The Journal of Agricultural Science


Evaluation of Cornell Net Carbohydrate and Protein System predictions of milk production, intake and liveweight change of grazing dairy cows fed contrast silages 1

A. V. CHAVES a1a2a3a4c1, I. M. BROOKES a2, G. C. WAGHORN a3a4, S. L. WOODWARD a3 and J. L. BURKE a2a4
a1 Agriculture and Agri-Food Canada, Lethbridge Research Centre, 5403-1 Avenue South, Box 3000, Lethbridge, AB, T1J4B1, Canada
a2 Massey University, Private Bag 11222, Palmerston North, New Zealand
a3 Dexcel Limited, Private Bag 3221, Hamilton, New Zealand
a4 AgResearch Limited, Grasslands Research Centre, Private Bag 11008, Palmerston North, New Zealand

Article author query
chaves av   [PubMed][Google Scholar] 
brookes im   [PubMed][Google Scholar] 
waghorn gc   [PubMed][Google Scholar] 
woodward sl   [PubMed][Google Scholar] 
burke jl   [PubMed][Google Scholar] 


The importance of mechanistic models for ration balancing with forages is indicated and physical limitations to intake emphasized, because these limit energy and nutrient supply to cows grazing forages, especially grass. Ration-balancing models using fresh or ensiled forages to complement pasture will need to accommodate intake limitations due to rumen fill, clearance, chewing or other criteria. The potential of the Cornell Net Carbohydrate and Protein System (CNCPS) model to predict milk production from diets based on pasture and forage supplements was tested using data from two experiments. Data were obtained from studies in which pasture was complemented with contrasting silages including maize, pasture, sulla, lotus and forage mixtures, comprising 0·30–0·40 of dry matter intake (DMI). Twelve diets were used in the evaluation. DMI, liveweight (LW), days in milk, and diet composition were determined during the trials and used as inputs in the model. Across all diets, a significant relationship existed between predicted and actual values for DMI (R2=0·58), milk yield (R2=0·59) and LW change (R2=0·51), but there were still large unexplained sources of variation. No significant mean bias was observed for any of the variables, but the slope of residual differences against predicted values was significantly different from zero for milk yield, LW change and for DMI (P<0·06). The results indicate a satisfactory prediction of milk production when cows are neither gaining nor losing weight, but that a systematic bias exists probably because of the failure of the CNCPS model to account for energy and nutrient partitioning.

(Published Online February 10 2006)
(Received October 21 2005)

c1 To whom all correspondence should be addressed. Email:


1 Part of these data was previously published in the Proceedings of the New Zealand Society of Animal Production (2003) 63, 91–95.