Evaluation of performance models for farm-specific optimization of pig production

Farm-specific optimization of pig production can be supported with a production function mechanistically derived from dynamic growth and feed intake information, hereafter called performance curves. Production-theoretical optimization requires, first, an accurate description of underlying processes and, second, possibilities for calibrations with available on-farm data. The objective is to evaluate accuracy and calibration requirements of seven models: the Gompertz, Monomolecular, Richards and Generalized Michaelis-Menten (GMM) model for growth and the cumulative feed intake and weight (CFIW), Bridges and Giesen models for both growth and feed intake. Evaluation is done with data from four trials with four sexes of a Piétrain x hybrid sow cross: boars, barrows, gilts and GnRH-vaccinated boars. Accuracy was evaluated with Root Mean Squared Errors (RMSE) of predictions versus observations and F-tests for differences in goodness-of-fit to discriminate between models. Calibration possibilities were evaluated through fitting the models through limited data. The Giesen, Bridges and GMM model, describing sigmoidal growth patterns, showed the highest accuracy. In terms of calibration possibilities, the Bridges model slightly outperformed the other. The Giesen model is accurate for describing feed intake of boars, barrows, gilts. GnRH-vaccinated animals showed unstable variance with increasing age of the animal, which could only partially been solved with extensions to the Bridges and Giesen model. The research showed that dynamic growth and feed intake curves can be reasonably estimated from limited on-farm collectable data and generic functional forms, and as such account for farm specificity in production-theoretical economic optimization.