Attributing the increase of atmospheric CO2 to emitters and absorbers

Abstract

Climate change policies need to consider the contribution of each emitting region to the increase in atmospheric carbon dioxide. We calculate regional attributions of increased atmospheric CO2 using two different assumptions about land sinks. In the first approach, each absorber region is attributed 'domestic sinks' that occur within its boundaries. In the second, alternative approach, each emitter region is attributed 'foreign sinks' that it created indirectly through its contribution to increasing CO2. We unambiguously attribute the largest share of the historical increase in CO 2 between pre-industrial times and the present-day period to developed countries. However, the excess CO2 in the atmosphere since pre-industrial times attributed to developing countries is greater than their share of cumulative CO2 emissions. This is because a greater fraction of their emissions occurred more recently. If emissions remain high over the coming decades, the share of excess CO2 attributable to developing countries will grow, and the sink service provided by forested regions - in particular those with tropical forest - to other regions will depend critically on future tropical land-use change.

Full text

ARTICLES
PUBLISHED ONLINE: 14 JULY 2013 | DOI: 10.1038/NCLIMATE1942
Attributing the increase in atmospheric CO
2
to
emitters and absorbers
P. Ciais
1,2
*
†
, T. Gasser
1†
, J. D. Paris
1
, K. Caldeira
3
, M. R. Raupach
4
, J. G. Canadell
4
, A. Patwardhan
5
,
P. Friedlingstein
6
, S. L. Piao
2,7
and V. Gitz
8
Climate change policies need to consider the contribution of each emitting region to the increase in atmospheric carbon dioxide.
We calculate regional attributions of increased atmospheric CO
2
using two different assumptions about land sinks. In the
first approach, each absorber region is attributed ‘domestic sinks’ that occur within its boundaries. In the second, alternative
approach, each emitter region is attributed ‘foreign sinks’ that it created indirectly through its contribution to increasing
CO
2
. We unambiguously attribute the largest share of the historical increase in CO
2
between pre-industrial times and the
present-day period to developed countries. However, the excess CO
2
in the atmosphere since pre-industrial times attributed
to developing countries is greater than their share of cumulative CO
2
emissions. This is because a greater fraction of their
emissions occurred more recently. If emissions remain high over the coming decades, the share of excess CO
2
attributable to
developing countries will grow, and the sink service provided by forested regions—in particular those with tropical forest—to
other regions will depend critically on future tropical land-use change.
D
ecreasing the risk of dangerous climate change requires
a decline in future emissions of CO
2
to the atmosphere.
There are three controls on the increase of atmospheric CO
2
(ref. 1): fossil fuel and cement emissions; land-use change emissions
mainly from deforestation (∼11% of all emissions from human
activity
1
); and land and ocean sinks. Climate policies mostly aim
at reducing fossil fuel emissions because they constitute the largest
flux driving the global increase of CO
2
and much of the emissions
come from point sources (for example, cities, power plants) that
are potentially verifiable or from sectors involving a relatively
small number of major emitters (for example, transportation). To
mitigate deforestation emissions, the focus is on reduced tropical
deforestation and degradation
2
.
Land and ocean sinks are usually not included in global
mitigation policy frameworks at the moment because they are
considered a ‘common service’. The goal here is to attribute CO
2
sinks to regions where sink fluxes occur. By the rules of the
Kyoto Protocol for instance, only a very small fraction of managed
land can be credited as carbon sinks by countries, and this has
already ignited controversies
3
, whereas the remainder of the land
is considered to be a non-human-induced sink
4,5
. Yet, there is
causality between sinks and anthropogenic CO
2
emissions. During
the Holocene epoch before the industrial era, the atmospheric
CO
2
concentration varied by small amounts, showing that in the
absence of anthropogenic emissions, land and ocean sinks must
be close to zero globally. It is the human-caused emission of
CO
2
from fossil fuel and land-use change that causes the present
ocean sink, mainly through increased partial pressure of CO
2
between atmospheric and ocean-dissolved CO
2
(ref. 6). It is also
1
Laboratoire des Sciences du Climat et de l’Environnement, CEA-CNRS-UVSQ, CE l’Orme des Merisiers, 91191 Gif sur Yvette Cedex, France,
2
Department
of Ecology, College of Urban and Environmental Science, Peking University, Beijing 100871, China,
3
Carnegie Institution Department of Global Ecology, 260
Panama Street, Stanford, California 94305, USA,
4
Global Carbon Project, CSIRO Marine and Atmospheric Research, Canberra, Australian Capital Territory
2601, Australia,
5
S J Mehta School of Management, Indian Institute of Technology, Powai, 400076 Mumbai, India,
6
College of Engineering, Mathematics
and Physical Sciences, University of Exeter, Exeter EX4 4QF, UK,
7
Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100085,
China,
8
Centre International de Recherche sur l’Environnement et le Développement, CNRS-CIRAD-ParisTech-EHESS 45 bis avenue de la Belle Gabrielle,
94736 Nogent Sur Marne, France.
†
These authors contributed equally to this work. *e-mail: philippe.ciais@lsce.ipsl.fr
anthropogenic activities that cause terrestrial sinks. Some sinks are
directly attributable to actions by individual countries, such as forest
management, land-use change and, to some extent, short-range
nitrogen deposition. Other sinks are attributable to the entirety
of greenhouse gas emissions through the effect of increased CO
2
concentrations and the resulting climate change.
Similarly, it is also anthropogenic activities such as forest
management, land-use change, nitrogen deposition, increased CO
2
and climate change that create sinks in terrestrial ecosystems
7
.
The role of CO
2
sinks provided by terrestrial ecosystems and the
ocean thus deserves more attention for determining the potential
for the success of CO
2
mitigation strategies. Terrestrial ecosystems,
mainly forests, and the oceans, remove on average each year 54% of
CO
2
emitted by deforestation and fossil fuel combustion
1
. Were it
not for these sinks, the concentration of atmospheric CO
2
would
increase more than twice as fast as observed. Here, we develop
two regionalized attribution approaches for natural sinks as services
that ameliorate the increase in atmospheric CO
2
. We attribute
the historical and future increase in atmospheric CO
2
to different
regions using two alternative approaches that both include sinks.
In one approach, each region is attributed the ‘domestic’ sink that
occurs within its territorial boundaries. In the second approach,
each region has a share of a global sink created indirectly through the
contribution of that region to the global rise in atmospheric CO
2
.
In our framework, this rise enhances vegetation uptake through
CO
2
fertilization and climate change, thereby enhancing the global
sink both in its territory and other regions. We apply these two
approaches to large regions to demonstrate the methodology and
to allow relatively accurate attribution with the available data. This
926 NATURE CLIMATE CHANGE | VOL 3 | OCTOBER 2013 | www.nature.com/natureclimatechange

NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE1942
ARTICLES
Table 1 | Attribution of 1CO
2
to regional fossil fuel and land-use emissions, and terrestrial sinks (column 2 and 3).
Contribution to
1CO
2
(ppm) in
2006 by: →
Fossil fuel +
land-use
emissions by this
region
Land sink within
the territory of
this region
Land sink
provided to
OECD by this
region
Land sink
provided to REF
by this region
Land sink
provided to ASO
by this region
Land sink
provided to ALM
by this region
OECD 97 −11 −6 −1 −2 −2
REF 30 −6 −3 −1 −1 −1
ASO 43 −11 −6 −1 −2 −2
ALM 76 −42 −23 −5 −8 −6
All land 246 −70 −38 −8 −13 −11
Ocean 0 −69 −37 −9 −13 −10
Globe 246 −139 −75 −17 −26 −21
Further breakdown of the ‘sink service’ provided by each absorbing region to others between 1850 and 2006 is given in columns 4 to 7, according to a choice of attribution based on emitters (see main
text). The attribution of the cumulative land sinks is expressed in ppm of 1CO
2
. A negative value is equivalent to a reduction of 1CO
2
. The 1CO
2
reduction ‘service’ provided by ocean uptake to each
region is shown as well. The corresponding 1CO
2
are given in Fig. 1a. All numbers are rounded to the ppm.
methodology could in principle be applied on the finer scale of
individual countries, but more reliable data on land biosphere
sources and sinks would need to be developed.
Here we extend the so-called ‘Brazilian proposal’ approach,
which first proposed to assign emission targets to nations on
the basis of their historical responsibility for the anthropogenic
greenhouse effect (http://www.match-info.net; refs 8–10). Because
of the importance of sinks in the global CO
2
budget, analyses
considering both emissions and sinks allow a more comprehensive
valuing of the contributing drivers and regions to the atmospheric
CO
2
imbalance. There is not a best scientific accounting framework
for sinks, as accounting choices necessarily follow value judgements
on which and how common resources might be shared. For
instance, should CO
2
fluxes into and out of unmanaged ecosystems
be considered a globally shared resource? Such choices represent
values. Values play a role in selecting accounting systems, as already
happens in international climate negotiations. However, once a
specific accounting system is decided, the ensuing attribution needs
to be scientifically based, consistent and with quantification of its
uncertainty. It is this factual attribution that is our primary focus.
Here, we show how the attribution of increased atmospheric
CO
2
since pre-industrial times differs between the historical period
(1850–2006) and the future (2010–2100), when using our two
distinct attribution approaches for sinks. Ocean sinks are attributed
in the same way in both accounting approaches.
Attributing the increase of CO
2
to regions of the globe
A simplified but comprehensive model of the carbon cycle (OSCAR;
ref. 11) is used that allows the increase of CO
2
above pre-industrial
levels to be attributed to emitting and absorbing regions (see
Methods). Fossil fuel and cement emissions are prescribed from
the Carbon Dioxide Information Analysis Center database
12
. The
model calculates the global ocean uptake and the regional the
land-use flux defined as the net CO
2
balance at any time t, of
terrestrial ecosystems affected by a land-use transition before t.
For undisturbed ecosystems, OSCAR has CO
2
sinks. The terrestrial
net primary productivity (NPP) is considered to scale with the
logarithm of atmospheric CO
2
content. The magnitude of this
CO
2
fertilization effect is calibrated within ecologically plausible
limits
13
to reproduce the observed trends in atmospheric CO
2
(see
Methods). In undisturbed ecosystems, the CO
2
fertilization sink
depends on the rate of increase in NPP and on the residence time of
carbon in biomass, litter and soil organic carbon pools. The excess
of NPP over heterotrophic respiration is increased or decreased by
processes such as temperature change, radiation quality changes
and nitrogen deposition. Some studies
14,15
have shown that for
high-emission scenarios, climate change reduces land sinks or can
even turn sinks into sources in some regions. These carbon–climate
feedbacks are not the focus here, but we tested a scenario where NPP
and respiration depend on temperature and found no qualitative
difference with the results in the absence of feedbacks.
We seek the share of each emitting region in the global excess of
atmospheric CO
2
above pre-industrial levels (1CO
2
). Thus, in the
OSCAR carbon cycle model, we decompose 1CO
2
into a sum of
individual positive (sources) or negative (sinks) contributions from
each emitting or absorbing region. This is achieved by numerically
‘tagging’ the modelled CO
2
molecules emitted from each region,
as if they were dye tracers. Carbon emitted each year by a given
region, be it fossil carbon (here fossil fuel burning and cement
emissions) or deforestation carbon, is subsequently spread through
the atmosphere–land–ocean carbon system. A fraction of each
region’s emissions stays in the atmosphere and adds to the increased
CO
2
burden. The rest is absorbed either by the ocean or by land in
the same, or another, region.
Attribution of the land sink follows two alternative approaches.
The first one attributes terrestrial sinks to the region where the sink
occurs. This idea of ‘domestic sinks’ is close to that of ‘full carbon
accounting’ in current international negotiations. The second one
estimates for each region the terrestrial sink that its historical
emissions have induced over all other regions, including itself.
This approach calculates the causal contribution of emitters to the
creation of sinks through CO
2
fertilization effects on increased
carbon storage (see Methods).
The ocean sink is also important, absorbing about 25% of fossil
and land-use emissions at present
16
. There are many ways in which
this ocean sink can be attributed. Here, we adopt the approach
that the ocean sink removes the share of excess CO
2
caused
by each emitting region in direct proportion to the total excess
CO
2
at the time of oceanic absorption (see Table 1), a physically
based concept in that physical processes at the ocean–atmosphere
interface are not discriminated as function of regionally tagged
atmospheric CO
2
molecules.
We note, however, that alternate approaches are possible. For
example, as the ocean is a shared resource among all people of the
world, it could be argued that its CO
2
uptake should be distributed
among regions on a per capita basis, or according to the nearest
ocean to a particular region. Under this assumption, people in
regions of low carbon emission would be attributed a service to
high-carbon-emission regions by allowing ‘their’ share of the ocean
to be used to absorb carbon from high-carbon-emission countries.
Here, for simplicity, we consider an attribution based on each
region’s share of excess atmospheric CO
2
.
We consider four land regions, plus the ocean. Each region
includes several countries that have similar projected economic
NATURE CLIMATE CHANGE | VOL 3 | OCTOBER 2013 | www.nature.com/natureclimatechange 927

ARTICLES
NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE1942
development pathways, hence similar fossil fuel and land-use CO
2
emissions trajectories. Projected fossil fuel emissions are from
the Intergovernmental Panel on Climate Change (IPCC) Special
Report on Emissions Scenarios A2 (SRES A2) scenario
17
. Future
land-use-change emissions are calculated by the OSCAR model,
given land-cover projections from the IMAGE 2.2 integrated
assessment model for the IPCC SRES A2 scenario
18
. The four
land regions are those used by the IPCC Fourth Assessment
Report (AR4), countries of the Organization for Economic Co-
operation and Development (OECD) in North America, western
Europe and Pacific developed, countries of Africa and Latin
America (ALM), all Asian countries and Oceania (ASO) and
the reformed economies of the former Soviet Union including
eastern Europe (REF).
Attribution of the historical increase of CO
2
Adopting the first attribution approach of only domestic sinks
for each emitter, Fig. 1a shows each region’s share of 1CO
2
between 1850 and the present day (2006), together with the separate
attributed contribution of fossil fuel and land-use emissions and
of domestic land sinks. The ocean sink, here not attributed,
is shown separately on the right. We estimate with OSCAR a
value of 1CO
2
equal to 107 ppm in 2006, close to the observed
value of 101 ± 5 ppm. Clearly, most of this historical increase is
attributed to OECD countries (+86 ppm) owing to their large
cumulated fossil fuel emissions, early century deforestation in
North America and their small land sinks. REF countries contribute
a small increase of 24 ppm to 1CO
2
and the ocean contributes
a decrease of −69 ppm. The ASO countries contribute +32 ppm
to 1CO
2
, about the same as ALM countries (+34 ppm) but for
different reasons. The contribution to 1CO
2
from ASO countries
is due to their cumulative fossil fuel emissions. In contrast, ALM
countries have small fossil emissions but high cumulative land-use
emissions (Fig. 1a). The attribution of 1CO
2
to ALM countries
is, however, attenuated by their high carbon sinks (tropical forest
biomes in OSCAR). The domestic carbon sink of ALM countries
is found to have offset 56% of their fossil fuel and land-use
emissions since 1850.
Using the second approach of attributing land sinks to the
emitters that caused them, Fig. 1c shows terrestrial sinks provided
by absorbing regions decrease 1CO
2
(negative bars in Fig. 1c) as
a service to emitters (see data in Table 1). With this attribution
approach, we estimate that since 1850 the ALM region has provided
an uptake of 23 ppm to the 1CO
2
of OECD, by far the largest ‘sink
service’ from an absorber to an emitter. An additional 9 ppm of
1CO
2
of OECD is removed by the combined terrestrial sinks of
REF and ASO. Not including the ocean in the attribution, the net
balance of OECD domestic sink (credits) compared with foreign
sinks (debits) is negative at −21 ppm (11 ppm minus 32 ppm). The
second region that is more indebted to others for terrestrial carbon
sink services is ASO, with other regions providing an uptake of
11 ppm, of which 8 ppm is provided by land sinks in ALM alone.
Over the historical period, the region that has provided the largest
net sink service to all others is ALM (36 ppm). Regions REF and
ASO are close to neutral with respect to the relative importance of
their sources and sinks.
Attribution of the future CO
2
increase
Given the SRES A2 fossil fuel emission scenario, the attribution
of the projected CO
2
increase in 2100 gives quite a different
picture compared with the historical period. Figure 1b shows a
modelled value of 1CO
2
in 2100 of 612 ppm with OSCAR, in
the range of three-dimensional (3D) carbon cycle model results
(450–740 ppm in ref. 14). ASO countries contribute 321 ppm to
1CO
2
in 2100, compared with 276 ppm from OECD countries,
233 ppm from ALM countries and 72 ppm from REF countries. The
107
¬69
59
90
7
¬38
22
26
30
23
20
¬13
64
16
60
¬12
2006
¬100
¬50
0
50
100
150
612
¬290
225
308
15
¬97
72
94
2
¬24
293
326
38
¬71
311
222
140
¬51
2100
¬400
¬200
0
200
400
600
107
¬69
86
90
7
¬11
24
26
4
¬6
32
23
20
¬11
34
16
60
¬42
ΔCO
2
(ppm)ΔCO
2
(ppm)ΔCO
2
(ppm)ΔCO
2
(ppm)
2006
¬100
¬50
0
50
100
150
612
¬290
276
308
15
¬47
72
94
2
¬24
321
326
38
¬43
233
222
140
¬129
2100
World OECD REF ASO ALM Ocean
¬400
¬200
0
200
400
600
Method 1
Method 2
4
¬8
World OECD REF ASO ALM Ocean
c
d
a
b
FF
LUC
Land
Ocean
FF
LUC
Ocean
Land sink:
Domestic
Foreign
Atmosphere
Atmosphere
Figure 1 | Attribution of the atmospheric CO
2
increase between 1850 and
2100, assuming that the carbon cycle was in equilibrium in 1850. a, First
attribution approach where regional terrestrial ‘domestic’ sinks are
attributed to each absorber region with the OSCAR carbon cycle model.
Hatched yellow bar denotes the modelled global 1CO
2
between 1850 and
2006, resulting from previous emissions and sinks during that period. This
modelled global 1CO
2
is 107 ppm, in line with the observed value of
101±5 ppm. Yellow bars, 1CO
2
values attributed to each region and their
apportionment to emissions and sinks further to the right of each bar; grey,
cumulative fossil fuel (FF) emissions; orange, cumulative land-use
emissions (LUC); green, terrestrial cumulative sinks, counted negatively
because they diminish 1CO
2
; blue, cumulative ocean uptake. b, Same but
for 1CO
2
simulated by OSCAR between 1850 and 2100, given the SRES A2
fossil fuel emission scenario and land-use emissions calculated by OSCAR
with prescribed land-cover-change forcing in each region (see main text).
c,d, Second attribution approach with regional sinks attributed to each
emitter region having caused both a domestic sink, in green, and foreign
sinks, in magenta. The ocean sink is shown separately on the right.
ASO countries are thus projected to have the largest contribution.
In the OSCAR model, NPP saturates at high CO
2
, therefore the land
sink per unit 1CO
2
weakens in the future, which tends to leave
more emissions airborne every year by 2100 under a high-emission
scenario such as SRES A2. For all absorbing regions together, the
overall sequestration of carbon on land is estimated to provide a
decrease in 1CO
2
of 243 ppm by year 2100 (Fig. 1b) and the overall
sequestration in the ocean a further decrease of 290 ppm. This
928 NATURE CLIMATE CHANGE | VOL 3 | OCTOBER 2013 | www.nature.com/natureclimatechange

NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE1942
ARTICLES
Year
FF
LUC
Atmosphere
Land
Ocean
Annual fluxes
FF
LUC
Atmosphere
Land
2010 2050 2100
1850 1900 1950 2000 2050 2100
0
10
¬10
20
¬20
30
¬30
OECD
REF
ASO
ALM
ΔCO
2
(Gt C yr
¬1
)
Figure 2 | Changes in global annual CO
2
sources and sinks between 1850
and 2100. Sinks in land and oceans, and the annual increase of CO
2
in the
atmosphere, are counted negatively and emissions are counted positively.
Annual fluxes are from fossil fuel emissions (grey), land-use change
(orange), terrestrial sinks (green), ocean sink (blue) and atmospheric
yearly increase (yellow). The four pie charts at the top show the attribution
of cumulated emissions, atmospheric increase and domestic sinks
occurring in each region for three time slices: 2006, 2050 and 2100.
mitigation of 1CO
2
is to be compared with the increase of 1CO
2
caused by emissions alone, in the absence of sinks, of 1,145 ppm.
In Fig. 2, the evolution of the global carbon budget illustrates the
fact that the fraction of fossil fuel and land-use emissions remaining
airborne increases progressively with time, reaching 64% in 2100
compared with 45% in 2006, owing to the saturation of land sinks
in the OSCAR model. For the three time periods of 2006, 2050
and 2100, the pie charts in Fig. 2 give the 1CO
2
attribution broken
down into cumulated fossil fuel emissions, land-use emissions and
land sink. For cumulated emissions, a shift of their attribution
occurs from OECD towards ASO countries between 2000 and
2100. Such a shift is already discernible in recent fossil fuel annual
emission data (Annex B countries having decreased from 64% of
global emissions in 1990 down to 44% in 2008; ref. 1). Land-use
emissions in the SRES A2 scenario, modelled with OSCAR driven by
the IMAGE 2.2 land-cover-change projections
18
and calculating net
CO
2
fluxes from land-cover transitions between forest, croplands
and grasslands, remain dominated by ALM countries throughout
the twenty-first century. These countries alone contribute 66%
of cumulated land-use emissions in 2006, but are projected to
contribute 72% in 2100 in the scenario studied here. Despite intense
land use, ALM countries remain the largest providers of land sink
service to emitting regions, including themselves. The share of ALM
into the land carbon storage change declines in the future, however,
whereas the share of OECD and REF regions increases. This is
because in IMAGE 2.2 land-cover-change scenario, deforestation
continues in ALM until late in the twenty-first century (resulting
in CO
2
emissions in OSCAR), whereas OECD and REF keep
stable areas of temperate and boreal forests
18
with long residence
times of carbon in soils, enabling carbon storage. This result
obviously depends both on the land-cover-change scenario and on
the terrestrial carbon cycle model used.
We also tested an attribution of idealized climate feedbacks, by
incorporating into OSCAR a linear response of NPP to temperature
change and a Q
10
response of heterotrophic respiration, with the
NPP response being calibrated to the results of a 3D Earth system
model (see Supplementary Information). An additional 1CO
2
of
38 ppm is calculated from the feedback. Once attributed, this term
does not qualitatively change the results shown in Fig. 1. In case
of a large positive feedback occurring in one region, however, the
contribution of climate change to the attribution of sinks to emitters
will be much more significant. In particular, a threshold-like
response to climate change could cause an abrupt drop in sinks,
or a large loss of CO
2
from natural reservoirs. In that case, one
could attribute to each emitter the 1CO
2
caused by passing such a
threshold, in proportion to the amount of climate change that this
emitter has created before the threshold is reached.
We show that it is possible to attribute the increased atmospheric
CO
2
burden to emitting regions, accounting for their sink function
and for the ocean, treated here as a global ‘sink service’. These
first results contribute towards comprehensive, consistent and
transparent frameworks for attributing the atmospheric CO
2
excess,
which could be further enriched by detailed analysis, for example,
using more realistic spatially detailed land and ocean carbon cycle
models. Our calculations (using a simplified carbon cycle model
and estimates of regional historical land carbon fluxes) are subject to
uncertainties. We do not claim that the results obtained here with a
simplified model are accurate enough to serve as a quantitative basis
for negotiating emission reductions or trading carbon sink services.
There are many other factors that could influence the results,
but these are unlikely to alter the sign or order of magnitude of
our findings. The nonlinear response of ecosystem carbon fluxes
and storage to climate is not accounted for in our approach. If
increased decomposition of soil carbon in response to warming or
if increased water stress on tropical forests
14,15
were incorporated in
our carbon cycle model, the role of land sinks in decreasing 1CO
2
by 2100 would be lessened. To account for these carbon–climate
feedbacks, our simple model could be calibrated to account for
the impacts of climate change on land sinks, from the results of
more complex Earth system sodels. Similarly, the ocean carbon
cycle sensitivity to climate would need to be accounted for in
future studies. Here, the only sources of nonlinearity in the
carbon cycle are the NPP saturation at high CO
2
, the land-use
effects on the carbon residence time in ecosystems and a weak
nonlinearity in ocean chemistry
19
. Another global adverse impact
of human activities that could be attributed to fossil fuel emitters
using our method is the acidification of the ocean, counted
proportional to CO
2
emitted from each region and absorbed by
the oceans. Beyond attribution of 1CO
2
to emissions, a higher
order attribution to emission socioeconomic drivers such as city
sprawl, energy production systems, manufacturing industries and
trade is possible, which would provide a more direct link between
policy action and atmospheric CO
2
. Finally, with a major upgrade
of carbon observing systems, using in situ networks
20
and satellite
observations
21
, regional carbon sinks could be measured with high
enough accuracy that the role of each region in the increase of CO
2
could be better quantified.
Ultimately, any discussion of attribution of 1CO
2
depends
on accurate estimates of emissions from human activities, an
appropriate representation of the carbon cycle and decisions about
how to treat carbon uptake by the land biosphere and the oceans.
Thus, attribution depends on both science and social values. Under
any reasonable set of assumptions, the largest share of 1CO
2
up
to the present can be attributed to the developed countries, but
the share attributed to the developing world is rapidly increasing.
For an intensive emission scenario such as SRES A2, developing
countries contribute the largest share of 1CO
2
by 2100 because
their emissions surge in the coming decades and sinks lag behind.
In this context, regionalized attribution of land and ocean sinks is
critical to provide options for policy, so that absorber regions can
benefit from incentives to maintain or enhance the ‘sink service’
they provide to themselves and to others.
NATURE CLIMATE CHANGE | VOL 3 | OCTOBER 2013 | www.nature.com/natureclimatechange 929

ARTICLES
NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE1942
Methods
Model description. The OSCAR model
11,22–24
used in this study includes:
prescribed emissions from fossil fuel combustion and cement factory based on
energy-use statistics since 1850 and economic projections of the SRES IPCC
scenarios up until 2100
17
; a calibrated ocean uptake module used for former
IPCC assessment reports
19
; a terrestrial ecosystems uptake module calculating the
imbalance between NPP and heterotrophic respiration, given residence times in
each reservoir
25
; and land-use flux forced by prescribed yearly land-cover-change
transitions
26
. The spatial resolution of the model is flexible and the configuration
used here is based on four large IPCC economic regions. Within each region, three
land-cover types and annual age classes (cohorts of vegetation affected by previous
land-use actions) of vegetation are considered. These land-cover-area changes
are prescribed from regional census
26
until the present and from the IMAGE 2.2
integrated assessment model scenario for the future
27
. For a specified land-use
transition, for example, when a tropical forest is cleared into cropland, the net
carbon balance of converted lands is calculated as the sum of losses incurring
from ‘old’ soil carbon decomposition and of ‘fresh’ gains of carbon by NPP and
subsequent litter delivery to the soil from the new vegetation. After such a land-use
change, ecosystem carbon pools can remain out of balance for several decades
28
.
One important feature is that the residence time of carbon, that is, the carbon
mass in a reservoir divided by the flow out of the reservoir, in vegetation is altered
in response to changing land use
11
. For example, after the conversion of tropical
forests to agriculture, the carbon residence time decreases greatly, which limits
future storage of CO
2
. The OSCAR-modelled CO
2
growth rate over the period
2000–2006 is 2.0 ppm yr
−1
, compared with 1.9 ppm yr
−1
in the observations
1
. The
modelled airborne fraction, defined as the ratio of growth rate to the sum of land
use and fossil emissions, is 0.45 in OSCAR against 0.46 analysed from actual CO
2
measurements
1
. There is a small bias in favour of the ocean in the proportion of
land versus ocean carbon global sinks, but the apportionment of land carbon fluxes
between northern and tropical biomes is in broad agreement with atmospheric
CO
2
gradients (see Fig. 7.7 in IPCC AR4 (ref. 7), reproduced with OSCAR estimates
in Supplementary Fig. S3). The northern hemispheric land sink in OSCAR lies
in the lower range of independent atmospheric inversion results
7
and the ocean
sink is 2.3 Pg C yr
−1
, comparable to IPCC AR4, but larger than inversion results,
possibly because inversions include natural outgassing of CO
2
by river-delivered
carbon to the ocean (0.3 Pg C yr
−1
) whereas OSCAR simulates the ocean sink
of anthropogenic CO
2
.
Attribution method. We modified the structure of the OSCAR model
(Supplementary Fig. S1) to calculate the fate of CO
2
emissions from each region,
which can either end up in the atmosphere or be absorbed by land ecosystems or the
ocean. OSCAR is divided into atmospheric, oceanic, plus managed and unmanaged
terrestrial reservoirs in each region. The ocean uptake of anthropogenic CO
2
is
calculated from an impulse response function model (HILDA model) and accounts
for the removal of anthropogenic CO
2
by physical mixing between the ocean
surface and deeper waters. For each region we numerically keep track of sinks in
unmanaged ecosystems and of land-use-induced sources or sinks for each emitter.
Structural nonlinearity in OSCAR implies that S
tot
6= 6sink(1C
i
) where S
tot
is
either the global land sink or the oceanic sink with all emitters, and S
i
= sink(1C
i
)
is the separate sink induced by the contribution to excess atmospheric CO
2
of the
ith emitter alone (1C
i
). We applied a linearization method to ensure that 100%
of S
tot
is attributed to each emitter, so that S
tot
= 6S
0
i
with S
0
i
= ρ
i
S
tot
being the
fully attributed sink induced by the ith emitter. The ρ
i
coefficients are first equal to
sink(C
tot
) − sink(C
tot
− 1C
i
) and then normalized so as to meet S
tot
= 6S
0
i
. This
corresponds to the so-called residual normalized attribution method described
by ref. 8. We tested other linearization methods discussed by ref. 8 and the
resulting ρ
i
coefficients were found to be similar to the second digit. Typically,
the residual fraction of 1CO
2
created by the difference S
tot
− 6sink(C
i
) due to
nonlinearity in both biospheric and oceanic sinks is 10 ppm in 2006 and goes
up to 155 ppm in 2100.
Received 8 February 2012; accepted 30 May 2013; published online
14 July 2013
References
1. Le Quéré, C. et al. The global carbon budget 1959–2012. Earth Syst. Sci. Data
Discuss. 5, 1107–1157 (2012).
2. Report of the Conference of the Parties on its thirteenth session, held in Bali from
3 to 15 December 2007 FCCC/CP/2007/6/Add.1 (UNFCCC, 2008); available at
http://unfccc.int/resource/docs/2007/cop13/eng/06a01.pdf#page=8.
3. Steffen, W. et al. The terrestrial carbon cycle: Implications for the Kyoto
protocol. Science 280, 1393–1394 (1998).
4. United Nations Framework Convention on Climate Change (United Nations,
1998); available at http://untreaty.un.org/cod/avl/ha/ccc/ccc.html.
5. Schulze, E-D., Valentini, R. & Sanz, M-J. The long way from Kyoto to
Marrakesh: Implications of the Kyoto Protocol negotiations for global ecology.
Glob. Change Biol. 8, 505–518 (2002).
6. Takahashi, T. et al. Climatological mean and decadal change in surface ocean
p
CO
2
, and net sea–air CO
2
flux over the global oceans. Deep-Sea Res. II 56,
554–577 (2009).
7. Denman, K. L. et al. in IPCC Climate Change 2007: The Physical Science Basis
(eds Solomon, S. et al.) 499–587 (Cambridge Univ. Press, 2007).
8. Trudinger, C. & Enting, I. Comparison of formalisms for attributing
responsibility for climate change: Non-linearities in the Brazilian Proposal
approach. Climatic Change 68, 67–99 (2005).
9. Den Elzen, M. et al. Analysing countries’ contribution to climate change:
Scientific and policy-related choices. Environ. Sci. Policy 8, 614–636 (2005).
10. Höhne, N. et al. Contributions of individual countries’ emissions to climate
change and their uncertainty. Climatic Change 106, 359–391 (2011).
11. Gitz, V. & Ciais, P. Amplifying effects of land-use change on future atmospheric
CO
2
levels. Glob. Biogeochem. Cycles 17, 359–391 (2003).
12. Boden, T. A., Marland, G. & Andres, R. J. Global, Regional, and National
Fossil-Fuel CO
2
Emissions (US Department of Energy, 2010).
13. Norby, R. J. et al. Forest response to elevated CO
2
is conserved across a broad
range of productivity. Proc. Natl Acad. Sci. USA 102, 18052–18056 (2005).
14. Friedlingstein, P. et al. Climate-carbon cycle feedback analysis; results from the
C4MIP model intercomparison. J. Clim. 19, 3337–3353 (2006).
15. Cox, P. M., Betts, R. A., Jones, C. D., Spall, S. A. & Totterdell, I. J. Acceleration
of global warming due to carbon-cycle feedbacks in a coupled climate model.
Nature 408, 184–187 (2000).
16. Le Quere, C. et al. Trends in the sources and sinks of carbon dioxide.
Nature Geosci. 2, 831–836 (2009).
17. IPCC, Special Report on Emissions Scenarios (eds Nakicenovic, N. & Swart, R.)
(Cambridge Univ. Press, 2000).
18. IMAGE team The IMAGE 2.2 Implementation of the SRES Scenarios; A
Comprehensive Analysis of Emissions, Climate Change and Impacts in the 21st
Century (RIVM, Bilthoven, 2001).
19. Joos, F. et al. An efficient and accurate representation of complex oceanic
and biospheric models of anthropogenic carbon uptake. Tellus 48(B),
397–417 (1996).
20. Ciais, P. & Moore, B. Integrated Global Carbon Observation Theme: A Strategy
to Realize a Coordinated System of Integrated Global Carbon Cycle
Observations (2001).
21. Crisp, D. et al. The Orbiting Carbon Observatory (OCO) mission. Adv. Space
Res. 34, 700–709 (2004).
22. Gitz, V. & Ciais, P. Future expansion of agriculture and pasture acts to
amplify atmospheric CO
2
levels in response to fossil-fuel and land-use change
emissions. Climatic Change 67, 161–184 (2004).
23. Gitz, V. Usage Des Terres et Politiques Climatiques Globales PhD thesis (in
French), Presses Académiques Francophones (2003).
24. Gitz, V., Hourcade, J. C. & Ciais, P. The timing of biological carbon
sequestration and carbon abatement in the energy sector under optimal
strategies against climate risks. Energy J. 27, 113–133 (2006).
25. Friedlingstein, P. et al. On the contribution of CO
2
fertilization to the missing
biospheric sink. Glob. Biogeochem. Cycles 9, 541–556 (1995).
26. Houghton, R. A. Revised estimates of the annual net flux of carbon to the
atmosphere from changes in land use and land management 1850–2000.
Tellus B 378–390 (2003).
27. Alcamo, J. et al. in Global Change Scenarios of the 21st Century, Results from
the IMAGE 2.1 Model (eds Alcamo, J., Leemans, R. & Kreileman, E.) 97–139
(Elsevier, 1998).
28. Houghton, R. A. Emissions of Carbon from Land-use Change (Cambridge Univ.
Press, 1997).
Acknowledgements
This paper is a contribution to the efforts of the Global Carbon Project, a joint project of
the IGBP, WCRP, IHDP and Diversitas, to track and analyse the interactions among the
carbon cycle, human activities and the climate system.
Author contributions
P.C. designed the study and wrote the text. T.G. prepared the model set-up, conducted
the simulations and contributed to the text. J.D.P. contributed to the model set-up and to
the text, and made the key figures. K.C., M.R.R., J.G.C., A.P., P.F. and S.L.P. contributed
to the interpretation of the results and to the text. V.G. developed the original OSCAR
model and contributed to the text.
Additional information
Supplementary information is available in the online version of the paper. Reprints and
permissions information is available online at www.nature.com/reprints. Correspondence
and requests for materials should be addressed to P.C.
Competing financial interests
The authors declare no competing financial interests.
930 NATURE CLIMATE CHANGE | VOL 3 | OCTOBER 2013 | www.nature.com/natureclimatechange