Science of the Total Environment 579 (2017) 1572–1580

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Scaling methane emissions in ruminants and global estimates in

wild populations
F.J. Pérez-Barbería
Grupo PAIDI RNM118, Estación Biológica de Doñana, CSIC, Sevilla 41092, Spain
Ungulate Research Unit, CRCP, University of Córdoba, Córdoba, Spain
James Hutton Institute, Craigiebuckler, AB15 8QH Aberdeen, Scotland, UK

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• CH4 vs body mass slope does not differ

from metabolic expenditure or isome-
try.
• Dietary fiber depresses CH4 as the level
of DMI increases.
• Cattle produce more CH4 per unit of
DMI than red deer, sheep or goats.
• For same diet larger ruminants produce
more CH4 per DMI unit than smaller ru-
minants.
• CH4 global emissions of wild ruminants
are between 1.094 and 2.687 Tg y− 1.

a r t i c l e i n f o a b s t r a c t

Article history: Methane (CH4) emissions by human activities have more than doubled since the 1700s, and they contribute to
Received 10 October 2016 global warming. One of the sources of CH4 is produced by incomplete oxidation of feed in the ruminant's gut. Do-
Received in revised form 8 November 2016 mestic ruminants produce most of the emissions from animal sources, but emissions by wild ruminants have
Accepted 24 November 2016
been poorly estimated. This study (i) scales CH4 against body mass in 503 experiments in ruminants fed herbage,
Available online 4 December 2016
and assesses the effect of different sources of variation, using published and new data; and (ii) it uses these
Editor: D. Barcelo models to produce global estimates of CH4 emissions from wild ruminants. The incorporation of phylogeny,
diet and technique of measuring in to a model that scales log10 CH4 g d−1 against log10 body mass (kg), reduces
Keywords: the slope, from 1.075 to 0.868, making it not significantly steeper than the scaling coefficient of metabolic re-
GHG inventory quirements to body mass. Scaling models that include dry matter intake (DMI) and dietary fiber indicate that al-
Meta-analysis though both increase CH4, dietary fiber depresses CH4 as the levels of DMI increases. Cattle produces more CH4
Methane emissions per unit of DMI than red deer, sheep or goat, and there are no significant differences between CH4 produced
Phylogeny by red deer and sheep. The average estimates of global emissions from wild ruminants calculated using different
Ruminants
models are smaller (1.094–2.687 Tg y−1) than those presented in the reports of the Intergovernmental Panel on
Climate Change (15 Tg yr−1). Potential causes to explain such discrepancy are the uncertainty on the world's wild
ruminant population size, and the use of methane output from cattle, a high methane producer, as representative
methane output of wild ruminants. The main limitation researchers' face in calculating accurate global CH4 emis-
sions from wild ungulates is a lack of reliable information on their population sizes.
© 2016 Elsevier B.V. All rights reserved.

E-mail address: j.perezbarberia@gmail.com.

http://dx.doi.org/10.1016/j.scitotenv.2016.11.175
0048-9697/© 2016 Elsevier B.V. All rights reserved.
 F.J. Pérez-Barbería / Science of the Total Environment 579 (2017) 1572–1580
1. Introduction
Methane is the most abundant organic trace gas in the atmosphere
and since the 1700s human activities, particularly agriculture, waste
disposal and fossil fuels extraction, have more than doubled its emis-
sions (Wuebbles and Hayhoe, 2002). Methane is a potent greenhouse
gas (Stocker et al., 2013) because of its higher global-warming potential
(ca. 21 times) and shorter atmospheric lifespan (1/5–1/20) than carbon
dioxide, which make methane reduction strategies a short term effec-
tive mean of slowing global warming (Moss et al., 2000).
Methane is an end-product of microbial fermentation of ingested
feeds in the rumen or gut of herbivorous mammals, especially rumi-
nants, but also in hindgut fermenters (Blaxter, 1962), and it has been
suggested that about one third of methane emissions worldwide come
from ruminants' enteric fermentation (Stocker et al., 2013). As methane
is the result of incomplete oxidation of feeds in ruminants' gut, much re-
search has been carried out to minimize its production to increase the
metabolisable energy content of feedstuffs, and most recently, as an al-
ternative way to mitigate global emissions by domestic ruminants
(Martin et al., 2010; Moss et al., 2000). A number of strategies have
been investigated for reducing ruminants' methanogenesis with differ-
ent results (propionate enhancers, acetogens, reduction in protozoal
populations by probiotics, alternative electron acceptors) (Moss et al.,
2000; Rira et al., 2015). In general, increasing animal productivity re-
duces methane production per unit of animal biomass emissions, but
at the cost of feeds containing high quality/lower fiber sources of carbo-
hydrates, their higher carbon foot print in production and transport, and
sometimes to the detriment of animal welfare (Paustian et al., 2006). It
can be argued that the strategy of reducing methane emissions by using
high quality/lower fiber diets defeats the function of ruminants in the
trophic chain, as resources high in fiber comprise most of the world veg-
etal biomass, and ruminants can make an efficient transformation of
this fiber into protein and milk sources for human consumption.
In comparison with intensive farming systems, extensive animal
farming is more energy and water efficient, and minimizes impact on
soil, surface waters, biodiversity and animal welfare (Mellor, 2015;
Wieren and Bakker, 2008). Despite the potential for the increase of ex-
tensive animal farming as a sustainable alternative to intensive produc-
tion systems, little research has focused on the effects of native plant
species as dietary components in methane emissions in grazing exten-
sive systems, in comparison with the plethora of work devoted to culti-
vated feedstuffs and supplementation. From a search on a peer-review
database on in vivo studies on methane production in ruminants
(n = 43) only one third exclusively used natural forages.
Methane emissions by ruminants might play a significant role in
global warming. It has been suggested that about 13,400 years ago, her-
bivore megafauna released 9.6 Tg yr−1 of methane to the atmosphere,
and the massive extinction event of America's megafauna
11,500 years ago could have been responsible for 12.5 to 100% of the
overall methane decline recorded at that period (Smith et al., 2010), al-
though this has been questioned by Brook and Severinghaus (2011).
Global methane inventories estimate that the main wild animal
sources of methane emissions are due to ruminants (15 Tg yr−1) and
termites (11 Tg yr− 1) (Stocker et al., 2013), which is a considerable
source of methane emissions compared to that estimated from domes-
tic ruminants (85 Tg yr−1) (Stocker et al., 2013). However, it is likely
that these figures are inaccurate, due to our surprisingly poor knowl-
edge on methane emissions by wild ruminants. For example, (i) the In-
tergovernmental Panel on Climate Change IPCC (Stocker et al., 2013),
the main global climate change report, presents poor estimates of meth-
ane emissions by wild ruminants, as pointed out by Smith et al. (2010)
and in this study (see Discussion); (ii) the number of ruminant species
(including the true pseudo-ruminants with three-chambered stomach)
for which information on methane emissions is available is very small,
to the best of my knowledge only 8 species (sheep, goat, cattle, yak,
llama, alpaca, water buffalo and red deer), which is a mere 4% of
1573
ruminants species and represent mixed feeders, only a part of the spec-
trum of feeding styles (Hofmann and Stewart, 1972; Pérez-Barbería and
Gordon, 2001) and only two studies that attempted to scale methane
emissions against a wide range of body masses (Franz et al., 2010;
Smith et al., 2010); (iii) there is no a unified technique to estimate
methane emissions (Moss et al., 2000), and the consequent potential
differences in emissions across techniques and limitations to estimate
methane output in free-ranging conditions (Pavao-Zuckerman et al.,
1999; Pinares-Patiño et al., 2008); and finally (iv) there is also a great
deal of heterogeneity across studies in how methane emissions infor-
mation is presented, which restricts the number of studies and informa-
tion that can be used for meta-analyses and modeling.
This study aims to (i) scale methane emissions against body mass in
ruminants fed only herbage, and study the effects of different sources of
variation, that is, animal species (both as a factor but also as a phyloge-
netic structure), technique of measuring CH4 emissions, dietary compo-
nents and their chemical characterization, and experimental conditions,
using published information collected by a comprehensive literature re-
view and new data from this study; (ii) use the methane yields of the
models of this study, combined with population statistics of wild rumi-
nant populations, to produce estimates of yearly input of methane emis-
sions to the atmosphere, and compare the estimates against those from
previous studies. The results are intended to improve methane invento-
ries and provide information to be used in bottom-up modeling exer-
cises on global warming.
2. Material and methods
2.1. Meta-analysis database
I created a database on methane emissions from ruminants and
pseudo-ruminants fed pure or mixed diets that only contained herbage,
and no supplements, as dietary components, by a comprehensive liter-
ature review (Online Sources 1 and 2). In total 77 studies were used,
from which I extracted 503 animal records of methane emissions. The
variables that produced the most comprehensive dataset (n = 360)
were: CH4 emissions (g·d− 1), daily feed dry matter intake (DMI,
kg·d− 1), dietary neutral detergent fiber content (NDF, % dry matter
basis), body mass of the experimental animal (BM, kg), animal species,
dietary components (Diet, with 5 levels: Compositae, grass, legumes,
grass – legumes, grass – heather), experimental technique used to mea-
sure CH4 emissions (Technique, with 7 levels: open circuit respiration
chamber OCC, closed circuit respiration chamber CCC, sulphur
hexafluoride tracer SF6, ethane trace C2H6, micrometeorological tech-
niques MM, polytunnel POL, and face mask methods FACE), and study
Fig. 1. Rooted tree showing the phylogenetic relatedness of the ruminant species used in
this study.
1574 F.J. Pérez-Barbería / Science of the Total Environment 579 (2017) 1572–1580
reference (Source). There were records of methane emissions for eight
species (sheep, Ovis aries n = 213; cattle, Bos taurus n = 108; llama,
Lama glama (n = 1); alpaca, Vicugna pacos n = 3; red deer, Cervus
elaphus n = 19; goat, Capra aegagrus n = 13; water buffalo, Bubalus
bubalis n = 3; yak, Bos grunniens n = 1, Fig. 1). A second dataset of
188 records had an extra variable with information on the duration, in
days, of each CH4 measurement (MPER), and it was used to assess its ef-
fect on CH4 emissions, as CH4 emissions can vary across experimental
days (Pinares-Patiño et al., 2013).
2.2. Body mass, wild ruminant population size and energy requirements
To estimate CH4 global emissions (see Statistical analysis) I created a
comprehensive database of the population sizes, body mass, estimated
intake, and a proxy of feed digestibility of 195 species by sex of wild ru-
minants across the world (Nowak, 1999), based on a literature review of
published and on-line information, together with information from per-
sonal communications from experts of IUCN Specialist Groups (Online
Sources 3 and 4).
Energy requirements depend on the animal condition and reproduc-
tive stage, which in multiples of maintenance can vary widely. For ex-
ample, in sheep in extensive systems, up to 1.31 in growing animals,
1.6–1.9 in late pregnancy (single and twins offspring, respectively),
2.8 during the first month of a lactating ewe with one lamb or 4.0
with twins (Ministry of Agriculture, 1975) have been reported. In farm-
ing systems feeding levels are for net production of animal product at a
high rate, in developing countries at a low rate, and in the wild, no net
production, except for animals growing as they reach maturity, and
body condition can vary widely across seasons, with body mass losses
due to negative energy-balance and use up of fat reserves (Mitchell et
al., 1977). As a pragmatic compromise I estimated methane emissions
in wild ruminants assuming all individuals at adult body mass and
maintenance food intake in the wild (i.e. 2.3 × basal metabolic rate)
[as in Crutzen et al. (1986)]. This was done in an attempt to average
(i) low methane emissions of suckling offspring and animals that
loose condition because old age or environmental harshness, and (ii)
higher methane emissions from growing, pregnant or lactating animals.
Furthermore, methane yield per unit of dry matter intake depends on
the negative interaction between the digestibility of feed and the level
of intake: as when intake increases, passage rate of food through the
gut increases, with a concomitant decrease of digestibility and methane
production (Blaxter and Clapperton, 1965; Krishna et al., 1978;
Vanderhoning et al., 1981; Wainman, 1977). To account for variability
in feed digestibility related to the body size of species, that is, smaller
species ingest more digestible food because of (i) constraints of digesta
retention associated with body size (Demment and Van Soest, 1985;
Geist, 1974; Robbins, 1993) [but for lack of evidence see (Clauss et al.,
2013; Pérez-Barbería et al., 2008, 2007)], or, (ii) because larger species
cannot afford the foraging time required to meet their energy require-
ments when selecting only high quality food (Clauss et al., 2007;
Jarman, 1974; Mueller et al., 2013); I used a scaling equation on NDF
against body mass to predict NDF (NDF being a proxy of plant digestibil-
ity, see Statistical analysis). Although dietary NDF varies seasonally, I
followed Hristov's (2012) approach of no controlling for seasonal
changes in diet digestibility as it would be very speculative and complex
across species and habitats.
2.3. Statistical analysis
Seven regression models on log10 methane emissions were fitted.
The first five models (namely r.0, r.1, r.2, r.3, r.4) were intended to
scale log10 CH4 emissions against log10 BM, and assess the effects of
using different random terms (i.e. study source, phylogeny, diet, tech-
nique) on the scaling coefficients, and these coefficients were compared
against those published in the literature. Model r.0: ordinary linear re-
gression with no random effects; Model r.1: Markov chain Monte
Carlo (MCMC) generalized linear model with study source (Source) as
random effect; Model r.2: MCMC generalized linear model with Source
and phylogeny (Sp) as random effects; Model r.3: MCMC generalized
linear model with Source, Sp and type of diet (Diet) as random effects;
Model r.4: MCMC generalized linear model with Source, Sp, Diet and
type of Technique to measure methane emissions (Technique) as ran-
dom effects.
The other two models were aimed to model methane emissions
against log10 DMI, BM, dietary NDF, species and their first and second
order meaningful interactions (no higher order interactions were fitted
to avoid over-parameterization within the data available). One of these
two models used species as a phylogenetic matrix structure in the ran-
dom effects (Phylogenetic model). The other model focused on differ-
ences between species (Sp) in methane emissions (Species model),
using only species with the larger number of measurements (i.e. cattle,
sheep, red deer, goat). I also attempted fitting models that included
experimental animal (Ind) nested within study source, as an additional
random effect, but this term was removed in the final model, as the
effect was negligible, and its inclusion reduced the sample size dramat-
ically, as in many published studies information on Ind was not available
or clearly specified.
I used generalized linear mixed models with Markov chain Monte
Carlo techniques in the statistical package MCMCglmm (Hadfield,
2010). This approach has the flexibility of fitting random and fixed ef-
fects, and makes the computation of the posterior distribution of any
function of the variance components straightforward, which allows
presenting information on the contribution of the different sources of
variability, one of the aims of this study.
The MCMC models also fitted, as a random effect, the structure of the
phylogenetic relatedness between species, as in a phylogenetically com-
parative method, because information from different species cannot be
considered statistically independent (Harvey and Purvis, 1991) and I
was interested in assessing its contribution to the total variance of the
data (details on MCMCglmm models in Online Source 5).
Estimates of CH4 emissions were conducted using the MCMC gener-
alized linear model (Phylogenetic model), that included CH4 as response
variable, and log10 DMI, log10 BM, log10 DMI x NDF as fixed effects, and
Phylogeny, Source, Technique and Diet as additive random effects.
Details on the parameterization of this model are in Online Source 6.
As the objective was to identify the main drivers of the dependent
variables, I correctly used, for model selection, p-values against mea-
sures based on information theory, such as ΔAIC or BIC (Murtaugh,
2014). I proceeded by first fitting the full models as described above,
and then using backward elimination by removing the non-significant
fixed-effects interactions, one at a time, following the principle of
marginality: the highest order interactions were tested first and if
they were significant, then the lower order effects were not tested for
significance. All analyses and graphics were conducted in R software
(R Development Core Team, 2015), mainly using packages MCMCglmm
(Hadfield, 2010) and ggplot2 (Wickham, 2009).
3. Results
3.1. Scaling methane emissions against body mass
The coefficients of the scaling equation of log10 CH4 against log10 BM
differed across models, mainly, depending on the structure of random
effects fitted. The highest values of slope, and smallest intercepts,
were those from the ordinary least square regression model (model
r.0: slope = 1.075, intercept = −0.572) and from the MCMC mixed lin-
ear model that included reference source as the only random effect
(model r.1: slope = 1.073, intercept = −0.540), with no significant dif-
ferences between these two models in slopes and intercepts (Table 1,
Fig. 2, Online Source 7). The inclusion of the phylogenetic relationships
between species, as an additive random effect, produced a significant
decrease in the slope and increase of the intercept (model r.2,
 F.J. Pérez-Barbería / Science of the Total Environment 579 (2017) 1572–1580 1575

upper values of the highest posterior density interval 0.95 of the estimate; eff.s: effective number of samples of the MCMC chain; var.: explained variance by each term of the random effects; number of sources used in the analyses = 66; number of
Parameters of five models on methane emissions in ruminants (log10 CH4 g d−1) against log10 body mass (kg) with different additive random effects and their explained variance. Model r.0: ordinary linear regression with no random effects; Model
r.1: MCMC generalized linear model with study source (Source) as random effect; Model r.2: MCMC generalized linear model with Source and phylogeny (Sp) as random effects; Model r.3: MCMC generalized linear model with Source, Sp and type of
diet (Diet) as random effects; Model r.4: MCMC generalized linear model with Source, Sp, Diet and type of Technique to measure methane emissions (Technique) as random effects. Mean: estimate of the posterior mean; l-CI and u-CI: lower and

b0.001
0.201
p

19,800
19,800
19,800
19,800
19,800

19,800
19,800
eff.s

0.022
0.117
0.015
0.033
0.013

0.148
1.025
0.535
0.878
0.247
0.372
u-CI

–0.645
0.0004
0.0002
0.0002
0.007

0.010

0.721
0.036
0.116
0.001
0.002
l-CI
Model r.4

Mean

–0.26
0.014
0.041
0.005
0.010
0.011

0.868
0.213
0.431
0.064
0.126
b0.001
0.158
p

19,800
18,978
19,800

19,800

19,800
19,800
eff.s

–
0.027
0.119
0.017

0.013

0.110
1.042
0.520
0.837
0.235
u-CI

–
0.0002

–0.666
0.010
0.001

0.010

0.737
0.067
0.107
0.002
l-CI

–
Model r.3

Fig. 2. Predictions of the five models in Table 1 on methane emissions in ruminants (log10
Mean

–0.28
0.018
0.041
0.006

0.011

0.893
0.288
0.458
0.075

CH4 g d−1) against log10 body mass (kg). Models differ in the number of random effects
–

(see Table 1 for the description of the models).

b0.001
0.134

slope = 0.912, intercept = − 0.308, Table 1, Fig. 2, Online Source 7).

–
p

Subsequent models that included Diet (i.e. dietary components)

19,800
19,800

19,800

19,800
19,800

(model r.3) and Technique of measuring CH4 (model r.4) did not pro-
eff.s

duce any significant change in the slope and intercept values in relation
–
–

to model r.2 (Table 1, Fig. 2, Online Source 7). Most of the variance ex-
0.030
0.132

0.013

0.088
1.061
0.597
0.851
u-CI

plained by the random effects (Table 1, model r.4) was due to Sp

–
–

–
–

(43.1%), followed by Source (21.3%), Technique (12.6%) and Diet (6.4%).

–0.70
0.012
0.000

0.010

0.756
0.078
0.102
l-CI

–
–

–
–
Model r.2

3.2. The effects of feed DMI and dietary NDF scaling methane emissions
Mean

0.020
0.046

0.012

0.912
0.327
0.487
–0.31

The fixed effect Technique was not significant in the Phylogenetic

–
–

–
–

model of Table 2 (full model), this term was moved to the random ef-
b0.001
b0.001

fects in the final model (Table 2). The final model clearly indicated
–
–
p

that log10 DMI (coefficient = 0.786, p b 0.001, Table 2, Fig. 3) and NDF
19,345

19,800

19,800
19,800

(coefficient = 0.004, p = 0.017, Table 2, Fig. 3) had both a significant

eff.s

positive effect on log10 CH4 emissions, but the significant interaction be-
–
–
–

–
–

tween log10 DMI and NDF pointed out that the rates of methane emis-
–0.40
0.035

0.013

1.132
0.767

sions were depressed in diets high in fiber in comparison to less fibrous

u-CI

–
–
–

–
–
–

diets (coefficient = − 0.005, p = 0.006, Table 2, Fig. 3). Interestingly,

0.015

0.010

1.011
0.577
–0.67

after controlling for DMI and NDF there was a significant positive effect
l-CI

of log10 body mass in methane emissions (coefficient = 0.303, p =

–
–
–

–
–
–
Model r.1

0.001, Table 2), which indicates that larger ruminants produce greater
Mean

0.025

0.012

1.073
0.671
–0.54

amounts of methane emissions than smaller ruminants per unit of DMI

–
–
–

–
–
–

and for diets with similar content of fiber. Most of the variance explained
b0.001
b0.001

by the random effects was due to species phylogeny (63.6%), followed by

Diet (10.8%), Source (8.8%) and Technique (5.8%) (Table 2).
–
–
–
–

–
–
–
p
Model r.0 (r2 = 0.8960)

63.3
–16

3.3. Species effect on methane emissions

–
–
–
–

–
–
–
–
t
experiments used in the analyses = 466.

0.036
0.017

The Species model used data of 197 experiments in 41 studies car-

se

–
–
–
–

–
–
–
–

ried out in cattle, sheep, goat and red deer (Table 3). The variance ex-
Estimate

plained by the random effects was 29.3%, 24.1% and 14.0% for Diet,
–0.573
1.075

Source and Technique, respectively. Consistently with the Phylogenetic

–
–
–
–

–
–
–
–

model, log10 BM (coefficient = 0.256, p = 0.006), log10 DMI (coeffi-

Random effects

cient = 0.793, p b 0.0001) and NDF (coefficient = 0.004, p = 0.010)

var Technique
Fixed effects

all had significant positive effects on log10 CH4 emissions, and dietary
var Source
Technique

Intercept
log10 BM

var Diet

NDF depressed log10 CH4 emissions as the levels of log10 DMI increased
Source

var Sp
Units
Table 1

Diet

(interaction log10 DMI × NDF: coefficient = −0.005, p = 0.004, Table 3,

Sp

Fig. 4).
1576 F.J. Pérez-Barbería / Science of the Total Environment 579 (2017) 1572–1580

Table 2
Parameters of the full and final MCMC generalized linear models on methane emissions in ruminants (log10 CH4 g d−1) against random and fixed effects. log10 dry matter food intake
(DMI), log10 body mass (BM, kg), dietary neutral detergent fiber (NDF), and four techniques used to measure methane emissions (MM: micrometeorological techniques, OC: open circuit
indirect calorimetry, SF6: sulphur hexafluoride; the reference technique is closed circuit respiration chamber); MPER: number of days during which methane emission measurements
were recorded. n exp: number of experiments. Other acronyms as in Table 1.

Full model (n exp = 188, n sources = 40) Final model (n exp = 203, n sources = 42)

Random effects Mean l-CI u-CI eff.s p var l-CI u-CI Mean l-CI u-CI eff.s p var l-CI u-CI

Source 0.006 0.002 0.011 19,800 0.092 0.003 0.253 0.006 0.003 0.011 19,086 0.088 0.002 0.279
Sp 0.088 0.001 0.264 15,737 0.671 0.239 0.979 0.082 0.0005 0.241 16,159 0.636 0.259 0.987
Diet 0.011 0.000 0.034 16,036 0.123 0.0003 0.388 0.010 0.0003 0.030 19,800 0.108 0.001 0.403
Technique – – – – 0.006 0.0002 0.018 19,800 0.059 0.0003 0.260
Units 0.008 0.006 0.010 18,017 0.008 0.006 0.010 22,179

Fixed effects
Intercept 0.637 0.073 1.223 19,800 0.029 0.621 0.100 1.147 18,650 0.019
log10 DMI 0.825 0.603 1.040 19,800 b0.001 0.786 0.580 0.988 19,800 b0.001
log10 BM 0.289 0.099 0.490 19,800 0.003 0.303 0.117 0.489 19,398 0.001
NDF 0.004 0.001 0.007 19,800 0.021 0.004 0.001 0.006 19,800 0.017
MM 0.046 –0.148 0.244 19,800 0.636 – – – – –
OC –0.026 –0.185 0.118 19,800 0.734 – – – – –
SF6 –0.001 –0.148 0.146 19,800 0.991 – – – – –
MPER 0.002 –0.009 0.013 19,800 0.688 – – – – –
log10 DMI × NDF –0.006 –0.010 –0.001 19,800 0.008 –0.005 –0.009 –0.001 19,800 0.006

After controlling for log10 DMI, log10 BM and NDF, cattle produced
the highest CH4 emissions of the four species (coefficient = 0.920,
95% CI = [0.431, 1.413]), followed by red deer (coefficient = −0.235,
95% CI = [− 0.388, − 0.076]) and sheep (coefficient = − 0.248, 95%
CI = [−0.427, −0.063]), the latter two species did not differ between
each other, and goat was the species with the lowest CH4 emissions of
the four species (coefficient = − 0.495, 95% CI = [− 0.718, − 0.281],
Table 3, Fig. 4).
3.4. Global estimates of methane emissions from wild ruminants
The mean estimates of methane emissions for a world population
size of 214 × 106 (min = 214 × 106, max = 219 × 106) wild ruminants
(Online Source 4) using the models of Crutzen et al. (1986), Smith et al.
(2010), Franz et al. (2010), and those of this study (r.0, r.1, r.2, r.3, r.4
and Phylogenetic) varied from a minimum of 1.094 Tg y−1 in the Phylo-
genetic model to a maximum of 2.687 Tg y−1 in Crutzen et al. (1986)
model. The difference in estimated emissions when using minimum or
maximum values of the global wild ruminant population size in
Crutzen et al. (1986), Franz et al. (2010) and Phylogenetic models was
about 9%, and between 1%–2% when using Smith et al. (2010), r.0, r.1,
r.2, r.3 and r.4 models. The Phylogenetic model produced the widest
confidence interval of the estimates of methane emissions across
models (Table 4). The mean methane emission estimates produced by
all models were 11 [range = 6–14] times lower than the 15 Tg y−1 re-
ported by IPCC (Stocker et al., 2013).
4. Discussion

4.1. Sources of variability scaling methane emissions

The scaling models on methane emissions against body mass, that

controlled for the effects of study source (model r.1), and study source
and phylogeny (model r.2), both produced a slope very close and no dif-
ferent to 1.0. The results of these two models are consistent with the
findings of Smith et al. (2010) in ruminants, and Franz et al. (2010) in
ruminants and equids, who reported scaling coefficients of methane

Table 3
Parameters of a MCMC generalized linear model on methane emissions in ruminants
(log10 CH4 g d−1) to assess the effect of four animal species (C. aegagrus, C. elaphus, O. aries,
the reference level is B. taurus) controlling for body mass, feed DMI, dietary NDF, and three
random effects. Other acronyms as in Tables 1 and 2.

Mean l-CI u-CI eff.s p var l-CI u-CI

Random effects

Source 0.006 0.003 0.010 19,800 0.241 0.047 0.431

Diet 0.011 0.0003 0.034 19,800 0.293 0.011 0.678
Technique 0.006 0.0001 0.017 19,800 0.140 0.002 0.457
Units 0.008 0.006 0.010 19,800

Fixed effects
Intercept 0.920 0.431 1.413 19,458 b0.001
log10 DMI 0.793 0.581 0.994 19,800 b0.001
log10 BM 0.256 0.081 0.446 19,800 0.006
NDF 0.004 0.001 0.007 20,313 0.010
C.aegagrus –0.495 –0.718 –0.281 19,800 b0.001
Fig. 3. Predictions of the phylogenetic MCMC model in Table 2 on methane production in
C.elaphus –0.235 –0.389 –0.076 20,524 0.003
ruminants species (log10 CH4 g d−1) showing the significant interaction between log10 dry
O.aries –0.248 –0.427 –0.063 19,382 0.009
matter intake DMI (kg) and dietary neutral detergent fiber NDF. Three arbitrary levels of
log10 –0.005 –0.009 –0.002 19,684 0.004
NDF have been plotted (20%, 50%, 70%), log10 body mass (kg) has been fixed to its mean
DMI × NDF
within species. See Table 2 for the description of the model.
 F.J. Pérez-Barbería / Science of the Total Environment 579 (2017) 1572–1580 1577

included study Source, Sp and Diet (model r.3), or, Source, Sp, Diet
and Technique (model r.4) produced 95% coefficient intervals for the
slope, that comprised both, 0.75 and 1.0 scaling coefficients. Analyses
based on better quality data sets, and preferably, using data based on
a measurement of digestible energy, are needed in order to confirm
whether the energy losses by methane scale linearly with body mass
or with the same allometry than energy expenditure.
I found significant differences in methane emissions between spe-
cies per unit of dry matter intake, after accounting for BM, dietary
NDF, and random variation due to Source, Diet and Technique, with cat-
tle being the species with higher yield of methane emissions, followed
by sheep and deer, that did not differ from each other, and goat that
had the lowest methane emissions. Why some species produce higher
emissions of methane after controlling for body mass and diet it is un-
clear, but it could be due to systematic differences in the microbial eco-
system between the species of different body mass. Food in the rumen
provides mainly proteins and polymeric carbohydrates, which are
fermented by the microbial community to volatile fatty acids, NH+ 4 ,
CO2, and H2. The inhibitory effect of H2 accumulation on this process is
mediated by methanogens, mainly Archaea, maintaining a favorable en-
vironment for the microbial community for fermentation (McAllister
and Newbold, 2008). The methanogens can be found free in the
rumen fluid, attached to the rumen epithelium or particulate material,
Fig. 4. Predictions of the model in Table 3 on methane production (log10 CH4 g d−1) in four or associated as endosymbionts within rumen protozoa. Different pop-
ruminants species to show the significant interaction between log10 dry matter intake DMI ulation growth rates are expected from the methanogens associated
(kg) and dietary neutral detergent fiber NDF (two arbitrary levels of NDF have been with these different fractions, since they will be removed from the
plotted: 20%, 70%), log10 body mass (kg) has been fixed to its mean within species. See rumen at different rates, which will be also influenced by the rate of pas-
Table 3 for details on the model.
sage of digesta through the rumen system, depending on the host spe-
cies and feed (Janssen and Kirs, 2008). Little is known about what
emissions against body mass closed to 1.0 (1.053 and 0.97, 95% CI = factors determine the methanogen community, and if this microbial
[0.88, 1.07], respectively) and significatively bigger than the allometric community is specific to different ruminant species. Studies performed
coefficient of scaling energy against body mass (BM0.75). Franz et al. in cattle and sheep from different locations, even continents, showed
(2010) pointed out that if this linear scaling could be confirmed in fur- intra-species similarities in the composition of the archaeal communi-
ther studies, it would suggest that methane production might be mainly ties (Wright et al., 2008, 2007). But other studies, carried out on cattle
a factor of gut capacity, as wet gut contents scale linearly with body from different sites and using the same methodology, have shown
mass (Clauss et al., 2007). However, the results of the models that clear differences in methanogen composition (Wright et al., 2006,
2004). Whether these findings are due to methodology, host species,
Table 4
or controlled by diet or animal management remains unclear. Whatever
Estimates of global methane emissions (Tg y−1) using different models for the mean the causes, the results of the present study indicate that modeling exer-
(214 × 106), minimum (min, 210 × 106) and maximum (max, 219 × 106) wild ruminant cises that aim for accurate estimates of methane emissions to use in
population size. Other acronyms as in Table 1. See Material and methods and Table 1 for GHG inventories should take into account animal species variability.
details of each model.
It is important to highlight that despite the claim that more accurate
Model Emissions l-CI u-CI measurements of methane emissions and better replicability are
Phylogenetic Mean 1.093 0.135 14.636 achieved by using respiration chambers in comparison with data pro-
Min 1.042 0.131 13.714 duced using gas tracers (SF6, C2H6), micrometeorological, polytunnel
Max 1.134 0.138 15.200 or face mask techniques (Blaxter and Clapperton, 1965; Pinares-Patiño
Crutzen et al., 1986 Mean 2.686 – – et al., 2011, 2007), in this study, Technique, fitted as a factor, did not
Min 2.559 – –
Max 2.783 – –
have a significant effect in the models. This is good news for meta-anal-
Franz et al., 2010 Mean 1.602 1.059 2.579 yses studies on methane emissions, as it makes easier the use of data
Min 1.515 1.005 2.431 from studies carried out using a variety of techniques (Storm et al.,
Max 1.663 1.099 2.678 2012). Although it could be possible that the effect of Technique was
Smith et al., 2010 Mean 1.347 0.868 2.089
confounded by Source, however, this was not shown up in my analyses.
Min 1.333 0.872 2.037
Max 1.359 0.864 2.136 Methane emissions depend on feed intake, which is affected by
r.0 Mean 1.311 1.311 1.311 animal behavior. Published studies indicate a reduction of feed intake
Min 1.304 1.304 1.304 in animals that are confined in respiration chambers. For example,
Max 1.317 1.317 1.317 Beauchemin and McGinn (2006) and Bickell et al. (2014) reported a re-
r.1 Mean 1.318 1.270 1.319
Min 1.311 1.264 1.311
duction between 15% to 25% in feed intake for sheep and cattle, respec-
Max 1.324 1.275 1.325 tively, that were confined in respiration chambers. Robinson et al.
r.2 Mean 1.327 1.198 1.331 (2011) discussed the importance of understanding how changes in be-
Min 1.320 1.195 1.324 havior affects the feed intake of sheep in chambers and its repercussion
Max 1.333 1.201 1.338
on methane emissions. Pinares-Patiño et al. (2013) reported high re-
r.3 Mean 1.328 1.197 1.328
Min 1.321 1.194 1.321 peatability values in CH4 (g/day) and CH4 DMI (g/Kg) across two con-
Max 1.334 1.200 1.334 secutive days, but there was a significant drop in repeatability when
r.4 Mean 1.326 1.199 1.333 comparing emissions recorded in the same conditions, but spaced inter-
Min 1.319 1.196 1.326 vals of 10 to 15 days. This is why acclimatization of animals to the exper-
Max 1.332 1.202 1.339
imental conditions is crucial for obtaining consistent measurements of
1578 F.J. Pérez-Barbería / Science of the Total Environment 579 (2017) 1572–1580
methane emissions (Hellwing et al., 2012; McLean and Tobin, 1987). In
the present study the experiment length had a negligible effect on
methane emissions, which suggests that its contribution was very
small and subsumed by experimental noise, or that the length of the ex-
periments included in the data set was appropriate to record average
estimates of methane emissions across the sources of variation used in
this paper.
This study reports for the first time in the literature that as feed DMI
increased, and so methane emissions, diets rich in fiber depressed
methane yield. This could be explained by a faster passage rate of food
through the gut, as a consequence of a higher food intake when using
diets rich in fiber, which could reduce rates of fermentation by the mi-
crobial community and so diminishing H2 availability to methanogens.
Findings of some studies support this explanation, as when intake
increases digestibility decreases and so methane production (Blaxter
and Clapperton, 1965; Krishna et al., 1978; Vanderhoning et al., 1981;
Wainman, 1977).
4.2. How accurate are global estimates of methane emissions in
wild ruminants?
Mean estimates of global methane emissions from wild ruminants
using the models in this study ranged between 1.094 and
2.687 Tg yr−1. These values are between 5 and 14 times smaller than
the value presented in the IPCC report (Stocker et al., 2013 Chapter 6,
Table 6.8), which was 15 Tg yr− 1. Hristov (2012) estimated
0.28 Tg yr− 1 the emissions of present United States population of
bison (0.5 × 106), elk (1 × 106), mule deer (4 × 106) and whitetail
deer (25 × 106), using average emissions of these species (21, 16, 10
and 10 CH4 g DMI Kg−1, respectively) and 2% DMI per body weight,
which for that population size is within the range of the estimate for
the world population of ruminants provided in this study. How can be
possible such a large discrepancy between the estimates of the present
study and those reported in the IPCC report? Detailed scrutiny of the
emissions value reported in the IPCC report (Stocker et al., 2013) re-
vealed a chain of inaccuracies to produce the reported value. First, the
authors were not faithful to the original sources from which the infor-
mation was taken, leading to wrong estimates of methane emissions;
and second, the value of the ruminants world population size was
taken from a non-existing literature reference, which makes impossible
to assess its accuracy. The chain of inaccuracies is described as follows.
Stocker et al. (2013) in Chapter 6, Table 6.8, reported a value of
15 Tg yr−1 of methane emissions from “wild animals excluding
termites” (but actually, it should have been cited as “wild ruminants ex-
cluding termites”) allegedly calculated by Denman et al. (2007) using a
bottom-up model. However, the figure was not originally estimated by
Denman et al. (2007), but cited in Denman et al. (2007) from a top-
down model from Houweling et al. (2000). Furthermore, the estimate
in Houweling et al. (2000) was reported in their Table 1 as the methane
emissions from preindustrial wild animals (buffaloes and bisons) calcu-
lated by Chappellaz et al. (1993). Chappellaz et al. (1993) calculated
methane emissions of “wild animals” in the preindustrial Holocene
using data from Crutzen et al. (1986), concretely, the methane emis-
sions from an estimated global population of buffaloes of 120 million,
6 Tg yr−1. Chappellaz et al. (1993), without any justification, assumed
that the 6 Tg yr−1 could be converted to 15 Tg yr−1 methane emissions
as an approximation of the preindustrial Holocene population of “wild
animals” (it is not sure if they meant “wild ruminants” or “wild rumi-
nants and nonruminant herbivores”). Crutzen et al. (1986) using the
methodology described in Methods in this paper (Eq. (2), 9% of meth-
ane yield per unit of gross energy), together with a world population es-
timate of 27 million of northern ruminants (except China), mainly red
deer and reindeer, and 100–500 million of wild ruminants in tropical
and sub-tropical regions, and their body masses and food intake, esti-
mated the present methane emissions from wild ruminants between
2 Tg yr−1 and 6 Tg yr−1. These emissions are between 2 and 7 times
lower than those reported by Stocker et al. (2013) in the IPCC report,
and as an average two times higher than those estimated in the present
study. My estimates were based on a wild ruminants population size of
214 million (range = 210–219), smaller than the population size used
by Crutzen et al. (1986), who used the population size reported by Mc-
Dowell (1976) (in Crutzen et al., 1986). Unfortunately, it could not been
verified the sources that McDowell used to estimate a population size of
wild ruminants of 500 million, as this reference does not seem to exist
in the literature, at least not in the way cited by Crutzen et al. (1986)
and other later studies. Based on the present paper literature research
on wild ruminants population size, it seems unlikely that the actual
population size can be 286 million bigger. Furthermore, although in de-
veloped countries there has been a trend of increase of some wild rumi-
nant populations, most reports from IUCN Specialist Groups state a
decrease of ungulate populations in the last few decades.
Crutzen et al. (1986) and Hristov (2012) stated how uncertain the
figures of the wild ruminants population sizes are, and consequently
how inaccurate are the estimates of CH4 emissions. I agree with this
statement, as during the process of collecting data on population size
of wild ruminants I experienced that the information on many popula-
tions was not up to date, or only population size approximations were
available, and for some species data was completely unknown. Another
important source of variation between CH4 estimates is species body
mass. Hristov (2012) comments on differences of 30% in the body
mass assigned to bison across studies on CH4 emissions. Despite these
difficulties when estimating CH4 emissions at global scale, the modeling
exercise presented in this paper is invaluable to assess methane emis-
sions from ruminant species fed herbage diets, accounting for phyloge-
ny, body mass and dietary fiber, when their population sizes are known,
which is possible in many local populations.
The results clearly indicate significant differences in methane emis-
sions between species per unit of dry matter intake, cattle producing
higher levels of emissions than other species. If cattle is a high level
methane producer, then, caution should be taken when using cattle
methane yield per unit of gross energy as a parameter to estimate meth-
ane global emissions of the wild ruminant and pseudo-ruminant world
population. Franz et al. (2010) and Hristov (2012) also mention how
important is to take into account differences between species in meth-
ane output to produce accurate estimates of enteric CH4 emissions in
wild ruminants and equids. For example, Crutzen et al. (1986) estimat-
ed global methane emissions from wild ruminants using 9% methane
yield per unit of gross energy, taken this value from experiments on
Indian cattle fed of low quality feed (Krishna et al., 1978). If this yield
value overestimates the emissions produced by other species of
ruminants, then, this could help to explain why the methane global
emissions presented here and calculated using Crutzen et al. (1986)
method are on average twice as high as those calculated by the other
models.
It is warned on the inaccuracy of the figures of global methane emis-
sions by wild ruminants reported in the literature, the risks of use them
in global warming models to predict climate change, and the limitations
faced to calculate accurate global emissions from wild ungulates. These
results are intended to improve methane emissions inventories, and to
provide information to be used in bottom-up modeling exercises on
global warming.
5. Conclusions
Scaling log10 CH4 g d− 1 against log10 body mass (kg) in ruminants
produces a slope very close to 1 (1.073). The incorporation of species
relatedness, diet and technique of measuring as random effects into
the scaling model, reduces the slope to 0.868 and also widens its con-
fident limits to such an extent that includes both, the scaling coeffi-
cient of metabolic requirements to body mass (0.75) and isometry
(1.0).
 F.J. Pérez-Barbería / Science of the Total Environment 579 (2017) 1572–1580
Scaling models that include DMI and dietary fiber indicate that
although both increase CH4 emissions, dietary fiber depresses CH4 as
the level of DMI increases.
Cattle produce more CH4 per unit of DMI than red deer, sheep or
goats. There are no significant differences between CH4 emissions
produced by red deer and sheep.
The global emissions of wild ruminants from this paper are smaller
(1.094–2.687 Tg y−1) than those presented in the reports of the Inter-
governmental Panel on Climate Change (15 Tg yr− 1). This could be
due to (i) inaccuracies when reporting and using data from the litera-
ture, (ii) the uncertainty of the world wild ruminant population size
(500 × 106 vs. 214 × 106 estimated in this study), and (iii) using the
mean methane output of cattle, a high methane producer, as represen-
tative value of the methane emissions of wild ruminants. Inventories on
methane emissions by wild ungulates should take into account the
effects of animal species and diet. The main limitation researchers'
face in calculating accurate global CH4 emissions from wild ungulates
is a lack of reliable information on their population sizes.
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.scitotenv.2016.11.175.
Acknowledgments
FJP-B was granted with a visiting professor fellowship by the Univer-
sity of Cordoba during the writing of this paper. FJP-B thank Dr Ramón
Soriguer-Scofet (Estación Biológica de Doñana, CSIC, Spain) for access
to on-line library facilities; and The European Union Lifelong learning
programme (Leonardo da Vinci) and Global Learning, for supporting
post-graduate students that assisted the work of this study. Dr Marcus
Clauss, Dr. Bob Mayes and one anonymous referee contributed with
useful comments on a draft of this study.
References
Beauchemin, K.A., McGinn, S.M., 2006. Methane emissions from beef cattle: effects of
fumaric acid, essential oil, and canola oil. Anim. Sci. 84, 1489–1496.
Bickell, S.L., Revell, D.K., Toovey, A.F., Vercoe, P.E., 2014. Feed intake of sheep when
allowed ad libitum access to feed in methane respiration chambers. Anim. Sci. 92,
2259–2264.
Blaxter, K.L., 1962. The Energy Metabolism of Ruminants. Hutchinson & CO, London, Great
Britain.
Blaxter, K.L., Clapperton, J.L., 1965. Prediction of amount of methane produced by rumi-
nants. Br. J. Nutr. 19:511–522. http://dx.doi.org/10.1079/BJN19650046.
Brook, E.J., Severinghaus, J.P., 2011. Methane and megafauna. Nat. Geosci. 4:271–272.
http://dx.doi.org/10.1038/ngeo1140.
Chappellaz, J.A., Fung, I.Y., Thompson, A.M., 1993. The atmospheric CH4 increase since the
last glacial maximum. Tellus B Chem. Phys. Meteorol. 45:228–241. http://dx.doi.org/
10.1034/j.1600-0889.1993.t01-2-00002.x.
Clauss, M., Schwarm, A., Ortmann, S., Streich, W., Hummel, J., 2007. A case of non-scaling
in mammalian physiology? Body size, digestive capacity, food intake, and ingesta
passage in mammalian herbivores. Comp. Biochem. Physiol. -Mol. Integr. Physiol.
148, 249–265.
Clauss, M., Steuer, P., Müller, D.W.H., Codron, D., Hummel, J., 2013. Herbivory and body
size: allometries of diet quality and gastrointestinal physiology, and implications
for herbivore ecology and dinosaur gigantism. PLoS One 8, e68714.
Crutzen, P.J., Crutzen, P.J., Aselmann, I., Seiler, W., 1986. Methane production by domestic
animals, wild ruminants and other herbivorous fauna, and humans. Tellus 38B,
271–284.
Demment, M.W., Van Soest, P.J., 1985. A nutritional explanation for body size patterns of
ruminant and nonruminant herbivores. Am. Nat. 125, 641–672.
Denman, K.L., Denman, K.L., et al., 2007. Couplings between changes in the climate system
and biogeochemistry. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M.,
Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), Climate Change 2007: The Physical Science
Basis. Contribution of Working Group I to the Fourth Assessment Report of the Inter-
governmental Panel on Climate Change. Cambridge University Press, Cambridge,
United Kingdom and New York, NY, USA, pp. 499–587.
R Development Core Team, 2015. R: A Language and Environment for Statistical Computing.
R Foundation for Statistical Computing, Viena, Austria.
Franz, R., Soliva, C.R., Kreuzer, M., Steuer, P., Hummel, J., Clauss, M., 2010. Methane
production in relation to body mass of ruminants and equids. Evol. Ecol. Res. 12,
727–738.
Geist, V., 1974. On the relationship of social evolution and ecology in ungulates. Am. Zool.
14, 205–220.
Hadfield, J.D., 2010. MCMC methods for multi-response generalized linear mixed models:
the MCMCglmm R package. J. Stat. Softw. 33, 1–22.
1579
Harvey, P.H., Purvis, A., 1991. Comparative methods for explaining adaptations. Nature
351, 619–624.
Hellwing, A.L.F., Lund, P., Weisbjerg, M.R., Brask, M., Hvelplund, T., 2012. Technical note:
test of a low-cost and animal-friendly system for measuring methane emissions
from dairy cows. J. Dairy Sci. 95, 6077–6085.
Hofmann, R.R., Stewart, D.R.M., 1972. Grazer or browser: a classification based on the
stomach-structure and feeding habitats of east African ruminants. Mammalia 36,
226–240.
Houweling, S., Dentener, F., Lelieveld, J., 2000. Simulation of preindustrial atmospheric
methane to constrain the global source strength of natural wetlands. J. Geophys.
Res.-Atmos. 105:17243–17255. http://dx.doi.org/10.1029/2000JD900193.
Hristov, A.N., 2012. Historic, pre-European settlement, and present-day contribution of
wild ruminants to enteric methane emissions in the United States. J. Anim. Sci. 90:
1371–1375. http://dx.doi.org/10.2527/jas.2011-4539.
Janssen, P.H., Kirs, M., 2008. Structure of the archaeal community of the rumen. Appl.
Environ. Microbiol. 74, 3619–3625.
Jarman, P.J., 1974. The social organisation of antelope in relation to their ecology. Behav-
iour 48, 215–266.
Krishna, G., Razdan, M.N., Ray, S.N., 1978. Effect of nutritional and seasonal-variations on
heat and methane production in Bos indicus. Indian J. Anim. Sci. 48, 366–370.
Martin, C., Morgavi, D., Doreau, M., 2010. Methane mitigation in ruminants: from microbe
to the farm scale. Animal 4:351–365. http://dx.doi.org/10.1017/S1751731109990620.
McAllister, T., Newbold, C., 2008. Redirecting rumen fermentation to reduce
methanogenesis. Aust. J. Exp. Agric. 48, 7–13.
McLean, J.A., Tobin, G., 1987. Animal and Human Calorimetry. Cambridge University Press,
Cambridge, United Kindom.
Mellor, D., 2015. Positive animal welfare states and encouraging environment-focused
and animal-to-animal interactive behaviours. N. Z. Vet. J. 63:9–16. http://dx.doi.org/
10.1080/00480169.2014.926800.
Ministry of Agriculture, F. and F., Great Britain, 1975. Energy Allowances and Feeding
Systems for Ruminants. Technical bulletin. Her Majesty's Stationery Office, London,
Great Britain.
Mitchell, B., Staines, B., Welch, D., 1977. Ecology of Red Deer. A Research Review Relevant
to their Management in Scotland. Institute of Terrestrial Ecology, Cambridge, United
Kindom.
Moss, A.R., Jouany, J.P., Newbold, J., 2000. Methane production by ruminants: its contribu-
tion to global warming. Ann. Zootech. 49, 231–253.
Mueller, D.W., Codron, D., Meloro, C., Munn, A., Schwarm, A., Hummel, J., Clauss, M., 2013.
Assessing the Jarman-Bell Principle: scaling of intake, digestibility, retention time and
gut fill with body mass in mammalian herbivores. Comp. Biochem. Physiol. -Mol.
Integr. Physiol. 164, 129–140.
Murtaugh, P.A., 2014. In defense of P values. Ecology 95, 611–617.
Nowak, R.M., 1999. Walker's Mammals of the World. The Johns Hopkins University Press,
Baltimore.
Paustian, K., Ravindranath, N.H., van Amstel, A., Gytarsky, M., Kurz, W.A., Ogle, S.,
Richards, G., Somogyi, Z., 2006. IPCC Guidelines for National Greenhouse Gas Invento-
ries. Agriculture, Forestry and Other Land Use. Volume 4.
Pavao-Zuckerman, M.A., Waller, J.C., Ingle, T., Fribourg, H.A., 1999. Methane emissions of
beef cattle grazing tall fescue pastures at three levels of endophyte infestation.
J. Environ. Qual. 28, 1963–1969.
Pérez-Barbería, F.J., Gordon, I.J., 2001. Relationships between oral morphology and feed-
ing style in the Ungulata: a phylogenetically controlled evaluation. Proc. R. Soc. B
Biol. Sci. 268, 1023–1032.
Pérez-Barbería, F.J., Robertson, E., Soriguer, R., Aldezabal, A., Mendizabal, M., Pérez-
Fernández, E., 2007. Why do polygynous ungulates segregate in space? Testing the
activity budget hypothesis in Soay sheep. Ecol. Monogr. 77, 631–647.
Pérez-Barbería, F.J., Pérez-Fernández, E., Robertson, E., Alvarez-Enriquez, B., 2008. Does
the Jarman-Bell principle at intra-specific level explain sexual segregation in polygy-
nous ungulates? Sex differences in forage digestibility in Soay sheep. Oecologia 157,
21–30.
Pinares-Patiño, C.S., D'Hour, P., Jouany, J.P., Martin, C., 2007. Effects of stocking rate on
methane and carbon dioxide emissions from grazing cattle. Agric. Ecosyst. Environ.
121, 30–46.
Pinares-Patiño, C.S., Holmes, C., Lassey, K., Ulyatt, M., 2008. Measurement of methane
emission from sheep by the sulphur hexafluoride tracer technique and by the calori-
metric chamber: failure and success. Animal 2:141–148. http://dx.doi.org/10.1017/
S1751731107000857.
Pinares-Patiño, C.S., Lassey, K., Martin, R., Molano, G., Fernandez, M., MacLean, S.,
Sandoval, E., Luo, D., Clark, H., 2011. Assessment of the sulphur hexafluoride (SF6)
tracer technique using respiration chambers for estimation of methane emissions
from sheep. Anim. Feed Sci. Technol. 166–67:201–209. http://dx.doi.org/10.1016/j.
anifeedsci.2011.04.067.
Pinares-Patiño, C.S., Hickey, S., Young, E., Dodds, K., MacLean, S., Molano, G., Sandoval, E.,
Kjestrup, H., Harland, R., Hunt, C., Pickering, N., McEwan, J., 2013. Heritability esti-
mates of methane emissions from sheep. Animal 7:316–321. http://dx.doi.org/10.
1017/S1751731113000864.
Rira, M., Morgavi, D.P., Archimede, H., Marie-Magdeleine, C., Popova, M., Bousseboua, H.,
Doreau, M., 2015. Potential of tannin-rich plants for modulating ruminal microbes
and ruminal fermentation in sheep. J. Anim. Sci. 93:334–347. http://dx.doi.org/10.
2527/jas.2014-7961.
Robbins, C.T., 1993. Wildlife Feeding and Nutrition. Academic Press, San Diego.
Robinson, D.L., Bickell, S.L., Toovey, A.F., Revell, D.K., Vercoe, P.E., 2011. Factors affecting
variability of feed intake of sheep with ad libitum access to feed and the relationship
with daily methane production. Proc.Assoc.Advmt.Anim.Breed.Genet. 19, 159–162.
Smith, F.A., Elliott, S.M., Lyons, S.K., 2010. Methane emissions from extinct megafauna.
Nat. Geosci. 3, 374–375.
1580 F.J. Pérez-Barbería / Science of the Total Environment 579 (2017) 1572–1580
Stocker, T.F., Qin, D., Plattner, G.K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y.,
Bex, V., Midgley, P.M., 2013. Climate Change 2013: The Physical Science Basis.
Cambridge University Press, Cambridge, United Kingdom.
Storm, I., Hellwing, A., Nielsen, N., Madsen, J., 2012. Methods for measuring and estimat-
ing methane emission from ruminants. Animals 2:160–183. http://dx.doi.org/10.
3390/ani2020160.
Vanderhoning, Y., Wieman, B.J., Steg, A., Vandonselaar, B., 1981. The effect of fat supple-
mentation of concentrates on digestion and utilization of energy by productive
dairy cows. Neth. J. Agric. Sci. 29, 79–92.
Wainman, F.W., 1977. Digestibility and balance in ruminants. Proc. Nutr. Soc. 36, 195–202.
Wickham, H., 2009. Elegant Graphics for Data Analysis (New York).
Wieren, S.E., Bakker, J.P., 2008. The impact of browsing and grazing herbivores on biodi-
versity. In: Gordon, I.J., Prins, H.H.T. (Eds.), The Ecology of Browsing and Grazing, Eco-
logical Studies. Springer, Berlin Heidelberg, pp. 263–292.
Wright, A.D.G., Williams, A.J., Winder, B., Christophersen, C.T., Rodgers, S.L., Smith, K.D.,
2004. Molecular diversity of rumen methanogens from sheep in western Australia.
Appl. Environ. Microbiol. 70, 1263–1270.
Wright, A.D.G., Toovey, A.F., Pimm, C.L., 2006. Molecular identification of methanogenic
archaea from sheep in Queensland, Australia reveal more uncultured novel archaea.
Anaerobe 12, 134–139.
Wright, A.D., Auckland, C.H., Lynn, D.H., 2007. Molecular diversity of methanogens in
feedlot cattle from Ontario and Prince Edward Island, Canada. Appl. Environ.
Microbiol. 73, 4206–4210.
Wright, A.D., Ma, X., Obispo, N.E., 2008. Methanobrevibacter phylotypes are the dominant
methanogens in sheep from Venezuela. Microb. Ecol. 56, 390–394.
Wuebbles, D.J., Hayhoe, K., 2002. Atmospheric methane and global change. Earth-Sci. Rev.
57:177–210. http://dx.doi.org/10.1016/S0012-8252(01)00062-9.