Problem & Solution

Currently there are few farms that actively consider resilience & efficiency (R&E) when making management decisions and commercially available software packages tend not to calculate or display herd level resilience and efficiency values effectively on farm. There is thus a need to raise awareness of the potential importance of assessing R&E.

Using commercial farm settings, GenTORE has delivered Champion Farm Events in various countries to show farmers that resilience and efficiency could be characterized and predicted using currently available sensor technologies.



Outcome

Dissemination to farmers and veterinarians was achieved through training workshops and guided farm walks that highlighted the use of current on-farm Precision Livestock Farming (PLF) technologies and sustainable farming practices. From the Champion Farm Events, delegates understood the definition of Precision Livestock Farming and how data from commercially available sensors can be used to predict herd level Resilience & Efficiency.



Recommendations

Ranking cows on Resilience & Efficiency using commercially available PLF sensor data can offer new and effective insights to inform herd level as well as individual animal management changes.

Problem & Solution

Defining and predicting 'different and changing environments’ is difficult but using Scenarios allows the testing of resilience and efficiency and so performance of proposed selection and management strategies.



What we need

Important aspects of scenario analysis include:



Baseline conditions:

• The starting point of any model that describes today's production environment.

• Lots of data is needed to test strategies.



A set of scenarios:

• Shared Socio-economic Pathways (SSPs) include scenarios describing major aspects of global human development.

• Because climate is affected by human development SSPs are linked to a set of climate scenarios called Representative Concentration Pathways (RCPs)

• RCPs include scenarios that describe how global climate is likely to evolve.

• Using SSPs and RCPs will ensure future scenarios include key drivers that will define many aspects of future food chains and production circumstances.



Selecting scenarios

SSPs and RCPs are global-scale scenarios and must be carefully modified for GenTORE's regional-scale production data to better understand the consequences of specific challenges to a farm system.



Outcomes

Researchers will be able to assess the impact of selection and management strategies on farm resilience and efficiency, gaining key insights into short-term decisions and longer-term outcomes.

Problem & Solution

The price paid for animals destined for beef is usually dictated by breed, weight, age, and visual appearance. Visual appearance relates to the conformation and fleshiness of the animal, but it’s not easy to assess in young dairy-beef calves. A decision support tool to predict both the revenue and costs associated with an animal up to slaughter will help assess what price should be paid for that animal.



Outcome

A framework was produced for a comprehensive decision-support tool (B.O.W – Beef’s Own Worth) providing the farmer with a single Euro value estimate of the expected profit potential of a growing animal destined for slaughter. This Euro value is a function of five traits:

1. Carcass weight

2. Carcass conformation

3. Carcass fat score

4. Animal docility

5. Animal feed intake

A ‘production value’ is estimated for each of these traits, which are then used to estimate a single animal’s potential (BOW - Beef’s Own Worth). Factors such as genetic merit, animal sex, singleton or twin, dam characteristics, and whether both the cow and calf are crossbred are considered in the estimation of the production values. The BOW can be applied to any breed or any age of animal. It is particularly useful in young calves, both pure- and cross-bred, especially in those from dairy herds.



Recommendations

The BOW updates in real-time as new data comes available, and its accuracy is improved if the animal is genotyped. Used in tandem with the beef female’s profit potential (BFPP) tool, the BOW can help to decide whether a beef female is more suited to become a replacement or should be destined for slaughter. The BOW is useful for producers to gauge what price to expect for their animals when selling, but also for purchasers to help determine animals’ values.

Problem & Solution

Identifying underperforming beef females for voluntary culling or beef heifers with greater lifetime profit potential to be kept as replacements is difficult due to its complex nature. Providing producers with a decision-support tool able to estimate the potential lifetime profit of beef females based on not only their current performance but also their expected future performance for various traits could be the solution.



Outcome

A framework was produced for a comprehensive decision-support tool (B.F.P.P – Beef Female’s Profit Potential) that provides farmers with a single Euro value estimate of the expected lifetime profit potential of all beef females in a herd. The euro value is broken down into four contributing factors:

1. the profit potential of the female as a heifer, provided she has not yet calved;

2. the profit potential arising from the beef female’s current parity, provided she has calved at least once;

3. the expected profit generated from her remaining future parities;

4. the retention value of the beef female, representing the cost-benefit of not voluntarily culling her.

Validation of the BFPP showed that the highest-ranked females calved earlier in the calving season, had a greater likelihood of surviving to the next lactation, and produced progeny with heavier, better conformed, and less fat-covered carcasses at slaughtering than the lowest-ranked females.



Recommendations

Quality data is fundamental for accurate ranking; producers are recommended to input all data into a national database. Based on the ranking, the producer should evaluate poorly ranked females for voluntary culling and highest-ranked heifers for graduating to the mature herd.

Problem & Solution

Feed efficiency is driven by both management and genetics. Gathering sufficient records to make proper genetic evaluations and to take into account biological changes over time is a challenge. Furthermore, current methodology does not consider that feed efficiency can change during lactation and across parities.

A multi-trait random regression model was applied to a large research farm data set with individual feed intake records on Holstein cows. This random regression model incorporates the knowledge of changes in variation over time into the modeling of the trait. Genetic parameters were estimated to assess how feed efficiency changes during lactation and across parities



Outcome

The approach showed that feed efficiency in early lactation is different from mid and late lactation in both first and second parity from a genetic point of view. The genetic correlations between mid and late lactation were strongly favourable in both parties. Across parities, the genetic correlations were around zero, which was unexpected. We will focus on improving the understanding of genetic correlations between parties and early versus mid and late lactation.



Recommendations

• The model requires frequent measurement of individual cows’ feed intake along with milk production and body weight across the lactation and parities.

• We recommend genetic evaluation centers to adapt the method for improving ‘saved feed’/‘feed efficiency’ indices

• We recommend breed associations to adapt or include ‘saved feed’/‘feed efficiency’ indices in their selection index to a higher degree.

• Model improvement is required better to understand the genetic correlation structure within and across parties.

Problem & solution

Resilient genotypes are able to better cope with environmental perturbations, which is very important when genotype by environment interactions (GxE) exist. Traditional genomic GxE models assume that GxE effects are similarly expressed across the genome. This assumption is not correct because specific regions in the genome harbor quantitative trait loci (QTL) and others do not, and loci may have different effects in different environments.

We have developed a protocol and analysis pipelines for genomic GxE models in which individual single nucleotide polymorphisms (SNP) effects may differ across environments.



Outcome

The analysis protocol consists of several steps and can be applied to dairy or beef cattle or other species. The data set of interest is split in two subsets and then follow a two-step approach: (1) estimate SNP effects in the first data set and calculate SNP (co)variances based on the estimated SNP effects and (2) weigh the SNP genotypes using the estimated (co)variances in (1) and simultaneously compute SNP effects and (co)variances in the second data set. The approach has been tested in simulated data showing a slight increase in accuracy of genomic breeding values of young selection candidates when allowing SNP (co)variances across the genome to be heterogeneous.



Practical recommendations

The analysis protocol can be applied to genomic GxE models like multi-trait or reaction norm models. The use of this protocol by genetic evaluation centres will aid in the selection of genotypes suited to their environment. The analysis protocol delivers genomic breeding values for animals in different environments, which are valuable selection tools for farmers to select the best animals for their environment.

Problem

Heat stress compromises dairy cows’ welfare leading to lowered productive and reproductive performances, to even death. Cows of the same herd can be more or less able to cope with the heat waves. Identifying and selecting the more resilient animals to heat stress would be desirable, but proxies for individual cows’ resilience to heat waves are still lacking.

Solution

By analyzing cows’ behavioral patterns during heat waves recorded at high-frequency by sensor systems installed on farms, it could be possible to distinguish between sensitive and resilient animals to heat stress.



Outcomes

Using GEE (Generalized Estimating Equations) models, we compared daily minutes of lying, rumination and activity recorded by Smartbow ear-tag accelerometers of ‘sensitive’ and ‘resilient’ cows to the heat waves in summer 2021. Heat waves were defined as at least 3 consecutive days with daily temperature-humidity index (THI) >75. Cows were ‘sensitive’ if they had at least one perturbation in the lactation curve linked to a heat wave, otherwise they were ‘resilient’.

No differences between the two groups of cows were detected during moderate heat waves (75 < THI ≤ 79), but ‘sensitive’ cows tended to spend more time ruminating and being active (and less time lying) during the severe heat wave (THI > 79) compared to ‘resilient’ cows.



Recommendations

The detected behavioral differences have to be interpreted carefully due to the modest sample size and observation period length.

In the face of the global warming issue, this kind of analysis could support farmers in identifying their most heat stress resilient cows, as well as could provide new indicators for the selection of heat resilient genotypes.

Problem

Whilst many farms now invest in on-farm sensor technologies to support their management, their application is often still limited to the detection of individual alerts that the farmer can react to with management adaptation at the individual cow level. By utilizing Precision Livestock Farming (PLF) approaches, big data outputs from commercially available biosensors have the potential to signal physiological, immunological, behavioural, and other variables in livestock, and assist decision-making for overall farm management, improving sustainable herd performance, resource efficiency and reducing labour requirements.

Currently there are few farms that actively consider herd level resilience & efficiency (R&E) when making management decisions and commercially available software packages tend not to calculate or display resilience and efficiency (R&E) values effectively on farm.



Solution

Using published algorithms and a small sample set of data, GenTORE has developed a demonstration dashboard that can visualize R&E ranking at a herd level with a simple and practical view for both farmers and vets. A second level of function highlights which of four performance areas is most likely responsible for R&E ranking changes: Production, Fertility, Health or Herd Inventory.



Outcome & recommendations

Ranking cows on R&E using commercially available PLF sensor data can offer new and effective insights to inform herd level as well as individual animal management changes. The quality and availability of data is of high importance when ranking cows in a herd to utilise PLF opportunities for sustainable production.

Problem

Economic, environmental, and ethical reasons suggest that a sustainably profitable dairy cow should stay in the herd until at least the third lactation, which is longer than the current average cows’ productive lifespan. Several advantages would arise from increasing cows’ longevity; for this purpose, farmers should be able to early identify the most resilient cows in their herds (i.e., those that better cope with their specific farm conditions) to optimize breeding and management decisions. Great support could come from sensor systems, which provide a constant flow of individual cow productive and behavioral measures.



Solution & Outcomes

This study explored the possibility of predicting cows’ probability of reaching at least the third calving (i.e., ‘mid-term resilience’) based on raw sensor data of cows’ daily milk yield, body weight, and rumination time recorded during the early stage (first 150 days) of the first lactation. A joint model for longitudinal and time-to-event data was applied to 9 datasets belonging to different commercial Holstein dairy farms equipped with automated milking systems. The model produced a survival probability with an average accuracy of 52% and a prediction error of 24%.



Recommendations

The developed algorithm should inform farmers about the prospects of their cows in their specific farming environment. The first 150 days of lactation were considered the maximum time that was useful to farmers making breeding decisions. Further research is needed to improve the model prediction performance, but it is flexible and can be used with additional on farm sensor measures.

Problem & Solution

There is a need for tailored solutions to achieve an optimal trade-off between resilience and efficiency in animal production. For these solutions, the impacts of the local production environment on production and efficiency need to be identified.

The European dairy sector was analysed using FADN economic data and AGRI4CAST climatic data to allow for regional differences to reflect different farm production environments. Individual farms in each region were analysed for production efficiency (comparing outputs vs inputs) and resilience (loss in performance and subsequent recovery).



Outcomes

The key results from the analysis are:

• European dairy production is efficient but has substantial regional- and farm system-related differences.

• The margin per cow is strongly correlated with the price of milk, creating a strong resilience challenge.

• Mediterranean region reacted less, thus showing increased robustness against price shocks.



Recommendations

• European dairy production should improve resilience through:

o Diversified income sources.

o Reduced costs of production.

o Reliance on homegrown feed.

• Differentiate the dairy policies to consider regional differences caused by climate and production systems.

• Milk prices require regulating through different instruments to increase resilience and sustainability of production, so a sustainable dairy sector in Europe can be maintained.

• The new CAP should consider different risk mitigation methods to support the diversification of income sources.

• Dual-purpose or crossbreeding with resilient breeds could increase the resilience of farms in regions expressing poor resilience and efficiency.

Problem & Solution

The need for resilient and sustainable cattle production systems involves improving animal resilience and efficiency against environmental challenges. Advanced herd management technologies already installed on dairy farms could provide relevant phenotypic data on individual cows, but more knowledge is needed on the distribution and type of such sensor systems among farms.



Outcomes

A large survey was performed on dairy farms of the Northeast of Italy in 2017 with information on the type of sensor system installed and parameters recorded.

The results of the study showed nearly 40% of farms were equipped with sensor systems to measure individual cow milk yield but the distribution of systems for monitoring cow behaviour was low.

The varying nature of sensor systems utilized on farms, accompanied by the diverse use the farmers can make of them, are constraints for the accurate collection of comparable data. However, sensor data has enormous potential for cow phenotyping purposes.



Recommendations

• The recording of cow behaviour by sensor systems is extremely useful to monitor real-time cow health and for the development of practical definitions for selecting new and complex traits such as resilience and efficiency and should therefore be encouraged.

• Data quality is fundamental to developing innovative algorithms for selecting more resilient and efficient cows.

• Proper use, regular maintenance of sensors with regular updating and backup of data are key for maximizing data quality.

Problem & Solution

Feed efficiency is a priority for the dairy sector and is traditionally estimated by residual feed intake (RFI). Yet, current methodology does not provide enough flexibility as predictors may vary during lactation. A new methodology based on a multi-trait random regression model that estimates RFI in a dynamic and continuous manner across the lactation was developed.



Outcome

This approach allows a continuous and more precise estimation of RFI over time, accounting for correlations between predictors and free from all time-related issues. RFI for each cow on each day of lactation is estimated as the difference between the actual intake and the intake predicted from three other traits using a multi-trait random regression model.



Practical recommendations

• The model needs continuous repeated measurements (every week or every day) across the lactation, ideally without missing data.

• The amount of data available is important: a sufficient number of cows with enough individual measurements is necessary for the model to work well.

• The model deals relatively smoothly with missing data. However, deduced individual effects that are outside the range of the actual measurements should not be used as a prediction.

• This approach is adaptable, and improvements are possible, e.g., by adding pedigree or genomic information that would allow separating the genetic from the environmental effect.



On-farm application

Individual intake measurements are mostly limited to experimental farms, hampering the application of the model to commercial farms. However, it can be adapted and used in a genetic or genomic selection context, with the aim of establishing a genomic selection on RFI.

Problem & Solution

When selecting for a livestock trait, there is no comprehensive view of the underlying biological mechanisms so one can only select on what is observed in a particular environment. However, whilst some biological strategies work well in one environment, they may not be well-adapted to another.

A computer simulation was used to test the performance of a population of virtual cows, all different in terms of resources acquisition and allocation, in different nutritional environments.



Outcome

The best combinations of acquisition and allocation were different for the different environments. Producing more milk during second lactation is genetically linked with short-term Feed Efficiency (FE). However, the link between these traits and long-term FE, recorded at a lifetime level, decreases as the environment becomes more constraining. In contrast, the link between body reserves and lifetime FE is positive and increases when the environment becomes more constraining.



Recommendations

• When using short-term FE as a selection criterium in dairy cattle, local farming conditions need to be considered.

• For high-quality environments, selecting cows that are the most efficient in the short-term (during second lactation) will lead to the selection of cows that are also the most efficient at lifetime level.

• For low-quality environments, the selection of cows with a lower production focused allocation will result in the selection of cows with a good lifetime FE.

Problem & Solution

Suckler cow efficiency can be estimated using data from growth and intake at young ages, so using a residual feed intake approach. But cow resilience (defined here as the ability to produce a heavy weaned calf every year) is difficult to be evaluated from real data without biases. So the trade-off between efficiency at young ages and resilience as adult suckler cows is not easy to be assessed. Data from an INRAE experiment was used to develop a cow-calf model in collaboration with UdL to simulate the performance and resilience of different groups of Charolais heifers under a scenario of nutrient restriction (decrease in quality and quantity of summer pastures expected by climate change).



Outcome

It was found that the most efficient heifers had more variability in the weight of calves at weaning under the climate change scenario and were, therefore, less resilient in that condition. The group that included the less efficient animals as heifers had higher production (higher weaning weight of their calves) and were also more resilient in terms of calving rate (less variability between years). On the contrary, the animals most efficient as heifers were also more efficient and more resilient in economic terms (showing less variability in herd profit before grants) as cows.



Recommendations

Efficiency at young ages and resilience as adult suckler cows could have a trade-off, but it depends on the output evaluated. This trade-off should be assessed under a herd evaluation approach to detect which animals are more suitable for a specific livestock system.

Problem & Solution

In extensive cattle production systems, farmers often lack actual information on the status of their herd regarding resilience and/or efficiency. Novel technologies such as drones and cow tracking sensors could help farmers by providing precise information about their cattle.



Outcome

Wageningen Research has shown that drone imagery combined with deep learning techniques and 3D analysis can provide useful information. Detecting cows in the field reached accuracies >95%, with detection in fields without shade reaching higher accuracies (99.9%) than in shaded fields (97.3%). Identifying Holstein cows reached an accuracy of 91% for cows in a small herd, having a distinct coat pattern. The characterization of cows into standing, grazing, and lying caused no difficulties between grazing and lying, but separating grazing and standing was challenging. The current machine learning algorithms can predict lying and standing with an accuracy between 0.82 and 0.95. Eating and rumination behaviour can be predicted with an accuracy between 0.85 to 0.99 with human annotation as reference.



Practical recommendations

The results suggest that camera-mounted drones along with cow tracker sensor technology implemented by Noldus are promising new tools for monitoring cattle activity and behaviour traits that can be used as input for proxies for cattle resilience and efficiency assessment in extensive rearing systems. Yet:

• The current cow tracker sensor and drones are limited in use due to battery life and weather conditions

• Drone information is often not real-time

• Sensor data is real-time but limited to a set of cows

• Identifying individual cows is only possible if they have a unique coat pattern

Problem & solution

Many multi-actor projects struggle with effective stakeholder engagement during the lifetime of the projects. Joining forces with projects that address similar challenges from different perspectives could increase the outreach and maximise the impacts of the project results.



Outcomes

Collaboration was encouraged with a free Common Dissemination Booster service initiated by the EU in which projects could form clusters to work together to improve their communication and dissemination activities. The GenTORE project led the cluster of 6 relevant projects (IMAGE, SAPHIR, Feed-a-Gene, SmartCow, GplusE) called “Fitter Livestock Farming”. The cluster produced a joint policy brief, flyers, and banners and held events through 2018 and 2019, including the joint workshop with the Animal Task Force (ATF) in Brussels on 6 November 2019. Because of this event, the outreach towards policymakers and other interest groups/representatives of livestock production increased with 23% and 38%.



Recommendations

• Relevant and complementary projects should join forces to improve communication and dissemination to common stakeholders, improving the impact of project results.

• Clusters should establish a common identity to increase the reach to different stakeholders by using media tools such as flyers, brochures, policy briefs, practice abstracts and social media.

• Joint events with academic and non-academic sessions should be organised to maximise the impact and reduce the financial burden on each project.

Problem and solution

Resilience is a cow’s capacity to respond to perturbations, safeguarding its ability to contribute genes to the next generation. However, the lack of a practical definition hinders the collection of phenotypes for resilience in commercial populations. Therefore, a practical definition of resilience that reflects a cow’s ability to re-calve is needed.



Outcomes

A tool to compute a cow’s lifetime resilience score, based on readily available farm data. This tool will facilitate phenotyping of resilience on a large scale, and in commercial populations.



Recommendations

A farmer wants a cow that is easy to take care of, therefore:

• The practical definition of resilience looks explicitly at the cow’s future (re)productive performance

• The practical definition depends on several traits, of which some strongly correlate with re-calving ability and others with health and production

• The practical definition is based on a scoring system that includes 5 categories - calvings, age at first calving/calving interval, inseminations, curative treatment days and 305-day milk yield, all with their own weighting

• plus points for each calving, fewer inseminations and higher milk production compared to peers

• minus points for increased number of inseminations, curative treatment days and if her milk yield is lower compared to peers.

• The practical resilience score can be assessed using readily available data

Problem & Solution

The time that dairy cows spend in the herd is mainly determined by management decisions. The identification of risk factors relevant to culling is of great importance in determining the length of productive life of dairy cattle. The latter is defined as the time between the date of first calving and the culling date.

To identify risk factors for culling, we investigated the influence of production, reproduction, morphology, and health traits on length of productive life using survival analysis.



Outcomes

• Health status, production, reproduction, and conformation traits affect the length of productive life.

• Milk yield plays a crucial role in culling decisions as farmers are more prone to cull cows with low milk production, even if they have other good characteristics.

• Cows selected as suitable for alpine pastures have a decreased culling risk in Swiss production systems.

• Among the most common dairy cattle breeds in Switzerland, Holstein Friesian cows exhibit the highest culling risk.



Recommendations

In the long term, resilient cows can maintain their normal productive life when exposed to environmental perturbations, reducing the risk to be culled.

So, length of productive life can be proposed as an indicator of long-term resilience.

Problem and solution

Breeding schemes can help farmers develop more resilient and efficient farming systems, but extensive cattle farmers show limited commitment to these schemes. To match breeding goals with farmers' objectives, a better understanding is needed of the traits they consider important and those they actually measure.



Outcomes

A survey of suckler cattle farming systems in the Central Spanish Pyrenees asked farmers to score the relative importance of a set of traits related to performance, if they recorded them and if they provided data to breeder associations. Results showed that despite 85% of the farmers belonged to breeder associations, only 21% provided data.

Despite these low recording rates, most traits were regarded as important for cow efficiency.

Controversies were found among traits rated by the farmers as important, traits they recorded, and traits included in breeding schemes. Differences were associated with difficulty of trait recording and lack of immediate profit, particularly for farmers selling weaners.

This survey shows a gap between farmers’ perceptions of relevant breeding traits for resilience and efficiency and their willingness to record data within their associations’ breeding schemes.



Recommendations

Farmers need encouragement to actively join and participate in breeding programmes by:

• designing participative breeding programmes

• developing easy phenotype recording protocols

• facilitate easy transfer of collected data and data analysis

Problem and solution:

The need for resilient animal production systems is urgent. We need tailored solutions to optimize the balance between resilience and efficiency for different livestock systems.

Cattle need to efficiently convert resources into products. However, the impact of conflicts between resilience and efficiency on improvement of these traits is unclear.



Outcomes:

A GenTORE stakeholders (i.e., farmers, vets, breeders, and consultants) survey indicated great confidence in genetic improvement technologies promoting resilient and efficient production, although our selection criteria may not be optimal yet. Stakeholder preference for the inclusion of specific traits in cattle breeding goals are similar across Europe, but perceptions show conflicting results between traits that support resilient production and traits that support efficient production.



Recommendations:

Important actions for efficient production for dairy include using genetic improvement tools and culling least adapted animals. For efficient beef production these are seeking technical advice, using genetic improvement tools, culling least adapted animals, and utilising group calving patterns are important.

Breeding and herd book organisations can provide easy-to-use apps and training that utilise GenTORE on-farm decision support tools so farmers can select better resilience and efficiency animals.

Problem & Solution

Candidate cows for culling are difficult to identify, especially in large herds, as a whole range of characteristics influence the future productive potential of cows.

A decision support tool is required which ranks cows on expected lifetime productivity. This productivity can be in monetary or carbon units. The ranking index must exploit the range of different data sources available and so useable in real-time



Outcomes

A GenTORE web-based real-time decision support tool to rank all cows in a herd on expected remaining lifetime profitability, thus facilitating a more accurate and rapid detection of culling candidates



Recommendations

• All data on individual animals should be put on national databases to ensure the most accurate ranking; information like milk test-day yield and calving dates automatically exist in the database, as do inseminations if undertaken by a service provider.

• An algorithm should be applied to the national database, drawing on all the relevant cow-level data within the database, allowing all cows in the herd to be ranked by expected remaining lifetime profitability.

• Decide which cows to cull but ensure first you have sufficient replacements to enable such a culling rate .

• Examine all cows estimated to have a negative remaining lifetime profit to confirm the predictions. In case of uncertainty, phenotypic data used in the prediction can be further analyzed to identify the reasoning behind the model outcomes.