Global Iron Connections Between Desert
Dust, Ocean Biogeochemistry, and Climate
T. D. Jickells,
Z. S. An,
K. K. Andersen,
A. R. Baker,
J. J. Cao,
P. W. Boyd,
R. A. Duce,
K. A. Hunter,
P. S. Liss,
J. M. Prospero,
A. J. Ridgwell,
The environmental conditions of Earth, including the climate, are determined by
physical, chemical, biological, and human interactions that transform and transport
materials and energy. This is the ‘‘Earth system’’: a highly complex entity characterized
by multiple nonlinear responses and thresholds, with linkages between disparate
components. One important part of this system is the iron cycle, in which iron-
containing soil dust is transported from land through the atmosphere to the oceans,
affecting ocean biogeochemistry and hence having feedback effects on climate and dust
production. Here we review the key components of this cycle, identifying critical
uncertainties and priorities for future research.
ron is an essential nutrient for all organisms,
used in a variety of enzyme systems,
including those for photosynthesis, respira-
tion, and nitrogen fixation (1, 2). However, iron
is very insoluble under oxidizing conditions
above pH 4 (3). For marine phytoplankton,
separated from the iron-rich sediment of the
ocean floor by considerable water depths,
physiological iron requirements must be met
from within the water column. Iron supply is a
limiting factor on phytoplankton growth over
vast areas of the modern ocean, although this
may not have been so in the distant past, when
prokaryotes first evolved in a less oxic ocean (1).
Iron supply reaches the oceans mainly from
rivers as suspended sediment in a vast global
transport system (Table 1). However, fluvial and
glacial particulate iron is efficiently trapped in
near-coastal areas (4), except where rivers dis-
charge directly beyond the shelf. Hydrothermal
inputs are rapidly precipitated at depth in the
oceans. Hence, the dominant external input of
iron to the surface of the open ocean is aeolian
dust transport, mainly from the great deserts of
the world. Currently hyper-arid areas such as the
Sahara desert occupy 0.9 billion hectares and
drylands occupy 5.2 billion hectares, which is
one-third of global land area. These environ-
ments are particularly sensitive to global change
pressures (5, 6), and such changes could alter
ocean productivity and hence climate. There are
other possible contributors to atmospheric iron
supply, including volcanic, anthropogenic, and
extraterrestrial sources (7, 8), whose iron may be
more soluble than iron in soil aluminosilicates
(7), and these merit further study.
Dust produced in arid areas has important
and disparate effects throughout the Earth
system, as illustrated in Fig. 1 and discussed
below. These need to be incorporated into
climate models to correctly predict impacts of
global change pressures. We first consider
each component of the system before attempt-
ing a global synthesis.
Climate Effects on Dust/Iron Fluxes
Satellite imagery has greatly increased our
knowledge of large-scale dust source regions,
emphasizing the importance of localized
sources, which vary seasonally. There are sim-
ilar climatic and geomorphological controls on
many source regions (6), and dried-out lake
systems such as the Bodele Depression in
North Africa appear to be particularly impor-
tant. Dust production depends on the supply of
wind-erodible material, which ironically usu-
ally requires fluvial erosion, often from
adjacent highlands, followed by subsequent
drying out and the loss or absence of veg-
etative protection (6, 9–11). Dust production
arises from saltation or sandblasting, when
winds above a threshold velocity transport soil
grains horizontally, producing smaller par-
ticles, a small proportion of which get carried
up into the atmosphere for long-range trans-
port. These processes depend on rainfall,
wind, surface roughness, temperature, topog-
raphy, and vegetation cover, which are inter-
dependent factors linked to aridity and climate
in a highly nonlinear way. Wind tunnel studies
show dust production to be proportional to the
cube of wind speed (5).
Desert dust aerosol is dominated by par-
ticles of diameter 0.1 to 10 mm, with the mean
size being around 2 mm. Such aerosols have a
lifetime of hours to weeks, allowing long-
range transport over scales of thousands of
School of Environmental Sciences, University of East
Anglia, Norwich NR47TJ, UK.
State Key Lab of Loess
and Quaternary Geology, Institute of Earth Environ-
ment, Chinese Academy of Sciences, AS, 10 Fenghui
South Road, Post Office Box 17, China.
Institute, University of Copenhagen, Juliane Maries Vej
30, 2100 Copenhagen, Denmark.
universitaire des Syste
mes Atmospherique, Universite
Paris 7 and Paris 12, UMR CNRS 7583, Paris, France.
National Institute of Water and Atmospheric Research
Centre for Chemical and Physical Oceanography, De-
partment of Chemistry;
Department of Chemistry, Post
Office Box 56; University of Otago, Dunedin, New
Departments of Oceanography and Atmo-
spheric Sciences, Texas A&M University, TAMU 3146,
College Station, TX 77843–3146, USA.
Survey of Japan, National Institute of Advanced
Industrial Science and Technology (AIST), Tsukuba-
higashi 1-1-1, Ibaraki 305–8567, Japan.
Marine Sciences, Middle East Technical University,
Post Office Box 28, Erdemli-Mersin 33731, Turkey.
Leibniz-Institute fu¨r Meereswissenchaften IFM-GEO-
MAR, Marine Biogeochemistry, Geba
Du¨sternbrooker Weg 20, 24105 Kiel, Germany.
tional Center for Atmospheric Research. Post Office
Box 3000, Boulder, CO 80307, USA.
School of Marine and Atmospheric Sciences, University
of Miami, 4600 Rickenbacker Causeway, Miami, FL
Department of Earth and Ocean
Sciences, University of British Columbia, 6339 Stores
Road, Vancouver, British Columbia V6T 1Z4 Canada.
Max-Planck-Institute for Biogeochemistry. Post Of-
fice Box 10, 01 64 07701 Jena, Germany.
dad de Conception, Departamento de Oceanografia,
Casilla 160C, Chile.
*Present address: Institute for Tropospheric Research,
Permoserstrasse 15 04318, Leipzig, Germany.
Table 1. Global iron fluxes to the ocean (in Tg
of Fe year
). From Poulton and Raiswell (4),
with modified atmospheric inputs from Fig.
2. ‘‘Authigenic fluxes’’ refer to releases from
deep-sea sediments during diagenesis. We
distinguish only separately dissolved and
particulate for fluvial inputs, because it is
clear that fluvial particulate iron, along with
iron from coastal erosion and glacial sedi-
ment sources, does not reach the oceans,
whereas authigenic, atmospheric, and hy-
drothermal iron all reach the oceans regard-
less of their phase.
Fluvial particulate total iron 625 to 962
Fluvial dissolved iron 1.5
Glacial sediments 34 to 211
Coastal erosion 8
www.sciencemag.org SCIENCE VOL 308 1 APRIL 2005
kilometers (5, 11) but producing strong gra-
dients of dust deposition and concentrations
that vary substantially on time scales of È1
day. Dust production, transport, and deposition
to the oceans again depend on climatic factors,
particularly atmospheric structure, which regu-
lates uplift, and wind speed and precipitation,
which influence removal. Much of the transport
of dust occurs at altitudes of several kilo-
meters, with subsequent removal by wet
deposition. Hence, satellite images of dust may
not reflect dust inputs to the oceans (7).
Dust removal occurs by wet and dry
deposition, processes whose efficiency varies
with aerosol particle size (7). Measurements at
a limited number of sites and modeling studies
suggest that 30 to 95% of total removal is by
wet deposition (7, 12). Wet deposition is
spatially variable, reflecting several climatical-
ly sensitive factors, including aerosol size
distribution, rainfall patterns, and transport
altitude. This results in considerable uncertain-
ties in estimating the contribution of wet
deposition to total deposition.
Dust fluxes can be estimated from direct
measurements and subsequent extrapolation
(13), models (11, 14, 15), and satellite obser-
vations (16). The various approaches all yield
similar dust deposition estimates of 1000 to
2000 Tg year
(1 Tg 0 10
substantially from year to year. However,
models are usually tuned to match observations,
and hence the agreement is not truly an
Existing models of global dust transport
(14, 15) include only first-order physical rep-
resentations of the key dust production
processes, largely because of the lack of
suitable data sets of global surface character-
istics. Despite this, global models seem able to
simulate dust deposition fluxes reasonably
well. We estimate production at 1700 Tg
, with almost two-thirds from North
Africa and 26% of the dust reaching the
oceans (Fig. 2). Changes in the hydrological
cycle and/or vegetative cover affect global dust
production (10) as recorded over glacial/
interglacial cycles in loess, ice core, and
marine sediment records. Dust fluxes were 2
to 20 times higher during the last glaciation
(17–19) because of stronger winds, aridity,
changes in vegetation cover, lowered sea level,
and reduced precipitation. It has been sug-
gested that changing land use practices over
recent decades have altered dust fluxes by up
to 50% (14, 20), although recent work suggests
lower values (15). Although the global impor-
tance of land use change as a dust source is
currently uncertain, effects at a regional scale
are clear, such as around the Aral Sea and the
in the U.S. 1930s Dust Bowl storms (6). Dust
storm frequency over the Sahel appears to
have increased since 1950. This may be related
to multidecadal-scale climate variability or
landusechange(21). Dust transport over
China, the United States, and North Africa
has been related to large-scale climatic cycles
(22–26), and the variability in dust transport
can be influenced by climatic cycles such as El
Nin˜o–Southern Oscillation and North Atlantic
Oscillation (23, 24). Some climate models
suggest that enhanced greenhouse warming
could ‘‘green’’ the Sahel and southern Sahara
(14, 27), drastically altering global dust pro-
duction. Different models of global dust flux
Fig. 1. Schematic view of global iron and dust connections. Highlighted are the four critical
components (clockwise from top): the state of the land surface and dust availability, atmospheric
aerosol loading, marine productivity, and some measure of climatic state (such as mean global
surface temperature). The sign of the connections linking these varies; where the correlation is
positive (for example, increased atmospheric aerosol loading Y increased marine productivity),
the line is terminated with a solid arrowhead. Where the correlation is negative (for example,
increased marine productivity Y lower CO
and a colder climate), the termination is an open
circle. Connections with an uncertain sign are terminated with an open arrowhead. The
mechanism by which the link acts (for example, the impact of a change in atmospheric CO
via the radiative forcing of climate) is displayed in italics. Finally, the ‘‘water tap’’ symbols
represent a secondary mechanism modulating the effect of a primary mechanism; for instance, a
change in global precipitation strength and distribution will alter the efficiency with which
entrained dust is transported to the open ocean. If a path of successive connections can be traced
from any given component back to itself, a closed or feedback loop is formed. An even number
(including zero) of negatively correlated connections counted around the loop gives a positive
feedback, which will act to amplify a perturbation and tend to destabilize the system. Conversely,
an odd number of negative correlations gives a negative feedback, dampening any perturbation
and thus stabilizing the system. For instance, atmospheric aerosol loading Y marine productivity Y
climatic state Y dust availability Y atmospheric aerosol loading contains two negative and two
positive correlations and thus is positive overall. In contrast, marine productivity looping back onto
itself contains a single negative correlation and thus represents a negative feedback.
1 APRIL 2005 VOL 308 SCIENCE www.sciencemag.org
(14, 15) predict similar present-day dust fluxes
(Fig. 2), but their predictions for the next 100
years range from a modest increase (þ12%) to
a significant decrease (–60%), with different
regional deposition patterns. These differences
reflect variations in the relative importance of
land use changes and CO
fertilization, as well
as differences in climate predictions.
Dust has important but uncertain direct
impacts on climate and radiative budgets
(20, 28) and possibly rainfall patterns (29).
We note the importance of these physical
effects but focus here on the biogeochemical
effects. In a biogeochemical context, the key
flux to the oceans is not dust, but soluble or
bioavailable iron. Although the iron content of
soil dust (average, 3.5%) is variable globally,
the uncertainty introduced by this variability is
small compared with other uncertainties in the
iron cycle (7). Iron solubility from soil dust is
low [G1to2%(7)]. Higher solubilities of
aerosol iron have been reported (12). The
controls on aerosol iron solubility include
photochemistry (photoreduction of Fe III to
Fe II) and acidity, particularly during aerosol
cloud processing (7). Emissions of acid
) have more than
doubled from the preanthropogenic state, and
emissions are expected to continue to
increase (20). Organic complexation may play
a role in regulating atmospheric iron solubility.
Consequently, emissions of organic matter
from natural sources (such as soil humic acids
and plant terpenes) and anthropogenic sources
(such as biomass burning and industrial/urban
emissions) may influence atmospheric iron
cycling (7). We know very little about the
organic chemistry of aerosols or of the active
microbial community identified in aerosols,
which may influence iron solubility (7, 30). All
these factors (acidity, organic complexation,
and photochemistry) will alter with global
Dust/Iron Impacts on the Ocean
The physicochemical environment of atmo-
spheric iron changes dramatically on entering
the oceans. At a seawater pH of 8, soluble ferric
iron rapidly reprecipitates, setting up a competi-
tion between adsorption to water column partic-
ulates, active biological uptake, and organic
complexation, which evolves over the surface
water residence time of dust [tens of days (31)].
Experimental measurements of the solubility of
aerosol iron have generally been conducted
over shorter time scales and hence may not
adequately predict the solubility of aerosol iron.
Measuring total and speciated iron concen-
trations in the ocean is difficult, but global data
are now emerging (32). Total dissolved iron
shows nutrient-like oceanic profiles, with low
surface water concentrations (0.03 to 1 nmol
where photochemically produced Fe II
may be significant) increasing to deep water
concentrations of 0.4 to 2 nmol liter
Significant colloidal iron is present in the
water column and is potentially biogeochemi-
cally labile (32). The impact of atmospheric
deposition on surface water iron concentrations
has been demonstrated (34), as has recycling
from sediments and coastal regions (35, 36).
Within the oceans, dissolved iron is predomi-
nantly organically complexed, stabilizing it
against rapid scavenging (37), although its
residence time is still probably only decades
(32, 33). The source, biological function, and
structure of these organic iron-complexing
ligands are essentially unknown. Electrochemical
titrations suggest that some have similar binding
strength to that of true siderophores: strong iron-
specific ligands (3, 33, 37). Siderophores have
been found in marine bacteria and coastal sea-
water (38). Although many species may be able
to use siderophore-bound iron, siderophore syn-
thesis systems are not readily identifiable in the
genomes of important picophytoplankton species
such as Synecochoccus and Prochlorococcus
(39, 40), though they may be present in
Trichodesmium and Crocosphaera (two marine
diazotrophs) and in uncultured heterotrophic
bacterial genomes from the Sargasso Sea (41).
Iron limitation reflects deep-water Fe/N
concentration ratios that are inadequate to
meet phytoplankton iron requirements (36)
because of scavenging of iron regenerated
from sinking organic matter in the deep ocean
at faster rates than N. Thus, sustaining open
ocean phytoplankton primary production
requires an additional input of iron to that
produced from upwelling, which is usually
atmospheric. The relative importance of atmo-
spheric and upwelling sources varies throughout
the oceans (36). Iron limitation of phytoplank-
ton primary production in as much as 30% of
the oceans has now been suggested (1, 36, 42).
In some areas, such as the Southern Ocean,
this results in incomplete use of macronutrients
(N, P, and Si) and relatively low algal abun-
dance, hence the term ‘‘high-nutrient low-
chlorophyll’’ (HNLC) regions. Recent studies
emphasize more complex interactions within the
ocean than simple iron limitation or sufficiency,
with evidence in some areas of simultaneous
limitation of primary production by iron, light,
Table 2. Effects of dust/iron (Fe) on ocean biogeochemistry. (In addition, there are dust effects on the climate system via albedo and the hydrological
cycle; see text.)
Interaction Mechanism Area* Reference
Primary productivity Reduction in Fe limitation allows more efficient use of macro-
nutrients and hence CO
HNLC and other Fe-limited areas (36, 42)
fixation Reduction in Fe limitation on nitrogen fixation increases primary
production and hence CO
Subtropical gyres (1, 43)
Changes in species
Species-selective relief of iron stress. Global (42)
Ballast effect Increases sinking rate of organic matter, reducing organic matter
regeneration within seasonal mixed layer; promotes CO
Probably only significant in areas
of high dust deposition
DMS Increased productivity leads to increased DMS emissions and
increased aerosol formation.
HNLC and other Fe-limited areas (54)
O and NO
Increased fluxes of organic matter to deep waters lower oxygen
concentrations and promote denitrification, release N
lower oceanic nitrate inventory.
Upwelling systems (53)
O and CH
Increased productivity leads to changes in euphotic zone methane
HNLC and other Fe-limited areas (54, 57)
S Increased fluxes of organic matter to deep waters lower oxygen
concentrations and promote sulfate reduction; sulphide produc-
tion lowers iron inventory.
Biogenic gases linked to primary productivity. These are greenhouse
gases, linked to aerosol formation and to the ozone cycle.
As for DMS (54)
Isoprene and CO Biogenic trace gases linked to primary productivity. These gases
influence atmospheric oxidizing capacity.
As for DMS (54, 57)
*Most impacts have effects throughout the oceans, but where appropriate, we identify here areas that are most sensitive to changes in dust/iron flux.
www.sciencemag.org SCIENCE VOL 308 1 APRIL 2005
macronutrients (42, 43), and trace nutrients
(suchasCoandZn)(2, 44). Furthermore
atmospheric inputs supply not only iron but also
other nutrients and carbonate, which may
influence ocean biogeochemistry (45, 46).
Luxury iron uptake has been demonstrated
for some phytoplankton, allowing them to better
adapt to episodic atmospheric supply (47). Iron
availability influences algal community struc-
ture as well as overall productivity. Open ocean
phytoplankton generally need less iron than
coastal species, which have evolved in a more
iron-rich environment, although iron-limited
coastal systems are known (36). A reduced
iron requirement can be achieved by reducing
cell size or minimizing the number of iron-
containing enzymes (39). The success of
Prochlorococus in HNLC areas depends on
both strategies. Relief of iron stress results in
the growth of phytoplankton taxa character-
ized by larger cells, particularly diatoms with
less dense opal skeletons (36). A similar
process may arise for coccolithophores as a
result of Fe/Zn co-limitation (44). Changes in
skeleton density should influence sinking
rates and hence carbon export to depth,
although this has not been seen in field
experiments (48). Changes in coccolithophore
abundance directly affect atmospheric partial
pressure of CO
), because their cal-
cification produces CO
(36, 42). In addition
to direct limitation of primary production in
the HNLC regions, iron may limit (or co-limit
with P) nitrogen fixation by photosynthetic
diazotrophs in tropical oceans, where stratifi-
cation creates high temperature and irradiance
and low nitrate concentrations in surface
waters, which favor this process (1, 43). The
best characterized photosynthetic diazotroph,
Trichodesmium, requires 5 to 10 times more
iron for growth based on nitrogen fixation, as
compared to ammonium (47).
The supply of dust to the oceans is very
important in maintaining oceanic primary pro-
duction and CO
uptake but is sensitive to
climate change, although the overall effect
will vary between ocean biogeochemical prov-
inces (Table 2). In HNLC regions, changes in
iron supply will directly affect primary pro-
duction and species composition, whereas in
subtropical/tropical oligotrophic regions, the
impact will be mainly via changes in nitrogen
fixation. The dust supply from the great North
African and Asian deserts directly affects the
tropical North Atlantic and temperate North
Pacific, respectively, and effects in the two
regions can be expected to be different. The
largest HNLC region, the Southern Ocean
(36), has the biggest potential to influence
. Here atmospheric dust
supply is low (Fig. 2), originating from small
dust sources in Argentina, Australia, and South
Africa (6). Changes in these small and little-
studied desert regions may have a dis-
proportionately large global impact.
Because the solubility of iron from dust is
low, it follows that there is a large flux of
particulate iron through the deep ocean,
particularly beneath the major dust plumes. If
some of this dust dissolves at depth, it will
increase abyssal dissolved iron concentrations
and, over the long term, productivity in
upwelling regions such as the Southern Ocean.
Deep-water dust dissolution will depend on
organic ligand concentrations and possibly
sediment redox (33).
Martin (49) proposed that increased dust
transport during the last glaciation reduced
iron limitation in HNLC regions, increasing
primary production and CO
complexity of iron biogeochemistry and
nutrient co-limitation means that higher gla-
cial dust loadings need not necessarily cause
increased productivity. Current models and
ice core data yield very different results,
predicting that glacial/interglacial changes in
dust fluxes will change atmospheric pCO
5 to 45 parts per million (ppm) as a con-
tribution to the total change of 80 to 100 ppm
(19, 50). Bopp et al.(50) reviewed much of
the existing marine sediment core data on
glacial/interglacial ocean productivity changes
and found no simple global pattern of change.
However, there are regional patterns (51)with
increases in productivity in the northwest
Pacific, South Atlantic, and Indian Oceans
north of the polar front, with decreases south
of it. South Pacific productivity appears to be
little changed. Some of these patterns can be
reproduced in ocean models (50).
Effect on Climate of Iron Inputs
to the Oceans
The oceans clearly exert a major influence on
climate via heat transport and related physical
processes (20). Large-scale reorganization of
oceanic circulation will also affect the trans-
port of iron, effects driven predominantly from
within the ocean. Climate change will induce a
variety of physicochemical changes in the
open ocean, particularly by changing stratifi-
cation and nutrient supply ratios (42), with
unpredictable effects. We acknowledge these
important issues but focus on the dust cycle,
considering now ways in which this can affect
the oceans and climate, beside the direct iron
limitation of primary production and nitrogen
fixation discussed above (Table 2).
Changes in iron fluxes can result in species
shifts and changes in phytoplankton size
distribution, changing oceanic CO
altering the efficiency of organic carbon export
to deep water. Dust may also play a direct role
in regulating export via the ballast effect (52).
In most areas, dust is a minor ballast compo-
nent compared to opal and calcite, but their
production is also influenced by dust/iron
supply. Changes in ocean productivity and
organic carbon export to deep water will
influence subsurface oxygen levels and thereby
denitrification in oxygen minima zones, oce-
anic nitrate inventories and productivity, and
nitrous oxide emissions (53). Changes in
S in such areas could affect deep-
ocean iron concentrations and productivity.
Up to eightfold changes in dimethyl sulfide
(DMS) concentrations are seen in iron addition
experiments (54). DMS oxidizes in the atmo-
sphere to form acidic sulfate aerosol, a highly
effective scatterer of solar radiation. Modeling
suggests that a twofold global rise in DMS
fluxes produces a global temperature decrease
of 1-C, proving a climate feedback and linking
Fig. 2. Dust fluxes to the world oceans based on a composite of three published modeling studies
that match satellite optical depth, in situ concentration, and deposition observations (11, 14, 15).
The models have been extensively compared to observations, and although individual models
show strengths and weaknesses, this composite appears to match observations well. Total
atmospheric dust inputs to the oceans 0 450 Tg year
. Percentage inputs to ocean basins based
on this figure are as follows: North Atlantic, 43%; South Atlantic, 4%; North Pacific, 15%; South
Pacific, 6%; Indian, 25%; and Southern Ocean, 6%.
1 APRIL 2005 VOL 308 SCIENCE www.sciencemag.org
the C, Fe, and S cycles (55). DMS is only one
of a group of trace gases that can influence
climate and whose emissions are sensitive to
iron concentrations (54, 56, 57). These include
gases that directly affect greenhouse gas
forcing (nitrous oxide and methane), ozone
cycling (halocarbons and alkylnitrates), and
atmospheric oxidizing capacity (isoprene and
carbon monoxide). Impacts on ozone are
important in radiative forcing (20) and via
ultraviolet-B impacts on phytoplankton com-
munity structure (42).
Global Iron Connections
Our analysis demonstrates the complexity of the
global iron cycle (Fig. 1). Low iron solubility
leads to limitation of marine productivity, with
potentially large-scale feedbacks within the
global climate system. These could act to either
amplify future global climate change via a
positive (destabilizing) feedback or diminish it
via a negative (stabilizing) feedback. There are
considerable uncertainties in our understanding
of these interactions, requiring research that
integrates across the whole Earth system. We
suggest the following research priorities: (i) dust
deposition processes, (ii) aerosol iron bioavail-
ability, and (iii) the impact of iron on marine
nitrogen fixation and trace gas emissions. These
should lead to improvements in global models,
allowing realistic predictive capability that can
be tested against improved results from the
paleo record of the biogeochemical response to
changing dust fluxes.
There are discussions about changing
terrestrial land uses to create carbon sinks to
help mitigate global change. Such changes
may reduce dust fluxes to the ocean and
thereby reduce primary productivity, offsetting
gains in terrestrial carbon storage (58). There is
also discussion about fertilizing the ocean with
iron to increase CO
uptake (59). Our analysis
demonstrates that such a scheme could pro-
duce many changes in marine biogeochemical
systems. Clearly, we need a comprehensive
understanding of the current and future dust/
iron cycle before we can contemplate such
engineering of the Earth system.
References and Notes
1. P. G. Falkowski, R. T. Barber, V. Smetacek, Science
281, 200 (1998).
2. F. M. M. Morel, N. M. Price, Science 300, 944 (2003).
3. S. M. Kraemer, Aquat. Sci. 66, 3 (2004).
4. S. W. Poulton, R. Raiswell, Am. J. Sci. 302, 774 (2002).
5. R. A. Duce, in Aerosol Forcing of Climate,R.J.
Charlson, J. Heintzenberger, Eds. (Wiley, Chichester,
UK, 1995), pp. 43–72.
6. J. M. Prospero, P. Ginoux, O. Torres, S. E. Nicholson,
T. E. Gill, Rev. Geophys. 40, 1002 (2002).
7. T. D. Jickells, L. J. Spokes, in The Biogeochemistry of
Iron in Seawater, D. R. Turner, K. Hunter, Eds. (Wiley,
Chichester, UK, 2001), pp. 85–121.
8. K. S. Johnson, Global Biogeochem. Cycles 15, 61 (2001).
9. N. Brooks, M. Legrand, in Linking Climate Change to
Land Surface Change, S. McLaren, D. Kniveton, Eds.
(Kluwer Academic, Dordrecht, Netherlands, 2000),
10. C. S. Zender, D. Newman, O. Torres, J. Geophys. Res.
Atmos. 108, 4543 (2003).
11. P. Ginoux et al., J. Geophys. Res. Atmos. 106, 20255
12. J. L. Hand et al., J. Geophys. Res. Atmos. 109,D17205
13. R. A. Duce et al., Global Biogeochem. Cycles 5, 193
14. N. M. Mahowald, C. Luo, Geophys. Res. Lett. 30, 1903
15. I. Tegen, M. Werner, S. P. Harrison, K. E. Kohfeld,
Geophys. Res. Lett. 31, L05105 (2004).
16. Y. J. Kaufman et al., J. Geophys. Res. Atmos.
17. Z. S. An, G. Kukla, S. C. Porter, J. L. Xiao, CATENA 18,
18. D. K. Rea, Rev. Geophys. 32, 159 (1994).
19. R. Ro
thlisberger et al., Geophys. Res. Lett. 31, L16207
20. Intergovernmental Panel on Climate Change, Climate
Change 2001: The Scientific Basis. Contribution of
Working Group1 to the Third Assessment Report of
the Intergovernmental Panel on Climate Change
(Cambridge Univ. Press, Cambridge, 2001).
21. J. M. Prospero, R. T. Nees, Nature 320, 735 (1986).
22. X. Y. Zhang et al., Geophys. Res. Lett. 30, 10.1029/
23. J. M. Prospero, P. J. Lamb, Science 302, 1024 (2003).
24. C. Moulin, I. Chiapello, Geophys. Res. Lett. 31,
25. S. D. Schubert, M. J. Suarez, P. J. Pegion, R. D. Koster,
J. T. Bacmeister, Science 303, 1855 (2004).
26. C. S. Zhao, X. Dabu, Y. Li, Geophys. Res. Lett. 31,
27. M. Claussen, V. Brovkin, A. Ganopolski, C. Kubatzki,
V. Petoukhov, Clim. Change 57, 99 (2003).
28. B. J. Huebert et al., J. Geophys. Res. Atmos. 108,
29. D. Rosenfeld, Y. Rudich, R. Lahav, Proc. Natl. Acad.
Sci. U.S.A. 98, 5975 (2001).
30. D. W. Griffin et al., Aerobiologia 19, 143 (2003).
31. T. D. Jickells, Mar. Chem. 68, 5 (1999).
32. H.J.W.deBaar,J.T.M.deJong,inThe Biogeochemistry
of Iron in Seawater,D.R.Turner,K.A.Hunter,Eds.
(Wiley, Chichester, UK, 2001), pp. 123–253.
33. P. Parekh, M. J. Follows, E. Boyle, Global Biogeochem.
Cycles 18, GB1002 (2004).
34. G. Sarthou et al., Deep-Sea Res. I 50, 1339 (2003).
35. V. A. Elrod, W. M. Berelson, K. H. Coale, K. S. Johnson,
Geophys. Res. Lett. 31, L12307 (2004).
36. A. J. Watson, in The Biogeochemistry of Iron in
Seawater, D. R. Turner, K. A. Hunter, Eds. (Wiley,
Chichester, UK, 2001), pp. 9–39.
37. K. W. Bruland, E. L. Rue, in The Biogeochemistry of
Iron in Seawater, D. R. Turner, K. Hunter, Eds. (Wiley,
Chichester, UK, 2001), pp. 255–289.
38. M. Gledhill et al., Mar. Chem. 88, 75 (2004).
39. B. Palenik et al., Nature 424, 1037 (2003).
40. G. Rocap et al., Nature 424, 1042 (2003).
41. J. C. Venter et al., Science 304, 66 (2004).
42. P. W. Boyd, S. C. Doney, in Ocean Biogeochemistry: A
Synthesis of the Joint Global Ocean Flux Study
(JGOFS), M. J. R. Fasham, Ed. (Springer, Berlin,
2003), pp. 157–193.
43. M. M. Mills, C. Ridame, M. Davey, J. La Roche, R. J.
Geider, Nature 429, 292 (2004).
44. K. G. Schulz et al., Nature 430, 673 (2004).
45. A. R. Baker, S. D. Kelly, K. F. Biswas, M. Witt, T. D.
Jickells, Geophys. Res. Lett. 30, 2296 (2003).
46. T. Oba, T. F. Pedersen, Paleoceanography 14, 34 (1999).
47. A. B. Kustka et al., Limnol. Oceanogr. 48, 1869
48. P. W. Boyd et al., Nature 428, 549 (2004).
49. J. H. Martin, Paleoceanography 5, 1 (1990).
50. L. Bopp, K. E. Kohfeld, C. Le Quere, O. Aumont,
Paleoceanography 18, 1046 (2003).
51. Z. S. An, J. J. Cao, K. K. Anderson, H. Kawahata, in
52. R. Francois, S. Honjo, R. Krishfield, S. Manganini,
Global Biogeochem. Cycles 16, 1087 (2002).
53. M. A. Altabet, M. J. Higginson, D. W. Murray, Nature
415, 159 (2002).
P. S. Liss, Geophys. Res. Lett. 31, L14307 (2004).
55. G. Zhuang, Z. Yi, R. A. Duce, P. R. Brown, Nature 355,
56. P. S. Liss, A. Chuck, D. C. E. Bakker, S. Turner,
57. O. W. Wingenter et al., Proc. Natl. Acad. Sci. U.S.A.
101, 8537 (2004).
58. A. J. Ridgwell, M. A. Maslin, A. J. Watson, Geophys.
Res. Lett. 29, 1095 (2002).
59. S. W. Chisholm, P. G. Falkowski, J. J. Cullen, Science
296, 467 (2002).
60. We acknowledge the leadership of the Surface Ocean–
Lower Atmosphere Study project (www.solas-int.org)
for conceiving this synthesis. We thank the Interna-
tional Council for Science, the United Nations Educa-
tional, Scientific and Cultural Organisation, NSF (grants
GEO-0402845 and OCE 0228863), the International
Geosphere-Biosphere Programme, and the Scientific
Committee on Ocean Research for partial support of
this work, and B. Sulzberger for her contribution to this
www.sciencemag.org SCIENCE VOL 308 1 APRIL 2005