Description
Monitoring livestock provides valuable insights into the spatial distribution, foraging patterns, and animal behaviour, which may lead to the improved management of livestock. This objective of study was to understand what variables were significant in determining where sheep spent the most time in paddocks of native (dominated by Poa spp., Stipa spp., and Hordeum leporinum), and improved (Phalaris aquatica, Festuca spp., and Trifolium subterraneum) pastures (~24 ha-1 in size). Castrated male sheep, wethers, were tracked using GPS collars on a property located in the Monaro region of Southern New South Wales, Australia. Trials were performed over four six-day periods in April, July, and November of 2014, and March in 2015. Thirty collars were randomly placed on 15 sheep in each pasture types. The GPS collars continuously recorded movements over the duration of each trial and were programmed to take five positional fixes over a minute. Data was analysed all together to understand various trends that may have occurred across the year, using random forest models (RFMs). Models were created for improved (IP) and native (NP) paddocks, producing R2 values of 0.97 and 0.94. Regardless of the pasture type, near distance to water (NW) was important in predicting where animals were located within paddocks, despite its statistical insignificance. Sheep spent more time further away from water troughs (at least ~600 m). Additionally, NDVI was another important variable in predicting sheep location for IP and NP (p < 0.01). The data suggests that RFMs are able to predict where animals will likely spend time on a paddock at a large scale, with water troughs and pasture quality being two key drivers.
Included in
Agricultural Science Commons, Agronomy and Crop Sciences Commons, Plant Biology Commons, Plant Pathology Commons, Soil Science Commons, Weed Science Commons
Determining ‘Wether’ Social Behaviour or Pasture Quality Drives Sheep Grazing Patterns using Random Forest Modelling
Monitoring livestock provides valuable insights into the spatial distribution, foraging patterns, and animal behaviour, which may lead to the improved management of livestock. This objective of study was to understand what variables were significant in determining where sheep spent the most time in paddocks of native (dominated by Poa spp., Stipa spp., and Hordeum leporinum), and improved (Phalaris aquatica, Festuca spp., and Trifolium subterraneum) pastures (~24 ha-1 in size). Castrated male sheep, wethers, were tracked using GPS collars on a property located in the Monaro region of Southern New South Wales, Australia. Trials were performed over four six-day periods in April, July, and November of 2014, and March in 2015. Thirty collars were randomly placed on 15 sheep in each pasture types. The GPS collars continuously recorded movements over the duration of each trial and were programmed to take five positional fixes over a minute. Data was analysed all together to understand various trends that may have occurred across the year, using random forest models (RFMs). Models were created for improved (IP) and native (NP) paddocks, producing R2 values of 0.97 and 0.94. Regardless of the pasture type, near distance to water (NW) was important in predicting where animals were located within paddocks, despite its statistical insignificance. Sheep spent more time further away from water troughs (at least ~600 m). Additionally, NDVI was another important variable in predicting sheep location for IP and NP (p < 0.01). The data suggests that RFMs are able to predict where animals will likely spend time on a paddock at a large scale, with water troughs and pasture quality being two key drivers.
