Conference PaperPDF Available

A robot amongst the herd: A pilot investigation regarding the behavioural response of dairy cows.

Centre for Carbon,
Water and Food,
University of Sydney,
Camden, NSW
26 – 27
Edited by
L Ingram, GM Cronin and LM Sutton
Centre for Carbon, Water and Food,
The University of Sydney, Camden, NSW
26 – 27 September 2013
Published by the Faculty of Agriculture and Environment, and the
Faculty of Veterinary Science, The University of Sydney
ISBN 978-1-74210-324-2 (PDF version)
Edited by: L Ingram, GM Cronin and LM Sutton
Edition: 4th
Cover photographs: LM Sutton & GM Cronin
Hyperspectral remote sensing technology as a tool for supporting precision
management of dairy pasture
SR Aarons, K Kawamura and E Perry
Virtual fencing A corrective or substantive paradigm changer for
managing animal dominated landscapes
DM Anderson
Virtual fencing: A dairy industry dream
LC Anderson
Spatial variability in animals, soil and nitrogen
K Betteridge
Digital homestead: Delivering end user value from real-time on-farm
G Bishop-Hurley, D Henry, L Overs, L Gonzalez, A Anderson, P
Pearce and C Fookes
Measuring resource use in sheep with proximity logger technology
J Broster and R Doyle
A robot amongst the herd: A pilot investigation regarding the behavioural
response of dairy cows
CEF Clark, SC Garcia, KL Kerrisk, JP Underwood, JI Nieto, MS
Calleija, S Sukkarieh and GM Cronin
Detection of pre-lambing and lambing behaviour with the use of GNSS
SJ Dickson, M Trotter and R Dobos
Classifying cattle ingestive behaviour 1. Sound wave analysis
R Dobos, M Trotter and H Oddy
Classifying cattle ingestive behaviour 2. Accelerometer analysis
R Dobos, M Trotter and H Oddy
Infrared thermography for animal health and welfare monitoring: Where to
from here?
S Dowling, M Stewart, J Webster, A Schaefer and T Landgraf
Examining the potential for virtual fencing in Merino sheep
Z Economou, M Trotter, R Dobos
The importance of spatial variability in pasture species in understanding
landscape management
C Edwards, M Trotter and Z Economou
GPS technology and its application for improved reproductive management
in extensive sheep systems
ES Fogarty, JK Manning, MG Trotter, DA Schneider, RD Bush and
GM Cronin
Methods to process large datasets from cattle behavioural monitoring
collars, remote weighing stations and infrared cameras
L Gonzalez, G Bishop-Hurley and D Henry
Pasture Base Ireland – National grassland database
V Griffith, A Geoghegan, M O’Donovan, B O’Brien and
L Shalloo
Ranking paddock performance using grazing events and milk yield data
J Haultain, I Yule, AJ Romera, B Dela Rue, D Clark, C Glassey and
J Jago
Attachment accuracy of a robotic rotary and investigation of two
management strategies for incomplete milked quarters
R Kolbach, KL Kerrisk, SC Garcia and NK Dhand
Effects of bail activation sequence and feed availability on cow traffic and
milk harvesting capacity in a robotic rotary dairy
R Kolbach, KL Kerrisk, SC Garcia and NK Dhand
The effect of premilking with a teat cup-like device, on attachment accuracy
and milk removal on a robotic rotary
R Kolbach, KL Kerrisk and SC Garcia
UNE SMART Farm: Showcasing the value of broadband connectivity in
livestock farming
D Lamb & M Trotter
Effect of milking permission refusal on cow traffic in a pasture-based
automatic milking system
NA Lyons, KL Kerrisk and SC Garcia
Manipulating pasture allocation to incentivate cow traffic in a pasture-based
automatic milking system
NA Lyons, KL Kerrisk and SC Garcia
Using GPS technology to quantify the behavioural responses of sheep
during simulated dog predation events
JK Manning, ES Fogarty, MG Trotter, DA Schneider, RD Bush and
GM Cronin
Optimum-N: The development of a variable rate nitrogen application
system for grazed pastures
JJ Roberts, A Werner, R Roten, K Irie, M Hagedorn, J Fourie, S
Woodward, I Woodhead, I Vogeler, KCameron, G Edwards, H Di,
I Yule, DW Lamb and MG Trotter
Dispersal of cattle as available feed declines determined by GNSS
JJ Roberts, MG Trotter, DW Lamb, DA Schneider, G Hinch, G
Falzon, R Dobos
Feed management strategy in a fully automated pasture-based robotic
milking system
Victoria E Scott, Peter C Thomson, Kendra L Kerrisk and Sergio C
Using RFID technology to collect sheep liveweight data on-farm
LM Sutton, LJ Gant, IK Gant, M Smith, GM Cronin and RD Bush
Evaluation of infrared thermography for predicting ovulation time in dairy
S Talukder, KL Kerrisk, L Ingenhoff, PC Thomson, SC Garcia and P
Accuracy of active optical sensors in estimating pasture biomass
MG Trotter and DW Lamb
A robot amongst the herd: remote detection and tracking of cows
JP Underwood, M Calleija, J Nieto, S Sukkarieh, CEF Clark, S C
Garcia, KL Kerrisk and GM Cronin
Using proximity loggers to measure contact structures to determine
potential modes of pathogen transmission in beef cattle
A Wildridge, A Smith, J Broster, G Lammers and J Heller
The effect of temperament on the behaviour and productivity of beef heifers
grazed extensively on pasture
LR Williams, RD Bush, RJ Kilgour, MG Trotter and GM Cronin
Sharon R. Aarons1, Kensuke Kawamura2 and Eileen Perry3
1DEPI Ellinbank, 1301 Hazeldean Road, Ellinbank Victoria 3820, Australia,, Email
2Department of Development and Technology, Graduate School for International Development and
Cooperation, Hiroshima University, 1-5-1 Kagamiyama, Higashi-hiroshima, Japan,, Email
3DEPI Bendigo, Epsom, VIC 3551, Australia,, Email
Precision fertiliser management is particularly important in grazed dairy systems due to the
heterogeneous distribution of nutrients by dairy cows at within and between paddock scales
(Aarons et al. 2013, Gourley et al. 2007). We trialed hyperspectral sensing techniques developed
in New Zealand dairy systems (Kawamura et al. 2009, Sanches et al. 2013) to estimate pasture
biomass and nutrient contents on an irrigated dairy pasture in northern Victoria.
Canopy reflectance spectra (350 to 2500 nm) and plant samples were collected from 40 grid
locations in a paddock (red dots in Figure 1a) on a commercial dairy farm on 4th June 2013,
using an ASD FieldSpec as well as the CAPP (canopy pasture probe) system, to allow collection
of spectra irrespective of cloud cover (Sanches et al. 2013). Spectral data were analysed using
algorithms and calibration models developed previously (Kawamura et al. 2009), based on
partial least squares (PLS) regression of first derivative reflectance (FDR) spectra and continuum
removed spectra. Spatial distribution maps of pasture biomass and concentrations of N, P and K
were generated from the predicted values for the 40 grid locations using variogram models and
ordinary kriging techniques (R software ver. 3.0.1 and ‘automap’ package ver.1.0-12).
Figure 1. Maps showing (a) grid locations used for kriging of predicted values of
(b) pasture biomass, (c) pasture N, (d) P and (e) K concentrations.
a b c
The ranges (minima, maxima) of predicted herbage biomass (1941, 4287 kg/ha), and herbage N
(0.62, 1.99%), P (0.34, 0.5%) and K (1.93, 4.02%) concentrations were smaller than previously
reported for New Zealand pastures, with smaller coefficients of variation (< 19%) due most
likely to the uniformity of this paddock (Kawamura et al. 2009). Predicted mean biomass (2790
kg/ha), and N, P and K concentrations (1.6, 0.43, 2.93%, respectively) were typical of actual
irrigated dairy pastures in Victoria (Jacobs and Rigby 1999).
Kriged data showed an accumulation of nutrients towards the irrigation outlet end of the
paddock, possibly due to movement of nutrients with irrigation water from the inlet end.
Analytical data from pasture samples collected at the grid locations where spectral data were
collected will be used to validate the predicted data.
Aarons SR, Gourley CJP, Hall M, White N Between and within paddock soil nutrient
variability and pasture production gradients in grazed dairy pastures (submitted).
Gourley CJP, Powell JM, Dougherty WJ, Weaver DM (2007) Aust. J. Exp. Agric. 47,
Jacobs JL, Rigby SE (1999) Minerals in dairy pastures in Victoria. Department of Natural
Resources and Environment, Melbourne, Victoria.
Kawamura K, Betteridge K, Sanches ID, Tuohy MP, Costall D, Inoue Y (2009) NZJ. Ag. Res.,
52, 417-434.
Sanches ID, Tuohy MP, Hedley MJ, Mackay AD (2013) Int. J. Rem. Sen., 34, 276-302.
Dean M. Anderson1
1The United States Department of Agriculture, Agricultural Research Service, Jornada Experimental
Range (USDA-ARS, JER), Las Cruces, New Mexico, USA,, Email
Grazing and browsing animals play a pivotal role in converting plant tissue into edible as well as
non-edible products and services that support man’s continually changing life styles. With
domestication came the task of providing animals an adequate plane of nutrition for growth and
reproduction while simultaneously managing vegetation for sustainable production. Attempting
to resolve these two seemingly opposing management goals has occupied much of range animal
ecology research to date. Today’s demands for multiple goods and services from rangelands
places even more of a focus on bringing about optimum livestock production with the smallest
“hoof print” possible. Virtual fencing (VF) is an evolving methodology of animal control that
requires at least a rudimentary understanding of the atmospheric sciences, soil science, plant
science, animal science, the global navigation satellite system (GNSS), geographic information
systems (GIS), ethology and 21st century electronics. Melding these diverse disciplines together
to produce a methodology for controlling free-ranging animals stands out as holding much
promise for managing animal dominated landscapes. When VF becomes a commercial reality the
use of manual labor to control livestock will largely be replaced with cognitive labor that will
result in prescription based livestock management that is robust, accurate, precise and flexible.
By managing the spatio-temporal aspects of foraging in real-time controlled utilization of
vegetation will be possible with prescription like precision. The end result will be ecosystem
management capable of providing optimum plant as well as realistic growing conditions for
animals without detracting from the additional demands being placed on rangelands today for
goods and services beyond just providing nutrition for free-ranging livestock and wildlife. The
methodology of VF should not be viewed as simply a new tool to carry on current management
but rather it offers the possibility of providing the basis for a new paradigm shift in managing
stocking density in real-time. This approach to livestock management can be thought of as
“virtual shepherding” for the 21st century.
Lindsay C Anderson1
1Kings Vista Jerseys, Athlone, Victoria, 3818, Australia Email
The majority of pasture based Australian dairy farmers spend at least two hours per day
fetching cows and setting up fences; for many this task could take up to four hours. In a basic
form, Virtual Fencing would allow us to send a signal so that the cows arrive at the cow shed
when you do, representing a large labor saving. Using this system, the herd would be
travelling at a slower pace, not harassed by people on motorbikes resulting in a reduction of
stress and lameness. Furthermore the cows could be brought home in multiple groups per
milking, avoiding long standing times on concrete yard. This may lead to decreased track
infrastructure, motorbike use and Occupational Health and Safety liabilities.
With permanent fencing the paddock size will invariantly not be right, just a bit too small or
big leading to over/under grazing that ultimately leads to loss of pasture utilization. Virtual
fencing would allow allocation of pasture accurately to meet the herd’s requirements and do
so remotely.
The benefits of using Virtual Fencing could also include more advance features such as:
Knowing where every animal is and alert the farmer when cows are not where they
should be
o Sprayed paddocks, fertilized paddocks particularly nitrogen areas etc.
Monitor animal movements
o Been stationary for too long leading to problems- milk fever, calving
o Too much movement- wild dog attack, cycling
Grazing pattern and if any link to soil types, fertility, grass species and wetness
Drafting animals into different paddocks as the herd walks down the track
Collecting data in the farm with additional sensors
While the above concepts are just some applications for Conventional Milking Systems
(CMS), Automatic Milking Systems (AMS) would also benefit. With voluntary AMS the last
cows are still required to be fetched. Furthermore some cows are milked too often for their
level of production whilst other cows not enough, thereby wasting valuable robot time and
utilization. Bad weather can cause the whole herd to come home at once which would lead to
long waiting period before milking. Virtual Fencing has the potential to overcome this. With
batch mode AMS, Virtual Fencing eliminates someone fetching cows which would enable
more batches of cows to be handed automatically. In summary Virtual Fencing offers labor
saving, productivity gains, improved utilization by integrating with other farming techniques
and lowering business risk.
Keith Betteridge1
1AgResearch Grassland, Palmerston North, New Zealand,
Many livestock farmers continue to manage their paddocks as though each was the same and
therefore requires the same level of inputs and stock management. However, increasingly
more farmers now recognise the productive and environmental value of applying variable
management to each of the land management units on the farm (Mackay et al., 2001). Despite
Lawrence (2013) showing substantial savings in fertiliser costs were achieved through soil
sampling each paddock on his dairy farm and applying nutrient requirements accordingly,
uptake of this simple, cost-effective strategy for managing spatial variability in soil fertility is
slow. Uptake of precision livestock farming technologies requires good knowledge of
variability in soil type and soil fertility. Soil maps of NZ are available at 1:50,000 scale,
though accuracy at the paddock scale can be poor since maps were created by interpolation
from lower resolution maps. However, the SUBS programme (Mackay et al., 2001) has been
developed for farmers to describe their own soils, in conjunction with national soil maps. EM
mapping of soils is now an accepted practice, at least on lowland.
Within paddocks, variable soil fertility is also created by stock use of gateways, shelter and
troughs, and by camping. High losses of P and N can occur from these sites. With the
imminent creation of regional water and air quality standards, NZ farmers may soon be faced
with caps on emissions and, therefore, productivity. Thus ‘proven’ mitigation strategies are
urgently required. In hill country, cattle camp on flat land usually at low elevation. GPS and
urine sensors were used to show that up to 50% of all cattle urination events in a paddock
occur in the campsites that occupy just 5-10% of the area (Betteridge et al., 2010). To
quantify the N deposited in urine patches a new sensor has been developed. Glued over the
vulva, the volume and N concentration of all excreted urine is estimated by this sensor. Data
are downloaded by telemetry and the position of the urine patch can be determined either by
triangulation using Zigbee, or by GPS on the cow. Diurnal patterns in urinary N loads will be
discussed, as will the change in N load of each urination event when cows are shifted to new
pasture. The data suggest that cows held for 20+ h on a stand-off area during winter are
unlikely to deposit much urinary N during the 2-3 h each day they are on the crop or pasture.
Paddock-scale prediction of annual N leaching using the lognormal distribution patterns of
urinary N content and volume, indicate that leaching is higher than if the same data were
input to the model as average values of N% and volume.
Betteridge K, Hoogendoorn C, Costall D, Carter M, Griffiths W (2010) Comp. Elect. Agric.
73, 66-73.
Lawrence H (2013) In: Accurate and efficient use of nutrients on farms. (Eds L.D. Currie and
C L. Christensen). Occasional Report No.
26. FLRC, Massey University, Palmerston North, New Zealand. 5 pp.
Mackay AD, Palmer AS, Rhodes AP, Cooper GK, Grant L, Withell B (2010) In: Precision
tools for improving land management. (Eds L D Currie and P Loganathan).
Occasional Report No. 14. FLRC, Massey University, Palmerston North, New
Zealand pp 79-87.
Greg Bishop-Hurley 1, David Henry 2, Leslie Overs3, Luciano Gonzalez4, Angela Anderson5,
Philip Pearce6 and Clinton Fookes7
1CSIRO Animal Food & Health Sciences, St Lucia QLD 4067, Australia, Email Greg.Bishop-
2CSIRO Animal Food & Health Sciences, Werribee, Vic 3030, Australia
3CSIRO ICT Centre, Pullenvale, QLD 4069, Australia
4CSIRO Animal Food & Health Sciences, Townsville QLD 4814, Australia
5Department of Agriculture, Fisheries and Forestry, Charters Towers, QLD 4820, Australia
6James Cook University, Townsville, QLD 4811
7QLD University of Technology, Brisbane, QLD 4001, Australia
Sustainable and viable primary industries must be capable of regularly producing a margin
above the costs of production. The real challenge is achieving this in an increasingly dynamic
and challenging environment where resources are limited whilst demonstrating improved
efficiency to the wider community with respect to environmental stewardship and animal
welfare. Viable and resilient farm businesses in the future will make use of a wide range of
data to make accurate and timely decisions. More accurate, timely and efficient management
(operational, tactical and strategic) across the farm business would be improved by the
timely, accurate and objective measurement of resources (from soil and water to feed,
animals and product quality and quantity) and the operating environment coupled with sound
interpretation and understanding.
In a joint initiative between CSIRO, James Cook University (JCU), Qld Dept Agriculture
Fisheries and Forestry (DAFF) and Queensland University of Technology (QUT), the Digital
Homestead project is investigating how electronic services enabled by connectivity to the
National Broadband Network can support greater productivity for farming enterprises, as well
as providing related support and social services to rural residents. Based at CSIRO’s
Lansdown Research Station near Townsville, QLD researchers are implementing sensor and
related technologies to provide information to simple and usable cloud-based decision
support systems for farmers and agriculture advisers. It is anticipated that key technological
solutions will then be evaluated on a commercial scale at QLD DAFFs Spyglass Beef
Research Station near Charters Towers, QLD.
A demonstration site has been established at Lansdown to monitor growing steers in an
extensive grazing environment. Three groups of 30 steers each graze one of three 15 ha
paddocks in rotation. Each group of three paddocks has one permanent water point that is
fenced off and has two spear gates, one for entry and one for exit. A walk over weigh station
connected to wireless sensor network is located behind the entry spear gate. The sensor
network relays data from a range of static sensors including animal live weight, climate data
and soil moisture and pasture/soil reflectance values. Livestock monitoring devices record
animal location and activity continuously. The data are uploaded to a central server and can
be viewed in real time via the web.
A web-based ‘dashboard’ has been developed to integrate and present information obtained
from both internal (e.g. LW, weather, animal location and behaviour) and external sources
(e.g. climate forecasts and market information). The key requirement is that information is
presented in a timely and informative way, can be tailored to individual users’ needs and
preferences, and enables more informed decisions. The design and functionality of the
dashboard were based on the ongoing input of industry stakeholders.
We gratefully acknowledge funding through the Queensland Government Smart Futures
John Broster1 and Rebecca Doyle1
1Graham Centre for Agricultural Innovation (NSW Department of Primary Industries and
Charles Sturt University) Locked Bag 588, Wagga Wagga, NSW 2678, Australia, Email;
Proximity loggers were attached to all 48 adult Merino ewes (1-9 yr old) in their home
paddock (3.04 ha) for a larger study looking at social relationships between individual sheep.
Proximity loggers were also placed at the only water trough in the paddock and the tree most
commonly used for shade between December 6 and 12, 2012 to investigate the use of these
resources. Whenever one unit was within approximately 4 m of another unit, a distance that is
common to published studies, the loggers recorded both the time and duration of that contact.
Over the seven days, the 48 sheep visited the trough on 1340 different occasions with over
98% of the visits occurring between 0600 h and 0900 h, and 1400 h and 2100 h, and no
animals visited the trough between 1000 h and 1300 h or 2100 h and 0400 h. The number of
visits per day per animal ranged from 0 to 14, with 81% of the 336 ewe/day combinations (48
ewes x 7 days) recording between 1 and 6 visits per day. The average time spent at the trough
per visit was 31.7 s (range 1 to 266 seconds) with only 231 visits (17%) lasting longer than
one minute. The animals were recorded as being under or near the tree on 16,861 occasions,
and 99% of these occurred between 0500 h and 1500 h. The number of visits to the tree per
day ranged from 0 to 233 (mean 50.6), with an average length visit of 16.5 s and the longest
25 min.
Older ewes (7 years and older) spent less time at the trough than younger ewes (4 years and
younger) (P < 0.01), but there was no effect of age on the number of trough visits, time spent
under the tree, or the number of visits. All ewes spent more time at the trough for the first
four days than the last three days (P < 0.001), while more time was spent under the tree
between December 8 and 10 than the four other days (P < 0.001). Higher daily temperatures
correlated to increased utilisation of the tree, with the temperature at 0900 h providing the
highest correlation for the time spent under the tree (r = 0.81, P < 0.05), and number of visits
(r = 0.91, P < 0.01). There was no significant correlation (P > 0.05) between daily
temperature and trough utilisation (visits r = -0.24; time r = -0.28). This was in opposition to
the hypothesis that as temperature increased, time spent at the trough increased.
It is probable however that the proximity loggers may have recorded more visits to the tree
than actually occurred, and under recorded the duration of the visits. When sheltering under
the tree the ewes were lying or standing very close to each other, possibly blocking the
signals between the loggers on the ewes and the logger placed in the tree. In order to further
investigate this hypothesis, behavioural observations need to be undertaken to accurately
correlate the proximity logger data with actual behaviour. The results nevertheless suggest
that proximity logger technology can be used to measure resource use, but that it needs to be
well-validated to ensure accuracy.
Cameron EF Clark1, Sergio C Garcia1, Kendra L Kerrisk1, James P Underwood2,
Juan I Nieto2, Mark S Calleija2, Salah Sukkarieh2 and Greg M Cronin1
1Dairy Science Group, Faculty of Veterinary Science, The University of Sydney, Camden NSW 2570,
Australia Email
2Australian Centre of Field Robotics, Faculty of Engineering and Information Technologies, The
University of Sydney Email
It is widely recognised that technology has a vast role to play in helping Australia’s farmers
reduce the time spent on repetitive tasks, increasing the attraction and retention of employed
labour in the industry and to provide and act on data to increase farm productivity to
sustainable levels. The continuous monitoring and movement of livestock between areas
defined for grazing or from these areas to procedural locations (i.e. yards, dairy facility) is a
repetitive task that is ideally suited to automation. A pilot study conducted at the University
of Sydney’s dairy research farm in Camden determined the behavioural response of dairy
cows to an unmanned ground vehicle (UGV) across time (Figure 1). The ability to use the
UGV in an operationally relevant way and the ability of the sensors and perception
algorithms on-board the UGV to automatically identify and track the motion of the dairy
cows are covered by Underwood et al. in the current proceedings. Following the morning
milking, the first 20 cows to be milked were separated from the main herd at 0830 h and
offered 0.5 ha of an ad-libitum kikuyu pasture allocation (50 kg DM/cow to ground level). A
pre-defined figure eight route was determined for the UGV within this 0.5 ha. After allowing
time for the cows to settle, the robot entered the pasture allocation at 0900 h and traversed the
figure eight route at a speed of 2.5 km/h (average traverse time was 7 min). Between
traverses the robot was parked outside the allocation until the process was repeated a further
four times at 15 min intervals. The 0.5 ha was virtually split into four sectors for observation
purposes, with four observers covering one sector each. To determine the interaction between
the UGV and cows, the number of cows exiting or entering each sector when the UGV was in
or out of the given sector was recorded. Data were analysed by GLMM within REML. The
model was as follows: Cows out = Fixed (Robot (presence/absence) * Traverse number) +
Random (Cow).
Figure 1. The unmanned ground vehicle
There was a significant effect of Robot (P = 0.02) and Traverse (P < 0.01) on the number of
cows exiting a sector, however, there was no interaction between these fixed effects. Twice
as many cows exited a sector when the robot was present (8%) as compared with absent
(4%). More cows exited a sector in traverse 1 (14%) as compared with all other traverses
(mean = 4%). The greater number of cows exiting a sector in the first traverse was likely
associated with an initial period of increased cow movement as cows foraged. These results
also indicate that dairy cows habituate to the moving UGV quickly. Future work will aim to
fully automate the process of herding and integrate this process with other data requirements
such as ground cover and soil moisture levels.
The authors would like to acknowledge the staff of both Corstorphine dairy and the
Australian Centre of Field Robotics for their excellent advice and support.
Sean J Dickson1, Mark Trotter 1 and Robin Dobos1,2
1Precision Agriculture Research Group, University of New England, Armidale, NSW 2351, Australia Email
2NSW DPI Beef Industry Centre of Excellence, Armidale, NSW, 2351, Australia
Lamb mortality is a major source of lost production in Australian sheep systems, with
estimates of actual losses between 5-30% (Everett-Hincks and Dodds, 2008). Lambing deaths
are mostly attributed to dystocia, starvation/exposure and mismothering (Geenty et al., 2013).
Of all these challenges dystocia has been identified as the most common cause, and has been
attributed to being a factor of starvation and exposure, as lambs that have longer than average
births have been found to have higher mortality (Cloete and Scholtz, 1998). Constantly
monitoring a flock of ewes around lambing is not practical, as a typical flock will lamb over a
6-8 week period. This opens an opportunity for an automated system that could detect the
initiation of lambing, and alert a producer if a dystocia event is occurring, providing a more
efficient technique for identification and ameliorating some causes of lamb mortality. The
objective of this experiment was to investigate the usefulness of GNSS location devices to
indentify changes in pregnant ewe behaviour around lambing to alert a producer.
Lambing behaviours have been well documented, such as isolation, circling, pawing the
ground and vocalisation (Arnold and Morgan, 1975). During the pre-lambing phase, it is
believed that the ewe will have a decreased level of activeness in terms of grazing and
distance travelled, (apart from the seeking of isolation). The current experiment used 20
GNSS collared grazing ewes set to record locations, set at 5 minute intervals. The ewes were
a mix in age and included maidens. The paddock size was 1.7 ha and there was limited
human interaction to mimic natural lambing behaviours. Observations were carried out twice
daily (morning and evening).
Data are currently being analysed using ArcGIS to determine ewe movement speeds before
and after lambing times, and is being correlated with lambing times. We will attempt to
identify a trigger point for lambing or the initiation of lambing from the correlation between
lambing time and ewe behaviour.
Arnold G, Morgan P (1975) Applied Animal Ethology, 2(1), 25-46.
Cloete S, Scholtz A (1998) Animal Production Science, 38(8), 801-811.
Everett-Hincks J, Dodds, K (2008) Journal of Animal Science, 86(14 suppl), E259 E270.
Geenty K, Brien F, Hinch G, Dobos R, Refshauge G, McCaskill M, Ball A, Behrendt R, Gore
K, Savage D (2013) Animal Production Science
Robin Dobos1,2, Mark Trotter1 and Hutton Oddy2
1Precision Agriculture Research Group, School of Science & Technology, University of New
England, Armidale, New South Wales, 2351, Australia, Email
2New South Wales Dept. Primary Industries Beef Centre of Excellence, Armidale, New South
Wales, 2351, Australia; Email;
Estimating intake of grazing ruminants is difficult and expensive. Previous approaches have
included behavioural observation, ratio techniques using indigestible markers, mechanical
recording of ingestive jaw motion and acoustic recording of ingestive behaviours. We present
preliminary results of an experiment designed to investigate activity of grazing ruminants to
improve our understanding of methane production and its prediction. In this experiment a
steer was fitted with a key-chain video with microphone and a tri-axial accelerometer to its
lower jaw to record ingestive behaviour. Recordings were taken while the steer was fed in
either a pen feeding or grazing situation. A total of 30 minutes video with sound was
collected on separate days for both feeding situations. This abstract reports on the preliminary
analysis of the sound waves recorded during a three-minute feeding bout. The grazing steer
had a higher proportion of ingestive sounds (mono) in the 0-1 kHz band than in when pen-
feeding (0.69 v 0.58; Figure 1). However, there was a higher proportion of sounds within the
1-2 kHz band for pen-feeding than for grazing (0.29 v 0.15). Shannon’s diversity index
indicated that there was little difference in the acoustic diversity for both feeding situations
(0.117 pen v 0.115 grazing). The Gini coefficient (a measure of statistical dispersion that is
zero if all the data are the same and increases as the data become more diverse) was
calculated using the occupancy per frequency band as input. There was a slightly higher Gini
coefficient for pen-feeding than grazing (0.75 v 0.72). Spectral analysis supported the
differences in sounds between pen feeding and grazing. Further analysis will determine the
utility of this technique in identifying eating behaviour in conjunction with tri-axial
accelerometer jaw movements.
Figure 1. Proportion of sounds from three-minute recording using key-chain video
device on a steer feeding in a pen (white bars) and grazing (black bars).
Robin Dobos1,2, Mark Trotter1 and Hutton Oddy2
1Precision Agriculture Research Group, School of Science & Technology, University of New
England, Armidale, New South Wales, 2351, Australia, Email
2NSW Dept. Primary Industries Beef Industry Centre of Excellence, Armidale, New South
Wales, 2351, Australia Email;
This abstract reports on the preliminary analysis of jaw movements measured using a tri-axial
accelerometer on the lower jaw of a steer either pen feeding or grazing. The steer also had a
key-chain video camera (see Dobos et al.; page 17 of these proceedings) to record
movements and sound. Data were collected over 36 h in both feeding situations on two
separate days. From a three-minute subset of the accelerometer data the inclination angles
were determined. This has identified differences in head and jaw movements between the two
feeding situations. During pen feeding the steer displayed a different feeding pattern than
when grazing, with head and lower jaw movements being more pronounced during pen
feeding than in the grazing situation (Figure 1). The higher peaks relate to head movements
(up/down) while the shorter peaks are lower jaw movements (up/down). In the pen feeding
situation, the head goes down to feed, a bite is taken and then the head is lifted to eat, while
during grazing the head stays in the down position while biting takes place over an extended
period of time. Lower jaw movement between head up/down states indicates both biting and
chewing taking place. Further analysis will determine the utility of this technique in
identifying eating behaviour in conjunction with video and sound recordings. This will enable
eating behaviour to be classified into chewing, biting, swallowing and chew-bite categories.
Figure 1. Steer head inclination (degrees) as determined from an accelerometer on the
lower jaw during (a) pen feeding and (b) grazing. Activities shown are not determined,
they are for information.
Head up?
Head down?
-50 0 5 0
Time (s)
Incli nation (degrees)
-50 0 50
Time (s)
Incli nation (degrees)
Suzanne Dowling1, Mairi Stewart1, Jim Webster1, Al Schaefer2 and Tom Landgraf3
1 AgResearch Ltd, Ruakura, Hamilton, NZ Email
2 Agriculture & Agri-Food Canada, Lacombe, Alberta, Canada Email
3Fraunhofer Institute for Transportation and Infrastructure Systems, Dresden, Germany Email
Sensing technologies capable of automated, non-invasive detection of stress and disease are
required for practical ‘farmer friendly’ assessments of animal health and welfare on-farm. For
example, as the average dairy herd size increases, so does the reliance on automated
monitoring systems to replace the human interaction with individual animals. Increased
capability such as early disease detection would be beneficial in terms of economic cost as
well as animal welfare. There is also an increasing pressure to provide ‘welfare friendly’
products in the animal-based agricultural industries.
Infrared thermography (IRT) currently has many applications ranging from measuring the
effectiveness of building insulation and electrical fault monitoring to human and animal
clinical diagnostics. IRT has been used in animal welfare as a non-invasive tool, to measure
stress during painful husbandry procedures such as disbudding of dairy calves (Stewart et al,
2008). The handheld FLIR S60 camera recorded the maximum eye temperature (ºC) at a 90º
angle (Figure 1) and at a distance of 0.5 m. An animal’s eye temperature rapidly decreases
immediately following a painful procedure without administration of local anaesthetic. The
drop in temperature is short-term (approximately 5 min post-procedure) and is then followed
by a prolonged elevation.
Figure 1. An infrared image of a calf’s eye. The cross indicates where the
maximum temperature is within the eye area which is then used for analysis.
More recently, IRT has successfully been applied in early disease detection in cattle (Schaefer
et al., 2012). Schaefer and his team developed an automated, RFID driven, IRT system where
an eye image was recorded each time a calf visited a water station. Calves that developed
bovine respiratory disease (BRD) were observed to have a higher eye temperature than calves
without BRD. Importantly, the BRD calves (identified by IRT) were not displaying any
clinical signs and stockmen handling the calves were not aware treatment was required. This
meant that the IRT technique enabled earlier treatment than was otherwise possible.
Last year, the Canadian system was set up in an automated calf rearing shed on a dairy farm
in Christchurch where we integrated the IRT system into an automated calf feeding station.
The idea was that the existing production and behavioural information that the system
currently gathers (e.g., number of visits, feed intake, milk removal velocity) in combination
with thermal changes may be a useful early predictive index for common diseases in dairy
calves, such as Rotovirus. Fifty calves were monitored daily from one day of age for
approximately 6 weeks. Three calves observed to have clinical signs by the farmer were
subsequently diagnosed with Rotovirus via faecal samples; however, predicted temperature
changes using the automated IRT system were not detected. This was thought to be mainly
due to only one of the two feeding stations available to the calves being monitored; therefore
we were unable to capture sufficient daily images for each calf that could be used for
analysis. One particular issue with IRT image detection in this way is the variability in the
data reducing the number of useful images.
Wirthgen et al. (2011) developed a s ystem for the automatic IRT monitoring of dairy cows on
a rotary parlour platform. The focus of this work was optimization of the IRT temperature
uncertainty and the development of automatic image processing algorithms. With the
introduction of an optimized reference body in the background of each image, the
temperature uncertainty could be reduced from typically ±2.3 K to below ±0.5 K (Wirthgen et
al., 2012). Furthermore a new correction model was introduced to reduce the influence of
changing ambient temperature and humidity on image quality. To enable automated image
processing, a new Active-Shape Model and a Level-Set approach were developed and tested
in a study where 450 dairy cows were monitored three times per day for 100 days. The
developed research platform is capable of collecting an IR dataset of complete herds over a
long time and is therefore a good tool for the development of health and welfare monitoring
systems. Preliminary results for diagnosis of individual animal inflammation from udder and
hoof temperatures are promising. Further investigations should optimize the modeled
temperature correction and transfer the image processing algorithms to other species (e.g.
goats and pigs).
For sheep and beef farmers, there would also be great benefits in early detection and
treatment of diseases (such as facial eczema, FE) and increased health and welfare
monitoring in extensive environments. However, monitoring of grazing animals is more
difficult than dairy cows, which are brought in to a milking shed daily for milking. In 2012,
we investigated the use of IRT as a method for early identification of the onset of FE. We
predicted that animals with high gamma-glutamyl transpeptidase (GGT) levels would have
increased eye temperature prior to clinical signs of FE. A similar water station to the
Canadian system (Figure 2) was built with the exception that the NZ design needed to be
mobile; therefore it was mounted on a trailer. At the conclusion of the study, no cows were
clinically affected by facial eczema but useful information was gathered on the technical
difficulties of mobile infrared monitoring in a grazing situation. Further studies in this area
would include solar power and a reference body.
In summary, IRT has the potential to solve a number of the biggest management problems
facing NZ pastoral agricultural industries such as monitoring health and welfare with reduced
supervision or in extensive environments. The opportunity to capitalise on this has increased
due to the reduced cost of the equipment, widening use of electronic identification systems
and advancing automated analysis methods. Continued work in this area of automated
monitoring will facilitate a major advance in decision-making abilities on-farm and assist
farmers with increasing compliance standards around animal health, welfare and
Schaefer AL, Cook NJ, Bench C, Chabot JB, Colyn J, Liu T, Okine EK, Stewart M, Webster
JR (2012) Research in Veterinary Science. 93, 928-935.
Stewart M, Stafford KJ, Dowling SK, Schaefer AL, Webster JR (2008) Physiology and
Behavior. 93, 789-797.
Wirthgen T, Zipser S, Franze U, Geidel S, Lempe G (2011) IEEE Sensors 2011, pp.1800-
1803, 28-31.
Wirthgen T, Zipser S, Geidel S, Franze U (2011) TM Technisches Messen, Oldenbourg
Verlag, 79, 168-174.
Figure 2. A drawing of the water station with automated IRT monitoring. Components:
(1) side panels, (2) two water troughs, (3) extension panels, (4) viewing windows, (5)
antennae (6) RFID control modules, (7) electromagnetic shielding (8) infrared camera
on motor mount within enclosure, (9) instrumentation cabinet (from Schaefer et al.
Zac Economou1, Mark Trotter1, Robin Dobos1,2
1The Precision Agriculture Research Group & CRC for Spatial Information, University of New
England, Armidale, NSW 2351, Australia.
2NSW Department of Primary Industries Beef Industry Centre of Excellence, Armidale, NSW 2351,
Managing livestock grazing behaviour is essential to understand grazing distribution and to
optimise grazing management. Many of the problems associated with grazing livestock in
extensive systems are related to their uneven patterns of use across the landscape. Fencing is
the most frequently used tool for influencing grazing distribution. Virtual fencing (VF) has
the potential to automate animal management and provide autonomous animal control in real
time. VF requires animals to wear an electronics package that includes hardware, software
and an antenna to receive radio frequency signals. However, there is limited information on
how VF might affect grazing distribution, animal behaviour and welfare in a large scale
commercial context.
The aim of this trial was to study animal responses to stimuli and evaluate the potential of VF
for sheep. Using a familiar paddock, eight sheep were fitted with radio frequency (RF) based
electronic containment devices (ECD’s) to simulate VF and were placed at one end of the
paddock. The paddock was subsequently divided in half by an RF ground wire. Within 1.5 m
of the ground wire an audible warning was delivered by the device, if the sheep continued to
move towards the ground wire an electrical stimuli was applied. The other side of the “virtual
fence” was made attractive by the presence of the animal’s campsite. The experimental
period was 7 h, after which the ECD’s and ground wire were removed. The sheep were
returned to the paddock to test if their spatial behaviour had been modified because of their
experience with VF.
The ewes demonstrated a strong ability for associative learning, one which could be further
exploited to reduce the need for electrical stimulation. No short term detrimental effects were
noticed, with sheep returning to graze within 10-20 seconds following electrical stimulation.
After ECD removal, there was no evidence of any behavioural changes, with the ewes
crossing the position of the ground wire and returning to pre-VF spatial utilisation
immediately upon returning to the paddock. This experiment found a number of problems
facing the fitting of VF units to sheep. The deployment of collars on sheep is not a long term
option as they interfere with fleece and skin.
The results suggest that there is potential for the application of VF technology to sheep to
modify their behaviour however further research is required into how sheep might actually be
fitted with VF devices.
This study has been supported by the Australian Wool Education Trust.
Clare Edwards1, Mark Trotter2 and Zac Economou2
1NSW Department of Primary Industries, Armidale, NSW 2351, Australia, Email
2The Precision Agriculture Research Group & CRC for Spatial Information, University of New
England, Armidale, NSW 2351, Australia, Email
Integral to understanding the spatial and temporal variability of landscapes is an investigation
of the pasture species component. While precision tools such as EM38, elevation, pasture
biomass and NDVI tools (e.g. CropCircle) can give rapid baseline and static data, pasture
species identification can enhance pasture quality and quantity predictability and help to
understand livestock behaviour.
As part of the ‘Kirby’ Smart Farm, University of New England, Armidale, pasture species
identification of paddocks was initiated in winter 2013. Using a 100 m grid point system,
botanical composition was assessed using the Botanal technique (Tothill et al. 1978). Visual
observations of herbage mass and ground cover were also recorded. Paddocks on the Smart
Farm have been previously surveyed using DGPS for elevation, EM38, and active optical
sensor for pasture biomass. Remote sensing imagery has also being compiled. Finally, the
spatial landscape utilisation by livestock has been recorded using GPS tracking collars and
Taggle ear tag tracking systems. This research quantifies the spatial variability in pasture
species and explores the relationship between this and other data layers.
Understanding the spatial variability in pasture species will provide a better awareness of the
interactions between soil, environment and livestock behaviour data. The majority of the
Smart Farm is based on perennial pastures, both temperate and tropical, with small areas of
forage cropping and intensive improved pasture systems. Results from a recent paddock
survey identified over 23 herbaceous species, with Weeping grass (Microlaena stipoides)
accounting for 20% of the biomass.
Results will refine Northern Tablelands data on pasture species growth curves, influences of
soil temperature and soil moisture on growth and inform a greater appreciation of livestock
behaviour, including their influence on pasture species distribution. Benefits also include
ground truthing of pasture biomass platforms such as Pastures from Space, especially in
regards to species, percentage green and ground cover. The results will also provide
information when considering sub-paddock scale pasture growth modeling.
Previous studies have indicated that landscape factors can produce pasture species diversity
and variability (Virgona and Hackney, 2008). However, the potential value for livestock
producers in terms of fodder budgeting, understanding grazing behaviour, landscape
management and predictability are enormous. Further, the ability to refine precision
agriculture in extensive grazing landscapes – including the use of site specific fertiliser
applications (Anderson et al., 2012), assisting also in animal productivity, welfare and health
is dependent on an appreciation of the underlying pasture species.
Anderson S, Trotter M, Haling R, Edwards C, Guppy C, Lamb D (2012) Australian and New
Zealand Spatially Enabled Livestock Management Symposium, 2012.
Tothill JC, Hargreaves JNG, Jones RM (1978) Tropical Agronomy Memorandum No 8.
Virgona J, Hackney B (2008) Proc. 14th Agronomy Conference, 2008.
Eloise S Fogarty1, Jaime K Manning1, Mark G Trotter2, Derek A Schneider2, Russell D Bush1
and Greg M Cronin1
1Faculty of Veterinary Science, The University of Sydney, Camden, NSW 2570, Australia,
2Precision Agriculture Research Group and CRC for Spatial Information, University of New
England, Armidale, NSW 2351, Australia
The Australian Merino industry has experienced substantial decline in recent years due to
falling consumer demand for wool products, extended drought, an aging farmer population
and increasing returns from prime lamb production (Australian Wool Innovation, 2010). The
resulting low national flock numbers highlights the need for improved sheep reproductive
rates for population rebuilding. Poor reproductive performance is considered a significant
economic burden within farming systems, with Australian specialist sheep enterprises
averaging 76.9 lambs born per 100 ewes (Hatcher et al. 2010). Recent advancements in
technologies such as GPS tracking have resulted in the incorporation of tracking and
biotelemetry data into animal behaviour studies. Specifically, accurate records of an animal’s
location and movement can be analysed to decipher behaviour patterns of interest to
researchers and producers. The incorporation of remote monitoring technology into current
sheep systems has the potential to improve both productivity and efficiency through increased
animal monitoring and reduced labour requirements. Furthermore, when GPS monitoring is
applied in a reproductive management scenario, remote oestrus detection could facilitate
increased reproductive rates and rate of genetic gain through improved accuracy of ewe-ram
The use of GPS monitoring for animal behaviour studies and how it may be applied to sheep
production systems to predict the onset of oestrus will be discussed. Potential indicators of
oestrus behaviour will be explored, including changes in diurnal activity patterns and daily
distance travelled between non-oestrus and oestrus days. The impact of sexual experience
will also be reported following a comparison of maiden and experienced ewe behaviour. This
research project will provide a solid foundation for the development of behavioural models
that explain ewe movement in the paddock and how these change during oestrus. These
models can then be incorporated into farm management allowing farmers to exploit real-time
GPS data for improved reproductive efficiency.
This study has been supported by the Australian Wool Education Trust (AWET).
Australian Wool Innovation (2010) Strategic plan 2010 to 2013: full version [Online].
Available at
revised310810.pdf (Verified 26 April 2013).
Hatcher S, Hinch GN, Kilgour RJ, Holst PJ, Refshauge PG, Shands CG (2010) Aust Farm
Business Mgt J 7(2), 65-78
Luciano Gonzalez1,4, Greg Bishop-Hurley 2,4 and David Henry 3,4
1CSIRO Animal Food & Health Sciences, Townsville QLD 4814, Australia, Email
2CSIRO Animal Food & Health Sciences, St Lucia QLD 4067, Australia
3CSIRO Animal Food & Health Sciences, Werribee, Vic 3030, Australia
4 CSIRO Sustainable Agriculture Flagship
New technologies, such as sensors to measure a variety of conditions, wireless sensor
networks and information and communication technologies are rapidly expanding and finding
a large range of applications in agriculture and livestock production. One of the greatest
advantages of these technologies is the ability to remotely collect and transmit information at
very fine spatial and temporal scales. New technologies are expected to produce enormous
benefits to the industry improving productivity, profitability, environmental outcomes and
quality of life in remote areas.
However, managing and making sense of large amounts of data and presenting them to the
end user is a challenge. The main challenges are related to the computational power and
methodologies required to process such large amount of information, and converting raw data
in useful measures to aid in monitoring and decision making. The present work will show
recent efforts around management and processing of remotely collected data from the
wireless sensor network deployed at CSIRO’s Lansdown Research Station. The main focus is
on data collected from GPS and motion sensors by cattle monitoring collars, animal live
weight collected by remote weighing stations, and body temperature collected with infrared
Location data were collected at 4 Hz and 3-axis accelerometer and magnetometer data at 10
Hz. Location data were used to determine distance travelled. Data from the motion sensors
were used to determine animal behavioural activities. Remote weighing stations were located
in the entry gate of a yard enclosing the water point in the paddock and the data collected
were used to determine live weight and live weight gain of individual animals. The weight of
each individual animal along with identification, date and time were collected as they walked
through the system to access the water. Grazing animals wearing RFID ear tags had to walk
through the entry gate and then the station containing load bars, a platform and a RFID reader
Data from the collars was processed to delete erroneous observations and then 10-seconds
mean and standard deviations were calculated for each variable collected. The methodology
developed for behavioural classification of data from the collars was based on the finite
mixture distributions methodology used to determine threshold values that formed part of a
decision tree. Direct visual behavioural observations were made for 5 activities, i.e. grazing,
ruminating, resting, travelling and other active behaviours including scratching and
grooming). The algorithm correctly classified 85 and 90% of the data points in the
development and validation dataset, respectively. Location data were both sub-sampled and
averaged over different time intervals to determine the effect of frequency of data collection
to measure distance travelled by animals (Euclidean distance between any two consecutive
points). Data from the remote weighing station were processed to delete erroneous values and
then rolling averages, rolling medians and rolling slopes were calculated to obtain final live
weight and live weight change. Infrared images were taken from the eye, the rump and the
trunk to measure the temperature in these locations. Images were processed with specialised
software to obtain maximum and average temperatures. The effects of several environmental,
management and physiological factors on body temperature were determined. Infrared
thermography was able to detect differences in body temperature as a result of several of
these factors.
Vincent Griffith1, Anne Geoghegan1, Michael O’Donovan1, Bernadette O’Brien1 and
Laurence Shalloo1
1Animal and Grassland Research and Innovation Centre,Teagasc, Moorepark, Fermoy, Co.
Cork, Ireland, Email
The future of an efficient low cost milk production system in Ireland will depend on the
development of new grassland technologies which can lead to greater efficiency in grass
utilisation. The creation of PastureBaseIreland (PBI) is an important step in the development
of such technologies. PBI represents a grassland management decision support tool which
incorporates a mechanism to capture background data on both research and commercial
farms. The database stores all grassland measurements in a common structure. This will
allow the quantification of grass growth and DM production (total and seasonal) across
different enterprises, grassland management systems, regions and soil types using a common
measurement protocol and methodology. Grass measurements are recorded on a regular basis
and reports (grass wedge, distribution of growth and paddock summary reports) are
automatically generated for management purposes. The reports are developed in a format that
allows individual farms to be benchmarked with other farmers in their discussion group or to
be benchmarked with farmers regionally. The background data such as paddock soil fertility,
grass cultivar, aspect, altitude, reseeding history, soil type, drainage characteristics and
fertiliser applications are also recorded. All farms on PBI are attached to their nearest Met
Eireann weather station to allow the linkage between meteorological information to grass
growth to be created. Information from different commercial grassland management software
packages will be allowed into PBI once it meets rigorous quality assurance standards.
Incorporating this data will increase the value of the database and would ensure that all
generated grassland data is stored in one national database. All of the information collated
within this new database can be used for future research projects to increase the
understanding around grass growth at farm level, which should ultimately contribute to
increased grass growth and utilisation at farm level. In summary, PBI is a new national
grassland database. This database will enable the collection of regional grassland data across
dairy, beef and sheep farms while providing decision support information for farmers and
collating the background research information into a centralised grassland database.
Jamie Haultain1, Ian Yule2, Alvaro J Romera1, Brian Dela Rue1, Dave Clark1, Chris Glassey1
and Jenny Jago1
1DairyNZ, Private Bag 3221, Hamilton 3240, New Zealand.
2Institute of Agriculture and Environment, Massey University, Private Bag 11222, Palmerston North
4442, New Zealand. Email
Knowledge of individual paddock performance can assist management decisions (e.g. which
paddocks to renovate) and contribute to improved productivity and profitability of dairy
farms. Annual paddock performance is typically assessed by estimating the net herbage
accumulation through regular ‘farm walks,’ but this is not a routine practice on many dairy
farms. The aim of this research was to test the hypothesis that automated daily records of
milk production could be used in conjunction with the number of cow grazing events
(measured using GPS units fitted to cows) to rank paddocks according to their annual net
herbage accumulation. This required the assignment of herd milk yield values at AM and PM
milkings to the corresponding paddock(s) grazed, along with records of the number of cows
present. This abstract describes progress in the project to date.
Prior to field trials, the DairyNZ Whole Farm Model (WFM) was used to test the hypothesis.
The WFM was run over two seasons (2001/02, 2002/03) and each paddock’s milk production
L/ha/yr and the number of grazing events/ha/yr were summarised. The number of grazing
events for each paddock was a good predictor of the ranking of herbage accumulation per
paddock, with an value of 0.84. Prediction using the milk production each year was less
reliable (R² = 0.67).
The hypothesis was then tested on a commercial farm for one year, between December 2011
and December 2012. Automated recording of grazing events was achieved using GPS devices
on a small number of cows per herd. The accuracy required of a GPS device was tested to
determine suitability for this task. Static testing of GPS devices and a simulation process
identified the minimum 95% circular error probable (CEP) was 11 m to obtain at least one
position fix sufficiently inside a 1 ha paddock and correctly identify the paddock the herd was
grazing. The GPS units used (Telemetry Solutions Ltd, USA) comfortably met these
requirements. Three GPS units were allocated per herd of cows, with each fixing one position
per paddock entered. During static testing, the GPS devices recorded a position at every
attempt, however whilst on the cows, the GPS devices regularly failed to fix a position,
possibly due to cow movement or the orientation of the device on the cow.
Using the same method of analysis as described for the WFM, the relationships between
annual net herbage accumulation, as determined by weekly measurements with a rising plate
meter, and each paddock’s grazing events or milk produced per ha were calculated. The
relationships were considerably lower than predicted (R2 of 0.1 and 0.26 for grazing and milk
respectively). However, these are preliminary results and are likely to improve as more
variables are included in the analysis, such as the effect of supplementary feed. Additionally,
accuracy may be improved if data were collected over two years as with the WFM.
René Kolbach1, Kendra L Kerrisk1, Sergio C Garcia1 and Navneet K Dhand2
1Dairy Science Group, Faculty of Veterinary Science, The University of Sydney, Camden
NSW 2570, Australia, Email
2J. L. Shute Building (C01), Faculty of Veterinary Science, The University of Sydney, Private
Mailbag 4003, Narellan, New South Wales 2567, Australia
Throughout 2009 and 2010, FutureDairy (Camden, NSW, Australia) was involved in testing
and co-developing a prototype robotic rotary (RR). The commercial version RR has the
technical capability of conducting in excess of 1500 milkings/day. This level of throughput
will only be possible through implementation of best practice management and suitable farm
and system design. To achieve the high throughput the rotary rotates individual cows to the
cup attachment robot(s) and then around the platform in a stop–start fashion. The robots do
not remain with the cow during the entire milking process. When not all teat cups are
attached or when one or more teat cups are removed prematurely the milking is deemed
incomplete. Recent data from a commercial installation of the RR with cows being batch
milked at relatively consistent intervals indicated a throughput potential that has averaged 58
milkings/hour and peaked at 99 milkings/hour. The proportion of incomplete milkings from
the same dataset averaged 5.8%. Whilst this level of performance is extremely encouraging,
the fact is that incompletely milked cows require some additional management. The study
presented here was conducted with the prototype RR and was designed to test whether or not
the extension of the interval between a first and second milking attempt improved the likely
success of the second attempt for incompletely milked cows. The 1 h milking interval
treatment (1 h) simulated cows being drafted directly back to the pre-milking waiting yard,
whilst the 3 h milking interval treatment (3 h) was designed to simulate cows being drafted
back after accessing post-milking supplementary feed on a feedpad. The results showed no
significant difference between the frequencies of successful attachment in the second attempt
between the 1 h and 3 h treatments indicating that a reasonable level of flexibility exists with
management of incompletely milked cows and dairy layout designs. Milk production level
affected the probability of success at second attempt, which was about 7.5 times higher in
cows with an average milk production level greater than 19.3 kg than those with less than
10.8 kg. Primiparous cows were more likely to be successfully milked after two attempts than
multiparous cows.
The authors acknowledge the Dairy Research Foundation for its support and the investors of
FutureDairy; Dairy Australia, NSW DPI, DeLaval and The University of Sydney. We also
gratefully acknowledge Mikael Karttunen for his technical support, the farm and technical
staff and Nick Dornauf for sharing data from their commercial installation of the RR.
René Kolbach1, Kendra L Kerrisk1, Sergio C Garcia1 and Navneet K Dhand2
1Dairy Science Group, Faculty of Veterinary Science, The University of Sydney, Camden
NSW 2570, Australia, Email
2J. L. Shute Building (C01), Faculty of Veterinary Science, The University of Sydney, Private
Mailbag 4003, Narellan, New South Wales 2567, Australia
This study was conducted to investigate the effects of different bail activation sequences in
combination with feed availability on cow traffic and harvesting capacity in a prototype
robotic rotary (RR; DeLaval AMR™, Tumba, Sweden). The throughput potential of the
commercial RR is 1500-1600 milkings/day. However, in voluntary cow traffic systems, the
number of cows presenting may be low at certain times of the day (or during certain months
or seasons in seasonal calving systems). In these circumstances, the ratio of active bails to the
number of cows available may be undesirably high, with consequential negative effects on
system efficiency and milk quality (the RR does not flush individual units automatically after
each milking). Activating only 50% of the bails may be a management strategy chosen to
cope with periods of underutilization. Four treatments with a total activation of 50% of bails
[8 bails with activation sequences of 8, 4, 2, or 1 consecutive bail(s)], with or without the
presence of feed on the RR, were observed during sixteen 4 h observation periods after a
system wash. The absence of feed resulted in a significant increase in the proportion of
available bails remaining idle, but no significant differences were observed across the 4 bail
activation sequences. Overall, the effect of bail activation sequence on cow traffic was
negligible, but the sequences that had more consecutive bail activations resulted in more
robot operations being conducted simultaneously and more milk being harvested per minute
of robot operation time. These results indicate that a feeding function upon entry to the RR
platform, in combination with bails activated sequentially, lead to a more efficient use of the
The authors acknowledge the Dairy Research Foundation for its support and the investors of
FutureDairy; Dairy Australia, NSW DPI, DeLaval and The University of Sydney. We also
gratefully acknowledge Mikael Karttunen for his support of the prototype, and farm staff.
Figure 1. Four bail activation sequences (dark coloured bails are inactive and light
coloured bails are active)
René Kolbach1, Kendra L Kerrisk1 and Sergio C Garcia1
1Dairy Science Group, Faculty of Veterinary Science, The University of Sydney, Camden
NSW 2570, Australia, Email
This study investigated the effects of premilking teat preparation on attachment accuracy and
milk removal characteristics for individual cows in a novel 16-bail prototype robotic rotary
(RR; automatic milking rotary system; DeLaval AMR™, Tumba, Sweden). The RR is
equipped with one or two teat preparation modules (TPM’s) for premilking stimulation and
cleaning of teats. Operators will be able to disable the TPM’s during periods of clean and dry
climatic conditions and may expect that higher throughput might be achieved when the TPM
is deactivated. However, it is expected that attachment of teat cups would be faster and more
successful with the use of a TPM, and that the efficiency of milk removal, in terms of average
and peak milk flow rates, would increase. A significant effect of treatment (no wash vs.
wash) and individual quarters on attachment success was observed; cows exposed to the wash
treatment had up to 1.5 times higher odds of being successfully attached. The attachment was
not only more successful but was also found to be 4.3 s faster after cows were exposed to the
wash treatment. Average milk flow rate was not affected by the wash treatment.
Nevertheless, a significant interaction was found between wash treatment and interval since
previous milking on peak milk flow (kg/min) of individual cows. This interaction showed
that cows with a milking interval 8 h subjected to the wash treatment had significantly
higher peak flow rates (300 g/min increase) compared with cows in the same milking interval
category with no wash treatment. Most Australian farmers do not routinely wash cows teats
in conventional milk harvesting facilities and may be inclined to disregard the benefits of
operating the TPM. The relationship between premilking stimulation and attachment success
shown in this study will increase awareness (for both farmers and developers of the
technology) of the importance of teat cleaning within the RR.
This work was carried out within the FutureDairy program. We acknowledge the investors
Dairy Australia, DeLaval, NSW Department of Primary Industries and the University of
Sydney. The authors also thank Mikael Karttunen and Thomas Olsson (DeLaval) for their
technical support of the RR and the staff of the Future-Dairy AMS research farm for their
assistance. We also acknowledge the assistance provided by Peter Thomson and Navneet
Dhand (both of the University of Sydney) with the analysis and reporting of the data.
David Lamb1 & Mark Trotter1
1Precision Agriculture Research Group, School of Science and Technology, University of
New England, Armidale NSW 2351, Email
The “SMART” in SMART Farm represents a vision to deliver Sustainable, Manageable and
Accessible Rural Technologies on farm which enhance the business and lifestyle of farming.
Connected via both satellite and fixed-wireless NBN (the same connectivity options currently
available to Australian farms), SMART Farm serves as an education and extension resource
for teachers of agriculture and environmental sciences, and as a key platform for up-skilling
the wider agriculture sector for NBN-enabled technologies.
The University’s 2,800 hectare, commercial, SMART Farm (Figure 1) is a livestock and
sheep enterprise. Spatially-enabled livestock management is highlighted through the
integration of tools such as remote and in-situ pasture and soil moisture monitoring, livestock
tracking, genetics and performance, providing to educators live and interactive data for use in
teaching agriculture and environmental science. At the same time, the SMART Farm aims to
demonstrate practical and realistic NBN-enabled pathways and tools for increasing
productivity, and improving environmental outcomes, safety and work-flow, business
resilience and social inclusion on farm.
Figure 1. (a) The NBN-connected SMART Farm serves as a (b) connected classroom providing
live data, as well as live video conferencing/data sharing (c) into and (d) out of the farm.
The SMART Farm project has been developed with a range of educators and industry
participants in mind and as a convergence of industry innovation and technology. In addition
to supporting a national outreach program for high school students, the SMART Farm is used
in supporting undergraduate teaching in Science as well as Rural and Environmental Science,
as well as in the Graduate Diploma in Precision Agriculture and the Diploma in Agriculture
(Precision Agriculture).
NA Lyons1, KL Kerrisk1 and SC Garcia1
1Dairy Science Group, Faculty of Veterinary Science, The University of Sydney, Camden
NSW 2570, Australia, Email
In a pasture-based automatic milking system (AMS) a proportion of milking events occur
with milking intervals (MI) greater than 16 h (extended MI). This can have a negative impact
on milk yield and udder health. Therefore the aim is to reduce their occurrence.
Additionally, cows necessarily walk longer distances than in indoor-based systems. The
decision to milk a cow is based on milking permission criteria, which is generally based on
time since last milking but can often take into account expected yield as well. Any cow
arriving at the dairy and that does not receive milking permission is drafted to a pasture
allocation, but it is not known if milking refusal influences total time of return and therefore
MI. Understanding the impact that a refusal can have on MI is key to establishing effective
management practices and thereby achieve production targets.
Data were collected over a 33-day period (February March 2011) from the FutureDairy
pasture-based AMS research farm using a prototype robotic rotary milking unit (Automatic
milking rotary, DeLaval, Tumba, Sweden). Data were collated and analysed to investigate the
hypothesis that a greater proportion of milking events occurring with extended MI would
correspond to cows that had a previous milking refusal. Milking permission was granted at
selection gates if time since the previous milking was > 4 hours for cows less than 70 days in
milk, or >8 hours for cows over 70 days in milk. Cows that were denied access to the milking
unit were drafted to the corresponding allocation. If this was the case, then management
practices could be put in place to deal with cows that visit the dairy soon after the last milking
Results indicate that one third of milking events had extended MI, although only 16% of
them had a previous milking refusal. The average refusal took place 3 h after the last milking
event and caused extended MI in over 60% of the cases.
This indicated that although special attention should be placed on cows that returned to the
dairy before milking permission (because they were likely to have an extended MI), milking
refusals were not the main cause of extended MI. Therefore cows that visit the dairy facility
earlier than expected could be sorted to a feeding area close to the dairy, yet the greatest
impact on overall MI will probably be achieved by reducing time spent in any one pasture
The authors acknowledge the Dairy Research Foundation for its support and the investors of
FutureDairy; Dairy Australia, NSW DPI, DeLaval and The University of Sydney. The
authors would also like to thank the farm staff (Terry Osborne, Shannon Bennetts and Grant
Oldfield), technicians (Mikael and Kattis Karttunen) and former students (Rene Kolbach and
Victoria Scott) for their assistance during the data collection period.
NA Lyons1, KL Kerrisk1 and SC Garcia1
1Dairy Science Group, Faculty of Veterinary Science, The University of Sydney, Camden
NSW 2570, Australia, Email:
In pasture-based automatic milking systems (AMS), cows usually have a lower milking
frequency than those reported in indoor housing systems. Cows milked in pasture-based
AMS have greater milking intervals than cows milked in indoor AMS. Long milking
intervals, greater than 16 h, have a negative effect on milk yield and udder health.
Feed is the main incentive to manipulate cow traffic in AMS, yet to date no research has been
published that quantifies the actual impact of number of feed allocations, on animal
performance in pasture-based systems.
Therefore, the impact of two systems of pasture allocation on milking interval and yield was
investigated at the FutureDairy AMS research farm in late November – early December 2010.
Two (2WG) versus 3-way grazing (3WG) allocations per 24 h period were compared in a
field study to test the hypothesis that an increase in the frequency of pasture allocation would
reduce milking interval and therefore increase milking frequency.
The study involved the entire milking herd of 145 mixed age and breed cows. The herd was
milked using two DeLaval VMS milking units. Cows were offered their daily pasture
allocation (18 kg/d) in equally sized portions according to the established treatments (2WG =
2 x 9 kg/d and 3WG = 3 x 6 kg/d). In addition to pasture, cows were supplemented with 4 kg
DM/d concentrate in the milking station.
Cows in the 3WG treatment had 31% reduced milking interval, 40% greater milking
frequency and 20% greater daily milk production in comparison to 2WG. There was a
reduction in the amount of feed made available under 3WG (in total kg DM/allocation),
which created the potential for feed depletion to occur more rapidly thereby creating an
incentive for cows to exit the allocation sooner in search of additional feed. Additionally,
increased milking frequency and milk production for 3WG were associated with greater
utilization of the AMS milking units throughout the day.
These results support the recommendation that wherever possible farmers installing AMS
should incorporate sufficient infrastructure to accommodate 3WG, which provides additional
flexibility with managing extremely long (and short) milking intervals.
The authors acknowledge the Dairy Research Foundation for its support and the investors of
FutureDairy; Dairy Australia, NSW DPI, DeLaval and The University of Sydney. We would
like to thank the farm staff (Terry Osborne, Shannon Bennetts, Daniel Dickeson and Grant
Oldfield) and Rene Kolbach, for their assistance in preparing and running the study.
Jaime K Manning1, Eloise S Fogarty1, Mark G Trotter2, Derek A Schneider2, Russell D Bush1
and Greg M Cronin1
1Faculty of Veterinary Science, University of Sydney, Camden NSW 2570, Email
2Precision Agricultural Research Group and CRC for Spatial Information, University of New
England, Armidale NSW 2350
The predation of sheep (Ovis aries) by wild and domestic dogs (Canis lupis) is a major issue
in Australian agricultural systems, negatively impacting financial returns and causing serious
welfare issues to inflicted animals. Predation costs an estimated $66 million a year (McLeod
and Norris, 2004), and without new control methods it is predicted the sheep industry may
disappear in some regions of Australia (Allen and West, 2013). The inability of some
traditional methods to control wild/feral dogs and dingoes in sheep areas highlights an
opportunity to investigate innovative technologies to reduce livestock losses from predation.
Sheep respond to predatory stimuli via different strategies including altering spatial
behaviour. The sheep in a flock commonly pack tightly together (termed flocking), with
individual animals appearing to be strongly motivated to be at the centre of the group. The
aggregated group may also circle, in a behavioural response also known as a centripetal
formation. The assumption is that those animals located on the periphery of the main group
are the most vulnerable to predators. A fleeing response may also occur, illustrated by sheep
breaking from the flock and running directly away from the threatening stimulus. This
behaviour potentially results in panicked individuals colliding with barriers such as fences,
causing serious injuries.
Precision Sheep Management (PSM) is the application of technology to improve sheep
welfare and ultimately production. This includes GPS technology which facilitates the ability
to remotely monitor sheep in the paddock, including during a dog attack. With on-farm
labour increasingly limiting, PSM technology capable of detecting and interpreting rapidly
altered spatial behaviour in sheep, measured via real-time GPS, could alert farmers of a
potential predation event. However, innovative research and development are needed before
systems can be developed that “recognise” the occurrence of predation. Our research
quantifies the behavioural responses of sheep during simulated dog predation events (Figure
1) using GPS technology. It is envisaged that this approach will provide improved
understanding of sheep movement patterns during a dog attack, thus highlighting the
potential of GPS devices for livestock monitoring.
This research has been supported by the Australian Wool Education Trust (AWET).
Figure 1. A sheep segregated during a simulated predation event. The sheep and the
dogs wore individual GPS tracking collars.
Allen B, West P (2013) Australian Veterinary Journal, 91, 261-267.
McLeod R, Norris A (2004) 'Counting the cost: impact of invasive animals in Australia'
(Cooperative Research Centre for Pest Animal Control: Canberra).
Jessica Roberts1, Armin Werner1, Rory Roten1, Kenji Irie1, Michael Hagedorn1, Jaco Fourie1, Simon
Woodward1, Ian Woodhead1, Iris Vogeler2, Keith Cameron3, Grant Edwards3, Hong Di3, Ian Yule4,
David Lamb5 and Mark Trotter5
1Lincoln Agritech, Lincoln, Christchurch, New Zealand,, Email
2AgResearch, Palmerston North, New Zealand,
3Lincoln University, Lincoln, Christchurch, New Zealand,
4Massey University, Palmerston North, New Zealand,
5University of New England, Armidale, Australia,
Increasing efficiency of nitrogen use as well as minimizing losses in pasture systems is of
major concern from both production and environmental perspectives. Variable rate nitrogen
fertiliser application (VRA-N) has the potential to improve production, while reducing
unwanted environmental effects.
Applying VRA-N technology on grazed paddocks could firstly lead to a reduction of overall
N application in intensive pasture systems by supporting farmer understanding of the actual
N-demand and N-supply. Secondly, VRA-N would allow a more precise response to the
spatio-temporal variation of supply and demand of a pasture canopy within a paddock.
Determining VRA-N schemes based on soil analysis is not a practical on-farm process for
eight to 10 N-applications a year to grazed pastures in intensive production systems.
Therefore, an online-sensing system is needed to inform the fertilizer spreader on the go.
There are currently automated variable rate fertiliser systems on the market for arable
cropping, however these are not developed for heterogeneous and grazed pastures. The
OPTIMUM-N project aims to bring together new pasture sensing tools and variable rate
fertiliser machinery to enable automatic optimum nitrogen application.
Optical N-sensing systems in arable crops primarily use indirectly sensed N-concentration of
the vegetative part of the canopy for differentiating N-fertilization, with the goal of optimal
grain fill. A sensor based N-fertilization strategy of a grass canopy where amount and quality
of the green biomass is the target needs to take into account the N-status (N-concentration) as
well as the total standing (and also the expected) vegetative biomass of the canopy. To
achieve this, a tool will be developed using sensors, including machine vision and spectral
reflectance, to detect nitrogen related pasture parameters including: (i) optical reflectance
characteristics and (ii) structural variables (plant height, vertical biomass distribution, grass-
clover proportion). Supported by simulated expected biomass and soil nutrients, this tool will
be linked to farming conditions with a variable rate fertiliser spreader. Based on the collected
information and a farmer chosen management approach, the system will then automatically
apply the appropriate amount of N fertiliser to pastures.
This project is primarily funded by the New Zealand Ministry of Business, Innovation and
Employment, and involves consultation with farmers, industry and public representatives
including C-DAX, Dairy NZ and Environment Canterbury.
Jessica J Roberts1, Mark G Trotter1, David W Lamb1, Derek A Schneider1, Geoff Hinch2,
Greg Falzon3, Robin Dobos4.
1Precision Agriculture Research Group, University of New England, Armidale, NSW, 2351, Australia, Email
2Animal Science, School of Environmental and Rural Science, University of New England,
NSW, Australia
3C4D, University of New England, NSW, Australia
4Beef Centre, Department of Primary Industries, Armidale, NSW, Australia
GPS tracking collars were used to monitor changes in social interactions between steers as
grazing forage oats biomass declined. This trial examined the hypothesis that as available
biomass declines cattle become more dispersed in their grazing environment and that this
metric might be developed as an early warning of overgrazing. The trial consisted of two
replicate mobs of 25 steers, 22 tracked in each, grazed in paddocks of approximately 2 ha for
24 and 29 days respectively. GPS collars collected positional data every 15 minutes and peak
grazing time was identified as 0700 h, at which time the positions of all animals were
examined for each day. These data were processed to create minimum convex polygons
(MCP) and average distance between points. Paddock biomass was monitored using a
calibrated active optical sensor (CropCircle), five times over the trial period and a model
developed to predict daily biomass.
As biomass decreased, MCP area and perimeter increased, area from 381 m2 to 6,079 m2 and
351 m2 to 15,761 m
2 in paddocks 1 and 2 respectively; however, there was considerable
variation from day to day. The minimum, maximum, and mean distance between points also
increased over time, these metrics also showed variation but to a lesser extent than the MCPs.
Figure 1 clearly demonstrates that cattle become more dispersed in a grazing environment as
available feed declines. This information will be valuable for the development of movement
metrics which may be integrated into producer-based livestock monitoring systems to alert
managers when biomass becomes limiting.
This project is supported by The AW Howard Memorial Trust; MLA and the CRC for Spatial
Figure 1. Within herd average mean distance apart (dispersion) as Green Leaf Biomass
decreased for each herd.
Victoria E Scott1, Peter C Thomson2, Kendra L Kerrisk1 and Sergio C Garcia1
1Dairy Science Group, Faculty of Veterinary Science, The University of Sydney, Camden NSW 2570,
Australia, Email
2ReproGen Animal Bioscience Group, Faculty of Veterinary Science, The University of Sydney,
Camden NSW 2570, Australia
Automatic, or robotic, milking systems (AMS) involve the automation of the entire milking
process (including teat cleaning/preparation, cup attachment, milking, cup removal and post-
milking teat disinfection). Detailed reports on individual cows (or groups of cows), such as
details regarding each milking session, cow performance, cow traffic and location on-farm,
are generated daily from such systems. Cows are generally managed with voluntary cow
traffic (whereby cows move around the entire farm system with minimal human interference)
to achieve the greatest benefits from AMS. Management of incentives is necessary to ensure
acceptable levels of cow traffic, and subsequent milk yield, is achieved. The development of
a robotic rotary (RR) with reported throughput potential of up to 1,600 milkings per day
could have the ability to milk up to 800 cows twice a day. With such high throughput and a
single entry point to the milk harvesting equipment, movement through the dairy must be
efficient in order to minimise the risk of congestion and any associated negative impacts.
Incentives (the strongest and most reliable being feed) are used to encourage cow traffic in
any robotic milking system. While most AMS have the ability to provide feed during milking,
this is not the case with the RR. Here we report on an investigation designed to determine the
impact of a feed reward at milking on pre-milking voluntary waiting times. Cows were
managed in an Australian pasture-based system with voluntary cow traffic and milked on a
prototype RR. Treatments were given as “Feed On” (pelleted concentrate supplied at milking)
and “Feed Off” (no concentrate supplied at milking) and were exclusive (i.e. treatments were
not given simultaneously to groups within the herd). The study was of 33 days duration, with
cow traffic data collected over 16 days. A survival analysis, with time-varying covariates, was
used to model pre-milking voluntary waiting times.
Cows spent a median time of 2.2 h in the pre-milking yard before volunteering for milking,
with just over 70% of cows having exited the yard (volunteered for milking) after 4 h. It was
found that during the “Feed On” treatment, cows spent less time in the pre-milking yard (0.53
× less time) than they did during “Feed Off”. Heifers were faster to exit the pre-milking yard,
with older cows spending at least 1.40× longer in the yard before milking. Time spent in the
pre-milking yard increased as the number of cows present increased.
Feeding a small reward on the RR platform reduced the time cows spent in the pre-milking
yard. As AMS becomes more common in Australia, it is essential that research into
management strategies that encourage good cow traffic and cow welfare occurs.
The authors acknowledge the Dairy Research Foundation for its support and the investors of
FutureDairy; Dairy Australia, DeLaval, NSW DPI and The University of Sydney.
Laura Sutton1, Linda Gant2, Ivan Gant2, Melanie Smith1, Greg Cronin1 and Russell Bush1
1The University Sydney, Faculty of Veterinary Science, Camden NSW 2570, Australia, Email
2‘Glenalvon’, Cassilis NSW 2329, Australia
Monitoring liveweight change is highly relevant to commercial wool and lamb producers for
assessing flock productivity, since liveweight relates directly to fleece quality, lamb survival
and reproductive efficiency. Frequent monitoring of liveweight change allows producers to
review production efficiency and make better-informed management decisions to maximize
commercial gains. RFID technology, combined with the use of automated drafting and
weighing equipment, allows producers to record production information of individual sheep
and facilitate improved farm efficiency through a reduction in the labour required to record
liveweights manually. RFID technology was combined with an autodrafting weigh scale to
assist in the assessment of farm productivity, through the compilation of the liveweight data
of prime lamb (i.e. meat), dual purpose (i.e. meat and wool) and wool producing sheep.
The study was conducted over two years at a mixed cattle and sheep enterprise in north-east
NSW, using White Suffolk (prime lamb), dual purpose Mega Merino® and medium-fine
Merino genotypes. All lambs were tagged in the left ear at weaning with an Allflex® RFID
tag, and each lamb’s unique 16 digit RFID number was entered into Tru-Test Link 3000™
computer software along with the date of birth, and genotype of the lamb’s dam and sire. In
both years, lambs from the three genotypes were randomly allocated into three mixed-flocks,
managed under a rotational grazing system. Liveweight of the individuals in each flock were
recorded on 3 or 4 occasions during the year using an autodrafting weigh scale and a Destron
Fearing Axiz SB-1 RFID reader. A linear mixed model analysis (REML) was performed to
investigate the relationships between genotype, liveweight, lamb age and year. Liveweight
was used to estimate carcass value, with the dollar values entered into gross margins to
determine the most profitable genotype, which was the White Suffolk ($66.05/DSE), then the
Merino ($53.27/DSE) and Mega Merino® ($51.92/DSE).
While liveweight differences were found between years for the Merino and Mega Merino®
genotypes (P < 0.001), these could be attributed to differences in environmental conditions
(e.g. rainfall) and concomitant poorer nutrient value of the pasture. Understanding the link
between nutrient availability and liveweight could assist producers in decision making when
aiming to achieve higher weaning weights. In this study, the weighing of lambs was timed to
coincide with farm husbandry procedures such as weaning, vaccination and shearing, when
the lambs were required to be mustered to the sheep yards. However, the use of a ‘walk over
weigh’ scale with RFID capability located in the paddock could greatly increase the value of
the data generated on liveweight change, since the liveweight data could be regularly
obtained over an extended period for more detailed analysis of trends. Labour savings would
also be achieved as the animals would not require mustering to be weighed.
This research was supported by the Australian Wool Education Trust (AWET).
Saranika Talukder1, Kendra L Kerrisk1, Luke Ingenhoff2, Peter C Thomson3, Sergio C
Garcia1 and Pietro Celi1
1Dairy Science Group, Faculty of Veterinary Science, The University of Sydney, Camden
NSW 2570, Australia, Email
2Livestock Veterinary Teaching and Research Unit, Faculty of Veterinary Science, The
University of Sydney, Camden NSW 2570, Australia.
3ReproGen Animal Bioscience Group, Faculty of Veterinary Science, The University of
Sydney, Camden NSW 2570, Australia.
The development and application of an algorithm to assess the ability of infrared
thermography (IRT) device to determine the occurrence of ovulation in dairy cows was
investigated. Twenty cows were synchronized using a controlled internal drug release (CIDR)
and prostaglandinF2. Vulva and muzzle temperatures were measured every 12 h from CIDR
insertion to 32 h post PGF2 injection (to develop baseline temperatures) and then every 4 h
until ovulation occurred or 128 h after PGF2 injection. Thermal images obtained with a FLIR
T620 series infrared camera were analysed using ThermaCAM Researcher Professional 2.9
software. Cows were also monitored for behavioural signs of oestrus and colour changes of
OestrotectTM Heat Detector devices applied to the tail head of each cow 36 h after PGF
injection. Algorithms were developed by adjusting the vulva temperature (for ambient
temperature and humidity) of individual animals during each observation period and were
expressed as a deviation (D, oC) from the baseline temperature. Out of the 20 cows enrolled
in this study 12 (60%) ovulated. An IRT oestrus alert was defined using different thresholds
(D = 1oC, 1.25oC, and 1.5oC). Rates of detection accuracy and error rate depended upon the
chosen threshold level. At a threshold D = 1oC, the highest percentage (92%, n = 11) of
sensitivity and the lowest specificity (28%) was observed. On the other hand, D = 1.4oC
resulted in sensitivity = 75%, specificity = 57% and positive predictive values = 69%. The
mean (±SD) intervals between onset and end of IRT oestrus alert to ovulation were 30.7 ± 8.2
and 13.3 ± 7.7 h respectively. Ovulation occurred 24-48 h after the onset of the IRT oestrus
alert for 8 out of the 11 ovulated cows (73%). Although the sensitivities of the IRT alert were
higher than visual observation (67%) and oestrotect activation (67%), the specificities and
positive predictive value were lower than these two aids (i.e. the IRT over-predicted the
incidence of ovulation). Further studies are required with a larger sample size and sampling
throughout different seasons to determine the true potential of IRT as a tool for ovulation
The authors acknowledge the Dairy Research Foundation for its support and the investors of
FutureDairy; Dairy Australia, DeLaval, NSW DPI and The University of Sydney. We also to
acknowledge the staff of Corstorphine dairy for their assistance and Dr Aaron Cowieson,
Director of Poultry Research Foundation for providing the infrared camera. The primary
author has been supported by The University of Sydney International Postgraduate Research
Mark Trotter1 and David Lamb1
1CRC for Spatial Information & University of New England Precision Agriculture Research Group,
Armidale NSW 2351, Australia, Email
Active Optical Sensors (AOS) are a relatively new class of sensor. These handheld devices
direct a beam of light, usually comprising both red and near infrared wavelengths, onto the
plant canopy and an on-board detector records the returning radiation and calculates the
optical reflectance of the target canopy in those specific wavelengths. The key advantage of
the technology over passive optical sensors (like radiometers and spectrometers) is that they
contain their own light source and readings can be taken under any illumination conditions
including at night. The combination of red and near infrared reflectance responds strongly to
the photosynthetically active biomass (PAB) component of the canopy being scanned (the
green fraction).
To date these devices have been developed for use in the cropping industry, ostensibly for
inferring crop nitrogen levels, however recent research has demonstrated the potential for
applying the same technology to estimate the green fraction of pastures (Trotter et al., 2010).
In addition to collecting static ‘point’ readings, which is a constraint of plate meters and
capacitance probes, AOS do not require physical contact with the canopy so can be operated
from a moving vehicle to collect transect averages, or when coupled with a global positioning
system, provide pasture maps. The large instrument manufacturer, Trimble, have recently
released a relatively low-cost device (TGH - $600AUD) which has driven the price point
down to a level accessible to mainstream users.
Now that these sensors are more widely available there is interest in determining the accuracy
with which they can predict the green fraction of a pasture sward. Previous studies suggest an
accuracy of RMSE (validation) = 288 kg/ha GDM can be achieved over a range of seasons
(Trotter et al. 2010). New research using a single site progressive cut and scan approach
suggests that AOS are even more accurate under constrained conditions with correlation
coefficients of greater than 0.99 being achieved for both forage oats and fescue swards. The
equivalent prediction accuracy of RMSE (validation) = 132 kg/ha GDM. Further research is
required to determine how well these correlations hold up over different sites, soil
background reflectance, seasonal and phonological variations. There is also a need for a data
management system that can be integrated with the new sensors to facilitate their application
for biomass measurement and fodder budgeting by producers.
This work was funded by the CRC for Spatial Information (CRCSI), established and
supported under the Australian Government’s Cooperative Research Centres Programme.
Trotter MG, Lamb DW, Donald GE, Schneider DA (2010). Crop and Pasture Science. 61,
James P Underwood1, Mark Calleija1, Juan Nieto1, Salah Sukkarieh1, Cameron EF Clark2,
Sergio C Garcia2, Kendra L Kerrisk2 and Greg M Cronin2
1Australian Centre of Field Robotics, Faculty of Engineering and Information Technologies, The
University of Sydney, NSW 2006, Australia Email
2Dairy Science Group, Faculty of Veterinary Science, The University of Sydney, Camden NSW 2570,
Australia Email
Recent advances in sensing, automation and information technology in Australia and globally
have resulted in commercially successful field-robotic applications and the timing is
appropriate to consider what additional roles these systems may serve in the dairy industry
during the next decade, to decrease the cost of milk production. A first test of existing
unmanned ground vehicle (UGV) technology was conducted to assess current capability in
this domain. Three aspects were studied: 1) the response of dairy cows to the presence of a
UGV (see Clark et al. in current proceedings), 2) the ability to use the UGV in an
operationally relevant way (remote controlled herding) and 3) the ability of the on-board
sensors and algorithms to automatically detect and track dairy cows. Success in all three is a
pre-requisite for UGVs to have a role in automated herding of dairy cows. All three aspects
were addressed using 3D LiDAR data. Raw data were displayed remotely for operator
situational awareness and control from a distance of up to 200 m. The data were geo-
referenced and processed through a perception pipeline, to estimate global cow trajectories
(with no instrumentation on the cows), including instantaneous position, velocity and a record
of the paths they followed (Figure 1). This enables quantitative analysis and modelling of
their response to the UGV, and provides real-time feedback required for future development
of autonomous herding. A pre-set route was driven five times, amongst 20 cows in a 0.5 ha
paddock. 3D LiDAR data showed mean cow velocities away from the UGV of [0.06, 0.04,
0.02, 0.01, 0.01] m/s, indicating that dairy cows habituate quickly to UGV movement as per
human observations (Clark et al. in current proceedings). The cows were then herded three
times from the same 0.5 ha paddock by remotely controlling the UGV. For each experiment,
every cow was successfully herded through the gate without human intervention. Summary
statistics are shown in Table 1. The mean velocity of the cows was at most 0.1 m/s, which
was considered a ‘calm herding’ from human observation, with potential animal welfare
benefits such as reduced lameness. In conclusion, this experiment showed remote herding to
be possible and that 3D LiDAR sensing and existing perception algorithms are able to detect
and track cows. Further work is required to build on these findings and automate the process
of herding.
The authors would like to acknowledge the staff of both Corstorphine dairy and the Dairy
Science group for their excellent advice and support.
Figure 1. Automatic detection and tracking of cows using 3D LiDAR data.
Test Time
speed of
vehicle (m/s)
Average velocity of
cows (m/s, +/-
away/to vehicle)
deviation of cow
velocity (m/s)
12:29 456 0.61 0.10 0.27
23:35 752 0.50 0.08 0.27
22:01 642 0.47 0.05 0.21
Table 1: Key performance indicators for three 20 cow herding experiments.
Ashleigh Wildridge1, Alistair Smith1, John Broster2, Geraldine Lammers1 and Jane Heller1
1 Charles Sturt University, School of Animal and Veterinary Sciences, Wagga Wagga 2650,
2 Charles Sturt University, School of Agricultural and Wine Sciences, Wagga Wagga 2650,
A herd of 23 mixed aged Hereford cows on a pasture based system were fitted with proximity
loggers (PL’s) to record the number and length of contacts between members of the herd.
Contacts were recorded when two or more PL’s came within a pre-set distance from each
other allowing measurements of herd contact structures to be collected.
Combined proximity and Global Positioning System (GPS) data were used to determine the
distance that a PL can receive a signal from another PL (Figure 1). This enabled precise
geographical coordinates to determine the approximate distance two animals are apart when a
contact is recorded.
The PL’s were fitted for 26 days and data were collected and used for analysis on all days
exclusive of when the cattle were mustered. The data were initially analysed descriptively.
Graphical representations of the data revealed peaks in both the number and duration of
contacts in a similar pattern over time (Figure 2, where each line represents the contacts of a
cow in the herd).
Figure 1. A Hereford cow fitted with a proximity logger
Initial analysis of the data does not appear to reveal a connection between contact patterns
and the feeding regime of the cattle. A full analysis of the data is yet to be completed to
reveal the statistical significance of age, pregnancy status, temperature and rainfall. Analysis
of these variables may be useful to predict peaks in social connectedness that may have the
potential to drive the transmission of infectious pathogens within this type of herd structure
(e.g. E. coli O157:H7).
Figure 2. (a) showing the median duration of contacts and (b) showing the
median number of contacts.
Lauren R Williams1, Russell D Bush1, Robert J Kilgour2, Mark G Trotter3 and Greg M Cronin1
1Faculty of Veterinary Science, The University of Sydney, Camden NSW 2570, Australia Email
2NSW Department of Primary Industries, Trangie NSW 2823, Australia
3University of New England, Precision Agriculture Research Group, Armidale NSW 2351, Australia
Beef production in Australia usually involves extensive grazing characterised by minimal
handling of the animals. Cattle may be fearful during handling by humans, and due to their size
and strength, fearful cattle have increased risk of injury and reduced welfare. Fearful cattle also
negatively impact profit, through decreased labour efficiency and increased production costs.
Temperament is an animal’s behavioural response to handling and is related to growth, feed
conversion efficiency and carcass quality. Yard tests have been developed to assess temperament
via measures of agitation and fear, but few studies have investigated whether yard tests predict
beef cattle behaviour under extensive grazing conditions.
This study utilised remote sensing technology to investigate the effects of temperament on the
behaviour and performance of extensively grazed beef cattle. Previously, Cronin et al. (2011)
reported that the behaviour cattle and sheep in the paddock was not altered by wearing a GPS
tracking neck collar and an accelerometer attached to the rear leg. For the present experiment,
the temperament of 64 Angus weaner heifers was assessed in two yard tests: flight time and fear
of humans, and 12 “Good” and 12 “Poorer” Temperament animals were selected. Three heifers
from each temperament treatment were randomly assigned to wear a neck collar with a GPS
tracker and an IceTag3D™ accelerometer on the rear leg. Another three heifers per treatment
were selected as controls. The 24 heifers were weighed and relocated to a 49 ha paddock for 51
days where their behaviour (posture and activity) and spatial distribution were measured using
four methods – GPS location points, GPS-based behaviour, IceTag3D™ activity and direct
behaviour observation. Heifer performance was determined by live weight change.
There was no effect of temperament on posture, activity or spatial utilisation in the paddock.
Heifers used some areas of the paddock preferentially and habitually congregated around water
points, which is typical of spatial distribution patterns in beef cattle. There was a strong effect of
time on activity (IceTag3D™ step count P < 0.001 and movement index P < 0.001) but no
differences between treatments. The local, daily weather conditions influenced standing posture
(P = 0.021), grazing behaviour (P < 0.001), activity (step count P < 0.001, motion index P <
0.001, GPS velocity P < 0.001) and spatial distribution. The herd participated in diurnal
behavioural activities consistent with observations of grazing beef cattle. Higher activity levels
and larger home range were recorded during the first week compared with subsequent weeks,
which is likely due to investigation of a novel environment. Temperament was not associated
with live weight change (P = 0.138). Although cattle experience acute stress during handling
procedures, extensively grazed cattle are generally removed from frequent human interaction and
unlikely to suffer compromised productivity from stress due to fear of humans. While the use of
remote sensing technology was found viable for measuring the behaviour of beef cattle on
pasture, yard temperament assessments do not appear to be reliable predictors of behaviour or
productivity of beef cattle grazing pasture. Between-animal differences in behaviour and
productive characteristics of animals in larger herds or more excitable breeds require further
Cronin GM, Williams LR, van der Smagt NE, Bush RD, Trotter MG 2011. Does wearing remote
monitoring technologies alter livestock behaviour in extensive grazing systems? Paper
presented at the 2011 Spatially Enabled Livestock Management Workshop, Eds. M
Trotter and LA González. In: Proceedings of the Biennial Conference of the Australian
Society for Engineering in Agriculture (SEAg), 28-30 Sept 2011, Surfers Paradise,
Queensland, p. 88.
ResearchGate has not been able to resolve any citations for this publication.
To describe the influence of the dingo (Canis lupus dingo) on the past, present and future distributions of sheep in Australia. The role of the dingo in the rise and fall of sheep numbers is reviewed, revised data are provided on the present distribution and density of sheep and dingoes, and historical patterns of sheep distribution are used to explore the future of rangeland sheep grazing. Dingoes are a critical causal factor in the distribution of sheep at the national, regional and local levels. Dingo predation contributed substantially to the historical contraction of the sheep industry to its present-day distribution, which is almost exclusively confined to areas within fenced dingo exclusion zones. Dingo populations and/or their influence are now present and increasing in all sheep production zones of Australia, inclusive of areas that were once 'dingo free'. Rangeland production of wool and sheep meat is predicted to disappear within 30-40 years if the present rate of contraction of the industry continues unabated. Understanding the influence of dingoes on sheep production may help refine disease response strategies and help predict the future distribution of sheep and their diseases.
A number of behavioural traits associated with the maternal instinct of sheep were studied under paddock conditions over a range of breeds, age, climate, nutrition and locations in five lambings in south Western and north Western Australia. There was very considerable variation between ewes in any group in the occurrence and timing of each trait. There were few differences due to breed, age of ewe, nutrition, climate or location in the behaviour of the ewe before and during parturition. Nearly all the differences that occurred were in the frequency of pawing the ground before and during labour. Fewer Southdowns and Merinos pawed the ground than ewes of other breeds. Also this trait occurred more in summer than in winter in south Western Australia, and less here than in north Western Australia.
  • K Betteridge
  • C Hoogendoorn
  • D Costall
  • M Carter
  • W Griffiths
Betteridge K, Hoogendoorn C, Costall D, Carter M, Griffiths W (2010) Comp. Elect. Agric. 73, 66-73.
In: Precision tools for improving land management
  • A D Mackay
  • A S Palmer
  • A P Rhodes
  • G K Cooper
  • L Grant
  • B Withell
Mackay AD, Palmer AS, Rhodes AP, Cooper GK, Grant L, Withell B (2010) In: Precision tools for improving land management. (Eds L D Currie and P Loganathan). Occasional Report No. 14. FLRC, Massey University, Palmerston North, New Zealand pp 79-87.
  • S Cloete
  • A Scholtz
Cloete S, Scholtz A (1998) Animal Production Science, 38(8), 801-811.
  • J Everett-Hincks
  • K Dodds
Everett-Hincks J, Dodds, K (2008) Journal of Animal Science, 86(14 suppl), E259 E270.
  • K Geenty
  • F Brien
  • G Hinch
  • R Dobos
  • G Refshauge
  • M Mccaskill
  • A Ball
  • R Behrendt
  • K Gore
  • D Savage
Geenty K, Brien F, Hinch G, Dobos R, Refshauge G, McCaskill M, Ball A, Behrendt R, Gore K, Savage D (2013) Animal Production Science REFERENCES
Tropical Agronomy Memorandum No 8
  • J C Tothill
  • Jng Hargreaves
  • R M Jones
Tothill JC, Hargreaves JNG, Jones RM (1978) Tropical Agronomy Memorandum No 8. Virgona J, Hackney B (2008) Proc. 14 th Agronomy Conference, 2008. REFERENCES
Counting the cost: impact of invasive animals in Australia' (Cooperative Research Centre for Pest Animal Control: Canberra)
  • R Mcleod
  • A Norris
McLeod R, Norris A (2004) 'Counting the cost: impact of invasive animals in Australia' (Cooperative Research Centre for Pest Animal Control: Canberra).
Australian and New Zealand Spatially Enabled Livestock Management Symposium
  • S Anderson
  • M Trotter
  • R Haling
  • C Edwards
  • C Guppy
  • D Lamb
Anderson S, Trotter M, Haling R, Edwards C, Guppy C, Lamb D (2012) Australian and New Zealand Spatially Enabled Livestock Management Symposium, 2012.