Automated Monitoring of Panting for Feedlot Cattle: Sensor System Accuracy and Individual Variability

Abstract

Heat stress causes significant economic losses by reducing the productivity and welfare of cattle whilst requiring a significant investment in resource for amelioration. Panting score (PS) is considered a robust indicator of cattle heat stress; however, individualised visual monitoring is impractical. Thermal index-based monitoring and mitigation decisions are applied at the herd level, but they have limited application for the individual animal. As such, an automated system to monitor the real-time animal response to heat stress is required for strategic mitigation. Our objectives were to validate an accelerometer-based ear tag sensor to monitor cattle panting and to determine individual variability in heat stress responses with reference to thermal indices. Two experiments were conducted: Experiment 1 validated the sensors, and Experiment 2 determined individual variability comparing sensor data against thermal indices. Ear tag sensors were fitted at feedlot entry to continuously monitor the behaviour of 100 steers of mixed breed in Experiment 1 and 200 steers and heifers of mixed breed in Experiment 2. Sensor-derived ‘heavy breathing’ was validated against visually observed PS. Sensor-derived behaviour bouts were analysed as ‘raw’, and single behaviour states were also converted to the preceding bout of ≥2 min, which was referred to as ‘fill’ data for the validation study. Our results demonstrate the sensors’ ability to accurately monitor panting in feedlot cattle. Sensor-recorded ‘heavy breathing’ duration per animal was highly correlated to observed panting duration for both raw (r = 0.89) and fill (r = 0.90) data; however, the concordance correlation co-efficient was lower for raw (0.45) as compared with fill (0.76). Predicted agreement for raw data were 75%, 45%, and 68% and predicted agreement for fill data were 65%, 54%, and 83% for PS0, PS1, and PS2, respectively. Sensitivity for raw data were 39%, 37%, and 45% and for fill data, they were 59%, 54% and 82% for all PS data, PS1 and PS2, respectively. Specificity and positive predictive values for both raw (77% and 79%, respectively) and fill (65% and 77%, respectively) data show the probability of reporting false positives by sensors to be low. Experiment 2 revealed that the duration of panting increased from 0800 to 1700 h alongside changes in thermal indices with significant differences between and within breed and coat colour categories of cattle, suggesting that grouping and allocating heat amelioration measures by breed and coat colour can be effective in commercial feedlots. However, there was high variability (CV > 80%) in the duration of panting between individuals within the same breed and same coat colour, revealing the potential for strategic management at an individual level, and with the same data, genetic selection for heat resilience.

Recent Advances on Early Detection of Heat Strain in Dairy Cows Using Animal-Based Indicators: A Review

“…Currently, few wearable devices on the market can measure BST and RR. Some commercially available accelerometer-based ear tags and collars can measure “heavy breathing” [ 8 , 124 ]. In practical farms, there is a need for measuring comprehensive respiratory dynamics including both frequency and depth.…”

“…It has been reported that both CBT and RR remained high even when THI was low at night [ 53 ]. In addition, there are large differences in the sensitivity to the thermal environment among individual animals [ 1 , 27 ], even within the same subgroup [ 124 ]. A large amount of energy used for cooling measures may be wasted on low-risk animals.…”

“…With the accumulation of heat load in dairy cattle (manifested as a rising RR), cows first exhibited drooling, following by open mouth and protruding tongue [ 131 ]. PS has been developed and refined as a numerical scale to grade the obvious respiratory dynamics and behaviors [ 59 , 88 , 123 , 124 ]. A recent study utilized a score consisting of RR and various panting characteristics to assess heat strain in grazing dairy cows [ 133 ].…”

“…Their results show that the proportion of cows breathing heavily was consistent with the change of mean CBT in both dry and lactating cows. An accelerometer-based ear tag from the same manufacturer was also validated in steers [ 124 ]. The visual measurement of PS and sensor-recorded “heavy breathing” were compared.…”

“…Therefore, the original output of these devices is probably much closer to RR, and “heavy breathing” can be seen as the extended results of RR. As proposed by both Bar et al [ 8 ] and Islam et al [ 124 ], these accelerometer-based sensors have difficulties in detecting slight panting (PS1) and only give a binary result. Although the detection of panting or not is enough for the start of heat abatement protocol, more sophisticated classification would be more informative for customized interventions.…”