NATURE CLIMATE CHANGE | VOL 6 | OCTOBER 2016 | www.nature.com/natureclimatechange 893
opinion & comment
the FACE experiments, which typically
increase CO2 from ~370ppmto ~550ppm)
(Fig.1). Furthermore, S15’s synthesis of
FACE data is incomplete as it omits several
years of published data10,11, and incorrectly
estimates an overall eect size by taking
the median across experiments, species
and years, rather than calculating a more
appropriate response ratio12.
S15 concludes that CESM1-BGC, the
ESM most consistent with the satellite
NPP estimates, is an improvement over
other ESMs, likely due to its inclusion of
explicit carbon–nitrogen interactions. We
agree that the inclusion of such interactions
in ESMs is a desirable objective, and
that neglect of these in ‘carbon only’
ESMs risks overestimating long-term
CO2 eects on NPP2. However, it is
premature to reach this conclusion given
the inability of CESM1-BGC to capture
the magnitude of recent CO2 uptake13 or
even (uniquely among models tested) the
‘sign’ of the relationship between tropical
land temperatures and CO2 uptake14. In
addition, the land surface model (CLM4) in
CESM1-BGC underestimates the measured
NPP response to elevated CO2 from the
two longest-running FACE experiments—
predicting a smaller response than ten
other ecosystem models that included
nutrient limitations on NPP15.
In summary the comparison of satellite
and FACE estimates of CO2 fertilization
is invalid, and the discussion of nitrogen
limitations is based on a single model
that poorly represents the response of
1. Smith, W.K. etal. Nat. Clim. Change 6, 306–310 (2016).
2. Hungate, B.A. etal. Science 302, 1512–1513 (2003).
3. Zhang, Y., Xu, M. & Adams, J. Global Ecol. Biogeogr.
18, 280–290 (2009).
4. Medlyn, B.E. Science 333, 1093 (2011).
5. Jung, M. etal. J.Geophys. Res. 116, G00J07 (2011).
6. Liu, Y.Y. etal. Glob. Ecol. Biogeogr. 22, 692–705 (2013).
7. Norby, R.J. etal. Prod. Natl Acad. Sci. USA 102, 18052–18056 (2005).
8. Iversen, C.M., Ledford, J. & Norby, R.J. New Phytol.
179, 837–847 (2008).
9. Franks, P.J. etal. New Phytol. 197, 1077–1094 (2013).
10. Talhelm, A.F. etal. Glob. Change Biol. 20, 2492–2504 (2014).
11. Reich, P.B. & Hobbie, S.E. Nat. Clim. Change 3, 278–282 (2013).
12. Hedges, L.V., Gurevitch, J. & Curtis, P.S. Ecology
80, 1150–1156 (1999).
13. Homan, F.M. etal. J.Geophys. Res. 119, 141–162 (2013).
14. Wenzel, S., Cox, P.M., Eyring, V. & Friedlinsgtein, P.
J.Geophys. Res. 119, 794–807 (2014).
15. Zaehle, S. etal. New Phytol. 202, 803–822 (2014).
Martin G. De Kauwe1*, Trevor F. Keenan1,
Belinda E. Medlyn2, I. Colin Prentice1,3 and
1Macquarie University, Department of Biological
Sciences, North Ryde, New South Wales
2109, Australia. 2Hawkesbury Institute for
the Environment, Western Sydney University,
Locked Bag 1797, Penrith, New South Wales,
Australia. 3AXA Chair Programme in Biosphere
and Climate Impacts, Department of Life
Sciences, Imperial College London, Silwood Park
Campus, Buckhurst Road, Ascot SL5 7PY, UK.
Emissions from cattle farming in Brazil
To the Editor — de Oliveira Silva
and colleagues1 have proposed that, if
decoupled from deforestation, increasing
beef consumption may reduce greenhouse
gas emissions, while at the same time
suggesting that reducing consumption
may not signicantly alter greenhouse gas
emissions. However, the analysis contains
unrealistic assumptions and disregards a
series of other analyses corroborated by
historical data, aecting the robustness of
the conclusions. Sustainable intensication
is presented as a feasible socioecological
solution, despite the fact that this concept
is still a matter of controversy. At the most
general level, it lacks any solid empirically
based mechanism. More specically, it fails
to address equity and local governance
aspects that ought to be inherent in
Furthermore, the authors assume a
scenario in which deforestation can be
decoupled from changes in pasture area,
something that has not happened in the
historical record of the Brazilian Cerrado.
is assumption is based on the idea that
increases in yield eciency will result
in spare land returning to its natural
state3. Historically, however, agricultural
productivity increases have usually been
accompanied by farmland expansion4,5,
tomeet growing demand: this is oen
referred to as the Jevons paradox by
agricultural economists6. e authors
may have reasons to doubt the substantial
empirical evidence supporting this issue,
but they should acknowledge their rejection
of it in their underlying assumptions.
Similarly, their assumptions of prot
maximization and construction of a
production-optimization model are
problematic and arbitrary, considering the
voluminous existing literature showing
the importance of deviations from the
maximization motive7 and the need to
explicitly grapple with the assumptions
made in any optimization analysis.
e analysis does not take into
consideration the local dynamics of
small farming and indigenous resource
management. Livestock production by
traditional peoples and small farmers
is generally regarded as less harmful
to biodiversity and more sustainable
than intensive livestock on exotic grass
monocultures, although the outcomes are
very context specic8. e assumption that
the Cerrado may behave as a single large
prot-maximizing farm does not reect
the socioeconomic diversity of extant
landholders or the remarkable gamma
diversity of its various ecosystems.
Another questionable assumption
is the idea that pasture recovery can be
accomplished with fertilization in most
of the Cerrado, which is implausible even
before accounting for its negative eects on
soil, water, and greenhouse gas emissions.
e model also assumes a xed value for
emissions as a result of deforestation in
the Cerrado, neglecting the ecological
heterogeneity of the biome. e authors
propose recovery of degraded areas using
exotic grass, even though such exotic
species have potentially profound eects
on the functioning and biodiversity of
the Cerrado9. Furthermore, the model
ignores the regrowth of woody vegetation
when pasture is taken out of production.
us, it eectively assumes that secondary
succession back to forest, which results
in carbon sequestration in biomass and
carbon soil, can never occur10. ese
assumptions limit the practical utility of
this modelling exercise. ❐
1. de Oliveira Silva, R. etal. Nat. Clim. Change 6, 493–497 (2016).
2. Loos, J. etal. Front. Ecol. Environ.12, 356–361 (2014).
3. Phalan, B. etal. Science 351, 450–451 (2016).
4. Rudel, T. K. etal. Proc. Natl Acad. Sci. USA 106, 20675–20680 (2009).
5. Perfecto, I. & Vandermeer, J. H. Proc. Natl Acad. Sci. USA
107, 5786–5791 (2010).
6. Polimeni, J. M. & Polimeni, R. I. Ecol. Complexity 3, 344–353 (2006).
7. Brown, C. etal. PLoS ONE 9, e114213 (2014).
894 NATURE CLIMATE CHANGE | VOL 6 | OCTOBER 2016 | www.nature.com/natureclimatechange
opinion & comment
8. Giroldo, A. B. & Scariot, A. Biol. Conserv. 191, 150–158 (2015).
9. Pivello, V. R., Shida, C. N. & Meirelles, S. T. Biodivers. Conserv.
8, 1281–1294 (1999).
10. Poorter, L. etal. Nature 530, 211–214 (2016).
Fernando F. Goulart1*, Ivette Perfecto2,
JohnVandermeer3, Doug Boucher4,
M.JahiChappell5, Geraldo Wilson Fernandes6,
Aldicir Scariot7, Marcelo Corrêa da Silva8,
Washington Oliveira9, Rebecca Neville10,
JamesMoore11, Mercedes Bustamante9,
SoniaRibeiro Carvalho1 and
1Universidade Federal de Minas Gerais,
Análise e Modelagem de Sistemas
Ambientais/Centro de Sensoriamento
Remoto, Belo Horizonte, MG, CEP 31270-
900, Brazil. 2School of Natural Resources
and Environment, University of Michigan,
AnnArbor, Michigan 48109, USA.
3Department of Ecology and Evolutionary
Biology, University of Michigan, Ann Arbor,
Michigan 48109, USA. 4Union of Concerned
Scientists, 1825 K Street, NW, Washington
DC 20006, USA. 5Institute for Agriculture
and Trade Policy, Minneapolis, Minnesota
55404, USA. 6Universidade Federal de Minas
Gerais, Ecologia Evolutiva & Biodiversidade/
DBG, 30161-970 Belo Horizonte MG,
Brazil. 7Embrapa Genetic Resources and
Biotechnology, Brasília, DF 70770-917,
Brazil. 8Universidade Federal de Grande
Dourados, Faculdade de Ciências Agrária,
Itahum, Postal code 533, Brazil. 9Universidade
de Brasília, Programa de Pós-graduação
em Ecologia, Departamento de Ecologia,
Brasília, 70910-900, Brazil. 10Department of
Teaching and Learning, Washington State
University Vancouver, Vancouver, Washington
98686, USA. 11School of Biological Sciences,
Washington State University Vancouver,
Vancouver, Washington 98686, USA.
Reply to ‘Emissions from cattle farming in Brazil’
de Oliveira Silva et al. reply —
Goulertetal. make some interesting
observations about the context of our
study, its modelling assumptions and data.
We clarify these issues but refute that our
study is unrealistic or misleading. Indeed,
we have been conservative with some
assumptions and it would be possible and
plausible to accentuate the counterintuitive
result we present.
In our reference to sustainable
agricultural intensication (SAI) we note
the contested nature of the concept and do
not imply a comprehensive characterization
of the term. is includes the equity
and governance trade-os undoubtedly
encountered in more granular research
on mitigation. Our contribution provides
one mathematical example of a plausible
SAI scenario developed at a meaningful
scale. We hope it partly lls a conspicuous
gap in the literature, largely populated by
normative conceptual papers rather than
‘empirically based mechanisms’ that might
form policy evidence.
We suggest that the scenarios are based
on sound empirical evidence, referenced
in our supplementary information.
is includes the recently observed
decoupled livestock deforestation (DLD)
scenario that resulted from more rigorous
deforestation control and a changing
market environment1,2. e DLD contrasts
with the coupled livestock deforestation
scenarios, which encompass worst case
assumptions about how deforestation
responds to demand. We suggest these are
likely to accommodate potential Jevons
e prot maximization assumption is
contestable, but we note that alternative
assumptions are no less subjective.
Furthermore, deviations from prot
maximization will not signicantly aect
the results or main conclusions. is is
because the level of intensication is not
based on prot maximization, as land
availability and demand are exogenous
to our model. In unreported analysis
other objective functions were tested (for
example, minimization of land use change)
with similar results.
While important, the heterogeneity
of local ecosystem dynamics and
gamma diversity are largely beyond the
resolution of the model we employed.
Nevertheless, we can draw some
conclusions in relation to the impacts of
intensication on biodiversity. We contest
the characterization of large intensive
farms versus smallholdings suggested by
Goulert and co-authors: recent monitoring
suggests the opposite3–5. Due to legal
enforcement, large ranchers are reducing
deforestation to avoid prosecution, while
signicant deforestation is attributable to
ere is considerable experimental
and practical evidence showing that
pasture recovery can be accomplished
with fertilization in much of the Cerrado6.
Moreover our scenarios account for all
related greenhouse gases using a life cycle
approach. Since little nitrogen is applied in
the Cerrado7, the issue of water pollution
is negligible. Water consumption for
intensication measures is also small,
demand being mostly for livestock. On
deforestation emissions, we suggest that
it is impossible to know in advance where
deforestation is going to happen in the
biome for the period of study. We are
condent that alternative assumptions on
which physiognomies would be converted
to grasslands would be at least as open to
e study proposes recovery of degraded
areas already planted with exotic grasses.
We stated that recovery strategies are based
on existing Brachiaria spp. pastures as
the preferred species for pasture recovery,
productivity and costs (see supplementary
information). We also note evidence that
degraded pastures have worse eects
on ecosystem function than productive
pastures8. e use of native Cerrado
species for cattle production is of minor
Finally, there is no reason to believe that
land would be abandoned or taken out of
production within the demand range we
studied. Note that the scenarios were based
on projections to 2030. Even in the demand
scenario of 30% below baseline (DBAU–30%)
productivity would remain approximately
at the current level. ❐
1. Lapola, D.M. et al. Nat. Clim. Change 4, 27–35 (2014).
2. Arima, E.Y., Barreto, P., Araújo, E. & Soares-Filho, B. Land Use
Policy 41, 465–473 (2014).
3. Agricultural Census 2006 (e Brazilian Institute of Geography
and Statistics, 2015); http://www.sidra.ibge.gov.br/bda/acervo/
4. FAO ST AT (Food and Agriculture Organization of the United
Nations, 2015); http://faostat3.fao.org/browse/R/RL/E
5. Nepstad, D. et al. Science 344, 1118–1123 (2014).
6. Assad, E.D. et al. Biogeosciences 10, 6141–6160 (2013).
7. Cederberg, C., Meyer, D. & Flysjö, A. Life Cycle Inventory of
Greenhouse Gas Emissions and Use of Land and Energy in
Brazilian Beef Production (Swedish Institute for Food and
Biotechnology, 2009); https://www.diva-portal.org/smash/get/
8. Parrotta, J.A. Agric. Ecosyst. Environ. 41, 115–133 (1992).
9. Ferraz, J.B. S. & de Felício, P.E. Meat Sci.
84, 238–243 (2010).
R. de Oliveira Silva1,2*, L. G. Barioni3, and
1Research Division, SRUC, West Mains
Road, Edinburgh EH9 3JG, UK. 2School of
Mathematics, The University of Edinburgh,
Mayﬁeld Road, Edinburgh EH9 3JZ, UK.
3Embrapa Agriculture Informatics,
CEP 13083-886 Campinas-SP, Brazil.