Prediction of 24-hour milk yield and composition in dairy cows from a single part-day yield and sample

Abstract The objective was to evaluate the accuracy of predicting 24-hour milk yield and composition from a single morning (AM) or evening (PM) milk weight and composition. A calibration dataset of 37,481 test-day records with both AM and PM yields and composition was used to generate the prediction equations; equations were validated using 4,644 test-day records. Prediction models were developed within stage of lactation and parity while accounting for the inter-milking time interval. The mean correlation between the predicted 24-hour yields and composition of milk, fat and protein and the respective actual values was 0.97 when based on just an AM milk yield and composition with a mean correlation of 0.95 when based on just a PM milk yield and composition. The regression of predicted 24-hour yield and composition on the respective actual values varied from 0.97 to 1.01 with the exception of 24-hour fat percentage predicted from a PM sample (1.06). A single AM sample is useful to predict 24-hour milk yield and composition when the milking interval is known.