Test of a sensor to estimate grazing and rumination time in dairy cow behaviour at pasture

Use of automated tools on farms is increasing worldwide and there are diverse applications available including optimization of grazing through monitoring rumination and ingestion times. The objective of this study was to evaluate the accuracy of the sensor developed by Medria to estimate grazing (G) and rumination (Ru) on pasture. This sensor includes a tri-axial accelerometer and provides data every 5 minutes. The trial lasted 12 days on 12 grazing cows in the CTA (Belgium), representing 380 h of observations. The data registered by the Medria device (M) were compared with visual observation (VO).
The datasets were compared using Fleiss-κ estimating the concordance of 5-min observations and linear regression analysis was used to estimate the Pearson correlation coefficients (rp) and relative prediction error (RPE). Moderate for G (κ=0.502) and poor agreement for Ru (κ=0.175) were observed. Linear relationship between VO and M was highlighted with rp: 0.793; P<0.001 for G and rp=0.32; P<0.05 for Ru. The relative error prediction was 0.16 and 0.44 for G and Ru respectively. With regards to these results, reliable data about grazing are provided on a daily basis. The reliability of rumination data was poor.