Southern Ocean iron enrichment experiment: carbon cycling in high- and low-Si waters.
ABSTRACT The availability of iron is known to exert a controlling influence on biological productivity in surface waters over large areas of the ocean and may have been an important factor in the variation of the concentration of atmospheric carbon dioxide over glacial cycles. The effect of iron in the Southern Ocean is particularly important because of its large area and abundant nitrate, yet iron-enhanced growth of phytoplankton may be differentially expressed between waters with high silicic acid in the south and low silicic acid in the north, where diatom growth may be limited by both silicic acid and iron. Two mesoscale experiments, designed to investigate the effects of iron enrichment in regions with high and low concentrations of silicic acid, were performed in the Southern Ocean. These experiments demonstrate iron's pivotal role in controlling carbon uptake and regulating atmospheric partial pressure of carbon dioxide.
- SourceAvailable from: Rachel U Shelley[Show abstract] [Hide abstract]
ABSTRACT: The atmospheric delivery of soluble and bioavailable iron (Fe) is essential for the biogeochemical functioning of many oceanic ecosystems where Fe is a limiting micronutrient for biological production. Aerosol samples associated with air masses characterized as European-influenced, primarily marine (no continental influence within 5 day back trajectories), or North African-influenced were collected along a cruise track in the eastern North Atlantic Ocean during a 2010 US GEOTRACES cruise. Aerosols were analyzed for total and soluble Fe and aluminum (Al) and organic matter (OM) loadings and OM chemical characteristics, to explore potential relationships between aerosol OM and Fe and Al that contribute to higher Fe and Al solubilities in combustion-influenced aerosols. Similar to the results from previous studies, North African-influenced air masses contained higher aerosol Fe (4.7–86 nmol m −3) and Al (13–240 nmol m −3) total loadings than European-influenced air masses (Fe: 0.63–2.7 nmol m −3 ; Al: 2.5–5.9 nmol m −3), but Fe and Al relative sol-ubilities were much higher for European (Fe: 2.1–4.6%; Al: 1.9–3.2%) versus North African-influenced aero-sols (Fe: 0.22–0.70%; Al: 0.39–1.1%). Water soluble organic carbon (WSOC) to trace metal ratios correlated positively with this trend in Fe and Al relative solubilities, as European-influenced WSOC/trace metal ratios ranged from ~2 to 32 while North African-influenced aerosol WSOC/trace metal ratios ranged from 0.04 to 0.51. Aerosols from primarily marine air masses showed the lowest Fe, Al, and OM loadings of all samples and Fe (0.71–2.5%) and Al (0.36–9.2%) solubilities that were variable and did not fit the patterns described for the continentally-influenced samples. Principal component analysis was employed on aerosol water soluble OM (WSOM) solution state 1 H nuclear magnetic resonance spectra and revealed the European-influenced aerosol WSOM to be characterized by higher contributions from acetic acid (a common photoproduct of atmospheric OM) and aliphatic hydrogens, while North African-influenced aerosol WSOM was characterized by carbohydrate-like compounds and compounds with unsaturations. The abundance of the acetic acid photoproduct in European-influenced aerosol WSOM suggests this WSOM to be rich in carboxyl groups that are thought to be strong Fe-binding ligands and provides evidence for the potential role of WSOM in maintaining aerosol Fe and Al solubilities.Marine Chemistry 08/2013; 154:24-33. · 3.00 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Diatoms of the iron-replete continental margins and North Atlantic are key exporters of organic carbon. In contrast, diatoms of the iron-limited Antarctic Circumpolar Current sequester silicon, but comparatively little carbon, in the underlying deep ocean and sediments. Because the Southern Ocean is the major hub of oceanic nutrient distribution, selective silicon sequestration there limits diatom blooms elsewhere and consequently the biotic carbon sequestration potential of the entire ocean. We investigated this paradox in an in situ iron fertilization experiment by comparing accumulation and sinking of diatom populations inside and outside the iron-fertilized patch over 5 wk. A bloom comprising various thin- and thick-shelled diatom species developed inside the patch despite the presence of large grazer populations. After the third week, most of the thinner-shelled diatom species underwent mass mortality, formed large, mucous aggregates, and sank out en masse (carbon sinkers). In contrast, thicker-shelled species, in particular Fragilariopsis kerguelensis, persisted in the surface layers, sank mainly empty shells continuously, and reduced silicate concentrations to similar levels both inside and outside the patch (silica sinkers). These patterns imply that thick-shelled, hence grazer-protected, diatom species evolved in response to heavy copepod grazing pressure in the presence of an abundant silicate supply. The ecology of these silica-sinking species decouples silicon and carbon cycles in the iron-limited Southern Ocean, whereas carbon-sinking species, when stimulated by iron fertilization, export more carbon per silicon. Our results suggest that large-scale iron fertilization of the silicate-rich Southern Ocean will not change silicon sequestration but will add carbon to the sinking silica flux.Proceedings of the National Academy of Sciences 11/2013; · 9.74 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: The growth dynamics of populations of interacting species in the aquatic environment is of great importance, both for understanding natural ecosystems and in efforts to cultivate these organisms for industrial purposes. Here we consider a simple two-species system wherein the bacterium Mesorhizobium loti supplies vitamin B12 (cobalamin) to the freshwater green alga Lobomonas rostrata, which requires this organic micronutrient for growth. In return, the bacterium receives photosynthate from the alga. Mathematical models are developed that describe minimally the interdependence between the two organisms, and that fit the experimental observations of the consortium. These models enable us to distinguish between different mechanisms of nutrient exchange between the organisms, and provide strong evidence that, rather than undergoing simple lysis and release of nutrients into the medium, M. loti regulates the levels of cobalamin it produces, resulting in a true mutualism with L. rostrata. Over half of all microalgae are dependent on an exogenous source of cobalamin for growth, and this vitamin is synthesised only by bacteria; it is very likely that similar symbiotic interactions underpin algal productivity more generally.The ISME Journal advance online publication, 13 February 2014; doi:10.1038/ismej.2014.9.The ISME Journal 02/2014; · 8.95 Impact Factor
28. W. W. Cochran, M. Wikelski, in Birds of Two Worlds, P.
Marra, R. Greenberg, Eds. (Johns Hopkins Press, Baltimore,
2004), in press.
29. R. Sandberg, et al., Auk 119, 201 (2002).
30. S. T. Emlen, Anim. Behav. 18, 215 (1970).
32. W. Wiltschko, R. Wiltschko, J. Comp. Physiol. 177,
33. T. Ritz, S. Adem, K. Schulten, Biophys. J. 78, 707 (2000).
34. T. Alerstam, J. Exp. Biol. 130, 63 (1987).
35. K. J. Lohmann, S. D. Cain, S. A. Dodge, C. M. F.
Lohmann, Science 294, 364 (2001).
36. K. P. Able, Nature 299, 550 (1982).
37. J. B. Phillips, F. R. Moore, Behav. Ecol. Sociobiol. 31,
38. K. P. Able, M. A. Able, J. Comp. Physiol. 177, 351 (1995).
39. R. C. Beason, Ethology 91, 75 (1992).
41. V. P. Bingman, W. Wiltschko, Ethology 77, 1 (1988).
42. E. Batschelet, Circular Statistics in Biology (Academic
Press, London, 1981).
inspiration; and G. Swenson, G. Swenson, and J. Brugge-
mann for help with light measurements and thank M.
Bowlin, N. Sapir, A. Medina, W. Cochran, J. Cochran, J.
Mandel, and S. David for their help, support, and under-
constructive comments on the manuscript. Supported by
sity (M.W.), Volkswagen Stiftung (H.M.), Oldenburg Uni-
versity (to H.M.) and NSF (GB 3155 and 6680 to W.W.C.).
Supporting Online Material
Materials and Methods
Figs. S1 and S2
Tables S1 and S2
20 January 2004; accepted 10 March 2004
Southern Ocean Iron Enrichment
Experiment: Carbon Cycling in
High- and Low-Si Waters
Kenneth H. Coale,1* Kenneth S. Johnson,2Francisco P. Chavez,2
Ken O. Buesseler,3Richard T. Barber,4Mark A. Brzezinski,5
William P. Cochlan,6Frank J. Millero,7Paul G. Falkowski,8
James E. Bauer,9Rik H. Wanninkhof,10Raphael M. Kudela,11
Mark A. Altabet,12Burke E. Hales,13Taro Takahashi,14
Michael R. Landry,15Robert R. Bidigare,16Xiujun Wang,1
Zanna Chase,2Pete G. Strutton,2Gernot E. Friederich,2
Maxim Y. Gorbunov,8Veronica P. Lance,4Anna K. Hilting,4
Michael R. Hiscock,4Mark Demarest,5William T. Hiscock,7
Kevin F. Sullivan,10Sara J. Tanner,1R. Mike Gordon,1
Craig N. Hunter,1Virginia A. Elrod,2Steve E. Fitzwater,2
Janice L. Jones,5Sasha Tozzi,8,9Michal Koblizek,8
Alice E. Roberts,6Julian Herndon,6Jodi Brewster,1
Nicolas Ladizinsky,1,6Geoffrey Smith,1David Cooper,1
David Timothy,12Susan L. Brown,16Karen E. Selph,16
Cecelia C. Sheridan,16Benjamin S. Twining,17
Zackary I. Johnson18
The availability of iron is known to exert a controlling influence on biological produc-
cycles. The effect of iron in the Southern Ocean is particularly important because of
its large area and abundant nitrate, yet iron-enhanced growth of phytoplankton may
be differentially expressed between waters with high silicic acid in the south and low
silicic acid in the north, where diatom growth may be limited by both silicic acid and
in regions with high and low concentrations of silicic acid, were performed in the
Southern Ocean. These experiments demonstrate iron’s pivotal role in controlling
carbon uptake and regulating atmospheric partial pressure of carbon dioxide.
partial pressure of carbon dioxide (pCO2) in the
atmosphere. Because rates of photosynthesis and
biological carbon export are low in Antarctic wa-
ters, macronutrients are largely unused, and up-
welled CO2entering the atmosphere (1, 2) sus-
tains the relatively high interglacial atmospheric
pCO2of the present day (3).
Southern Ocean surface waters contain ex-
tremely low iron concentrations (4, 5), and the
low rates of primary production have been at-
tributed to iron deficiency. Recent open-ocean
iron enrichment experiments demonstrate the
validity of this hypothesis in the Southern
Ocean (6, 7). Martin (8) proposed that natural
variations in the atmospheric iron flux ultimate-
ly regulate primary production in the Southern
Ocean and influence the pCO2of the atmo-
sphere, thereby potentially affecting the radia-
tive balance of the planet. Syntheses of models,
field observations, and paleoceanographic data
(3, 9, 10, 11, 12) support a role for iron-
regulated changes in Southern Ocean macronutri-
between iron-rich dust, marine production, and
atmospheric pCO2over the past four glacial cy-
cles as recorded in Antarctic ice cores (13). These
spheric pCO2remains uncertain.
Although all Southern Ocean surface waters
have high concentrations of nitrate and phos-
phate, silicic acid concentrations differ marked-
ly from north to south. Subantarctic waters
north of the Antarctic Polar Front Zone (APFZ)
have low Si concentrations (1 to 5 ?M), where-
as high Si (?60 ?M) is found to the south (fig.
S1). Diatoms, which require Si for growth, are
believed responsible for much of the carbon
export from the surface to the deep sea (14). In
1Moss Landing Marine Laboratories, 8272 Moss Landing
Road, Moss Landing, CA 95039–9647, USA.2Monterey
Bay Aquarium Research Institute, 7700 Sandholdt Road,
Moss Landing, CA 95039, USA.3Department of Marine
Chemistry and Geochemistry, Woods Hole Oceano-
University, 135 Duke Marine Lab Road, Beaufort, NC
28516, USA.5Marine Science Institute and the Depart-
ment of Ecology, Evolution, and Marine Biology, Univer-
sity of California, Santa Barbara, CA 93106, USA.6Rom-
berg Tiburon Center for Environmental Studies, San
Francisco State University, 3152 Paradise Drive, Tiburon,
CA 94920–1205, USA.7Rosenstiel School of Marine and
Atmospheric Research, University of Miami, 4600 Rick-
enbacker Causeway, Miami, FL 33149–1098, USA.8En-
vironmental Biophysics and Molecular Ecology Program,
Institute of Marine and Coastal Sciences and Depart-
ment of Geology, Rutgers University, 71 Dudley Road,
New Brunswick, NJ 08901–8521, USA.9Virginia Insti-
tute of Marine Science, Route 1208 Greate Road,
Gloucester Point, VA 23062, USA.10Atlantic Ocean-
ographic and Meteorological Laboratory/National
Oceanic and Atmosphereic Administration, 4301
Rickenbacker Causeway, Miami, FL 33149, USA.
11University of California at Santa Cruz, 1156 High
Street, Santa Cruz, CA 95064, USA.12School for Ma-
rine Science and Technology, University of Massachu-
setts, 706 South Rodney French Boulevard, New Bed-
ford, MA 02744–1221, USA.13College of Oceanic and
Atmospheric Sciences, Oregon State University Cor-
vallis, OR 97331, USA.14Lamont-Doherty Earth Obser-
vatory, Columbia University, 61 Route 9W, Palisades,
NY 10964–1000, USA.15Scripps Institution of Ocean-
ography, University of California, San Diego, 9500 Gil-
man Drive, La Jolla, CA 92093–0227, USA.16Depart-
ment of Oceanography, University of Hawaii at Manoa,
1000 Pope Road, Honolulu, HI 96822, USA.
University of New York, Stony Brook, NY 11794, USA.
18Massachusetts Institute of Technology, 48-336A MIT,
15 Vassar Street, Cambridge, MA 02139, USA.
*To whom correspondence should be addressed. E-
R E S E A R C H A R T I C L E S
16 APRIL 2004VOL 304SCIENCEwww.sciencemag.org
previous iron enrichment experiments, diatoms
have responded with the greatest increase in bio-
mass (6, 7, 15, 16). The large silicic acid gradient
across the APFZ (17) suggests that iron enrich-
ment would cause diatoms to bloom to the south
of the APF, whereas nonsiliceous phytoplankton
species would likely dominate in waters to the
north. The majority (?65%) of the Southern
Ocean surface waters lie in the northern subant-
arctic region with low silicic acid but high nitrate.
If increased macronutrient use does play a role in
based photosynthesis and carbon export should
occur after iron enrichment in both the high and
low silicic acid waters of the Southern Ocean.
Two previous open-ocean iron enrichment
experiments have been conducted in the South-
ern Ocean, one in the Australian sector [South-
ern Ocean iron release experiment (SOIREE)
(6)] and the other in the Atlantic sector [Eisen
(iron) experiment (EisenEx) (7)] (Fig. 1). Both
experiments demonstrated that Southern Ocean
phytoplankton respond directly to iron addition
by increasing primary production and biomass,
with a corresponding reduction in pCO2and
nitrate concentrations. However, both experi-
ments were performed in waters of moderate
south of the APFZ. Considerable Si depletion
occurs each year during natural phytoplankton
blooms in this zone (18, 19). The prior experi-
ments do not clarify the potential for Fe and Si
interactions to regulate the carbon cycle in the
very low–Si waters to the north of the APFZ or
no Si depletion occurs (18). This experiment,
the Southern Ocean Iron Experiment (SOFeX),
was designed to address that uncertainty.
Experimental conditions. We enriched
two distinct regions. The north patch was char-
acterized by high nitrate (?20 ?M) and low
silicic acid concentrations (?3 ?M), whereas
the south patch location had high nitrate (?28
?M) and high silicic acid concentrations (?60
?M). The average mixed layer temperatures
and depths at the north and south patch sites
were 5° to 7°C and 40 m and –0.5°C and 45 m,
respectively. Three ships were used to perform
the experiments. Both initial site surveys and
iron injections were carried out by research
vessel (RV) Revelle with the use of methods
previously described (20). Multiple enrich-
ments were performed on areas of 15 km by 15
km (three additions of iron to ?1.2 nM in the
north and four additions of iron to ?0.7 nM in
in the experimental areas (21) that exceed sat-
uration for phytoplankton growth (22). Similar
to other Southern Ocean iron enrichments (6,
7), iron persisted in the patch at high concen-
throughout the experiment (23). The iron-
enriched patches were subsequently sam-
pled in the north and south by RV Revelle
and RV Melville and in the south by U.S.
Coast Guard Research Ice Breaker Polar
Star. The details of the northern and south-
ern patch enrichments are shown in tables
S1 and S2, respectively, and the ship tracks
with regional nutrient fields are shown in
fig. S1, A and B, with local hydrography
and ship tracks shown in fig. S1C.
Patch dilution and mixing. The north
patch was deployed in a region with many
fronts and evolved from a square into a narrow
filament 7 km wide by at least 340 km long by
day 38 (Fig. 2A). In contrast, the south patch
slowly expanded in all directions throughout
the observational period to an area of ?2380
km2after 20 days (Fig. 2B). Such mixing will
act to dilute products of bloom development
[chlorophyll a (Chl a), particulate organic car-
bon (POC), etc.] yet enrich the reactants for
bloom formation (Si, NO3
were used to calculate the resultant dilution of
the iron-enriched waters with the unenriched
waters surrounding each patch: (i) The physical
–, etc.). Two methods
Fig. 1. The Southern Ocean showing the waters with low (?5 ?M), intermediate (5 to 25 ?M), and
high (25 to 70 ?M) silicic acid concentrations together with the locations of the recent Southern
Ocean enrichment experiments (SOIREE, EisenEx, and SOFeX).
Fig. 2. (A) MODIS
ometer) image of the
north patch on day 28
showing an extremely
elongated and narrow
band of chlorophyll
extending over 250
km. (B) SeaWIFS (Sea-
Viewing Wide Field-
of-View Sensor) im-
age of the south patch
on day 20 of the south
R E S E A R C H A R T I C L E S
www.sciencemag.org SCIENCEVOL 304 16 APRIL 2004
strain was calculated from satellite images of
patch deformation (24). (ii) The loss of SF6,
injected as a conservative tracer and corrected
for outgassing, was used to calculate dilution.
Both methods yielded similar results. The rate
of dilution of the north patch was 0.11 day?1
(strain) and 0.10 day?1(SF6). Dilution esti-
mates for the south patch were 0.086 day?1
(strain) and 0.03 day?1to 0.07 day?1(SF6).
SF6-derived mixing rates varied because of inho-
mongeneities within the patch and analytical vari-
ability. Dilution rates of 0.11 and 0.08 were used
for north and south patches, respectively. These
rates are substantially lower than those observed
during SOIREE (0.07 day?1) (24).
Dilution of the patch with unenriched wa-
ters will reduce the magnitude of geochemical
signals that are produced inside the patch as a
result of iron-stimulated phytoplankton growth.
For a dilution rate of 0.1 day?1, the initial
concentration of an inert tracer is reduced by
some 20-fold over the course of 30 days, cor-
responding to a mixing loss of 0.1 day?1. How-
ever, dilution has a much smaller impact on a
geochemical signal such as the accumulation of
POC, which is produced by an exponentially
increasing phytoplankton population (fig. S2).
Most of that geochemical signal will be pro-
duced near the end of the 30-day period, be-
cause the phytoplankton population will be
largest at that time. As a result, a simple nu-
merical simulation of the SOFeX patches,
based on the observed rates of growth and
mixing, indicates that a 0.1 day?1dilution rate
reduced geochemical signals by two to four
times. Further, dilution will act on all geo-
chemical signals in about the same manner.
Ratios that are determined from the temporal
change in chemical concentrations remain un-
changed. However, dilution of a patch may
have a much larger impact on physiological
processes that are operating in a near-limiting
condition. Horizontal dilution of the northern
patch appeared to be a major source of silicic
acid that sustained diatom growth (see below).
This effect is analogous to silicic acid supply
from vertical mixing and upwelling, the domi-
nant supply over much larger areas. Except in
Table 1, we report all geochemical changes
without correction for dilution.
After each addition of iron, its concentration
decreased (Fig. 3, A and B) at a rate that was
nearly equal to that of SF6. The SF6concentra-
tion decreased because of dilution and because
of gas exchange with the atmosphere (?0.07
day?1). Given the similarity of the total loss
rates for each chemical, dissolved iron must
also have had additional sinks other than dilu-
tion, which are presumably biological uptake
and scavenging. These processes must have
occurred at a rate of ?0.05 day?1.
The biological response. Before iron re-
lease, photosynthetic competency [Fv/Fm(25)]
was low in both north and south patches (0.20
and 0.25, respectively), indicating severe nutri-
ent limitation of phytoplankton photosynthesis.
After iron enrichment, Fv/Fmincreased to 0.5
indicating that the photosynthetic community
was relieved from iron limitation (26) (fig. S3).
Maximum rates of photosynthesis (Popt) also
increased with iron addition, from 0.29 to 6.9
0.29 to 4.6 mmol C m?3day?1in the south
patch (Fig. 4). The increase in photosynthetic
than in IronEx I and II (2 to 3 hours) (27), but
similar to the response measured during the
SOIREE experiment (2 to 3 days) (28) and
consistent with temperature being the main fac-
tor driving this difference (29). Both iron en-
richment and temperature effects were apparent
in phytoplankton community growth rates (30),
day?1to ?0.3 day?1in the south patch and
from 0.3 day?1to ?0.5 day?1by day 12 in the
north patch. During SOIREE, phytoplankton
growth rates increased similarly from 0.08
day?1to 0.20 day?1(31), and the relative
magnitude of the increase was comparable to
those observed in IronEx II (32).
Although phytoplankton biomass outside
the patches remained roughly constant through-
out the observational period, Chl a concentra-
tion in the euphotic zone increased by factors of
10 to 20 inside north and south patches, respec-
tively (Fig. 3, C and D, and fig. S4). Both
patches were easily observed from space by
ocean color satellite sensors when clouds per-
mitted (Fig. 2, A, day 28 north patch, and B,
day 20 south patch). Such conditions only pre-
vailed for 1 day during the south experiment,
whereas a composite image of the northern
patch could be constructed from multiple satel-
lite passes. The response of the community
Fig. 3. (A to L) Evolu-
tion of major geo-
chemical signals (iron,
chlorophyll, Si, nitrate,
NTCO2, and POC) in
north (left) and south
(right) patch experi-
mental sites as a func-
tion of day of experi-
ment. The lines in (A)
and (B) refer to the ex-
pected dissolved iron
concentrations based on
dilution as determined
by SF6. Multiple addi-
tions explain the saw-
tooth behavior of these
curves. Whereas good
sured iron exists imme-
diately after most in-
jections, the difference
between expected and
observed reflects parti-
cle formation and bio-
logical uptake. The de-
tails of the injections
can be found in tables
S1 and S2. Gray dots in
(G) are continuous ni-
made with pumping
SeaSoar in transects
across the patch. Gray
dots in (I) and (J) are
normalized (to salinity
35), and TCO2values
were determined from
served alkalinity values
in the north and south
patches, respectively. It
should be noted that
RV Melville reoccupation of the north patch (last data points) encountered only slightly elevated values
of SF6and thus these latter points do not reflect maximum “inside” stations. These latter values do not,
therefore, reflect a decrease in biological activity. Open circles, outside patch; solid circles, inside patch.
R E S E A R C H A R T I C L E S
16 APRIL 2004 VOL 304SCIENCE www.sciencemag.org
structure in the northern patch was toward larg-
er cells (?5 ?m), whereas the size structure of
the phytoplankton community remained un-
changed in the south. In the north patch, en-
hanced growth was evident for flagellated
phytoplankton groups (prymnesiophytes, pel-
agophytes, and dinoflagellates) as well as the
diatom Pseudonitzschia spp. By day 38, di-
atom biomass showed the largest change rela-
tive to initial values but represented somewhat
less than half of the total (fig. S5).
The south patch was dominated (at both
inside and outside stations) by diatoms, and Chl
a increased by a factor of 20. Despite high
silicic acid concentrations in the south patch
(Fig. 3F), the diatoms in both patches were
thin-walled, poorly silicified, and therefore,
lightly ballasted (14). Moreover, on the basis of
single-cell, x-ray fluorescence analysis (33), the
silicic acid cell quotas of diatoms in the south
patch decreased about 50% after the initial iron
addition and remained lower than silicic acid
quotas in diatoms outside of the patch. Whether
this corresponds to an adjustment of frustule
thickness by individual species or a shift to
species with inherently thinner frustules is pres-
ently unknown. A shift to lower silicic acid
high iron (34). In contrast, iron-limited diatoms
can generally increase silicification, perhaps to
facilitate access to higher iron concentrations in
deeper water (34, 35).
Phytoplankton growth using nitrate requires
greater cellular iron requirements as compared
to growth on more reduced forms of nitrogen
(36). Alleviation of iron deficiency after fertil-
ization resulted in enhanced rates of nitrate
uptake at both sites. Absolute rates of nitrate
patch and ?25 in the south patch relative to
outside control regions. Addition of Fe also
increased biomass-specific NO3
by factors of 5 and 10 in the north and south
patches, respectively (Fig. 5, A and B), a result
similar to those reported for iron-amended bot-
tle experiments conducted previously in the
Southern Ocean (37, 38).
There were clear decreases (?2 ?M) in the
the north and south patches (Fig. 3, G to H).
The proportion of nitrate to total nitrogen
measured during day-long15N incubations as a
proxy for new production. This value (f ratio)
increased from a range of 0.1 to 0.2 to a ratio of
0.5 to 0.6 in the southern patch and 0.3 to 0.4 in
the northern patch, clearly indicating that suffi-
cient iron allows for the greater use of nitrate
reserves (Fig. 5, A and B).
Absolute rates of silicic acid uptake increased
of 4 in the south patch. Specific uptake rates
ic rates observed near the end of each experiment
?, and urea) uptake was
imply maximum doubling times for diatoms in
those for the south patch (1 day), despite the
significantly warmer surface waters in the north,
possibly because of Si limitation in the low-Si
waters of the subantarctic.
The addition of iron caused a steep decrease
in silicic acid:nitrate uptake ratios in the dia-
tom-dominated south patch, from an average of
8.1 ? 1.5 outside the patch to 2.1 ? 0.5 inside
the patch, consistent with observations that re-
lease from iron stress shifts silicic acid–to–
Fig. 4. Primary production in the north and south patches. (A) Depth profiles indicating differences between
inside and outside stations. For north patch black symbols, initial pre and final out stations were sampled on
patch days –1.7 and 27, respectively; for north patch red symbols, earliest in, subsequent in (s), and final in
out were sampled on patch days 9 and 32; for south patch blue symbols, earliest in, subsequent in (s), and
final in stations were sampled on patch days 14, 17, 19, 27, and 33, respectively d, day. (B) All productivity
measurements are relative to the values obtained during the JGOFS AESOPS study, indicating iron enrich-
ment enhances primary production beyond values normally seen for any time of year. Poptis the maximal
north in; red open circles, north out; blue solid circles, south in; blue open circles, south out. Ed(0–) is the
downwelling irradiance just below the surface of the ocean.
Table 1. Concentration changes observed within the experimental sites relative to values observed outside
each patch on day 30. South patch values were calculated from the difference in regression lines fit to inside
and outside patch measurements at patch day 30 (Fig. 3). North patch values were calculated from the
difference in measurements made inside and outside the patch on patch day 28. DIC:NO3
from the TCO2/NO3
in carbon due to vertical export of POC or accumulation of DOC and were estimated from the model increase
of POC and an export ratio of 0.5 (Fig. 5) for north and south patches. Units are in ?g l?1for Chl a, ?mol kg?1
for TCO2, and ?M for all other values except DIC:NO3
confidence levels. Values in parentheses represent estimated dilution corrected values as per the model
presented in table S2. For the dilution-corrected Chl a values, a net growth rate of 0.18 day?1(versus 0.11
day?1for POC) was used in the north patch, and a net growth rate of 0.16 day?1(versus 0.08 day?1for POC)
was used in the south patch. GE indicates gas exchange.
–ratios are calculated
–depletion. Correction for gas exchange was not made. The balance terms reflect changes
–, for which the units are mol mol–1. Error limits are 90%
North patchSouth patch
Silicic acid depletion
Biogenic silica accumulation
Balance (–TCO2– POC ? GE)
–15 ? 2 (–60)
–2.0 ? 0.2 (–12)
7.7 ? 1.2 (9.0 ? 0.1)†
8.5 ? 0.8 (60)
9 ? 3 (32)
–16 ? 6 (–42)
–1.9 ? 0.7 (–10)
8.9 ? 5
11 ? 3 (42)
2 ? 1
7 ? 7 (21)
–molar depletion ratio
*There was not this much initial silicic acid present at the north patch, yet dilution with outside waters provided a
the upper 30 m with the pumping SeaSoar on patch day 28 (Fig. 6).
–value in parentheses for the north patch was derived from measurements made in
R E S E A R C H A R T I C L E S
www.sciencemag.org SCIENCEVOL 30416 APRIL 2004
nitrate uptake ratios toward values closer to
Redfield proportions (34, 39). In the north patch
where the phytoplankton was dominated by non-
diatoms, silicic acid:nitrate uptake ratios were
nearly identical inside and outside the patch
(0.86 ? 0.35 inside versus 1.0 ? 0.38 outside).
Although rates of silicic acid uptake increased
by a factor of 4 in the northern patch, the low
silicic acid concentrations in the north (Fig. 3E)
strongly limited silica production, which likely
diminished the diatom response to iron addition.
These results have their basis in shipboard exper-
iments in which diatoms received enrichments of
silicic acid and their subsequent uptake was mea-
sured (40) (Fig. 5, C and D). Whereas only one
patch silicic acid enrichments, the north patch
consistently showed 180 to 380% increases in
uptake when silicic acid was added, indicating
sufficient ambient silicic acid in the southern but
not the northern patch.
The demand for silicic acid in the north
patch appeared to be met by dilution of the
patch with outside waters by the end of the
experiment. Silicic acid uptake rates measured
directly in north patch waters on patch day 39
were 0.22 ?mol Si l?1day?1. A strain rate of
0.11 day?1and a Si concentration difference of
3 ?M outside and 0.8 ?M inside will add 0.24
?mol l?1day?1to the patch. Thus, biological
uptake of Si was nearly balanced by the amount
of Si added by mixing. It appears that biogenic
silica production in the north patch was regu-
lated by Si added by mixing with surrounding
waters. The supply of silicic acid via mixing
was thought to have contributed to the longev-
ity (?50 days) of the SOIREE bloom [(6),
based on satellite data], but in situ data were not
available to confirm this. Silicic acid limitation
the increase in biogenic silica in the north patch
were each less than half those observed in the
south patch (Table 1) despite similar initial
siliceous biomass in each area and the warmer
temperatures in the north.
Despite the low silicic acid concentrations,
patch were about double those in the south patch
(Fig. 4A). Much of this difference likely stems
from the higher temperatures in the north patch
and demonstrates that the low silicic acid concen-
trations do not necessarily limit primary produc-
iron addition to low silicic acid waters.
The photosynthetic and biomass response to
iron additions were compared with observations
from the U.S. Joint Global Ocean Flux Study
(JGOFS) Antarctic Environment and Southern
Ocean Process Study (AESOPS) that sampled
largely between the APFZ and the Southern Ant-
arctic Circumpolar Current Front along 170°W,
where blooms are more frequent (Fig. 4B). Phy-
toplankton production rates inside both
patches were clearly above the maximum
values observed during AESOPS (18), and
production continued to increase linearly
throughout the experiment. This would in-
dicate that conditions in the Southern
Ocean rarely exist naturally to promote the
maximum production that was observed un-
der iron-replete conditions.
The profiles of primary production in Fig. 4A
are consistent with the development of light lim-
itation also observed during EisenEx (7). The
highest primary production values are seen near
the surface. As the bloom developed, subsurface
production rates decreased to values lower than
ambient. Measurements of in situ irradiance indi-
cate that the diffuse attenuation of solar radiation
roughly doubled in each patch over the course of
the experiments, with the depth of the 1% light
level shoaling from 83 m to 32 m in the north (28
days) and from 58 m to 23 m in the south (33
of self-shading may slow production. Further-
more, the ratio of photoprotective to photosyn-
thetic carotenoids in the south patch decreased by
a factor of 3 during bloom development, suggest-
ing a transition from iron- toward light-limited
growth inside the patch (41).
ological processes driven by photosynthesis may
shift CO2from the atmosphere into the ocean by
four mechanisms: (i) increased air-to-sea CO2
flux from reduced pCO2because of increased
ing from an increase in the ratio of carbon to
the mixed-layer remineralization ratios whereby
nitrogen is regenerated more efficiently than car-
bon, and (iv) a change in the species composition
of the export community. Iron influenced all of
these mechanisms in our experiments.
were clearly depleted in both the north and south
patches (Fig. 3, I and J), as iron increased photo-
synthetic rates and nitrate consumption. The
direct measurements of TCO2in bottle samples
collected over the course of the experiment, was
16 ? 6 [90% confidence interval (CI)] ?mol
kg–1. Continuous measurements of pCO2from
the Revelle survey were combined with mea-
surements of titration alkalinity to estimate the
change in TCO2in the north patch of 15 ? 2
ing decrease in pCO2within both of the fertil-
ized patches was ?40 ?atm.
A budgeting approach can be used to esti-
mate export from both patches. After correction
for the flux of CO2across the air-sea interface,
the decrease in TCO2exceeds the accumulation
kg?1in the south patch (Table 1). The balance
increase in the dissolved organic carbon (DOC)
Fig. 5. (A and B) Depth profiles of
biomass (PN)-specific NO3
take rates (V), determined during
24-hour incubations in Plexiglas
(Rohm GmbH and Company Kg.
glass incubators under simulated
in situ light and temperature con-
ditions and with the use of ultra-
clean trace-metal techniques for
samples collected within and out-
side (control waters) of the Fe-
enriched patch north (A) and
south (B) of the APFZ. The ƒ
values [ƒ ? VNO3/(VNO3
VNH4? VNO2? VUrea) were de-
termined at the 47 and 16%
light depths with the use of
tracer-level isotopic enrichments
and are not corrected for the
effects of isotopic dilution. Error
bars represent the range of du-
plicate samples (n ? 2). Hatched
bars indicate inside stations;
open bars indicate outside sta-
tions. North patch inside station
was sampled on patch day 39;
outside station was sampled on
patch day 38. South patch inside
stations were sampled on patch
day 17; outside station was sam-
pled on patch day 18. (C and D)
Enhancement of the specific rate of biogenic silica production, Vb(40), by saturating additions of silicic acid
[?21 ?M Si(OH)4in the north patch (C) and ?42 ?M Si(OH)4in the south patch (D)]. Values ? 1 indicate
silicic acid uptake. Analytical and experimental uncertainties require that the enrichment ratio exceed 1.2 to
be significant. Solid fill denotes in stations; open, out patch stations. Circles denote experiments at the 47%,
and triangles, 16% light depths. The dashed line indicates a ratio of one.
R E S E A R C H A R T I C L E S
16 APRIL 2004VOL 304SCIENCEwww.sciencemag.org
pool or by vertical export via sinking particles
from the patch. Buesseler et al. (42) measured
increased POC export (?10 mmol m?2day?1)
from the southern patch at the end of the ex-
periment. This export corresponds to a loss of
?6 ?mol C kg?1from a 50-m mixed layer
over 30 days and accounts for our observed
imbalance between TCO2reduction and POC
production. Although insufficient POC and
DOC measurements for the north patch are
available from this experiment to resolve the
fate of the missing carbon at this time, aggre-
gates of the diatom Pseudonitzchia spp. and
colonial Phaeocystis spp. were observed by day
38 in the north patch, consistent with the po-
tential for significant carbon export. Further-
more, observations from an autonomous drifter
(43) indicate that a large flux event resulted
from iron enrichment in the north patch. Al-
though northern production and biomass were
dominated by nonsilicious phytoplankton, the
communities driving export production shifted
toward diatoms in the north but remained
diatom-dominated in the south.
ern Ocean suggest that ecosystems may export
carbon in greater-than-Redfield proportions
(44). The drawdown of TCO2in each iron-
fertilized patch exceeded values expected from
the observed amount of nitrate consumption
and a Redfield ratio of 6.6 (Table 1). The best
evidence for elevated C:N ratios of exported
particles was obtained during a survey of the
north patch with a pumping SeaSoar system
(45) on patch day 28. This survey shows a
dissolved inorganic carbon (DIC):NO3
down ratio of 9.0 ? 0.12 (95% CI, model II
6A), which is conservative because it was uncor-
rected for gas exchange. The DIC:NO3removal
ratio (6.9 ? 0.6) estimated from hydrographic
measurements in the Pacific sector of the South-
This is unexpected given that C:N ratios are usu-
ally observed to decline in populations recently
relieved from iron deficiency (38).
In contrast, the ratios of C:N in suspended
particles from the north patch and south patch
mixed-layer stations were 6 and 5.2, respective-
ly, values near Redfield and consistent with the
preferential nitrate uptake by iron-stimulated
phytoplankton (Fig. 6B). These ratios are con-
sistent with removal ratios observed elsewhere
yet are much lower than those observed in the
water column during both experiments. If par-
ticulate material is responsible for the majority
of the vertical flux of carbon from the mixed
layer, then these results strongly suggest that
shallow differential remineralization of nitro-
gen relative to carbon is the likely explanation.
Such differential remineralization has been ob-
served for nitrogen relative to silicon during
AESOPS experiments in this same area (46).
Synthesis. The SOFeX experiments pro-
duced several unexpected results: (i) Because of
silicic acid limitation, a strong differential re-
sponse to added iron was observed, with nonsili-
north and diatoms dominating in the south. (ii) In
radioactive proxies for flux (42), and free-vehicle
observations (43) indicate carbon flux from both
sites, suggesting a stronger role for iron-limited
carbon removal from these waters. (iii) Differen-
tial remineralization of nitrogen relative to carbon
predicted from available nitrate.
Relative to iron fertilizations in equatorial
waters (22), both north and south patch blooms
increase throughout the observational period.
This rate of biomass increase is consistent with
previous iron enrichments in Antarctic waters
(6, 7), indicating that temperature as well as
patch dilution may act together to slow biomass
accumulation. Only 10% of the nitrate available
for phytoplankton growth was consumed. In
contrast, phytoplankton consumed 20 ?M ni-
trate during an iron enrichment experiment in
the subarctic Pacific (16). In that experiment,
the mixed layer was only 10 m deep, with a
temperature of 9.5°C. We speculate that light
limitation produced by a combination of a deep
mixed layer (?40 m) and self-shading attenu-
ated the development of blooms in the north
and south patches. Much shallower mixed layers
(?20 m) were observed during the U.S. JGOFS
experiments near the south patch site (20), yet
seasonally and spatially variable mixed layers are
characteristic of these waters (47). Thus, it is
possible that much greater consumption of nitrate
can be achieved during periods with lower wind
speeds. Loss of iron does not appear to have
limited production. Iron concentrations in the
north patch were 0.2 nM above ambient concen-
trations on patch day 28 (23), which is consistent
with loss predominantly by mixing.
These results stand in contrast to some of
the recent JGOFS findings that indicate phyto-
plankton communities in the Southern Ocean
south of the APFZ are not limited by iron
availability (48). Further, some of us have re-
and carbon export. Low silicic acid concentra-
tions do appear to limit diatom growth in the
plays the dominant role in phytoplankton rate
processes in the vast subantarctic region north
of the APFZ. Low silicic acid, however, would
curtail the contribution of diatoms to new pro-
duction under prolonged iron enrichment. The
?1 ?M silicic acid concentrations in the north
patch would prevent diatoms from consuming
more than 5% of the remaining nitrate, given
the silicic acid:nitrate molar depletion ratio of
0.8:1 occurring in the north patch (Table 1).
Although differences in silicic acid concen-
trations may lead to differences in the export
community that develops under iron-replete
conditions, iron remains the proximal control
on total phytoplankton biomass. Both north and
south of the APFZ enhanced growth of the
larger phytoplankton taxa followed iron addi-
tion. Large blooms led to a decoupling between
production and grazing, resulting in a draw-
down of the major nutrients and carbon. How-
ever, the development of large blooms is inti-
mately linked to the mixed layer depth and light
limitation. Although the larger diatoms blooming
as a result of iron enrichment appear weakly si-
licified with C:N ratios slightly lower than Red-
field, differential remineralization of nitrogen
from the sinking particles resulted in a depletion
ratio from the water column much larger than
Redfield. Such a shift in DIC:NO3
aerosol iron input would have a significant effect
on atmospheric carbon dioxide amounts.
The average decreases in pCO2, corrected
–ratio over the
of TCO2and nitrate in the upper 30 m of the north
patch. TCO2was determined from continuous
pCO2measurements and observed alkalinity values
for the north and south patches, respectively, nor-
through the pumping SeaSoar was analyzed on
board Revelle. These measurements indicate that
non-Redfieldian drawdown ratios of carbon:nitro-
gen (9.0 ? 0.12) could have supported greater ex-
port than expected on the basis of the depletion of
nitrate alone. (B) Carbon (C) versus nitrogen (N) in
suspended particulate material in the mixed layers
from both enrichment experiments. North: C ?
5.98, N ? 0.11, and R2? 0.9819. South: C ? 5.19,
N ? 0.24, and R2? 0.9899.
R E S E A R C H A R T I C L E S
www.sciencemag.org SCIENCE VOL 30416 APRIL 2004
glacial-scale enrichment would result in an air-
to-sea flux of about 4.6 mol C m?2year?1or
about 2 Pg C year?1over an oceanic area of
3.6 ? 107km2(50°S to 65°S). Seasonal ice
cover may reduce this estimate by a factor of 2.
These estimates are several times larger than
comparable to the current global ocean net up-
take of atmospheric CO2. Dilution-corrected
estimates of POC export (Table 1) extrapolated
to an annual basis suggests a similar flux on the
order of 8 mol C m?2year?1.
These results demonstrate that iron addition
to the Southern Ocean increases primary pro-
ductivity and decreases pCO2. It remains
difficult to extrapolate these findings with con-
fidence to their impact on atmospheric compo-
sition because the large-scale impacts of iron
enrichment on midwater processes and the
length scales of POC remineralization are not
yet known. The results strongly suggest, how-
ever, that the Southern Ocean was more pro-
ductive and exported more carbon during peri-
ods of higher atmospheric iron input, which
occurred during the last glacial maximum.
References and Notes
1. T. Takahashi et al., Deep-Sea Res. Part II Top. Stud.
Oceanogr. 49, 1601 (2002).
2. K. L. Daly et al., J. Geophys. Res. 106, 7107 (2001).
3. D. M. Sigman, E. A. Boyle, Nature 407, 859 (2000).
4. K. S. Johnson et al., Mar. Chem. 57, 137 (1997).
5. S. E. Fitzwater et al., Deep-Sea Res. Part II Top. Stud.
Oceanogr. 47, 3159 (2000).
6. P. W. Boyd et al., Nature 407, 695 (2000).
7. F. Gervais et al., Limnol. Oceanogr. 47, 1324 (2002).
8. J. H. Martin, Paleoceanography 5, 1 (1990).
The Effects of Iron Fertilization
on Carbon Sequestration in the
9. D. Archer et al., Rev. Geophys. 38, 159 (2000).
10. J. K. Moore et al., Global Biogeochem. Cycles 14, 455
11. A. J. Watson et al., Nature 407, 730 (2000).
12. R. F. Anderson, Z. Chase, M. Q. Fleisher, J. Sachs, Deep-
Sea Res. Part II Top. Stud. Oceanogr. 49, 1909 (2002).
13. J. R. Petit et al., Nature 399, 429 (1999).
14. R. A. Armstrong et al., Deep-Sea Res. Part II Top. Stud.
Oceanogr. 49, 2265 (2002).
15. K. H. Coale et al., Nature 383, 495 (1996).
16. A. Tsuda et al., Science 300, 958 (2003).
17. R. Schlitzer, EOS 81, 45 (2000).
18. M. R. Hiscock et al., Deep-Sea Res. II 50, 533 (2003).
19. T. Trull et al., Deep-Sea Res. Part II Top. Stud. Oceanogr.
48, 2439 (2001).
20. K. H. Coale et al., Deep-Sea Res. Part II Top. Stud.
Oceanogr. 45, 919 (1998).
21. Iron enrichments were conducted by R/V Revelle. The
northern patch was deployed 10 to 12 January 2002 at
56.23°S and 172°W by injecting 631 kg of iron (as acidic
patch of 631 kg and 450 kg were repeated on 16 January
and 10 February, respectively. The southern patch was
deployed 24 to 26 January at 66.45°S and 171.8°W, with
repeated infusions on 29 January, 1 February, and 5 Feb-
area. For both patches, initial iron infusions were supple-
mented with infusions of SF6and3He as inert chemical
and provided estimates of mixing and gas exchange with
the atmosphere. Lagrangian drifter buoys were deployed
both inside and outside the enriched areas.
22. K. H. Coale, X. Wang, S. J. Tanner, K. S. Johnson, Deep-
Sea Res. Part II Top. Stud. Oceanogr. 50, 635 (2003).
23. K. Johnson et al., EOS 83 (suppl.), F799 (2002).
24. E. R. Abraham et al., Nature 407, 727 (2000).
25. P. G. Falkowski, J. A. Raven, Aquatic Photosynthesis
(Blackwell Scientific, Malden, MA, 1997).
26. M. Behrenfeld et al., Nature 371, 508 (1996).
27. Z. Kolber et al., Nature 371, 145 (1994).
28. P. W. Boyd, E. R. Abraham, Deep-Sea Res. Part II Top.
Stud. Oceanogr. 48, 2529 (2001).
29. P. W. Boyd, Deep-Sea Res. Part II Top. Stud. Ocean-
ogr. 49, 1803 (2002).
30. M. R. Landry, R. P. Hassett, Mar. Biol. 67, 283 (1982).
Res. Part II Top. Stud. Oceanogr. 48, 2571 (2001).
32. M. R. Landry et al., Mar. Ecol. Prog. Ser. 201, 57 (2000).
33. B. S. Twining et al., Anal. Chem. 75, 3806 (2003).
34. S. Takeda, Nature 393, 774 (1998).
36. M. T. Maldonado, N. M. Price, Mar. Ecol. Prog. Ser.
141, 161 (1996).
37. K. R. Timmermans et al., Mar. Ecol. Prog. Ser. 166, 27
38. W. P. Cochlan, D. A. Bronk, K. H. Coale, Deep-Sea Res.
Part II Top. Stud. Oceanogr. 49, 3365 (2002).
39. D. A. Hutchins, K. W. Bruland, Nature 393, 561 (1998).
40. M. A. Brzezinski, D. R. Phillips, Limnol. Oceanogr. 42,
41. R. R. Bidigare et al., EOS 83 (suppl.), F798 (2002).
42. K. O. Buesseler et al., Science 304, 414 (2004).
43. J. K. B. Bishop et al., Science 304, 417 (2004).
44. C. Sweeney et al., Deep-Sea Res. Part II Top. Stud.
Oceanogr. 47, 3395 (2000).
45. B. Hales, T. Takahashi, J. Atmos. Ocean. Technol. 19,
46. W. O. Smith et al., Deep-Sea Res. Part II Top. Stud.
Oceanogr. 47, 3073 (2000).
47. S. Levitus, U.S. World Ocean Atlas (U.S. Department
of Commerce, 1998).
48. R. J. Olson et al., Deep-Sea Res. Part II Top. Stud.
Oceanogr. 47, 3181 (2000).
ogr. 47, 3315 (2000).
50. We wish to thank the entire SOFeX group, crew, and
officers of RV Revelle, RV Melville, and U.S. Coast
Guard Research Ice Breaker Polar Star and two very
thorough reviewers. This research was supported by
grants from NSF, Chemical and Biological Oceanog-
raphy, and U.S. Department of Energy, Office of
Science, Biological and Environmental Research.
Supporting Online Material
Materials and Methods
Figs. S1 to S5
Tables S1 and S2
29 July 2003; accepted 25 March 2004
Ken O. Buesseler,* John E. Andrews,
Steven M. Pike, Matthew A. Charette
An unresolved issue in ocean and climate sciences is whether changes to the surface
ocean input of the micronutrient iron can alter the flux of carbon to the deep ocean.
During the Southern Ocean Iron Experiment, we measured an increase in the flux of
particulate carbon from the surface mixed layer, as well as changes in particle cycling
of natural blooms in the Southern Ocean and thus small relative to global
carbon budgets and proposed geoengineering plans to sequester atmospheric
carbon dioxide in the deep sea.
As the largest high nutrient–low chlorophyll
region, the Southern Ocean was chosen for a
SOFeX (Southern Ocean Iron Experiment).
The experiment was conducted at two sites
both north and south of the Antarctic Polar
Front, in low- and high-silicate waters, re-
spectively. We focus here on the “southern
patch” where the inert tracer SF6and four
enrichments of iron were added to a 15 km by
15 km patch (66°S, 172°W), which was
tracked and monitored by three ships in a
Lagrangian fashion for 1 month in January to
February 2002. As in previous experiments
(1–3), the addition of the essential micronu-
trient iron led to measurable decreases in
dissolved inorganic carbon and nutrients
within the surface mixed layer (upper 40 to
50 m) associated with enhanced growth of
marine phytoplankton, the details of which
are described in the accompanying article by
Coale et al. (4).
Department of Marine Chemistry and Geochemistry,
Woods Hole Oceanographic Institution, Woods Hole,
MA 02543, USA.
*To whom correspondence should be addressed. E-
R E S E A R C H A R T I C L E S
16 APRIL 2004VOL 304SCIENCEwww.sciencemag.org