Infrared thermography is a tool to investigate the welfare of cattle. This study aimed to identify a sampling strategy for recording infrared thermograms in dairy cows, in order to most efficiently determine biologically relevant changes in the maximum infrared temperature (IRT) of the eyes and coronary band of forelimbs. Thirty-one dairy cows were used for the study. They were assessed with four replicates of thermograms for each of the head and lower forelimb per cow for 6 mostly consecutive days (sessions). The data obtained were subjected to random effects Analysis of Variance which was used to estimate the variance components for this sampling model, using maximum IRT of both eyes; (left + right eye)/2 and both limbs; (left + right coronary band of forelimb)/2 as dependant variables. The variance components were used to calculate least significant differences (LSD) between two theoretical treatment groups under different sampling scenarios. Analysis showed that there was minimal improvement in precision beyond 2 thermograms within a session but there was improvement with increasing the number of sessions from 2 to 3. The LSD of both eyes and both limbs reached a biologically relevant difference (0.4 and 0.9 °C, respectively) at a minimum number of 14 - 16 cows monitored for 2 consecutive thermography sessions, or 10 - 12 cows for 3 sessions. We conclude that no more than 2 replicate IRT measures are required per session but that measuring on 3 consecutive days should be considered, depending on whether time or number of cows used is the primary limitation.
Keywords: Infrared thermography; Least significant difference; Variances; Welfare.