Intake estimation in dairy cows fed roughage-based diets: An approach based on chewing behaviour measurements

Abstract Chewing behaviour of 23 lactating Swiss Fleckvieh cows was analysed in order to evaluate the predictive potential for quantitative dry matter intake in a roughage-based indoor cattle feeding system. Cows were fed total mixed rations (TMR) based on silages and hay with different concentrate supplements. They were kept in a tie stall enabling individual feed intake measurements. Two measurements were conducted within one month. Chewing behaviour was recorded with RumiWatch ® sensor collars, based on pressure tubes in the collar's noseband. Cows were equipped with collars for 96h per measurement period. First 24h were accounted as adaptation time; data of the subsequent 72h were used for analysis. Data included ruminating, eating (min/day), rumination boli (n per day), chewing frequency and intensity during ruminating (chews/min and chews/bolus), and activity changes (switching between ruminating, eating and idle; n per h). The constancy of parameters within cows across measurement days was tested with linear regression models. A linear mixed-effects model was applied to estimate a regression on measured feed intake. Average feed intake per day across all measurements was 19.7kg dry matter per cow, average eating time was 389min/day and ruminating time was 551min/day. For most of the chewing behaviour variables, factor ‘cow' was significant, while ‘day' was not, indicating a between-animals variance but good consistency of the data within animal. After a stepwise backward procedure in the mixed-effects model, the remaining significant variable was ‘chewing frequency' (chews per minute during rumination). Inclusion of ‘animal' as a random factor resulted in an equation with conditional R 2 =0.7. The model without random factor revealed a very low R 2 . In conclusion, the random factor model allowed estimation of individual changes in feed intake within animal but not across animals. Chewing behaviour measurements proved to have a potential for the detection of relative intake alterations with roughage-based TMR diets but data were not sufficient for quantitative estimations.