in the book ‘Grass to milk’ (1960, Pg 25) reported that ‘Not one calf was lost over the whole of this period under the rotational grazing system. In some years, the mortality rate rose as high as 25% under set-stocking’.

Figure 6. Live weight gain profile of animals from weaning (9 weeks) until post-calving in their third lactation: A =

Rotational grazing or =

Set-stocking up to pre-calving and then rotational grazing as lactating cows; B =

Rotational grazing or =

Set-stocking up to pre-calving and then set-stocked grazing as lactating cows (Experiment 1).

The difference in Lwt between the grazing strategies declined during the first lactation, with the animals that were changed from Set Stocking during their replacement heifer growth phase to Rotational Grazing as lactating cows undergoing compensatory growth, such that by the second and third calving, the Lwt difference between the Rotational Grazing and Set Stocking strategies had decreased to 26 and 18 kg, respectively. Interestingly, grazing management strategy for young stock did not affect their subsequent milk production (Table 6). The results of this experiment, therefore, did not support an investment in growing heifers beyond the need for them to reach puberty before a 15-month breeding age and to ensure they were adequately conditioned to withstand the challenge posed by winter.

Table 6. Effects of grazing management strategy^a as replacement heifers (pre-calving) and during lactation (Lactating) on annual milk fat yield/cow (kg; estimated^b MS (fat and protein) yields in parentheses).

	Lactating		SED	P-value
Pre-calving	Rotational Grazing	Set-Stocking		Pre-calving	Lactating	Pre-calving × Lactating
Rotational grazing	160.0 (280)	141.3 (280)	3.25	0.97	<0.001	0.99
Set-Stocking	160.1 (247)	141.4 (247)

^a Grazing management strategies were either Rotational Grazing (i.e. Controlled) or Set-Stocking (i.e. uncontrolled), either as young stock (pre-calving) or post-calving (i.e. lactating cows).

^b Estimated milk fat and protein yields = milk fat yield × 1.75.

From the results of this experiment, McMeekan developed an ‘optimal’ Lwt profile for replacement stock (similar to Figure 7), on which he reported at the Ruakura Farmers’ Conference in 1954 (‘Good rearing of young stock’; McMeekan 1954). In his paper, he stated that:

This type of graph will be similar to every farmer whose wife has reared her children under the Plunket system. The white middle line represents the weights … . and what might be called ‘good economic growth’. This is the target which every farmer should hit. The shaded portion above this to the top curve represents ‘very good growth’ … the shaded portion below the middle curve to the lowest curve represents ‘poor growth’.

This ‘Plunket’-like graph became the recommended curve against which replacement heifers’ development should be gauged.

Figure 7. Live weight targets for Jersey calves from 9 wks of age until pre-calving at 24 mo (i.e. the ‘Plunket graph’; Adapted from McMeekan 1954). The middle line depicts the desired growth rate with the bottom being below the target and the top line being above the target.

More than 40 years later, one of the largest heifer-rearing experiments undertaken globally, confirmed McMeekan’s results, that the effect of heifer growth rates and, by association, pre- and post-pubertal nutrition, had modest and somewhat temporary effects on mature Lwt and milk production, although the effects were dependent on the timing of the nutritional treatment (Macdonald et al. 2005). This experiment was established in the early 1990s to define the optimum way to rear heifers and the most appropriate Lwt for heifers at first calving. At that time, recommended target Lwts were less than those suggested by McMeekan in his ‘Plunket Curve’ 40 years previously (Bryant and McRobbie 1991).

Consistent with the results of McMeekan’s experiment, heifers on high feed allowances pre- and post-puberty were heavier and had a larger frame at first calving, than their slower grown herd mates, irrespective of breed. However, as reported by McMeekan (1960), the slow growth rate heifers experienced compensatory growth during their first lactation. The effect of growth rate was, however, affected by timing of the nutritional treatments, a result worth noting. For example, if the growth rate differences were achieved pre-puberty, the difference in Lwt was retained through to first calving (i.e. there was no compensatory growth during their first pregnancy): a 66 and 46 kg Lwt difference at puberty in Holstein-Friesian (HF) and Jersey (J) animals, respectively, was 68 and 55 kg, respectively, at first calving. This had declined to 29 and 22 kg, respectively, 3 months post-calving, but a residual ∼20 kg difference in Lwt remained through to lactation 3, when measurements ceased. In comparison, if the growth rate differences were achieved post-puberty, compensatory growth during lactation 1 was almost complete, such that 65 kg (HF) and 58 kg (J) difference in first calving Lwt had declined to 17 and 15 kg, respectively, 3 months post-calving and 8 and 7 kg, respectively, after the second calving. The results highlight the importance of growth rate during the first year of life in defining mature size in HF and J heifers.

From a milk production perspective, the timing of different growth rates was also important. A high feed allowance prepuberty did not affect milk production during the first two lactations, despite significant nutrient investment in compensatory growth in the low ADG treatments; but, arguably more important, it reduced (P < 0.01) yield of milk, fat and protein production in Jersey cows during the third lactation. Growth rate post-puberty, however, was positively correlated with first lactation milk production, with nutrients being diverted to compensatory growth in the low ADG treatment heifers and away from milk production. However, the effect was small (milk yield was 7% greater in first lactation heifers on the high post-pubertal feed allowance) and milk production during subsequent lactations was not affected. These results imply that this effect of heifer post-pubertal growth rates on milk production resulted from a direct swap of nutrients from lactation to compensatory growth. Growth rate pre- or post-puberty did not affect reproduction during the first, second or third lactations.

The negative effects of high pre-pubertal growth rate on milk production may indicate that accelerated pre-pubertal growth above a threshold reduces functional mammary development in grazing dairy cows, but this does not affect milk production in early lactations because of:

1. the heifer’s superior size, and an associated ability to consume more, and

2. the diversion of nutrients to compensatory growth in lower Lwt heifers.

This is important because of the longevity of cows in pasture-based systems. Reduced milk production associated with accelerated pre-pubertal growth rate has been extensively reported (Harrison et al. 1983; Capuco et al. 1995). The mammary gland undergoes three distinct growth phases (allometric before puberty and following breeding, and isometric during the peripubescent period); although previously hypothesised (Valentine et al. 1987), feeding level during the phases of allometric growth does not affect mammary development (Meyer et al. 2004). This may be why post-pubertal accelerated gain did not negatively affect long-term milk production. However, the onset of puberty is advanced in heifers experiencing accelerated growth in the first 12 months of life (Macdonald et al. 2005), shortening the period of mammary parenchymal development during the first phase of allometric growth. This is hypothesised to reduce potential milk production, as was evident in Lactation 3 in the heifers with greatest pre-pubertal Lwt gain.

Behaviour and DMI

McMeekan also reported that behaviour of replacement stock was affected by grazing strategy. Grazing ruminants are diurnal animals with crepuscular tendencies (Hafez 1969; Krysl and Hess 1993; Sheahan et al. 2011); they eat very little during darkness and consume their main meals immediately post-sunrise and pre-sunset. Nevertheless, despite consistent reports of these dominant behaviours, set-stocked calves spent more time grazing (9 hours/day) and, therefore, presumably, standing and/or walking than calves in the rotational grazing strategy treatment (7.5 hours/day); furthermore, set-stocked calves had 27% less bites/minute than their rotationally grazed peers (35 and 48 bites/min, respectively; McMeekan 1954). Collectively, these data indicate that set-stocked animals had ~12.5% less bites/day than rotationally grazed animals, despite spending 17% more time grazing.

It is likely that these differences in feeding behaviour resulted in a greater energy requirement for grazing in set-stocked animals and, also, resulted in less DMI and the associated lower ADG presented in Figure 6. Experimental results in beef cattle (O’Sullivan 1984) subsequently confirmed the effects of grazing strategy on animal behaviour. O’Sullivan (1984) reported that length of time spent grazing and the distance walked in the process was greater for continuously grazed animals than for the rotationally grazed animals; set-stocked cattle had a 44% greater energy expenditure in grazing than their rotationally grazed counterparts. He concluded that the efficiency of feed conversion was greater under rotational grazing because of lower maintenance energy requirements. It could be concluded from this work, therefore, that rotational grazing systems resulted in more efficient grazing practices and a greater DMI than set-stocking systems. Grazing strategy may also have affected the quality of the pasture available. The set-stocked calves exhibited a more selective grazing habit, avoiding grazing close to dung and urine patches (McMeekan 1954). These behaviours are likely to reduce pasture quality during mid-season (Stakelum and Dillon 1991), as available pasture is increasingly around previous dung and urine patches, with a lower proportion of leaf to stem, an associated greater fibre content and lower fibre- and DM-digestibility. Such a reduction in pasture quality, along with the greater energy requirement for grazing (i.e. reduced grazing efficiency) and less total bites/day (i.e. a proxy for lower DMI), is probably the reason for the recorded lower Lwt gain in set-stocked heifers compared with those in a rotational grazing system.

Animal health

Despite the greater selectivity in grazing practices, set-stocked calves had a similar intestinal parasite burden to their peers in the rotational grazing treatment, based on faecal examination (McMeekan 1954). Surprisingly, however, the set stocked heifers were more negatively affected by intestinal parasites than their peers managed under rotational grazing. None of the rotationally grazed stock were treated for intestinal parasites; in contrast, set-stocked animals were treated every three weeks and McMeekan (1954) stated that without this, animal deaths were up to 50% in severe winters. Considering the reduced exposure to internal parasites by faecal patch avoidance, the similar parasite burden, and, yet, a more severe effect of internal parasites on mortality statistics (and, probably, morbidity), it is likely that the improved nutrition of the rotationally grazed cattle, as is evidenced by the greater Lwt gain, provided greater immunological resilience to intestinal parasites (Coop and Holmes 1996).

Conclusions and implications

The results presented by Macdonald et al. (2005) are consistent with the previous results of McMeekan (1954) and help quantify the importance of heifer growth rate management pre- and post-puberty. Greater Lwt heifers reach puberty earlier and, providing the greater Lwt gain is attained post-puberty, produce more milk. However, both experiments indicate that the effect on milk production is small in seasonal calving, grazing dairy cows. In fact, accelerated Lwt gain pre-puberty, increases Lwt at first calving and mature Lwt, but can reduce lifetime milk production. Thus, there is a balance between achieving growth rates that ensure the heifer reaches:

• puberty 6–8 wk before planned start of mating;

• a mature size that allows her to calve unaided and maximise feed intake as a cow; with the risk of lower milk production capacity from an accelerated growth rate that shortens the first period of mammary allometric growth.

Considered on balance, however, this means that an optimum Lwt gain can be determined for all breeds of cow. Holstein-Friesian and Jersey heifers reach puberty at ∼250 and 180 kg Lwt, respectively. If calves are weaned at 100 kg (HF) and 80 kg (J) at 12 wk, heifers must, therefore, gain 150 and 100 kg Lwt, respectively, in ∼39 wk, if they are to reach puberty at 12 months, allowing 6–8 wk on average between the onset of puberty and breeding. This provides for a recommendation of 0.55 and 0.37 kg/d ADG pre-puberty for HF and J heifers, respectively.

Growth rate post-puberty is less important to mature size and has only minor effects on milk production. Nevertheless, it is an important heifer characteristic to ensure calving ease, although a lower ADG during pregnancy would, likely, reduce calf size and, so, the effect of heifer post-pubertal growth rate may not be significant. The results do indicate, however, that a low ADG pre-puberty would reduce first calving Lwt, and combined with a high ADG, post-puberty, could result in calving problems. So, care is necessary in rectifying a situation through increased feeding post-puberty where heifers are undersized at 12-15 months.

Theme 4: evaluating new ryegrass varieties and the role of prairie grass in rotational grazing systems

Background

New Zealand dairy pastures are dominated by perennial ryegrass, mainly because it grows in a wide range of conditions, is easy to establish, and is, largely, forgiving of historical poor management. In addition, the plant has a high digestibility, capable of supporting high levels of animal performance. Reported pasture yields from research farms published in the 1950s range between 13 and 19 t DM/ha (Melville 1953; Brougham 1957) and at No. 2 Dairy ranged from 14.0 to 17.7 t DM/ha/yr (average 16.0) in the 1980s (Figure 8). Mitchell (1963) estimated that potential pasture production was 24 t DM/ha for Waikato and Manawatu, 19 t DM/ha in Southland, and up to 28 t DM/ha/yr in Northland, in ryegrass pastures not suffering any moisture or nutrient deficiencies, and that were pest- and disease-free; but, this was clearly not being attained. A subsequent experiment at No. 2 Dairy in the 1990s (reported in Theme 6) identified that, with the application of 200–400 kg N/ha/yr at No. 2 Dairy, on farm pasture production could be increased to 18 and 20 t DM/ha, respectively. The lack of increase in pasture yield through time was confirmed in a review by Hodgson (1989), who reported little advance in ceiling pasture production in New Zealand over the preceding 50 years.

Figure 8. Average annual pasture growth (t DM/ha/yr; ) for No 2 Dairy over 5 yr periods (estimated for 1960s & 1970s using 15.0 kg DM/kg MS; from 1979 growth calculated from weekly farm walks).

It was generally surmised, that the lack of increase in pasture yield between the 1950s and 1980s indicated that the increase in MS per hectare on farms over the 30–40 years was more a result of better pasture utilisation than of any major changes in inherent DM yield potential for well-managed pastures. In fact, in a review of New Zealand dairy production over the previous 50 years, Holmes (1989) reported a three-fold increase in MS per hectare, with the main contributing factors being increased pasture consumption and utilisation from higher SRs.

The experiments discussed in this Theme came about because it was generally thought by farmers and scientists that perennial ryegrass DM production had reached a ceiling yield, despite the theoretical potential. Even though the annual pasture growth from No. 2 dairy in the best years was ∼17.5 t DM/ha, it was believed that there was a need to investigate the potential of new cultivars and/or species to increase the amount of pasture grown and resultant cow production in a farm system situation rather than as a component experiment.

Experimental methods

New ryegrasses and Matua prairie grass

A series of experiments were established at No. 2 dairy in the mid-1980s by Errol Thom. The first (Experiment 17) was established in autumn 1986 to investigate the effect of sowing the farm with either: (1) New ryegrasses with clovers or (2) New ryegrasses with Matua prairie grass; these were compared with (3) existing pasture, which acted as the ‘Control’. The experiment was replicated, with 36 Holstein-Friesian cows randomly allocated to one of two replicates (18 cows/replicate) and 18 Jersey cows randomly allocated (within replicate) to one of the three treatments mentioned. Each farmlet consisted of 4.9 ha, was stocked at 3.7 cows/ha and managed to optimise pasture harvest and MS production/ha as per recommended best management practices at the time.

Pasture cultivar comparison

The second in this series of experiments (Experiment 20) was established in 1991 for two years and investigated the potential for increasing the amount of feed grown on the farm by using perennial and annual ryegrasses. Sixty Holstein-Friesian cows and 15.8 ha were randomly allocated to one of three farmlets (n = 20 cows and 5.26 ha; SR = 3.8 cows/ha): (1) Existing pasture, which acted as the ‘Control’; (2) a newly sown perennial ryegrass cultivar: 85% of the farmlet was undersown with cv. Yatsyn and 15% of the farmlet was sprayed and drilled with a mixture of perennial ryegrass cultivars (cv. Yatsyn, Embassy, Vedette, Grasslands Pacific, and Banks); and (3) an Annual ryegrass-dominant (Lolium multiflorum L.) farmlet: 85% of the farmlet was undersown with cv. Concord and 15% sprayed and drilled with cv. Concord in autumn.

Results and discussion

The prairie grass developed a fungal infection (Anthracnose) in autumn 1987 (end of Year 1), which reduced available herbage by causing excessive rotting. This was followed in the spring by a severe infestation of insects (Argentine Stem Weevil; Listronotus bonariensis; and Hessian Fly; Mayetiola destructor), which further reduced seedling and plant survival (Thom et al. 1992). As a result of these challenges, there was a 10-fold decline in the sown prairie grass population over two years – from establishment in autumn 1986 to autumn 1988 (255 to 25 plants/m²), which resulted in lower Matua/ryegrass farmlet pasture DM production (12.8 t DM/ha per year) relative to the ‘Control’ (13.5 t DM/ha per year) or the New ryegrass farmlet (14.3 t DM/ha per year; Thom et al. 1990). Despite, these differences in pasture DM production, treatment differences in milk fat production in the first season (1987/88) were small; 187, 191 and 184 kg/cow milk fat for the Existing, New ryegrass and Matua/New ryegrass treatments, respectively (Thom and Prestidge 1988). These equate to ∼324, 328 & 318 kg MS/cow, respectively: 324 kg MS/cow for Jersey and 323 kg MS/cow for Holstein-Friesian (~1230 kg MS/ha).

In Experiment 20 (Perennial ryegrass-Annual ryegrass comparison), during the renovation and establishment period, cows on both the pasture-renewal farmlets produced less MS: 0.90, 0.92 and 1.03 kg MS/cow/day for the Perennial ryegrass-dominant, the Annual ryegrass-dominant, and the ‘Control’ farmlets, respectively. This difference was, most likely, a result of lower pasture DM production in the pasture renewal farmlets. For example, in the first year, pasture DM yield on the Perennial ryegrass- and Annual ryegrass-dominant farmlets was reduced by about 250 kg DM/ha in autumn-early winter compared with the ‘Control’ farmlet. Consistent with these pasture production results, cows on the Annual ryegrass-dominant ‘Renewal’ farmlet produced less milk and MS than those grazing the ‘Control’ farmlet and the Perennial ryegrass-dominant farmlet in summer/autumn (January-April) of Year 1, although the latter two treatments did not differ from each other.

Thom and Bryant (1996) reported that as anticipated, average pasture DM yield in late winter-early spring was greater for the Annual ryegrass-dominant farmlet, compared with the ‘Control’ farmlet; this advantage was offset, however, by a lower DM yield in this farmlet in summer-autumn. The two-year average pasture DM production was 16.3, 16.8, and 15.5 tonnes DM/ha/year for the ‘Control’, Perennial ryegrass- and Annual ryegrass-dominant farmlets, respectively. There were no treatment differences in milk production during Year 2 (Thom and Bryant 1996; Thom and Prestidge 1996), but average MS yield (kg/cow) for the 2 yrs was 318, 309 & 299 for the ‘Control’, Perennial ryegrass-dominant and Annual ryegrass dominant farmlets, respectively (Thom and Bryant 1996), a difference which was translated to per ha differences because of the consistency of SR across treatments.

The lower MS production response in pasture renewal farmlets could be partially attributed to the practical restriction of being able to spray and sow no more than 15% of the farmlets, without reducing available feed for the herd and compromising autumn cow production. It was also reported that the annual ryegrass pasture had a lower DM content (%), which has been postulated to reduce cow DMI (John and Ulyatt 1987). This, in addition to lower yield in summer-autumn and the reduced productivity during establishment, may have increased the time and energy required to harvest low DM pasture by grazing (Parsons et al. 1994).

Conclusions and implications

Both experiments highlighted that although some new pastures show promise in component experiments, the results are not always translated into system-level changes and other farm system-related issues can be highlighted (e.g. fungal infection and insect damage-Experiment 17). The value of having the correct endophytes for the pasture were also highlighted (Prestidge and Thom 1994). For Experiment 20, it was concluded that the annual ryegrass cultivar was unable to provide sufficient increases in feed to improve seasonal or total milk production from cows on intensively grazed farms. Thom and Bryant (1996) warned that this experiment highlighted the danger of making incorrect assumptions about the milk production potential of pasture cultivars based on differences in herbage accumulation in plot experiments.

Theme 5: the role of SR in farm profitability and environmental sustainability

Background

Stocking rate is defined as the number of animals allocated to an area of land (i.e. cows/ha), with increases in SR reducing pasture allowance/cow, and vice-versa. It has been recognised as a primary driver of milk production/ha and profitability in grazing systems for over half a century (McMeekan 1956) and is a key factor in deriving the benefits of rotational (controlled) grazing (McMeekan and Walshe 1963): “there is no greater force, for good nor evil, than the control of stocking rate in grassland farming” (C.P. McMeekan 1960).

Increasing SR negatively affects the amount of pasture available to each animal and, therefore, milk production per cow; but, it increases the amount of pasture harvested per hectare and, therefore, milk production/ha (McMeekan and Walshe 1963). These relationships, however, are curvilinear, with DMI/cow increasing at a declining rate with increasing pasture allowance (Poppi et al. 1987) and the increase in pasture utilisation/ha with increasing SR similarly finite. As approximately 60% of the costs in dairy systems reliant on grazed pasture are fixed costs associated with each cow (Macdonald et al. 2011), there must be a point at which the revenue associated with greater milk production/ha is less than the costs associated with more cows (i.e. cost of marginal milk is greater than price received). The curvilinear nature of the relationships and the direct increase in costs with increasing cow numbers suggests that there is an optimum SR and pasture allowance/cow for both biological and economic efficiency.

Researchers had previously tried to define an optimum SR, particularly from an economic perspective (Wright and Pringle 1983). But, optimum SR will vary with location; differences in land class, soil type and fertility, aspect, climate, climate variability, and pasture species, as all affect the amount of pasture produced/ha. These variables and the availability and price of supplements relative to milk price, influence the amount of feed available/ha and, therefore, per cow at any given SR. Furthermore, genetic effects (both breed and strain) on DMI ability affect feed demand/ha. These factors vary across experiments, making it difficult to extrapolate and compare results. Furthermore, optimum is defined by the variable most limiting the capacity of the farm to produce. For example, the optimum economic SR is likely different than the optimum SR for greenhouse gas emissions (GHG) or N lost to ground water.

The ability to consider all the variables associated with pasture and animal production and both the environmental and economic consequences of changes to stocking rate was facilitated with the initiation of the Whole Farm Efficiency (WFE) experiment in 1998, an experiment that was to significantly increase our understanding of SR on pasture production and utilisation, animal production and reproduction, profitability, and environmental outcomes. Wright and Pringle (1983) had recommended that future SR experiments needed to have a sufficient representation of stocking rate as to enable the defining of an optimum SR. In the WFE and for the first time, a large range of SR treatments would be compared in one location and across multiple years.

Experimental methods

Experiments 2, 3, 19, and 26 had SR treatments and the experimental details have been summarised in Themes 1 & 7. To specifically define an optimum SR that could be transferred to multiple environments, the Whole farm efficiency experiment (Experiment 24) was established in the late 1990s by John Penno. The experiment was also used to generate data to assist in the development of a ‘Digital Twin’ for a pasture-based dairy farm system; the resultant product was called the ‘Whole Farm Model’ (Sherlock et al. 1997). The primary objective of the experiment was to investigate the effect of five SRs (2.2, 2.7, 3.1, 3.7 and 4.3 cows/ha) over three lactations on pasture production, utilisation, and quality, milk production/cow and per ha, reproduction and herd health, nitrate leaching, and to provide the biological information required to calculate the optimum SR for profitability and environmental sustainability.

The experiment used 94 Holstein-Friesian cows that were randomly allocated to one of 5 SR (2.2, 2.7, 3.1, 3.7, and 4.3 cows/ha) in a completely randomised design. Treatment farmlets were 8.1, 7.3, 6.1, 4.9, and 4.5 ha in size, with 18, 20, 19, 18, and 19 cows, respectively. The SRs were chosen to give a wide range either side of what would be considered optimum for the farming location, which is relative to the amount of pasture it can produce. Details of the experimental methods are provided in Macdonald et al. (2008b) and a brief description is presented in Table 7.

Table 7. Summary of Whole Farm Efficiency experiment: experimental design, and summary pasture growth, and production results (Experiment 24).

	Stocking rate					P-value
	2.2	2.7	3.1	3.7	4.3	Linear	Quadratic
Planned CSR^a kg Lwt/t DM	62	76	90	103	120
Actual CSR	60	70	76	89	91
Pasture grown kg DM/ha/yr	18,048	18,050	19,484	18,538	20,394	0.11	0.74
Days in milk	291	274	258	234	221	<0.001	0.74
Milksolids kg/cow	407	365	338	295	265	<0.01	0.01
Milksolids kg/ha	895	972	1047	1092	1141	<0.01	<0.01

^a CSR-Comparative stocking rate (kg Lwt/t/DM/ha).

Results and discussion

Since the 1950s, No. 2 Dairy was home to various experiments that investigated differences in SR along with interacting factors (e.g. grazing strategy, N fertiliser or supplement use, cow breed and genetic merit for milk production). McMeekan and Walshe (1963) reported that Jersey cows stocked at 2.9 cows/ha produced 11% less 4% fat corrected milk (FCM) and milk fat/cow on average over a 4-year experiment (Experiment 3; Theme 1) than comparison herds stocked at 2.3 cows/ha, but 12% more 4% FCM and milk fat/ha. Furthermore, they reported an interaction between SR and grazing strategy (rotational grazing or set-stocking), wherein the high SR treatment under rotational grazing resulted in only a 3% reduction in FCM and milk fat production/cow, but a 25% greater yield of FCM and milk fat/ha when compared with a low SR set-stocked treatment. These results were subsequently confirmed by Carter (1964; Experiment 5; Theme 7), although he reported that the advantages of high SRs were even greater in high genetic merit cows: high genetic merit and low genetic merit cows produced 7% and 9% less milk fat, respectively, under a high SR (2.95 cows/ha) compared with a low SR (2.35 cows/ha), but 36% and 31% more milk/ha, respectively. The results presented by Carter (1964) dispute the recommendation that a farmer should reduce SR as herd average genetic merit improves.

Despite comparing only two SRs, however, McMeekan and Walshe (1963) discussed the concept of an optimum SR by comparing across years within treatment, as they changed SR with time. Although not an ideal methodology, it did allow them to conclude when an optimum SR was surpassed, as the decline in milk fat production/cow eroded the per ha advantage associated with increasing SR. Despite design limitations, the results caused them to speculate that the optimum SR would result in lower milk production/cow (−10%), but greater milk production/ha (+15%). The results reported by Carter (1964) seemingly confirmed these effects, with per cow production declining by 7% and 9%, but per ha production increasing by 17% and 14% in the greater SR treatment in High and Low genetic merit cows, respectively. Nonetheless, neither of these experiments were designed to identify a biological or economic SR.

The results presented by Carter (1964) highlighted that the optimum SR was likely to interact with genetics and cow-level physical characteristics. Cow size was recognised to impact on ‘potential DMI’ and, therefore, breeds differing in average Lwt would, likely, have a different optimal SR. As the national herd transitioned from Jersey to Holstein-Friesian in the 1960s and 1970s (Anonymous 1981), there was a recognised need to test the hypothesis that SR can be adequately defined by cow Lwt/ha. Arnold Bryant established an experiment in 1990 to determine if Holstein-Friesian cows and Jersey cows when stocked at the same Lwt/ha had the same production/ha (Spaans et al. 2018; Experiment 19; Discussed in Theme 7).

Spaans et al. (2018) confirmed the effect of exceeding the optimum SR discussed by McMeekan and Walshe (1963), as milk production/cow was 17% lower in Jersey cows at the High SR, and milk production/ha was only 5% greater than in the lower-stocked farmlet. However, in addition, they identified a breed × SR interaction that could not be accounted for by adjusting for cow Lwt, with milk yield/cow 27% less in the Holstein-Friesian-high SR treatment compared with the Holstein-Friesian-low SR. This per cow effect led to a lower milk production/ha in the high SR Holstein-Friesian farmlet than the lower SR comparison. The higher SR reduced pasture production in both breeds, but this effect was exacerbated in the Holstein-Friesian treatments: the Holstein-Friesian farmlets produced 6% less pasture/ha than the Jersey comparison, but this effect was much greater (−17%) at the higher SR. This effect was particularly noticeable during summer and autumn, when the Holstein-Friesian cows ate to a lower (P < 0.05) post-grazing residual, thereby reducing (P < 0.05) regrowth and, ultimately, annual DM production (Spaans et al. 2018).

Further confirmation of the effect of exceeding the ‘optimum’ SR was reported in the 1.75 t Milksolids experiment (Macdonald et al. 2017; Experiment 21; Theme 6), wherein the interaction between SR, N fertiliser use, and supplementary feed were investigated. Increasing SR from 3.3 to 4.4 Holstein-Friesian cows did not reduce annual pasture production, as reported in Spaans et al. (2018), but 4% FCM and MS production per cow declined 24% and per ha milk production was not increased.

Collectively, these experiments provided a better understanding of the role for an optimum SR in successful grazing systems, but they didn’t provide a clear and transferable definition of what would constitute ‘an optimum SR’. In the 1990s, farmers, recognising the superior genetics that had been developed through decades of selected breeding, began to question the traditional ‘high SR’ system recommended by McMeekan, Carter and Bryant over the previous 40 years, believing instead, that fewer higher yielding cows would be more financially astute. An experiment was established at No. 2 Dairy in 1998 that was to achieve two things:

1. define an optimum SR; and

2. develop a method for disseminating the results to multiple situations;

This experiment was called the ‘Whole Farm Efficiency’ Experiment, and the results were to give a greater understanding of the role of tactical grazing management in optimising farm systems, the linear and quadratic effects of SR on pasture production, milk production/cow and per ha, reproduction (Macdonald et al. 2008a), body condition score change (Roche et al. 2007a), economics (Macdonald et al. 2011), and nitrate leaching (Roche et al. 2016). This was the most complete SR experiment ever undertaken, with five SRs (2.2, 2.7, 3.1, 3.7, and 4.3 Holstein-Friesian cows/ha) compared for three years, and pasture and animal-level measurements facilitating a greater understanding of the effects of SR, and enabling an optimum SR to be defined at the biophysical, economic, and environmental-level.

Pasture-level results

Per ha pasture grown, utilised, and consumed directly by the cow increased 986, 1486, 2073 kg with every 1 cow/ha increase in SR; the amount of pasture conserved as silage decreased 607 kg/ha with each additional cow/ha. In addition, average pasture quality improved with SR, with neutral detergent fibre (NDF) (r² = 0.97) and acid detergent fibre (ADF) (r² = 0.93) declining 1.3% and 0.8% DM, and OM digestibility (r² = 0.87) and ME (r² = 0.85) increasing 1.2 % DM and 0.19 MJ ME/kg DM for every additional cow/ha increase in SR. The growth that was measured was pasture accumulation, (i.e. the difference between the pre- and post-grazings). With the higher SR grazing lower there is less material to rot and decay compared with a much more lax grazing at the lower SRs.

These results are noteworthy for two reasons, in particular:

(1) Previous studies had indicated either no effect (Experiment 21; Macdonald et al. 2017) or a negative effect (Experiment 19; Spaans et al. 2018) of SR on pasture growth, the negative effect particularly evident during summer and autumn. However, during the tactical grazing experiments (Theme 2), the importance of lengthening rotation length to facilitate pasture regrowth and associated animal production was quantified and decision rules changed in the 1990s to lengthen rotation when pasture growth declined relative to animal demand.

In the WFE experiment, rotation length during summer, autumn, and winter increased (P < 0.05) 3.4, 5.2, and 7.6 days, respectively, for every extra cow/ha. This increase in rotation length led to greater pasture growth rate and, as a result, a greater net pasture accumulation with increasing SR. In addition, in recognising the negative effect of over-grazing during summer on pasture DM accumulation, decision rules were developed during the 1990s to reduce SR in response to declining pasture growth rates (Macdonald and Penno 1998). These rules are designed to be proactive in reducing SR (i.e. culling early and reducing lactation length) to reduce pressure on pasture and cows. This ensures pasture growth is not reduced under higher peak SRs and that cow BCS is optimised at the start of calving.

(2) A significant proportion of the additional SR was fed by changes to tactical management, through increased pasture DM production/ha and the quality of the feed available. At the lowest SR, cows consumed 5438 kg DM/year. Each additional cow/ha resulted in an additional 986 kg DM grown and 2073 kg DM/ha consumed directly by the cow (1486 kg DM increase in utilisation, when pasture conserved as silage was accounted for). This means that for every extra cow/ha increase in SR, changes to tactical grazing management provided 38% of the feed consumption of the low SR cows.

Furthermore, average ME increased 0.19 MJ/kg DM with each additional cow/ha. When this is factored, in ME consumed as pasture increased by 25.9 GJ/ha per year for every extra cow/ha; this is equivalent to 43% of the ME consumed by a cow at the lowest SR. Therefore, although increasing SR reduces the amount of pasture available/cow, the reduction is much less than calculated theoretically because of the advances in tactical pasture management, which led to increased pasture DM yield and quality.

The effect of SR on annual pasture DM accumulation is a combination of increased growth, due to the sigmoid nature of pasture DM accumulation (Brougham 1957; Voisin 1959), and the reduced loss of grown pasture through senescence (Campbell 1966a); the greater NDF and ADF and lower OM digestibility and ME in the lower SR treatment pasture probably reflects an increased proportion of senescent pasture in the sward.

In the WFE experiment (Experiment 24), the fact that the high SR (HSR) farmlets did not accumulate less pasture than the low SR (LSR) farmlets as happened in Experiment 21, may be attributed to the use of evolved decision rules from the experiments of the 1980s (Macdonald and Penno 1998). In mid- and late lactation these rules are designed to be proactive in reducing SR (culling early) and ensuring optimised cow BCS and pasture cover on farm at the start of calving in the following lactation, as reported in Theme 2.

Animal-level results

As previously reported, FCM and MS production/cow declined with increasing SR, but FCM and MS production/ha increased; for every extra cow/ha increase in SR, FCM and MS production/cow declined 865 and 67 kg, respectively, while FCM and MS production/ha increased 1687 and 116 kg, respectively. The data did not support a biophysical optimum, however, where production/ha did not increase with SR. Nevertheless, the percentage increase in milk production/ha declined as SR increased. Increasing SR from 2.2 to 2.7 cows/ha reduced milk production/cow by 12%, while increasing milk production/ha by 8% to 9%; an additional cow/ha reduced FCM and MS production/cow by 18% and 13%, respectively, but only increased milk production/ha by 4% to 5%. These data confirm that increasing SR leads to a declining per ha advantage. Nevertheless, in this experiment, changes to grazing management to increase available pasture DM/ha facilitated SRs as high as 4.3 Holstein-Friesian cows/ha on a farm that traditionally produced 18 t DM/ha under more moderate SRs; SR did not affect reproductive outcomes or calving pattern.

Economics

Using the biophysical data reported by Macdonald et al. (2008b), an economic analysis was undertaken to better understand the financial trade-offs between the lower per cow production and greater per ha production (Macdonald et al. 2011). Milk production, gross revenue, operating expenses, and operating profit per cow all declined with increasing SR. Milk production and, therefore, gross revenue per hectare increased with increasing SR, as did operating expenses; however, there was a tendency for gross revenue to increase quadratically, with the increase getting smaller as SR increased, while operating expenses increased linearly with each additional cow. As a result, operating profit/ha increased quadratically, with optimum profitability achieved at ∼3.3 Holstein-Friesian cows/ha. These effects were irrespective of milk price but did change with milk pricing system. In fluid milk systems, operating profit continued to increase, up to 4.3 cows/ha, because the reduction in milk composition with increasing SR did not disadvantage the higher-stocked systems. The results confirmed the conclusions of both McMeekan and Walshe (1963) and Carter (1964) that, in a system where farmers are paid for milk components, there was an optimum SR, above which the increase in milk production/ha was eroded by the reduction in milk production/cow, as, above a certain point, the revenue/cow did not cover the cost of the marginal milk associated with the increase in SR.

Effect on environment

An assumed positive relationship between increasing SR and N loss to the environment and methane (CH₄) emissions/ha are often discussed as reasons to reduce SR (i.e. de-intensify). The 1.75 t MS experiment (Experiment 21) and WFE experiment (Experiment 24) allowed the isolation and quantification of the effects of SR on these important environmental concerns, removing the effects of other system-level intensification factors (e.g. N fertiliser, imported feed use).

From a GHG perspective, increasing SR without any increase in feed availability increased the emissions of the dominant GHG, CH₄, by 8% for every additional cow/ha. By applying better grazing management practices to increase pasture grown/ha, as in the WFE experiment, increasing SR increased CH₄ emissions by 14%/ha (Figure 9). In comparison, however, maintaining pasture grown at the different SRs and importing either maize grain or maize silage (Experiment 21) to maintain feed DM allowance/cow, increased CH₄ emissions/ha by 31%/ha and 38%/ha for the maize silage and maize grain supplemented farmlets, respectively.

Figure 9. The effect of stocking rate (cows/ha) on pasture harvest (t DM/ha;

) and estimated (est.) methane (CH₄) emissions/ha (kg CO₂-eq;

), milksolids production/ha (kg;

) and operating profit/ha (NZ$;

), and nitrate leached/ha (kg;

). Methane emissions/ha were estimated from the equation: kg DM × 21.6 (Clark 2008, Macdonald et al. 2008b, 2011. Roche et al. 2016).

Nitrogen use efficiency is low in grazing systems (Ledgard et al. 2009; Huebsch et al. 2013), because intensively managed temperate pastures contain substantially more rumen-degradable protein than is required for the milk production levels achieved (Kolver and Muller 1998; Roche et al. 2009b, 2009d; Roche 2017); the excess N is excreted in high concentrations in urine and can contribute to increased soil solution and groundwater nitrate-N (NO₃-N) concentrations (Huebsch et al. 2013; Shepherd et al. 2014; Selbie et al. 2015).

Julian et al. (2017) reported that between 1990 and 2015, NO₃-N concentrations increased in 35% of New Zealand’s monitored rivers, which they attributed to increased dairy cattle density and legacy nutrients that have built up on intensively managed grasslands since the 1950s and are slowly leaking into the rivers. Because of this, it is often assumed that an increase in SR leads to an increase in NO₃-N leaching, because of the low N-use efficiency in grazing animals and the significant contribution of the urine patch to N loss (i.e. more cows = more urine patches). The results from No. 2 Dairy were, therefore, surprising, as there was a linear decline in NO₃-N leached/ha with increasing SR (P = 0.06; NO₃-N leached/ha = −12.7 SR + 76.7), which coincided with a decline (P = 0.06) in the concentration of NO₃-N in leachate with increasing SR.

Stocking rate did not affect N-use efficiency/ha, as measured by the output of N in milk and meat relative to total N intake (13% to 14%) or net importation of N (i.e. fertiliser N ± silage imported/exported; 35%). These levels of N-use efficiency are similar to those reported elsewhere for intensively managed pasture-based systems (Huebsch et al. 2013). Huebsch et al. (2013) did acknowledge a reduction in NO₃-N leaching with increasing pasture harvest in Ireland, but the cause of the lower NO₃-N leached in their study was multifactorial. Nevertheless, in the No. 2 experiment, NO₃-N leached/ha did decline 8.4 and 5.9 kg with each 1 t of DM pasture harvested/ha (i.e. grazed plus conservation) and consumed directly/ha (i.e. grazed only), respectively, highlighting the importance of grazing management in sustainable pasture-based dairying.

The timing of N deposition in urine patches is critical to the risk of N loss (Shepherd et al. 2014). When pasture is growing actively, most of the plant-available ammonium (NH₄) and NO₃-N in the soil are readily taken up by the plants. However, when pasture growth declines, which occurs particularly during autumn, NO₃-N can accumulate in the topsoil (Whitehead 1995; Shepherd et al. 2014). In free-draining soils, when the amount of precipitation exceeds evapotranspiration (i.e. drainage season), surplus water moves downward through the soil profile and any NO₃-N positioned in the topsoil under these conditions has a high risk of being leached (Shepherd et al. 2014).

Urine N deposition during autumn and early winter, as pasture growth slows and the balance of precipitation and evapotranspiration swings in favour of drainage, is a greater risk factor for NO₃-N leaching than urinary N deposited during late winter, spring, and summer, when pasture is at peak growth (Roche et al. 2009b) and evapotranspiration exceeds precipitation (Roche et al. 2009a; Shepherd et al. 2014). In dairy systems that don’t import much feed, lactation length is a way of managing asynchrony between pasture supply and animal demand and, importantly, for ensuring that animal BCS and pasture cover at start of calving are optimised (Macdonald and Penno 1998). As a result, average lactation length declined with increasing SR: 34.5 d for each additional cow/ha increase in SR. Nonlactating cows consume approximately 50% of the DM and, more importantly, the N, of a lactating cow on the same calendar day. With increasing SR, N intake/lactating cow declined because of lower DMI and a greater proportion of cows were nonlactating through autumn (because of the shorter lactation length/cow). Both factors result in a decline in N intake/cow with increasing SR and a corresponding reduction in N excreted in urine. In addition, cow BCS in mid- to late lactation declines with SR (Roche et al. 2007a); a greater proportion of the N consumption during this period, therefore, is partitioned to muscle accretion as SR increases because of the stock-management strategies employed to ensure optimum BCS at calving. The greater OM digestibility of the pasture in the higher SR treatments might further facilitate this anabolism (Mandok et al. 2014).

Importantly, if we accept that the primary reason for the negative association between SR and NO₃-N leaching/ha is a reduction in surplus N intake and urinary N excretion/cow during the ‘sensitive’ months for N accumulation before the drainage season, and that reducing days in milk (DIM)/cow with increasing SR is the management strategy that facilitates this, the negative association between SR and NO₃-N leached will not hold true. However on low SR farms, where cows are milked into late autumn/early winter to maximise DIM, in autumn-calving scenarios, where early calving cows receive high N diets, and in spring-calving systems in which supplementary feeds are provided in autumn to extend DIM/cow, DIM/cow, N intake/cow, or both during the autumn do not decline with SR and, in fact, NO₃-N leached/ha will, very likely, increase (Ledgard et al. 2006; Peyraud and Delaby 2006); for example, Ledgard et al. (2006) reported that the NO₃-N leached increased by approximately 18 kg/ha per year for every additional cow/ha increase in SR, when imported feed was used to extend DIM/cow in autumn in a spring-calving system.

Genotype by feeding level interaction

Bryant et al. (2003), a Ph.D. student with the Massey University farm systems expert, Professor Colin Holmes, analysed the No. 2 data to better understand whether there was an interaction between SR (i.e. per cow feeding level) and the genetic merit of the cows. They reported that at low levels of feeding (highest SRs), Estimated Breeding Values (EBV) for production traits accurately predicted differences between cows; in comparison, however, at high levels of feeding (lowest SRs), the EBVs significantly under-represented the milk yield differences between genetic groups. The SR at which the EBV accurately predicted phenotypic differences between cows were 4.1, 3.6, 4.9, and 3.8 cows/ha for milk, protein, fat, and MS yields, respectively; this is equivalent to 4.4, 5.0, 3.7, and 4.7 t DM allowance/cow at 18 t pasture DM/ha.

As the focus of selection through the 1990s was for a balance of milk fat and protein (MS) and EBVs are predicted from daughter test results for sires, with these daughters managed at average SRs, it is unsurprising that the SR where EBV most accurately predicts the phenotype for production traits is between 3.0 and 4.0 cows/ha (between 4.0 and 5.0 t DM/ha), although the poor prediction of fat yield is surprising. In fact, the BW (New Zealand measure of genetic merit) was originally developed in 1996 and was based on the profitability of an animal at a feed allowance of 4.5 t DM (it increased to 5.0 t DM/cow when the economic model was rewritten in 2013; pers comm Bevan Harris).

The relationship between EBV and phenotype and the interaction with feed allowance are, somewhat, consistent with those of Fulkerson et al. (2008) in Australia. They reported that the EBV was much greater than the phenotypic difference between cows in farm systems receiving very little concentrate supplement (∼0.3 t DM/cow), and only reflected the phenotypic difference when cows received >0.8 t concentrate DM/cow, a system more similar to the systems under which daughters were evaluated. The results reflect a linear ‘scaling’ of phenotypic performance relative to the cows’ EBV (a genotype × environment interaction), but no re-ranking, and are, therefore, consistent with Carter (1964) that high genetic merit cows perform better in all systems. This will remain valid as long as the most important phenotypic characteristics for sustainable profit are included and appropriately weighted in the selection index. This was to be highlighted explicitly in the Strain Experiment at No. 2 (and Scott Farm; Experiment 26; Theme 7), wherein animals selected on production and without appropriate weighting on important functional traits for pasture-based systems (e.g. reproductive function) re-ranked in their performance at a system-level (Kennedy et al. 2002; Horan et al. 2005; Macdonald et al. 2008b).

Conclusions and implications

The near 60 years of research, investigating changes in SR in conjunction with other key variables (McMeekan 1950; Bryant 1990; Macdonald et al. 2008a, 2011, 2017; Roche et al. 2016; Spaans et al. 2019), provided farmers confidence to increase SR and maximise milk production from pasture. The WFE experiment was to exceed all expectations, however, demonstrating increased pasture grown and utilised/ha, and harvested directly by the cow, with very high SRs manageable with the discipline founded on the DSS developed in the 1980s (see Theme 2). Furthermore, the economic and water quality implications of increasing SR were investigated and the fable that SR needed to be reduced to ‘achieve the genetic potential of the cow’ was put to rest.

In addition, the level of measurement undertaken facilitated an extension of the work beyond No. 2 Dairy and beyond the Waikato, with the introduction of the concept of Comparative Stocking Rate (CSR); CSR is a metric that acknowledges that different farms have different pasture growth potential, different cows have different intake abilities, and different systems of farming made greater or lesser use of off-farm grazing and supplementary feeds (see Theme 5). The CSR uses the Lwt of the cow as a measure of cow intake capacity and added the total feed DM together (i.e. pasture and non-pasture) to provide an index: kg Lwt/t feed DM available. It is simplistic in that it assumes all feeds average 11 MJ ME/kg DM, but is a very effective DSS for estimating the optimum SR for a farm and farm management system.

The collated data have facilitated an informed discussion on optimum SR. Focussing on farm financial performance, a detailed assessment indicated that the optimum CSR was 77 kg Lwt/t DM feed available (Table 7; Macdonald et al. 2011), although, admittedly, the effect of a wide range of CSR on profitability was small. In that analysis, however, the positive effect of SR on pasture growth was included in the analysis, something that farmers are unlikely to consider when estimating what is likely to be the optimum SR for their farm. A subsequent reassessment of these results (Macdonald et al. 2017) led to the conclusion that the optimum CSR for profitability is closer to 90 kg of Lwt/t of DM feed available (see Table 7 for planned CSR).

The work was undertaken prior to the recent but significant emphasis on farm management effects on environmental outcomes and the analyses to estimate ‘optimum’ SR assumed that the variable most limiting production from pasture and, therefore, profitability, was land area. Increasingly, environmental sustainability and the need to reduce GHG emissions and contaminant loads to receiving waterbodies have become important objectives for the public and future farm systems may be limited by the amount of biogenic CH₄ emitted/ha and/or the amount N they can lose through sub-surface drainage. In Figure 9, we present the different biophysical, financial, and environmental parameters and how they are affected by stocking rate. For example, Roche et al. (2016) reported that operating profit in a seasonal, spring-calving, pasture-based dairy farm in a N-sensitive zone would increase linearly with SR (NZ$16.90/kg of NO₃-N leached for every extra cow/ha increase in SR), challenging the wisdom of mitigation strategies proposed for seasonal-calving dairy farms in New Zealand (Vogeler et al. 2013). However, methane emissions increase with each additional cow/ha, suggesting that optimum SR will be lower if the factor most limiting production/ha is the amount of GHG a farmer is allowed to produce unabated. In the future, optimum SR will be determined by the factor most-limiting production, but the WFE (Experiment 24) has contributed detailed measurements that will facilitate modelling to optimise different scenarios as the need arises.

Theme 6: productive and economic responses to non-pasture supplementation of dairy cows

Background

From a nutrition perspective, it was recognised almost a century ago that moderate-yielding cows could adequately nourish themselves on immature pasture; for example, Woodward (1936), in the USA, concluded that immature pasture was a relatively well-balanced feed. He also concluded, however, that there was a threshold level of milk production above which grazing cows would produce more milk if supplemented with ‘hard’ or ‘concentrate’ feeds (i.e. processed cereal grains or the co-products of cereal grain or oil seed processing; Roche et al. 2017a).

Consistent with this premise, Boutflour (1928), in the UK surmised that poor-quality forage was one of four key variables limiting milk production in early lactation and Cooper (1941), in New Zealand, reported that peak milk production occurred later in grazing cows than it should from a physiological standpoint; instead, he claimed, it coincided with peak pasture growth, suggesting feeding level in early lactation was limiting per cow milk production. Therefore, it would seem, farmers were losing milk production opportunities in systems reliant on grazed pasture. To redress some of this ‘lost milk production opportunity’, farmers were interested in using supplementary feeds and considerable research was undertaken to understand component responses to supplementary feeds. This interest in, and argument for greater milk production/cow continues to the present day (Roche et al. 2017b).

In accepting that fresh pasture is a nutritionally balanced feed for moderate-yielding ruminant animals (Woodward 1936; Roche 2017), there are two reasons for providing grazing dairy cows with additional feeds (i.e. supplementary feeding):

1. to nourish cows during periods when there is insufficient pasture growth to meet herd demand (i.e. stabilise feed supply: feed robustness); and

2. to increase the milk production of the herd above that achievable from pasture alone, by facilitating a greater SR, greater milk production/cow, or both (i.e. increase production from the farm: maximise nutritional efficiency).

Both justifications for additional feeding appear sensible. Although grazing systems are synonymous with cows self-harvesting forages, the profiles of pasture supply and herd demand are not perfectly aligned (Roche et al. 2009a, 2009d, 2017), creating periods of pasture surplus and deficits relative to requirements (see Figure 1). Furthermore, the supply of fresh pasture is heavily dependent on the prevailing weather, particularly soil moisture, temperature, and solar radiation; although temperature and radiation are reasonably repeatable within fortnight across years in New Zealand’s temperate zone, rainfall and, therefore, soil moisture, are not (∼3% repeatability within fortnight across 6 years; Roche et al. 2009b). As a result, pasture growth can vary and, even with optimal management, deficits in supply relative to herd demand can occur and, as a result, herd milk production and farm revenue can be compromised.

Although much research was being undertaken to understand component-level (i.e. short-term responses to supplementary feeds) and N fertiliser, there was limited, if any, information on the farm system-level responses to supplementary feed, especially when some advantages can pass across years (e.g. cow BCS, pasture cover). Several experiments were established at No. 2 dairy to investigate the biophysical and econometric effects of providing supplementary feeds to grazing dairy cows in addition to pasture. These experiments were designed specifically to investigate both previously mentioned strategies: increase MS production/cow and/or increase MS production per ha through facilitating greater SRs.

The experiments – (Experiment 7, 8, 11, 21, and 22)

No. 2 Dairy was to host four experiments over 30 years, which were to unequivocally answer the vexing question on the role of supplementary feeds on farm productivity (i.e. milk production from a resource) and profitability (i.e. earnings before interest and tax (EBIT) from the same resource). The experiments were designed to investigate the interaction between calving date, SR, and supplementary feeding, whether that supplementary feed was grown on farm, harvested and fed back to cows during periods of pasture deficit, or purchased from off-farm and offered to cows during periods of pasture deficit. The experiments also investigated differences in the effects of silage vs. concentrate feed; an additional component experiment (Experiment 22) was designed to investigate the role of protein nutrition when a high proportion of the cow’s diet was low crude protein maize silage.

The first experiment was undertaken in 1967 by Archie Campbell, with the objective of quantifying the effect of the early lactation feed deficit on milk production by supplementing cows in early lactation with two levels of a concentrate meal and comparing them with an unsupplemented ‘Control’ herd. In the second experiment, the interaction between SR, calving date and supplementary feeding was investigated. Campbell, an advocate of greater SRs to maximise pasture utilisation, established a one-year experiment to investigate altering herd PSC to determine whether there was an interaction between SR, PSC, and any advantage that the use of supplementary feeds in spring-calving herds might provide. This design recognised that both SR and calving date influenced the extent of any early lactation feed deficit and, in so doing, both variables would likely affect responses to supplementary feeds.

In the third experiment, Campbell acknowledged there had been little gain in pasture production during the previous decades and if milk production/ha from pasture were to increase further, he would need to devise a strategy that increased DM production/ha. He also recognised the potential for superior DM production from maize planted for silage in spring and harvested in early autumn, a period when pasture production and quality were variable and, generally suppressed by temperature and moisture deficit. In this experiment, Campbell investigated whether per cow and per ha productivity could be improved by comparing the established system (Control) over two seasons with systems in which either 20% (MS-20%) or 50% (MS-50%) of the farm area was planted in spring with high density, direct-drilled forage maize for ensiling in autumn (Campbell et al. 1978).

In a final tour de force (Experiment 21), Arnold Bryant and colleagues were to establish, arguably, the most complete supplementary feeding experiment ever undertaken in grazing systems. In its design, the researchers were able to investigate:

1. The role of N fertiliser in increasing pasture available/cow and per ha and associated increases in milk production at both moderate and high SRs (see Theme 10);

2. The role of a ‘balanced’ concentrate meal, cracked maize grain, and maize silage in increasing feeding level per cow or in increasing SR while maintaining feeding level per cow; and

3. If different supplementary feeds resulted in different per cow and per ha responses.

The experiments were sufficiently long and detailed to investigate the productivity and economic effects of farm intensification through greater SRs and/or the provision of additional feed/ha.

Experimental methods

In the 1st experiment (Experiment 7), Archie Campbell compared feeding two concentrate meal allowances (1.25 and 2.50 kg/DM/cow per day) to dairy cows for six weeks after calving with an unsupplemented ‘Control’ herd. All three herds were stocked at 3.56 cows/ha and subjected to an ‘apparent shortage’ of grazing in the spring (the experimental descriptions do not provide further detail).

In the 2nd experiment, 192 Jersey cows were randomly allocated to one of six herds (n = 32 cows/herd), each of which were randomly allocated to Low and High SR treatments (3.1 and 4.3 cows/ha). Within each SR treatment, each herd was randomly allocated to an early-calving (EC) treatment with (EC+) and without (Control) supplementary feed, and a late-calving treatment (LC), without supplementary feed in a 2 (SR) × [2 (calving date) + 1 (supplementary feeding)] arrangement.

Early calving cows (Control and EC+) were bred in October for a July PSC, as per industry recommended best practice; LC cows were bred in November for an August PSC, reducing the early lactation feed pressure and the assumed ‘requirement’ for supplementary feed. Cows in the EC+ treatment received 1.8 kg supplementary feed DM/d from calving until 19 September (i.e. approximately 60 days), in addition to their pasture allocation (supplement was ∼15% of estimated DMI).

In the 3rd experiment, Campbell sought to increase feed production/ha by growing maize silage on either 20% (MS-20%) or 50% (MS-50%) of the farm area and comparing these two treatments over two seasons with a ‘Control’ herd. The Control and the MS-20% herds were stocked at 3.95 cows/ha (42 Holstein-Friesian × Jersey cows plus replacements on 12.14 ha; 3.45 milking cows and 0.5 cow equivalents as replacements per ha), while the MS-50% herd had a SR of 4.7 cows/ha (50 Friesian × Jersey cows plus replacements on 12.14 ha; 4.1 milking cows and 0.6 cow equivalents in replacements per ha), in acknowledgement of the expected additional DM production from the maize silage.

In the maize silage farmlets, 20 per cent (MS-20%) or 50 per cent (MS-50%) of the farm area was sprayed with paraquat in late spring (November) to kill off the vegetation and was planted with high density, direct-drilled forage maize for ensiling in autumn and feeding during periods of pasture shortage (Campbell et al. 1978). After harvesting, the maize areas were re-sown (i.e. direct drilled) with Italian ryegrass (Lolium multiflorum, cv. Manawa). Pasture was used as the only protein and mineral supplement for the maize silage-fed cows.

The 4th experiment (Experiment 21; 1.75 t MS experiment) was the last experiment established by Arnold Bryant and ran in two phases. In the first two years of the project (Phase 1), there were 7 farmlets, each of 6.47 ha with Holstein-Friesian cows and organised as 2 SR (3.2-LSR and 4.5 c/ha-HSR), each with one of 3 levels of N (N: 0, 200, 400 kg N/ha) and with or without maize grain (Table 8). Farmlets 1–4 were stocked at 3.2 and farmlets 5 to 7 at 4.5 cows/ha, and Farmlets 1 & 5 received no N or purchased supplements and were the ‘Control’ at the two SRs. Farmlet 2 received 0 kg N fertiliser/ha plus maize grain. Farmlets 3 and 4 had a SR of 3.4 and were designed to investigate the effect of use of N and supplementary feed on MS production at a low SR, so received either 200 or 400 kg N/ha plus supplement (Table 8). Farmlets 5 (No N) and 6 & 7 were stocked at the higher SR to investigate the effects of N fertiliser and supplementary feed/cow on MS production when extra cows were used to ensure any additional feed is well utilised. The cows in the supplement farmlets received enough supplement (maize grain) to ensure they were fully fed. The amount supplemented to each farmlet is presented in Table 8.

Table 8. Treatment details for 1.75 t MS/ha experiment (Experiment 21; Phase 1; 1993–1995 adapted from Penno et al. 1996).

Farmlet	1	2	3	4	5	6	7
Cows/ha	3.24	3.24	3.24	3.24	4.48	4.48	4.48
Number of cows	21	21	21	21	29	29	29
N kg/h/yr (nominal)	0	0	200	400	0	200	400
N kg/h/yr (applied)	0	0	215	380	0	206	370
Maize grain supplement	No	Yes	Yes	Yes	Yes	Yes	Yes
Maize grain offered (kg DM/c/yr)^a		548	139	105	1460	1043	958

^a Average of 2 years.

After two years, the experiment was changed (Phase 2) and increased in size to 8 farmlets (Table 9; this experiment is comprehensively reported by Macdonald et al. 2017). One farmlet (LSR-0N) served as a ‘Control’ farmlet receiving no N fertiliser or purchased supplements (SR was 3.35 cows/ha). Farmlets 2 to 5 (LSR-200N, LSR-400N, HSR-200N and HSR-400N) were designed to investigate the effects of N fertiliser (either 200 or 400 kg N/ha) on MS production per cow and per hectare at a low (LSR – 3.35 cows/ha) and high (HSR – 4.41 cows/ha) SR. Farmlets 6–8 (HSR-200N-MG, HSR-200N-MS, and HSR-200N-BR) were designed to investigate the use of large amounts (>1 t DM/cow per annum) of different supplementary feeds to increase MS per cow and per hectare at a high SR, and included the application of 200 kg N/ha/year. In the three-supplementary feed farmlets, cows were fed maize grain (MG-200N), maize silage (MS-200N), or a ration designed to provide what a dairy cow nutrition model deemed lacking in the pasture diet (Balanced ration: BR-200N).

Table 9. Experimental design, and summary production and economic results (for the 1.75 t MS/ha experiment ( Least square means over 3 years: 1995/96–1997/98). Effects of nitrogen fertiliser use, stocking rate, and supplementary feeding on modelled per ha greenhouse gas (GHG) emissions are presented (Experiment 21; Phase 2; adapted from Macdonald 1999 & Macdonald et al. 2017).

Herd	1	2	3	4	5	6	7	8
Treatment	LSR^a 0N^c	LSR 200N	LSR 400N	HSR^b 200N	HSR 400N	HSR 200N MG^d	HSR 200N MS^e	HSR 200N BR^f
Stocking rate (cows/ha)	3.4	3.4	3.4	4.4	4.4	4.4	4.4	4.4
Nitrogen used (Kg N/ha)	0	202	418	203	416	207	209	211
Milksolids (Kg/cow)	311	357	390	267	296	396	359	400
Milksolids (Kg/ha)	1042	1196	1306	1177	1305	1746	1574	1764
Total GHG –(kg CO2 eq/ha)	11,779	15,431	17,753	16,758	19,080	22,590	19,928	22,823
Operating profit^g ($/ha)	$3138	$3722	$4112	$3004	$3572	$3116	$3369	$1650^h

^a LSR-Low stocking rate = 3.3 cows/ha.

^b HSR-High stocking rate = 4.4 cows/ha.

^c N-Nitrogen.

^d MG-Maize grain-1.25 t DM/c/yr.

^e MS-Maize silage-1.28 t DM/c/yr.

^f BR-Balance ration-1.46 t DM/c/yr.

^g Operating profit calculated as outlined in Macdonald et al. (2017), but with a MS price of $6.50/kg MS. All expenses remained the same.

^h Operating profit estimated from Macdonald et al. (2017), with the assumption that the balanced ration supplement cost $750/t and the MS price was $6.50.

The experimental design (see Macdonald et al. 2017) allowed a number of farm system comparisons to be evaluated:

1. The effect of increasing SR, with a declining feed allowance/cow, partially offset by increasing N fertiliser use (LSR-200N and LSR-400N vs HSR-200N and HSR-400N) in a 2 × 2 factorial arrangement, and compared with a LSR farmlet that didn’t receive N fertiliser (LSR-0N);

2. The effect of increasing feed allowance/cow and per ha at high SRs by importing supplementary feed (HSR-200N vs HSR-MG, HSR-MS, and HSR-BR);

3. The effect of increasing SR while maintaining feed allowance/cow, by importing an equivalent amount of supplementary feed/cow, to increase production/ha (LSR-200N vs HSR-MG, HSR-MS, and HSR-BR); and

4. A response curve to N fertiliser (LSR-0N vs LSR-200N vs LSR 400N).

A 5th nutrition experiment (Experiment 22) was undertaken over one year to investigate the effects of increasing the protein content of pasture and maize silage diets for dairy cows, using urea, soybean meal, or fishmeal. In this short-term component experiment, there were 4 treatment groups of 10 Holstein-Friesian cows. Pasture allowance was manipulated to achieve an average pasture DMI of 6.6 kg/cow/day and supplemented with 75 MJ ME/cow/day as either 7.0 kg DM of maize silage, 7.0 kg DM of maize silage plus 90 g of urea, 5.2 kg DM of maize silage plus 1.4 kg DM of soybean meal, or 5.8 kg DM of maize silage plus 1.0 kg DM of fish meal (Macdonald et al. 1998). Milk production was measured in spring, summer and autumn at what were deemed to be critical times of the year.

Results and discussion

While pasture is an adequate feed for moderate-producing dairy cows and it provides for an effective and low-cost system of milk production, it was recognised that cows were not being adequately fed on pasture to fulfil their 'genetic potential'. Consistent with this, McMeekan and Walshe (1963) and Carter (1964) (Theme 1) reported greater milk production/cow in lower SR systems because of the higher feed allowance/cow. This effect was irrespective of genetic merit (Theme 3; Carter 1964). Therefore, importing feeds to offer cows a higher feed allowance in a high-stocked system was expected to result in greater milk production.

Short lactation lengths relative to international benchmarks was also identified as a factor limiting productivity from pasture-based systems. For example, the average lactation length at No. 2 Dairy was only about 240 days, compared with a recommended optimum internationally of ∼305 days; in other words, New Zealand farmers were missing out on two months of milk production/cow because of the limitations in the pasture production profile. Furthermore, with the recommendation to further increase SR, lactation length was only likely to decline further.

Early experiments at No. 5 Dairy Ruakura (Wallace 1957) highlighted an increase in milk production with concentrate feeding, but acknowledged that ‘concentrate feeding should not be undertaken in the hope of obtaining immediately profitable returns’, reflecting too high a concentrate to milk price ratio in New Zealand at the time. Nevertheless, there was still an interest in ensuring cows were ‘fully fed’ in spring and autumn and, if this could not be achieved with pasture, then the cows should be supplemented with a non-pasture feed. Farmers, therefore, sought the use of supplementary feeds or high-yielding forage crops, such as maize, to increase both feed availability/cow and per ha.

Responses to supplementary feeds in early lactation

The first two experiments in this theme were a recognition that very early lactation milk production in grazing dairy cows was, likely, compromised by low winter pasture growth rates and the need for high SRs to maximise pasture utilisation during spring (i.e. low feed allowance/cow in early lactation). Campbell, therefore, sought to strategically utilise supplements during the first 40–60 days of lactation or delay calving date to reduce the size of the early lactation feed deficit.

The effects of supplementary feeding were extraordinary by the lack of any material effect on milk production. In both Experiment 7 and 8, there was no milk production response from meal feeding (Annual Report 1966–67, pp. 1967–68), in agreement with the earlier report by Wallace (1957). In Experiment 7, the lack of a significant production advantage from meal feeding was reported as even evident at the higher SR (Annual Report 1967–68), consistent with results from the HSR treatment in Experiment 8.

Although surprising, these results are not inconsistent with experiments decades later that tried to understand the reasons for variable responses to supplementary feeds in grazing dairy cows (see Stockdale 2000; Penno 2002; Poole 2018). Responses to supplements are primarily affected by the size of feed deficit (Penno 2002), the genetics of the cow (Horan et al. 2005; Fulkerson et al. 2008), and the stage of lactation (Poole 2018), but there are also some less well-understood associations with the quality of the base pasture (i.e. the better the pasture quality or the greater the clover content, the smaller the milk production response to supplement; Stockdale 2000). Consistently, responses to supplements were least in early lactation and greatest in late lactation (Poole 2018).

It is unclear what causes this lactation-stage effect, but it is unlikely related to physiological state. It may, however, relate to the feed quality on offer at the time supplements are offered. When cows consume supplements, the presence of energy-dense ingredients in the intestine and products of digestion circulating in blood result in neuroendocrine feedback signals that cause a cessation of eating (Roche et al. 2007d, 2008); this is referred to as ‘substitution’ of pasture with a supplementary feed and the amount of pasture refused relative to supplement eaten is referred to as the substitution rate. Substitution rate is consistently greatest in spring/early lactation and, as a result, the marginal milk production response (MMPR) to supplements is similarly least (Stockdale 2000; Poole 2018). This is because the net increase in DM and ME intake is small, because substitution rate is greatest and the quality of the pasture being refused is also greatest. It could also be that the level of supplementation was too low for any response to be detected. These factors may explain the lack of response to supplements in Experiment 7 and 8.

The delayed calving treatment also failed to increase annual milk production/cow. Even though the early lactation feed deficit was less in this treatment, relative to the early calving ‘Control’ group, and this was reflected in a greater peak milk production, a summer drought and facial eczema conditions resulted in all cows being dried off on 5 April and, therefore, a much shorter lactation in the LC treatment. In fact, the LC cows produced considerably less than those that calved earlier (Annual Report 1967–68).

These results highlighted the importance of milk production/ha before summer (before Christmas), as something that can be relied upon because of reasonable climatic repeatability and, therefore, pasture growth in spring (Roche et al. 2009b). The near trebling in the variability of pasture DM production between January and June, relative to during spring (DairyNZ, unpublished data), means that post-Christmas milk production from pasture is much less certain. So, although higher SRs reduce the amount of feed available/cow over the year and DIM/cow, the feed availability/cow pre-Christmas is not greatly affected, and DIM/ha tend to be much greater. The results of Experiment 8, therefore, highlight the importance of early-winter PSC, along with a higher SR, to ensure that maximum DIM per ha are achieved before Christmas.

These studies underscore the advantages of a high SR, supporting the previous work outlined in Theme 5. The higher SR treatments in Experiment 3 produced 54 kg milk fat/ha (∼90 kg MS/ha; +12%) more than the low-stocked treatments, primarily because the potential advantage of longer lactations associated with a lower SR did not occur. Although not a principal objective of these experiments, this finding also highlights the importance of farm systems experiments to fully contextualise component results. For example, although the LC treatment resulted in greater peak milk yields and, if the experiment were to have been undertaken in early lactation, the researchers could justifiably have concluded a benefit from delaying calving, the full system implications of the decision to delay calving date provided for a different conclusion.

Total milk production responses to additional feed or supplementary feeds

Although earlier experiments identified limited, if any, milk production responses to reducing the early lactation feed deficit, with supplementary feeds in spring or later PSC, the 3rd (Experiment 11) and 4th (Experiment 21) experiments in this theme allowed researchers to investigate the full farm system MMPR to significantly increasing the amount of feed available per cow and per ha.

The 1.75 t MS experiment design was particularly useful in understanding the MMPR to the different strategies for providing and using additional feed, whether grown or imported (Macdonald et al. 2017). Additional feed resulted in, on average, 0.8–1.2 kg of milk/kg of feed DM (or 73–97 g of MS/kg of feed DM); these ranges are consistent with published MMPR (Stockdale 2000; Bargo et al. 2003; Horan et al. 2005; Roche 2017). Some of the variation in MMPR (i.e. milk/kg feed DM) could be explained by differences in the ME density of the additional feed offered, some was a result of overcoming a nutritional imbalance associated with ME supplementation, but most reflected the size of the feed deficit.

Source of ME

Although MMPR per kg DM of additional feed varied with feed source, the source of ME (pasture, maize grain, or maize silage) did not affect the MMPR response to additional feed on an energy-equivalent basis. Cows, on average, produced 0.08 kg of milk/MJ of ME offered, irrespective of the feed source (i.e. the source of the ME); the effect of feed type was, therefore, a reflection of the amount of ME in each kg of supplementary feed. These results are consistent with previous nutrition experiments that reported only small effects of ME source on net energy output in milk in pasture-based dairy cows (Carruthers et al. 1997; Roche et al. 2010; Higgs et al. 2013).

Balancing the ration

There were exceptions to this simplistic approach to nutrition and these were evident in both the component experiment (Experiment 22) investigating the effects of protein supplementation in a pasture-maize silage diet and in the Balanced ration treatment in the 1.75 t MS experiment (Experiment 21). In these experiments, MMPR to supplementary ME was greater when a high-quality rumen by-pass protein source was provided in addition to the maize silage. For example, the addition of urea as a source of non-protein N had no effect on milk, milk fat or milk protein production in Experiment 22; milk production, therefore, was not limited by supply of rumen degradable protein. In comparison, however, soybean meal increased milk protein production by 0.06 kg/cow/day in both summer and autumn, while fishmeal increased milk fat production by 0.07 and 0.08 kg/cow/day in summer and autumn and increased milk protein by 0.06, 0.10 and 0.08 kg/cow/day in spring, summer and autumn, respectively (Macdonald et al. 1998). As the diets were near isoenergetic, the results indicated that the additional oil in fishmeal and the protein composition in both soybean and fishmeal provided some diet component that was inadequate in the maize silage supplemented cows. Similarly, the ‘Balanced ration’ treatment in the 1.75 t MS experiment resulted in ∼2% greater annual milk production/cow and per ha than the HSR-200-MG treatment, with all of this production difference occurring in summer.

The milk production responses to rumen bypass protein sources in Experiment 22 probably highlight an amino acid deficiency when maize silage is more than 50% of the cow’s diet, with the effect most evident in summer. During summer, lower protein concentrations in pasture as well as the supply of low protein, maize-based supplements, can result in a deficiency of the amino acid lysine (NRC 2001), which limits MMPR to additional energy. This is probably the reason for the greater response to fishmeal over soybean meal (NRC 2001) and the reason why the MMPR to supplements in the HSR-200N-BR treatment was greater than HSR-200N-MG during summer in Experiment 21. Nevertheless, these small increases in milk production would rarely justify the additional expense of lysine-rich ingredients, leaving the previous conclusion about ME source pertinent from the perspective of pragmatic implications. The small difference in milk production from balancing the ration is also consistent with the seminal work of Kolver and Muller (1998), wherein they were able to show that ∼90% of the difference in milk production between grazing cows and those being fed a total mixed ration in confinement did nnot rrelate to nutritional compositiion of the diet, highlighting that intake of ME was the primary reason for the difference in milk production between the two systems.

Extent of the relative feed deficit

The third and, arguably, more important reason for the variation in MMPR to additional feed was in the extent of the feed deficit. Penno (2002) introduced the concept of relative feed deficit, hypothesising that milk production responses to supplementary feeds increased as the size of the relative feed deficit increases. Poole (2018) confirmed this, reporting an increase in MMPR to supplements as post-grazing residual declined: a decline in post-grazing residual reflects a scenario where cows are grazing more severely and is, therefore, indicative of a greater relative feed deficit. These results are an acknowledgement that the slope of response curves in biology are rarely 1.0; in general, the largest response to alleviating a deficit is in the initial investment, with the response declining with each new addition: the law of diminishing returns.

In Experiment 21, Macdonald et al. (2017) reported the MMPR to additional feeds was strongly associated (i.e. r² > 90%) with CSR (Theme 5); the size of the MMPR to additional feed declined with declining CSR (i.e. severity of the initial feed deficit). The MMPR from reducing CSR from 95 to 86 (1.6 kg of milk/kg of pasture DM or 108 g of fat and protein/kg of pasture DM), for example, was 2- to 3-fold greater than the MMPR achieved from reducing CSR from 86 to 79 (0.5 kg of milk/kg of pasture DM or 55 g of fat and protein/kg of pasture DM). These differences reflect a decline in the milk production response to supplementary feed as the size of the feed deficit declines: to quantify this, MS yield increased by 55 (LSR-400N), 77 (LSR-200N), and 91 (HSR-200N corn grain) g/kg of additional feed DM (4.7, 6.6, and 7.3 g of fat and protein/MJ of ME, respectively) when the initial deficit was 0.5, 1.0, and 1.6 t of DM/cow, respectively. Importantly though, the source of the ME (i.e. pasture, maize grain (MG), or maize silage (MS)) did not materially influence the MMPR.

Growing more feed

In the mid-1970s, No. 2 dairy averaged 612 kg milk fat/ha/yr (∼1020 kg MS/ha) with young stock carried on the farm. Campbell et al. (1978) estimated that this was equivalent to 730 kg milk fat/ha allocated to the dairy herd (∼1250 kg MS/ha). They also estimated that about 90% of the pasture was being utilised and that ‘fine tuning’ of the No. 2 Dairy system would not increase production by more than 10%; it appeared that pasture production was setting the limit on MS production. If milk production per ha was to increase, feed availability/ha also needed to increase. This was investigated in two ways:

1. Planting an alternative feed that had greater DM yields/ha than the traditional perennial ryegrass-white clover pastures (Experiment 11); or

2. Through strategic use of N fertiliser to grow more pasture when supply was limiting production (see Theme 10).

In 1970, Campbell and Matthews undertook (Matthews 1971, 1972, 1973; ) component experiments on what they believed was a new and more highly productive dairy farm system, based on incorporating a maize crop for silage into the milking platform. Direct drilling of seed was being advocated to avoid soil disturbance and, hence, loss of N. The other advantage of direct drilling was that some clover was retained, providing some N for the resown pasture in autumn and the following spring. Maize was selected as it produced large yields (~18 t DM/ha) and, climatically, was very suited to most North Island dairy farms. Furthermore, as silage, it could be used at any time of the year, facilitating early spring supplementation, robustness to dry summers, and as an autumn feed to extend lactations or gain cow BCS. Their early component experiments led Campbell to investigate the benefits of growing maize on the farm without cultivation (direct drilled) in a full systems experiment. The experimental aims in Experiment 11 were, therefore, two-fold; the aim with the MS-20% system was to increase lactation length (DIM) by providing a greater amount of DM/ha in autumn-winter at the same SR, while the MS-50% system would further increase feed supply, allowing the dual benefits of greater SR and more DIM.

In Experiment 11, DM production from the maize (12–16 t DM/ha) was lower than was being achieved in commercial settings (~18 t DM/ha). This may have reflected ineffective weed control in the establishment phase, but was also because the small research paddocks (0.4 ha) created disproportionately large paddock edges, and water trough and gateway effects. Additional maize was, therefore, imported to increase DM yield to the equivalent of 18 t DM/ha, so that the true productivity responses to additional feed could be measured if the agronomic issues associated with direct drilling maize could be overcome. This change to the experimental design allowed the hypothesis to be tested, despite the sub-optimal maize yields, and dissociated the animal production objectives of the experiment from the need to optimise agronomic practices.

With an assumption that approximately 40% of the annual pasture DM yield (16 t DM) was achieved on the maize silage area, both before tillage and in the annual ryegrass autumn production post-maize harvest, and DM yields of maize for silage was 18 t DM/ha, then 16, 18, and 20 t DM/ha of available feed were expected in the Control, MS-20%, and MS-50% treatment, respectively. The proportion of maize silage in the cows’ diets was different in the three treatment groups throughout the year. The Control, MS-20% and MS-50% diets, respectively, were 0%, 40%, and >50%, 0%, 40%, >60%, and 0%, 15%, and 25% to 30%, maize silage in summer-autumn, winter, and spring periods, respectively.

Annual MS production was 314, 332, and 302 kg MS/cow and 962, 1017 and 1100 kg MS/ha for the Control, MS-20% and MS-50% farmlets, respectively; this was equivalent to 1083, 1145, and 1239 kg MS from the milking platform (i.e. removing young-stock area). Based on 2 and 4 t DM extra feed/ha, on average, in the MS-20% (+62 kg MS/ha) and MS-50% (+156 kg MS/ha) farmlets, this equated to a MMPR of 37.2 kg MS/t DM additional feed. This is a relatively small response to supplements and probably reflects a small relative feed deficit in the ‘Control’ treatment (feed allowance of >4 t DM/cow). This is confirmed by the small extension in lactation length (9d) in MS-20% herd compared with the ‘Control’. Days in milk/ha were increased by 17% in the MS-50% treatment, but this was due to the 19% greater SR and not an increase in lactation length/cow. The objective of substantially extending lactation, by adding in large inputs of maize silage was, therefore, not achieved.

Environmental sustainability measurements associated with intensification

The impact of SR on nitrate leaching was summarised in Theme 5, with nitrate leached declining linearly with increasing SR, primarily, it is hypothesised, because of the reduction in lactation length and the associated N surplus in the cows' diet during autumn (Shepherd et al. 2014; Roche et al. 2016). Supplementing cows on the 360 kg N fertiliser treatment with 1 t DM of maize grain did not reduce nitrate loss. Importantly, the analysis did not account for any nitrate loss in the cropping and harvesting activities, nor any 'fallow' period associated with cropping for maize grain.

A 33% increase in SR increased GHG emissions/ha (t CO₂-eq) by ∼8%, because there was no increase in feed; this effect merely represents a decline in per cow energetic efficiency, with a greater proportion of ME used for maintenance than for productive purposes. If, however, SR was increased, but feed allowance/cow kept similar through the importation of supplementary feeds, GHG emissions/ha (t CO₂-eq) increased by 29% and 46% for maize silage and maize grain treatments, respectively, not including soil C lost through tillage.

Economics of providing additional feed into the pasture system

The topic of whether the intensification of the farm system away from a sole reliance on pasture and silage/hay and towards increased inclusion of supplementary feeds leads to greater profit has been controversial. Recognition that cows can produce more milk when offered more feed and partial budgets that include revenue, feed, and the cost of feeding appear to justify the system intensification, but analyses of national farm databases (Ramsbottom et al. 2015; Ma et al. 2018; Neal et al. 2018; Neal and Roche 2019) highlight either no economic gain at the farm level or a reduction in EBIT or Return on Assets (RoA). The systems experiments undertaken at No. 2 Dairy were to explain this apparent contradiction in outcomes of different economic analyses.

Because grain was so expensive relative to milk in the 1950s and 1960s, early analyses (Wallace 1957) suggested that any revenue gain from supplementing cows with grain-based concentrates was insufficient to justify the costs of purchasing feeds. Furthermore, farm system experiments in which supplementary feed was used to increase cow DMI in early lactation did not result in increased lactational milk production. It was hypothesised, however, that low cost, direct drilled maize for silage grown on the milking platform and offered to cows throughout the year could potentially be justified. Milksolids/ha and, therefore, milk revenue/ha increased by 6% and 15% in the MS-20% and MS-50% farmlets, respectively, but EBIT declined by 4% and 10% ($903 to $871 and $816 for the Control; MS-20% and MS-50% farmlets, respectively; Campbell et al. 1978). Although only three maize silage inclusion levels (0%, 20% and 50%), the results indicated a linear decline in profitability with increasing reliance on maize for silage; for every 10% of the farm area used to grow maize for silage, milk revenue increased 3%, but EBIT declined 2%.

The results questioned the validity of the partial budgeting approach to evaluating practice change. Low production responses, possibly indicative of too low a SR to take advantage of supplementary feeding and greater operating expenses may have eroded the revenue advantage of intensification of the farm system. However, the studies, until now, were not designed in a way that allowed the reasons for the disparity between partial budgeting and farm systems analysis to be unpicked. This was to change with Experiment 21: the 1.75 t MS experiment.

The design of the 1.75 t MS experiment allows us to unpick the effects of N fertiliser (Theme 10), SR, and supplementary feeding in a way that no prior experiment facilitated. A full farm systems economic analysis of this experiment was reported by Macdonald et al. (2017). Comparing:

• LSR-0N, LSR-200N, and LSR-400N and LSR-0N, HSR-200N, and HSR-400N allowed them to establish the cost-benefit of N fertiliser, with or without an increase in SR (see Theme 10);

• LSR-200N, HSR-200N-MG, and HSR-200N-MS and HSR-200N, HSR-200N-MG, and HSR-200N-MS allowed them to estimate the cost of marginal milk from feed intensification.

Cost–benefit of feed intensification

The design of the experiment was such that supplementary feeds provided to the HSR treatments resulted in a CSR similar to the LSR-200N treatment; this allowed Macdonald et al. (2017) to investigate the cost of marginal milk produced from using supplements to fill feed deficits (i.e. in the HSR-200N treatment) and the cost associated with using supplements to intensify the farm system (i.e. against the LSR-200N) treatment. Serendipitously, this uncovered, at least in part, the reason behind the vexing discrepancy between partial budgets and the whole farm analyses.

Macdonald et al. (2017) reported that the cost of a kg of marginal MS was NZ$6.33 and NZ$5.54 when HSR-200N-MG and HSR-200N-MS, respectively, were compared with HSR-200N, reflecting the cost of marginal milk when supplementary feeds are used to fill genuine feed deficits and extend lactation from 217 days to 280 days. However, the cost of the marginal milk was NZ$7.97 and NZ$7.91, respectively, when these treatments were compared with LSR-200N; this, therefore, is the cost of intensifying the system: using supplements to increase SR, only marginally increasing lactation length (i.e. from 266 days to 280) and milk production/cow (357, 396, and 359 kg MS in LSR-200N, HSR-200N-MG, and HSR-200N-MS, respectively). Despite the MMPR to supplementary feeds being identical (i.e. 0.08 kg of milk/MJ of ME offered), irrespective of whether the feed was used to feed cows better or to increase SR and maintain feeding levels, the cost of the marginal milk was 26% to 43% more when the supplementary feed was used to increase SR compared with when supplementary feed was used within a system to increase DIM and the milk production of existing cows.

Conclusions and implications

Milk production responses to supplements offered throughout lactation were consistent with international studies, irrespective of whether they were offered in a high-stocked system to increase feed DM allowance/cow or used to increase SR while maintaining feed DM allowance/cow. When used in a feed deficit scenario, responses could be economic, particular with reasonable milk to feed price ratios (e.g. 1.5; i.e. price of 1 kg milk is greater than the price of 1 kg DM supplement). Responses were uneconomic, however, when:

1. the MMPR to the supplements used are low (e.g. in spring), primarily because they replace pasture already grown (i.e. substitution), which is therefore wasted; or

2. when the SR of the farm is near optimal for pasture management (i.e. 85-90 kg Lwt/t DM pasture grown), as the additional fixed costs associated with the increased SR (Experiment 10; Experiment 21) significantly increased the cost of the marginal milk.

The experiments undertaken at No. 2 Dairy, Ruakura, to investigate the effect of intensifying pasture-based dairy farm systems by using supplementary feeds to strategically fill feed gaps or increase SR have provided a plausible reason for the difference between results of partial budgets and whole farm analyses. Responses to supplements in early lactation are small and can be dwarfed by effects later in lactation that are outside the control of the farmer. Whole farm analyses generally involve analysing a system that has used supplementary feeds to both fill feed shortages and increase SR (Ramsbottom et al. 2015; Ma et al. 2018). The cost of this milk is different. Strategically using purchased feed to increase milk production during periods of feed deficit results in marginal milk of moderate cost, often less than the milk price and, therefore, profitable. However, using purchased feeds to increase SR, introduces all of the costs associated with the additional animals as well as the feed and feeding costs (∼60% of the costs in a pasture-based dairy system are per cow costs; Macdonald et al. 2011). This marginal milk is, therefore, substantially more expensive. These additional cow costs, however, are generally excluded from partial budgets, misleading farmers and advisers as to the revenue-cost implications of system change.

Finally, the experiments undertaken highlight the high quality of fresh pasture as a feed ingredient, with increased ME intake resulting in greater milk production, irrespective of the source of the energy. Roche (2017) discussed this in great detail. The exception to this is when a large proportion of the cow’s diet is provided as low crude protein-high energy supplements, particularly in summer. In these situations, the primary dietary factor limiting production switches from ME to amino acids (e.g. lysine if the supplementary feed of choice is maize-based) and the milk production response to the supplement will increase if this limitation is removed. Such an intervention, however, is generally cost-prohibitive in a pasture-based system.

A wise message was provided by Campbell et al. (1978) in the final paragraph of the paper outlining the results of Experiment 10:

There is still ample scope for most dairy farmers to increase production from pasture alone by employing higher SRs and improved management. Those already in the 500 kg milk fat/ha (∼860 kg MS/ha) and over bracket should, for the time being, view cropping systems with caution, having due regard to the net returns.

The results of the 1.75 t MS experiment suggest a similar caution should be used for purchased supplementary feeds.

Theme 7: ‘cows for courses’ – selecting the correct cow for grazing systems

Background

Although milk recording services became available in many countries in the early twentieth century (Weigel et al. 2017), there was poor uptake and most ‘dairy cattle’ possessed dual-purpose characteristics to supply both replacement female calves for milk production and bull calves with good beef characteristics (Stewart 1946; Roche et al. 2017a). In fact, according to some commentators, the lack of genetic improvement in production stock was identified as ‘the one factor wherein the charge that farmers are careless in business was justified’ (Anonymous 1924); for example, 50% of champion milk producers in New Zealand were reportedly purchased for a fraction of their value because the previous owner did not milk record (i.e. measure milk yield and sample composition; Anonymous 1924).

‘Herd-testing’ (i.e. milk recording) became an official national scheme in 1936 in New Zealand (Atkins 2016) and farmers were encouraged to ‘milk record’ to identify the ‘top-notch’ and ‘boarder’ cows: the term ‘boarder’ was used for a cow that ‘boarded’ but provided very little value in the herd (Anonymous 1924; Roche et al. 2017a). However, Stewart (1946) reported that the failure to improve cow productivity over the previous decade was due to ‘the quality of the animals used for breeding purposes’. He suggested that farmers needed to select ‘the animals to be parents’. The selection of the dam was within the farmers’ control and was aided, quantitatively, by milk recording; but, genetic improvement was still restricted, as the ability to ‘share’ superior sires among farmers was logistically restrictive and a biosecurity hazard (Moore and Hasler 2017).

During the 1930s and 40s, considerable research was undertaken to overcome this problem, through optimising the collection, dilution, and cryopreservation of sperm, and AI. Used in conjunction with herd-testing, AI could facilitate the use of one bull across tens of thousands of cows, providing the platform for rapid genetic gain and dairy herd expansion (Roche et al. 2017a). In addition, cryopreservation meant that superior genetics could, potentially, be sourced from around the world. Despite the technological improvements, however, and the accumulating evidence that similar pregnancy results were achieved with AI as with natural mating (James 1941), uptake of AI was slow in New Zealand, partially because of a farmer preference for ‘line-breeding’ (i.e. continuing to breed from within a line of stock; inbreeding). By the 1950s, the advantages of ‘outbreeding’ were becoming accepted (Hancock 1952), with scientists and breeders conceding that inbreeding, except in a very mild form, had ‘detrimental effects on most characters of economic importance’.

Outbreeding can occur by the crossing of inbred lines within a breed (i.e. sourcing genetics from other herds, but within a breed) or by crossbreeding (i.e. crossing one dairy breed with another). During its 60-year tenure, No. 2 Dairy was to be the home of several experiments comparing different breeds and strains within breeds in different environments. The results were critical to understanding the importance of genetic × environment interactions in grazing dairy systems and in the advancement of herd improvement nationally. These experiments are summarised here.

The experiments – (Experiment 5, 10, 14, 19, 26). Lifetime project, grazing method, SR and genetics

Experimental methods

The first experiment (Experiment 5) to investigate the interaction between cow genetics and environment was previously introduced in Theme 1. It was undertaken over three full lactations and involved a 2 × 2 × 2 factorial arrangement of treatments: the interaction between grazing strategy (Rotational Grazing vs Set Stocking), SR (2.35 vs. 2.95 cows/ha), and cow genetics (High vs. Low Genetic Merit) were evaluated (Carter 1964). Half of the cows in the experiment were sourced from commercial dairy herds that used milk recording, but had not used AI (i.e. low-genetic-merit cows; LGM). These cows were compared with cows from the research station, where cows had been bred by AI to superior sires for more than 15 years (i.e. high-genetic-merit cows; HGM). The objectives of the experiment were four-fold (Carter 1964):

1. To continue the previous comparisons of the two extreme types of grazing management under two SRs;

2. To obtain a measure of the actual superiority of AI-bred cows over those naturally mated, but herd recorded;

3. To test whether the results derived to-date applied to ‘the average’ dairy cow; and

4. To determine whether the advantage of the superior genetics was dependent on feeding and management.

The latter two objectives were to test whether genetic merit × environment interactions were sufficiently material to be considered by farmers in their selection of genetics for their farms.

A second experiment undertaken at No. 2 Dairy investigated the effect of cross-breeding, as a way of accelerating genetic changes (Experiment 10). This experiment was established by Archie Campbell in 1972 as an extension of Experiment 9 (integrating dairy and beef breeds, Theme 8). He compared the effects of changing from Jersey to Holstein-Friesian-Jersey crossbred on milk production/cow and per ha. Two farmlets were established, one stocked with Jersey cows (n = 42) and the other with Holstein-Friesian-Jersey crossbred cows (n = 42; Campbell 1977) and were randomly allocated to one of the two breeds. Each farmlet was 12.1 ha, such that cows were stocked at 4.12 cow equivalents per ha (3.38 cows + 0.74 cow equivalents as replacement stock). The experiment was undertaken for five successive years, with replacements managed within their parent farmlet (i.e. Jersey heifers within the Jersey farmlet and Crossbred heifers with the Holstein-Friesian-Jersey Crossbred farmlet).

A third experiment (Experiment 14) was established in 1986 to investigate the effect of pasture allowance on cow production. This was a short-term component experiment undertaken by Phil L’Huiller, comparing four herds of Jersey cows (12 cows/herd) with four Holstein-Friesian herd comparisons. Herds were offered nominal herbage allowances of 10, 20, 30 or 40 kg DM cow/d during a 21-d period in early to mid-lactation (October-November). This experiment was supported by an energy metabolism experiment undertaken in Open Circuit Calorimeters at No 5 Dairy Ruakura, in which six Jersey and six Holstein-Friesian cows were offered a maintenance allowance of pasture (L’Huiller et al. 1988).

A fourth experiment was established by Arnold Bryant in 1990 (Experiment 19) to investigate the interaction between cow breed and SR. This three-year experiment compared Holstein-Friesian and Jersey cows at two SR and was published in full by Spaans et al. (2019). The experiment recognised that the breeds differed phenotypically in Lwt (420 and 360 kg average Lwt/cow for Holstein-Friesian and Jersey breeds, respectively) and, therefore, in maintenance requirements and, probably, in intake capacity. To account for these potential confounding features, cow Lwt/ha was equivalent for both breeds at the two SRs (i.e. SR = 1270 and 1650 kg Lwt/ha), but cows/ha differed between breeds: SR was 3.0 and 4.0 cows/ha for the Holstein-Friesian farmlet and 3.6 and 4.5 cows/ha for the Jersey farmlet.

This was one of the first farm systems-level experiments globally to recognise that cow Lwt was an important component of DMI and production potential, adjusting SR for Lwt/ha parity. This cow characteristic was later incorporated into CSR (kg Lwt/t DM feed available; Theme 5; Macdonald et al. 2008a), to improve the comparability of experiments undertaken with different breeds, different within-breed genetic strains, and, importantly, differences in pasture DM production and supplementary feed availability. Utilising that metric, the Holstein-Friesian and Jersey treatment herds were established with a CSR of 80 or 100 kg Lwt/t DM feed available.

The four farmlets each consisted of 18 paddocks of 0.41 ha; herds within breeds were balanced for expected calving date, age, Lwt, BCS, and EBV for traits in the national selection index at the start of the experiment. Mean calving date was the first week in August and the mean breeding index (BI; the breeding worth calculation of the day) score was 136. The objective of the experiment was to investigate the interaction between cow breed and SR on pasture productivity, per cow and per hectare production performance, and reproductive outcomes within a closed farm system.

A fifth experiment investigating the effect of genetics on farm productivity and profitability began at No. 2 Dairy during its final years as an experimental station (Experiment 26) under the stewardship of John Penno and Kevin Macdonald, before being transferred to Scott Farm on Vaile Road, a short distance from the No. 2 Dairy site. Growth rates, onset of reproductive maturity, milk production, reproductive function, and economics were evaluated in three divergent strains of Holstein-Friesian between 2001 and 2005 (McNaughton et al. 2002; Macdonald et al. 2007, 2008b; Lucy et al. 2009). The strains included a 1970s (low BW) strain of NZ Holstein-Friesian (NZ70), a high BW Holstein-Friesian strain of NZ origin (NZ90), and a high BW Holstein-Friesian of North American or Dutch origin (NA90), all managed at a range of feed allowances following their first calving (Macdonald et al. 2008b). In addition, a companion study (Horan et al. 2005, 2006) was simultaneously undertaken at the Moorepark Research Centre in the Republic of Ireland, comparing three strains of Holstein-Friesian, representing animals of Irish, New Zealand (similar to NZ90), and NA (similar to NA90) origin.

In New Zealand, NZ Holstein-Friesian (NZ90) and North American Holstein-Friesian (NA90), with a similar genetic merit at the beginning of the experiment, as estimated using the National Breeding Objective (BW: $86 and $84 for NZ Holstein-Friesian and North American Holstein-Friesian, respectively), were compared in a multi-lactational whole farm system comparison. Both strains were compared under a range of feed allowances (i.e. between 5.5 and 7.0 t DM/cow; Macdonald et al. 2008b), allowing them to test the ‘old chestnut’ previously considered by Carter (1964): is the performance of a genetic strain dependent on its feed allowance? The experiment in Ireland was similarly designed, although with fewer feed allowances considered. The New Zealand experiment, known colloquially as ‘The Dexcel Strain Trial’, was to have an additional advantage; a strain comparison representative of the genetics available in the early 1970s (NZ70). This unique comparison enabled the quantification of 25 years of genetic progress and the advantages that husbandry changes over the intervening period had provided.

Results and discussion

An old proverb states that ‘an ounce of breeding is worth a pound of feeding’, highlighting the importance and permanence of animal quality in farm productivity (Roche et al. 2017a), but also the importance of choosing the correct animal for the situation (i.e. ‘horses for courses’). McMeekan supported this focus on choosing the appropriate cow for the system (i.e. genetic selection), when he recommended using ‘animals that will process the grass efficiently’ (McMeekan 1960). But change was slow. As a result, by 1960, cows at No. 2 dairy were significantly superior to the average New Zealand dairy cow (see Theme 1). This undermined the extension of some research results, as farmers believed that the findings didn’t apply to their herd. An experiment to investigate the interaction between grazing strategy (i.e. Rotational Grazing or Set Stocking), SR, and cow genetic merit was to change all that (introduced in Theme 1).

Intuitively, different systems challenge different cows in different ways. Grazing cows are required to walk long distances compared with housed cows, and spend more time consuming their daily feed, leaving less time for lying (Hendriks et al. 2019). Housed cows are rarely restricted in DMI, whereas pasture-based cows may, occasionally, be restricted in feed supply or in their ability to consume allocated feed, due to the variability in climate (Roche et al. 2009a, 2009b, 2009c, 2009d, 2017a, 2017b; Sheahan et al. 2011). To capture the potential from the cow, a housed cow may be milked three- to six-times a day (Hale et al. 2003); in a pasture-based system, cows are generally milked twice a day and occasionally only once a day. Finally, the grazing system is defined around a biological clock; calving is timed to coincide with the onset of spring growth and peak herd DMI with peak pasture supply. These factors mean that the animal of choice for a New Zealand dairy system is quite different to that of other countries: the cow must be mobile, able to graze pasture effectively and, importantly, get in calf within ∼80 days of calving to ensure a calving every 365 days (Holmes 1995; Roche et al. 2017b). During its 60-year tenure, experiments undertaken at No. 2 Dairy Ruakura quantified the effects of genetic change on production, reproduction, and profit. These were pivotal in supporting the uptake of artificial breeding technologies, herd testing, and in providing direction on the functional traits to include in the national breeding objectives.

Putting the advantage of AI over natural breeding beyond doubt

Although the importance of cow milk yield (i.e. genetic merit for production) was recognised in the early twentieth century (Anonymous 1924) and a national Herd Testing Service was widely used by the mid-1930s (Stewart 1946), New Zealand farmers failed to take advantage of AI in accelerating genetic progress. Failure to adopt AI not only reduced farm productivity and the productivity of the national herd, but also, ironically, led many to question the relevance of research results for their farms (McMeekan 1960), because they recognised the superiority of the cows being used in experiments at No. 2 Dairy. The investigation of genetic merit (AI versus natural mating) × environment (SR and grazing strategy) interactions (Carter 1964) was to change this, by:

• quantifying the size of the lost opportunity farmers incurred by failing to adopt AI; and

• highlighting that the productivity improvements associated with greater SR and rotational grazing occurred irrespective of cow genetic merit (see Theme 1).

The experiment compared cows from No. 2 dairy, which had used AI intensively for 15 years, with cows from local dairy farms that herd tested and, therefore, presumably selected replacement animals from ‘top-notch’ cows, but had not availed of AI. The design, therefore, separated out the effect of AI (i.e. bull selection) from the herd testing of dams.

Across high and low SRs and different grazing strategies, both known to affect cow milk yield, the HGM herd produced, on average, 25 kg more milk fat/cow and 66 kg more milk fat/ha (Table 10) than the LGM herd (+17%). This equated to ∼1.5 to 1.8 kg/yr increase in milk fat production/cow for each year of using AI. This was very similar to the predicted BI differences estimated from the known ratings of the sires entering into the pedigrees of the Ruakura cows (Carter 1964; Wickham et al. 1978), providing significant support for the development of a national BI.

Table 10. The effects of grazing strategy (Controlled vs Uncontrolled), Stocking rate (Low: 2.35 cows/ha vs High: 2.95 cows/ha), and cow genetic merit (HGM vs LGM) on milk fat production/cow and per ha and the % increase in milk production from genetic merit, stocking rate, and grazing strategy (Experiment 5).

	CL	CH	UL	UH
Per cow
HGM	189	184	170	149
LGM	162	155	148	127
Difference	26	29	21	22
Per ha
HGM	443	543	398	440
LGM	381	457	348	375
Difference	62	86	50	64
%increase/ha
GM	16%	19%	14%	17%
SR		39%		39%
Grazing strategy			23%	33%

The experiment, however, had a much more important evidentiary effect than, merely, quantifying the effects of AI on cow production and supporting the BI. At the time, farmers questioned whether the superiority of genetics was maintained across different farming systems and regions or whether a genetically superior animal could only express their genetic advantage in a good environment. The inference was, therefore, that in low-producing herds, where presumably the nutritional or management environment is below optimum, dairy merit was less important (Carter 1964; Wickham et al. 1978). If such a genetic × environment interaction existed, how could a farmer have confidence that a bull evaluated as superior in the Waikato was the most suitable sire for cows in Northland or Southland?

The complexity of the experimental design allowed Carter to explore whether feed allowance (i.e. SR) or timing of feed surpluses/deficits (i.e. grazing strategy), not to mention the unreported effects of these system-level factors (e.g. BCS, parasite burden, etc.; see Theme 1 and 3), interacted with genetic merit and undermined the value proposition from using AI. Although HGM cows produced more milk fat/cow, irrespective of SR (i.e. feeding level) or grazing strategy (i.e. consistency of feed availability; Table 10; Carter 1964), the effect of genetic improvement was greater under rotational grazing: HGM cows produced 28 kg more milk fat/cow under rotational grazing, compared with 22 kg in the Continuous Stocking treatment (+28% more milk fat under rotational grazing). This was particularly evident in the High SR treatment, as HGM produced 23% more milk fat/cow and per ha under rotational grazing in the Low SR treatment, but 33% more milk fat/cow and per ha under rotational grazing in the High SR treatment (Table 11). This was the first quantitative assessment of the interaction between genetics and the environment for pasture-based production systems and provided confidence in a national BI (developed in 1974, replacing sire ratings) and the transferability of research results beyond the farming system in which it was undertaken. Following the presentation of results, uptake of AI by farmers increased swiftly, with the number of cows inseminated increasing from ~17% in 1958 to ~45% in 1968 (Anonymous 1981).

Table 11. Principal results summarising the effects of stocking rate and breed on pasture and animal production variables and economic outcomes (Experiment 19; from Spaans et al. 2018).

Breed	Jersey		Holstein-Friesian			P-value
Stocking Rate	CSR80	CSR100	CSR80	CSR100	SED	Breed	CSR	Breed × CSR
Pasture production, kg DM/ha	17,267	16,214	16,273	13,424	533.9	<0.01	<0.01	0.05
Lactation length, d	262	223	261	226	3.3	0.64	<0.001	0.38
Pasture disappearance, kg/cow/yr	4340	3590	4856	3541	88.8	<0.01	<0.001	<0.01
Supplementary feed, kg/cow/yr	154	159	160	167	7.5	0.24	0.3	0.88
DMI, kg/cow/yr	4494	3749	5016	3708	87.9	<0.01	<0.001	<0.01
4% FCM, kg/cow/yr	4265	3520	5000	3635	69.1	<0.001	<0.001	<0.001
Fat, kg/cow/yr	198	163	211	153	3.2	0.6	<0.001	<0.01
Protein, kg/cow/yr	136	108	161	115	2.7	<0.001	<0.001	<0.01
Lwt post-calving, kg	364	356	450	416	5.8	<0.001	<0.001	<0.05
BCS at calving, 1–10 scale	4.9	4.8	4.9	4.6	0.06	0.16	<0.05	0.17
Calf birth weight, kg	25	26	36	36	1.2	<0.001	0.95	0.72
Total ME/cow, MJ	42,636	37,530	49,308	40,946	413.4	<0.001	<0.001	<0.01
Total ME/ha, MJ	152,230	168,884	147,923	163,783	1834.40	<0.01	<0.001	0.77
DIM to first heat	28	28	38	56	4.4	<0.001	<0.05	<0.05
Non-cyclers^3 (%)	3	1	18	13	2.8	<0.001	0.11	0.35
Submission rate^4 (%)	97	96	98	89	3.5	0.25	0.07	0.15
Conception (%)	60	62	68	64	5	0.21	0.72	0.43
Services/conception	1.6	1.6	1.5	1.5	0.12	0.36	0.94	0.88
AI pregnancy rate	74	78	85	75	6.3	0.46	0.5	0.18
In Calf (%)	91	94	98	91	4.3	0.5	0.5	0.14
Milk price	$5.88	$5.85	$5.97	$5.96
Gross Farm Revenue	7522	7767	7216	7005
Total Dairy Operating Expenses	4186	4891	3710	4515
Dairy Operating Profit	3336	2876	3505	2490
Expenses – cost/kg fat and protein	3.51	3.97	3.29	4.24

Another important point from this study was the additive effects of investment in genetic merit. Carter (1964) reported that the production advantages from AI sires increased with time. In other words, the benefit of genetic improvement was permanent (i.e. the farmer only paid for the advantage once for each animal but reaped the benefits in each lactation of the cow’s life) and, importantly, its value compounded with the production of heifer calves that entered the herd with even more superior genetics.

In summary, the No. 2 experiment quantified the production gains derived from using daughter-proven superior sires and AI, highlighting the production foregone by progressive farmers (i.e. those that were using herd testing) that failed to adopt AI for breeding replacement animals. It also demonstrated that interactions between genetic merit and environment, although evident, were ‘not of primary importance’ (Carter 1964), at least within the practical range of dairy farming conditions in New Zealand at the time (i.e. before the importation of germplasm from abroad). The latter results, however, were to be challenged in the ‘Dexcel Strain Trial’ undertaken almost 40 years later, when semen imported from very different production systems was found to have unintended consequences on the biophysics and economics of grazing systems.

What breed should I milk?

The dominance of breeds and breeding objectives changed in New Zealand over the 60 years that No. 2 Dairy was in operation. Although the original dairy cows introduced to New Zealand were of the Shorthorn breed (Peden 2008), by the time the Ruakura research station was established, more than 85% of dairy cows were purebred or ‘grade’ Jersey cattle (Jhala 1952). Cheese was the first milk product exported (in 1846, to the UK), followed by butter, providing value for a cow that produced high solids and, in particular, fat.

The Jersey breed, therefore, continued to dominate the dairy cattle breed statistics until the 1960s (Anonymous 1981), when the bulk collection of whole milk, instead of fat, became more common, and the recognised value of the non-fat solids in milk led to a shift towards Holstein-Friesians through crossbreeding. For example, in 1962/63, Jersey and Holstein-Friesian cows constituted 81% and 12% of the national herd, respectively; but, within 15 years, the purebred Holstein-Friesian breed overtook Jerseys in national breed statistics, with 46% Holstein-Friesian in 1977/78 compared with 45% for Jerseys (Anonymous 1981), and this trend was to continue to the 1990s (Harris and Kolver 2001).

With the introduction of Holstein-Friesian genetics and the recognition that genetic change took time and was not easy to reverse, there were questions about whether a crossbred or purebred animal would be superior in dairy systems. A series of experiments at No. 2 Dairy was to establish the value of the crossbred and purebred options, with Campbell investigating the transition from Jersey to Holstein-Friesian (Campbell 1977), to identify if there were advantages to be had in farming crossbred cows or purebred, and Bryant comparing Jerseys with Holstein-Friesians under different SRs (Spaans et al. 2019).

Comparison of Jersey and Holstein-Friesian-Jersey crossbreds

On average, Jersey × Holstein-Friesian crossbred cows were 46 kg heavier than similar age Jersey cows; this was exactly half the difference in Lwt between purebred Jersey and Holstein-Friesian cows reported by Macdonald et al. (2005). The results, therefore, are mathematically consistent with the average of the two breeds. Profiles of Lwt change through lactation were similar for the two herds, except that the crossbred 2 yr old heifers grew more (i.e. increased more in Lwt) in their first lactation than their Jersey contemporaries. It was suggested that this may be because the Holstein-Friesian animal is a later-maturing animal than the Jersey and, therefore, the crossbred of the two breeds would increase in size more in its first lactation. However, a comparison of growth characteristics for both breeds by Macdonald et al. (2005) identified no such delay in growth in the Holstein-Friesian animals. The difference could reflect a nutritional disadvantage in the larger breed as young stock, with compensatory growth during the first lactation well-established in grazing cattle (Macdonald et al. 2005).

Lactation lengths were not different between breeds (247 days), but the Crossbred cows produced 16% more milk in all years and age groups (+424 kg milk/cow/year). Jersey milk was, on average, 0.5 percentage units higher in milk fat, but 0.2 percentage units lower in protein than the crossbred milk. There was no effect of breed on lactose %. The crossbred cows produced ∼5% more milk fat than their Jersey contemporaries (+7 kg). A simple economic analysis indicated no advantage to crossbreeding during the transition: -gross -margin was $765 and $770/ha for the Jersey and the Jersey/Friesian crossbred farmlets, respectively, with little to no difference in costs because of the same stocking rate/ha.

The results did indicate, however, an interaction between the breeds when in the same herd. The Jerseys in the crossbred herd (Base Jerseys) were lighter than their contemporaries in the Jersey herd and produced less milk fat. Because of this, whole-herd fat productions were not different (5-year average: 164 and 162 kg milk fat/cow for the Jersey and Holstein-Friesian-Jersey crossbred herd, respectively), despite the milk fat advantage of Crossbred cows over Jerseys. The results identified an important point of between breed competition: the Jerseys were at a disadvantage when grazing with the Jersey × Holstein-Friesian crossbred cows. The competitive characteristics of the Holstein-Friesian will be discussed again in the Holstein-Friesian-Jersey comparison experiment.

It was concluded that crossbreeding resulted in no financial advantage during the transition between a pre-dominantly Jersey and Holstein-Friesian herd. This does not mean that crossbreeding does not result in advantages. Depending on ‘failings’ of the ‘local breed’, crossbreeding can be used to quickly improve cow characteristics, such as production, health or reproduction. For example, Prendiville et al. (2009) and Coffey et al. (2016) reported significant evidence for hybrid vigour in the cross of Holstein-Friesian and Jersey breeds for milk yield, solids-corrected milk (SCM), fat and protein yield, Lwt, BCS, and the net energy intake per kg fat and protein produced. Buckley et al. (2014) reported 19% more Holstein-Friesian × Jersey crossbred cows pregnant after 6-wk breeding than purebred Holstein-Friesian cows. Similarly, an analysis of the New Zealand national herd highlighted benefits in reproduction in crossbred cows when compared with purebred Holstein-Friesian (Jayawardana et al. 2023). Such advantages can lead to short-term economic superiority of crossbreds over cows from either pure breed (Lopez-Villalobos et al. 2000), but the heterosis advantage cannot be sustained. The recent crossbred animal in New Zealand has better fertility outcomes than the Holstein-Friesian from which she is derived, once again changing the makeup of the national herd from predominantly Holstein-Friesian in 2000 to 50% Holstein-Friesian × Jersey crossbred in 2020.

In these recent examples, however, the transition was from Holstein-Friesian to crossbred. Campbell (1977), however, reported on the transition from Jersey, in which there was no evident fertility, health, or longevity issues, to Holstein-Friesian, and did not uncover evidence of hybrid vigour and, in fact, the Holstein-Friesian competitiveness appeared to disadvantage the purebred Jerseys remaining in the herd. The question remained, therefore, was whether the transition to a Holstein-Friesian herd from the Jersey was a wise decision, if there was no transition-period advantage to the change and it was subsequently identified that Holstein-Friesian genetics imported from North American and Europe would lead to significant disadvantages. Experiments at No. 2 dairy were to provide answers to these questions.

By the 1980s, No. 2 dairy had transitioned from Jerseys to Holstein-Friesian cows and farmers were curious about which breed was most efficient. To provide some guidance, Bryant et al. (1985) compared Holstein-Friesians at No. 2 Dairy and Jerseys at the adjoining research farm (No. 5 Dairy). Jerseys at No. 5 Dairy were in an experiment investigating the effect of BI on farm productivity and as such were managed at a range of SRs (2.8 to 4.3 cows/ha). Regression analysis was used to derive estimates of the performance for Jersey cows at the same SR and Lwt/ha as the Holstein-Friesian cows at No. 2 Dairy.

The analysis identified that at a common SR of 3.7 cows/ha, Holstein-Friesian cows outproduced Jersey cows by 7%, 15% and 13% more fat, protein and SCM, respectively. At the same Lwt/ha, it was estimated that the Holstein-Friesian cows outproduced Jersey cows by 3%, 13% and 10% for fat, protein and SCM, respectively. It was concluded that Holstein-Friesian cows were superior due to a higher feed conversion efficiency at the same DMI. The comparison, however, received a lot of criticism, mainly because two of the Holstein-Friesian sires were disproportionately represented in the Holstein-Friesian herd. So, the differences identified could be more of a sire effect than an effect of breed. This criticism and uncertainty about possible breed differences resulted in a short-term experiment being established by L’Huiller (Experiment 14) to investigate aspects of this, and a long-term farm systems experiment comparing the two breeds, which took place at No. 2 dairy from 1990 to 1993 (Experiment 19).

Comparative performance of Jersey and Friesians in mid-lactation

L’Huiller established a short-term experiment in mid-lactation to better understand the interaction between dairy cow breed and feeding level; the experiment consisted of a 2 (breed: Jersey and Holstein-Friesian) × 4 (feeding level: 10, 20, 30, and 40 kg DM allowance/cow/day) factorial arrangement of treatments. Holstein-Friesian cows produced 26% more milk, 6% more milkfat, 13% more protein and 24% more lactose. But at the lowest level of feeding (10 kg DM allowance/cow/d) the Holstein-Friesian cows produced more milk but less milk fat and protein than the Jersey cows. The Holstein-Friesian cows consumed more pasture DM (average +13%), lost more Lwt, and had a lower feed conversion efficiency (61 v 67 g milk fat/kg DM consumed) than Jersey cows. It was calculated that milkfat/ha, at a common Lwt was lower for the Holstein-Friesian breed than Jerseys, particularly at high SR.

The results suggested that, in mid-lactation, the Holstein-Friesian breed can produce more milk and MS per cow than the Jersey breed, because of their greater feed intake ability. In contrast, however, Jersey cows had a higher feed conversion efficiency, greater utilisation of ME for milk production, and a greater DMI per kg metabolic Lwt. This meant they could potentially produce more MS per hectare than the Holstein-Friesian cows. The improvement in the efficiency of ME use, however, was only apparent in a grazing environment with restricted DMI (High SR), where Jersey cows produced 20% more milk/kg of DMI (L’Huiller et al. 1988); under ad libitum feeding, this ME conversion advantage disappeared. These results were supported by the short-term calorimetry experiment at No 5 Dairy (L’Huiller et al. 1988). In this part of the experiment, partitioning of gross energy into digestible energy or ME was not significantly affected by breed, but the efficiency of utilisation of ME for milk and tissue (energy balance) was greater for the Jersey cows than the Holstein-Friesian comparison (0.58 v 0.48 respectively).

The accumulated results appeared to indicate that Jersey cows would outperform Holstein-Friesian cows in systems where feed supply was less certain (e.g. a high SR, less consistent climate for pasture production), while Holstein-Friesian cows would outperform Jersey cows in production systems where feed supply per cow was more certain (e.g. low SR, greater use of supplementary feed). This was the first time that there appeared to be a material genotype (breed) × environment (feed supply certainty) interaction in grazing systems and, considering the variation in repeatability of pasture production in different locations, particularly through mid-lactation, the interaction between breed and feed supply (i.e. SR) required investigation in a multi-year farm systems experiment.

Holstein-Friesian vs Jersey comparison

Arnold Bryant established a 2 (breed) × 2 (SR) experiment in 1990/91 to investigate the interaction between breed and feed supply on pasture and animal production, reproduction, and, ultimately, farm profitability. The experimental design ensured a consistent metabolic Lwt/ha, to ensure equivalent comparison. Both breeds had a similar comparative SR (CSR; 80 and 100 kg Lwt/t feed DM available) at their different SR comparisons (3.6 and 4.5 cows/ha and 3.0 and 4.0 cows/ha for Jersey and Holstein-Friesian treatment herds, respectively).

Ahlborn and Bryant (1992) reported (on the first year of the experiment), that the Holstein-Friesians produced greater milk yields and gross and net milk incomes per cow, while Jerseys produced more milk fat and protein, and greater gross and net milk incomes per ha, when both breeds were stocked at the same Lwt or for maximum production or optimum economic performance. Maximum net income, based on 1990/91 milk price and expense costs, occurred at 3.6 and 3.0 cows/ha for Jersey and Holstein-Friesian herds, respectively. At these SRs, net income for Jerseys was 5% greater than for Holstein-Friesians: $2866/ha and $2717 for the Jersey and Holstein-Friesian treatments, respectively. A subsequent analysis of 3-years’ data (Penno 1998) concluded that the optimum SR was slightly lower, at 3.5 and 2.7 cows/ha for the Jersey and Holstein-Friesian treatments, respectively, but the Holstein-Friesian had a greater economic farm surplus (i.e. a standardised metric for operating profit): $1770 and $1820/ha for the Jersey and Holstein-Friesian treatments, respectively, or ∼3% advantage to the Holstein-Friesian treatment (Table 11).

Despite the earlier preliminary reporting of results, the entire dataset was not compiled, analysed, or interpreted for a further 20 years (Spaans et al. 2018). Importantly, however, Jersey and Holstein-Friesian cows at a CSR of 80 kg Lwt/t feed DM available (close to economic optimum; Macdonald et al. 2011) produced 334 and 372 kg MS/cow, respectively, and 1192 and 1124 kg MS/ha, respectively, which, from pasture alone, is better milk production than the top 10% of Waikato farmers in 2020/21. The results, therefore, should be considered applicable in the modern scene.

Summary results are presented in Table 11. Spaans et al. (2019) reported on the pasture and animal production results and modelled the economic outcomes of the different systems and Edwards et al. (2019) modelled the effects of milk price on these outcomes. There was an interaction between breed and CSR in the pasture and animal production variables measured. At the CSR100, the Holstein-Friesian farmlet produced less pasture, probably because of persistent over-grazing, particularly during late-spring and summer, and, as a result, produced less milk and milk components/cow and per ha than the Jersey-CSR100 treatment. There was no material effect of treatment on reproduction outcomes, although Jersey cows had shorter post-partum anoestrous intervals and less non-cycling cows at the beginning of the seasonal breeding period.

Spaans et al. (2019) also estimated the efficiency with which ME was used for maintenance and productive purposes at the level of the farm system, utilising pasture production and estimated pasture quality variables, metabolic Lwt and production data, per cow and per ha. On average, Jersey cows used less ME for maintenance, activity, pregnancy and production/cow, but more ME/ha, with the exception of pregnancy. A greater percentage of ME consumed by CSR100-Jersey cows was used for milk production (+3%) and a lesser percentage (−2%) was used for maintenance, but there was no effect of breed at CSR80. These results are consistent with those of L’Huiller et al. (1988), who reported a 10% lower efficiency for Holstein-Friesian cows converting feed into milk in a short-term experiment in mid-lactation, a result subsequently confirmed in open-circuit calorimeters and by Prendiville et al. (2009), who reported that Jersey cows required 7% to 8% less total feed for every kg of fat and protein produced in a pasture-based system, when compared with Holstein-Friesian cows.

The milk production results, in conjunction with the expenses associated with a grazing system and the different numbers of cows/ha at the same CSR, were used to model the financial performance of each of the farmlets. The modelling was performed using a milk fat and protein price of NZ$4.04 and NZ$7.35, respectively, which reflected both the milk price and value differential between milk fat and protein at the time of analysis. This resulted in an interaction in the modelled operating profit/ha (Table 11), with the Holstein-Friesian herd marginally (NZ$169/ha) more profitable than the Jersey herd at the lower SR (CSR = 80), but much less profitable (-NZ$386) at the higher SR (CSR = 100). Edwards et al. (2019) investigated the effect of changing milk fat and protein values on the profitability of the breeds. They concluded that CSR80-Jersey herds were more profitable than Holstein-Friesian herds when milk fat price >NZ$5.67, irrespective of milk protein price, although the difference was only 1%/ha to 5%/ha.

In conclusion, milk price aside, Holstein-Friesian cows were more profitable at low SRs and less profitable at high SRs. As Arnold Bryant putatively quipped: “if you’re short of land milk Jerseys; if you’re short of labour milk Friesians”. Nevertheless, the milk price signals that valued milk protein more than fat, and the higher milk yields of Holstein-Friesian cows resulted in a shift from a predominantly Jersey national herd to one predominantly Holstein-Friesian. There was also a demand for Holstein-Friesian and Friesian × Jersey cross bull calves to rear for beef, as there was a reluctance to rear purebred Jersey stock for this. This change did not result in an increase in profitability and, in fact, in low-input production systems probably reduced the resilience of the system to feed supply variability.

‘Cows for courses’ – the importance of genetic suitability for farming systems

The current New Zealand Holstein-Friesian (NZHF) cow was bred from the Jersey background by crossbreeding the Jersey with New Zealand Holstein-Friesian semen, commencing in the early 1960s and continuing into the 1980s. With the acceptance of AI and the move from linebreeding to outcrossing, and improvements in the cryopreservation of semen, opportunities for using semen from other countries increased, expanding the bull pool from which farmers could select. Intensive genetic selection and the associated increases in milk yield in Europe and North America (the United States and Canada) after World War II as well as an assumed selection pressure on fat yield in a ‘closed herd’ in New Zealand encouraged farmers to use Holstein-Friesian genetics from these countries (Roche et al. 2017a).

Some North American Holstein-Friesian (NAHF) genetics were imported from Canada during the 1960s and 1970s, although the volume was small. Importation increased in the 1980s, particularly from the United States and Europe (Harris and Kolver 2001). The demand was so great that between 1980 and 1999, the average percentage of NAHF in NZHF cows increased from 2% to 38%, and the percentage of Holstein-Friesian cows with some NAHF ancestry increasing from 7% to 96%.

By the early to mid-1990s, however, farmers had begun to question the suitability of this ‘modern’ Holstein-Friesian breed for temperate grazing systems. Anecdotally, pregnancy rates declined, empty rates at the end of the seasonal breeding period increased, and the survival of cows with greater proportions of NAHF ancestry was less than herd mates with less NAHF ancestry. At the time, the national breeding objective (i.e. BW) did not include non-production, but functional traits important for profitability, like reproduction, primarily because such traits were known to have low heritability; it was therefore assumed that selection for these traits would not influence offspring performance. However, the cacophony of farmer voices claiming this was not correct were to result in a pair of farm systems experiments, one at No. 2 Dairy (Macdonald et al. 2007, 2008b; Lucy et al. 2009) and one at Moorepark in Ireland (Horan et al. 2005, 2006), that were to change how the world viewed functional traits of low heritability.

Genetic ancestry affected heifer and cow Lwt

Strain effects on birth weight were small, although NZ70 heifer calves were approximately 10% (4.5 kg) lighter than NZ90 and NA90 calves, which did not differ from each other. Despite consistency in rearing practices, NA90 heifers grew faster than NZ90, which grew faster than NZ70 heifers during the first 6 months. The NA90 heifers had a greater ADG until 12 months than either NZ strain, which did not differ from each other. Average daily gain during year 2 was not affected by genetic ancestry, but the legacy differences remained, so that NZ70 were lighter (10 kg) than NZ90, which were lighter (24 kg) than NA90 heifers at 24 months of age (Table 12). These differences were to remain as adult cows: 480, 498, and 529 kg Lwt for NZ70, NZ90, and NA90, respectively.

Table 12. Summary data for the three strains of Holstein-Friesian from birth to end of second lactation (Experiment 26).

	NZ70	NZ90	NA90	P-value
Age at puberty, d	329	356	373	<0.05
Lwt – 6 month, kg	153	162	169	<0.05
Lwt – 12 month, kg	239	248	258	<0.05
Lwt – puberty, kg	230	253	274	<0.05
Lwt – 24 month, kg	481	491	515	<0.05
Total feed allowance, t DM/cow	6.0	6.0	6.0
Stocking rate, cows/ha	3.1	3.1	3.1
CSR, kg Lwt/t DM	86	83	85
Milk, kg/cow	4811	5593	5479	<0.001
Milk fat, kg/cow	218	264	228	<0.001
Milk protein, kg/cow	162	204	187	<0.001
Milksolids, kg/cow	380	468	415
Efficiency, kgmilksolids/kg Lwt^0.75	4.24	4.39	3.93	<0.001
Postpartum anoestrous interval, d	32.2	38.7	28.4	<0.001
Conception to first AI, %	45	46	39	NS
Conception to second AI, %	54	48	44	NS
Interval 1st to 2nd AI, d	23.7	22.8	26.2	<0.05
Pregnant in 6 wk, %	70	69	54	<0.001
Pregnant in 8 wk, %	80	75	62	<0.001
Mean final Pregnancy rate, %	93	93	87	<0.05
Yr 3 calving date	Jul-30	Jul-28	Aug-07	<0.05

Notes: Milk production and reproduction data are presented from herds receiving 5.5 t DM pasture and 0.5 t DM supplementary feed per cow (i.e. total feed allowance of 6 t DM/cow) (adapted from Macdonald et al. 2007, 2008a).

The NA90 heifer also had a larger frame than the NZ strains, which did not differ from each other. Body length, heart girth, and wither height were 2% to 4% greater in NA90 heifers, with differences increasing with age: from 2.3 to 4.0 cm, 2.7 to 4.3 cm, and 2.1 to 4.9 cm greater in NA90 compared with NZ90 between 6 months and 36 months, respectively. Age and Lwt at puberty were also affected by genetic ancestry: NZ70 cows were younger (33 days; P < 0.05) and lighter (23 kg; P < 0.05) than NZ90 heifers, which were younger (17 days; P < 0.05) and lighter (21 kg; P = 0.07) than NA90 cows. The heavier mature Lwt and older age at puberty in the NA90 heifer may suggest that either they mature later or that, under pastoral conditions, growth rate was limited by their inability to consume sufficient nutrients from grazed pasture, with a consequent delay in puberty, or both.

The genetic ancestry effects on age and Lwt at which sexual maturity is reached can have significant effects on the success of these strains in grazing systems. Twenty-one percent of NA90 heifers had not reached puberty at planned start of mating, compared with only 3% of the NZ90 heifers and 0% NZ70 heifers. The NA strain were, therefore, at an increased risk of not conceiving as a nulliparous heifer or calving later in the herd, a factor known to reduce the likelihood of reproductive success in their first lactation (Burke et al. 1995).

Cows for courses: rewriting the textbook on genotype × environment interactions

Genetic strain affected the profiles of BCS change (Figure 10; Macdonald et al. 2008b), with NA90 cows losing more BCS and for longer in early lactation than NZ90 cows, which lost more BCS and for longer than NZ70 cows. The profile of BCS recovery in mid-late lactation was also different, with the effect increasingly exaggerated with age. As a result, average BCS was greater in NZ70 cows (5.1) than NZ90 cows (4.5), and NA90 cows (4.1; 1 to 10 scale; Roche et al. 2004; Macdonald and Roche 2011). These results were supported by Roche et al, (2006a), when they fitted mathematical functions to the BCS and BW profile, providing quantitative parameters for the slope of the post-calving decline, the slope of the post-nadir incline, and the height of the BCS and BW profiles for the NZ90 and NA90 genotypes.

Figure 10. Body condition score for the three strains of Holstein-Friesians over consecutive lactations that were evaluated at a range of feeding allowances in Experiment 26: 'The Dexcel Strain Trial'. (

NZ70 = a 1970s high Breeding Worth strain of New Zealand Friesian,

NZ90 = a 1990s high Breeding Worth Holstein-Friesian of New Zealand origin,

NA90 = a 1990s high Breeding Worth Holstein-Friesian of North American origin).

Summary results for milk production are presented in Table 12. NZ90 cows had a higher milk protein concentration than either NA90 or NZ70 cows, and a higher milk fat concentration than the NA90 cows. NZ90 produced greater yields of fat, protein, and lactose than the NA90 and NZ70 cows. Similar to the previous results of Carter (1964), there was no genotype × environment interaction for MS, with NZ90 cows producing 52 kg more MS/cow than NA90 cows at all feeding levels (Macdonald et al. 2008b). Lactation length was the same for the two NZ strains, but was 34 days shorter for NA90 cows to ensure they were able to achieve the appropriate BCS for calving. Differences in milk production/cow at the same SR resulted in differences in estimated Operating profit: $2755, $3408 and $3052/ha for the NZ70, NZ90 and NA90 strains, respectively (milk fat = $2.36/kg; protein = $6.76/kg; volume = -$0.04/kg) (Macdonald et al. 2005).

Genetic ancestry also affected reproductive outcomes. Despite the delay in age at puberty, on average, NA90 cows began oestrous activity 4 and 10 days earlier than either the NZ70 or NZ90 cows, respectively; the earlier oestrous activity, however, was not associated with an increase in either conception rate or pregnancy rate. Seventy per cent of cows in NZ70 and NZ90 herds were pregnant within six weeks of the start of the seasonal breeding period, compared 54% of NA90 cows. Final pregnancy rates also reflected lower conception rates in the NA strain; % cows pregnant after 12 weeks of breeding were not different in the New Zealand strains, but 6% less cows were pregnant in the NA90 herd. It would appear that the active management for a compact calving system to maximise the use of pasture (Roche et al. 2017) had led to a form of ‘artificial selection’ for superior reproductive performance, as cows not pregnant within the defined window were culled and there were no offspring to further the trait. Because dairy selection objectives and farm production systems in USA and Europe were different from those in New Zealand (Miglior et al. 2005), however, there was no such selection pressure within their systems and, it was plausible that these strains of Holstein-Freisian would take longer to conceive, despite this change in phenotype having significant economic consequences for New Zealand farming.

Investigations into physiological and behavioural differences between the strains identified potential reasons for the differences reported. Lucy et al. (2009) reported that the uncoupling of the somatotropic axis in early lactation, a homeorhetic adaptation to support milk production, was more pronounced in the NA90 cows and was extended in duration compared with housed cows being fed total mixed rations. This resulted in greater circulating concentrations of growth hormone, a potent lipolytic agent and consistent with the differences in BCS profiles, and lower concentrations of both insulin-like growth factor (IGF-1) and insulin, important hormones in the preparation of the follicle/oocyte for ovulation. These results were supported by functional genomics assays of liver tissues from the strains, with reduced expression of growth hormone receptor-1A and IGF-1 in liver of the NA90 strain than either of the New Zealand strains out to 56 days post-calving. The extended period of uncoupling could be a key reason for poor reproductive outcomes, with circulating IGF-1 concentrations a reported marker of improved reproductive outcomes.

The greater extent of somatotropic axis uncoupling would be expected to increase milk production. But, the NA90 cows produced less fat and protein than the NZ90 cows. This may be a result of inherent selection for superior foraging capability in the New Zealand strains, giving them an advantage over the North American strain in DMI, despite the greater physiological stimuli (i.e. greater growth hormone concentrations) to eat and produce milk. Rossi et al. (2005), compared the strains grazing behaviour and reported that the NZ strain consumed more pasture/kg Lwt. In their analysis, the NA cow required a leafier pasture and a higher pre-grazing mass before intakes and production was greater than that of the New Zealand cow. But these traits are opposing, with % leaf, in general, declining with pre-grazing mass (Holmes 1989). Alternatively, to provide this leafy, but longer pasture, the cow must be able to have a greater degree of selection, leaving higher post-grazing residuals and reducing pasture quality and intake capacity in subsequent grazings. It would appear, therefore, that cow selection within a grazing system resulted in greater grazing ability and a higher DMI/kg Lwt than when cows were selected within a system with unlimited supply of feed and limited competition (NA90).

Conclusions and implications

Intensive selection for important traits (genetics) increased production and profitability, irrespective of feed supply (environment) and cow breed. Crossbreeding did not offer significant advantages when both parent breeds were suitable for pasture-based milk production. However, it is a fast way to outcross when the parent breed is not suitable.

Genetic improvement is slow, but it is permanent. Hence, it is critical to ensure that the traits considered are most important for the sustainable profitability of the farming system. Although the results from No. 2 dairy do not support a genotype × environment interaction, they highlight the need for multi-trait selection objectives, even if the trait is lowly heritable. In less than 20 years of selection for NAHF genetics, the resultant animal was less fertile and, as a result, had a lower survival rate in grazing systems. Despite being born in New Zealand and raised under a grazing system, even its grazing behaviour was different to a cow whose ancestry was pasture-based. There was also an indication of an effect of genetics on grazing behaviour in the Jersey-HF comparison experiment (Spaans et al. 2018), with the NZHF cow exhibiting more aggressive grazing tendencies than the Jersey cow, with perverse outcomes for mid-season pasture growth and overall feed supply.

Collectively, the results were profound. Farmers should select the best bull for their system, irrespective of feed supply or breed preference. Furthermore, the recognition that traits other than production were important to profitability was instrumental in global conversations about the need to consider functional traits, no matter the heritability, in addition to production traits and spearheaded a desire for easily acquired phenotypic measurements for health and reproduction traits.

Theme 8: integrating dairy and beef production

Background

To produce milk, a cow must calve; the male to female ratio in offspring calves is 52:48 (Roche et al. 2006b). With survival rates and replacement rates of ∼95% and ∼20%, respectively, a 400-cow farm will have more than 300 calves (male and female) surplus to requirements for the dairy business. These animals make a valuable contribution to the beef sector, with 70 per cent of beef cattle slaughtered in New Zealand coming directly from the dairy sector (NZ Livestock and Products Annual 2019).

In the late 1960s, with the anticipation of the UK joining the European Economic Community (now, the European Union or EU), there was concern that New Zealand farmers would not have a market for all of the milk produced. There was also concern with the processing facilities ability to handle peak milk flow, and the main way to alleviate this seemed to be a capital cost of adding to processing plants or building new plants. To alleviate the issue of surplus milk for processing, farmers were encouraged to diversify their farm operations, with one option being to grow beef cattle to finishing in addition to the dairy business. Campbell investigated alternative management systems on dairy farms that would reduce the amount of milk produced, but bring in alternative sources of income for the farmer (Experiment 9).

Experimental methods

A 5-year project (Experiment 9) investigating the integration of dairy and beef rearing and finishing was established by Archie Campbell in 1968 to follow on from the grazing strategy, SR, and genetic merit research undertaken by McMeekan and Carter. The experimental objective was to compare the output of saleable animal products from Jersey and Holstein-Friesian crossbred cows, with calves suckled as weaners and either fattened for sale as 6-month old calves, therein reducing SR in the second half of the season, or reared to finishing. A particular objective was to identify and overcome problems associated with combining dairy and dairy-beef enterprises on the one farm, especially problems of fostering, weaning, suckled cow and calf health, grazing management, and the fattening of beef stock on the dairy farm.

Jersey cows bred to AI were randomly allocated to four 12 ha farmlets, with herds balanced for cow age and genetic merit. One herd was maintained as straight-bred Jerseys using the best available Jersey AI sires. In the other 3 herds, semen from Holstein-Friesian AI sires, of similar genetic merit to the Jersey sires used in the Control, was used, with the breed make-up of three-quarter of the herd transitioning to Holstein-Friesian × Jersey cross-bred cows. All farmlets reared 25% of the milking cow number for replacements, with an annual replacement rate of 20%, and were stocked with the equivalent of 4.1 cows/ha. On the farmlets that had cows suckling calves the mature cows suckled 4 and the heifers 3. Treatments were:

1. ‘Control’: Forty-two Jersey cows under a normal dairy farm system;

2. Farmlet 2: Thirty-nine Holstein-Friesian-Jersey cross bred cows with 25% of the cows rearing calves for 10 weeks, with surplus calves sold in mid-December;

3. Farmlet 3: Similar to Farmlet 2, but a 20% lower SR (33 cows), with 25% of cows rearing calvesd to slaughter at 385 kg or on the 30th April in their 2nd year (∼20 to 21 months), whichever came first; and

4. Farmlet 4: Twenty-seven Holstein-Friesian-Jersey cross bred cows with all cows rearing calves for 10 weeks. Extra calves were purchased to facilitate the rearing of 4 calves/cow and 3 calves/heifer. The calves that were surplus to replacement needs were sold in mid-December.

Results and discussion

Due to the differing number of cows milked within each treatment and the date cows started producing milk for sale, the milk fat/ha supplied was 524, 430, 265, and 302 for Treatment 1, 2, 3 and 4 respectively, with peak milk production supply occurring on day 75 for Treatment 1, and on day 93–95 post-PSC for the other 3 treatments. These data demonstrate a significant effect of differing stock policies on the timing and volume of peak milk supplied to the processor. The use of a system rearing some calves (Treatments 2, 3 & 4) would have improved the processing facilities ability to handle peak milk flow, which would be a positive in some regions because of the concern about the need for added processing capital and the cost of processing the extra peak milk, with no ability to capitalise on the asset beyond peak and, potentially, a less lucrative market for its sale.

There was more mastitis in the cows that were suckled; most of this occurred when the calves were weaned and milk harvesting was twice daily at the shed. The reason for this effect was unclear, but potentially related to post-weaning stress.

Treatment 4 was deemed to be impractical because of the very high percentage of cows that failed to become pregnant or had a delayed calving date due to the late onset of oestrus associated with lactational amenorrhea, high labour requirements associated with managing suckling, rejection of calves by cows, a larger Lwt loss of cows during suckling, and added grazing pressure creating lower hay yields as a result of grazing the weaners with their dams.

A partial economic analysis was undertaken by Campbell et al. (1974), concluding that, on the basis of gross margin, the ‘Control’ dairy system was most profitable; however, there was only a small difference between ‘Control’ farmlet and Farmlet 2 (i.e. 25% of calves reared for 10-wk and surplus calves sold at Christmas): gross margins were $526 and $507/ha in the ‘Control’ Farmlet and Farmlet 2, respectively. In comparison, Farmlet 3 and 4 had a significantly smaller gross margin than either the ‘Control’ or Farmlet 2: $430/ha and $365/ha, respectively. Although Farmlet 4 had a greater income from the sale of weaners than either Farmlet 2 or 3, it was not enough to make up for the decrease in milk income and the greater costs, including the additional cost of purchased calves. The effect of treatment was dependent on year, wherein Farmlet 2 was more profitable than the ‘Control’ farmlet in drought years, where mid- and late lactation milk production was significantly reduced relative to a normal climate year.

Conclusions and implications

The general conclusion was that specialising in dairy was the most economic optiion on Ruakura soils and with the prevailing Waikato climate. In general, when milking cow numbers were reduced and beef numbers increased, the gross margin was, generally, lowered. This decline was less in drought years because milk production was reduced more than beef production, indicating that when farming in a drought-prone area, some diversification (e.g. Farmlet 2 or variations on it) may be a prudent strategy for cashflow resilience, as the revenue from the weaner sales is received in the January-February months.

Theme 9: evaluating seasonal calving: the effect of calving date on farm management and profitability

Background

In pasture-based dairy systems, the cost of milk produced is negatively related, in quadratic fashion, to the amount of grazed pasture in the diet of the cow (Dillon et al. 2005; Roche et al. 2017b). In comparison, as presented in Theme 6 (supplementary feeding), milk produced from imported (i.e. supplemental) feeds can have a very high marginal cost, linearly increasing the cost of marginal milk produced (Ramsbottom et al. 2015). Therefore, a key aim of profitable grazing systems is to maximise the amount of milk produced directly from pasture.

As described earlier, in temporal regions, pasture growth rates vary throughout the year, with peak growth during spring and troughs in winter; thus, pasture-based herds are traditionally managed such that the PSC is mid to late winter, to ensure that the high nutrient demand at peak lactation (∼80 DIM) coincides with seasonal peaks in pasture yield and digestibility in spring.

With an increased demand for liquid milk supply in New Zealand from the mid-1940s, ‘town milk’ farms were established. This was the result of an Act of Parliament in 1943 that saw those supplying milk for liquid consumption allocated a quota and receiving approximately 50% more than those supplying milk for manufacturing into product. In the 1960s, with increasing whole milk demand all year, a small number of specialty ‘autumn-calving’ farms were established, where cows were calved in autumn and milked through winter and spring. The cost of milk production was greater than for a spring-calving herd, so there was a premium for the milk and farmers were paid on a volume of milk basis and not fat. Additionally, there were a small number of farmers who chose to calve in the autumn because their farms were at greater risk of dry summers. As numbers of these farms increased there was a range of systems implemented from all the cows on a farm calving in either the spring or autumn, to various proportions of the herd calving in autumn and spring.

In the 1980s, there was growing interest in these changes to calving patterns as there was a concern about the increasing requirement for processing capacity to accommodate peak milk flow, with manufacturing plants underutilised for a good portion of the year (Davis and Kirk 1984), and difficulties manufacturing high-quality products early and late in the season (Auldist et al. 1997, 2002). Cows calving at different times of the year and differentiating supply into early and late lactation milk could alleviate these manufacturing problems. Furthermore, the increasing availability and use of N fertiliser and imported supplements meant feed supply was not as aligned with the grass growth profile as before, particularly for milk that attracted a premium price.

Although farmers had been calving cows in spring and autumn for decades, little research had been undertaken to quantify the biophysical and economic implications of changing calving season, nor to define the potentially optimum time to begin calving. An experiment was established at No. 2 Dairy Ruakura in the early 1990s to quantify the biophysical and economic effects of calving at different times of the year.

Experimental methods

Martin Auldist established 4 farmlets of 20 mixed-aged Holstein-Friesian cows that were bred such that one herd calved during a 6-week period in each of January (JAN), April (APR), July (JUL), and October (OCT) (Auldist et al. 2002) and is described in Table 13. All farmlets were managed at a SR of 3.1 cows/ha on 6.5 ha farmlets. On each farmlet, 200 kg N/ha was applied annually, but the time of application differed for each herd and was determined by the feed requirements of the herds and expected pasture responses. The amount of silage made on each farmlet, and the time at which it was fed, was also determined by the prevailing pasture cover and nutritional requirements on each farmlet. The biophysical and economic effects of season of calving were comprehensively reported by Spaans et al. (2019).

Table 13. Lactation lengths, yields of milk, milk fat and protein, and operating profit for herds of cows calving in July, October, January and April (Experiment 25; adapted from Spaans et al. 2019).

Planned start of calving	July	October	January	April	SED	P
Annual pasture intake, kg DM/cow	4073	4306	4679	4205	256.6	0.28
Annual DMI, kg DM/cow	4772	4677	5026	4905	160.4	0.32
Lactation length (d)	261	262	292	287	12.3	0.17
Milk kg/cow	4018	3681	3726	3843	188.2	0.07
4% Fat corrected milk, kg/cow	4057	4236	4878	4002	218.9	0.08
Milkfat (kg/cow)	184	169	171	180	8.3	0.08
Protein (kg/cow)	136	123	128	134	6.5	0.06
Milksolids (kg/cow)	320	292	299	314	14.8	0.07
Operating Profit ($/ha)	3137	2199	2455	2850

Results and discussion

Traditionally, it has been assumed that operating profit is maximised by establishing the PSC during winter, such that the peak DMI of the herd coincides with peak growth of highest digestibility pasture, reducing the need for imported supplements and the increased cost of marginal milk associated with that strategy (Roche et al. 2009b, 2017b; Macdonald et al. 2017; Roche 2017). However, very few studies have evaluated this assumption, and none have ensured that SR was identical under the different calving date treatments. This study offered a unique opportunity to establish a platform to compare four alternative calving seasons, using a quantitative case study approach to evaluate the biophysical and economic performance of altering calving date in the Waikato region of New Zealand and, in so doing, provide context for its applicability in other regions.

Annual pasture growth (t DM/ha) was not affected by season of calving, but the proportion of fresh pasture versus pasture silage in the herd’s diet differed at different stages of lactation. For example, cows in the APR calving treatment had the least pasture available to them during 1 to 90 d post-calving, the period of greatest herd feed requirements, but most feed available to them 181 to 270 d post-calving. Garcia and Holmes (2001) also reported that similar levels of pasture production and utilisation and milk production could be achieved in contrasting pasture-based calving systems if common grazing, conservation, and feeding management criteria were applied to all systems. Therefore, in theory, the same level of milk production is possible from autumn and spring-calving herds.

The No. 2 results do not support the assertion that both autumn and spring calving dates are equivalent in terms of milk production, despite being in agreement that total pasture production and utilisation was not affected by calving date. Furthermore, the No.2 results were able to provide some insight into calving dates during the intervening ∼9 months. Despite similar pasture production and utilisation, irrespective of calving date, diet composition during the different stages of lactation was affected by treatment (Spaans et al. 2019). As a result, there was a trend (P < 0.10) for lower milk production/cow in the JAN, APR, and OCT herds, when compared with the JUL calving herd, despite numerically longer lactations in the JAN and APR treatments (Table 13).

The superior milk production of autumn-calving cows has been often reported, but, in general, the effects are confounded by differences in the amount of supplementary feeds provided to the different calving date herds. For example, Garcia et al. (1998) reported that autumn-calving cows produced 52 kg more milk fat and protein compared with spring-calving cows and that the greater production was due to a 50-d-longer lactation (291 vs. 241 d). However, the autumn-calving treatment in their experiment had both a lower SR (2.0 vs. 2.4 cows/ha) and received almost 10% of their diet from purchased maize silage. Any experimental differences were, therefore, at least in part, a result of a different allowance and type of feed/cow and cannot be attributed solely to calving date. When only calving date was changed, the No. 2 experimental results indicate a trend for greater milk production per cow and, by extension, per hectare in JUL cows compared with cows in the APR, OCT, or JAN calving treatments.

Calving date also affected BCS and Lwt during the peripartum period, with management protocols to ensure cows achieved target BCS pre-calving more difficult to implement in the OCT and JAN calving treatments; cows were, on average, 0.5 BCS units fatter than recommended for New Zealand grazing systems (Roche et al. 2009e; Macdonald and Roche 2011). By comparison, APR and JUL calving cows were able to achieve the recommended pre-calving BCS targets, but APR cows lost more BCS between one month pre-calving and post-calving (1.2 units) than cows in any of the other calving herds (0.5–0.8 units). These treatment effects on BCS may reflect a greater amount of pasture silage in the diet of the APR-calving cows immediately post-calving, especially relative to the JUL and OCT treatments, whose cows lost the least amount of BCS, maybe reflecting the nutritional superiority of spring pasture over autumn pasture for maintenance and BCS gain (Mandok et al. 2014). Irrespective, the experimental results highlighted advantages for JUL calving in optimising BCS targets for milk production, reproduction, and animal health when compared with alternative calving dates.

Calving date also affected profitability. In recognising the challenges associated with changing calving date away from mid-winter, milk companies pay a premium for ‘out of season’ milk. They also tend to ‘dis-incentivise’ peak milk production to help manage processing capacity. Following adjustment for winter milk premium, JAN, APR, and OCT treatments, respectively, earned an extra $409, $428, and $344/ha. Furthermore, the gross farm revenue for the JUL treatment decreased by $213/ha because of the downward adjustment on milk price for milk produced during peak supply. Compared with the MS payment received by the other 3 treatments ($6.60, $6.60, and $6.54/kg MS for JAN, APR, and OCT, respectively), the JUL-calving treatment received the lowest average milk price ($5.97/kg MS). Nevertheless, the JUL-calving treatment had the greatest operating profit/ha, followed by the APR-calving treatment. Spaans et al. (2019) used stochastic modelling techniques and probability distribution functions for input costs to determine the long-term profitability of the different calving dates. Even with the inclusion of the winter milk premium, the JUL treatment remained stochastically dominant over the other 3 treatments, and the APR treatment remained stochastically dominant over the JAN and OCT treatments. Spaans et al. (2019) concluded, in fact, that their modelling was conservative, with anecdotal evidence of increased repairs and maintenance expenses for machinery and lameness in grazing herds producing winter milk. It is likely, therefore, that the probability of stochastic dominance of a mid- to late-winter calving treatment group is greater than portrayed in the No. 2 experiment.

The JAN and APR herds had much flatter lactation curves than the JUL and OCT herds, implying a more consistent daily milk supply to factories. Manipulating the time of calving could be a potential tool for overcoming problems caused by the seasonality of supply, but the No. 2 experimental results suggest that current premiums for ‘out of season’ milk and disincentives for peak milk production are not adequate to make up the shortfall in operating profit. The econometric approach to modelling the farm system implications of change in calving date could be used to determine the milk price incentives needed to justify farmers changing calving patterns to help smooth manufacturing capacity and dairy product specifications.

Conclusions and implications

In summary, the results indicated that a PSC in late winter in a temperate climate is most profitable in a grazing system that is not importing feed, with or without a realistic price incentive scheme. The No. 2 experiment indicated that a calving date in mid-winter increases the proportion of fresh pasture in the diet of the lactating cow, which, in turn, increased milk production/cow relative to the other calving dates. Operating expenses were not materially affected by calving date treatment in this experiment, but this is likely conservative; at least anecdotally, autumn-calving cows have greater fixed costs than those calving in winter-spring. When the variability in annual pasture growth, input prices, and milk prices was considered, the JUL-calving treatment was consistently the most profitable, even following the inclusion of a commercially realistic winter milk price premium.

Theme 10: determining the value and environmental consequences of N fertiliser in grazing dairy systems

Background

Plants require N to grow; but the natural cycles of mineralisation (converting organic N to inorganic forms suitable for uptake by plant roots) and fixation (the process by which some soil microbes convert atmospheric N to plant-available N) are often out of step with cow requirements under production management settings. Rates of soil mineralisation and fixation are very low at temperatures below 10°C, but perennial ryegrass and other temperate pasture grasses grow at shallow soil temperatures of ≥4°C (Whitehead 1995). Pasture growth during winter is limited, therefore, by restricted availability of N.

During the 1970s, concern about consistency of winter pasture supply constrained the uptake of management advances developed at No. 2 Dairy over the previous three decades. Campbell et al. (1977) had identified that the primary reasons for the vastly superior milk production at No.2 dairy, when compared with the regional average, were the higher SR and the earlier and more concentrated calving of the No. 2 herd. The earlier and more compact calving alone resulted in the No. 2 Dairy herd having an extra 21 DIM/cow by 31st December, when compared with the regional average, which when multiplied by a 55% greater SR and a 10% greater milk production/cow, amounted to ~100% more milk production/ha before summer.

If farmers were to emulate the No. 2 example, they would need confidence of a consistent supply of pasture during winter months to cater to the increased feed demand in early lactation when pasture growth was slow. Earlier experiments had indicated that pasture growth responses to N could be good (see Whitehead 1995), but were not economic because of the milk to N price ratio. However, the plan to build a urea plant at Kapuni in Taranaki in 1982, thereby reducing the price of N fertiliser to New Zealand farmers, led to experiments evaluating the role of N fertiliser as a system-level tool to provide a predictable level of feed during winter consistently and enable a better match of feed supply and herd demand.

Experimental methods

Experiment 12 – effects of using nitrogen to increase pasture growth on production of milksolids

An experiment was established in 1979 to examine the interactions between milk production, calving date, SR, and N fertiliser (Experiment 12). This was the first of a series of experiments at No. 2 Dairy established by Arnold Bryant to examine the role of nitrogen fertiliser in improving out-of-season feed supply for dairy systems.

The experiment ran for three years. Initially, 52 hectares were randomly allocated to 1 of 8 farmlets (6.5 ha each) in a 2 × 2 factorial management with 2 PSC dates (early July and early August) and 2 N fertiliser strategies rates of N (with or without urea; +N and -N). The farmlets and cows (n = 206) were randomly allocated to one of the 8 treatments, thus there was replication of the treatments. After the first season, the design was changed to a 2 × 2 × 2; with 2 SRs (SRs; 3.9 and 4.3 cows/ha) and the replication dispensed with. Mean calving dates during the three years averaged 21 July and 14 August. Nitrogen was applied at ∼90 kg/ha per yr and was applied in 2 or 3 dressings over the whole farm during April to August.

Experiment 21–1.75 t MS

In Experiment 21 (See Theme 6 and Macdonald et al. 2017 for full details), Arnold Bryant and colleagues established a long-term (5 year) experiment (colloquially known as the 1.75 t MS/ha experiment) to investigate the role of N fertiliser in increasing pasture production/ha and, therefore, pasture available/cow, and associated changes in milk production at both moderate and high SRs. The experiment ran in two phases: (1) in the first two years of the project, there were 7 farmlets, each of 6.47 ha with Holstein-Friesian cows in two SR, each with one of three levels of N (0, 200, 400 kgN/ha). After two years, the experiment was changed, with one ‘Control’ farmlet (LSR-0N) receiving no N fertiliser or purchased supplements (SR was 3.35 cows/ha) compared with four farmlets (LSR-200N, LSR-400N, HSR-200N and HSR-400N) designed to investigate the effects of N fertiliser (either 200 or 400 kg N/ha) on MS production per cow and per hectare at low (LSR – 3.35 cows/ha) and high (HSR – 4.41 cows/ha) SR.

Results and discussion

Pasture and animal production from N fertiliser

In Experiment 12, the use of 90 kg N/ha per year increased MS production by 60 kg/ha, on average over the 3 years (1152 and 1212 kg MS/ha for the -N farms and +N farms, respectively); this equated to a response of 667 g MS/kg N fertiliser applied (or 67 kg MS/100 kg N fertiliser). However, responses (kg MS/kg N applied) varied with year and across treatments being 391, 919 & 707 g MS/kg N applied for seasons 1979/80, 1980/81 and 1981/82, respectively. Thus they were not consistent across years; neither were they consistent for early calving, where the greatest feed shortage occurred, being 545, 233 & 957 and 236, 1605 & 457 for the three seasons for the Low SR and High SR farmlets, respectively.

The ability of the cows on the -N farms to overcome feed shortages in early lactation was a major factor acting to reduce the overall animal response. While N significantly increased pasture production in the spring in two of the three years, this was followed by significantly less pasture being grown in the summer and a reduction in clover content and, presumably, feed quality as a result, for much of the lactation on the +N farms. Milk yields tended to follow a similar trend.

Later calving dates conserved a greater amount of pasture as silage and had a higher peak milk production; but, total seasonal milk production was similar to the earlier calving groups. The high SR farmlets produced 4% more milk/ha than the low SR farmlets in one of the three years, but there was no difference in per ha milk production in the other two years, suggesting that the point of optimum SR had been exceeded. There was no silage conserved on the high SR farmlets, irrespective of calving date or N fertiliser input (nil versus 8 bales hay/cow for the low SR). An analysis of feed grown and harvested estimated that a SR of 3.8 cows/ha was optimum for No. 2 Dairy.

Although the average pasture grown and milk production response to N fertiliser was good, responses were variable across years and insufficient to justify the cost relative to revenue. In fact, the impact of feed shortages was generally of insufficient severity or duration to warrant the use of N. Bryant and colleagues concluded that farmers could achieve more on their own farms by adopting higher standards of pasture management, rather than relying on the uncertainties of using N fertiliser. The decline in clover as a per cent of pasture composition and associated lower mid-season production in the +N treatment led to questions about the requirement for N fertiliser beyond the autumn-winter period if the technology were to successfully contribute to increased feed supply and milk production.

In the early 1990s, the price of N decreased significantly, with the availability of New Zealand made urea from the Kapuni plant; this and a period of relatively stable milk prices led to an increase in the use of N fertiliser to improve MS production per cow and per hectare. Experiment 21 was designed to investigate whether the increased pasture associated with N fertiliser use was transferred to an increase in MS production and per ha profitability (Tables 8 and 9) and whether any effect interacted with SR.

Consistent with the first experiment, results indicated a negative effect of N fertiliser on clover morphology and more detailed measurements indicated a negative effect on the amount of N₂ ‘fixed’ from the atmosphere. Harris et al. (1996) reported that application of 400 kg N/ha per year had a significant effect on clover plant morphology, although number of plants on the farmlet LSR-200N was no different to those on the LSR-0N farmlet. Plants on farmlet LSR-400N had lower stolon length and stolon dry weight, developed fewer auxiliary buds, and, therefore had fewer stolons; these morphological changes would be expected to reduce clover survival in the sward, reducing its ability to compete with the rapidly growing ‘companion’ grass species and probably suggest a maximum-use threshold for N fertiliser to avoid negatively impacting on clover growth. Consistent with this, total clover leaf dry weight and number of leaves/plant were lower on farmlet LSR-400N; number of leaves/stolon and individual leaf dry weight were not, however, significantly different. Ledgard et al. (1996a) reported DM yields of clover on the 0, 200, and 400 kg N farmlets as 2770, 3020, and 1650 kg DM/ha, highlighting the negative effect of 400 kg N/ha on clover growth in a mixed sward and supporting the idea of an optimum use amount. As well as these morphological changes to the clover plants, Ledgard et al. (1996a) quantified the well-published (see Whitehead 1995) negative effects of N fertiliser on N-fixation from legumes in New Zealand. In farmlets receiving no N fertiliser, annual inputs from N₂ fixation varied between 99 and 231 kg N/ha; with every 100 kg N fertiliser/ha applied, however, N₂ fixation declined, on average, 33 kg/ha, (P < 0.01; r² = 0.67; Ledgard et al. 1999).

Despite the negative effects of N fertiliser on clover morphology and N₂-fixation by the clover-rhizobia symbiont, and on clover yield at 400 kg N/ha per year, total pasture production increased linearly with N fertiliser by 8.8 kg DM/kg N applied up to 400 kg N/ha (Macdonald et al. 2017). However, the usefulness of this pasture was dependent on the amount of N fertiliser applied and farm SR. With no increase in SR (i.e. all treatments 3.35 cows/ha), each kg N up to 200 kg/ha resulted in an increase in 5.3 kg DM/ha pasture grazed directly by the cow and 4.3 kg DM/ha conserved as silage. If SR increased from 3.35 to 4.41 cows/ha, each kg N up to 200 kg/ha resulted in an increase in 14.0 kg DM/ha pasture grazed directly by the cow, with no pasture conserved as silage. In comparison, each kg of N between 201 and 400 kg/ha only increased pasture grazed by 0.5 kg DM at 3.35 cows/ha, while 4.6 kg DM were conserved as silage; the additional pasture grown at 400 kg N in the HSR treatment was not utilised. To our knowledge, this was the first time that the interaction between N fertiliser and SR was investigated in a farm systems experiment, with the results highlighting that although the pasture DM production response to N was linear (up to 400 kg in this experiment), % utilisation of pasture/ha declined when >200 kg N/ha was used and, unless the farm increased SR significantly, any additional pasture utilised was as conserved silage.

The increase in pasture production with increasing N use resulted in a linear increase in milk production/cow and per ha. Each 100 kg N increased:

• lactation length by 9 days/cow;

• 4% FCM and MS by 256 and 19.8 kg/cow, respectively; and

• 4% FCM and MS by 859 and 66 kg/ha, respectively.

The MS response/ha in Experiment 21 (66 kg/100 kg N applied/ha) is identical to that reported in Experiment 12 (67 kg MS/100 kg N applied/ha). Although the response was linear (P < 0.01), 60% of this effect was produced from the first 200 kg N, with 40% from increasing N fertiliser applied from 200 to 400 kg.

Environmental sustainability measurements associated with N fertiliser use

Although concern regarding agriculture’s effect on environmental sustainability was not as prominent during the 60-year tenure of No. 2 Dairy as it is today, Ledgard et al. (1996a, 1996b, p. 1999) investigated the effects of N fertiliser on N loss through sub-surface drainage. Increasing N fertiliser from 200 to 400 kg/ha per year increased nitrate loss by 70% to 100% (+41 to 102 kg nitrate/ha lost as leachate). For every 100 kg N/ha applied as fertiliser, N lost in leachate increased by 5 kg/ha (low drainage year; 170–210 mm/yr; Ledgard et al. 1996a) to 34 kg/ha (high drainage year; 550–620 mm/yr; Ledgard et al. 1996b).

Increasing N fertiliser-use also increased GHG emissions; for every 100 kg N applied as fertiliser, greenhouse gas emissions/ha increased by ∼1500 kg CO₂-eq/ha, although the effect was greatest when going from 0 N to 200 kg N (31% increase in kg CO₂-eq; 3651 kg CO₂-eq/ha) and diminished slightly with greater fertiliser use (200 to 400 kg N increased GHG emission by 15% and 2322 kg CO₂-eq/ha) because the GHG emissions from the application event did not increase.

Cost-benefit of N fertiliser

The pasture production and milk production responses to N fertiliser were large, with 100 kg N increasing pasture DM production by 880 kg/ per year, on average, over three years, 530 kg DM/ha per year was consumed directly by the cow, assuming no change in SR, and an additional 430 kg DM were conserved as silage; pasture utilisation increased from 93% to 95%, explaining the difference between additional pasture grown and pasture utilised. With an increase in SR from 3.35 to 4.41 cows/ha, 100 kg N resulted in an additional 1400 kg DM/ha pasture being consumed directly by the cow. In short, 100 kg N results in enough grazed pasture to feed ~0.3 cows/ha. The additional pasture produced resulted in 859 and 66 kg/ha more 4% FCM and MS, respectively, and the effect of N fertiliser on milk production was linear up to 400 kg/ha per year.

In their economic analysis using period costs, the authors estimated the base cost of producing 1 kg MS to be NZ$3.91. The cost of producing each additional kg MS (i.e. marginal milk) using N fertiliser was NZ$3.16 and NZ$3.25 for 200 and 400 kg N fertiliser. Since their analysis, the cost of N fertiliser has increased substantially. At the time of writing, urea is NZ$1300 (NZ$2766/t N); with $200/ha application charges, the cost of marginal milk production is estimated at NZ$5.35 and NZ$6.86/kg MS at 200 and 400 kg N fertiliser/ha per year, respectively. This comparison highlights the sensitivity of the cost of milk production to key variable costs, with the cost of the marginal milk increasing by NZ$2.09-NZ$3.61 because of the increase in fertiliser prices.

Macdonald et al. (2017) also reported on the cost of marginal milk, when N fertiliser was used to facilitate an increase in SR. Although the pasture production response to N fertiliser was the same and the milk production response to the N fertiliser was the same, the cost of the marginal milk from using 200 kg N fertiliser/ha to increase SR from 3.35 to 4.41 cows/ha was NZ$8.89 (NZ$11.35 using the current price for N fertiliser).

Conclusions and implications

Within the ranges tested at No. 2 Dairy (i.e. up to 400 kg N/ha per year), N fertiliser increased pasture production and MS production, but reduced the amount of atmospheric N₂ fixed by clover and, at high rates, the DM production of clover. Although profitable at the time because of the fertiliser-milk price ratio, the cost of the associated marginal milk has increased by 69% and 111% at 200 and 400 kg N/ha respectively; furthermore, nitrate leaching and GHG emissions increase linearly with N fertiliser use, adding to the importance of considered and strategic use. The amount lost through leaching, admittedly, is very dependent on rainfall/drainage during winter.

Theme 11: use of farm systems experiments in developing a digital twin of the dairy system, evaluate farm systems models and as a platform for other experiments, component studies and the value of well-kept records

Background

Farm systems experiments are important for understanding the interacting features of a system, which helps to fully quantify the effects of component change. This characteristic of well-designed farm systems experiments, and the data collated over many years made No. 2 Dairy a valuable asset in the development of a Digital Twin (the Whole Farm Model; WFM; Sherlock et al. 1997; Beukes et al. 2008) and the testing of computer models for accuracy in simulating farm system change.

No. 2 Dairy provided a platform for evaluating the effects of cow-level inputs within a farm system, thereby enabling a greater understanding of interacting effects that were material to on-farm decision making. From the 1970s, there were a lot of experiments superimposed on the existing experiment. To collate all of these would be a review in its own right, but to truly reflect the legacy of No. 2 Dairy we believed it appropriate to briefly present a few of the key component experiments.

The development and evaluation of a digital twin – the whole farm model

As computers became more mainstream and processing power improved, complex mathematic models were developed to describe the behaviour of components of the farm system. For example, Lee Baldwin produced a detailed mathematical representation of post-digestive metabolism in the cow (Baldwin 1995) and many researchers developed models to describe how climate, soil type, and fertiliser inputs predicted pasture growth (McCall 1984; Woodward 1999). But, the ‘holy grail’ of simulation modelling is to develop a ‘digital twin’ of the whole system – a mathematical representation that accurately predicts the outcomes of the changes in the many interacting features of a system. In reviewing 9 models (Bryant and Snow 2008), they identified that very few attempts had been made to combine animal, pasture, climate, and management models into a whole farm scenario because of the complexity of a pasture-based dairy system, relative to one where feed is purchased and provided to housed animals; for example,:

• feed is fresh and cannot be stored for long periods;

• supply is heavily dependent on prevailing climatic conditions, and animal intake is influenced by the amount of pasture that has been produced and the physiological state of the cow;

• total utilisation is further complicated by management decisions about conservation, young stock placement, cow culling and timing of lactation end; and

• the financial and environmental outcomes are a result of all of these variables.

Models of whole farm systems fall into two general categories, either management or research (Plant and Stone 1991). Management models (e.g. UDDER; Larcombe 1990), are designed to use mathematical equations representing empirical relationships within a system to test the outcome of various management strategies; for example, using the knowledge of how much pasture is grown and the profile of growth to estimate how much additional feed would be required if the number of cows on a farm were increased. In contrast, research models use a mathematical representation of a mechanism at the cellular or tissue level to simulate situations for which the answer is not immediately apparent; for example, a knowledge of how climate influences pasture growth and quality and animal health and production to understand the projected effects of climate change on future farm systems. In effect, they are used to improve our understanding of the most likely outcome of hypothetical situations. Both model types have advantages and disadvantages, and most models have both mechanistic and empirical components, depending on the complexity of a relationship being simulated.

It was recognised in the 1990s that the traditional experimental approach could be improved upon by utilising the advancements in hardware and software development and the increasing knowledge of tissue physiology. For example, because of the cost and time commitment of multi-year farm systems experiments and the limited number of treatments that can be logistically undertaken in vivo, an accurate in silico digital twin of the farm system would offer significant opportunities to investigate multiple changes within a farming system on different soil types, under different climates, with different genotypes, and, potentially, with different levels of management expertise.

As discussed in Theme 5, the WFE experiment was established with the aim of developing a digital twin (a WFM; Sherlock et al. 1997; Beukes et al. 2008) to aid in simulating changes in the management of a New Zealand dairy farm. From 1996, Rob Sherlock led a team who set about developing the WFM. The WFM was designed with a framework that allowed the incorporation of existing (and future) sub-model components, and to readily allow contributions from other sources. Data from the first year of the WFE experiment were used to develop the WFM and data from the second and third year used to evaluate the model. The WFM was designed as a research resource, with the intention that it would contain the latest research information, be able to simulate detailed questions relating to pasture management and cow metabolism, and represent any realistic farm management scenario. Although there have been many further additions to the WFM and its ability to predict financial and environmental consequences of farm system change has improved dramatically with time and further experimentation, its inception and consequential value proposition of a Digital Twin of a pasture-based farm system was the result of the detailed measurements at No. 2 Dairy of the biophysical variables underpinning a pasture-based farm system.

UDDER – evaluating a desktop dairy farm for extension and research

The WFE experiment (Experiment 24; Theme 5) was undertaken over 3 years, with 10 farmlets investigating 5 SRs (2.2 to 4.3 cows/ha) in a 5 × 2 factorial arrangement; one set of five SRs was managed by a set of decision rules (DR; discussed in Theme 5 and by Macdonald et al. 2008a; Macdonald et al. 2011; and Roche et al. 2016) developed from farm systems research undertaken at No. 2 Dairy (Macdonald and Penno 1998); the second set of five farmlets were managed with the use of a computer simulation model ‘UDDER’ (Larcombe 1990). UDDER simulations were undertaken fortnightly for each SR to simulate the immediate management predicted to optimise performance during the rest of the season.

When the UDDER and the DR farmlets were compared (Macdonald et al. 2010), the UDDER farmlets produced more milk yield/cow per day (P < 0.01, 0.5 kg) and annually (P < 0.05, 114 kg). UDDER farmlets, therefore, produced slightly more milk per ha annually (P < 0.05, 301 kg/ha per season). Fat and protein content were not affected by management method, so MS/cow per day (P < 0.05, 33 g) and per ha/year (P = 0.08, 19 kg) were also greater in the UDDER treatment. Lactation length was not affected by treatment (256 and 254 lactation days for DR and UDDER, respectively). Estimated pasture grown tended (P = 0.07) to be greater on DR farmlets; but pasture consumed was similar for both management systems. The additional pasture grown was conserved as silage on DR farmlets (P < 0.01, 319 vs 220 kg DM/cow for DR and UDDER farmlets, respectively); however, this effect diminished with increasing SR (management system × SR interaction: P < 0.05). A greater proportion of the UDDER farmlet paddocks were topped after grazing to remove clumps more often than those on DR farmlets (0.79 v 0.47).

Why the DR farmlets produced more pasture/ha than UDDER pastures is unclear, but probably relates to the rotation lengths recommended within the different management decision systems. On average, UDDER recommended rotation lengths 11% (±7.8%; standard deviation) shorter than those employed on the DR farmlets; in particular, average rotation lengths were 8% shorter in spring. Pasture growth follows a sigmoidal pattern, with growth rate increasing with rotation length up to a plateau (Voisin 1959). Shortening rotation length would, therefore, be expected to reduce total pasture production, as more days are spent at a slow phase of pasture growth rather than in the exponential phase (Macdonald et al. 2008b). The decision rules reported by Macdonald and Penno (1998) were designed to maximise pasture production and conserve surplus for periods of insufficient pasture growth. Such a plan is vital in the high SR systems for which the DR were developed. However, in low SR systems, deficits in pasture supply are much less likely and so shorter rotations may be strategically more sensible to reduce the expense associated with pasture conservation.

The two management systems were compared financially using the Red Sky dairy farm financial analysis package. The analysis indicated a tendency (P < 0.1) for UDDER farmlets to have $88/ha greater revenue and $75/ha greater operating profit, primarily because of the greater milk production and the reduced costs of conservation. At the Low SR where DM allowance per cow was high, cows were able to graze more selectively than in the managed herds and a lot of feed was not utilised. The experiment demonstrated that UDDER is an effective decision support tool to aid farm management decisions when supplied with adequate data.

The use of bovine somatotropin to increase cow production

Background

Early research in the 1940s reported that injecting unfractioned extracts of anterior pituitary tissue into cows increased their milk production. Some of the seminal research in this field was undertaken by Peter Brumby and John Hancock at Ruakura in the 1950s (Brumby and Hancock 1955). The effect was isolated to the endocrine factor, growth hormone or somatotropin and became commonly known as bovine somatotropin (bST). However, it wasn’t until the 1980s that it became technically possible and economically feasible to produce large commercial quantities of bST using biotechnology. The bST derived by this process is typically called ‘recombinant’ bST or ‘rbST’.

By the 1990s, rbST was approved for use in 25 countries (Davis et al. 1999) to increase milk production, but it was not approved for use on New Zealand dairy farms and there had been little evaluation of efficacy in grazing systems when compared with the vast numbers of experiments undertaken in housed animals with unlimited access to mixed rations. Studies with pituitary-derived somatotropin indicated that bST was effective in enhancing milk yield in cows on a pasture diet (Brumby and Hancock 1955; Peel et al. 1985) and research with rbST given by daily injection for a whole lactation, demonstrated a milk yield response of +18% in the spring and early summer, when adequate pasture was available; the response declined to zero during the summer drought, but then increased again when autumn rains revived pasture growth (Hoogendoorn et al. 1990); over the entire lactation, rbST increased milk and MS production by 10% to 11%.

There was a reluctance at the time to approve use of rbST in New Zealand and many other countries, primarily because of perceptions of food safety concerns and the fact that the New Zealand economy relied heavily on the export of milk products. But it was recognised that if attitudes changed, there would be a need to evaluate rbST efficacy under the different management conditions and identify the optimum way to integrate rbST into pastoral systems. A consideration was the effects, if any, there might be on animal health and reproduction, to enable the calculation of its potential cost-benefit. An experiment was, therefore, established to attempt to increase cow efficiency with the use of rbST and to determine the potential for its use within the New Zealand dairy industry (Experiment 16).

Experimental methods

The experiment was established by Steve Davis (Davis et al. 1999) and used the commercially available controlled-release preparation of rbST (Somidobove^®). The experiment was superimposed over the New ryegrass cultivar and Matua prairie grass experiment (Experiment 17; 1986–1987), with six cows within each farmlet acting as placebo-treated ‘Controls’ while another two groups of six cows were treated with 320 and 640 mg on four occasions during lactation. The cows were injected with the Somidobove every 28 d. The first treatment was applied in mid-November, about 6 wks post-start of breeding, as there was a concern that the use of bST any earlier may negatively affect cow reproduction.

Results and discussion

The effect of the Somidobove on milk yield was sustained for only 14-d of each 28-d cycle, with mean increases of 18% and 23% for the 320 & 640 mg Somidobove, respectively. The increase in milk production was associated with an increase in blood fatty acid concentrations, indicating that at least some of the increased production was coming from mobilised body fat. During a period of drought (summer), milk yield during week 3 and 4 following treatment was 10% lower in the 640 mg group, relative to the ‘Control’, which may have been in response to lowered feed quality and/or physiological drivers in mid-lactation to replenish used body reserves. There were no adverse effects of Somidobove treatment on any aspect of animal health, including calf birth weight and calving Lwt at the cow’s next parturition. Soon after this experiment was completed, the New Zealand Dairy Board made the decision that rbST was not to be used within the New Zealand dairy industry, so the line of research enquiry was not continued.

Effect of avoparcin on the yield and composition of milk in pasture-fed dairy cows

Background

The predominance of pasture in the cows’ diet in New Zealand is the basis of the world-recognised low-cost dairying system in which there was minimal use of feed supplements or additives. At the time, however, it was believed that additives or supplementary feeds would be more widely used if they were cost-effective or if they provided other benefits, including a reduction in nutritional disorders such as bloat. Recent changes in the New Zealand dairy industry around the use of supplementary feeds and perceived economic benefits have since proven this postulate correct.

Antibiotic growth promoters, such as Avoparcin, are used widely in diets of poultry and pigs to improve efficiency of utilisation of energy and protein components of the feed. Improved feed conversion ratios and Lwt gains have also been demonstrated in beef cattle on a range of diets including silage, concentrates, and low-quality roughages with and without protein and energy supplements (Cuthbert and Thickett 1984; Dodemaide et al. 1988; Gulbransen and Standfast 1988; Lindsay et al. 1988).

Experimental methods

Vicki Carruthers established an experiment that was aimed at investigating the effects of avoparcin on milk production. Sixty early lactating Holstein-Friesian cows of mixed age and with an average BI of 133 were randomly allocated to an untreated ‘Control’ or a treatment group receiving 125 mg avoparcin/cow as an oral drench once daily following a two-week uniformity (covariate) period. Treatments started in mid-September 1988 and continued for 14 weeks until 22 December. The cows grazed on pasture containing ryegrass and white clover and received a new area of pasture after each morning milking, returning to the same area in the evening. Cows were milked twice daily and were drenched each day with bloat oil and magnesium oxide, separately from the avoparcin for the treated cows.

Results and discussion

Neither yield of milk nor MS were increased by avoparcin treatment and there was no difference between the treatment groups in changes in Lwt or BCS over the treatment period.

The effect of sodium bicarbonate on yield and composition of milk from grazing cows in early lactation

Background

Sodium bicarbonate is widely used in concentrate rations to buffer the reduction in rumen pH associated with the rapid fermentation of starch and sugar. Level of inclusion had ranged from about 0.7% to 2.5% of total DMI. Effects on MS production, however, had been inconsistent, with reported improvements in milk yield but not fat content (Kihner et al. 1980; Fisher and MacKay 1983), fat content but not milk yield (Snyder et al. 1983; Rogers et al. 1985; Solorzano et al. 1989), or no effect on either milk yield or fat content (Donker and Marx 1985). Some farmers in New Zealand had claimed improved MS yields through supplementing cows grazed on pasture with sodium bicarbonate in the drinking water (Dairy Exporter 1990) and there was increasing concern that the low rumen pHs reported in cows consuming high quality ryegrass-white clover pastures may be indicative of subclinical rumen acidosis (SARA) and that cows might benefit from supplementation with buffers or alkalising agents (de Veth and Kolver 2002).

Experimental methods

Vicki Carruthers initiated a series of 4 experiments at both No. 2 and No. 5 Dairies. The experiment established at No. 2 Dairy was to evaluate the effects of supplementing cow with sodium bicarbonate by drenching the cow once daily or ‘fortifying’ drinking water with 100 g/cow per day of sodium bicarbonate.

Treatments were administered to each group according to a Latin square design, with the first 21-d period starting 41 days after calving. The 3 groups of 20 cows were grazed as three separate herds, with each offered 0.27 ha of fresh pasture after the morning milking. The only source of water for the cows on the trough treatment was a portable tank and trough placed within the area of pasture being grazed. The water volume consumed each day by each treatment group was estimated using flowmeters inserted in the trough inlet lines. The tank contained a solution of sodium bicarbonate sufficient to provide 100 g/cow/day. An estimate of the amount of water likely to be drunk by cows each day was made based on previous day’s intakes and on the weather, and the amount of water added to the tank was adjusted accordingly. Thus, in an attempt to keep daily intake of sodium bicarbonate the same, the concentration varied from day to day depending on expected water intake.

Results and discussion

In the No. 2. Experiment, adding sodium bicarbonate to the water supply increased milk yield and yields of protein and lactose, though the results from the other experiments at No 5 dairy were variable. Water intake was increased by about 26% when sodium bicarbonate was added to the drinking water.

Salt supplementation of lactating dairy cows

Background

Sodium (Na) is an essential element for animals but is not required by plants. Soil reserves of Na were declining in many New Zealand soils (Mike O’Connor, Personal Communication) and it was expected that an inevitable consequence of this decline would be to reduce pasture Na concentration. High- producing dairy cows require more than 20 g Na/day (Underwood and Shuttle 1989) and supplementation with Na can be either direct (via oral supplements) or indirect through fertiliser and animals ingesting it through increased Na in pasture.

Reported responses to Na supplementation varied, with O’Connor et al. (2000) reporting an increase of 12.8% in MS production in New Zealand when cows grazing pastures containing 0.05% Na were supplemented with 15 g Na/d. Responses were immediate and persisted throughout the experimental period (November 1999 to February 2000). Furthermore, higher-producing cows responded more to Na supplementation compared with low-producing cows.

Some pasture plant species (e.g. white clover and perennial ryegrass-) readily take up Na, but others like lucerne (Medicago sativa), browntop (Agrostis tenuis) and Kikuyu (Pennisetum clandestinum) have low rates of uptake. Sodium concentrations in perennial ryegrass-dominant pastures at No. 2 Dairy in the 1990s averaged 0.16% DM, sufficient to provide 26 g/d for cows consuming 16 kg DM but varied from 0.1 in summer to 0.2 in spring (i.e. 16 to 32 g/d, respectively, for a cow eating 16 kg DM; Roche et al. 2009b). So, deficiencies are possible within a year and, certainly, in farms with pastures containing lower concentrations of Na/kg DM.

Experimental methods

An experiment was established in 2001 by Kevin Macdonald and utilised the 188 Holstein-Friesian cows on the 10 farmlets that were part of a multiyear WFE experiment previously discussed in Theme 5 (Macdonald et al. 2008a; Macdonald et al. 2011). Cows were randomly allocated to a salt supplement or ‘Control’ group (within farmlets), ensuring that there was an even distribution of age, genetic merit and pre-experimental milk and MS production. The experiment ran from the 31st January to 21st February 2001, the period predicted by Roche et al. (2009b) for pastures to be lowest in Na at No. 2 Dairy. Treated cows received an oral supplement of 14 g Na/cow/day in the form of sodium chloride (NaCl; 35 g/cow/day) at the morning milking.

Results and discussion

Following a period of low rainfall, pasture quality declined from an average of 10.7 ± 0.5 MJ ME during the pre-experimental period. Herbage sodium concentrations varied within farmlets (0.10% Na ± 0.04). Pasture potassium (K) averaged 3.34% DM and ranged from 1.65% to 5.47% DM. Calculated pasture Na consumed ranged from 8 to 23 g/cow/day. Supplementation with Na had no significant effect on milk yield, milk fat, milk protein or MS yields, with the general conclusion being that there was no evidence to indicate that cows in mid-lactation grazing pastures containing 0.06% to 0.14% Na required Na supplementation.

Magnesium supplementation of lactating dairy cows in the summer

Background

The range of magnesium (Mg) concentration in New Zealand pastures is wide and this wide range is in part due to the many different soil types and the variations in botanical composition of New Zealand pastures. Seasonally, pasture Mg levels tend to be least in late-winter/ early-spring and greatest in summer and autumn, in apparent disharmony with K (Roberts 1987; Roche 1999). This change is often attributed to increasing amounts of clover in the sward as clover has a greater Mg concentration than ryegrasses and superior growth rates in high temperature settings. The uptake of Mg (from the soil) by the plant involves both passive and active mechanisms. There are many factors that affect this uptake and its subsequent availability to the animal. The most important of these is high K concentration, which can reduce both herbage Mg concentration and the absorption of Mg by the animal. Both these factors lead to increased risk of metabolic disease (Roberts 1994).

In New Zealand, dairy cows are generally supplemented with Mg several weeks pre-calving until late-spring to overcome the risk of hypomagnesaemia (Young et al. 1981) and associated hypocalcaemia (Roche and Berry 2006; Roche et al. 2013). The use of Mg supplements has reduced the prevalence of hypomagnesaemia, but it still results in a large economic loss, which was estimated by Towers (1994) to be $18 million annually in reduced milk production and $1 million-$10 million/year in cow deaths.

Young et al. (1979) reported milk fat increases of 10% to 15% for cows supplemented with Mg and there was anecdotal evidence from dairy farmers that Mg supplementation in the summer increased milk production and that cows were less nervous during milking. In 1994, Ellison (1994) reported that there had been a steady increase in the number of herds found to be hypomagnesaemic since 1988, especially in the autumn, and unpublished results from No.2 Dairy identified a decline in blood Mg with increasing SR. Therefore, there was a justification to investigate the effects of summer Mg supplementation on cow behaviour and production.

Experimental methods

The aim of the Mg supplementation experiments established in 1999 by Kevin Macdonald, was to measure MS response to Mg supplementation during the summer and autumn and to determine if serum Mg concentrations differ in herds managed at different SRs. The effect of Mg supplementation on cow temperament in summer and autumn was also of interest.

The experiment was superimposed on 188 cows that were already being investigated in the multi-year WFE experiment previously described in Theme 5. Cows were allocated within farmlets, balancing for age, pre-experimental blood Mg concentration and MS production, to receive either 0, 10 or 20 g Mg supplement/day, in the form of Magnesium Chloride (MgCI².6H²O). This was prepared as a solution and given to the cows orally (0, 120 or 240 ml solution) at the morning milking. The experiment was designed as a Latin Square, with three periods of 14-day, covering the period of 20th January to 24th May 1999, and each cohort of cows being exposed to each Mg supplementation rate.

Results and discussion

Within the range of Mg intakes of the cows, there was no effect of supplementation on production. Even at a blood concentration of 0.75 mmol/l (the lowest measured mean for any of the farmlets), cows did not display any signs of clinical or subclinical Mg deficiency. The effect of additional Mg supplied to the cows could have been negated by a high pasture K concentration (3.29% DM, SEM 0.102), which is known to reduce Mg absorption in the rumen; however, the effect is not complete and some of the supplemental Mg would be expected to be absorbed either ruminally or post-ruminally (Schoneville et al. 1999).

A greater Mg absorption is consistent with evident changes in urinary Mg to creatinine ratio, which increased linearly with increasing dietary Mg (1.38, 1.87, 2.01 for the 0, 10 and 20 g treatments, respectively). These results confirm greater Mg absorption, but that Mg surplus to requirement was being voided by the cow. The results highlight that it was unlikely that even at higher SRs, a milk production response to Mg supplementation will occur in the summer/early-autumn period when pasture Mg concentration is ≥0.18%. Supplementation with Mg did not affect cow behaviour in situations where they would be expected to show additional signs of nervousness.

Experiments to improve on-farm reproduction outcomes

Background

The success of seasonal grazing systems, such as used in New Zealand, is dependent on each cow re-calving every 365 d, in theory, and the herd calving within a short timeframe (i.e. ∼60 d, including replacement heifers; Holmes 1995; Roche et al. 2017b). Although this is a successful strategy for maximising pasture utilisation (Roche et al. 2017a, 2017b) it also means that the period of breeding cows is equally short and requires the herd to become pregnant in an equally short timeframe and requires cows to achieve a successful pregnancy within 83 days of calving. Cows, therefore, must return to oestrus quickly after calving, be identified by the farmer as being in oestrus, and be highly fertile.

In the 1970s, as SR and cow numbers on farms increased, and before tactical pasture management (discussed in Theme 2) and BCS management was optimised, there was an increase in the duration of post-calving anoestrus. Reproductive problems had, over the years, limited the performance of the national herd. Non-cycling cows, late-calving cows, and poor heat detection all reduced submission rates, spread calving and reduced overall reproductive performance.

Experiments on heat detection

These limiting factors played an important part in Jock Macmillan’s approach to tackling the problem, as his research at No. 2 dairy was based around a series of experiments aimed at understanding the physiology behind the non-cycling cow and developing technology to assist with achieving an early and successful breeding event.

With increasing herd size came a reduction in rates of oestrous detection. This resulted in the development of tail-painting, arguably one of the most simple and, yet effective solutions to a critical problem in larger herds. The concept was tested at No. 2 dairy as well as on commercial herds (Macmillan and Curnow 1977), which provided an estimate of potential effect in extremely well-resourced and well-monitored research herds and in the commercial world of large herds being managed by few people.

Results and discussion

When tail-painting was introduced to No. 2, Des Clayton (Technician in charge) believed that its use would not make any difference in the No. 2 herds because cow monitoring was undertaken so diligently in the research setting. However, when Tony Day (Ruakura Veterinarian) examined the unmated cows after 3 weeks of AI, he identified that most were non-cyclers, but some had ovulated (and had rub marks; pers comm Jock Macmillan). These results identified one of the key problems associated with poor reproductive problems in New Zealand farming systems and provided farmers with a simple way of identifying cows that should be examined by the vet in a timely manner. The technique was quickly adopted widely in grazing systems all over the world. No. 2 Dairy was also used for testing new brands of tail-paint that were coming onto the market.

Experiments on cow nutrition for improved reproduction and the development of the controlled internal drug releasing insert (CIDR)

Cow nutrition, through its role in BCS at calving, is a major factor determining reproductive efficiency (McDougall 1993; Burke et al. 1995; McDougall et al. 1995; Roche et al. 2009e). End of lactation BCS declines linearly with SR (Roche et al. 2009e). So, as SR increased through the 1960s and 1970s, there was a risk that resultant poor cow BCS in early lactation would negatively affect reproductive success; this was confirmed in Experiment 19, wherein increasing SR from 3.0 to 4.0 Holstein-Friesian cows per ha increased calving to mating interval from 35 to 50 days because of the effect of SR on extending the post-partum period of anoestrus.

A number of studies were undertaken in an effort to identify treatments that would overcome the problem of anoestrus in pasture-based herds. The development of the CIDR to administer progesterone vaginally for periods of 5 to 14 d in dairy cows and heifers that were identified as anoestrus (Macmillan and Peterson 1993) was started by Bob Welch. A development team from Ruakura was formed with Bob Welch and Jock Macmillan (field trials) in collaboration with Doug Millar (AHI Plastic Products Ltd) and Graham Duirs (commercialisation by AHI Agri Division). This collaboration between Ruakura research and a commercial company was a major achievement for Jock and his team. Much of the underpinning research was undertaken at No. 2 Dairy and on many commercial dairy farms. The insert was used in combinations with injected Prostaglandin F2α (PGF2α), Gonadotropin-releasing hormone (GnRH), and oestradiol benzoate. Its use became the standard treatment for anoestrous cows in seasonally bred dairy herds. Jock and his team identified that there were several prerequisites for treatment to produce consistent results.

1. cows must be a BCS of 4 or better;

2. cows should be on improving nutrition; and

3. treated cows that don’t return to service should be pregnancy tested to ensure that they haven’t relapsed into anoestrum.

The use of CIDRs was extended to use in maiden 15 mth old heifers, so that oestrus of the group could be synchronised. This facilitated greater use of AI in heifers and, as a result, increased rates of genetic progress and the number female calves to select for replacements. The yearling heifers at No. 2 dairy were a major part of this project for Jock and his team.

Proving the value of data

From the beginning, meticulous records were kept at No. 2 Dairy. In the early 1990s, Jim Lancaster collated the old hand-written records of McMeekan’s experiments into Microsoft® Excel files and we used these for reanalysis of the data from 1944 to 1964. As Dexcel moved its operations from No. 2 Dairy to Scott Farm on Vaile Road in 2001, John Roche and Julia Lee uploaded pasture and cow data from 1970 to 2001 into a Microsoft^® Access database. The Lancaster and Roche & Lee datasets have been the basis of much of this compendium, with additional information drawn and collated from Department of Agriculture Annual Reports, the annual Ruakura Farmers’ Conference proceedings, and Science Journals.

The bulk of the data collated by Roche & Lee were collected between 1986 and 2001 (15 yr) from more than 60 research treatment herds. Daily climate data from Ruakura station (∼1 km from No. 2 Dairy) and data on 2635 lactations from 897 cows were available. Of the 2635 lactations, 374 (14%) were Jersey and the remainder were Holstein-Friesian. The dataset included more than 50,000 BCS records; a total of 2594 service records and 2463 lactation records, where each cow had at least 20 test-day records.

The Roche & Lee database was to prove the value of data collection and collation, as it was used to:

• underpin the BCS management strategies consistently applied in New Zealand, including in the Dairy Cattle Code of Welfare; and helped illuminate

• the role of climate in pasture quality and the role of both climate and resulting pasture quality in animal production;

• the role of climate and cow-level factors in the risk of milk fever;

• the association between climate and cow body condition score on secondary sex ratio; and

• it was used for reanalysis of Experiments 19, 22 and 25, and these data were used in a number of Masters’ Theses and in the preparation of papers by Spaans et al. (2018, 2019) and Edwards et al. (2019).

Equally importantly, however, it was to demonstrate the relevance of historical experiments in answering pressing research questions and impress upon scientists the hidden value of analysing large databases to understand the most appropriate hypothesis for testing in intervention experiments. In this section, we briefly summarise some of these papers validating our management systems, but which were only possible because of the long-term and comprehensive datasets collated.

Factors affecting BCS and BCS change and the association among BCS variables and milk production, reproduction, health and welfare, dystocia and birthing ease, and the secondary sex ratio of dairy calves

Seven manuscripts were published in the Journal of Dairy Research and the Journal of Dairy Science from the Roche & Lee Database. These described the unique features of the BCS profile in grazing dairy cows in New Zealand (Roche et al. 2007a), the effects of cow age, SR, calving condition score, and feeding system on the BCS profile (Roche et al. 2007a), as well as the relationship between BCS and Lwt (Berry et al. 2006), and the association between BCS and BCS change in early lactation and mastitis (Berry et al. 2007), dystocia (Berry et al. 2007), milk production (Roche et al. 2007b), reproduction (Roche et al. 2007c), and with the secondary sex ratio (male to female ratio at the time of birth) at the following calving (Roche et al. 2006c). They culminated in the invitation to produce a comprehensive review in the Journal of Dairy Science, ‘Body condition score and its association with dairy cow productivity, health, and welfare’ (Roche et al. 2009e), for which the award for the most cited review in the Physiology and Management section of the Journal between 2010 and 2012 was received from the American Dairy Science Association.

Briefly, it was well known that cows, like all mammals, lose condition post-parturition, although that loss has been accentuated through genetic selection for production traits. The profile of BCS change differs in New Zealand, however, in that it follows a ‘W’ shape, with cows losing BCS in mid-lactation, before regaining it again in late lactation (Roche et al. 2007a). It is unclear why this occurs, although less-accentuated changes in BCS have been reported in pasture-based systems in Ireland, wherein BCS gain ceases through mid-season, and the same profile has been presented in Lwt in cows on tropical pastures in the USA. It would appear, therefore, that the effect is mediated, at least in part, by mid-season pasture quality and, potentially, some degree of heat stress.

A first derivative analysis of the BCS profile quantified, for the first time, the daily rate of change in BCS through lactation in grazing dairy cows (Figure 11); despite an inability to accurately measure DMI in grazing cows, this provided a graphical representation of energy balance. Cows have a maximum rate of BCS loss in the first 10 days post-calving; in fact, cows lose ∼25% of the BCS they will lose in the first week post-calving. From Day 10, the negative energy balance becomes increasingly smaller until between 50 and 70 d post-calving, when the cow enters a positive state of energy balance and BCS begins to increase. Between 120 and 210 d post-calving, cows again entered a state of negative energy balance (W-shaped profile; Roche et al. 2007a), losing BCS through mid-lactation, although the rate of loss is only 25% of the rate identified in early lactation. Interestingly, the shape of this profile was not materially affected by calving BCS, cow age, or SR, although the size of the peaks and troughs in the profile were impacted. Key lessons in quantifying the profile (Roche et al. 2007a):

• Higher BCS cows lose more condition in early lactation, but remain fatter at both the spring and summer nadir; importantly, however, the difference in BCS evident at calving had shrunk by more than 50% by mid-lactation (from 2.5 BCS units difference to 1 BCS unit difference within 150 days);

• Despite calving at a greater BCS, young cows (≤3 year old) lose no more BCS than older cows in early lactation; but they are at a greater risk to BCS recovery in late lactation;

• Except in extreme circumstances (e.g. 4.3 cows/ha), cows in low SR systems (i.e. greater feed allowance/cow) tend to calve fatter, but lose an equal amount of BCS in early lactation to cows in higher SR systems and the shape of the lactation BCS profile is similar; and

• Nutrition has little influence on the rate of BCS loss in early lactation, but starch-based feeds increase the rate of BCS gain in mid-lactation, leaving less BCS to be gained during the dry period;

Figure 11. Daily rate of change in BCS in pasture-based cows. Profile derived from the average BCS profile of 1172 cows across 3209 lactations at No. 2 Dairy between 1986 and 2004. Source: Roche et al. (2007a).

In addition to mathematically describing the BCS profile, the database facilitated the quantification of the role of BCS in milk production, reproduction, and aspects of animal health. Milk and MS yields were nonlinearly associated with calving BCS, increasing at a declining rate up to BCS 6.0 to 6.5 (10-point scale), and declining thereafter. There was very little increase in milk or MS yield, however, above a calving BCS of 5.0: MS production was 18 kg greater in cows calving at BCS 4 vs BCS 3, but only 12 and 6 kg greater in cows calving at BCS 5 vs BCS 4 or BCS 6 vs BCS 5, respectively.

Calving BCS was very important in the timing of onset of oestrus, with the odds of a cow cycling by the planned start of the seasonal breeding period positively associated with calving BCS. Although calving BCS was not positively associated with pregnancy outcomes, it was positively associated with nadir BCS and was, in fact, the only variable influenced by management that affects nadir BCS in a healthy cow. As nadir BCS was positively associated with pregnancy at 21, 42, and 84 d post-planned start of mating, calving BCS was identified as a critical management factor influencing reproductive success.

The association between calving BCS and BCS change through early lactation and health traits measured was not straight forward. Increased BCS at calving was associated with reduced somatic cell count (an indicator of udder health) in first and second parity cows, but was associated with greater somatic cell count in cows of third parity or greater. Increased BCS loss in early lactation was associated with lower somatic cell count and a reduced probability of a high test-day somatic cell count. Nevertheless, despite the statistical significance of the associations between BCS and udder health, most lacked biological significance within the ranges of BCS and BW generally observed on-farm.

Body condition score 8-wk precalving or at calving or BCS change precalving did not significantly affect the odds of a difficult calving or stillbirth. Cows that experienced dystocia lost, on average, more BCS and BW between calving and nadir and had lower nadir BCS and BW and reduced 60-d milk yield; pregnancy rates to first service and throughout the 12-wk breeding season were also compromised.

In a curious, but little reported hypothesis, the database also provided evidence in support of the Trivers-Willard hypothesis (Trivers and Willard 1973): maternal condition at or around conception affects the secondary sex ratio (male to female ratio at the time of birth) in mammals. There was a linear relationship between the probability of a male calf and BCS change between calving and conception, the rate of BCS change over this period (BCS divided by days in milk), and BCS at the calving event immediately before conception. The birth of a bull calf was 1.85 times more likely in cows that lost no BCS from calving to conception compared with cows that lost one BCS unit (1–5 scale: approximately 2.25 BCS units in the 10-point scale) from calving to conception. This increase in odds was equivalent to a 14% unit increase in the probability of a male calf (from 54% to 68%). From a farmer perspective, the data suggested that fatter cows at calving were more likely to show oestrus before the planned start of mating, be fatter at mating and, therefore, were more likely to successfully conceive, and, were more likely to have a heifer calf at the following calving.

Conclusions and implications

Collectively, the data provided a quantitative basis for the mid- and late-lactation management of cow condition to ensure mature cows were a BCS of 5.0 at calving and 2- and 3-year olds were prioritised to ensure a calving BCS of 5.5. These recommendations are consistent with those of Campbell (Theme 2). The results also provided farmers with evidence of the futility of attempts to manage BCS in early lactation. The results were compiled in a farmer-friendly resource (DairyNZ body condition scoring. The reference guide for New Zealand dairy farmers), helping farmers to understand the reasons for the established recommendations. The results of these data analyses also provided a strong defence for the proposed minimum standard for BCS in the Code of Welfare for Dairy Cattle: ‘When the body condition score of any animal falls below 3 (on a scale of 1–10), urgent remedial action must be taken to improve condition’. The code both acknowledged the normal distribution of minimum BCS within a dairy herd (i.e. thin individuals are natural within a herd) and recognised that individual cows at BCS 3.0 or greater were not at more risk of animal health problems.

Climate and its role in pasture quality, their collective role in animal production, and the role of climate and animal-level factors in secondary sex ratio and the risk of milk fever

All plant growth, and hence animal production in grazing systems, is the cumulative result of the factors promoting or inhibiting growth over the production cycle and Fulkerson et al. (1998) suggested that the first step in balancing rations for pasture-based dairy cows should be to acquire accurate data on the nutrient content of the pasture and the likely changes in nutrient availability as prevailing environmental conditions change. Most individual experiments are too small to investigate the inter- and intra-relationships between climate, pasture quality, and animal production, despite collecting all of the relevant data, but the large dataset of No. 2 data compiled by Roche and Lee provided the first real opportunity to investigate these relationships.

Weather data were recorded daily at 0900 hours at Ruakura Research Station, Hamilton, New Zealand (37°46′S, 175°18′E and 40 m above sea level), such that a total of 2215 daily records were available between from 1 July 1995 to 30 June 2001. Samples of herbage offered to cows were collected weekly from 1995 to 1997, and monthly from 1997 to 2001, providing almost 1000 data records for pasture quality. Individual cow milk yields and composition was measured on one day each week and BCS and Lwt every fortnight. This provided almost 120,000 test-day records and more than 60,000 BCS and Lwt records.

Air and soil temperatures, radiation, and potential evapotranspiration followed distinct and quantifiable sinusoidal profiles and were highly repeatable within fortnight across years; but, the repeatability of rainfall was ≤7%. This provided confidence that patterns of pasture growth would be consistent, at least during winter and spring when soil moisture was more likely to be optimal for plant growth, but unpredictable in summer and autumn. As expected, air and soil temperature measurements were highly positively correlated with each other (r = 0.53–0.99), and with evaporation (r = 0.40–0.68) and potential evapotranspiration (r = 0.43–0.79). Maximum air temperature was also positively correlated with radiation (r = 0.61). Fitted sinusoidal functions demonstrated cyclic temporal trends across pasture quality variables, but there was little cyclic temporal variation in the majority of pasture mineral concentration variables; the within fortnight repeatability of pasture quality measurements was low to moderate (22% for ether extract to 54% for ME). Many of the pasture quality variables were associated (e.g. the different measures of fibre were strongly correlated; r = 0.87) and negatively correlated with pasture digestibility (r = –0.64 to –0.74), water-soluble carbohydrate concentration (r = –0.52 to –0.68) and ME content (r = –0.60 to –0.75). With a few exceptions, correlations among most pasture minerals were poor.

Weather explained up to 14% of the variation in herbage nutrient content over and above that explained by time of year and farmlet. There were moderate correlations between some weather variables and pasture quality measures and mineral concentrations, but the strength of the absolute correlations decreased following adjustment of the pasture-related variables for month of year, suggesting any associative effects were more likely a result of season than climate. Small, but significant negative correlations existed between rainfall and pasture water-soluble carbohydrate (r = –0.19) and organic matter digestibility concentration (r = –0.13) and ME content (r = –0.14), independent of time of year.

The analysis indicated moderate relationships between some weather- and pasture-related variables and dairy cow production, although most relationships weakened following adjustment for animal parity, stage of lactation, and week of the year at calving, suggesting correlations were likely an artefact of temporal variation rather than a true relationship. Even so, following adjustment for the confounding effects, all associations between DMI and weather were rendered nonsignificant. This probably indicates that management can overcome the effect of weather on DMI. Indeed, the decision rules discussed in Macdonald and Penno (1998) were designed to remove subjectivity from the management of farm systems research. Notwithstanding the effect of accounting for season of the year on the associations, weather, pasture quality and mineral concentration still explained up to 22% more variation in dairy cow production variables than farmlet and time of year. Milk yield was positively associated with sunlight hours (r = 0.14), however, and there was a negative relationship between temperature-related variables and milk protein concentration (r = –0.08), regardless of time of year. Milk protein concentration was positively associated with pasture ME content (r = 0.06), water-soluble carbohydrate (r = 0.11), and organic matter digestibility (r = 0.06) concentrations, and negatively associated with ether extract (r = –0.07), acid detergent fibre (r = –0.06), and neutral detergent fibre (r = –0.05) concentrations. Although it is unsurprising that milk protein is positively associated with weather variables associated with grazing and total DMI (i.e. sunlight hours) or pasture quality, this was the first study to quantify the amount of variation in animal production that could be explained by changes in weather and nutritional composition. The results could be used to further develop the predictive ability of mechanistic models.

The climate and animal databases were also interrogated to examine the likelihood of an unusual relationship – whether weather impacted the secondary sex ratio (SSR, i.e. the proportion of male to female offspring born). Although probability theory dictates that the SSR should be 50:50 in situations of evolutionary equilibrium, substantial evidence exists that both the primary sex ratio (male-to-female ratio at the time of conception) and the SSR (male to female ratio at the time of birth) can be strikingly skewed from this balance (see Roche et al. 2006b). Various factors have been associated with a positive or negative effect on secondary sex ratio (e.g. maternal age, pre- and post-conception nutritional impacts on the uterine environment, latitude, and social dominance), but analyses have yielded inconsistent results. For example, Cameron (2004) reviewed the association between season of the year and SSR and reported that the number of studies reporting an effect for season was approximately equal to those reporting no effect. Some researchers have reported an effect of climate on the SSR in mammals, including humans. For example, Lerchl (1998) reported a bimodal seasonal distribution of the SSR in humans born approximately 6 mo apart, indicating that the human SSR may be influenced by some climatic variables. It is plausible that environmental conditions (climate) near conception may influence the secondary sex ratio, either through alterations in the internal milieu or physiological feedback mechanisms associating climate with the future feed supply altering the primary sex ratio. However, the limited examples presented were from high latitudes, where daylength varied significantly from summer to winter, or tropical latitudes, where temperatures and humidity varied significantly between dry and wet seasons. In a seasonal calving system like that dominating dairy farms in New Zealand, climate extremes within the breeding season are moderate. It, therefore, offered a unique opportunity to determine if subtle changes in climate were associated with SSR.

Data on cow breed, parity, calving date, BW immediately post-calving, sex of the calf, and incidence of multiple births were available for 8716 lactations from 1897 cows between the years 1970 and 2003 in the database compiled by Roche and Lee. Results indicate that climatic factors, even within a relatively short 12-wk seasonal breeding period in late spring/early summer in temperate latitudes, are associated with SSR in dairy cattle. A male calf was more likely to be born following periods of elevated air temperature, greater evaporation, or both. A 1°C increase in average maximum or minimum air temperature, during the week immediately prior to conception, was associated with a 1-percentage unit or 0.5% unit increase, respectively, in the probability of a male calf being born (i.e. from 52% to 53%). An additional millimeter of evaporation per day increased the probability of a male calf being born by 2.9 percentage units. Furthermore, the probability of a male calf being born were 3.74 times greater when the immediately previous calf born was male. Breed of cow, year of conception, and parity at conception did not affect the SSR. These results have become very interesting in the era of sexed semen, where farmers are aiming to improve the chances of a female dairy replacement calf from their highest genetic merit cows and the probability of a male calf with better beef characteristics; although the farmer cannot control the weather, they and their advisers can use the relationships identified, along with the effect of calving BCS, to improve their selection of suitable animals for the different calf choices. The results may also have implications for other mammalian species.

Summary

The New Zealand dairy system of farming is recognised globally for its efficiency, wherein milk is produced at scale and at operational costs vastly superior to any other country. Less well known, however, are the experiments that underpin the DR and decision support resources so important to the systems success. The No. 2 Dairy farm unit at Ruakura is globally recognised for the research work supporting the development of the New Zealand dairy system; but, because of a failure to publish experimental results in the traditional scientific channels, the foundational research and underpinning rationale for the system design are at risk of being lost. In an effort to avoid this, we have presented a compendium of the farm system experiments undertaken at No. 2 Dairy between the arrival of Campbell Percy McMeekan (Mac) as the Research Station Superintendent in 1944 and its closure as an experimental station in 2004. We describe 21 full-year or multi-year farmlet experiments, plus 10 component experiments that were designed to complement the farm system work in progress or further investigate aspects of pasture and cow management.

From McMeekan’s ground-breaking work in the 1950s with ‘Controlled’ and ‘Uncontrolled’ grazing systems, he postulated that there were three principles involved in the efficient conversion of pasture into milk:

1. The efficiency of the process depends on the amount and seasonal distribution of the feed grown;

2. The proportion of the feed grown that is actually harvested by the animal; and

3. The efficiency with which the animal uses the feed it consumed.

These principles were to underpin the design of the modern pasture-based dairy farm system and were to guide experimental priorities at No. 2 Dairy until its closure in 2004.

The superiority of Rotational (‘Controlled’) Grazing over Set-stocking (‘Uncontrolled’) in milk production and young-stock health and growth was demonstrated repeatedly at No. 2 Dairy, but it was the interaction with SR that was to underpin Mac’s two first principles with greatest effect: rotational grazing combined with greater SRs led to greater pasture yield and utilisation directly by the cow, such that milk production/ha increased by ∼25%. Further research qualified the effect of pasture growth potential, N fertiliser, and imported supplements on optimum SR (i.e. Comparative SR; kg Lwt/t DM feed available) and quantified the optimum SR for operating profit (i.e. ∼85–90 kg Lwt/t DM feed available) and, importantly, the environmental implications of greater and lesser SRs. Carter convincingly demonstrated a lack of interaction between cow genotype and SR, promoting the importance of genetic improvement, as per Mac’s third principle, irrespective of farming location or system. Forty years of research was to follow that brought greater resolution to the three principles:

• testing the value of different pasture species and cultivars and grazing strategies for DM production, persistence and the seasonality of growth;

• developing strategies to optimise the transfer of feed from periods of surplus relative to demand (i.e. autumn and spring) to periods of deficit (i.e. winter and summer) and whether importing feed from outside the farm would improve efficiency and profitability;

• defining the most suitable cow genotype and breed for a grazing system; and

• determining the impact of farm management and farm system change on environmental sustainability.

Another important feature of No. 2 Dairy was in its use for extension and education. Frequent visits to ‘No. 2′ by individual farmers and discussion groups, and attendees at the Ruakura Farmers’ Conference Field Days were important in ground-truthing research results and the ensuing discussion important in convincing farmers of the need for change. The impact of No. 2 Dairy on the success of the New Zealand dairy industry would be impossible to quantify, particularly as the component research undertaken elsewhere was equally important in designing the farm systems experiments embarked upon; however, the system developed over this ∼60 year period, through trial and error, is what has become known synonymously as the New Zealand system of dairy production globally, highlighting the legacy of the experimental station and the brilliant minds that dedicated their lives to the pursuit of system optimisation. We hope that the compendium helps farmers, their advisers, and researchers to understand the scientific basis for many of the practices that have become routine in the management of successful dairy farms (e.g. the SRP, or the BCS decision rules) and stimulates further hypotheses on how we can improve upon an excellent platform to deal with the challenges of the future (e.g. climate change, nutrient losses).

Acknowledgements

The authors acknowledge the commitment of all of the principal science investigators and the dedicated technical and farm staff at No. 2 Dairy. Although fraught to identify individuals among such an illustrious team, Jim Keir, Des Clayton, Henry Walker, and Jim Lancaster deserve special mention for their dedication as technicians to science excellence through the ages.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Appendix. General materials and methods for No. 2 Dairy

The No. 2 Dairy Research farm at Ruakura was located on the eastern outskirts of Hamilton city, New Zealand (37°47′ S, 175°19′E and approximately 40 m above sea level) with average rainfall from 1979 to 2004 being 1200 (±164) mm/y. The area was, at the time of establishment, considered as being of marginal quality for dairy production, as the land was on the edge of a peat swamp. Depending on the experiment, between 30 and 60 ha were devoted to permanent grassland and divided into small paddocks (i.e. defined grazing area), each of which was contained by electrifying galvanised wire, and serviced by laneways (i.e. cow tracks) that facilitated cow and machinery movement, and water for drinking troughs (Roche et al. 2017). During the 1965/66 season, research at the dairy was disturbed by a re-fencing and re-roading programme. The farm was permanently fenced with either netting or 6 wire fences into 4-acre (1.6 ha) blocks, and within these, subdivided by 2 wire electric fences into 1-acre (0.4 ha) paddocks, with each paddock had a water trough, allowing the cows unrestricted access to water. From 1991 to 1996 a experiment involving rearing of calves was established (and managed by the No. 2 staff) at the ‘Grazing Unit’, a farm that was contiguous with No. 2 Dairy. In 1998, the area that encompassed No. 3 Dairy was brought into No. 2 dairy and all cows were milked at No. 2 dairy; this meant that 2 major farmlet experiments (14 farmlets; Experiments 24 & 25) were being managed simultaneously on 82 ha.

Soil was predominantly a Te Rapa peaty silt loam, which is a Humic Aquic Haplorthod in soil taxonomy or a Humose Groundwater-Gley Podzol in the New Zealand classification.

Fertiliser was applied annually:

• up to 1965, 250 kg/ha of serpentine superphosphate/year (N:P:K:S 0:6.8:0:8.4);

• In the early 1960s, some of the paddocks had the lowest soil K levels ever recorded in the Waikato (Campbell et al. 1977);

• from 1965 to the mid 1980s, 500 kg/ha of 15% potassic superphosphate/year (N:P:K:S 0:7.7:7.5:9.4);

• 1985, 600 kg/ha of superphosphate (N:P:K:S 0:9:0:11) and 100 kg/ha of muriate of potash (N:P:K:S 0:0:50:0) were applied annually;

• Nitrogenous fertilisers were not applied until 1979;

• 100 kg/ha muriate of potash (50 kg K/ha) applied to any pasture in which silage had been conserved.

• Milksolids production/ha without feed importation increased from approximately 575 kg (average for 1945–1955: McMeekan 1956), with young stock reared on the farm, to about 925 kg 35 years later (Bryant et al. 1981), with young stock reared off farm, and, consistently, to just over 1100 kg/ha two decades later, with the incorporation of up to 200 kg/ha of N from urea fertiliser (Macdonald 1999; Macdonald et al. 2017).

• The original herd was predominantly Jersey and was one of the first to use the New Zealand Dairy Board's artificial breeding service and early experiments on the use of AI were undertaken on cows at the dairy. From 1974, the herd was inseminated with Holstein-Friesian semen and gradually transitioned from Jersey to NZ Holstein-Friesian.

The farmlets were managed as seasonal calving systems, with cows generally calving over an 8–12-wk period in spring, with breeding starting in early October, except for Experiment 25 (Table 1). Approximately 20% of cows from each farmlet were culled each lactation, on the basis of reproductive failure, health, age, and genetic merit, and were replaced with primiparous cows 1 mo before the planned start of calving. Within experiments, age structure did not differ across treatments.

In Experiments 2 & 3 (Table 1) the Uncontrolled farmlet had 2 paddocks and during lactation the cows spent between the am and pm milking in one paddock and between the pm and am milkings in the other paddock. The Controlled farmlet had 15 paddocks, ranging in size from 0.65 to 1.70 ha. The paddocks were grazed based on DM available rather than a strict rotational grazing pattern, with the cows in the allocated area for 24 hrs. This was the farming system adopted from the mid-1960s, but with the defined grazing area (paddocks) grazed in rotational order and with cows only returning to the same area when more than two leaves had appeared on more than 66% of perennial ryegrass tillers. Leaf appearance rate was not monitored.

Generally, farmlets were established by balancing the paddocks for geographic location, soil type, distance from the dairy shed, and previous experimental treatments, such that farmlets were evenly spread over No. 2 Dairy in a checker-board fashion. Each farmlet was then randomly allocated to one of the treatments. Once farmlets were established they were unchanged throughout the experiment.

Cows were milked twice a day until milk production declined during late lactation, when they were often milked once a day, with drying off being determined by level of production. All cows were milked through the same dairy and farmlets were generally being milked in the same order at each milking. The initial dairy was a walk-though, with 8 sets of cups. In the early 1960s, a Herringbone with 15 sets of cups was built and it was extended to 20 sets in the late 1990s. In the 1950s and 1960s, the milking involved a standard preparation of the cow’s udder to stimulate milk let-down (described by Whittlestone, 1949). Phillips (1978) reported that there was a reduced need for udder stimulation; so, udder preparation was changed to a brief wash to clean the teats and udder. Then, in the 1980s, udder preparation ceased, except when excessively soiled teats were cleaned prior to cup attachment.

Milk volume and milk fat composition of all cows was measured by weekly herd test, with protein and lactose also measured from the mid-1980s. Milk volume was measured on a pm and am basis for one day each week;, initially, the milk was weighed in a bucket, but when the herringbone dairy was built, in-line milk meters were installed.

Cows were weighed in the morning every second wk and from the early 1970s, BCS was assessed. During periods of feed shortage in the summer/autumn, cows were weighed and BCS assessed weekly. At calving, cows were weighed and BCS assessed as soon as practical after parturition.

In experiments 2 & 3 (Table 1), 25% to 40% of the farmlet area were conserved annually for silage or hay. The aim being to provide as much winter supplementary feed as possible. Following the research results identifying that conserving too much feed for supplements could reduce production in the spring (Campbell and Clayton 1966), dedicated conservation areas were abondoned and only pasture that was surplus to requirements was conserved.

In the early 1970s, from 3 wk prior to PSC until calving was completed in each farmlet, the pastures grazed by the pregnant cows were dusted with magnesium oxide (70 g/cow per d) for prevention of hypomagnesaemia. After calving, the cows were orally drenched with 20 g of Mg supplement/d until late November (generally, once daily, but, occasionally, twice daily) in the form of magnesium chloride (MgCI₂6H₂0). During periods of bloat in the spring, an anti-bloating compound (Bloatenz 2 in 1, Ecolab, Hamilton, NZ) was given to them (orally), either as a single drench or added to the magnesium chloride solution.

From the late 1960s, in summer and autumn – periods of increased vulnerability to facial eczema, as determined by pasture fungal (Pithomyces chartarum) spore counts – the pasture was sprayed with Benlate (140 g/ha active ingredient; C₁₄H₁₈N₄O₃) to control the spores, and from the late-1970s, the cows were orally supplemented with zinc sulphate (8 g of ZnSO₄7H₂O/kg of Lwt).