Abstract
This study was conducted to evaluate and compare three rumen models with observed experimental data covering a wide range of feeding situations. Molly (MOL), developed by Baldwin in 1987, the Cornell Net Carbohydrate and Protein System (CNCPS), developed in 1992, and the model developed by Lescoat and Sauvant (LES) in 1995 were evaluated. These mechanistic models were compared on their ability to predict various rumen parameters and digestive characteristics. A database of 47 references (194 treatments) was built with animal characteristics, detailed rations and in vivo measurements on dairy cattle. A feedstuff library with all input parameters needed to evaluate the models was created for 73 feedstuff. Thus, the comparative simulations were based on identical and consistent feed inputs. Global regressions and general linear models within experiments were used to determine the ability of the models to predict global and within experiment variation. Evaluations involved four parameters: coefficient of determination, residual standard deviation, slope of the regression and mean deviation to the bisector ( Y= X). Results underlined the fairly good capacity of the model of Lescoat and Sauvant to predict starch digestion in the rumen, with a residual standard deviation of 0.06 kg/kg starch and slope of 0.70. The duodenal flow of microbial N was best predicted by the CNCPS with a residual standard deviation of 28.6 g/day and slope of 0.91. Rumen pH was best predicted by LES with a residual standard deviation of 0.10 and slope of 0.90. Alternatively, the models did not accurately predict fiber digestion in the rumen or volatile fatty acids concentrations. The study demonstrates the strengths and weaknesses of the models. Future improvements of rumen modeling can be considered by pooling the advantages of each model.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have