Progress in developing grazing behaviour models from spatio-temporal data

Abstract

Autonomous Spatial Livestock Monitoring (ASLM) systems that will enable livestock producers to monitor the location and movement of their animals in real time has the potential to revolutionise the grazing industry. Technology developers have already produced ear tag devices which allow graziers to remotely locate their animals. As these monitoring systems provide regular location information on the animals being tracked it is possible to use this information to generate measures of movement such as animal speed, turn-angles and other quantitative measures related to spatial activity. These movement metrics will provide the potential for correlation with animal behaviour. One of the key behavioural activities is grazing, as alterations in animal feeding behaviour can be related to the early onset of disease and/or changes in welfare status. In addition, accurate records of grazing behaviour will allow us to quantify spatial variability in pasture utilisation which may support site specific management strategies. Those ASLM systems targeted at commercial applications which are in development are currently not able to distinguish between the different behavioural activities of livestock. The current behavioural models derived from spatio-temporal data only which are reported in the literature are rudimentary and are based on either velocity or diurnal activity. This presentation reports on a project using new temporal sampling strategies (e.g. multipoint interval tracking) and more advanced statistical modelling processes (e.g. dynamic Bayesian network modelling) to develop suitable grazing behaviour models from simple movement data. This project has been supported by Meat and Livestock Australia through the Department of Agriculture Forestry and Fisheries Science and Innovation Awards for Young People.