On the Deployment of Sensor-Based Technologies for Monitoring Supplement Intake Behaviours in Grazing Cattle

Author(s)
Simanungkalit, Gamaliel
Hegarty, Roger
Barwick, Jamie
Cowley, Frances
Publication Date
2022-09-07
Abstract
Please contact rune@une.edu.au if you require access to this thesis for the purpose of research or study.
Abstract
<p>Monitoring licking behaviour and time spent licking by grazing cattle offer a means to estimate intake of block supplements. The advancement of sensor-based technologies to measure these makes harnessing information from individual animals possible in an extensive grazing environment. A literature review (Chapter 2) exhaustively discusses the potential of using stationary and wearable (on-animal) sensor-based devices to monitor supplement intake by cattle. Five experimental studies (Chapters 3 – 7) evaluate the effectiveness of ear-tag accelerometers and radio-frequency identification (RFID) integrated into automatic feeders to monitor licking behaviour and time spent licking, thus estimating block supplement intake in grazing beef cattle.</p> <p>The first two studies assessing gravimetric and blood marker-based assessment of supplement intake (Chapters 3 and 4) indicated that:</p> <p>• The total time spent by individual grazing cattle at supplement blocks recorded by an RFID equipped automatic supplement weighing unit (ASW) exhibited a linear relationship with block intake (<i>R<sup>2</sup></i> = 0.93 [Experiment 1] and 0.70 [Experiment 2]; <i>P</i> < 0.001). However, the high mean error of the linear models (RMSE = 34% [Experiment 1] and 42% [Experiment 2]) indicated substantial between- and within-animal variability in block intake.</p> <p>• A linear relationship between the animal's cumulative weighted time spent at the ASW unit and its plasma level of the block intake marker (oxfendazole sulfone) was found (<i>R<sup>2</sup></i> =0.75; <i>P</i> < 0.001).</p> <p>• Using a complex on-ground supplement weighing system over a long period in a large herd was technically prohibitive (Chapter 4). Hence, using an on-animal sensor was considered necessary to monitor time spent licking by individuals at the block supplements.</p> <p>Due to the challenges of using remote weighing platforms in arid Australia (Chapter 4), a pilot study evaluated accelerometers deployed on a neck-collar, or an ear-tag were as alternative supplement intake proxies (Chapter 5)</p> <p>• Both the neck collar and ear-tags were capable of identifying the licking behaviour at a supplement block in Angus steers confined in individual pens. </p> <p>• A behaviour classification model developed using the Random Forest (RF) machine learning algorithm performed well in classifying licking behaviour, with an accuracy ranging between 92% and 98% for neck-collar and between 88% and 98% for ear-tag. </p> <p>• Further research is required to test the ear-tag accelerometer model under actual grazing conditions for predicting licking state (LS) duration or time spent licking.</p> <p>An experiment with a group of cattle in a 900 m<sup>2</sup> yard confirmed that the ear-tag accelerometer performed well in classifying licking behaviour at a supplement block in grazing steers. In this experiment, the block supplement intake was individually accessed through the raceway of an automatic supplement feeder equipped with an RFID system.</p> <p>• Accuracy, kappa coefficient, and F1 scores of the ear-tag accelerometer classification model developed using the Extreme Gradient Boosting (XGB) algorithm for licking behaviour were 93%, 0.88, and 0.88, respectively. The accelerometer model and the RFID system predictions for LS duration were acceptable, with a mean absolute error (MAE) of 22% and 10%, a ratio of root mean square prediction error (RSR) of 0.33 and 0.14 and a modelling efficiency (MEF) of 0.89 and 0.98. </p> <p>A validation study was conducted on Angus (<i>n</i>=7) and Brahman (<i>n</i>=7) heifer groups to test the accelerometer classification model and RFID system to predict an individual’s time spent licking at supplement blocks provided at an ASW unit for each group. </p> <p>• The ear-tag accelerometer behaviour classification model developed using the Support Vector Machine (SVM) algorithm satisfactorily categorised licking behaviour and predicted the time spent licking at the lick-block, with a MAE of 22% and 11%, MEF of 0.81 and 0.94, the concordance correlation coefficient (CCC) of 0.88 and 0.96, and RSR of 0.44 and 0.25, for Angus and Brahman heifers, respectively.</p> <p>• However, the RFID system predictions of time spent licking for both breeds were unacceptable as the low-frequency RFID system (134.2 kHz) was unable to capture the presence of multiple RFID tags accurately.</p> <p>This thesis revealed that an ear-tag accelerometer in conjunction with an RFID system offers the possibility of estimating the lick-block supplement intake of individual grazing cattle by determining the time spent licking. This may apply in remote rangelands where on-ground measurement of block intake was found too difficult (Chapter 4). Ear-tag based supplement intake estimates could be downloaded from the tags at key locations such as water points and transmitted to the farm manager.</p>
Link
Publisher
University of New England
Title
On the Deployment of Sensor-Based Technologies for Monitoring Supplement Intake Behaviours in Grazing Cattle
Type of document
Thesis Doctoral
Entity Type
Publication

Files:

NameSizeformatDescriptionLink