Content uploaded by Valier Galy
Author content
All content in this area was uploaded by Valier Galy on Oct 21, 2015
Content may be subject to copyright.
LETTER doi:10.1038/nature14400
Global carbon export from the terrestrial biosphere
controlled by erosion
Valier Galy
1
, Bernhard Peucker-Ehrenbrink
1
& Timothy Eglinton
1,2
Riverine export of particulate organic carbon (POC) to the ocean
affects the atmospheric carbon inventory over a broad range of
timescales
1–5
. On geological timescales, the balance between
sequestration of POC from the terrestrial biosphere and oxidation
of rock-derived (petrogenic) organic carbon sets the magnitude of
the atmospheric carbon and oxygen reservoirs
6,7
. Over shorter
timescales, variations in the rate of exchange between carbon reser-
voirs, such as soils and marine sediments, also modulate atmo-
spheric carbon dioxide levels
1
. The respective fluxes of biospheric
and petrogenic organic carbon are poorly constrained, however,
and mechanisms controlling POC export have remained elusive,
limiting our ability to predict POC fluxes quantitatively as a result
of climatic or tectonic changes. Here we estimate biospheric and
petrogenic POC fluxes for a suite of river systems representative of
the natural variability in catchment properties. We show that
export yields of both biospheric and petrogenic POC are positively
related to the yield of suspended sediment, revealing that POC
export is mostly controlled by physical erosion. Using a global
compilation of gauged suspended sediment flux, we derive separate
estimates of global biospheric and petrogenic POC fluxes of
157z74
{50 and 43z61
{25 megatonnes of carbon per year, respectively.
We find that biospheric POC export is primarily controlled by
the capacity of rivers to mobilize and transport POC, and is largely
insensitive to the magnitude of terrestrial primary production.
Globally, physical erosion rates affect the rate of biospheric POC
burial in marine sediments more strongly than carbon sequest-
ration through silicate weathering. We conclude that burial of bio-
spheric POC in marine sediments becomes the dominant long-
term atmospheric carbon dioxide sink under enhanced physical
erosion.
The atmosphere is a small reservoir of carbon in comparison with
rocks, soils, the biosphere and the ocean
1
. Its size is therefore sensitive
to small imbalances in the exchange of C with and between these larger
reservoirs. Over long timescales, the continental biosphere is mostly at
equilibrium with the atmosphere, because most C fixed by terrestrial
photosynthesis is quickly returned to the atmospheric reservoir
through respiration
1
. However, rivers deliver to the oceans a fraction
of this net primary production (NPP) as POC and dissolved organic
carbon (DOC)
2–5
. Although most DOC is quickly returned to the
atmosphere through oxidation in estuaries and the ocean, a significant
fraction of riverine POC is buried in marine sediments and stored over
long timescales. This ‘leakage’ of carbon from the biosphere–atmo-
sphere loop represents a net sequestration of atmospheric C (ref. 6).
Rivers also transfer POC from the rock reservoir (petrogenic organic
carbon, OC) to marine sediments, thereby transferring C between two
reservoirs disconnected from the atmosphere
7
. During this transfer,
oxidation of petrogenic OC represents another leakage of C, in this
case towards the atmosphere
8
. The nature and efficiency of riverine
export of POC to the ocean thus fundamentally affect the long-term
atmospheric C inventory. Despite its importance, global riverine
export of POC to the ocean has until now remained poorly con-
strained
2–5
. In particular, the respective global fluxes of biospheric
and petrogenic POC remain largely unconstrained. More importantly,
the sensitivities and relative magnitudes of global biospheric and pet-
rogenic POC export are not well defined, impeding our ability to
quantitatively predict POC fluxes and their impact on the long-term
global C cycle under different forcing scenarios.
Developing accurate constraints on fluvial transfer of biospheric
POC requires the quantification of, and correction for, petrogenic
OC in river sediment. Although the presence of petrogenic OC in river
sediments and river-dominated margin sediments has been inferred
for decades, its direct and unambiguous detection is quite recent
7,9,10
.
Consequently, few quantitative reconstructions of petrogenic OC
fluxes exist. However, those reconstructions encompass such diverse
river systems as the Amazon, Taiwanese rivers and the Ganges–
Brahmaputra system
7,8,11
. Radiocarbon (
14
C) measurements have pro-
vided key constraints on petrogenic OC concentrations and fluxes.
Exploiting the absence of
14
C in petrogenic OC and its presence in
biospheric OC,
14
C measurements on riverine POC combined with
additional constraints on the chemical composition of either constitu-
ent allow these two key constituents to be differentiated
7,12–18
.Herewe
use published POC compositional data (including
14
C measurements)
to derive a direct, global estimate of the petrogenic OC flux (Methods).
These new results, together with published petrogenic OC fluxes, pro-
vide a unique compilation of riverine export of petrogenic and bio-
spheric POC to the ocean from 43 river systems that account for 20%
of the sediment discharge to the oceans.
Petrogenic OC is an integral component of sedimentary and other
rocks. Its export by river systems is therefore tightly linked to that of
sediments. Indeed, both Komada et al.
17
and Hilton et al.
16
have shown
that yields of petrogenic POC (the petrogenic POC flux normalized to
catchment area) are positively correlated with corresponding yields of
suspended sediment in the Santa Clara River (California) and
Taiwanese rivers, respectively, with the latter property serving as a
measure of spatially averaged physical erosion rate. Our data set
extends this observation to a broad range of river systems, covering
more than four orders of magnitude in both catchment size and sedi-
ment yield (Supplementary Table 1). These data broadly follow a
power-law relationship characterized by a power exponent close to
unity (1.11 60.13 ) (errors are 1 s.d.; see Methods). Petrogenic POC
yield thus varies roughly linearly with sediment yield, implying that the
behaviour of petrogenic OC during erosion and transport is similar to
that of the other mineral phases (Fig. 1). In particular, this finding
implies a generally uniform depth-distribution of petrogenic OC in
soils and rocks. The average petrogenic POC concentration in river
sediments, however, varies considerably (0.02% to 0.6%). This can be
explained by variations in the average petrogenic OC content of rocks
and/or by variable oxidation of petrogenic OC during sediment trans-
fer to the ocean. Recent studies of large river basins with extensive
floodplains (Ganges–Brahmaputra and Amazon) indicate that up to
50% of petrogenic POC initially present in rocks can be oxidized dur-
ing sediment transport and temporary storage in intermediate reser-
1
Woods Hole Oceanographic Institution, Department of Marine Chemistry and Geochemistry, 360 WoodsHole Road, Woods Hole, Massachusetts02543, USA.
2
Geological Institute, Department of Earth
Sciences, Sonneggstrasse 5, Eidgeno
¨ssische Technische Hochschule, 8092 Zu
¨rich, Switzerland.
G2015 Macmillan Publishers Limited. All rights reserved
204 | NATURE | VOL 521 | 14 MAY 2015
voirs
7,8
. In contrast, petrogenic OC is very efficiently preserved in
fluvial systems characterized by rapid sediment transfer to the ocean,
such as Taiwanese rivers
19
and the Eel River (California)
20
. Enhanced
oxidation of petrogenic OC could explain up to about 50% of the
observed order-of-magnitude difference in petrogenic OC concentra-
tion. In addition, average petrogenic OC concentrations in rocks have
been reported to vary from catchment to catchment by at least an order
of magnitude
10,21
. Together, these observations suggest that whereas
initial contents in rocks and subsequent oxidation of petrogenic OC
during sediment transport together are the dominant controls of pet-
rogenic OC concentration in river sediments, sediment yield—that is,
physical erosion rate—is the primary control on petrogenic OC export
efficiency.
Unlike petrogenic OC, biospheric POC is not an indigenous mineral
component of the sediment; instead, it is added during vegetation
growth, soil formation and processes associated with the movement
of materials from source to sink (for example landsliding or overland
flow). The controls on its behaviour could therefore differ from those
affecting the mineral load. Because it is the other component of
riverine POC, biospheric OC fluxes can be obtained by subtracting
petrogenic OC contributions from riverine POC fluxes. Here we
use a compilation of riverine POC fluxes from 70 river systems
(Supplementary Tables 1 and 2), covering 42.7 310
6
km
2
(that is,
40% of the total exorheic continental area) and accounting for about
45% of the global freshwater discharge to the ocean. For 27 of these
river systems we lack direct estimates of petrogenic OC contributions.
We therefore use the relationship between sediment yield and petro-
genic OC yield (Fig. 1) to estimate petrogenic contributions to POC in
these systems (Methods). This permits the calculation of biospheric
POC fluxes and yields for the entire set of 70 river systems. Our data
show that biospheric POC yield is positively correlated with suspended
sediment yield, following a unique power-law relationship (r
2
50.78)
(Fig. 2). Although relationships between riverine POC and suspended
sediment concentrations have been reported previously
2
, we extend
this observation to biospheric POC, accounting for its dilution by
petrogenic OC at high sediment yield. The singularity of the relation-
ship between biospheric POC and sediment yield is remarkable, con-
sidering the very broad range of climate, vegetation, geomorphology
and anthropogenic disturbance characterizing the drainage basins
considered. It implies that, globally, the rate of biospheric POC export
is primarily controlled by sediment export processes. The exponent of
the power relationship between biospheric POC and sediment yield is
significantly smaller than 1 (0.56 60.03), reflecting an increasing dilu-
tion of biospheric POC by mineral phases (that is, decreasing bio-
spheric POC concentrations) at high sediment yield. This reflects
the well-documented decrease in OC concentration with depth in soil
profiles, which results in an increase in POC stock with depth that
globally follows a power law characterized by an exponent of 0.4 (ref.
22). In general, at low sediment yield, erosion proceeds mostly by
means of overland flow, exporting surface material (such as soil litter
and O horizons) characterized by high biospheric POC concentra-
tions. Conversely, at high sediment yield, erosion proceeds by means
of deep gully erosion and/or landslides, thereby lowering the overall
biospheric POC concentration of the eroded material
23
.
To further evaluate to what extent productivity and associated soil
OC content control biospheric POC export, we used the MOD17
database
24
to extract basin-scale estimates of NPP for 40 systems.
Biospheric OC yield and NPP are weakly positively correlated (power
law with exponent of 1.16 and r
2
50.30; Extended Data Fig. 1), sug-
gesting that productivity does not impose a strong control on bio-
spheric OC yield. However, the calculated fractions of the NPP
exported from catchments by rivers vary by more than three orders
of magnitude (0.01% to 2.1%) and are positively correlated with the
suspended sediment yield, following a power-law relationship (Fig. 3)
that is characterized by an exponent (0.50 60.05) statistically identical
to that of the relationship defined by biospheric POC yield. We there-
fore conclude that the rate of biospheric POC export depends prim-
arily on the capacity of rivers to mobilize and transport POC out of
catchments, rather than on POC production within the watershed.
Our data set also reveals significant secondary variations (Figs 2 and 3),
suggesting the existence of additional control mechanisms. Among
the possible mechanisms, we postulate that, for a given sediment yield,
higher frequency and/or deeper landslides result in decreased bio-
spheric POC yield, illustrating the critical role of physical erosion
processes in biospheric POC export. Other possible mechanisms
include the sorption of DOC onto mineral phases in sediment-starved
rivers (for example the Congo river) that are often characterized by
high DOC concentrations. However, the lack of correlation between
DOC concentration and biospheric POC yield suggests that sorption
of DOC does not exert a significant control on the efficiency of bio-
spheric POC export. The increasing relative variance of biospheric
POC yield at a low concentration of suspended sediment—conditions
that promote aquatic primary production—suggests that within-river
biological productivity could explain part of the secondary variability
of biospheric POC yield. Transient non-steady-state erosion of soil
and biosphere reservoirs can also introduce some variability in bio-
spheric POC yield that might not be accounted for by either sediment
yield or NPP. Finally, human activities such as damming and agricul-
ture currently affect virtually all modern river systems and probably
influence biospheric POC yields.
Biospheric OC yield (t km
–2
yr
–1
)
Suspended sediment yield (t km–2 yr–1)
0.01
0.1
1
10
100
1
1,000
10010 100,000 10,0001,000
Figure 2
|
Relationship betweenbiospheric POC yield (
Y
bios
) and suspended
sediment yield. Data obtained by subtracting measured petrogenic OC fluxes
from riverine POC fluxes (black dots) and those obtained using petrogenic OC
fluxes inferred from the relationship shown in Fig. 1 (grey dots) plot on the
same trend. The regression line is Y
bios
50.081Y0:56
sed ;r
2
50.78; P,0.001.
0.01
0.1
1
10
100
100
Petrogenic C yield (t km
–2
yr
–1
)
Suspended sediment
y
ield (t km–2
y
r–1)
1,000
0.001 10 100,000 10,0001,000
Figure 1
|
Relationship betweenpetrogenic OC yield (
Y
petro
) and suspended
sediment yield (
Y
sed
). Catchments larger (black diamonds) or smaller (grey
dots) than 100,000 km
2
plot on the same trend. Most of the variability in
petrogenic OC concentration derives from variableinitial OC concentrations in
rocks and petrogenic OC oxidation during sediment transport. The regression
line is Y
petro
50.0007Y1:11
sed ;r
2
50.82; P,0.001.
LETTER RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved
14 MAY 2015 | VOL 521 | NATURE | 205
The strong relationships between suspended sediment yield and
POC yield (both petrogenic and biospheric) allow global POC fluxes
to be inferred from the better-constrained global sediment flux. Here
we use recent estimates of the suspended sediment flux to the ocean
25–27
to estimate global biospheric and petrogenic POC fluxes. We use a
global suspended sediment flux of 19,000 6500 Mt yr
21
and a corres-
ponding suspended sediment yield of 17668tkm
22
yr
21
(Methods).
In turn, using the overall relationship between suspended sediment
yield and biospheric POC yield (Fig. 2), we estimate a global bio-
spheric POC yield of 1:46z0:68
{0:47 tCkm
22
yr
21
, which translates into a
global biospheric POC flux of 157z74
{50 Mt C yr
21
. A similar approach
applied to the relationship between suspended sediment yield and the
fraction of the NPP exported by rivers (Fig. 3) gives a global value of
0.18% of terrestrial NPP being exported to the ocean. Combined with
a mean estimate of the global terrestrial NPP of 77.6 Gt Cyr
21
(ref.
24), this gives a global biospheric POC flux of 140z96
{57 Mt C yr
21
that
is statistically identical to our estimate based on biospheric POC
yield data. Annually, about 0.02% of the total mass of C present in
the atmosphere is thus transferred to the ocean as POC, showing
that biospheric POC sequestration can affect the size of the atmo-
spheric reservoir over timescales as short as 10
3
–10
4
years.
The dependence of petrogenic OC yield on OC concentrations
in rocks complicates the estimation of the global petrogenic POC
flux from our estimate of global suspended sediment yield. Ideally,
the distribution of rocks characterized by variable OC concentra-
tions as well as intrinsic geomorphic characteristics (such as the
size of the floodplain and the spatial distribution of physical ero-
sion) need to be taken into account, because they both exert strong
control on petrogenic OC yields. In the absence of such a model,
we can only assume that the 43 river systems that we characterized
are representative of the natural variability (that is, rock types and
catchment morphology). Using our global suspended sediment
yield of 176 68tkm
22
yr
21
, we estimate a global petrogenic
POC flux of 43z61
{25 Mt C yr
21
(Methods).
Finally, we derive a combined global flux of terrestrial POC to the
ocean of 200z135
{75 Mt C yr
21
, of which about 80% and 20% are bio-
spheric and petrogenic POC, respectively. This direct estimate of these
two fluxes provides an assessment of the magnitude of POC transfer
from the terrestrial biosphere to the ocean, and reveals the global
significance of petrogenic OC as a component of POC export by rivers
to the ocean. However, these fluxes do not take bedload transport into
account and must therefore represent a lower bound of actual petro-
genic and biospheric POC fluxes to the ocean. Indeed, bedload mater-
ial can be dominated either by petrogenic OC
7,8
or biospheric POC
28
,
implying that bedload transport contributes globally to the fluvial
export of both petrogenic and biospheric POC.
On geological timescales, petrogenic and biospheric OC have
opposing roles in the global C cycle: the net transfer of C between
atmospheric and terrestrial reservoirs is set by the balance between
petrogenic OC oxidationand biospheric POC burial. We show that the
rate of both petrogenic and biospheric POC export from the conti-
nents is controlled primarily by the rate of sediment export; that is,
physical erosion. The preservation of both petrogenic and biospheric
POC in the ocean is up to three times higher at high physical erosion
rates
10,16,19
. Thus, increased physical erosion rates favour efficient
transfer and burial of biospheric POC coupled with enhanced preser-
vation (that is, decreased oxidation) of petrogenic POC, both acting to
limit the return of carbon to the atmospheric reservoir. The small
fraction of NPP exported even at very high erosion rates (a few per
cent; Fig. 3) suggests that enhanced export of biospheric POC is sus-
tainable over long timescales, as the terrestrial biospheric OC reservoir
is continuously replenished by photosynthetic C fixation. Using avail-
able estimates of terrestrial POC burial efficiency, we show that bio-
spheric POC burial yield is positively correlated with sediment yield
(Extended Data Fig. 2). Globally, the rate of C sequestration through
silicate weathering has a weaker sensitivity to sediment yield
(Extended Data Fig. 2) as a result of kinetic limitation of weathering
reactions at high physical erosion rates
29
. Biospheric POC burial is thus
predicted to become the dominant long-term atmospheric CO
2
sink
under a fourfold increase in global physical erosion rate at constant
temperature. We conclude that tectonic and climatic forcing of phys-
ical erosion favours biospheric POC sequestration over silicate weath-
ering.
Online Content Methods, along with any additional Extended Data display items
and SourceData, are available in theonline version of the paper;references unique
to these sections appear only in the online paper.
Received 2 September 2014; accepted 9 March 2015.
1. Sarmiento,J. & Gruber,N. in Ocean Biogeochemical Dynamics(eds Sarmiento, J. &
Gruber, N.) 392–453 (Princeton Univ. Press, 2006).
2. Ludwig,W., Probst, J.-L. & Kempe,S. Predicting the oceanicinput of organic carbon
by continental erosion. Glob. Biogeochem. Cycles 10, 23–41 (1996).
3. Schlu
¨nz, B. & Schneider,R. R.Transport of terrestrial organiccarbon to the oceans
by rivers: re-estimating flux and burial rates. Int. J. Earth Sci. 88, 599–606 (2000).
4. Meybeck, M. in Interactions of C, N, P and S: Biogeochemical Cycles and Global
Change (eds Wollast, R., Mackenzie, F. T. & Chou, L.) 163–193 (Springer, 1993).
5. Degens, E.T., Kempe, S. & Richey, J. E.Biogeochemistry of Major WorldRivers (Wiley,
1991).
6. Berner, R. A. Burial of organic carbon and pyrite sulfur in the modern ocean: its
geochemical and environmental significance. Am. J. Sci. 282, 451–473 (1982).
7. Galy, V., Beyssac, O., France-Lanord, C. & Eglinton,T. I. Recycling of graphite during
Himalayan erosion: a geological stabilization of carbon in the crust. Science 322,
943–945 (2008).
8. Bouchez, J. et al. Oxidation of petrogenic organic carbon in the Amazon floodplain
as a source of atmospheric CO
2
.Geology 38, 255–258 (2010).
9. Blair, N. E., Leithold, E. L. & Aller, R. C. From bedrock to burial: the evolution of
particulate organic carbon across coupled watershed–continental margin
systems. Mar. Chem. 92, 141–156 (2004).
10. Galy, V. et al. Efficient organic carbon burial in the Bengal fan sustained by the
Himalayan erosional system. Nature 450, 407–410 (2007).
11. Hilton, R. G. et al. Climatic and geomorphic controls on the erosion of terrestrial
biomass from subtropical mountain forest. Glob. Biogeochem. Cycles 26, GB3014
(2012).
12. Bouchez, J. et al. Source, transport and fluxesof Amazon River particulate organic
carbon: insights from river sediment depth-profiles. Geochim. Cosmochim. Acta
133, 280–298 (2014).
13. Drenzek,N. et al. A newlook at old carbon in active margin sediments. Geology 37,
239–242 (2009).
14. Galy, V. & Eglinton, T. I. Protracted storage of biospheric carbon in the Ganges–
Brahmaputra basin. Nature Geosci. 4, 843–847 (2011).
15. Hilton, R. G. et al. Tropical-cyclone-driven erosion of the terrestrial biosphere from
mountains. Nature Geosci. 1, 759–762 (2008).
16. Hilton, R. G., Galy, A., Hovius, N. & Horng, M. J. Efficient transport of fossil organic
carbon to the ocean by steep mountain rivers: an orogenic carbon sequestration
mechanism. Geology 39, 71–74 (2011).
17. Komada,T., Druffel, E. R. M. & Trumbore,S. E. Ocanic export of relictorganic carbon
by small mountainous rivers. Geophys. Res. Lett. 31, 1–4 (2004).
0.01
0.10
1
10
NPP exported (%)
Suspended sediment yield (t km–2 yr–1)
0.001
1 10010 100,000 10,0001,000
Figure 3
|
Relationship between the proportion of NPP exported annually
(NPP
exp
) and suspended sediment yield. Normalization of biospheric POC
export to NPP does not remove its dependence on suspended sediment yield,
illustrating the overarching control exerted by physical erosion on biospheric
POC export. The regression line is NPP
exp
50.013Y0:50
sed ;r
2
50.71; P,0.001.
RESEARCH LETTER
G2015 Macmillan Publishers Limited. All rights reserved
206 | NATURE | VOL 521 | 14 MAY 2015
18. Leithold, E. L., Blair, N. E. & Perkey, D. W. Geomorphologic controls on the age of
particulate organic carbon from small mountainous and upland rivers. Glob.
Biogeochem. Cycles 20, GB3022 (2006).
19. Kao, S.-J. et al. Preservation of terrestrial organic carbon in marine sediments
offshore Taiwan: mountain building and atmospheric carbon dioxide
sequestration. Earth Surf. Dyn. 2, 127–139 (2014).
20. Blair, N. E. et al. The persistence of memory: the fate of ancient sedimentary
organic carbon in a modern sedimentary system. Geochim. Cosmochim. Acta 67,
63–73 (2003).
21. Hilton,R. G., Galy, A., Hovius, N., Horng,M. J. & Chen, H. E. The isotopiccomposition
of particulate organic carbon in mountain rivers of Taiwan. Geochim. Cosmochim.
Acta 74, 3164–3181 (2010).
22. Jobbagy, E. G. & Jackson, R. B. The vertical distribution of soil organic carbon and
its relation to climate and vegetation. Ecol. Appl. 10, 423–436 (2000).
23. Hilton,R. G., Meunier, P., Hovius,N., Bellingham,P. J. & Galy, A. Landslide impacton
organic carbon cycling in a temperate montane forest. Earth Surf. Process. Landf.
36, 1670–1679 (2011).
24. Zhao, M., Nemani, Z. & Running, S. (ed. NASA).
25. Milliman, J. D. & Farnsworth, K. River Discharge to the Coastal Ocean: a Global
Synthesis (Cambridge Univ. Press, 2011).
26. Peucker-Ehrenbrink, B. Land2Sea database of river drainage basin sizes, annual
water discharges, and suspended sediment fluxes. Geochem. Geophys. Geosyst.
10, Q06014 (2009).
27. Larsen,I. J., Montgomery,D. R. & Greenberg, H. M. Thecontribution of mountains to
global denudation. Geology 42, 527–530 (2014).
28. Bianchi, T. S., Galler, J. J. & Allison, M. A. Hydrodynamic sorting and transport of
terrestrially derived organic carbon in sediments of the Mississippi and
Atchafalaya Rivers. Estuar. Coast. Shelf Sci. 73, 211–222 (2007).
29. West, A. J., Galy, A. & Bickle, M. Tectonic and climatic controls on silicate
weathering. Earth Planet. Sci. Lett. 235, 211–228 (2005).
Supplementary Information is available in the online version of the paper.
Acknowledgements We thank Y. Godderis, J. Hemingway and G. Soulet for comments
on early versions of the manuscript. G. Fiske generated the NPP data. Supportfor this
project was provided by US National Science Foundation (NSF) grant OCE-0851015
(to B.P.-E., T.E. and V.G.),NSF grant OCE-0928582 (to V.G. and T.E.) and SwissNational
Science Foundation grant 200021_140850 (to T.E.).
Author Contributions V.G. designed the study, performed the analysisand drafted the
manuscript with inputs from B.P.-E. and T.E.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of the paper. Correspondence
and requests for materials should be addressed to V.G. (vgaly@whoi.edu).
LETTER RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved
14 MAY 2015 | VOL 521 | NATURE | 207
METHODS
Quantitative apportionment of biospheric and petrogenic POC in river sedi-
ments. Several methods have recently been used to apportion petrogenic and bio-
spheric POC quantitatively in river sediments. All of these methods are based on
the unique property of petrogenic OC that it does not contain any
14
C, whereas
biospheric POC does. However, the
14
C content of biospheric POC is difficult to
predict a priori, because itdepends on a complex arrayof processes such as physical
erosion, soil formation and the dynamics ofthe biosphere itself. Therefore, in most
cases, additional constraints on the chemical composition of biospheric and/or
petrogenic OC are needed. Galy et al.
12
proposed that the different hydrodynamic
propertiesof petrogenic and biospheric POC in largestreams and rivers enables the
use of bulk radiocarbon measurements on POC in suspended and bed sediments
collected along depth profiles to estimate both petrogenic OC concentration and
biospheric POC
14
C content. The underlying assumption is that petrogenic OC is
uniformlydistributed in the water column (owing to its physicalproperties and size
distribution), whereas the relative concentration of biospheric POC—which is
preferentially associated with fine-grained sediments—decreases predictably with
depth. This approach has been tested and yields robust results for the Ganges–
Brahmaputra
7,14
,Amazon
8,12
and Fraser
30
rivers. Here we reanalyse published
14
C
data and show that the approach developed by Galy et al.
7
for suspended sediment
depth profiles can also be applied to size-fractionated and density-fractionated
fluvial sediments as well as to suspended sediments sampled across a wide range
of flow conditions. The
14
C content of density-fractionated sedimentsderived from
the Mississippi river
31
and of time-series suspendedsediments of the Ishikari river
32
provide excellent examples (Extended Data Figs 3 and 4). This type of binary
mixing approach is, however, not always appropriate, eitherbecause adequate data
may not be available or because of non-systematic behaviour of biospheric POC in
the water column. Indeed, biospheric POC often reflects the mixing of several
components such as soil and fresh plant debris, which may associate with the
mineral load in different ways and have different hydrodynamic properties. Here
the Mackenzie river provides a good example. Recently fixed plant-derived bio-
spheric OC and very old permafrost-derived biospheric OC are preferentially
associated with coarse and fine sediments, respectively, undermining the binary
mixing approach
33
. In these circumstances, additional constraints on the composi-
tion of the petrogenic and biospheric end-members are needed. These can be
obtained by using bulk (for example N/C, d
13
Cord
15
N) or compound-specific
(
14
C) data, as demonstrated for Taiwanese rivers
21
, the Eel River
13
and the
Mackenzie River
34
. Once a priori compositions of the petrogenic and different
biospheric end-members have been established, simple mixing models can be used
to apportion petrogenic and biospheric POC quantitatively. We used a combina-
tion of the above-mentioned techniques toevaluate petrogenic OCconcentrations
in river sediments by using published and newly acquired POC characterizations.
Supplementary Table 1 summarizes the sources of the data, the methods employed
and the results obtained.
In some cases, POC fluxes have been measured but adequate data to quantify
petrogenic and biospheric POC are not available. In these cases (27 rivers;
Supplementary Table 2) we estimate the petrogenic OC yield by using the rela-
tionship between suspendedsediment yield and petrogenic OC yield defined by all
rivers for which we could quantify petrogenic and biospheric OC (43 rivers; Fig. 1
and Supplementary Table 1). The significant scatter around this relationship
introduces uncertainty in the estimation of petrogenic OC yield from suspended
sediment yield. We therefore estimated the uncertainty of calculated petrogenic
OC yield on the basis of the uncertainty of the relationship between suspended
sediment yield and petrogenic OC yield. Biospheric POC yield is obtained by
subtracting inferred petrogenic OC yield from riverine POC yield. The relation-
ship between biospheric POC yield and suspended sediment yield is identical
when the two groups of data—that is, data determined from geochemical char-
acterization versus data inferred from the relationship between suspended sedi-
ment yield and petrogenic OC yield—are considered separately (Fig. 2).
Specifically, the exponents and multiplying terms of the two relationships are
statistically identical (within 61 s.d.): 0.51 60.04 and 0.59 60.09 for the expo-
nents, and 20.93 60.13 and 21.1960.15 for the multiplying term. This shows
that inferring petrogenic OC yield from the relationship between suspended sedi-
ment yield and petrogenic OC yield does not introduce a systematic bias in the
determination of biospheric POC yield.
Estimating global riverine petrogenic and biospheric POC fluxes. Robust rela-
tionships between both petrogenic and biospheric POC yields and suspended
sediment yield enablethe use of global suspended sediment yield to estimate global
petrogenic and biospheric POC fluxes. Global suspended sediment fluxes have
been the subject of extensive research over the past several decades. Suspended
sediment fluxes for individual gauged rivers have been compiled
25,26,35,36
and used
to derive estimates of global suspended sediment fluxes. Usually, fluxes were first
extrapolated regionally (for example by grouping rivers according to the oceanic
basin they drain into) to account for regional differences in average suspended
sediment yield (for example very high in small ocean islands versus small in the
Russian Arctic). Global fluxes were then obtained by summing the regional fluxes
for all areas draining into oceanic basins (in other words, endorheic systems are
excluded from the global estimate). Peucker-Erhenbrink
26
and Milliman and
Farnsworth
25
have provided the two most comprehensive recent compilations
of gauged suspended sediment fluxes. They estimated the global suspended sedi-
ment flux to the ocean at 18,548 and 19,000 Mtyr
21
, respectively. Recently,Larsen
et al.
27
estimated the global exorheic denudation flux at 19,000 Mt yr
21
, using an
empirical denudationmodel based on the strong relationship between denudation
rates and topography. Here we use a global suspended sediment flux of
19,000 6500 Mt yr
21
. Graham et al.
37
estimated the global exorheic land surface
at 110 310
6
km
2
, whereas Syvitski et al.
38
proposed a slightly lower estimate of
106 310
6
km
2
. The difference between these two estimates probably derives from
differences in corrections for endorheic drainage areas
26
. Here we use an average
value of (108 62) 310
6
km
2
, resulting in an estimated globalsuspended sediment
yield of 176 68tkm
22
yr
21
. It should be noted that this value attempts to correct,
as far as possible, for recent damming of river basins, because pre-dam sediment
fluxes were used to estimate the global suspended sediment fluxes whenever
possible both by Peucker-Ehrenbrink et al.
26
and by Milliman and Farnsworth
25
.
Finally, we use the calculated value for global suspended sediment yield (176 68t
km
22
yr
21
) and our relationships between petrogenic and biospheric POC yields
and suspended sediment yield to derive estimates of the global petrogenic and
biospheric POC fluxes. To estimate the uncertainties associated with calculated
global fluxes we first use Monte Carlo simulations (10,000 repetitions) to account
for the uncertainty associated with the determination of the relationships between
petrogenic and biospheric POC yields and suspended sediment yield. Then we
propagate the uncertainty (68tkm
22
yr
21
) in the global suspended sediment
yield. Last, we compare the lower bound of the calculated global petrogenic and
biospheric POC fluxes with the sum of all petrogenic and biospheric POC fluxes,
respectively, and use the highest of the two values as the lower bound of our final
estimate.
30. Voss, B. M. Spatialand Temporal Dynamics of BiogeochemicalProcesses in the Fraser
River, Canada: a Coupled Organic–Inorganic Perspective. PhD thesis,
Massachussetts Institute of Technology and Woods Hole Oceanographic
Institution (2014).
31. Wakeham, S. G. et al. Partitioning of organic matter in continental margin
sediments among density fractions. Mar. Chem. 115, 211–225 (2009).
32. Alam, M. J.,Nagao, S., Aramaki, T., Shibata, Y. & Yoneda, M.Transport of particulate
organic matter in the Ishikari River, Japan during spring and summer. Nuclear
Instrum. Meth. Phys. Res. B 259, 513–517 (2007).
33. Hilton, R. G. et al. Erosion of organic carbon in the Arctic as a geological carbon
dioxide sink. Nature (submitted).
34. Drenzek, N. J., Montlucon, D. B., Yunker, M. B., Macdonald, R. W. & Eglinton, T. I.
Constraints on the origin of sedimentary organic carbon in the Beaufort Sea from
coupledmolecular
13
Cand
14
C measurements. Mar. Chem. 103,146–162 (2007).
35. Milliman,J. D. & Meade, R. H. World deliveryof river sediment to the oceans.J. Geol.
1, 1–21 (1983).
36. Milliman,J. D. & Syvitski, P. M. Geomorphic/tectonic controlof sediment discharge
to the ocean: the importance of small mountainous rivers. J. Geol. 100, 525–544
(1992).
37. Graham, S. T., Famiglietti, J. S.& Maidment, D. R. Five-minute, 1/2 degrees, and1
degrees data sets of continentalwatersheds and river networks for use in regional
and global hydrologicand climate system modeling studies. Wat. Resour. Res. 35,
583–587 (1999).
38. Syvitski,J. P. M., Vorosmarty,C. J., Kettner, A. J. & Green,P. Impact of humanson the
flux of terrestrial sediment to the global coastal ocean. Science 308, 376–380
(2005).
39. Rosenheim, B. E. et al. River discharge influences on particulate organic carbon
age structure in the Mississippi/Atchafalaya River System. Glob. Biogeochem.
Cycles 27, 154–166 (2013).
40. Gaillardet,J., Dupre
´, B., Louvat,P. & Alle
`gre, C. J. Globalsilicate weathering and CO
2
consumption rates deduced from the chemistry of large rivers. Chem. Geol. 159,
3–30 (1999).
RESEARCH LETTER
G2015 Macmillan Publishers Limited. All rights reserved
Biospheric POC Yield (t/km2/yr)
NPP (t/km2/yr)
0.01
0.1
1
10
100
10 100 1000 10000
Ybios = 0.0004 x NPP1.16
r2 = 0.30 ; p<0.001
Extended Data Figure 1
|
Global relationship between biospheric POC yield and NPP. Basin-averaged NPP estimates were derived from the MOD17
database
24
.
LETTER RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved
0.01
0.1
1
10
100
1 10 100 1000 10000 100000
Sediment yield (t/km
2/yr)
CO2 sequestration yield (tC/km2/yr)
YCsil = 0.26 x Ysed0.28
YCorg = 0.022 x Ysed0.66
BE = 100%
BE = 30%
Extended Data Figure 2
|
Global relationship between long-term CO
2
sequestration yield and sediment yield. CO
2
sequestration throughterrestrial
biospheric POC burial (black dots; Y
Corg
) is more sensitive to sediment yield
than CO
2
sequestration through silicate weathering (grey crosses; Y
Csil
). At
high physical erosion rates (that is, high sediment yield), the burial of terrestrial
biospheric POC becomes the dominant long-term atmospheric CO
2
sink. The
dotted lines show CO
2
sequestration through terrestrial biospheric POC burial
for the entire set of biospheric POC export data (Fig. 2), assuming constant
burial efficiencies (BE) of 30 and 100%. CO
2
sequestration data through silicate
weathering are from Gaillardet et al.
40
.P= 0.001; r
2
= 0.80.
RESEARCH LETTER
G2015 Macmillan Publishers Limited. All rights reserved
0
0.5
1
1.5
2
2.5
3
0 0.5 1 1.5 2 2.5 3
Biospheric C Fm = 0.88
[Petrogenic C] = 0.05%
[OC] (%)
[Modern OC] (%)
Extended Data Figure 3
|
Organic carbon and radiocarbon contents of bulk
suspended sediments and grain size fractions in the Mississippi River.
Results are expressed as modern organic carbon (that is, the product of modern
fraction (Fm) and organic carbon content).The linear best fit gives the absolute
petrogenic OC content (0.05%) as well as the Fm of the biospheric POC (0.88).
Data from Wakeham et al.
31
and Rosenheim et al.
39
.P= 0.001; r
2
= 0.99.
LETTER RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved
0
0123456
1
2
3
4
5
6
[OC] (%)
[Modern OC] (%)
Biospheric C Fm = 1.00
[Petrogenic C] = 0.54 %
Extended Data Figure 4
|
Organic carbon and radiocarbon contents of bulk
suspended sediments from the Ishikari River, collected over a wide range of
flow regimes. Results are expressedas in Extended Data Fig. 3. The linear best
fit gives the absolute petrogenic OC content (0.54%) as well as the Fm of the
biospheric POC (1.00). Data from Alam et al.
32
.P= 0.001; r
2
= 0.99.
RESEARCH LETTER
G2015 Macmillan Publishers Limited. All rights reserved
