Introduction

The information technology revolution is impacting on most aspects of modern life, and agricultural production is no exception. Precision farming is about measuring and managing the variability that is inherent in the crops and animals we farm. For example, in arable farming, crop growth and yields can be mapped and then fertiliser and other treatments applied only where they are needed. The same approach of targeting inputs only where they are needed, when applied to livestock production, is called Precision Livestock Farming (or PLF). This paper will focus the application of PLF approaches to dairy farming, giving an overview of current dairy PLF technologies and considering the likely developments in the future.

EID and oestrus detection

Electronic identification (or EID) is the core technology in PLF, as it enables individual animals to be automatically identified, helping link data to individuals and allowing them to receive individual treatment. Current EID systems are generally ‘passive’ i.e. the tags do not contain their own power source, but are energised when they are close to the reader. Although ‘active’ EID tags are larger and more expensive, they are more reliable over longer ranges and can store data. Given the desire by dairy farmers to improve herd genetics using artificial insemination and the fact that getting a cow pregnant is a crucial step in milk production, one of the first applications of PLF in dairying was in oestrus (heat) detection. There are several changes in cow behaviour during oestrus, including an increase in activity (up to four times higher) and a reduction in daily feed intake. Leg-mounted pedometers can be used to detect the increase in activity levels, and rumination monitors can detect the drop in food intake, as time spent ruminating is closely linked to fibre intake. But there are times when cows ovulate but fail to show any of these behavioural signs of oestrus (known as a ‘silent heat’), and these cannot be detected by systems that rely on behaviour changes. Recently, a commercial system(DeLaval’s Herd Navigator™) that monitors progesterone levels in milk has been developed, and this has the potential to detect both normal and silent heats and so improve oestrus detection.

Physiological and health monitoring

As well as the veterinary and production costs of lameness, it has a negative impact on animal health and welfare. Clearly, technology that can help in the early detection of the lameness can bring considerable benefits. The Boumatic Step Matrix™ system can detect gait abnormalities in cows as they walk over it. Another approach to detecting the gait abnormalities associated with lameness is to use automated analysis of video images of the cows as they walk past the camera.

The development of low-cost accelerometers for video game controllers and smart phones has made the technology cheap enough for other uses, including animal-mounted sensors. Unlike the earlier generation of pedometers which simply gave activity levels, accelerometer-based leg-mounted sensors (e.g. IceRobotics IceQubes) can record the number of steps taken and time spent lying down. As well as facilitating oestrus detection, these parameters can also be used to detect health conditions such as lameness. Another spin off from the video game industry is comparatively cheap three-dimensional (3D) camera technology. This is used in a system supplied by DeLaval that can give an automated body-condition score for each cow that walks under it. Walkover-weighing systems allow the liveweight of cows to be recorded as they walk over the weigh platform when e.g. they leave the milking parlour.

Rumen pH boli are also available, although the challenging conditions in the rumen mean that the sensor only gives reliable pH data for about five months. Consequently, pH monitoring is typically used either as a veterinary diagnostic tool or is used in a few ‘sentinel’ animals rather than continuous monitoring of all the animals in the herd. A rumen bolus can also contain accelerometers and temperature sensors, the latter not only helping detect heat stress but can also be used to monitor water intake. As well as monitoring progesterone in milk samples, DeLaval’s Herd Navigator™ also uses lactate dehydrogenase to detect mastitis, beta hydroxybutyrate to detect ketosis and urea levels to monitor feed protein balance.

Animal positioning

Although satellite positioning (e.g. GPS) has been used since the 1990s to track domestic animal movement, the high power requirements and comparatively poor indoor performance of GPS receivers mean they are currently impractical for on-farm animal tracking. However, several companies have recently brought to market systems that allow the on-farm tracking of cow movements using radio-location. As well as helping a farmer find a particular cow in the herd when e.g. she needs to be inseminated, position data has the potential to provide valuable management information. One challenge here is that the on-farm deployment of such technology has jumped ahead of scientific research, and more research is needed to understand and maximise the benefits of cow position data. For example, knowing the relative movements of all the animals in the herd should give insight into cow social behaviour, allowing more active management to reduce instances of aggression between cows.

Robotics

Robotics have also made a big impact in dairying, with robotic milking systems gaining in popularity. Most of the commercial robotic milking systems currently on the market accommodate a single cow and a laser-guided robotic arm attaches the milking cluster. Cows choose when to be milked, with the reward of concentrate feed attracting them to the robot. Robotic arms can also be used with rotary and conventional parlours to reduce labour requirements, and it is predicted that half of the cows in NW Europe will be milked by robot by 2025. Alongside milking robots, fullyautomated feeding systems are now available, as well as robotic feedpushers and robotic slurry scrapers. Likely restrictions on the movement of foreign labour in post-Brexit Britain will probably accelerate the adoption of robotic dairy systems in the UK.

Future developments

One area that is currently being researched is improved data integration. The reliability of the detection of oestrus, health issues and other conditions can be improved if data from a range of sensors are integrated. Although the larger dairy equipment manufacturers sell systems which include such multi-sensor data integration, farmers do not always want to purchase complete systems from a single supplier. Many farmers want to buy different technologies from different suppliers as this allows them to add new capability to their existing equipment as well as allowing them to pick and choose the best sensor technologies. However, at the moment, equipment from different manufacturers rarely communicates with each other, so multi-manufacturer systems do not benefit from multi-sensor integration. Ideally, common data exchange standards will be established and adopted by equipment manufacturers, as this would allow third-party software to integrate data and improve the reliability of management decision support information provided to farmers.

Recording reliable estimates of on-farm individual feed intake has been a goal for some time. Although there are systems such as Grow-Safe, which utilise EID and weigh cells under the feed to record individual feed intake, they are expensive and are only really viable in progeny testing or scientific research. Neck or head mounted accelerometers can give estimates of feeding time, but overall feeding time is not a reliable indicator of feed intake as it includes time spent searching (or sorting) feed as well as eating it. However, it should be possible with more advanced sensing and data analysis to discriminate between searching/sorting and eating using accelerometers or bioacoustics. As well as giving more reliable estimates of intake (based on eating time), the amount of time spent sorting could also be a useful diagnostic. For example, changes in sorting time may indicate an improperly mixed Total Mixed Ration due to a malfunction of the mixer wagon or operator error. Sorting time may also be influenced by health status, although further research is needed to explore the diagnostic potential of eating behaviour data.

Although the majority of PLF technologies developed to date are used in intensively managed dairy systems, there is no fundamental reason why the technology should not be used in more extensive systems. There is considerable scope to use sensor technologies and robotic control to enhance grazing management, although again further research is needed to develop the technologies and management systems.

Impact of PLF on animal welfare

There are several ways in which PLF systems are helping to promote the welfare of dairy cows. By providing continuous, 24/7 monitoring of cows, PLF technologies are already speeding up the detection of health problems and so improving animal welfare. It is important to emphasize that these technologies are a tool to help the stockperson and not replace them or their skill. However, PLF systems are arguably getting better than humans at detecting the subtle changes in behaviour associated with the early stages of disease. Robotic milking systems also reduce the stress associated with being gathered by humans and moved into the close proximity of other cows in the collection yard before being milked. Automated estimates of feed intake, liveweight and body condition score can be used to help optimise the nutrition of each cow, and this should help reduce hunger.

PLF systems need to be fail-safe i.e. when they fail, and it is likely most systems will fail at some point, they need to fail in a way that does not compromise animal welfare and has minimal impact on production. Ideally, monitoring systems will alert the farmer as soon as there is a problem, and rapid technical support will be able to fix the problem quickly. However, the farmer needs to have plans in place to ensure they have the backup capability to provide water, feed and to milk the animals until the system can be fixed.

Conclusion

Precision livestock technologies are already having a big impact on commercial dairy farms, improving production efficiency, enabling more rapid detection of health problems and helping promote animal welfare. Future developments in enhanced sensing capabilities and data integration should lead to further improvements in these areas, and reductions in costs should enable these benefits to be applied to beef and sheep production.