The existence of genotype×environment interactions means that the extremes of dairy production systems will benefit from individualised system-level breeding objectives. For systems to remain robust to economic and consumer-led drivers, future grazing systems will require cows that have a lower environmental footprint and thrive in an environment that has periodic interruptions to feed supply, while being ‘easy care’ and efficient converters of foraged feed into milk.
Robustness, in any system, reflects the ability of the system to absorb and withstand change and rebound from challenges to achieve the same or greater performance (Roche and Horan, Reference Roche and Horan2013). This is true for businesses, animals and the people who manage them. Historically, the aim of agricultural production systems was to ensure adequate amounts of high quality foodstuffs for people. However, with the exponential increase in food production since the 1960s and the associated improvement in the nutrition of most of the world’s population, the focus has shifted from a need for food, to a desire for a more multi-faceted approach to food production. Future global food production systems, therefore, face a four-fold challenge:
i) meet the rapidly changing demand for food within a resource-constrained environment (e.g. including as a result of climate change);
ii) do so in an environmentally and socially sustainable manner for both consumers and producers;
iii) ensure that the products meet the highest standards of safety and nutritional value for increasingly discerning consumers; and
iv) ensure sufficient financial return to producers.
Producing more food from the same land area, while reducing environmental impacts and addressing animal welfare concerns, will require what has been referred to as the ‘sustainable intensification’ (Pretty, Reference Pretty1997; Pretty and Bharucha, Reference Pretty and Bharucha2014) or the ‘ecological intensification’ (Bommarco et al., Reference Bommarco, Kleijn and Potts2013; Hochman et al., Reference Hochman, Carberry, Robertson, Gaydon, Bell and McIntosh2013; Tittonell, Reference Tittonell2014) of agricultural production. The challenge, therefore, is to improve the efficiency with which natural resources are used, so that the amount of food produced from existing farmland can be increased while, at the same time, reducing environmental degradation, maintaining or improving the animal’s quality of life and providing a sustainable economic return to the farmer. Although sustainable intensification in all livestock-based production systems will require innovative blueprints for production, in grazing production systems, it will fundamentally depend on increased pasture production and quality and improved pasture utilisation. This has implications for feed allowance per animal and the associated environmental footprint. In the following passages, we will provide insight into the peculiarities of intensive grazing systems that expect a large proportion of the cow’s diet to be self-harvested pasture, how genetics influences the cow’s ability to function and survive in these systems and how genetic selection priorities have evolved to accommodate genotype×environment (G×E) interactions, what a robust grazing systems looks like and how it affects farm profitability and environmental sustainability, and, finally, the most important traits for cow robustness.
Peculiarities of a grazing system
Among dairy production systems, grazing systems are peculiar in that (1) they are designed to work within the immutable and immediate natural forces of climate (i.e. weather); (2) the majority of feed management is undertaken autonomously by the cow, with minimal use of mechanisation; and (3) the base forage in grazing systems is fresh and, therefore, perishable; if it is not used, it decays in the paddock and is wasted. As such, grazing systems differ from housed systems in that in the latter, the farmer must manage stocks of feed and apportion them relative to cow demand, whilst, in the former, the farmer is required to manage the flux between cow demand and an unpredictable supply, including supplementary feeds in the daily allocation when pasture growth is insufficient to meet cow demand, and conserving pasture as hay or silage in times of surplus growth. As a result, grazing systems tend to be situated primarily in the centre of the temperate climatic zones (between latitudes 35 and 50 ° North and South), where air and soil temperature and precipitation follow relatively repeatable within-year sinusoidal profiles (Roche et al., Reference Roche, Turner, Lee, Edmeades, Donaghy, Macdonald, Penno and Berry2009b and Reference Roche, Turner, Lee, Edmeades, Donaghy, Macdonald, Penno and Berry2009c). In these latitudes, the climate facilitates the production of relatively large volumes (12 to 20 t DM) of high quality (>70% DM digestibility) perennial and annual grazeable forages (i.e. pastures) that re-grow following defoliation (Roche et al., Reference Roche, Turner, Lee, Edmeades, Donaghy, Macdonald, Penno and Berry2009c).
The cost of producing 1 kg of milk or 1 kg of fat and protein has been identified as the key determinant of operating profitability in grazing production systems, with milk production/cow or per ha explaining less than half as much of the variability in operating profit/ha (Ramsbottom et al., Reference Ramsbottom, Horan, Berry and Roche2015). In an analysis of dairy production systems globally, Dillon et al. (Reference Dillon, Roche, Shalloo and Horan2005) reported a strong quadratic relationship between the quantity of grazed pasture in the diet and the cost of producing 1 kg of milk: operating expenses/kg milk declined with increased reliance on grazed pasture. This relationship was particularly steep in intensive grazing systems (i.e. more than 50% of the cow’s diet as grazed pasture). Therefore, to minimise expenditure, more than 80% of the pasture is harvested directly by the cow (Dillon et al., Reference Dillon, Roche, Shalloo and Horan2005), reducing the need for mechanisation and associated costs.
A further peculiarity of grazing systems to consider when comparing with housed systems where cows are fed mixed rations is the requirement to use forage quickly. In either system, if the level of concentrate fed is increased, the base forage content of the diet is reduced. In a housed system, however, the base forage is already preserved in a silo, bunker or bale and, if not used at that point in time, it can be used at a later date without a material loss in quality. In comparison, in a grazing system, the forage is fresh; it is perishable and cannot be stored long in situ for later use. Therefore, if a supplement is substituted for pasture, the feed is wasted and the associated costs expended without financial return. In fact, when pasture is not consumed, but left in the paddock, the digestibility and metabolisable energy of pasture subsequently presented to the cow is less (Stakelum and Dillon, Reference Stakelum and Dillon1991) and milk production from pasture declines. The recognition of a ‘use it or lose it’ principle for the base forage is important in farm system-level nutritional management in a pasture-based system (Roche, Reference Roche2017).
Grazing systems – the coordination of feed supply and demand
In milk or meat production from pasture, the dual aims of successful grazing systems are to:
i) maximise the utilisation of pasture grown/ha; and
ii) have the cow consume as much of her diet from grazed forages, with minimal use of labour, mechanisation and capital infrastructure for feeding.
With intensive grazing systems being traditionally focussed in temperate climatic zones, these aims are complicated by the seasonal nature of pasture supply (i.e. zero to low pasture growth in winter, very high growth rates in spring, variable growth and feed quality during summer and moderate, but declining growth rates in autumn; Figure 1; Roche et al., Reference Roche, Turner, Lee, Edmeades, Donaghy, Macdonald, Penno and Berry2009b). Although not as extreme in shape, the temporal profile of cow dry matter intake (DMI) is similar to that of temperate pasture production. Taking advantage of this, and the fact that pasture can be stored for relatively long periods in situ during autumn and winter (up to 100 days), without much loss of feed quality, seasonal production systems have been developed to ensure lactating cows are well fed on:
∙ stored pasture, with or without concentrates, during early lactation;
∙ pasture alone during mid-lactation;
∙ pasture and pasture/cereal silage, with or without grains or co-product feeds (i.e. concentrates), during late lactation; and
∙ pasture, forage crops or pasture/cereal silage during the non-lactating period.
Simply, if there is sufficient pasture available, the cows are fed on pasture alone. If there is insufficient pasture to meet the cows’ nutrient requirements, silages, silages and concentrate feeds or root crops will be offered. If pasture growth exceeds herd requirements, surplus pasture is conserved as hay or silage in late spring and summer or ‘deferred’ for later grazing in the autumn (Figure 1).
To facilitate the synchrony between feed supply and demand, cows need to calve during a short 8- to 12-week window before peak pasture growth (i.e. seasonal calving system; Figure 2; Berry et al., Reference Berry, Kearney, Twomey and Evans2012). This ensures that peak herd demand coincides with peak pasture supply and that the period of minimum feed demand of the herd (i.e. when cows are not lactating) coincides with the period of lowest pasture production (i.e. during winter). However, it also means that the period of breeding cows is equally short, and requires cows to establish a successful pregnancy within ∼80 days of calving (Berry et al., Reference Berry, Kearney, Twomey and Evans2012; Roche et al., Reference Roche, Washburn, Berry, Donaghy and Horan2017b).
A further consideration in comparing grazing systems with housed, mixed-ration systems is that:
∙ cost of the base milk produced from forage in grazing systems is very low and marginal milk from imported feeds is more expensive (Ramsbottom et al., Reference Ramsbottom, Horan, Berry and Roche2015; Macdonald et al., Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017);
∙ cost of base milk produced from forage in mixed-ration systems is very high and marginal milk from concentrate feeds is much less expensive.
This difference has led to confusion about the appropriateness of imported (i.e. non-pasture feeds) feed in grazing systems. In housed, mixed-ration systems, where machinery and labour are already in place, additional milk from feed merely requires the replacement of an inferior feed ingredient with a superior feed ingredient, within the requirements of acceptable ration balancing and as long as the total ration cost fits within the strategic ‘milk to feed price ratio’ for the business. In fact, in this scenario, the additional milk helps to reduce the fixed costs, the majority of which have already been incurred.
The same logic has been expressed for grazing systems; however, the system-level ramifications are materially different. First, in grazing systems, imported feeds must be used in situations of genuine feed deficit; therefore, it is necessary to increase stocking rate (i.e. cows/ha) to achieve large responses to imported feeds (Macdonald et al., Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017). However, 50% to 60% of a grazing farm’s costs are associated with each cow (i.e. ‘fixed cost’; Macdonald et al., Reference Macdonald, Beca, Penno, Lancaster and Roche2011). Therefore, the use of imported feeds to increase milk production/ha increases fixed costs/ha and the marginal cost of this milk is often greater than the cost of base milk and, sometimes, greater than the milk price (Ramsbottom et al., Reference Ramsbottom, Horan, Berry and Roche2015; Macdonald et al., Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017). Macdonald et al. (Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017) reported that filling a feed deficit with imported feed or introducing high yielding forage crops, such as maize, would result in a reduction in profitability/ha. In fact, Macdonald et al. (Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017) estimated that it was prudent only to fill 90% of the feed deficit, except in situations of very high milk prices or very low feed prices. On average, for every €1 spent on feed in grazing systems, total costs increase by €1.60 to €2.00 (DairyCo, 2013; Ramsbottom et al., Reference Ramsbottom, Horan, Berry and Roche2015; New Zealand DairyBase, unpublished data; Figure 3).
Although grazing systems can vary from the archetypical grazing system portrayed in Figures 1 and 2 (Roche et al., Reference Roche, Washburn, Berry, Donaghy and Horan2017b) due to climatic challenges (i.e. less conducive to longer pasture growing seasons), variations in milk price and the price and availability of non-pasture feeds, and differences in farmer goals and aspirations, the system-level challenges and, in particular, the challenges for the dairy cow are similar across grazing systems, although, admittedly, the scale of the challenge can vary from farm to farm and location to location.
What is a robust grazing dairy system?
While it has always been difficult to predict international commodity prices or foresee production risks, the reduction in product stores and government intervention policies in many countries and increasing wealth in countries previously regarded as ‘developing’ has led to price volatility, arguably, not witnessed previously (Roche and Horan, Reference Roche and Horan2013). Future milk production will, therefore, be set against a background of increased farm business uncertainty. Furthermore, the lack of certainty around mean and extreme temperatures, rainfall and the risk of severe weather events with future climate change scenarios could also have considerable impact on future dairy systems. As a result, future farm systems must be sufficiently robust to respond positively and rapidly to change.
Robust businesses are designed to utilise their competitive and comparative advantages. For grazing systems, this involves:
∙ maximising the utilisation of grazed pasture, an inexpensive, highly balanced feed that was, largely, paid for in the purchase or lease cost of the land (i.e. competitive advantage);
∙ ensuring that the management systems equate to good custodians of the environment and the animals under our care, thereby furthering the consumer’s desire for products derived from grazing systems (i.e. comparative advantage).
For dairy systems, Roche and Horan (Reference Roche and Horan2013) went further, defining robust dairy systems as those that (1) provide a reasonable rate of return on equity invested; (2) are environmentally sustainable and animal welfare compliant; (3) allow for an enjoyable and rewarding lifestyle; and (4) allow opportunities for training and personal development.
In summary, robust grazing dairy systems need to be profitable and biologically, socially and environmentally acceptable. The cow has a significant influence on these parameters, affecting business profitability, the farm’s environmental footprint, whether her own quality of life is acceptable in the defined system, and whether she is easy to manage (i.e. ‘easy care’) or requires significant time investment.
Profitability and biological robustness
For a farming business to be robust, it must be consistently profitable, even in the face of challenges. Profitability is defined as the difference between revenue and costs (i.e. operating profit) relative to the value of the asset (i.e. return on asset) and relative to the ability of the business to pay debt (i.e. solvency) or reward investment (i.e. return on equity invested). Although all are unique business metrics, they are, ultimately, a reflection of operating profit/ha, which is explained, in large part, by the unit cost of production (c/kg of milk or c/kg fat and protein; Ramsbottom et al., Reference Ramsbottom, Horan, Berry and Roche2015).
In grazing systems, operating expenses/kg of milk or per kilogram of milk fat and protein produced are strongly related to pasture utilisation/ha (Ramsbottom et al., Reference Ramsbottom, Horan, Berry and Roche2015) and the proportion of the cow’s diet that is grazed pasture (Dillon et al., Reference Dillon, Roche, Shalloo and Horan2005). Therefore, robust grazing dairy farm systems are designed to utilise a large proportion of the pasture grown/ha, but also require that pasture constitutes a high proportion of the cow’s diet. These suggestions for robustness were supported by the recent analysis of farm systems experiments by Macdonald et al. (Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017); they reported a decline in profitability when non-pasture feeds constituting >30% of the cow’s total diet were used either to increase the stocking rate (i.e. expansion) or fill feed deficits in an already high stocking rate system (i.e. intensification).
The archetypical grazing system was associated with self-contained farms that nourished cows with fresh pasture when growth was plentiful and with conserved pasture (i.e. silage or hay) when demand exceeded growth rate. These systems were common in temperate climatic zones until the 1970s and typical for New Zealand until the 2000s. However, longer periods of low rainfall in some temperate zones (e.g. Australia), larger cows, increased stocking rates, improved genetic merit for milk production, availability of seemingly inexpensive grains and co-product feeds and social awareness around animal welfare (real and perceived) have all led to the increased importation of feeds into grazing systems as a way of improving ‘biological robustness’. For example, genetic improvements for milk production resulted in cows mobilising greater amounts of body condition (BC) in early lactation and a resistance to replenishing BC stores while lactating (Roche et al., Reference Roche, Berry and Kolver2006a; Macdonald et al., Reference Macdonald, Verkerk, Thorrold, Pryce, Penno, McNaughton, Burton, Lancaster, Williamson and Holmes2008b). Although imported feeds did little to arrest BC loss in early lactation (Roche et al., Reference Roche, Berry and Kolver2006a, Reference Roche, Friggens, Kay, Fisher, Stafford and Berry2009a; Roche, Reference Roche2007), they facilitated BC replenishment during late lactation and during the non-lactating period. The importation of feed or the growing of high yielding crops to be harvested and offered to cows as a supplementary feed were, therefore, viewed as a component of robustness and biological sustainability (i.e. ensuring cows were returned to the same BC before the next calving event). However, they were also used to increase stocking rate, increasing, further, the need for imported feeds (Macdonald et al., Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017) and increasing the farm’s environmental footprint (Huebsch et al., Reference Huebsch, Horan, Blum, Richards, Grant and Fenton2013; Roche et al., Reference Roche, Ledgard, Sprosen, Lindsey, Penno, Horan and Macdonald2016a).
Although component- and system-level studies have resulted in a greater understanding of marginal milk production responses to supplementary feeds, the literature does not facilitate the investigation of system-level and profitability-related effects of increasing levels of supplementation from zero input. Therefore, it is not possible to determine an optimum level of supplementation to accommodate the seemingly conflicting economic and biological requirements of robustness. However, Ramsbottom et al. (Reference Ramsbottom, Horan, Berry and Roche2015) reported a linear decline in profitability when feed importation exceeded 500 kg DM/cow. This amount is approximately equivalent to 3.5 kg DM feed for 150 days. A greater feed allowance/cow would require an increase in stocking rate to ensure pasture utilisation is maintained (Roche et al., Reference Roche, Washburn, Berry, Donaghy and Horan2017b), with an increase in fixed costs/ha associated with the increase in cows milked (Macdonald et al., Reference Macdonald, Beca, Penno, Lancaster and Roche2011 and Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017), and a likely increase in the farm’s environmental footprint (Roche et al., Reference Roche, Ledgard, Sprosen, Lindsey, Penno, Horan and Macdonald2016b). Macdonald et al. (Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017) questioned the economic viability of this strategy, with the cost of the marginal milk produced considerably greater than the milk price except when milk price was exceptionally high or feed prices were exceptionally low. In support, Ramsbottom et al. (Reference Ramsbottom, Horan, Berry and Roche2015) reported that the cost of the marginal milk produced from imported feeds was greater than the milk price received in systems utilising >500 kg DM/cow purchased feeds. Recent results from farm systems experiments in Ireland (Ruelle et al., Reference Ruelle, Delaby, Wallace and Shalloo2018) and New Zealand Roche J. R., unpublished data) concur; the cost of marginal milk produced from inexpensive purchased feeds was greater than a moderate to good milk price, despite very good marginal milk production responses to the purchased feed.
From a cow perspective, there is one further complication. If supplements are to be used profitably, they cannot be used to fill the feed deficit completely. In an assessment of farm system intensification, Macdonald et al. (Reference Macdonald, Penno, Lancaster, Bryant, Kidd and Roche2017) concluded that to improve profitability, farms using purchased supplements needed to achieve high marginal milk production responses and, therefore, feed allowance/cow per year (i.e. pasture plus supplement) needed to be ~10% less than on farms not importing feed. This conclusion explains some of the discrepancy on optimal stocking rate from research undertaken in similar grazing systems in New Zealand and Ireland (Macdonald et al., Reference Macdonald, Penno, Lancaster and Roche2008a; McCarthy et al., Reference McCarthy, Delaby, Pierce, Journot and Horan2010); optimum stocking rate in Ireland, where feed was imported for early and late lactation cows, was defined as ~10% greater than in New Zealand, where no feed was imported (Macdonald et al., Reference Macdonald, Beca, Penno, Lancaster and Roche2011; Roche and Horan, Reference Roche and Horan2013). This has implications for cow robustness; it means that grazing systems, irrespective of availability of purchased feed, require a cow robust to short-term feed shortages.
In summary, profitability in grazing systems is strongly associated with pasture utilisation/ha and low unit costs of production; low unit cost of production is strongly associated with the proportion of the cow’s diet that is grazed pasture. Imported feeds support biological robustness, allowing BC gain in late lactation and facilitating the management of feed shortages. However, they can undermine profitability, at least in systems designed to require >500 kg DM of imported feed/cow per year. Profitable use of supplementary feeds requires that the feed allowance/cow per year is less than in grazing systems not importing feeds (i.e. stocking rate must be higher). Therefore, an essential characteristic of the most suitable cow for a grazing system is robustness to periodic, small, feed deficits.
In addition to the need for economic and biophysical robustness, a relatively recent requirement is the need to reduce the farm’s negative impact on the environment (i.e. the farm’s environmental footprint). The historical and cynical adage that ‘you have to be in the black to be green’, insinuating that you cannot be environmentally responsible unless you are profitable, has been replaced by public voice that farmers will not be allowed to operate unless they can prove that their environmental footprint is sustainable. From a grazing system perspective, the greatest risks to environmental sustainability and, therefore, business robustness are the contribution of nitrate (NO3 −) to ground water degradation and methane (CH4) and nitrous oxide (N2O) to climate change. In environmental impact, CH4 can be considered as a net loss of energy, whereas NO3 − is, at farm level, manageable and, in some instances, recyclable, and can, actually, contribute to reducing inputs in the form of mineral fertilisers (Huebsch et al., Reference Huebsch, Horan, Blum, Richards, Grant and Fenton2013).
Methane is a natural consequence of the ruminant’s ability to digest fibre and corresponds to 5% to 10% of gross energy intake (Van Soest, Reference Van Soest1982). The primary factor influencing CH4 yield is the cow’s intake of digestible organic matter (Wims et al., Reference Wims, Deighton, Lewis, O’Loughlin, Delaby, Boland and O’Donovan2010; Hammond, Reference Hammond2011), although the composition of the pasture must also be considered (Wims et al., Reference Wims, Deighton, Lewis, O’Loughlin, Delaby, Boland and O’Donovan2010; Enriquez-Hidalgo et al., Reference Enriquez-Hidalgo, Gilliland, Deighton, O’Donovan and Hennessy2014; Williams et al., Reference Williams, Moate, Deighton, Hannah, Wales and Jacobs2016). Nevertheless, there is considerable variation in, and reasonable heritability for, CH4 production (g/day; heritability of 0.26) and CH4 intensity (g/l milk; heritability of 0.21); Pinares-Patiño et al., Reference Pinares-Patiño, Hickey, Young, Dodds, MacLean, Molano, Sandoval, Kjestrup, Harland, Hunt, Pickering and McEwan2013; Lassen and Løvendahl, Reference Lassen and Løvendahl2016), supporting a possible genetic solution to reduce CH4 emissions. There is no evidence that there are differences in the expression of the trait in different environments (Robertson and Waghorn, Reference Robertson and Waghorn2002; Van Der Nagel et al., Reference Van Der Nagel, Waghorn and Forgie2003). However, there are possible differences in the importance of the breeding goal because of differences in the contribution of livestock emissions to individual countries’ national inventory. This G×E must be considered in future breeding objectives for the robust cow.
The nitrogen footprint of grazing systems is more complex. Ruminant production systems are particularly inefficient in their conversion of ingested nitrogen to exported product (Huebsch et al., Reference Huebsch, Horan, Blum, Richards, Grant and Fenton2013). A key contributor to this is dietary intake of nitrogen, with urinary nitrogen yield increasing exponentially with nitrogen intake above 400 g/day (i.e. CP of ~17% DM) and ~80% of nitrogen consumed above this threshold lost in urine (Kebreab et al., Reference Kebreab, France, Beever and Castillo2001). As fresh pasture tends to be very high in nitrogen throughout the year (Roche et al., Reference Roche, Turner, Lee, Edmeades, Donaghy, Macdonald, Penno and Berry2009c), the potentially leachable nitrogen load in each urine patch is very high (Selbie et al., Reference Selbie, Buckthought and Shepherd2015; Roche et al., Reference Roche, Ledgard, Sprosen, Lindsey, Penno, Horan and Macdonald2016a). It is further complicated by soil type (Hack-ten Broeke et al., Reference Hack-ten Broeke, De Groot and Dijkstra1996) and the timing of urinary N excretion (Selbie et al., Reference Selbie, Buckthought and Shepherd2015; Roche et al., Reference Roche, Ledgard, Sprosen, Lindsey, Penno, Horan and Macdonald2016a), with free-draining soils and urine deposited in autumn, both risk factors for greater nitrogen loss to ground water.
Genetic selection options to reduce urinary nitrogen concentrations have been proposed as a possible strategy for building robustness into grazing systems from an environmental sustainability perspective, but identifying a measurable and informative phenotype is key to success. The use of milk urea nitrogen (MUN) has been proposed as a potential phenotype to select animals for low urinary nitrogen concentrations. There is a positive correlation between MUN and urinary nitrogen yield (Jonker et al., Reference Jonker, Kohn and Erdman1998; Kauffman and St-Pierre, Reference Kauffman and St-Pierre2001; Nousiainen et al., Reference Nousiainen, Shingfield and Huhtanen2004) in housed cows fed mixed rations, which is consistent with our data from grazing dairy cows (Roche J. R., unpublished data).
It appears, therefore, that MUN could, potentially, be used as a phenotype to select bulls for lower urinary nitrogen excretion. From a genetics perspective, Miglior et al. (Reference Miglior, Sewalem, Jamrozik, Bohmanova, Lefebvre and Moore2007) reported that average daily heritability for MUN was moderately high (from 0.38 to 0.41), providing confidence of a genetic-level solution to the nitrogen footprint of grazing systems. In New Zealand, modelling indicates that reducing the national herd’s average breeding value for MUN by 2.4 units could, if successful, reduce nitrogen leaching by 17%, assuming the phenotypic MUN to urinary nitrogen relationship represents a genetic correlation (Mark Camara, personal communication).
From a G×E perspective, there is no evidence that the expression of the genotype is different in different environments. However, where nitrogen-loss limits are imposed (i.e. so called, NO3 − sensitive zones), profitability will be defined as margin/kg NO3 − leached (Roche et al., Reference Roche, Ledgard, Sprosen, Lindsey, Penno, Horan and Macdonald2016a) rather than per cow or per ha. Therefore, breeding objectives for these environments require emphasis on this trait (or a correlated trait) to ensure that the breeding goal remains holistic.
In summary, although the expression of the genotype for environmental measurements does not appear to change with environment, the importance of environmental traits will vary with regional and national priorities. Therefore, there is likely to be differences in the definition of the breeding goal in different environments and, as such, the G×E interaction needs to be accounted for in the ‘environment’ breeding objective.
There is little evidence of a large G×E for individual cow traits; but, there is clear evidence that differences exist in the importance of key animal-level traits in different farming systems (e.g. fertility in grazing systems, the economics of imported supplements, the importance of cow urinary nitrogen output or CH4 yield). Differences in the definition of the breeding goal for different environments, therefore, constitute a G×E interaction that has consequences for the type of cow required for different systems to ensure they are robust to the periodic challenges of milk price and the environment in which they operate.
Genotype×environment interactions – do we need to choose different cows for different systems?
Genotype by environment interactions are defined as ‘the phenomenon that performances of different genotypes are not equally affected by different environments’ (Falconer, Reference Falconer1952). Two types of G×E interactions exist: (1) re-scaling, which is a change in the relative magnitude of the difference between genotypes (e.g. individual animals, strains and breeds) in different environments; and (2) re-ranking, where the ranking of genotypes differs by environment. The greater the genetic diversity between genotypes and/or the larger the differences between environments, the greater the likelihood of G×E interactions existing (Falconer, Reference Falconer1989). The phenotype of interest also impacts the extent of genetic variability; of particular note, in the absence of re-ranking for any trait in a breeding objective, re-scaling in the breed goal traits can result in a re-ranking for the index (Namkoong, Reference Namkoong1985).
Quantification of the extent of G×E interactions is important, since it affects genetic gain. If only a scaling effect exists, then only the magnitude of genetic gain differs from expectation: the same genotypes rank highest irrespective of environment. If, however, re-ranking occurs and is not considered, genetic gain will be reduced and could be unfavourable if the genetic correlation between the traits in the two environments is negative.
Genotype×environment interactions can be due to:
1) true differences in the expression of a genotype in different environments;
2) differences in the definition of the trait or the statistical model(s) applied in the environments compared; or
3) differences due to the relative emphasis on traits within an overall breeding objective.
Differences in the expression of a genotype in different environments
Genotype×environment interactions in dairy cows in grazing systems have been the subject of much debate and experimentation for more than 50 years (Roche et al., Reference Roche, Berry, Bryant, Burke, Butler, Dillon, Donaghy, Horan, Macdonald and Macmillan2017a). For example, McMeekan (Reference McMeekan1960) reported that New Zealand farmers believed that the stocking rate research undertaken at Ruakura in the 1950s was not applicable commercially because the cows used in the research were genetically superior to the average commercial cow, since they had been bred using artificial insemination with genetically superior sires. This criticism was subsequently proven to have little validity, with the higher genetic merit cows proving their superiority, irrespective of the stocking rates implemented (Carter, Reference Carter1964). The difference between the systems compared, however, was not great, reflecting only a difference in feed allowance/cow (i.e. small divergence in environment); so, the lack of an interaction is not surprising.
Until the 1990s, the accepted paradigm, globally, was that environment did not re-scale the relative expression of dairy cow genetic merit, at least not for production-related traits. However, production systems evolved to be increasingly diverse during the 1970s and 1980s (Roche et al., Reference Roche, Berry, Bryant, Burke, Butler, Dillon, Donaghy, Horan, Macdonald and Macmillan2017a), with increased availability of grains and co-products from the human food, textile and cosmetic industries. In many systems, cows were offered more than grazed and conserved pasture to increase milk production/cow (Stockdale, Reference Stockdale2000; Bargo et al., Reference Bargo, Muller, Kolver and Delahoy2003). This increased the likelihood of G×E interactions and, especially, if germplasm was transferred between countries. Nevertheless, several experiments indicated no G×E interaction for milk production. For example, the Canadian–New Zealand G×E study (CANZ) used 20 Canadian and 20 New Zealand Holstein-Friesian bulls as AI sires for cows in 20 New Zealand and 10 Canadian herds in a 2×2 factorial arrangement (Peterson, Reference Peterson1988). Holmes (Reference Holmes1995) reported that there was no interaction between the origin of the sires and the daughter’s environment. There was, however, a significant interaction between individual sires within strain and the daughters’ environment; as a result, the correlations between the sire estimated breeding values in the two countries were approximately half the expected values. Nevertheless, there was no evidence that the correlation was negative. These results were supported subsequently by controlled experiments in research herds in Ireland (Kennedy et al., Reference Kennedy, Dillon, Faverdin, Delaby, Buckley and Rath2002; Horan et al., Reference Horan, Dillon, Faverdin, Delaby, Buckley and Rath2005b) and New Zealand (Roche et al., Reference Roche, Berry and Kolver2006a; Macdonald et al., Reference Macdonald, Verkerk, Thorrold, Pryce, Penno, McNaughton, Burton, Lancaster, Williamson and Holmes2008b), wherein intensity of genetic selection for milk yield influenced the size of the milk production response to purchased feeds. In all of these studies, cows of North American origin genetically selected almost exclusively for milk production in a non-grazing system produced significantly more milk/kg concentrate offered than cows of New Zealand origin selected for a balance of production and functional traits within grazing production systems. Results from an experimental study in Australia (Fulkerson et al., Reference Fulkerson, Davison, Garcia, Hough, Goddard, Dobos and Blockey2008) supported these findings. Fulkerson et al. (Reference Fulkerson, Davison, Garcia, Hough, Goddard, Dobos and Blockey2008) reported that cows did not express their genetic potential for milk production unless they were offered ⩾800 kg DM of concentrates. Below this level of supplementation, the association between cow genetic merit and milk production was less than predicted; however, above this level of supplementation, the cows did not exceed their predicted difference for milk production. These latest experiments indicated the existence of a G×E interaction, but that the effect on milk production was predominantly scaling, at least at the macro-level, although some re-ranking of individual sires did occur. Therefore, if only milk production is considered, differences in the expression of genotypes did exist, but the genotypes did not substantially re-rank between the production systems evaluated.
Breed comparisons during this period also supported the existence of a G×E interaction, but it was limited to a re-scaling effect; breeds selected in an environment with more feed available produced more milk, subsequently, when in an environment with greater access to feed. For example, the greater milk production drive of the Holstein-Friesian and the associated greater ‘hunger’ (Roche et al., Reference Roche, Sheahan, Chagas and Berry2006b) resulted in lower post-grazing residuals (L’Huillier et al., Reference L’Huillier, Parr and Bryant1988) compared with Jersey cows. This effect reduced pasture production/ha in a high stocking rate environment through over-grazing during summer and autumn and this negatively affected the profitability/ha of the Holstein-Friesian, when compared with Jersey cows under the same conditions (Spaans et al., Reference Spaans, Macdonald, Lancaster, Bryant and Roche2018). However, under a lower stocking rate, when feed allowance per cow was greater, those same breed attributes resulted in greater milk production/cow in the Holstein-Friesian and an increase in profitability/ha relative to the Jersey.
In summary, comparing both across and within breeds, there is considerable evidence of a G×E interaction. Genotypes selected within dairy systems that have greater access to imported feed produce more milk when offered additional feed or when compared with genotypes selected from systems that import less feed. However, when ranking is based on milk production, this G×E interaction is a scaling effect, with little evidence of large-scale genetic re-ranking.
Differences in the definition of the trait or the statistical model applied in the environments compared
Genotype x environment interactions in the form of re-scaling or re-ranking, as well as being a result of biological differences in the expression of the variome (i.e. the whole set of genetic variations in a species), through phenotypic differences in contrasting production systems, can also be, simply, an artefact of differences in the definition of the trait or the underlying statistical modelling or data edits/collection mechanism in the contrasting environments (Berry, Reference Berry2015 and Reference Berry2018). Although one may believe that testing for the presence of G x E in a well-defined trait, like lactation protein yield, for example, is unlikely to suffer from “statistical G x E”, this may not, necessarily, be true. For example, most countries evaluate 305-day equivalent milk yield; New Zealand, however, bases its dairy cow genetic evaluations on 270-day yield, and, thus, probably contribute a re-scaling G x E when compared with other countries. Moreover, many countries (e.g. the United States) present their 305-day yield genetic evaluation on the scale of a mature equivalent, whereas others (e.g. Ireland) present the data on the scale of a first lactation animal, with others (e.g. the Netherlands) using a hybrid of the two. Furthermore, many countries now use test-day models in their genetic evaluation of milk production that facilitate the modelling of different lactation profiles for different animals; some countries, however, (e.g. the United States, Ireland) use repeatability models. Hence, even for something as ubiquitous in genetic selection indices as protein yield, which appears to be a well-defined trait, G x E can be detected, and some of this is attributable, simply, to the underlying mathematics of how the trait is defined and modelled.
The phenomenon is generally exacerbated for more complex traits such as fertility. Different countries use a plethora of different definitions and statistical models, or editing criteria, in defining reproductive performance. Ireland and New Zealand, both predominated by seasonal-calving, grazing production systems, use calving interval and calved in the first 42 days of the calving season, respectively, as the goal traits in their breeding objectives. Even within Ireland, alone, the genetic correlation between these two traits for the same animals in the same lactation is, on average, 0.45 (Berry et al., Reference Berry, Kearney, Twomey and Evans2012). Therefore, when interpreting any documented G x E for a given trait across contrasting production systems, and especially across countries, the likely contribution of the mathematical and statistical properties of the comparison must be considered.
Differences in the definition of the breeding goal (i.e. cow type by production system interactions)
Because there was no large-scale re-ranking of genotypes across production systems when milk production was the primary trait of interest, Holstein-Friesian germplasm from farm systems that did not involve grazing or seasonal calving were increasingly used through the 1980s and 1990s because of their superior milk production ability (Harris and Kolver, Reference Harris and Kolver2001). In the 1990s, farmers increasingly reported concerns that the ‘modern’ Holstein-Friesian cow had inferior fertility performance when compared with her predecessor. Much of those protestations ‘fell on deaf ears’ because it was largely regarded that breeding for or against fertility traits was not possible due to its low heritability (Berry et al., Reference Berry, Wall and Pryce2014). Numerous experiments were undertaken in New Zealand (Kolver et al., Reference Kolver, Roche, de Veth, Thorne and Napper2002; Roche et al., Reference Roche, Berry and Kolver2006a; Macdonald et al., Reference Macdonald, Verkerk, Thorrold, Pryce, Penno, McNaughton, Burton, Lancaster, Williamson and Holmes2008b), Australia (Fulkerson et al., Reference Fulkerson, Wilkins, Dobos, Hough, Goddard and Davison2001 and Reference Fulkerson, Davison, Garcia, Hough, Goddard, Dobos and Blockey2008), France (Bedere et al., Reference Bedere, Delaby, Ducrocq, Leurent-Colette and Disenhaus2015, Reference Bedere, Delaby, Leurent-Colette and Disenhaus2016, Reference Bedere, Disenhaus, Ducrocq, Leurent-Colette and Delaby2017a and Reference Bedere, Disenhaus, Ducrocq, Leurent-Colette and Delaby2017b) and Ireland (Buckley et al., Reference Buckley, Dillon, Rath and Veerkamp2000; Kennedy et al., Reference Kennedy, Dillon, Faverdin, Delaby, Buckley and Rath2002; Horan et al., Reference Horan, Dillon, Berry, O’Connor and Rath2005a) to determine if different genetic strains performed differently under different feeding systems, in a holistic or overall economic sense. Although there was evidence that feeding regimen influenced the marginal milk production response to non-pasture feed and the ability of the dairy cow to exhibit her genetic potential, compared with her peers (Kennedy et al., Reference Kennedy, Dillon, Faverdin, Delaby, Buckley and Rath2002; Fulkerson et al., Reference Fulkerson, Davison, Garcia, Hough, Goddard, Dobos and Blockey2008), the greatest differences between the strains were in the extent of their negative energy balance in early lactation (Buckley et al., Reference Buckley, Dillon, Rath and Veerkamp2000; Roche et al., Reference Roche, Berry and Kolver2006a; McCarthy et al., Reference McCarthy, Berry, Dillon, Rath and Horan2007) and in their pregnancy rates in a seasonal calving system (Horan et al., Reference Horan, Mee, O’Connor, Rath and Dillon2005c; Macdonald et al., Reference Macdonald, Verkerk, Thorrold, Pryce, Penno, McNaughton, Burton, Lancaster, Williamson and Holmes2008b; Bedere et al., Reference Bedere, Disenhaus, Ducrocq, Leurent-Colette and Delaby2017a). In brief, the 6-week in-calf rate for North American-derived Holstein-Friesian cows was 15 percentage points lower than for New Zealand-derived Holstein-Friesian cows (Horan et al., Reference Horan, Mee, O’Connor, Rath and Dillon2005c; Macdonald et al., Reference Macdonald, Verkerk, Thorrold, Pryce, Penno, McNaughton, Burton, Lancaster, Williamson and Holmes2008b), and the North American-derived Holstein-Friesian cows were less likely to survive to successive lactations (Harris and Kolver, Reference Harris and Kolver2001). These differences were irrespective of feeding regime, thus indicating no G×E interaction.
As evidence accumulated in seasonal-calving grazing systems of the inferior reproductive performance and survival of Holstein-Friesian strains selected heavily for milk production, breeding objectives evolved rapidly. In 2001, the Economic Breeding Index (EBI; Berry et al., Reference Berry, Shalloo, Cromie, Veerkamp, Dillon, Amer, Kearney, Evans and Wickham2007) was introduced in Ireland and included the two fitness traits of calving interval and survival, in addition to the milk yield traits. Similarly, fertility was included in the New Zealand national breeding goal in 2001. By 2017, the EBI included 18 traits encompassed within seven sub-indices: (1) milk production; (2) fertility and survival; (3) calving performance; (4) maintenance; (5) beef performance; (6) health; and (7) management (Figure 4). From the analysis of 1053 high-reliability Holsteins sires, Berry (Reference Berry2018) reported a correlation between the Net Merit Index (from the United States) and the EBI (from Ireland) of 0.66, suggesting considerable re-ranking of sires on overall profit when comparing breeding objectives designed for high input confinement production systems against breeding objectives designed for grazing production systems. The effects of changing the Irish dairy cow breeding objective (i.e. EBI) on genetic trends for production of milk fat and protein and fertility are presented in Figure 5. Recognition of the success of this multi-trait approach in designing the ideal cow for seasonal calving and grazing systems has led to the development of the US Grazing Merit Index (GM$; Gay et al., Reference Gay, Widmar, Nennich, Schinckel, Cole and Schutz2014) and the UK Seasonal Calving Index in 2014, with both indices putting increased emphasis on functional traits associated with grazing and the need to achieve pregnancy within ~80 days of calving.
As only ~10% of the world’s milk production originates from grazing systems, most ruminant livestock globally have not been directly selected for grazing systems. Until recently, most experimental results indicated little or no importance of G×E interactions on a trait by trait basis (Holmes, Reference Holmes1995); however, adoption of more diverse genotypes and/or production environments has increased the likelihood of such interactions, especially at the level of overall profitability (Falconer, Reference Falconer1952 and Reference Falconer1989). Strong evidence now exists to show that the animals genetically best suited to non-grazing systems are not best suited to grazing systems, predominantly because of their compromised reproductive performance, which is independent of feeding system (Horan et al., Reference Horan, Mee, O’Connor, Rath and Dillon2005c; Macdonald et al., Reference Macdonald, Verkerk, Thorrold, Pryce, Penno, McNaughton, Burton, Lancaster, Williamson and Holmes2008b; Delaby et al., Reference Delaby, Horan, O’Donovan, Gallard and Peyraud2010; Roche et al., Reference Roche, Berry, Bryant, Burke, Butler, Dillon, Donaghy, Horan, Macdonald and Macmillan2017a).
Cows for courses: choosing the right cow for grazing systems?
Whether a separate breeding programme for grazing systems is required is a function of: (i) the intensity of selection for the breeding goal, (ii) the maximum genetic correlation between the breeding objective in a grazing system and all breeding objectives in other populations from where semen can be imported and (iii) the cost of implementing a separate breeding programme tailored for grazing systems.
The requirement for a separate breeding programme can be illustrated using an example of two dairy systems: a housed system, in which cows are offered a mixed ration, and an intensive grazing system, in which the primary source of feed is fresh pasture. For simplicity, the generation interval, genetic variance and accuracy of selection are assumed to be identical in both environments. The accuracy of an animal’s estimated breeding value in the grazing system, having been evaluated in the housed system, is the accuracy in the housed system multiplied by the genetic correlation between the systems. The number of sires that must be evaluated in the grazing system to achieve similar genetic progress to using sires evaluated in the housed production system, therefore, declines as the genetic correlation between the breeding objectives in both production systems weaken. For example, when the correlation between breeding objectives across the systems is 0.6, 0.7, 0.8 or 0.9, the number of sires that must be evaluated in the grazing system to achieve the same genetic progress as importing semen from the breeding programme with 100 sires evaluated under housed production systems is 28, 37, 51 and 71, respectively. This is because the greater the dissimilarity between breeding objectives (i.e. weaker genetic correlations), the less applicable the breeding objective for the housed production system is for selecting elite animals under grazing production; thus, the greater the difference between farming systems, the fewer sires that must be tested under grazing production systems to achieve the same gains. Therefore, the justification of a separate breeding programme is dependent on the extent of G×E with other breeding programmes and the actual size and structure (i.e. ability to achieve high accuracy of selection) of those breeding programmes.
Not considered here, however, is the long-term security of having a grazing-specific breeding programme and the risks associated with relying on an external breeding programme. Although genomic selection will reduce the benefit (and burden) of progeny testing in the future, large training population sizes used to estimate genetic marker effects for inclusion in genomic predictions are still required and the estimates of genetic merit in these reference animals should still be based on information generated from grazing production systems (and breeds).
Characteristics of a grazing cow
Systems based on greater use of ‘service offerings’ from nature (i.e. low supplementary feed input systems, organic systems) must cope with more variability in feed availability and quality and climate. Such systems, therefore, require animals that are robust, autonomous and ‘easy care’, and capable of high levels of performance from a predominantly grazed pasture diet. As the genetic correlation between breeding objectives in grazing systems and systems where cows are housed and offered mixed rations tend to be weak (i.e. correlation of 0.66 between the US and Irish dairy cow breeding objectives, as described earlier) and the overwhelming evidence from experiments undertaken over the last two decades is that cows selected from within housed systems are not appropriate for grazing systems, there is a good justification for selecting a ‘grazing-appropriate’ cow.
Several countries have focussed their breeding objectives on the cow characteristics important for profitability in a grazing system (e.g. New Zealand, Ireland, UK, United States and Australia). Grazing systems require a cow with good grazing characteristics and a propensity for high pasture DMI, that has good fertility and is ‘easy care’ and docile, that can walk long distances, that is resistant to some of the diseases of grazing systems (e.g. hypomagnesaemia, foamy bloat) and that is robust to fluctuations in feed quality and quantity (Berry, Reference Berry2015; Friggens et al., Reference Friggens, Blanc, Berry and Puillet2017; Delaby et al., Reference Delaby, Buckley, McHugh and Blanc2018). These traits are fundamental for grazing systems to be successful and robust from 1 year to the next, as:
∙ cow demand is set strategically (i.e. stocking rate) and is not easily changed annually, despite variability in pasture supply;
∙ the cow must be able to walk reasonable distances to and from their feed source and harvest large quantities quickly;
∙ the cow must establish and maintain pregnancy within ~80 days of calving to maintain the seasonal calving pattern and the alignment between supply and demand; and
∙ must not require an inordinate amount of individual attention (i.e. ‘easy care’), as, in general, there are two to three times more cows/person in an intensively managed grazing system compared with a housed system.
In addition to these functional traits, traits associated with the environmental footprint of the cow must also be considered. A future requirement of robust systems will be that they produce less CH4 and will have a lower nitrogen footprint (i.e. lower NO3 − leaching and N2O emissions). Further research efforts need to investigate a genetic solution for these problems as well as the management mitigation strategies being explored.
Robust grazing cows: adapting to changes in feed supply and quality
Economically robust grazing systems are defined by the high levels of pasture utilisation and a reduced exposure to external price fluctuations (Ramsbottom et al., Reference Ramsbottom, Horan, Berry and Roche2015). Pasture utilisation is primarily determined by stocking rate (Macdonald et al., Reference Macdonald, Penno, Lancaster and Roche2008a; McCarthy et al., Reference McCarthy, Delaby, Pierce, McCarthy, Fleming, Brennan and Horan2016), with high stocking rates resulting in a high proportion of the pasture being directly harvested by the cow and low stocking rates resulting in either lower pasture utilisation or a greater proportion of pasture utilised via silage or hay. High stocking rates increase the likelihood of short-term feed deficits, particularly in the change between winter and spring and between autumn and winter. The cow for this system, therefore, must be able to absorb short-term feed deficits and be able to preserve BC, through reducing milk production, and respond to re-feeding when feed supply resumes. This could be achieved by selecting for cows that are able to transfer mammary secretory cells from an active state to a ‘quiescent’ state, from which they can return to an active state when feeding resumes.
Robust grazing cows: fertility
As robust grazing systems are dependent on high pasture utilisation, they are designed so that the herd’s demand for feed is matched as closely as possible with the pasture growth profile (Roche et al., Reference Roche, Washburn, Berry, Donaghy and Horan2017b). This requires all cows to calve over a short period (~10 weeks) and, by necessity, become pregnant in an equally short period and within ~80 days of calving. Achieving this requires the quick resolution of a range of physiological, endocrine and tissue re-modelling processes during a particularly nutrient demanding and both a physiologically and immunologically challenging period.
Robust grazing cows: walking, pasture harvesting and the efficiency of grazing
Grazing dairy cows are required to walk long distances daily and harvest their feed in a very competitive environment. In fact, cows in grazing systems will spend up to 10 h grazing each day (Linnane et al., Reference Linnane, Horan, Connolly, O’Connor, Buckley and Dillon2004; Sheahan et al., Reference Sheahan, Kolver and Roche2011), whereas those indoors on mixed rations rarely need to eat for more than 5 h (Thorne et al., Reference Thorne, Jago, Kolver and Roche2003). Genotypes have been reported to graze differently. For example, animals selected for high feed conversion efficiency spend less time grazing/kg DMI (Gregorini et al., Reference Gregorini, Waghorn, Kuhn-Sherlock, Romera and Macdonald2015) and animals selected exclusively for milk production exhibit less of a reduction in their time spent grazing when offered concentrate feeds than cows also selected for functional traits (i.e. G×E interaction; Linnane et al., Reference Linnane, Horan, Connolly, O’Connor, Buckley and Dillon2004). Similarly, Thorne et al. (Reference Thorne, Jago, Kolver and Roche2003) reported that some genotypes of cows selected exclusively for milk production were more aggressive grazers, probably indicative of a greater hunger state, a neuroendocrine basis for which was reported by Roche et al. (Reference Roche, Sheahan, Chagas and Berry2007). In summary, an ability to graze efficiently and walk long distances are necessary characteristics for efficient grazing dairy cows, which may not be as necessary for cows consuming mixed rations.
A G×E interaction is defined as the phenomenon that performances of different genotypes are not equally affected by different environments. Although there are instances in which the expression of a genotype differs in different environments, the greatest G×E interaction is due to different breeding objectives in different production systems and, as a result, the suitability of different cows for different production systems. Successful grazing systems require cows capable of achieving large intakes of forage so that they can achieve their nutritional requirements almost entirely from grazed pastures, with limited inputs of conserved forage and imported feed supplements. They must be able to produce reasonable quantities of milk (i.e. 5500 to 7500 kg; 350 to 500 kg milk fat and protein), primarily from grazed pasture, walk long distances and become pregnant within ~80 days of calving. Such autonomous grazing animals must also be able to maintain adequate BC to avoid ill health within a periodic restricted feed environment (Delaby et al., Reference Delaby, Faverdin, Michel, Disenhaus and Peyraud2009; Piccand et al., Reference Piccand, Cutullic, Schori, Weilenmann and Thomet2011). In addition, in the context of ‘sustainable intensification’, grazing cows must produce less CH4 and must excrete a lower concentration and amount of nitrogen in urine. Many of these requirements are different to the priorities in housed dairy systems and this G×E needs to be considered in breeding objectives as researchers identify traits for selection or suitably precise proxies, so that farmers can choose cows to suit their production system. Each country must evaluate the factors most influencing the robustness traits we’ve discussed, place a weighting on their importance to system or cow robustness and develop an appropriate breeding objective to ensure the cow of the future underpins a dairy system robust to climatic and financial challenges.
J. R. Roche acknowledges New Zealand dairy farmers, through DairyNZ Incorporated (RD1405), and the New Zealand Government, through the Ministry of Business, Innovation, and Employment (DRCX1302) for their financial contribution towards compiling this manuscript.
Declaration of interest
Software and data repository resources