Sow replacement : Reduction of state space in Dynamic Programming model and evaluation of benefit from using the model

Only few theoretical efforts have been made in order to support decisions of culling/replacement in sow herds, despite obvious possibilities for affecting productivity. In this paper different culling strategies are compared: 1) culling according to parity (CP) and 2) culling according to a policy determined by a Dynamic Programming (DP) model using litter sizes of each sow (CS). The latter replacement model utilizes a hierarchic Markov process and maximizes average litter size. A statistical model is used describing expected litter size as depending on a sow effect, an autoregressive random error, and fixed effects. The model is used directly in the state description of the DP model. Only expectations of sow dependent model parameters are included in the state space, reducing the state space significantly. Improvement in average litter size was estimated with varying model parameters. The method proved successful in terms of computing ressources. Despite full use of information, only minor benefits could be obtained from CS, and CP gave comparable improvements, indicating that pig producers need only take parity into account. If culling strategies should be used, it is imperative to estimate herd specific model parameters. Future efforts should be focused on improving methods for these estimations.