Impact of hot weather on animal performance and genetic strategies to minimise the effect

Keywords: complex traits, environmental impact, genomics, genotype by environment interactions, heat tolerance, resilience, sensors, thermotolerance.

Climate change due to global warming is likely to lead to more weather variability (CSIRO 2016). Weather variability affects animals directly by affecting their performance (production and reproduction) and indirectly by affecting the environment, including prevalence of disease and parasites, availability of quality and quantity of feed. Dairy cattle are likely to be especially sensitive because of the elevated metabolic load associated with lactation and high feed demand.

The impact of weather on animal performance is almost entirely due to warm and hot weather conditions, even in temperate regions that experience cold winters and springs. This is especially true for pasture-based systems (typical of Australia and New Zealand) where there is more opportunity for cows to be directly affected by ambient temperature and solar radiation.

Many indices have also been developed to account for the thermal flow mechanisms (ambient temperature, relative humidity, solar radiation, wind speed, precipitation) on the animal and then used to define thresholds beyond which heat stress begins. Due to limited publicly available data on solar radiation and wind speed, most studies have focused on ambient temperature and relative humidity to quantify environmental heat load. The most frequently used environmental heat load index in dairy cattle is the temperature–humidity index (THI), originally developed to measure discomfort in humans (Thom 1959). THI is a single value that combines ambient temperature and relative humidity.

The thermo-neutral range for dairy animals is from 16°C to 25°C; at this temperature a body temperature of 38.4–39.1°C can be maintained (Yousef 1985). In Australia, the proportion of days in a year when dairy cows are expected to experience conditions outside of this range which lead to reductions in performance as a result heat stress varies between ∼300 days per year (Qld) to ∼60 days (Tas.; Nguyen et al. 2016). Variability in weather is expected to continue and the intensity, frequency and duration of heat events are predicted to increase (CSIRO 2016).

Normally animals are well adapted to their production environment. However, genetic selection for increased productivity, particularly in dairy cattle over the past five decades, appears to have had an unintended negative effect on the ability of cows to withstand heat stress (Nguyen et al. 2016). In more recent times, there has been a greater focus on traits other than production, such as survival, fertility, longevity and feed efficiency in the breeding objective (Byrne et al. 2016; Cole and VanRaden 2018). This trend is likely to continue and is necessary to ensure that cattle are suited to the production environments most common in dairying worldwide.

Heat tolerance is defined as the ability to maintain thermal stability at high temperatures and humidity (Carabaño et al. 2019); this process is largely controlled by an animal’s capacity to dissipate the heat generated by metabolic heat production. Heat stress decreases production, growth, sexual behaviour and fertility, with extreme heat stress sometimes resulting in death (Hansen 2020). In addition to reducing production losses, selecting for heat tolerance is likely to also have favourable effects on animal welfare, especially as heat stress is a growing welfare concern (Polsky and von Keyserlingk 2017).

One major concern for farmers is the reduction in productivity during hot weather. Dairy cows respond to heat stress through behavioural, physiological and cellular adaptations that are initiated to reduce the internal temperature of the animal when exposed to hot conditions (Dunshea et al. 2013). The main biological determinants of heat tolerance in animals include the relative body surface area (Berman 2003), sweating rate (determined by the morphology, density, and water transfer capacity of sweat glands), coat characteristics (length, thickness, colour, weight per unit surface, etc.; Collier et al. 2008) as well as the rate of metabolic heat production and dissipation (Kadzere et al. 2002). These adaptations to heat stress may be responsive not only to immediate nutritional or management modification, but also to long-term targeted genomic selection of animals for improved heat tolerance (Nguyen et al. 2016).

The use of sprinklers and other cooling devices and the provision of shade have been shown to provide relief from hot weather (Dunshea et al. 2013). While there are significant benefits to these mitigation options for cow welfare and productivity, many are not practical in pasture-based, or extensively managed environments. Temperatures are projected to continue to rise and management solutions may not on their own be able to prevent heat stress in high-producing cows, even where mitigating infrastructure (e.g. cooling equipment) exists. The use of genetic selection is expected to provide an effective long-term solution since the effects of genetic improvement are permanent and cumulative over generations.

In this review we discuss (1) exploiting genetic variation to improve resilience to weather variability (restricted to heat tolerance), (2) genotype by environment interactions, i.e. suitability of high genetic merit cows to heat stress and (3) how novel phenotyping and genomics can help improve resilience to heat stress.

One of the enabling technologies for improving novel traits, such as heat tolerance, has been the development of genomic selection, which has transformed livestock and plant breeding internationally (VanRaden 2020). Genomic selection uses a reference population that comprises genotyped individuals that also have phenotypes of interest (known as a genomic reference population) to develop genomic prediction equations. These predictions can then be applied to individuals that are genotyped but may not have phenotypic records. The dairy industry was an early adopter of genomic selection, with the first countries implementing it in 2009 (Hayes et al. 2009).

Australia introduced genomic breeding values for Holsteins and Jerseys in 2010, initially through male-only genomic reference populations. The inclusion of females to the Australian reference population was initiated in 2013, through ∼100 herds recruited on the basis of excellent data recording, to participate in the Genomic Information Nucleus (Ginfo) research project and has been shown to increase the reliability of genomic breeding values (Pryce et al. 2018). Ginfo is now approaching 200 herds (∼60 000 cows) and is now administered by DataGene Ltd (Bundoora, Australia), so that it is integrated into routine genetic evaluations of Australian dairy cattle. The reference population was designed to improve the reliability of traditional breeding values such as production, conformation, fertility, etc. (Pryce et al. 2018), but also provides a dataset for genomic breeding values of new traits, such as heat tolerance, mastitis resistance, metabolic resilience, etc. Ginfo is designed to be used to also provide data on novel measures of heat tolerance, such as mid-infrared spectral data (from routine milk recording) and this can also be expanded to sensor body temperature measurements, or even inferred behaviours indicative of heat stress. When these phenotypes are matched to genotypes on the same cow, they could provide rich future opportunities for genomic prediction.

There are several strategies that should be considered for genetically altering heat tolerance; these include (1) choice of breed or cross, as some breeds are more heat tolerant than others, (2) introgression of genes that are associated with thermotolerance, and (3) selection criteria associated with heat tolerance, as follows:

1. Choice of breed or use of crosses. In tropical climates, Bos indicus breeds are frequently crossed with temperate dairy breeds so as to combine the heat resilience and parasite resistance of Bos indicus with the production qualities of temperate breeds (Davis et al. 2017; Burrow et al. 2019). For example, the Girolando breed in Brazil is a cross between Gir (more heat tolerant) and Holstein (higher production). Temperate breeds also exhibit a greater reduction in heat tolerance; for example, New Zealand Holsteins appear to exhibit greater reductions in milk yield (less tolerant) in hotter climates than do Jerseys or crossbreds (Bryant et al. 2007a).

2. Introgression of useful locus that increase tolerance to heat. Introgression involves introducing known alleles of favourable effect, usually from a completely different breed. Traditionally, introgression involves crossing back to the original breed to restore the favourable dairy characteristics from the base breed, this process takes several generations of crossing and is therefore time-consuming and inefficient. A mutation that confers heat tolerance and a smooth coat phenotype is a deletion in an exon of the prolactin receptor in Senepol cattle (Littlejohn et al. 2014; http://omia.angis.org.au/OMIA001372/9913/). The so-called slick mutation appears to be associated with the type of coat, being slick or smooth, with some possible alteration to sweating ability (Davis et al. 2017). The slick variant is generally fixed in most breeds, meaning that it needs to be introgressed to benefit from this variant. Breeding programs in some countries, including Puerto Rico and the USA, have already started incorporating the slick mutation in Holstein cattle (Carabaño et al. 2019; Hansen 2020).

The slick mutation is also a target for gene-editing, which will enable more rapid introgression of the desirable mutation, with little to no background DNA from the donor breed. Other variants that could also be useful in introgression programs for heat tolerance have been identified, such as a deletion in the promoter regions of a heat shock protein gene (HSPA1) that confers cellular protection to high temperatures (Hansen 2020). Other genes associated with hair and skin colour (Collier et al. 2008) could also be candidate target genes.

Gene editing provides a faster alternative to introgression and is especially useful for precisely introducing single alleles and haplotypes from locally adapted breeds (Van Eenennaam 2019). Regardless of the technology used, the introgression of the slick locus from the Senepol breed, or other loci identified as conferring heat tolerance, could produce cattle that perform better in hot climates, without environmental modifications. This would be of great value to dairy farmers in many dairying regions around the world.

• (3)

• Selection within breeds. Although the slick variant confers an ability to regulate body temperature, heat tolerance is highly polygenic, meaning that a breeding value for heat tolerance (that includes many variants) could be an alternative, or even additional solution. The first step in developing a breeding value for any trait is to identify suitable selection criteria. To achieve improved heat tolerance, the selection criteria chosen could include one or several traits and where selection on these traits leads to a favourable outcome in the breeding objective. Methods to describe heat tolerance have been an active area of research for a considerable period. Measuring changes in core body temperature in addition to animal performance in relation to bioclimatic indices has commonly been used (Carabaño et al. 2019). Selection criteria for heat tolerance could include body temperature, respiration rate, heart rate, sweating rate, reduction in intake or milk yield and impacts on reproductive performance in relation to THI.

Rectal or vaginal temperature is a direct measure of body temperature regulation that is heritable (Dikmen et al. 2012), with evidence that cows selected to be heat tolerant have lower core body temperatures (Garner et al. 2016). A limitation of this approach is that it is often invasive and labour-intensive and, therefore, likely to be available only on a small number of selection candidates. Some devices such as vaginal probes that can measure intra-vaginal temperatures continuously over short periods can be useful for small-scale reference populations and especially in grazing systems (Garner et al. 2016; Koltes et al. 2018).

Changes in production and reproductive performance of cattle are associated with heat stress (West 2003; Das et al. 2016) and measurable on large numbers of cows, and, because of this, are likely to be the most convenient breeding targets (Carabaño et al. 2019). Since heat tolerance has an unfavourable genetic correlation with milk production traits (Aguilar et al. 2009; Nguyen et al. 2017), it is likely that selection objectives focused on high yielding dairy cows will continue to be detrimental to heat tolerance. However, it is important to note that a reduction in yield to rising ambient temperatures is partially a consequence of a voluntary reduction in feed intake. The proportion of milk yield reduction directly attributed to the decline in feed intake during heat stress is between 30% and 50% (Rhoads et al. 2009; Wheelock et al. 2010). Therefore, identifying the genetics that confers an ability to maintain a high level of production, while minimising losses due to heat stress, is important.

Ravagnolo et al. (2000) pioneered the use of dairy cow milk yield combined with temperature–humidity data from weather stations to measure variability in the rate of milk yield decline related to heat stress. Most dairy cow studies have used the so-called broken stick model to quantify the level of milk production traits under thermo-neutral conditions, which is the slope of decay (in milk production traits, for example) after passing a threshold associated with a transition to heat stress (Bernabucci et al. 2014). An alternative is to model using a reaction norm; this approach uses polynomials and, therefore, there is more flexibility than with the broken stick approach (Nguyen et al. 2016; Carabaño et al. 2019), such as, for example, if the slope gets steeper at higher THI values (non-linear relationship).

Nguyen et al. (2016) found the the heritability of heat tolerance traits (i.e. slopes of milk, fat and protein) was approximately 0.17–0.23. Using high-density single-nucleotide polymorphism (SNP) genotypes, heat tolerance modelled with a random regression models assuming the THI threshold of 60 (equivalent to a daily temperature of 20°C and 45% relative humidity) gives a genomic prediction accuracy of 0.42–0.61 (Nguyen et al. 2016). Heat tolerance breeding values were unfavourably correlated with milk production traits but were favourably associated with fertility (Nguyen et al. 2017), confirming the observations of Ravagnolo and Misztal (2000). Since these studies, additional animals have been added to the reference population (mainly genotyped females from Ginfo), enabling the release of the first genomic breeding values worldwide for dairy cow heat tolerance in 2017 (Nguyen et al. 2017). Reliabilities published by DataGene (the Australian national genetic evaluation unit for dairy) for bulls with no progeny are currently ∼40%. Nguyen et al. (2017) found that the ability of dairy cattle in Australia to tolerate heat has been declining over the past years at a rate of 0.3%/year, in part, due to continued selection for milk production traits. This mirrors more recent data for heat tolerance from DataGene, the organisation responsible for routine genetic evaluation in Australia, with a steeper genetic decline of heat tolerance following the introduction of genomic selection in 2010 and approximating a 1 s.d. (five units of heat tolerance ABV) reduction over two decades (Fig. 1).

Fig. 1.  Genetic trend of Australian average balanced performance index (BPI) and heat tolerance breeding values (ABVg) over the past years for Holstein cows (Date source; DataGene, https://datagene.com.au/; accessed September 2021).

Animals identified to be divergent for heat susceptibility using the genomic breeding values for Australia developed by Nguyen et al. (2016) had significant differences in milk yield losses (P < 0.023, at Day 4), rectal (P < 0.01, Days 2–4) and vaginal (P < 0.01, on all 4 days) temperatures under a 4-day climate-controlled challenge designed to emulate a mild heatwave in south eastern Australia (Garner et al. 2016). More recently, heat tolerance genomic breeding values (developed using the same Australian reference population) were validated in a Californian dairy herd. Here, Jensen et al.(2021) found that for groups that were approximately 1 s.d. apart in heat tolerance ABVs, heat tolerant cows had significantly lower rectal and vaginal temperatures than did heat-sensitive cows (P < 0.03 and P < 0.001 respectively).

Farmers operate very different farming systems; for example, while grazing dominates in Australia, some farmers have feeding systems that might be considered as more aligned to North American systems. Additionally, dairy farms in Australia are scattered over a wide range of climatic regions from Tasmania to Queensland. Therefore, producers are often interested in identifying sires that they believe will suit their production environment. Fortunately, this has been an area of extensive study for several decades, more specifically, quantifying the extent of genotype interactions with climatic and nutritional environments (Veerkamp et al. 1994; Ravagnolo and Misztal 2000; Hayes et al. 2003; Bryant et al. 2007b).

Generally, genetic variation in environmental sensitivity manifests itself as a scaling effect, where there are phenotypic differences in daughter performance on the basis of the environment they are in, i.e. greater levels of production for example in more favourable environments. Of greater importance from a breeding perspective is whether sires re-rank across environments. This can be investigated by looking at genetic correlations of the same trait evaluated in different environments. For example, while genetic correlations between countries often imply significant re-rankings between sires, most studies within countries have concluded that re-ranking is expected at only extreme environments (Hayes et al. 2003; Bryant et al. 2007b; Haile-Mariam et al. 2008; Cheruiyot et al. 2020).

While estimated breeding values (EBVs) for milk production traits already to some extent include heat stress effects, this is true only when all cows are evaluated under heat stress conditions. Quite often this is not the case and could become more of an issue in the genomic era where this information is unlikely to be captured in conventional production EBVs. Zwald et al. (2003) found that the heritabilities of production traits in colder climates in the USA were lower than in hot environments (0.26 and 0.39 respectively), and the genetic correlation between these two groups was 0.66. This implies that heat stress may play a role in genotype and environment interactions (G × E). However, Bohmanova et al. (2008) found that considering heat stress as part of EBVs estimated using northeastern and southeastern data increased the correlation between the two regions by 0.01 and that heat stress explains only a small amount of regional differences. One of the reasons for this is that test-day records capture only a snap-shot of heat stress. Bryant et al. (2006) found that Holstein-Friesian genotypes originating from North America are environmentally more sensitive than are New Zealand Friesian and Jersey genotypes to variation in the feeding level.

One option for sire selection decisions is selecting for robustness, i.e. similar EBVs across environments could be an option for ensuring stable performance (König and May 2019). For example, using Australian Holstein data, Cheruiyot et al. (2020) found variation in EBVs estimated as the slope of reaction norms for milk production traits against THI, indicating that there are some sires that could be considered to have ‘robust’ performance, i.e. whose daughters had stable levels of production regardless of THI. Robustness is often considered to be the capacity to perform in a wide range of current environments (Colditz and Hine 2016). Similarly, Bryant et al.(2007b) classified bulls into generalist and specialist bulls and most bulls that originate in the New Zealand production system were generalist and are capable of copying with changing production environment better than are specialist bulls. Improving the robustness of animals does not necessarily improve their resilience, as future environments may not be represented adequately by current environments.

There is a need to identify potential environmental stressors that dairy cattle may confront in the future. For example, Berry (2015) identified the following nine traits that are of likely future importance: (1) high production, (2) good reproductive performance, (3) good health status, (4) good longevity, (5) feed efficiency (6) easy to manage, (7) good conformation, (8) low environmental footprint, and (9) resilience to external disturbances. Generally, selection indices are used to appropriately weight multiple selection objectives and a single index value is presented for selection decisions. While indices have traditionally included traits that contribute to farmer profitability, there is a growing move towards including farmer preferences for selection pressure (Martin-Collado et al. 2015) and consumer preferences. This is a theme that is likely to continue and it may mean that traits associated with animal welfare and environmental impact have higher weights in selection indices than do those from solely from economic drivers. Most of the traits (Berry 2015) are already considered in most national dairy cattle selection indexes. Extensive work by Berghof et al.(2019) showed that if production and the most important health-related traits are already included in the selection index, the benefit of adding resilience indicator traits is expected to be small. Resilience, which is the capacity of an animal to be minimally affected by disturbances or to rapidly return to the status before exposure to disturbance, is suggested as additional trait to select for (Berghof et al. 2019). Most likely resilient animals, as well as animals considered to be generalists, are likely to be able to withstand weather variability compared with specialist animals (e.g. Bryant et al. 2007a; Cheruiyot et al. 2020).

Currently, heat tolerance is not included in Australia’s national selection index (balanced performance index; BPI), which means that farmers need to consider the heat tolerance ABV as a secondary selection criterion. One of the challenges with including heat tolerance in the BPI is that Australia is a large country with very diverse climates, so the economic value of selecting for heat tolerance may vary by region. For instance, where the environmental heat load is higher, there may be more justification in selecting for heat tolerance than in regions with a lower heat load. Using projected climate data from Commonwealth Scientific and Industrial Research Organisation (CSIRO; Canberra Australia) and The Bureau of Meteorology, Nguyen et al. (2016) determined levels of heat load for various dairying regions in Australia and used this information as weights for heat tolerance in an adapted version of the BPI. The Pearson correlation between the two indices, i.e. BPI and the BPI with heat tolerance (with weights appropriate for a high THI environment) was ∼0.95, which implied only a small amount of re-ranking of sires and there is currently no justification to have regionally specific BPI.

To summarise this section, a question that is often posed is whether it is better to select the cow to suit the system, or alter the system to suit the cow? Most short-term challenges from a changing environment are best addressed by using management strategies and long-term problems can be addressed using genetic selection. Sires described as generalists have daughters with comparable performance across systems/environments. In effect, the heat tolerance ABV should be helpful in identifying generalists, as animals with higher ABVs have less reduction in yield in hotter and more humid conditions. Furthermore, it is often impractical to implement cooling equipment in grazing situations, making genetic selection a very cost-effective approach.

Currently, the Australian breeding objective for heat tolerance includes three measurements (usually referred to as selection criteria in animal breeding); they are the decline in milk, fat and protein yields as THI rises. It is possible to increase the reliability of breeding values by incorporating other selection criteria. This is an approach that has been used effectively for other traits; for example, traits such as fertility are complex and multi-factorial and lend themselves well to multi-trait models that include different aspects of the biology of fertility (González-Recio et al. 2016). The same is likely to be true for heat tolerance, that there are different heritable aspects of heat tolerance that can potentially be used to improve heat tolerance EBVs, such as sweat gland density, body temperature, respiration rate etc. Understanding of genetic (co)variance components, for example, with other heat tolerance traits in addition to other traits under selection is important to quantify the possible impact of additional selection criteria. As mentioned earlier, the pre-requisite is that these traits need to be measurable in a genomic reference population, such as Australia’s Ginfo (females) and obviously heritable. It should be noted that even lowly heritable traits, such as fertility, often have considerable genetic variation and even larger variation affected by management, environment, chance etc. (Berry et al. 2014). Here, we discuss the potential role of wearable sensor devices, either for continuously measuring internal body temperature or wearable devices that can provide behavioural data associated with heat tolerance in addition to promising molecular phenotypes.

Many modern internal (adapted intra-vaginal or rectal) devices are capable of continuously logging body temperature data; however, the limitation is that these can be used only for short periods (currently up to ∼14 days) and are comparatively expensive, making them unsuitable for large-scale measurement. Nevertheless, Dikmen et al. (2012) generated a large enough dataset to determine the heritability of rectal temperatures under heat stress conditions of 0.17. More expensive and longer-term alternatives include surgically implantable temperature measurement devices, including loggers that require subsequent removal from the animal and, more attractively, those that transmit data via telemetry. Ruminal temperature boli with telemetry capability also show promise and are less invasive but may be influenced by water intake (Bewley et al. 2008; Cantor et al. 2018) and, unless the animal is fistulated, remain in the animal forever, rendering them single use and reliant on adequate battery life. Ruminal temperature is approximately 0.5–0.6°C greater than is rectal temperature (AlZahal et al. 2011; Timsit et al. 2011).

Recent advances in sensor device technology might make devices that are wearable on the outside of an animal a viable alternative, such as panting scores (Gaughan et al. 2008) and tympanic temperature measurement devices. Many wearable sensor devices are now capable of measuring heavy breathing on an almost continuous basis. Bar et al. (2019) found good concordance between accelerometer measures of lactating cows breathing heavily and vaginal temperatures.

As wearable devices are becoming more popular, this could be an excellent opportunity for developing large datasets for estimating breeding values on the ability of animals to tolerate heat. However, simply relying on these sorts of behavioural measures without validation can be risky. For example, Garner et al. (2016) observed that cows with the greatest respiration rate/panting score did not always have the greatest core body temperature or the level of heat stress. This is probably due to cows dissipating more heat through respiratory evaporative cooling. Garner et al. (2016) found that cows can have only slightly elevated respiratory rates, but very high core body temperatures, which was the case for most of the heat susceptible cows since they have less efficient evaporative cooling systems. One of the significant gaps in current knowledge is how to use sensor data to develop phenotypes for heat tolerance that could be used for animal breeding purposes. However, at the very least, breeding values developed using solely panting scores need to be carefully validated, ideally with internal body temperatures.

Other selection criteria that could be worth considering include metabolites that change under heat stress. Initial work by Agriculture Victoria colleagues (Liu et al. 2017) has shown that heat alters the lipid composition of milk. The review by Ríus (2019) describes how the breakdown of whole-body protein and synthesis of urea increases under heat stress, suggesting that amino acids are mobilised to meet the animal requirements. The limitation with this type of approach has been to obtain phenotypes on a sufficiently large population for genetic evaluation purposes. One option is to use MIR, as this should detect the biomarkers associated with variation in heat tolerance present in milk (König and May 2019). Hammami et al. (2015) found that several milk fatty acids predicted using MIR were sensitive to heat stress, most notably C18:1 cis-9, which is known to reflect body tissue mobilisation and could indicate cows with depressed appetites and yields.

A rise in body temperature modifies cellular function (Hansen 2020). In this regard, biomarkers of oxidative stress could also be useful, especially if these underpin pathways that are associated with genetic variation, which seems plausible. Specific heat shock protein genes have been proposed as promising biomarkers for heat stress and thermotolerance in dairy cattle (Zachut et al. 2017; Garner et al. 2020). A potential biomarker of oxidative stress is heat shock protein 70 (Basiricò et al. 2011) which has been found to be upregulated in the mammary epithelial cells of dairy cows during heat stress to improve the thermotolerance of the mammary gland (Garner et al. 2020). In addition, how the immune system responds to an environmental stressor, such as heat, could also be useful to know. Garner et al. (2020) found that genes highly upregulated during heat stress were involved in increasing the rate of inflammation and the function of the innate immune system via the action of cytokines. Additionally, it appears that there may be endocrine and paracrine signals to heat stress that modify intracellular responses to heat stress (Collier et al. 2008). These systems can also lead to an altered physiological state (acclimation).

On the basis of our recent work of genomic selection applied to serum biomarkers of animal health (Luke et al. 2019), it appears that developing genomic predictions using female reference populations where biomarkers are measured in serum is promising. Furthermore, research in phenotypes predicted using mid-infrared spectroscopy, where a spectrum is obtained from a routine herd-testing of milk sample, could pave the way for expanding the genomic reference population at a minimal extra cost (van den Berg et al. 2021). Elucidating the physiological systems controlled by genes involved in thermoregulation is an area worthy of further research, as understanding the biological control will help prioritise variants in genomic selection strategies in addition to understanding how G × E is modulated.

There are other ways that heat affects dairy cows. For example, Dahl et al. (2016) stated that heat stress reduces dry-matter intake, which in turn reduces yield and compromises immune function. Alternatively, indicators of recovery after a heat event could also be a useful selection criteria, such as an analysis of the slope of the return to baseline milk yield following heat stress, sensors may improve our understanding of the recovery period. There is evidence that dairy cows varying in genomic heat tolerance recover their feed intake and milk production differently following periods of heat stress (Garner et al. 2016). Many of these traits are correlated. For example, there is evidence that traits associated with disease or parasite resistance are favourably correlated with heat tolerance (Mackinnon et al. 1991; Burrow et al. 2019). This is an important concept in defining new breeding objectives and corresponding selection criteria for traits associated with adaptation and robustness across a broad range of environments. Opportunities to extend selection criteria to include the impact of heat stress on health and fertility will help capture the complexity of heat stress and, potentially, lead to the selection of better adapted dairy cattle.

The more accurate the measure of microenvironment surrounding animals is, the more reliable is the breeding value estimated. Most studies have relied on public weather stations for temperature and humidity data, which can be scarce in some areas or located in uneven terrain and are unlikely to represent microclimate at a farm level. That makes matching the environmental with phenotypic records less than ideal. One option is to have mini-weather stations installed on farms, such as locations of Ginfo herds. Another option is to obtain higher resolution/downscale weather data from meteorological agencies. The above will bring about better representation of farm microclimates and more variance on heat tolerance could be captured.

An approach of combining genomics, and traditional and novel measures of heat tolerance with intermediate metabolic biomarkers and prioritised genetic variants is likely to be an effective strategy to capture the complexity of thermotolerance in next-generation heat-tolerance breeding values. Reducing the impact of heat stress on animal performance can be achieved by using both genetic selection as well as improving herd management; in this regard, selection and modifying the conditions we keep cows are complementary strategies.

Australian breeding values for heat tolerance are publicly available for active sires and can be accessed from the DataGene website (https://datagene.com.au/BreedingValues). Other requests should be directed to DataGene Ltd (Melbourne, Australia) as custodians of data on Australian dairy cows. Research-related requests for access to the data may be accommodated on a case-by-case basis.

The authors declare no conflicts of interest.

This study was undertaken as part of the DairyBio research program, which is jointly funded by Dairy Australia (Melbourne, Australia), the Gardiner Foundation (Melbourne, Australia), and Agriculture Victoria (Melbourne, Australia).