Tool for assessing the intervention effect on milk production in an evolutionary operation setup.

Modern dairy herds resemble factories. Cows, organized in production units, are manufacturingmilk from many components (e.g. concentrates, silage). However, both production units andcomponents can greatly differ between each other. Therefore, production optimization based ongeneral recommendations might be inefficient. Instead, as in manufacturing industry, decisionsupport could be based on systematic experimentation within ongoing production system. Thisconcept, known as Evolutionary Operations (EVOP), is based on small changes to the productionsystem. However, a challenge here is lack of a tool which would allow a farmer to assess how smallchanges, for example in feeding, influence productivity. The objective of this study was to constructa multivariate dynamic linear model (DLM) to assess the intervention effect on milk production.The DLM was built to account for intervention at individual and herd level. It consisted of anobservation and a system equation. The observation equation links the observations to parametersdescribing the herd (lactation curve), individual cows and an intervention effect. The systemequation expresses how the parameters may change over time. The lactation curve was modeledby two linear expressions and was parameterized using: milk yield 60 days after calving, slope overthe first 60 days in milk and slope after 60 days in milk. The variance components of the DLM wereestimated using a maximum likelihood method. The application of the model was demonstratedon a field experiment in a commercial herd with 4 automatic milking systems (AMS). The herdwas split into 2 groups based on the AMS. The experiment relied on two steps. The first step wasto reduce the feed energy given to cows in the AMS and instead supply the feed energy to the cowsat the feed bunk. The second step was to reduce the feed energy given in two of the four AMS. TheDLM presented here was successful in providing estimates of the effects on milk yield of change infeed energy gi
Modern dairy herds resemble factories. Cows, organized in production units, are manufacturing milk from many components (e.g. concentrates, silage). However, both production units and components can greatly differ between each other. Therefore, production optimization based on general recommendations might be inefficient. Instead, as in manufacturing industry, decision support could be based on systematic experimentation within ongoing production system. This concept, known as Evolutionary Operations (EVOP), is based on small changes to the production system. However, a challenge here is lack of a tool which would allow a farmer to assess how small changes, for example in feeding, influence productivity. The objective of this study was to construct a multivariate dynamic linear model (DLM) to assess the intervention effect on milk production. The DLM was built to account for intervention at individual and herd level. It consisted of an observation and a system equation. The observation equation links the observations to parameters describing the herd (lactation curve), individual cows and an intervention effect. The system equation expresses how the parameters may change over time. The lactation curve was modeled by two linear expressions and was parameterized using: milk yield 60 days after calving, slope over the first 60 days in milk and slope after 60 days in milk. The variance components of the DLM were estimated using a maximum likelihood method. The application of the model was demonstrated on a field experiment in a commercial herd with 4 automatic milking systems (AMS). The herd was split into 2 groups based on the AMS. The experiment relied on two steps. The first step was to reduce the feed energy given to cows in the AMS and instead supply the feed energy to the cows at the feed bunk. The second step was to reduce the feed energy given in two of the four AMS. The DLM presented here was successful in providing estimates of the effects on milk yield of cha