Phenomics for sustainable production in the South African dairy and beef cattle industry

Phenotypic and pedigree recording of livestock remains a challenge in developing countries with informal agricultural sectors. Automated technologies would assist commercial farmers to become more efficient and productive, and to make full use of genomics. Farmers will benefit from the use of automated technologies in terms of improved animal welfare and economic sustainability in resource-poor areas. The impact of the rapidly growing human population on natural resources and in particular on food security has been well documented. As an example of the growth expected in Africa, the current South African human population of approximately 57.7 million people (Statistics South Africa, 2018) are expected to increase to 72.8 million people by 2050 (United Nations, 2017). Nutritional demand, specifically the demand for animal protein, would follow a similar pattern, and the responsibility of meeting this demand will place additional burdens on livestock producers. Climate change will add to this already demanding task, with extreme environmental changes expected for developing countries south of the equator. Modern genetic improvement in livestock relies on the availability of genomic data (e.g., Single-Nucleotide Polymorphism genotyping and sequencing), as well as accurate recording of relevant phenotypes. The combination of the available phenotypes and genomic data have contributed to identification of specific genes and quantitative trait loci of economic importance and generation of genome-enhanced estimated breeding values for application in genomic selection. Although the field of molecular genetics has advanced rapidly over the past two decades, accurate phenotyping still remains a serious constraint, and has become the overriding bottleneck in improving livestock production and efficiency. In South Africa, there are almost 14 million cattle which constitute 1.6 million dairy (604,781 cows in milk) and 12.5 million beef cattle. Of the latter, approximately 53% are kept in commercial systems and the remaining 47% in informal systems (DAFF, 2019). Phenotypic and pedigree recording of livestock in Africa faces constraints in terms of the extensive nature of the farming systems and the large informal livestock sector consisting of communal and small-holder farmers, which is characterized by a general lack of resources such as financial, infrastructural and extension support. Within the commercial sector, participation in collection of data from animal production systems ranges from compulsory participation (100% participation for stud breeders) to extremely poor (mainly within the informal and small-stock industry). Identifying breeding objectives across this wide spectrum of production levels is crucial, followed by the collection of relevant data to address the various issues. Livestock phenomics is defined as obtaining high-dimensional phenotypic data on an animal-wide scale, which can capture missing heritability, break down composite traits into their components and transform hard-to-measure traits into easily measurable traits (Greenwood et al., 2016). Several state-of-the-art technologies (mostly making use of remote sensing techniques) to enhance phenotyping have been developed and are in various stages of incorporation within the crop production field (Mir et al., 2019). Plant phenomics have the distinct advantage of plants being stationary, and basic crops can be produced rapidly and artificially, with short generation intervals (Greenwood et al., 2016). Object tracking is more complicated in livestock than in either crops or humans, as similar colors and shapes between animals, as well as background clutter often lead to failures (Kim et al., 2017). The use of radio-frequency identification also has limitations, as the range of these sensors is quite restricted. There is a growing interest in low-cost sensor solutions and the use of mobile platforms to address these challenges. This review aims to discuss the use of precision phenotyping in the beef and dairy cattle industries of South Africa, highlighting the challenges, limitations and possible contribution of these technologies towards the livestock sector. A key enabling technology for precision phenotyping is sensor networks. A sensor is a device that converts an observed property into an electrical signal. In wireless sensor networks, multiple sensors consolidate their observations wirelessly to provide fine-grained monitoring and automation in challenging environments. An important aspect in designing wireless sensor networks is in selecting suitable sensors, which is typically application driven. Sensors are classified as attached or nonattached. Attached sensors are composed of sensors inserted inside an animal, and wearable sensors that are fitted to the outside of an animal. Nonattached sensors are off-animal sensors which observe an animal, animal by-products, or the environment. Figure 1 provides an illustration of the general framework of a sensor network which could be applied in the South African dairy and beef industries for precision livestock farming. Illustration of the general framework of a sensor network-based precision livestock farming application. Although a clear abundance of various sensors exists in the literature (corresponding to the physical/communication blocks in Figure 1), there is a lack of intelligence in sensor systems (which is provided through the cloud processing/local processing blocks in Figure 1), as sensor measurements are presented to farm personnel directly, or with minor processing, who are then required to draw their own conclusions (Rutten et al., 2013). However, this becomes problematic when there are a large number of animals producing excessive amounts of sensor data that must be interpreted by farm personnel; or when farm personnel do not have the required level of education to interpret sensor readings. This has led to poor adoption rates of intelligent systems by farmers. Poor adoption rates have also been attributed to exorbitant costs of solutions which are difficult to operate, and a lack of solutions that provide meaningful alerts. This is consistent with multiple farmer surveys on drivers of decision making which highlighted economic feasibility, and usability and technical support among others (Eckelkamp, 2019). In the context of Southern Africa, it was also found that technology adoption is linked to farmer’s participation with an innovation platform, which is described as a multisectoral and multi-institutional coalition of actors in specific value chain systems (Hanyani-Mlambo et al., 2017). These platforms provide greater access to agricultural advisory services and other support services. Genetic progress in dairy cattle over the past three decades has been substantial compared to beef cattle or sheep. The extensive use of reproductive technologies in dairy cattle has given this industry an advantage in making genetic progress in production traits and more recently in traits with low heritability with the application of genomics (Miglior et al., 2017). Decades of selection pressure on milk yield has resulted in a number of unfavorable outcomes in fitness traits. DNA marker technology has solved many challenges regarding the identification of recessive genes using diagnostic tests, while genomic selection is contributing to genetic progress in fertility and feed efficiency (Miglior et al., 2017). The trend worldwide is towards larger dairy farms with economics as the main driver (Rutten et al., 2013). A similar tendency is observed in South Africa where the number of milk producers decreased from 3,551 in 2009 to 1,253 in 2019 (Lacto Data, 2019), while herd sizes have increased. Pasture-based production accounts for up to 70% of dairy production in South Africa, and the remaining 30% are on total mixed ration systems (Meissner, personal communication). The larger dairy units have adopted automated milking systems, which require less labor and have the advantage of automated recording of a large number of phenotypes. The automated milking systems include AfiFarm, DeLaval-Alpro and DeLaval-Delpro management systems as shown in Figure 2. Traditional milk recording schemes are used by less than 10% of South African dairy farmers (SA Stud book, http://www.sastudbook.co.za/p116/services/logix-milk.html). Percentage dairy farmers using the three automated milking systems in South Africa (Gresse, 2018). Production, fitness, health, and workability traits have been included in genetic evaluations providing producers with estimated breeding values, genomic enhanced breeding values, and selection indices (Van Marle-Köster & Visser, 2018). Welfare traits in dairy cattle, which include claw health, lameness, mastitis, and other health traits are more difficult to improve due to limitations in effective routine recording, compared to production traits, as well as the low heritability of these traits. Precision or smart farming has presented advanced technologies for automated recording of a large range of welfare phenomes using automated monitoring systems (Hansen et al., 2018; Alsaaod et al., 2019), with the primary focus on early detection of a potential problem for timely interventions. Table 1 describes an array of sensors currently used in dairy farms. Nonexhaustive list of sensors used for dairy cattle health management (Adapted from Rutten et al., 2013 and Mottram, 2016) Nonexhaustive list of sensors used for dairy cattle health management (Adapted from Rutten et al., 2013 and Mottram, 2016) The average herd size (in terms of cows in milk) in South Africa varies from 918 in the pasture-based provinces to as low as 119 in the drier regions (Lacto data, 2019). The larger operations making use of automated milking systems are primarily found in the coastal regions. Gresse (2018) demonstrated the use of data from automated milking systems for improved cow production, emphasizing the potential of the data for improving a range of dairy production traits. Implementation of precision dairy monitoring technologies in South Africa will be dictated by herd size and feasibility of automated milking systems. Improvement of fertility relies largely on sound reproductive management and accurate record keeping. Estrus detection is often a primary constraint and various studies have shown that pedometers can improve detection compared to using only observations by herdsmen. The reliability of pedometers, however, has been shown to vary and can therefore not yet be implemented as a fool-proof method (Mottram, 2016). Rutten et al. (2017) explored the automatic behavior detection of calving, where each cow was fitted with an ear tag device consisting of sensors which monitored cumulative activity, rumination activity, feeding activity, and temperature on an hourly basis. Based on the sensor data, a predictive model predicted 43.5% of the calving events with 1% false positive alerts. A range of precision dairy monitoring technologies were tested against the reference gold standard (progesterone patterns) and the accuracy of estrus detection was improved by between 15% and 35% compared to human observations (Mayo et al., 2019). The sensors in this study ranged from attachments to the front or rear legs, the ear or neck or a bolus in the reticulorumen, which all monitored activity, eating, rumination and lying times. A primary advantage of precision dairy monitoring technologies is the continuous measurement of these activities which have been linked to health and metabolic disturbances (King et al., 2018). These measurements hold the potential for early detection of estrus and enhanced management as dairy units become larger and the farmer has less time available to inspect the herd. Development of tri-axial accelerometers and sensors attached to ear tags also holds potential for more effective recognition of other cow activities which include behavioral changes, eating patterns, and locomotion (Mottram, 2016). Rahman et al. (2018) placed accelerometers on a neck collar at the back of the head, on a halter position at the side of the head behind the mouth, and on the ear using a tag. The data were used to train a machine learning-based model to classify three animal behaviors (grazing, standing, or ruminating), and attained accuracies of approximately 90%. In South Africa, collars and pedometers are used for detection of rumination and eating patterns, activity, and estrus, while milk meters and analyzers are popular for milk traits such as milk flow, conductivity and composition. Substantial data are generated in automated milking systems in South Africa, but the data are primarily used for short-term management and economic decisions (Gresse, 2018). Somatic cell counts are used as an indicator trait for incidence of mastitis. An animal with a cell count of less than 50,000 cells/mL is regarded as a noninfected animal (Seegers et al., 2003). Mastitis being a complex trait and influenced by several factors, individual herd data are not available. The Milk Producers’ Organization (MPO) of South Africa estimated a loss of 1 billion ZAR for 2019 due to sub-optimal udder health (Chris van Dijk, MPO, personal communication, https://www.mpo.co.za). The consequences of lameness in dairy cows have been well documented, including adversely affecting genetic improvement, welfare, and farm profitability. Visual inspections and scorings are both time consuming and subjective. Automated methods for lameness detection include kinematic sensors (on the legs and feet) for monitoring body movement with three-dimensional video analyses or accelerometers and pressure-sensitive walkways. There are also kinetic methods using walkways with pressure plates based on ground force reaction (Alsaaod et al., 2019). Walkways with pressure sensors can distinguish between lame and non-lame cows and provide an indication of potential claw lesions (Volkman et al., 2019). Lameness cannot be considered without recognizing the role of claw quality and health, which are both influenced by housing and environmental factors. Literature indicates that claw traits should be considered for improvement of lameness and moderate genetic correlations were reported between claw health and lameness. However, accurate scoring of claw traits and lameness remains the main challenge. In South Africa, claw data are limited to producers who make use of private hoof trimmers, who record the claw lesions on paper and the data is not necessarily captured in an electronic recording system (Mhlongo, 2019). Claw quality and health are not included in goal driven selection and available data have not been explored in research for lameness in South African dairy cattle. Widespread regions of South Africa are experiencing a state of drought, which is the result of the strong El Niño event that occurred in 2015 to 2016 (South African Weather Service, 2016). The shortage of water in the country has raised concerns, among other things, about the amount of water it takes to finish cattle in feedlots (Meissner et al., 2013). Approximately 68.6% of South African land is available for grazing, which is an ideal situation for extensive livestock production systems that rely on natural veld as a feed source. Global warming, however, will be responsible for fluctuations in the nutritional value of the natural veld (Scholtz et al., 2014). There is pressure from consumers to produce beef with less greenhouse-gas emissions and limited exploitation of natural resources, while considering the health and well-being of the livestock. Most of the traits relevant to these objectives are difficult and expensive to measure, expressed late in life, and for some new equipment are still being developed. Phenomics aims to make use of advanced technologies and management systems to develop labor saving automated data collection. This would result in the collection of large amounts of phenotypic data, which could be used for genomic prediction and faster genetic improvement. The level at which phenotypes could be measured varies between traits (i.e., behavioral vs. physiological), and systems (i.e., in-field walk-over vs. animal-attached collars). Traits that are routinely recorded by South African beef breeders commonly reflect easily measured traits that can be measured on farm with no additional such as to growth and and a of traits at calving and as by Marle-Köster & Traits such as feed feed efficiency and are extremely difficult to in extensive production systems. is the for many efficiency and the measurement of of would also be important as a and have been used for this but have limitations in terms of over of walk-over as well as sensors to behavior have been developed to this but vary in terms of et al. (2018) that the number of that can be measured is and this a challenge with to developing a phenotyping across systems. A number of individual breeders only beef cattle for traits. This is not available to all farmers at an additional and farmers should have an economic to such additional South Africa has a which the of the animal, of the and to the & 2017). a of the and and not make for of based on However, the has recently the South African beef and in this is with the for the A of is the use of a The used by measurement of and shown in Figure and will also be used to and of quality using the of the automated have been to the South African beef of the to would be in-field walk-over systems, such as used in the system (Greenwood et al., 2016). sensor data could also be used for phenotyping and of feed on as well as and behavior such as and would be the application of These require such as or but could become more the used for genotyping could be applied for phenotyping as This would assist extensive commercial farmers in more and as precision phenotyping would be without additional costs to the A illustration of some phenotyping is given in Figure A of some traits that would benefit from precision phenotyping in both beef and dairy cattle in South precision phenotyping would also improve animal health and welfare in resource-poor by early animals, and et al., 2019). it could have a impact on by and the technologies were commonly several challenges in terms of and quality would have to be this could become a The a new in human where of the past decades are to change and improve the and 2019). the of the and and it is to the The of networks, and with a of available sensors, will the for improving the efficiency and of the agricultural sector. These technologies will to be available to the farmer in and automation of key farming which will lead to efficiency and while the of The will be key in affecting and of and farmers, enabling and Development of intelligent cattle monitoring and behavior systems will identification and tracking of individual cattle in the as well as and various can the and of an individual animal over in to it to the of measurement is in the developing sector where is not available. could provide into the nutritional of the of the the of or and measurement would be to the of estrus, which could improve the of and therefore production rates and the of genetic improvement. monitoring systems would typically make use of the to provide a video of the cattle. identification at the of the could individual animals through the The data and from this system could be to a using various wireless from where total farm would be livestock monitoring and behavior a number of for commercial and farmers to the health of their animals, which can be used to make decisions to sustainability and improve This would farmers to by providing to the farmer at given systems could be to number of farms for large and farming there are a number of in of in the cattle farming industry with the potential to increase and farmers. The impact that such a system could have is it is provided at and farmers will be to in the which will have a of and on their generation and also to the and The use of measurement technologies would assist South African commercial producers to become more efficient and profitability. relevant phenotypes will the of which will assist in making full use of genomic the other side of the it could assist in improving animal welfare and of farmers, as well as contributing to economic sustainability in areas. and and in in the of and of South for the of to the and is currently a research focus is primarily on the of including population genetic studies on and and As food security is a also on the genetic improvement of livestock which can to the of this in in and from the of South has been at this in the of and where is the head of the research current research are in wireless systems, sensor machine and mobile is the of the and solutions and is a of the van Marle-Köster the of & of in as a in the agricultural industry for applied for the South African study on genetic of developed and used in Southern Africa have been included in genetic studies and in other current research interest is on the application of genomic technology for genetic improvement of livestock with a for beef and dairy cattle in South has interest in genomics to improve of the genetic traits of economic importance and to traits with animal welfare and production in a South African the and in from the of South a from the and at the of is a in the of and at the of are in the of processing and machine in sensor networks, and complex systems.