Archivos Latinoamericanos de Producción Animal. 2024. 32 (1)
Effect of additional herbage areas on grazing dairy cows in commercial
farms: A GPS and LoRaWAN based case study on herbage intake and
milk yield
Received: 20230209. Reviewed: 20230517. Accepted: 20240305
1Chalcombe Ltd, Mountwood House, Biddenfield Lane, Shedfield, Southampton, Hants, SO32 2HP, United Kingdom.
2Rothamsted Research Ltd, West Common, Harpenden, Herts AL5 2JQ, United Kingdom.
3Correspondence author: M. Jordana Rivero jordana.riveroviera@rothamsted.ac.uk Net Zero & Resilient Farming, Rothamsted Research Ltd,
North Wyke, Okehampton, Devon, EX20 2SB, United Kingdom.
4University of Nottingham, School of Biosciences, Sutton Bonington Campus, Loughborough, Leics, LE12 5RD, United Kingdom.
37
A. Tom Chamberlain1
Abstract: Although grazing systems are widely used for lactating dairy cattle, feed intake is generally lower than in a
fully housed system even when the feed quality and animals’ nutritional requirements are similar. Running trials in
commercial settings, long range wide area network (LoRa) technology and GPS positioning were used to track animal
activity and position to investigate whether allocating additional herbage at a time linked to the cow’s behaviour
could increase productivity. In particular, we examined whether additional herbage allowance increases daily herbage
intake and milk production without compromising grazing efficiency. Fourteen trials were undertaken on eight
commercial dairy herds in 2019, 2020 and 2021 generally with cows in mid to late lactation. The ‘GrazeMore’ addi
tional grazing was compared to a standard daily herbage allocation. The ‘GrazeMore’ treatment period always
followed an initial control period, sometimes with a subsequent control period. The composition of the grazing groups
was largely consistent over the duration of each trial, enabling the responses to be compared directly. Cow location
could be tracked while they grazed and their grazing activity determined allowing the timing of additional grazing
allocation to be linked to grazing behaviour. Responses to additional ‘GrazeMore’ pasture allocations were
inconsistent. The pattern of grazing was changed, but increased intakes during day grazing periods were sometimes
balanced by reduced intakes in the following night periods, suggesting that factors other than the quantity of herbage
on offer and the timing of its allocation during day grazing were responsible for limiting total 24 h herbage intake and
milk production. Synchronising additional pasture allocation with grazing behaviour does not always increase
herbage intake and milk production. We have also highlighted some of the challenges encountered while conducting
research in commercial settings, as opposed to controlled experiments in research facilities.
Keywords: grazing management; Long range wide area network; grazing efficiency; grazing behaviour; Global
Positioning System.
https://doi.org/10.53588/alpa.320104
Rothamsted Research Ltd, West Common, Harpenden, Herts AL5 2JQ, United Kingdom
Andrew Mead2
Efecto de áreas de pasto adicionales en el pastoreo de vacas lecheras
en granjas comerciales: Un estudio de caso basado en GPS y
LoRaWAN sobre el consumo de forraje y la producción de leche
Resumen: Aunque los sistemas de pastoreo se utilizan ampliamente para el ganado lechero lactante, el consumo de
alimento es generalmente menor que en un sistema completamente estabulado, incluso cuando la calidad del
alimento y los requisitos nutricionales del animal son similares. Se utilizaron pruebas en entornos comerciales,
tecnología de red de área amplia (LoRa) de largo alcance y posicionamiento GPS para rastrear la actividad y la
posición de los animales e investigar si la asignación de forraje adicional en un momento relacionado con el
comportamiento de la vaca podría aumentar la productividad. En particular, examinamos si la asignación adicional
de forraje aumenta el consumo diario de forraje y la producción de leche sin comprometer la eficiencia del pastoreo.
Se realizaron catorce ensayos en ocho rebaños lecheros comerciales en 2019, 2020 y 2021, generalmente con vacas en
lactancia media o tardía. El pastoreo adicional 'GrazeMore' se comparó con una asignación diaria estándar de
forraje. El período de tratamiento con 'GrazeMore' siempre siguió a un período de control inicial, a veces con un
período de control posterior. La composición de los grupos de pastoreo fue en gran medida consistente durante la
M. Jordana Rivero3 J. Michael Wilkinson4
38
Introduction
Rivero et al.
Global competition for limited arable land that can
produce either human food or animal feed highlights a
key role for grazed livestock in producing human food
from less productive pastureland that cannot be
cultivated to produce crops (Wilkinson and Lee, 2018).
Prediction and management of herbage growth,
allocation of herbage to grazed livestock and
optimising production per animal and per hectare are
critical challenges facing pasturebased farmers
(Wilkinson et al., 2020). A major objective in pasture
management is to provide an adequate daily supply of
dense herbage comprising young vegetative growth
throughout the grazing season to meet the nutritional
requirement of the grazing animal (McGilloway et al.,
1996). This must be achieved despite potentially large
variations in the rate of plant growth due to season,
Efeito de áreas adicionais de forragem em vacas leiteiras pastando em fazendas
comerciais: Um estudo de caso baseado em GPS e LoRaWAN sobre consumo de
forragem e produção de leite
Resumo: Embora os sistemas de pastoreio sejam amplamente utilizados para bovinos leiteiros em lactação, o
consumo de ração é geralmente menor do que em um sistema totalmente alojado, mesmo quando a qualidade da
ração e as exigências nutricionais do animal são semelhantes. A realização de testes em ambientes comerciais, a
tecnologia de rede de longa distância (LoRa) e o posicionamento GPS foram usados para rastrear a atividade e a
posição dos animais para investigar se a alocação de forragem adicional em um momento ligado ao comportamento
da vaca poderia aumentar a produtividade. Em particular, examinamos se a oferta adicional de forragem aumenta a
ingestão diária de forragem e a produção de leite sem comprometer a eficiência do pastoreio. Quatorze ensaios
foram realizados em oito rebanhos leiteiros comerciais em 2019, 2020 e 2021, geralmente com vacas no meio ou no
final da lactação. O pastoreio adicional ‘GrazeMore’ foi comparado com uma alocação diária padrão de forragem. O
período de tratamento do ‘GrazeMore’ sempre seguiu um período de controle inicial, às vezes com um período de
controle subsequente. A composição dos grupos de pastoreio foi bastante consistente ao longo de cada ensaio,
permitindo que as respostas fossem comparadas diretamente. A localização das vacas poderia ser rastreada
enquanto pastavam e a sua atividade de pastoreio determinada, permitindo que o momento da atribuição adicional
de pasto fosse ligado ao comportamento de pastoreio. As respostas às alocações adicionais de pastagens
‘GrazeMore’ foram inconsistentes. O padrão de pastoreio foi alterado, mas o aumento do consumo durante os
períodos de pasto diurno foi por vezes compensado por consumos reduzidos nos períodos noturnos seguintes,
sugerindo que outros fatores para além da quantidade de forragem oferecida e o momento da sua distribuição
durante o pastoreio diurno foram responsáveis por limitar o consumo total de forragem. Consumo de forragem em
24 h e produção de leite. Sincronizar a alocação adicional de pastagens com o comportamento de pastoreio nem
sempre aumenta o consumo de forragem e a produção de leite. Também destacamos alguns dos desafios
encontrados durante a realização de investigação em ambientes comerciais, em oposição a experiências controladas
em instalações de investigação.
Palavraschave: manejo de pastagem; Rede de longa distância; eficiência de pastoreio; comportamento de pastoreio;
Sistema de Posicionamento Global.
duración de cada prueba, lo que permitió comparar las respuestas directamente. Se podría rastrear la ubicación de
las vacas mientras pastaban y determinar su actividad de pastoreo, lo que permitiría vincular el momento de la
asignación de pasto adicional al comportamiento de pastoreo. Las respuestas a las asignaciones adicionales de
pastos de 'GrazeMore' fueron inconsistentes. Se cambió el patrón de pastoreo, pero el aumento del consumo
durante los períodos de pastoreo diurno a veces se vio compensado por un consumo reducido en los períodos
nocturnos siguientes, lo que sugiere que factores distintos de la cantidad de forraje ofrecido y el momento de su
distribución durante el pastoreo diurno fueron los responsables de limitar el total de pastos. Consumo de forraje 24
h y producción de leche. Sincronizar la asignación de pastos adicionales con el comportamiento de pastoreo no
siempre aumenta el consumo de forraje y la producción de leche. También hemos destacado algunos de los desafíos
encontrados al realizar investigaciones en entornos comerciales, a diferencia de los experimentos controlados en
instalaciones de investigación.
Palabras clave: manejo del pastoreo; Red de área amplia de largo alcance; eficiencia del pastoreo; comportamiento
de pastoreo; Sistema de Posicionamiento Global.
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
39
Additional herbage areas on grazing dairy cows
temperature, water and nutrient supply. In addition,
the feed intake of the dairy cow varies with live
weight, milk yield and stage of lactation (Chamberlain
and Wilkinson, 1996). The diurnal grazing pattern
emerges from a series of grazing decisions such as
‘when’ to begin, the intensity (i.e., herbage intake rate),
‘what’ frequency and ‘how’ to distribute the grazing
events in time (Gregorini, 2012) and, for a milking cow,
what other activities she has to perform such as visiting
the milking parlour. Therefore, the causes for a grazing
cow to initiate and terminate the grazing event are
complex and multifactorial (Chilibroste et al., 2015).
Although grazing plays a central role in dairy cow
nutrition in many regions of the world, low herbage
intake by the grazing animal is a major limitation to
herd productivity (Bargo et al., 2003). Target levels of
daily herbage allowance in temperate intensive
perennialryegrassbased dairy grazing systems (e.g.,
New Zealand, France) are 20 to 30 kg DM/head to
support a daily intake of 15 to 17 kg DM/head
(Wilkinson et al., 2020). In contrast, typical daily DM
intake by dairy cows given total mixed rations is some
33 % higher than that of cows given grazed pasture
(Wilkinson and Lee, 2018). Further, the intake response
by grazed cows to increased daily herbage allowance is
at the expense of efficiency of pasture utilisation
(Baudracco et al., 2010). Previous research by
MacDonald et al. (2001) in New Zealand examined
farmlets with varying stocking rates and demonstrated
that milk yields increase as stocking rate and grass
supply rise. However, it was observed that grazing
efficiency decreases as cows become more selective.
Thus, a balance has to be struck between maximising
herbage intake and maintaining an acceptable
proportion of available herbage that is actually
consumed by the animal. The area of pasture to be
allocated to a herd of cattle for grazing varies
according to size of herd, quantity of herbage available
and planned grazing intensity (proportion of the sward
to be removed). Having calculated the appropriate area
of land to be allocated, traditionally animals are
introduced to new pasture at a time convenient to the
herd manager. In the case of rotationally grazed
swards, this occurs after fences have been moved to
allow access by livestock to a new area of ungrazed
herbage. Timing of new pasture allocation to dairy
cows normally occurs when the herd is withdrawn
from the field to be milked. Consequently, allocation of
new herbage may not be in synchrony with the
natural pattern of grazing behaviour (Abrahamse et al.,
2009; O’Driscoll et al., 2010).
Development of global positioning systems (GPS)
(Turner et al., 2000; Rivero et al., 2021), sensors
(accelerometers) capable of monitoring/measuring
animal behaviour, and longrange wide area networks
(LoRaWAN) (Miles et al., 2020) introduces the
possibility of controlling grazing on commercial dairy
farms. GPS data can be used to determine when the
cows are in the grazing field and accelerometers fitted
to the cow can model when the cow is grazing. Such
data can be collected in realtime over the LoRa
network and processed in the ‘cloud’ to determine
when onfarm actions (such as gate opening) should be
triggered with information relayed back through the
LoRa network to infield actuators.
Preliminary visual observations (ATC) of cows in
commercial herds showed an intensive period of
grazing at the start of the grazing bout followed by a
period of rumination, confirmed by the findings
reported by Sheahan et al., (2013). The cows then
commenced a second grazing bout ranging over
ground they had already grazed but with less intense
grazing and more sward selection to avoid
contaminants, e.g., dung pats. This behaviour of
modifying the grazing (e.g., biting rate) due to the
presence of contaminants has been previously re
ported by Bao et al., (1998) who concluded that
selective grazing exists due to the presence of dung
and is conditioned by dung distribution. Ad hoc trials
in 2017 allocating fresh grazing at the start of the
secondary bout led to the herd of cows rapidly moving
to the new area followed by a bout of intensive grazing
(Chamberlain and Kodam, 2019) . The objective of
these trials was to determine if the starting time of the
secondary grazing bout could be determined
automatically, and a new grazing allocation opened up
without any human intervention. Such automation
would be needed for any commercial application.
The work reported here comprised a feasibility study
to test two hypotheses under commercial farm settings:
1) Allocation of a new area of herbage to be grazed can
be synchronised with natural grazing behaviour,
determined by GPS and accelerometer data collated
through LoRaWAN, and 2) Provision of additional
herbage when a new grazing bout is imminent
increases daily herbage intake and milk production in
commercial dairy herds without compromising
grazing efficiency. In addition, the suitability of
LoRaWAN technology as a tool to collate data
effectively was assessed. We have also highlighted
some of the challenges encountered while conducting
research in commercial settings, as opposed to
controlled experiments in research facilities.
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
40
Materials and Methods
The ethical implications of the project were con
sidered by ATC, who was a Named Veterinary
Surgeon with the UK Home Office. Procedures that
directly involved animals were the fitting and removal
of neck collars. Commercially available weighted cow
neck collars were used (Kerbl, UK) with a gross weight
of 850 g, and these were considered indistinguishable
from other commercial collars such as activity
monitors. The procedures were therefore classed as
‘nonregulated’ under the Animals (Scientific
Procedures) Act 1986 (ASPA), and further ethical
consideration was not applicable.
Implementation of Grazing Trials
Fourteen trials were undertaken on eight com
mercial dairy units between April and September 2019,
JulyAugust 2020 and AprilMay 2021 in the counties
of Dorset, Shropshire, Somerset, Wiltshire, located in
the south of England (UK). The country is
characterised by a temperate climate. The accumulated
annual precipitation for the region where the trials
were conducted is approximately 720 mm pa (evenly
spread through the year), with an average minimum
and maximum summer air temperature of 8 and 21 °C,
respectively (Met_Office, 2023). Pastures are usually
mixed swards comprising perennial ryegrass (Lolium
perenne L.) and white clover (Trifolium repens L.).
Eight collaborating farmers (different dairy units,
see Table 1) allowed temporary installation of a single
LoRaWAN ‘gateway’ on their farm to enable grazing
behaviour to be monitored remotely. Dairy units were
initially selected by phone interview on the basis of
their grazing infrastructure, i.e., rotational paddock
grazing systems, good water supply and trough pro
vision, temporary fencing within paddocks, little or no
buffer feeding, regular plate meter readings and milk
recording of individual cow’s yield. Each unit
operated separate ‘day’ and ‘night’ grazing allocations
with twicedaily milking. Day fields were grazed
between the end of morning milking and the start of
evening milking and night fields were grazed between
the end of evening milking, and the start of morning
milking (Sheahan et al., 2013).
The ‘Control’ treatment consisted of the usual pas
ture allocation for the farm, herd and stage of the
grazing season as determined by the farm
management. To test hypothesis 2, an additional area
of pasture for the ‘GrazeMore’ treatment was allocated
daily during the day grazing period, at the start of the
herd’s second major grazing period (usually mid
morning), calculated as a percentage of the initially
allocated amount; the percentage increase varied
between trials (Table 1). The timing for the allocation
of the new ‘GrazeMore’ area was not set at a fixed time
of the day (see section 2.4.). Both treatments were
applied to the same groups of animals at different
times (e.g., ‘OffOn’: ‘Control’ then ‘GrazeMore’ or
‘OffOnOff’: ‘Control’, then ‘GrazeMore’ and then
‘Control’ again; Table 1). In each trial, the initial area of
pasture allocated for the day grazing period, which
commenced after the morning milking, was
determined based on existing farm practices including
levels of pregrazing herbage mass and expected
residual covers.
As the trials were run on commercial dairy farms,
group selection was initially conservative, where
possible, to limit the impact of any milk yield
depression on herd output. Where cows were grouped
within the herd for management purposes, the ‘Low’
yielding group was generally selected. The milk yield,
stage of lactation and supplemental feeding allocation
at the start of each trial are shown in Table 2.
Supplementary feeding was a pelleted compound feed
offered twice a day in the milking parlour that was
formulated to 16 % or 18 % crude protein in the fresh
weight. The supplementary feed allocation was
determined by the milking parlour software based on
individual cow milk yields at the start of the trial and
was held at a fixed amount throughout the trial. To
avoid milk yields becoming too separated from
supplementary feed allocations the trials were limited
to 20 days and then the parlour software reset the
feeding allocations.
Identification of Grazed Fields and Individual
Animals
Selected farms were geomapped using a com
bination of Scribble Map software (52 Stairs Studio Inc.
Ontario, Canada) and ground truthing with a
handheld GPS device (iPhone 6 augmented with a
Garmin Glo2 portable GPS/GLONASS sensor) to
identify latitude/longitude coordinates of the apex
points of individual fields.
Eight cows in each trial, selected from each herd
(Table 1) as being representative in terms of age, stage
of lactation and milk yield, were each fitted with
weighted (500g) neck collars (Kerbl UK Limited,
Oakham, UK), on which were mounted longrange
(LoRa) nodes developed in a previous project
(InnovateUK project, 132355). The nodes weighed
approximately 350 g and contained a battery (Saft
33600, 3.6V, 17Ah, Saft Groupe SAS, LevalloisPerret,
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
Rivero et al.
41
Figure 1. Neck collar and ‘node’ box for GPS transmission.
France) and a custombuilt printed circuit board
(Figure 1). The LoRa nodes recorded GPS position and
accelerometer data.
Cow Behaviour and Milking Times
Using the node accelerometer data, cloudbased
algorithms were developed to identify a change from
the intensive grazing that is seen when cows enter a
new grazing area to a ruminating behaviour (around
60 to 120 minutes post field entry). Cows will generally
ruminate for 30 to 90 minutes before starting to graze
again (Figure 2). A logistic regression was fitted
relating the accelerometer reading variability to
observed animal eating behaviour throughout each 120
second interval to predict the probability that a cow was
grazing (pGraze) as described by Chamberlain and
Kodam (2019). GPS and pGraze data were (nominally)
reported for each period of 120 seconds.
Individual cow position and behaviour data were
combined to summarise herdlevel activities. The GPS
fix for each cow was related to the field apex database
to determine in which field the cow was grazing. Once
five of the eight cows were in a particular field it was
concluded that the entire group had entered that field.
Milking times were defined at the period when five
out of eight cows were in the milking complex area.
Similarly changes in grazing behaviour were
examined at the individual level and once five out of
eight cows had changed behaviour then it was
concluded the group had changed behaviour. This
removed issues with outliers due to particular animals
being kept in the milking buildings for husbandry
interventions or unusual behaviours such as fighting
or oestrus. If monitoring halted for any of the eight
animals, the decision threshold was modified
accordingly, always ensuring that at least 60 % of the
monitored animals had changed behaviour before
decisions were made.
Table 1. Collaborating dairy units, dates of trials, dairy cows herd size, treatment allocation design and numbers of days in each
treatment period.
Dairy Trial Start End Grazing group size Annual milk yield Design2 Number of days3Comments4
Unit No. date date (type1) (litres/cow)
2019 Control GrazeMore Control
A 1 26 April 10 May 100 (lows) 9300 OffOn +20 % 12 7 Inadequate
grazing
availability
B 2 26 April 10 May 250 (whole herd) 10000 OffOn + 20 % 8 6 Inadequate
grazing
availability
No individual
cow milk
yield data
C 3 20 May 7 June 150 (lows) 9000 OffOn +20 % 8 8
D 4 20 May 7 June 200 (whole herd) 9000 OffOn + 20 % 13 4
Variable
pasture
quantity and
quality. No
individual
cow milk
yield data
E 5 17 June 5 July 190 (whole herd) 7000 OffOn +20 % 9 10 No
individual
cow milk
yield data
A 6 17 June 5 July 100 (lows) 9300 OffOn +20 % 11 5
C 7 22 July 3 August 150 (lows) 9000 OffOn +20% n/a n/a4 Inadequate
grass due to
drought
Additional herbage areas on grazing dairy cows
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
42
Table 1. Collaborating dairy units, dates of trials, dairy cows herd size, treatment allocation design and numbers of days in each
treatment period. (continuación)
Dairy Trial Start End Grazing group size Annual milk yield Design2 Number of days3Comments4
Unit No. date date (type1) (litres/cow)
2019 Control GrazeMore Control
D 8 22 July 3 August 200 (whole herd) 9000 OffOn +20 % n/a n/a4 Inadequate
grass due to
drought
E 9 2 Sept 16 Sept 190 (whole herd) 7000 OffOn +20 % 9 6 No
individual
cow milk
yield data
2020
F 10 6 July 25 July 90 (lows) 10200 OffOn +40 % 7 10 Limited gra
zing availa
bility
No
individual
cow milk
yield data
G 11 3 August 23 August 190 (lows) 8900 OffOn +40 % 7 11
2021
G 12 12 April 2 May 240 (lows) 8900 OffOn +30 %Off 11 7 3
G 13 12 April 2 May 145 (mids) 8900 OffOn +30 %Off 12 6 4
H 14 10 May 30 May 250 (whole herd) 8500 OffOn +30 %Off 10 7 1 Intermittent
grazing data
capture due
to terrain
1 ‘Lows’ = cows in late lactation; ‘mids’ = cows in mid lactation.
2 ‘On’ = additional day grazing allocated in ‘GrazeMore’ period as percentage of total day grazing allocation in ‘Off” period
3 ‘Off’‘On’ or ‘Off’‘On’‘Off.
Table 2. Main performance and feeding characteristics of the cows selected in trials used for analysis of feed intakes or milk
yields (from parlour software or commercial milk recording).
Milk yield (l/day) in Days in milk at start of trial Supplementary feed allocation
group at start of trial (kg/day) at start of trial
Trial No Group Stage1 Median IQR12 Median IQR Median IQR
3 150 lows 21.1 8.9 241 103 2.6 3.1
5 190 whole herd 14.6 6.5 283 35 2 0
6 100 lows 15.4 8.4 270 63 3.5 1
9 190 whole herd 22.1 10.2 18 15.5 6 0
11 190 lows 20.1 6.8 258 97 3.2 3.8
12 240 lows 23 7.5 228 79.5 2.3 3
13 145 mids 29 7.3 127 66 2.6 3.3
14 250 whole herd 29.1 10.8 150 136 5.6 4.3
1‘lows’ = cows in late lactation; ‘mids’ = cows in mid lactation.
2IQR: inter quartile range.
Movement of Fences and Daily Herbage Allocations
A recoil gate (Gallagher UK) and battery/solar
powered opening device (modified Batt Latch, Novel
Ways Limited, Taupo, New Zealand) were fitted to
sections of paddock fencing to facilitate opening and
allow herd access to the additional grazing area each
day during the ‘GrazeMore’ treatment period of each
trial. The trigger of gate opening varied according to
milking times and amounts of grazing allocated in the
initial area but was constrained to be at least 90
minutes before the start of (planned) afternoon
milking and was triggered by changes in the
behaviour of the eight selected (monitored) cows
(detailed information on grazing allocations is shown
in Chamberlain et al., (2022b)).
Sending outwardbound messages to the BattLatch
gate opener was unreliable due to lineofsight issues
in many fields. As an alternative a solarpowered SMS
Modem enabled BattLatch was installed in Trial 10,
but the power draw required for regular checks for
trigger messages necessitated frequent retrieval and
battery recharging. In Trials 9 to 14, the gate opener
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
Rivero et al.
43
Figure 2. Typical plot of pGraze against time for one cow on one day in one trial (Cow 901, 8 Aug 2020, Trial 11) during the
‘Control’ treatment. pGraze shown as a rolling 10 min average to aid graphical clarity, timing of milking determined from cow
location, timing of ‘Gate Trigger’ determined from activity analysis.
battery recharging. In Trials 9 to 14, the gate opener
was programmed manually. Node data was
downloaded daily at 8 pm, analysed and assessed that
evening. The next day, the final setting of the gate
trigger time was manually entered into the standard
BattLatch firmware.
Field sizes were measured and grazing plot
allocations were calculated using anticipated herbage
intakes and grazing group sizes (determined from
farm records). Grazing areas were defined with electric
fencing, and herbage residuals were measured after
each grazing for both day and night fields (see
Supplementary Table 1 and 2 in Chamberlain et al.,
2022d).
Herbage Mass, Dry Matter Intake and Grazing
Efficiency
Quantities of herbage DM mass per hectare were
estimated with a rising plate meter (F200 model,
Jenquip, Feilding, New Zealand) daily during each
trial to provide estimates of both pre and postgrazing
for all allocated day and night grazing areas. At each
measurement occasion, a total of 40 readings were
taken at random across the paddock (excluding dung
patches) and averaged. Duplicate sets of 40 readings
were taken and if the average values of the two sets
differed by more than 200 kg DM/ha a further
replicate set of 40 readings was collected. Herbage DM
intake per cow was estimated using the pre and post
grazing herbage DM mass estimates, the area allocated
for grazing, and the number of animals grazing each
day (Lukuyu et al., 2014). Grazing efficiency was
calculated from herbage DM allowance per animal and
DM consumed over each grazing period.
Milk Yield
Milk yield data were collected from herds that used
GEA (GEA Group AG, Düsseldorf, Germany)
recording equipment in milking parlours (herds A, C,
G, H in Table 1). Data were collated from all the cows
grazing in the groups (including the eight selected
with GPS collars as shown in the Results section;
detailed information can be found in Chamberlain et
al., 2022d). Software routines (Visual Basic, Microsoft
2016) were written to extract daily milk yield data
from data feeds and converted to a standard format
showing daily milk yields for each cow in the trial.
Data were filtered as follows:
•Daily records were set to missing, but animals still
retained for the other days where yields wer
e
> 200 % or < 50 %
of the previous day’s yield,
accounting for double or missing recordings at one
milking,
•Data associated with animals with an unknown age/
lactation were removed,
Additional herbage areas on grazing dairy cows
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
44
•Data for animals from recording more than 405 days
after calving or less than 10 days post calving were
removed, and
•Data for animals with missing milk yields on two or
more consecutive days were removed.
Most of the trials were carried out on cows in mid
to late lactation (Table 1) as the authors did not feel
they could ask the participating farmers to expose
their best, highest yielding cows to a novel trial
procedure. As such, most animals would have been
past peak yield and yields would be naturally falling.
As the ‘GrazeMore’ treatment always followed the
‘Control’ period, this could have consistently reduced
milk yields in the ‘GrazeMore’ period. To counter this
effect, milk yields were corrected for stage of lactation
as follows. Daily individual cow milk yields were
determined for the 60d period before each trial began
(the GEA software system only holds individual milk
yields each milking for 60 days). Primiparous and
multiparous animals were then analysed separately.
Milk yields through lactation were referenced to the
milk yield at 150 days (around peak yield). To reduce
the impact of random variation between days around
150 days, the average milk yield for each animal three
days before and 3 days after 150 days was averaged to
give a 7day average for each parity group to
determine the reference yield. Average yields for each
parity group at other stages of the lactation were then
expressed as proportions of the 150th day reference
yield and these proportions were used
multiplicatively to correct individual daily milk yields
for all animals during the trial.
Experimental Design
To enable the assessment of hypothesis 1 trials were
implemented using commercial dairy herds to ensure
that the trial results were representative of potential
impacts on other commercial dairy herds. Animals
were exposed to each treatment for a number of days
to enable the elimination of any carryover effects of the
previous treatment. A ‘crossover’ design approach
using pairs of commercial dairy units in close
geographical proximity and with similar
characteristics (environments, animals) was not
possible as suitable pairs of farms could not be
identified from the pool of available candidates.
Instead, ‘an ‘OffOnOff’ treatment structure was
selected: period of time (days) under the Control
treatment, followed by a period of time (days) under
the GrazeMore treatment, followed by a further period
of time (days) under the Control treatment. However,
available grazing resources and timescales for the
early trials meant that this approach could not be
implemented initially, so a simpler ‘OffOn’ treatment
structure was implemented for most of the trials. In
this structure, the group of animals first experienced
the ‘Control’ treatment for a number of days, followed
by same group experiencing the ‘GrazeMore’
treatment for a number of days.
Fourteen grazing trials were undertaken on eight
commercial dairy units during the 2019, 2020 and 2021
grazing seasons (details of individual trials are shown
in Table 1), with the aim of having equal length
periods for the ‘Control’ and ‘GrazeMore’ treatments
in each trial (including where the ‘Control’ treatment
was applied in two periods at the start and end of each
trial). In all trials a minimum period of 7 days was
included for each treatment. Herds were allocated a
fresh grazing paddock each morning and evening. The
experimental unit was assumed to be the group of
grazing animals allocated to the combination of a day
grazing paddock and a night grazing paddock during
a particular 24hour period. Response variables
associated with herbage intake could only be
measured (estimated) for the whole group of animals
(i.e., for the selected paddocks), whilst milk yields
could be measured for each individual animal, and
behavioural data for only the 8 (nominally) selected
animals monitored using the nodes. Where data were
collected for individual animals, animal was used as a
blocking (random) term in any analyses, or data were
summarised as mean values per animal prior to
analysis (with the number of animals constant across
all days of a particular trial), and treatment differences
were still assessed relative to the identified experi
mental unit. Responses on the first day on which a
treatment was applied were omitted from the
analyses, as these responses might have been
influenced by the previous treatment, particularly
given the expected lag in milk yield productivity
relative to grazing intake. The temporal ordering of
the treatments (‘Control’ followed by ‘GrazeMore’)
meant that some adjustment in milk yield productivity
was needed to account for the usual changes in
productivity with increasing “days in milk” (DIM).
Herd lactation curves were obtained for each trial and
observed yields multiplicatively adjusted for each
animal separately to 150 DIM.
Statistical Analysis
Due to logistical constraints, different subsets of the
trials were used to investigate different aspects of the
treatment responses. For the analysis of the temporal
dynamics of grazing activity, the highest quality data
were regarded to come from the later trials (10, 11, 12,
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
Rivero et al.
45
14); when the analysis of these four trials failed to
show a consistent increase in feed intakes if was
considered that there was little to be gained from
analysing further trials. Reliable individual daily milk
yield data were only available on farms A, C, G and H
and hence for 6 trials (3, 6, 11, 12, 13, 14). In some trials
the availability of grazing or the management of the
herd and grazing allocations were considered to have
limited, and as such, reliable grazing intake data were
only available for 8 trials (3, 5, 6, 9, 11, 12, 13, 14).
Data for pregrazing herbage mass, postgrazing
residual herbage mass, herbage intake and grazing
efficiency were analysed for treatment differences in
three different temporal periods:
•Day grazing period – directly assessing the impact of
providing the additional grazing area in the
‘GrazeMore’ treatment.
•Night grazing period – to assess possible compen
satory changes in herbage intake.
•24hour grazing period – combined data, considered
likely to influence energy intake, and hence milk
yield.
Differences in these grazing intake responses
between the two treatments (‘Control’ vs.
‘GrazeMore’) were assessed for each trial separately
using twosided, twosample ttests, preceded by F
tests to assess the assumption of equal variance (i.e.,
equal daytoday variation within each treatment
period). Where the Ftest provided no evidence (at the 5
% significance level) for differences in variances
between the treatments, a standard ttest was applied.
Where the Ftest provided evidence for differences in
the variances, an alternative Welch test was applied.
Results of these tests are presented in terms of the
difference between the treatments, a standard error of
the difference (SED) for this comparison, the tstatistic
(tstat), the number of degrees of freedom (df) and a p
value (tprob). Tests with noninteger degrees of
freedom indicate where the alternative Welch test was
applied.
Node efficiency in terms of data transmission and
cloud capture was assessed for Trials 10, 11, 12 and 14
where cows were grazed day and night with no buffer
feeding. Efficiency was calculated as a percentage of
the maximal number of data packets that could have
been received in each 24hour period. In practice, the
number of data records was usually much lower than
the nominal 30 per hour, and records were not all
reported at the same time points in each hour for each
animal, such that the data needed filtering to provide a
reliable assessment over a longer period (e.g., an hour)
of the mean probability of grazing in that period.
Grazing data from node transmission were analy
sed for Trials 10, 11, 12 and 14. For each trial data are
pGraze values which were reported, nominally, every
2 minutes through the duration of the trial. Linear
interpolation was applied to fill gaps between
reporting points and provide a regular sequence of 30
pGraze estimates per hour. Interpolated datasets were
obtained for each of the monitored animals in each
trial and mean values per hour calculated from the 30
interpolated values for each hour. If there were 10 or
fewer reporting points in an hour or where the
maximum gap between reporting points was more
than 20 minutes the hourly mean was set to missing
(and hence the data excluded from the analysis).
The analysis of these data allowed for differences
between animals (nodes) and between days, assessing
for differences between the treatments (assigned to the
different days of the trials), between hours (times of
day – expected to be the dominant source of variability
in the data), and the interaction between treatment and
hour, using analysis of variance (ANOVA), again, with
separate analyses for each of the trials. Interest was in
whether there were overall differences in the prob
ability of grazing (i.e., pGraze; Chamberlain and
Kodam (2019)) between the two treatments, and
whether the distributions of grazing probabilities over
a day were affected by the treatments. To cope with
the anticipated variance heterogeneity of the bounded
probability values, the data were logit transformed
prior to analysis. All analyses were implemented in
Genstat 21st Edition (VSN International, 2021).
Results
Grazing Patterns Through 24h Periods
Issues with node efficiency (see section 3.5) meant
that there were parts of some of these trials where
pGraze data were not recorded frequently enough to
allow interpolated datasets to be produced, impacting
on the reliability of treatment comparisons. A
summary of pGraze data for Trials 10, 11, 12 and 14 is
presented in Table 3, with differences in withinday
grazing patterns shown through plots of the back
transformed means in Figure 3. Red diamonds in
Figure 3 identify hours when the treatment difference
was significantly different from zero. Yellow bars show
average milking periods over the trial period, and the
red star indicates the average time when access to the
‘GrazeMore’ additional grazing was provided.
Additional herbage areas on grazing dairy cows
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
46
For Trials 10 and 12, where full days of reporting
points were missing in the middle of the trials, the
proportion of missing hourly means included these
periods (67 consecutive missing hours for each node in
Trial 10, amounting to 17.4 % of the total number of
hourly means; 65 or 66 consecutive missing hours for
each node in Trial 12, amounting to 15.2 % of the total
number of hourly means), primarily impacting the
‘GrazeMore’ treatment summaries. For Trial 14,
records were unavailable for the last 6 days of the trial
(data omitted for 3 days due to low frequencies of
observations), including much of the ‘GrazeMore’
treatment period, so the comparison of the treatments
was based on only two reliable days for that treatment.
It is also important to note the much higher percentage
of missing hourly means for this trial and the much
higher level of background variability, reflecting less
consistency both between animals within a day and
between days for each animal. For Trial 11, the missing
data were for several hours at the start of the ‘Control’
treatment period.
Table 3. Grazing activity based on node transmission (Trials 1012, 14) of dairy cows submitted to two herbage allocation
methods (‘GrazeMore’ and ‘Control’) in commercial farms. Separate analyses (ANOVA) were applied to the data for each trial.
Trial 10 11 12 14
Days – Control 6 7 [1]110 9
Days – GrazeMore 10 [2]111 8 [2]15 [3]1
Nodes (animals) 7 8 7 8
No. of hourly means 2688 3264 3024 2112
Percentage hours missing 22.4 13.6 18.3 48.1
Treatment effect: Ftest 0.63 0.09 2.40 0.87
df31, 12 1, 15 1, 14 1, 9
pvalue 0.444 0.766 0.143 0.375
Control mean pGraze 0.420 (0.323)2 0.367 (0.547) 0.345 (0.639) 0.343 (0.652)
Grazemore mean pGraze 0.403 (0.391) 0.379 (0.496) 0.328 (0.718) 0.409 (0.369)
1 Number of missing days for all animals. 2 Logit transformed data shown in parentheses. 3df = degrees of freedom. 4 SED = standard error of the difference. 5LSD =
least significant difference.
Figure 3. Mean probability of grazing of dairy cows submitted to two herbage allocation methods (‘GrazeMore’ and ‘Control’) in
commercial farms; a) Trial 10, b) Trial 11, c) Trial 12, and d) Trial 14. Red diamonds indicate hours where there were significant
(p < 0.05) differences between the ‘Control’ and ‘GrazeMore’ treatment means, yellow bars indicate the average milking periods,
and red star indicates the average time when the gate to additional grazing was opened for the ‘GrazeMore’ treatment.
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
Rivero et al.
47
Despite the numerical differences in the overall
mean proportion of time grazing (trials 11 and 14 in
favour of ‘GrazeMore’, and trials 10 and 12 in favour of
‘Control’), for none of the trials were the differences
significantly different from zero. The largest difference
was seen in Trial 14, but the greater background
variability stopped this difference being significant,
though the same difference would have been
statistically significant in the other trials. Overall mean
pGraze values can be used to estimate the total time
grazing per day, so, for example, for Trial 10 the
estimated total daily grazing time for the ‘Control’
treatment was 10.08 hours (10 hours and 5 minutes),
reducing to 9.67 hours (9 hours and 40 minutes) for the
‘GrazeMore’ treatment.
In all four trials there was a highly significant effect
of the hour, as was expected given the usual pattern of
grazing, and a highly significant, though smaller,
interaction effect. These interaction effects are best
interpreted by identifying hours where the
‘GrazeMore’ treatment either significantly increased
or decreased the probability of grazing compared with
the ‘Control’ treatment. Two of the trials (11, 12)
showed a significantly higher mean pGraze for the
‘GrazeMore’ treatment for a couple of hours
immediately after the access to the additional pasture,
but a significantly lower mean pGraze for the
‘GrazeMore’ treatment in the immediately following
hour. There were further, smaller treatment
differences, often with the ‘Control’ treatment mean
pGraze being significantly higher, at other times of
day. For Trial 10, there was more of a shift in the
timing with the ‘GrazeMore’ treatment bringing
grazing bouts earlier. In Trial 14, a rather different
pattern was seen, with the ‘GrazeMore’ treatment
having significantly lower mean pGraze values in the
period immediately after the access to additional
grazing, and significantly higher mean pGraze values
in the subsequent 4 or 5 hours (after the evening
milking). However, the ‘GrazeMore’ treatment data
were less reliable for this trial, based only on the first
two days after the change of treatments.
Herbage Mass PreGrazing (Cover) and PostGrazing
(Residual)
Herbage mass pregrazing ranged from 2000 to 4500
kg DM/hectare (see Supplementary Table 1 in
Chamberlain et al., 2022d), with most values within the
target of 3000 to 4000 kg DM/hectare recommended to
be offered to dairy cows grazing temperate grasslands
(Wilkinson et al., 2020). Significant differences in
quantities of herbage mass pregrazing were recorded
between ‘Control’ and ‘GrazeMore’ treatments in some
trials, but differences between trials were inconsistent
(see Supplementary Table 1 in Chamberlain et al.,
2022d). Mean 24 h herbage mass pregrazing was
significantly higher for ‘GrazeMore’ than ‘Control’ in
Trials 5, 12 and 14 (p < 0.05) and tended to be higher
(p < 0.10) in Trial 11 and lower in Trials 9 and 13. Levels
of postgrazing residual herbage mass ranged from 1500
to 3000 kg DM/hectare, with most values close to 2000
kg DM/hectare (see Supplementary Table 2 in
Chamberlain et al., 2022d). Mean 24 h residual herbage
levels were significantly higher for ‘GrazeMore’ than
‘Control’ in Trials 12 and 14 (p < 0.05).
Herbage Intake and Grazing Efficiency
Average daily herbage intake and grazing efficiency
(expressed as intake as a % of kg DM allowance per
cow per day) are shown in Table 4. Offering additional
herbage in the ‘GrazeMore’ treatment periods had
inconsistent effects on herbage intake. Total 24 h
intakes were significantly higher for ‘GrazeMore’ than
‘Control’ in Trials 5 (p = 0.022), 6 (p = 0.001), 12 (p =
0.020) and 14 (p = 0.002), significantly lower in Trial 3
(p = 0.044) and similar in Trials 9, 11, and 13. In Trials 9
and 13, increased intakes during the day (when
additional herbage was offered in ‘GrazeMore’
periods) were balanced by reduced intakes in
subsequent night periods. In Trial 11, intakes were
similar in both day and night periods (Table 4).
The largest increases in 24 h herbage intake for the
‘GrazeMore’ treatment were 3.9 kg DM/cow in Trial 5,
3.1 kg DM/cow in Trial 12 and 6.3 kg DM/cow in Trial
14. In these trials, increased intakes during the day
were not balanced by reduced intakes at night, as seen
in Trials 9 and 13. Overall, daily herbage intake
averaged 13.8 kg DM/cow for the ‘Control’ treatment
and 15.1 kg DM/cow for the ‘GrazeMore’ treatment.
There were no consistent effects of treatment on
grazing efficiency (Table 4). The lower grazing
efficiency recorded in ‘GrazeMore’ periods in Trials 3,
9, 11, and 13 reflected the greater amount of herbage
DM allowance per cow during day periods which were
not reflected in higher levels of herbage intake.
Milk Yield
Average daily milk yields, corrected for the stage of
lactation to 150 days in milk for individual animals, are
shown in Table 5 for six trials where data for
individual daily yields per cow were available. Data
for milk yields on the first day of each treatment
Additional herbage areas on grazing dairy cows
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
48
regime were omitted as a transition between
treatments. The number of animals in Table 5 was
fewer than the number shown in Table 1 due to the
removal of data during the filtering process (see section
2.6). The number of days in each period varied between
trials due to differences in herbage mass and rate of
daily herbage growth. Treatment differences in mean
daily milk yields per animal were generally small, with
higher yields for the ‘Control’ treatment than for the
‘GrazeMore’ treatment in 4 of the 6 trials, the
difference of 1.9 litres/cow/day in Trial 11 being
statistically significant. The numerical differences
registered between the treatment mean yields in Trials
6 and 12 were not statistically significant.
Table 5. Mean corrected milk yield (litres/cow per day) and power analysis of achieved experimental designs (replicate days for
each treatment) for assessing differences in milk yield of dairy cows submitted to two herbage allocation methods (‘GrazeMore’
and ‘Control’) in commercial farms (trials 3, 6, 1114). A separate analysis was applied for each trial. The least significant
difference (LSD) indicates the smallest differences that would have been statistically significant at the 5 % level.
Trial N°. of Average DIM Milk yield (litres/cow/day) Power Calculations
animals at start1 Overall
Control GrazeMore SED2 tstat3 df4 tprob5 LSD df
mean mean
3 95 239 23.2 22.1 0.530 1.99 8.82 0.079 1.221 8.82
6 79 273 27.2 28.3 0.679 1.59 14 0.134 1.456 14
11 174 256 23.5 21.6 0.364 5.07 16 <0.001 0.771 16
12 166 210 26.8 27.6 0.376 2.13 13.49 0.052 0.813 13.49
13 97 126 27.8 27.7 0.365 0.15 16 0.879 0.774 16
14 142 168 29.8 29.6 0.328 0.44 15 0.666 0.700 15
1DIM = days in milk; 2 SED = standard error of the difference; 3 tstat = tstatistic; 4df = degrees of freedom. 5 tprob = probability of a more extreme tstatistic (two
sided test) from the appropriate tdistribution.
Figure 4. Node efficiency (percentage of data packets received): a) by node identity, and b) by day of trial, for each of Trials 10,
11, 12 and 14.
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
Rivero et al.
49
Table 4. Herbage intake and grazing efficiency (Trials 3, 5, 6, 9, 1114) of dairy cows submitted to two herbage allocation methods (‘GrazeMore’ and ‘Control’) in commercial
farms. Separate analyses were applied for each trial and grazing period. Day and Night analyses omit any backgrazing, 24 h analyses include backgrazing.
Trial Grazing Number of days Daily herbage intake (kg DM/cow) Grazing efficiency (%)5
period Control GrazeMore Control GrazeMore Difference SED1 tstat2 df3 tprob4 Control GrazeMore Difference SED tstat df tprob
mean mean (GrazeMoreControl) mean mean (GrazeMoreControl)
3 Day 8 9 9.27 5.89 3.38 1.540 2.19 15 0.044 43.4 32.4 11.1 4.23 2.62 15 0.019
5 Day 5 9 6.24 9.11 2.87 0.880 3.26 12 0.007 44.2 48.5 4.34 1.35 3.22 4.72 0.025
5 Night 5 8 6.51 7.69 1.18 1.002 1.18 11 0.264 45.9 47.2 1.32 2.54 0.52 7.49 0.617
5 24h 5 8 12.8 16.7 3.91 1.460 2.67 11 0.022 45.2 47.8 2.64 1.63 1.62 11 0.134
6 Day 11 6 2.54 6.72 4.18 0.620 6.74 15 <0.001 32.8 45.6 12.7 2.45 5.21 13.88 < 0.001
9 Day 9 6 7.93 8.34 4.08 0.999 0.41 12 0.690 36.7 26.0 10.8 3.38 3.19 12 0.008
9 Night 9 6 8.61 6.97 1.64 0.800 2.05 13 0.061 37.9 27.6 10.2 4.25 2.41 6.14 0.052
9 24h 9 6 16.1 15.3 0.81 1.300 0.63 12 0.542 26.6 21.3 5.24 2.55 2.06 12 0.062
11 Day 7 11 8.97 8.09 0.88 1.435 0.62 16 0.546 34.4 22.4 12.0 3.55 3.37 7.09 0.012
11 Night 7 11 7.97 8.10 0.14 1.100 0.12 16 0.903 34.5 29.1 5.39 3.87 1.39 16 0.183
11 24h 7 11 17.0 16.7 0.34 2.320 0.15 16 0.886 30.7 20.6 10.1 4.13 2.44 16 0.027
12 Day 11 8 6.35 9.10 2.75 0.691 3.98 14.71 0.001 44.3 45.5 1.19 3.62 0.33 17 0.747
12 Night 12 8 6.59 6.24 0.35 0.798 0.44 18 0.662 47.6 43.4 4.13 3.90 1.06 17 0.305
12 24h 12 8 12.5 15.6 3.09 1.210 2.56 18 0.020 42.3 40.2 2.03 4.09 0.50 17 0.626
13 Day 11 8 4.75 6.64 1.89 0.690 2.74 17 0.014 33.5 29.1 4.43 3.07 1.45 17 0.166
13 Night 12 8 5.75 4.86 0.89 0.684 1.30 17 0.210 39.5 28.6 10.9 3.39 3.23 18 0.005
13 24h 11 8 10.4 11.5 1.12 1.060 1.05 17 0.307 34.7 27.7 6.99 3.19 2.19 18 0.043
14 Day 11 8 6.46 7.33 0.86 0.787 1.10 17 0.287 30.4 34.7 4.22 2.81 1.50 17 0.152
14 Night 11 8 2.13 7.05 4.92 1.139 4.32 17 < 0.001 9.79 38.0 28.2 5.24 5.39 13.16 < 0.001
14 24h 11 8 8.70 15.0 6.29 1.740 3.60 17 0.002 16.0 29.6 13.6 5.33 2.55 17 0.021
1 SED = standard error of the difference
2 tstat = tstatistic
3 df = degrees of freedom
4 tprob = probability of a more extreme tstatistic (twosided test) from the appropriate tdistribution
5 Efficiency = Intake as % of kg DM allowance/cow/day
Additional herbage areas on grazing dairy cows
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
50
This paper presents results from 11 trials on eight
commercial farms, acknowledging the influence of
farm and trial variations on the outcome. While using
commercial farms has limitations, conducting 11 trials
in controlled facilities was impractical. However, the
study aimed to demonstrate the innovation's
applicability across diverse farm settings for potential
commercial use.
Test of hypotheses 1 and 2
In grazing ruminants, the meal and ingestive
behaviour patterns are circadian (Gregorini et al., 2012).
As stressed by Gregorini et al., (2013) ruminants prefer
to graze during daylight hours but may also graze for
short periods during the night (515 % of daily grazing
time). The main meals are concentrated around
twilight hours, dawn, and dusk, with dusk being the
most intense and prolonged. However, the grazing
bout at dawn may have been curtailed by the morning
milking routine. In temperate climates, they typically
have three to five major meals per day, but this can
vary based on external factors like grazing
management (Gregorini et al., 2013). In our case study,
we observed a diurnal pattern with three main grazing
bouts during the day. The evening grazing bout was
the biggest and on one farm trial it was split into two
smaller bouts (Trials 11 and 12, herd G). Generally,
there was minimal grazing activity occurring
overnight. The more intense grazing events, indicated
by a larger area below the pGraze curve, were typically
observed in the late afternoon and evening. The
methodology employed successfully triggered gate
opening after the end of the morning grazing bout and
created new grazing bouts in trials 11 and 12 and
increased the size of the bout in trial 10. It had no effect
in trial 14 possibly due to limited grass growth in the
first part of the trial. However, evening grazing activity
was reduced under the ‘GrazeMore’ treatment,
limiting the impact across the 24 h period.
With regard to the second hypothesis, i.e., the effect
of synchronised allocation of additional herbage on
relevant variables, our findings were inconsistent in
magnitude and direction between trials. Regarding
herbage intake over a 24 h period, four trials showed
greater values for ‘GrazeMore’, whereas ‘Control’
showed greater herbage intake in one trial. For two of
the three trials where no differences were found
between allocation methods, the apparent herbage
intake associated with the increased allocation of
herbage during the day was balanced by reductions in
intake during the subsequent night period. This would
suggest that in some grazing situations there is a limit
to the total time cows are able to allocate to the act of
grazing (Kilgour, 2012), i.e., grazing time is limited by
the time requirements to ruminate and idle (i.e., non
grazing and nonruminating activity) (Chilibroste et al.,
2015). Interestingly, for the three trials (out of four)
with the greatest differences in herbage intake in
favour of ‘GrazeMore’, this increased intake during the
day was not balanced by reduced intake at night. This
would suggest that even though there is a time
limitation for grazing (Kilgour, 2012), the grazing
activity could have been more intense during the
'GrazeMore' stage, possibly involving a higher biting
rate or larger bite sizes.
It was expected that a higher herbage DM intake
would lead to higher milk yields. However, dif
ferences in milk yield between treatments were
generally small and not statistically significant.
Overall, three of the six trials with sufficient milk yield
data showed changes of interest (p < 0.10). However,
two trials showed that the ‘GrazeMore’ treatment
slightly reduced milk yields and only one showed an
increase. This lack of consistent response could be due
to the relatively advanced stage of lactation of cows in
some trials, who might be diverted a proportion of the
additional energy consumed to recover body condition
instead (Moran, 2005). For instance, in Trial 12, an
increase in 24h herbage intake of 3.1 kg DM/day for
the ‘GrazeMore’ treatment was reflected in a positive
milk yield response of 0.8 litres/cow per day; however,
similar responses to the ‘GrazeMore’ treatment were
not seen in Trial 13, which comprised another group of
cows on the same farm at the same time of year. These
inconsistent results could be also explained by the
differences in individual levels in herbage intake of the
cows, since cattle show considerable group
synchronicity in the initiation of grazing activity,
although slightly less so in terminating this activity
(Chilibroste et al., 2015), highlighting the complexity of
factors driving the ingestive behaviour of dairy cows.
During the trials it was evident that although cows ate
more herbage after the initial ‘GrazeMore’ allocation,
they possibly compensated by grazing less during the
following evening period. Possibly, the timing of
additional pasture allocation influenced the response
(Abrahamse et al., 2009) and introduction of additional
pasture later in the day might have been more
beneficial. In order to reduce the impact of
confounding effects, the milk yield data was corrected
for DIM (i.e., to compare all the cows as if they were in
150 DIM), and the amount of inparlour feeding was
held constant through each trial, but it would seem
that there are other factors such as changes in weather,
pasture age, botanical species content, pregrazing
Discussion
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
Rivero et al.
51
herbage mass, cow type and management that have af
fected the milk yields. However, since three trials were
in the same geographical location, two of which were
over the same time period, assessment of
environmental factors influencing the response would
have been unlikely to reveal significant effects.
Another factor that could have affected the results is
the method used for estimation of herbage intake (i.e.,
rising platemeter). This indirect technique is reliable
over a short time scale (e.g., 24h) (Smith et al., 2021).
The rising platemeter technique is useful for obtaining
herd estimates of pasture intake for management
decisions and for the determination of pasture
parameters associated with intake (Reeves et al., 1996).
The accuracy of the estimations would increase with a
prediction equation developed for each farm (Lukuyu
et al., 2014), and for pregrazing and postgrazing
herbage (Reeves et al., 1996), though this task was
beyond the objective of this case study. Macoon et al.,
(2003) states that, when appropriate for the research
objectives, herbage disappearance method may be
useful and less costly alternatives to using the pulse
dose method. Another disappearance method such as
the classical sward cutting method can give a good es
timate of herbage intake by grazing animals, but often
a large variation in the estimation of herbage mass is
found. Variation of both pre and postgrazing
measurements are added; hence, the herbage intake
values become even more variable (Smit et al., 2005),
which was likely the case for the present study.
Consequently, the total 24hour grazing efficiency
values were not particularly reliable for some trials
because these figures included the impact of back
grazing, where, often, the estimate of herbage intake
done with the raising plate meter was close to zero, but
a relatively substantial area of herbage was available
which magnified any errors in estimated feed intake.
However, separate day and night analyses were not
impacted by these backgrazing assessments.
The statistical power of the design to test the
‘GrazeMore’ hypothesis with regard to milk yield
(hypothesis 2) was assessed by considering the size of
the least significant difference (LSD, at the 5 %
significance level) for each trial, obtained from the
information about daytoday variation in responses
having allowed for any differences in the mean yields
for the two treatments. In all cases, a milk yield
difference of 2 litres/animal/day would have been
statistically highly significant should such a difference
have been observed.
Main Practical Limitations
This paper reports results from 14 trials across 8
commercial farms. The variability in the results will
have been influenced by differences between farms and
between trials that we could not control. This is a
disadvantage of using commercial farms but the costs
and resources to run one trial on an experimental
facility were beyond this project and to run 14 such
trials is probably not possible in any country. This
issue was anticipated at the start of the project, but it
was felt that if this innovation was to be of commercial
interest to the industry it would have to work and
show benefits across a range of farms and
environments under commercial conditions.
The initial experimental design was to identify
suitable pairs of farms (within 20 miles of each other)
where, at least, some cow groups were grazed without
buffer feeding, that used a rising plate meter and the
AgriNet system (Irish Farm Computers Ltd., Ireland)
and recorded daily milk yields through the parlour.
Farms were to be identified in pairs that were near
each other so that two farms could be enrolled in each
trial and the treatment/control periods reversed
between the two farms. However, it was only possible
to identify a few farms that fit the selection criteria, and
it was not possible to work with pairs of farms
simultaneously. In addition, trial farms were identified
several months in advance of the actual trial; however,
at the time of running the trial, various factors had
changed that impacted the running of and results from
some of the trials. A further issue was that the farms
needed to have surplus grass at the time of trial to
enable the additional herbage allowance to be
provided during the GrazeMore treatment period.
Unfortunately, droughts and heatwaves limited grass
growth. For instance, trial 7 was in midsummer on a
farm on welldraining chalkbased soils and hence
suffered very poor grass growth in the dry summer
such that the farm management would not allow an
additional allocation of pasture. In Trial 10, grass
covers had been managed to be very high and outside
the predictive range of the rising plate meter.
Additionally, heterogeneous sward structures
developed through the grazing season and back
grazing made it difficult to assess forage mass and
allocate grazing areas accurately (Merino et al., 2018).
Overall, data was incomplete due to poor grazing
infrastructures in Trials 1 and 2 and low grass
availability restricting the length of grazing periods in
Trials 1 to 8 and also in Trial 10. Although the
experimental design could not be fully implemented
Additional herbage areas on grazing dairy cows
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the study;
in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the
results.
Bargo, F., L. D. Muller, E. S. Kolver, and J. E. Delahoy,
2003, Invited review: Production and digestion of
supplemented dairy cows on pasture: Journal of
Dairy Science, v. 86, no. 1, p. 1–42, doi:10.3168/
jds.S00220302(03)735814.
Conclusions
Acknowledgments: The authors acknowledge the kind and generous support from the farmer members of WD
Farmers Ltd. We thank E. Harper and S. McCrudden from WD Farmers Ltd, J. Daniel from Precision Grazing Ltd,
and S. Kodam from Hoofprints Technologies Ltd for their contributions to the execution of the project.
Literature Cited
Funding: This work was cofunded by the UK’s innovation agency, Innovate UK, Project 105150: ‘The use of GPS
tracking and the LoRaWAN network to improve productivity of grazing dairy cows’. This work was part of the Soil
to Nutrition Institute Strategic Program Grant BBS/E/C/000I0320 at Rothamsted Research funded by the UK
Biotechnology and Biological Sciences Research Council (BBSRC). Support in writing up the work was greatly
received by BBSRC through the strategic program Growing Health (BB/X010953/1).
Data Availability: Data are available from Rothamsted Research (Chamberlain et al., 2022a, 2022b, 2022c, 2022d)
Abrahamse, P. A., S. Tamminga, and J. Dijkstra, 2009,
Effect of daily movement of dairy cattle to fresh
grass in morning or afternoon on intake, grazing
behaviour, rumen fermentation and milk
production: The Journal of Agricultural Science, v.
147, no. 6, p. 721–730.
doi:10.1017/S0021859609990153.
Edited by: Julio Galli
52
on these farms, the trials were still very valuable for
developing the data capture and processing aspects
and refining the cloudbased algorithms.
The Covid19 pandemic severely restricted field work
due to lockdown measures in 2020 and 2021 that
prevented access to farms by the research team, and
several potential trials were lost. In the early stages of
the project, we were unable to fabricate LoRaWAN
nodes due to severe global component shortages. Early
trials (Trials 1 to 4) were therefore carried out using
inferior generic Chinese LoRaWAN nodes. These
nodes had performance issues related to transmission,
and poor battery management firmware limited
battery lives to only 2 to 3 weeks. In later trials, we
were able to revert to fabricating our nodes, which
gave better performance and battery life (68 months).
LoRaWAN signal transmission is limited by line of
sight. In several trials, woodlands and ancillary
buildings limited signal transmission, and on farm H,
the hilly topography of the farm made some fields
inaccessible to LoRaWAN signals. Secondary
LoRaWAN gateways can be used to cover ‘dead spots’,
but the logistics of identifying areas requiring
additional gateways, supplying mains electricity or
large solar panels and the associated security required
made this impractical for a shortterm trial.
In conclusion, the variable and inconsistent responses
to the ‘GrazeMore’ treatment indicate that the concept
of allocating additional areas of pasture automatically
at specific times requires further development. In these
trials it was possible to change the pattern of feeding,
but in the current format, it was not possible to
consistently increase grazing intakes and milk yields.
However, the timing of additional daily pasture
allocation may be critical in determining the extent to
which dairy cows are likely to respond in terms of
increased herbage intake and subsequent milk yield.
For example, additional pasture allocation in the early
evening, coincidental with the main grazing period,
may be worthy of investigation. The LoRaWAN
system performed well in the farm environment when
using our node design, but additional LoRa gateways
may be necessary to cope with hilly terrains especially
when sending data from the gateway to the gate
actuator node in the grazing paddocks.
Bao, J., P. S. Giller, and G. Stakelum, 1998, Selective
grazing by dairy cows in the presence of dung and
the defoliation of tall grass dung patches: Animal
Science, v. 66, no. 1, p. 65–73, doi:10.1017/
S1357729800008845.
Baudracco, J., N. LopezVillalobos, C. W. Holmes, and
K. A. Macdonald, 2010, Effects of stocking rate,
supplementation, genotype and their interactions
on grazing dairy systems: a review: New Zealand
Journal of Agricultural Research, v. 53, no. 2, p. 109–
133, doi:10.1080/00288231003777665.
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
Rivero et al.
53
Chamberlain, A., and S. Kodam, 2019, The use of GPS
tracking and the LoRaWAN network to improve
the productivity of grazing dairy cows: preliminary
results, in B. O’Brien, D. Hennessy, and L. Shalloo,
eds., 9th European Conference on Precision
Livestock Farming: ECPLF, p. 779–782.
Macoon, B., L. E. Sollenberger, J. E. Moore, C. R.
Staples, J. H. Fike, and K. M. Portier, 2003, Comparison
of three techniques for estimating the forage intake of
lactating dairy cows on pasture: Journal of Animal
Science, v. 81, no. 9, p. 2357–2366.
doi:10.2527/2003.8192357x.
Chamberlain, A. T., A. Mead, M. J. Rivero, and J. M.
Wilkinson, 2022a, GrazeMore grazing activity
data: doi:10.23637/rothamsted.988xq.
Chamberlain, A. T., A. Mead, M. J. Rivero, and J. M.
Wilkinson, 2022b, GrazeMore grazing intake data:
doi:10.23637/rothamsted.988xv.
Chamberlain, A., A. Mead, M. Rivero, and J. Wilkinson,
2022, GrazeMore grazing intake data. doi 10.23637/
rothamsted.988xv: doi:10.23637/rothamsted.988xv.
Chamberlain, A. T., A. Mead, M. J. Rivero, and J. M.
Wilkinson, 2022c, GrazeMore milk yield data:
doi:10.23637/rothamsted.988xw.
Chamberlain, A. T., A. Mead, M. J. Rivero, and J. M.
Rivero, and J. M. Wilkinson, 2022d, GrazeMore
Supplementary material Table 1 an 2: doi:10.23637/
rothamsted.988xv.
Chamberlain, A. T., and J. M. Wilkinson, 1996, Feeding
the Dairy Cow: Lincoln, Chalcombe Publications, 421 p.
Chilibroste, P., M. J. Gibb, P. Soca, and D. A.
Mattiauda, 2015, Behavioural adaptation of grazing
dairy cows to changes in feeding management: Do they
follow a predictable pattern? Animal Production
Science, v. 55, no. 3, p. 328–338, doi:10.1071/AN14484.
Gregorini, P., 2012, Diurnal grazing pattern: Its
physiological basis and strategic management:
Animal Production Science, v. 52, no. 7, p. 416–430,
doi:10.1071/AN11250.
Gregorini, P., P. C. Beukes, A. J. Romera, G. Levy, and
M. D. Hanigan, 2013, A model of diurnal grazing
patterns and herbage intake of a dairy cow,
MINDY: Model description: Ecological Modelling,
v. 270, p. 11–29, doi:https://doi.org/10.1016/
j.ecolmodel.2013.09.001.
Gregorini, P., B. DelaRue, K. McLeod, C. E. F. Clark, C.
B. Glassey, and J. Jago, 2012, Rumination behavior
of grazing dairy cows in response to restricted time
at pasture: Livestock Science, v. 146, no. 1, p. 95–98,
doi:https://doi.org/10.1016/j.livsci.2012.02.020.
Kilgour, R. J., 2012, In pursuit of “normal”: A review of
the behaviour of cattle at pasture: Applied Animal
Behaviour Science, v. 138, no. 1–2, p. 1–11,
doi:10.1016/j.applanim.2011.12.002.
Lukuyu, M., D. R. Paull, W. H. Johns, D. Niemeyer, J.
McLeod, B. McCorkell, D. Savage, I. W. Purvis, and
P. L. Greenwood, 2014, Precision of estimating
individual feed intake of grazing animals offered
low, declining pasture availability. Animal
Production Science, v. 54, no. 12, p. 2105–2111.
Macdonald, K. A., Penno, J. W., Nicholas, P. K., Lile, J.
A., Coulter, M. & Lancaster, J. A. S. (2001). Farm
systems – Impact of stocking rate on dairy farm
efficiency. Proceedings of the New Zealand
Grassland Association, 63, 223227.
McGilloway, D., C. S. Mayne, P. C. Garnsworthy, J.
Wiseman, and W. Haresign, 1996, Importance of grass
availability for the high genetic merit dairy cow.:
Recent advances in animal nutrition, p. 135–169.
Merino, V., O. Balocchi, and R. Pulido, 2018, Effect of
daily herbage allowance restriction on pasture
characteristics and milk production by grazing dairy
cows in spring: Ciencia e Investigacion Agraria, v.
45, no. 1, p. 21–34, doi:10.7764/rcia.v45i1.1841.
Met_Office, 2023, UK regional climates: <https://
www.metoffice.gov.uk/research/climate/mapsanddata/
regionalclimates/index> (accessed February 2, 2023).
Miles, B., E.B. Bourennane, S. Boucherkha, and S.
Chikhi, 2020, A study of LoRaWAN protocol
performance for IoT applications in smart agriculture:
Computer Communications, v. 164, p. 148–157,
doi:https://doi.org/10.1016/j.comcom.2020.10.009.
Moran, J., 2005, How feed requirements change during
lactation: Feeding management for small holder dairy
farmers in the humid tropics. CSIRO Publishing.
O’Driscoll, K., B. O’Brien, D. Gleeson, and L. Boyle,
2010, Milking frequency and nutritional level affect
grazing behaviour of dairy cows: A case study: Applied
Animal Behaviour Science, v. 122, no. 2, p. 77–83,
doi:https://doi.org/10.1016/j.applanim.2009.11.014.
Reeves, M., W. J. Fulkerson, R. C. Kellaway, and H.
Dove, 1996, A comparison of three techniques to
determine the herbage intake of dairy cows grazing
kikuyu (Pennisetum clandestinum) pasture:
Australian Journal of Experimental Agriculture, v.
36, no. 1, p. 23–30, doi:10.1071/EA9960023.
Rivero, M. J., P. GrauCampanario, S. Mullan, S. D. E.
Held, J. E. Stokes, M. R. F. Lee, and L. M. Cardenas,
2021, Factors Affecting Site Use Preference of
Grazing Cattle Studied from 2000 to 2020 through
GPS Tracking: A Review: Sensors, v. 21, no. 8,
doi:10.3390/s21082696.
Sheahan, A. J., R. C. Boston, and J. R. Roche, 2013,
Diurnal patterns of grazing behavior and humoral
factors in supplemented dairy cows: Journal of
Dairy Science, v. 96, no. 5, p. 3201–3210, doi:10.3168/
jds.20126201.
Additional herbage areas on grazing dairy cows
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
54
Smit, H. J., H. Z. Taweel, B. M. Tas, S. Tamminga, and
A. Elgersma, 2005, Comparison of techniques for
estimating herbage intake of grazing dairy cows:
Journal of Dairy Science, v. 88, no. 5, p. 1827–1836,
doi:10.3168/jds.S00220302(05)728575.
Wilkinson, J. M., and M. R. F. Lee, 2018, Review: Use of
humanedible animal feeds by ruminant livestock.:
Animal : an international journal of animal
bioscience, v. 12, no. 8, p. 1735–1743, doi:10.1017/
S175173111700218X.
ISSNL 10221301. Archivos Latinoamericanos de Producción Animal. 2024. 32 (1): 37  54
Smith, W. B., M. L. Galyean, R. L. Kallenbach, P. L.
Greenwood, and E. J. Scholljegerdes, 2021,
Understanding intake on pastures: how, why, and a
way forward: Journal of Animal Science, v. 99, no. 6,
p. skab062, doi:10.1093/jas/skab062.
Turner, L. W., M. C. Udal, B. T. Larson, and S. A.
Shearer, 2000, Monitoring cattle behavior and pasture
use with GPS and GIS: Canadian Journal of Animal
Science, v. 80, no. 3, p. 405–413, doi:10.4141/A99093.
Wilkinson, J. M., M. R. F. Lee, M. J. Rivero, and A. T.
Chamberlain, 2020, Some challenges and
opportunities for grazing dairy cows on temperate
pastures: Grass and Forage Science, v. 75, no. 1, p. 1–
17, doi:10.1111/gfs.12458.
Rivero et al.