An adaptive decision-making aid for dairy cow replacement

A decision aid is developed for the retention or replacement of individual dairy cows. An important determining factor is the cow’s milk-yield potential. However, the potential is difficult to assess from recorded yields in previous lactations because temporary environmental factors mask the underlying yield potential. An optimal replacement strategy can be found using stochastic dynamic programming. In this paper the approach, is extended to allow for the Bayesian re-evaluation of yield potential on the basis of previous recorded yields. The sensitivity of the replacement decision to variance estimates determined for a sample of commercial dairy herds in England and Wales is tested in an adaptive dynamic programming model. The paper concludes by demonstrating the way in which the model results could be used for helping to identify which cows should be called without replacement if the expected milk production of the herd is above the quota limit.