Validation of models for predicting milk urea nitrogen concentrations of Holstein cows herd

Abstract

The objective of this study was to validate prediction models for milk urea nitrogen, aiming at evaluating the efficiency utilization of diets of a dairy cows' herd. Data (8,833 observations) from Holstein cows were used, belonging to a commercial herd. Milk yield, body weight, days in milk and parity were recorded on the milk sampling days. From the database, monthly averages were calculated, attempting to study the herd. Model 1 for predicting milk urea nitrogen was developed by Jonker et al. (1998) and models 2 and 3 by Kauffman & St-Pierre (2001). In order to evaluate models, accuracy, precision and robustness of those three models were tested. Lack of accuracy was found for models 1 (mean bias=2.60mg/dL) and 2 (mean bias=-1.95mg/dL), while model 3 was accurate (mean bias=-0.89mg/dL). However, precision among models 1, 2 and 3 did not differ (residual error=3.72, 2.68and 2.64mg/dL, respectively). Models 1 and 2 did not show robustness for days in milk, nor model 1 for fat percentage. Among the 3 models studied, model 3 was the best evaluated when the aim is to estimate milk urea nitrogen concentrations or to predict urinary nitrogen excretion of a Holstein cows’ herd in the field conditions evaluated.