R ES E A RC H

CLIMATE CHANGE ly initiating coastal upwelling along Antarctica,

in areas with downwelling today (Fig. 1 and figs. S1
and S2). Ocean heat content and stratification
Sustained climate warming increase globally, and deep mixing in the North
Atlantic collapses, reducing North Atlantic Deep

drives declining marine Water (NADW) formation from 30 to 5 sverdrups

by 2200 (1 sverdrup = 106 m3 s−1) (21, 22). Global
sea surface temperature and stratification (0 to
biological productivity 500 m) peak by 2200 (22). Deeper down, the ocean
is still warming in 2300, increasing stratification
J. Keith Moore,1* Weiwei Fu,1* Francois Primeau,1 Gregory L. Britten,1
between intermediate depths and the deep ocean.
Density differences between 500 and 1500 m and
Keith Lindsay,2 Matthew Long,2 Scott C. Doney,3 Natalie Mahowald,4
between 1000 and 2000 m more than double pre-
Forrest Hoffman,5 James T. Randerson1
industrial differences by 2300, with most (>80%)
of the change occurring after 2100 (table S1).
Climate change projections to the year 2100 may miss physical-biogeochemical feedbacks
Vertical exchange with cold, deep waters contrib-
that emerge later from the cumulative effects of climate warming. In a coupled climate
utes to a slower warming trend in the Southern
simulation to the year 2300, the westerly winds strengthen and shift poleward, surface
Ocean at depth, though surface waters warm
waters warm, and sea ice disappears, leading to intense nutrient trapping in the Southern
considerably (Fig. 1 and fig. S4).
Ocean. The trapping drives a global-scale nutrient redistribution, with net transfer to the
Biological export in the Southern Ocean in-
deep ocean. Ensuing surface nutrient reductions north of 30°S drive steady declines in
creases by 2100 but declines at lower latitudes
primary production and carbon export (decreases of 24 and 41%, respectively, by 2300).

Downloaded from http://science.sciencemag.org/ on March 8, 2018

and in the high-latitude North Atlantic (table S1).
Potential fishery yields, constrained by lower–trophic-level productivity, decrease by
Both patterns intensify after 2100, leading to nu-
more than 20% globally and by nearly 60% in the North Atlantic. Continued high levels of
trient trapping, with subsurface nutrient concen-
greenhouse gas emissions could suppress marine biological productivity for a millennium.
trations near Antarctica increasing substantially

T
by 2300 (Fig. 2, fig. S3, and table S2). Concentra-
he Southern Ocean strongly influences increases in Southern Ocean productivity, can tions of macronutrients (phosphate, nitrate, and
Earth’s climate and biogeochemistry (1, 2). develop nutrient trapping that boosts Southern silicic acid) decrease in the northward-subducting
Deep ocean waters upwell to the surface at Ocean nutrient concentrations and decreases waters, decreasing thermocline concentrations
the Antarctic Divergence. Subantarctic Mode northward lateral nutrient transport, reducing and depressing low-latitude NPP and export
and Antarctic Intermediate waters form as subsurface nutrient concentrations and decreas- (Fig. 2, figs. S3 and S4, and tables S1 to S3). Low-
northward-drifting surface waters sink and con- ing biological productivity at low latitudes (9–16). latitude productivity steadily drops as stratifica-
tinue northward at mid-depths, transporting Southern Ocean nutrient trapping, modulated tion increases and both surface and subsurface
nutrients into the low-latitude thermocline. The by circulation, can potentially transfer nutrients nutrient concentrations decline (Fig. 2, fig. S3,
Southern Ocean increasingly dominates ocean from the upper ocean to the deep ocean (12, 15). and tables S1 and S3). The equatorial upwelling
uptake of heat and CO2 with strong climate warm- We found intense Southern Ocean nutrient flux of phosphate declines sharply (41%), even
ing because of a poleward shift and intensifica- trapping as a result of climate warming in a fully though the mean equatorial upwelling rate de-
tion of the mid-latitude westerly winds (3–5). coupled simulation to the year 2300, with the clines modestly (3%). The sharp drop in nutrient
Earth system models (ESMs) in the fifth phase Community Earth System Model forced with rep- flux is due to the decrease in subsurface nutrients
of the Coupled Model Intercomparison Project resentative concentration pathway 8.5 (RCP8.5) after 2100, driven by reduced lateral transport
(CMIP5) show consistent declines in global ma- and extended concentration pathway 8.5 scenar- from the Southern Ocean (Fig. 3, fig. S3, and
rine net primary production (NPP) during the ios. The prescribed atmospheric CO2 concentra- tables S1 and S3).
21st century in scenarios with high fossil fuel tions increase to 1960 parts per million by 2250, The nutrients stripped out of surface waters
emissions, often with increasing Southern Ocean before leveling off (17, 18). We previously used by enhanced productivity in the Southern Ocean
NPP (4, 6–8). this ESM to examine marine biogeochemistry to are redistributed through the deep ocean by large-
Biological export of organic matter transfers the year 2100 (19–21) and century-by-century scale circulation (Fig. 4). Nutrient concentrations
nutrients vertically as sinking particles decom- changes in the climate-carbon feedback to 2300 steadily increase in the Southern Ocean and the
pose, releasing nutrients. Where surface currents (22). Southern Ocean nutrient trapping has not global deep ocean after 2100, while declining
diverge (and subsurface currents converge), been simulated previously without imposed NPP everywhere to the north, from the surface down
nutrients are transported upward, but some of increases (arbitrarily modifying biological or phys- to the depth of Antarctic Intermediate Water
the nutrients subsequently rain down as a result ical forcings). In our simulation, nutrient trapp- (~1500 m) (Fig. 4). We found similar global re-
of biological export, instead of being advected ing develops naturally after centuries of climate distribution patterns for nitrate and silicic acid
away laterally at the surface. If the time scale warming. This nutrient trapping drives a global but not for iron (figs. S5 to S7), because iron is
for downward transfer by sinking particles is reorganization of nutrient distributions, with a removed on time scales too short to permit long-
fast relative to the flushing time, nutrients be- net transfer to the deep ocean, leading to a steady range transport (23, 24) (supplementary materials).
come trapped, increasing concentrations locally decline in global-scale marine biological produc- Three distinct processes drive the transfer of
and reducing lateral transport of nutrients out of tivity. This climate-biogeochemistry interaction nutrients to the deep ocean. First, Southern Ocean
the area. Idealized model studies, with imposed amplifies the declines in productivity due to in- nutrient trapping lowers the nutrient flux from
creasing stratification projected previously for the deep ocean to the upper ocean within the
1
Department of Earth System Science, University of California, the 21st century, and its negative effects on pro- northward-subducting Antarctic Intermediate
Irvine, CA, USA. 2Climate and Global Dynamics Division,
Natural Center for Atmospheric Research, Boulder, CO, USA.
ductivity eventually exceed those of increasing and Subantarctic Mode waters. This is the pri-
3
Department of Environmental Sciences, University of stratification (6–8). mary pathway for nutrients to return to the upper
Virginia, Charlottesville, VA, USA. 4Department of Earth and The Southern Hemisphere westerly winds ocean (12, 15). Second, increasing stratification
Atmospheric Sciences, Cornell University, Ithaca, NY, USA. strengthen and shift poleward with climate warm- globally decreases vertical mixing and exchange
5
Oak Ridge National Laboratory, U.S. Department of Energy,
Oak Ridge, TN, USA.
ing, approaching Antarctica by 2300 (Fig. 1 and between the upper and deep ocean (table S1).
*Corresponding author. Email: jkmoore@uci.edu (J.K.M.); fig. S1). The Antarctic Divergence upwelling zone Third, reduced vertical mixing and reduced
weiweif@uci.edu (W.F.) also strengthens and shifts poleward (1), ultimate- NADW formation (21, 22) decrease the main

Moore et al., Science 359, 1139–1143 (2018) 9 March 2018 1 of 4

R ES E A RC H | R E PO R T

source of lower-nutrient waters to the deep ocean maximum phytoplankton growth rates by 52% concentrations increase by 34% south of 60°S by
(12), driving nutrient declines that are larger in (table S2 and supplementary materials). The 2300, with particularly high subsurface concen-
the Atlantic basin and high northern latitudes mean surface mixed-layer depth shoals with in- trations near Antarctica (table S2 and figs. S7 and
than in other regions (Fig. 4, figs. S5 to S7, and creasing stratification, declining from 75 m in 1850 S10). Volumetric upwelling rates remain ~25%
supplementary materials). to 40 m by 2300 (table S2). Sea ice cover reduces above preindustrial levels after 2150, and the up-
Increasing NPP in the Southern Ocean (south radiation to the contemporary Southern Ocean, welling phosphate flux follows this temporal pat-
of 60°S) is driven by the poleward shift of the but this shielding is weakened considerably by tern, boosted modestly by the nutrient trapping
westerlies, warming surface waters, and vanish- 2300 because the ice-covered area declines by effect (table S2 and supplementary materials).
ing sea ice, all of which enhance phytoplankton 96% (fig. S9 and tables S1 and S2). Mean light In contrast, the iron upwelling flux continues to
growth. Initially, the rise in NPP is driven by the levels in the surface mixed layer increase 245% rise to the year 2300, increasing 276% relative to
shifting westerlies, with upwelling rates increas- by 2300 as a consequence of near-complete sea preindustrial levels (Fig. 3 and table S2). This
ing to the year 2150 before leveling off (Fig. 3A). ice loss and shoaling mixed-layer depths (table S2 large increase is due to the southward shift in
Surface stratification intensifies to the year 2300, and supplementary materials). the upwelling zone, which entrains more margin-
driven by strong surface warming and decreases Iron availability modulates phytoplankton ca- influenced, high-iron waters, further boosting pro-
in surface salinity (Fig. 3, fig. S3, and table S2). pacity to take advantage of improving light and ductivity (Fig. 2; figs. S2, S3, and S9 to S12; and
The 6°C warming of polar surface waters increases temperature growth conditions. Subsurface iron supplementary materials).

Downloaded from http://science.sciencemag.org/ on March 8, 2018

Fig. 1. Climate change shifts surface winds and
warms sea surface temperatures. (A) Mean zonal wind
speed vectors (arrows) for the 1990s are plotted over
the mean sea surface temperatures. Colors represent
temperatures in degrees Celsius as indicated by the color
bar. (B and C) Changes in zonal wind speeds and sea
surface temperatures in comparisons of the 2090s with
the 1990s and the 2290s with the 1990s, respectively.
Colors in (C) represent changes in temperature
(in degrees Celsius) as indicated by the color bar.

Fig. 2. Climate change effects on

biological export and nutrient
distributions. Global maps of
(A) particulate organic carbon (POC)
flux at a depth of 100 m (expressed
in grams of carbon per square meter
per year) and (B) mean phosphate
concentrations at depths of 200
to 1000 m (expressed as micromolar
concentrations). The left column
shows 1990s means, the middle
column shows the difference
between the 2090s and the 1990s,
and the right column shows the
difference between the 2290s and
the 1990s. Phosphate concentrations
increase around the Antarctic by 2300 because of nutrient trapping.

Moore et al., Science 359, 1139–1143 (2018) 9 March 2018 2 of 4

R ES E A RC H | R E PO R T

Downloaded from http://science.sciencemag.org/ on March 8, 2018

Fig. 3. Regional time series of ocean physical forcing and biogeochemical response. Fig. 4. Southern Ocean nutrient trapping
(A) Evolution of biogeochemical and physical variables over time (1850 to 2300) for the Southern transfers phosphate to the deep ocean. Zonal
Ocean and the tropics. The mean upwelling rate increases in the Southern Ocean but slows in mean phosphate concentrations (micromolar)
the tropics, as surface stratification increases in both regions. (B) Upwelling rates for phosphate are shown for the 1990s (A), along with the
(expressed as the micromolar concentration per meter of upwelled water per day) and dissolved iron differences in zonal phosphate concentrations
(expressed as the nanomolar concentration per meter of upwelled water per day), (C) changes in in comparisons of the 1990s with the 2090s
temperature and salinity [in parts per thousand (ppt)], (D) surface (0 to 100 m) and intermediate- (B) and the 1990s with the 2290s (C). Phosphate
depth (200 to 1000 m) phosphate concentrations, and (E) sinking POC flux and NPP [both in concentrations increase over time in the high-
petagrams of carbon per year (Pg C/yr)]. The fractional sea ice cover is overlain in (C) (left column). latitude Southern Ocean (at all depths south
Upwelling flux was averaged for 10°S to 10°N for the tropical time series, and stratification was of 60°S) and in the global deep ocean (below
estimated from the density difference between the surface and 200 m for both regions. 2000 m and north of 60°S) (D).

Phytoplankton biomass and community com- stantial recycled production, changes in export in NPP (−15%) and export (−30%) by 2300 (figs.
position do not change greatly, but growth rates more directly reflect the decreased flux of nu- S12 and S13 and table S1). The declines above 30°S
increase and the growing season is longer, lead- trients to surface waters (8) (tables S1 to S3 and are 24% for NPP and 41% for particulate organic
ing to a doubling of the annual biological surface supplementary materials). The smallest phyto- carbon export (tables S1 to S3). The largest re-
phosphate drawdown by 2300 (table S2). Addi- plankton benefit most from increasing temper- ductions occur in the North Atlantic, western
tional productivity increases are possible with atures and nutrient depletion in surface waters Pacific, and southern Indian oceans, with zoo-
additional iron input, because surface phosphate (27), outcompeting larger phytoplankton and plankton productivity closely tracking phyto-
concentrations are far from depleted. One pos- reducing the efficiency of biological export (8, 28). plankton (figs. S12 and S13 and tables S1 to S4).
sible iron source (not included in our simulation) Loss of the sea ice biome at both poles will Production at higher trophic levels (including
is from Antarctic glaciers (25). The strong climate considerably modify biological communities, potential fishery yield) is limited by lower–trophic-
warming would greatly increase glacial discharge with reduced competitiveness and, potentially, level production and trophic transfer efficiency
(26), increasing iron inputs, allowing for even extinction of some polar-adapted, ice-dependent (30–32). To estimate how changing zooplankton
more efficient nutrient trapping, and modifying organisms, including the Antarctic krill central productivity influences higher trophic levels and
freshwater dynamic forcing of the oceans. to Southern Ocean food webs (29) (supplemen- maximum potential fishery yields, we used an
Marine food webs will shift with increasing nu- tary materials). empirical model with optimized transfer efficien-
trient stress outside the Southern Ocean. Relative Decreasing NPP and biological export north of cies, constrained by fishery data sets (32) (supple-
changes in biological export are larger than those 30°S more than offset productivity increases in mentary materials). We found that for higher
in NPP (table S1); because NPP can include sub- the Southern Ocean, driving global-scale declines trophic levels, production declines by more than

Moore et al., Science 359, 1139–1143 (2018) 9 March 2018 3 of 4

R ES E A RC H | R E PO R T

20% globally and by nearly 60% in the North peratures, and iron limitation depress biological 9. J. L. Sarmiento, N. Gruber, M. A. Brzezinski, J. P. Dunne, Nature
Atlantic (fig. S14 and table S5). Spatial patterns production in this region. Warming of surface 427, 56–60 (2004).
10. S. Dutkiewicz, M. Follows, P. Parekh, Global Biogeochem.
are similar to zooplankton production shifts but waters and removal of sea ice greatly improve Cycles 19, (2005).
are modified by decreasing transfer efficiency phytoplankton growth conditions, inducing 11. I. Marinov, A. Gnanadesikan, J. R. Toggweiler, J. L. Sarmiento,
where the warm-water biome expands (32) (figs. strong regional nutrient trapping that further Nature 441, 964–967 (2006).
12. I. Marinov et al., Global Biogeochem. Cycles 22, GB3007 (2008).
S13 to S15). boosts productivity. If the climate warms enough 13. A. Oschlies, W. Koeve, W. Rickels, K. Rehdanz, Biogeosciences
The processes leading to the climate-driven to remove Southern Ocean sea ice, the nutrient 7, 4017–4035 (2010).
transfer of nutrients to the deep ocean are scarce- redistribution and depression of global-scale 14. M. Holzer, F. W. Primeau, J. Geophys. Res. Oceans 118,
ly apparent by 2100 but dominate ocean bio- productivity outlined here seem inevitable. Thus, 1775–1796 (2013).
15. F. W. Primeau, M. Holzer, T. DeVries, J. Geophys. Res. Oceans
geochemistry by 2300. Little macronutrient the loss of Southern Ocean sea ice marks a crit- 118, 2547–2564 (2013).
accumulation occurs in the upper ocean around ical biogeochemical tipping point in the Earth 16. B. Bronselaer, L. Zanna, D. R. Munday, J. Lowe, Global
Antarctica by 2100 (Fig. 2 and figs. S3, S5, and system. More research is needed on the physical Biogeochem. Cycles 30, 844–858 (2016).
17. D. P. van Vuuren et al., Clim. Change 109, 5–31 (2011).
S6). This trend is consistent with the finding of forcings and the degree of warming necessary 18. Materials and methods are available as supplementary materials.
only a modest increase in biological carbon stor- for removal of southern sea ice cover and the ini- 19. K. Lindsay et al., J. Clim. 27, 8981–9005 (2014).
age in the Southern Ocean across multiple CMIP5 tiation of Southern Ocean nutrient trapping, in- 20. M. C. Long, K. Lindsay, S. Peacock, J. K. Moore, S. C. Doney,
models in 2100, driven by a slowdown of the deep cluding simulations with more moderate future J. Clim. 26, 6775–6800 (2013).
21. J. K. Moore, K. Lindsay, S. C. Doney, M. C. Long, K. Misumi,
circulation rather than by nutrient trapping (4). trajectories of fossil fuel emissions. Rapid southern J. Clim. 26, 9291–9312 (2013).
In 2300, tropical subsurface phosphate concen- sea ice decline begins in about the year 2050 in 22. J. T. Randerson et al., Global Biogeochem. Cycles 29, 744–759
tration is still declining and the deep ocean phos- our simulation, when mean surface air temper- (2015).
23. J. K. Moore, O. Braucher, Biogeosciences 5, 631–656 (2008).
phate level is still increasing linearly (Figs. 3 and ature has risen by ~2.5°C, with near-complete re- 24. A. Tagliabue et al., Global Biogeochem. Cycles 30, 149–174 (2016).
4). Thus, the transfer of nutrients to the deep moval of sea ice by 2200 (Fig. 3 and fig. S9). This 25. L. J. A. Gerringa et al., Deep-Sea Res. II 71, 16–31 (2012).

Downloaded from http://science.sciencemag.org/ on March 8, 2018

ocean will continue well past 2300, further de- sea ice trend and the biogeochemical responses 26. J. Hansen et al., PLOS ONE 8, e81648 (2013).
pressing upper-ocean nutrient concentrations documented here support emission reduction tar- 27. P. Flombaum et al., Proc. Natl. Acad. Sci. U.S.A. 110,
9824–9829 (2013).
and global-scale productivity. ESM simulations gets that keep climate warming below 2°C (26). 28. C. Laufkötter et al., Biogeosciences 13, 4023–4047 (2016).
must extend well past 2100 to capture key ocean The ongoing depletion of upper-ocean nutri- 29. A. J. Constable et al., Global Change Biol. 20, 3004–3025 (2014).
physical and biogeochemical responses to multi- ents in 2300 will not reverse until the climate 30. K. M. Brander, Proc. Natl. Acad. Sci. U.S.A. 104, 19709–19714
century climate warming. cools and sea ice returns to the Southern Ocean, (2007).
31. K. D. Friedland et al., PLOS ONE 7, e28945 (2012).
Southern Ocean nutrient trapping appears to depressing high-latitude productivity and enhanc- 32. C. A. Stock et al., Proc. Natl. Acad. Sci. U.S.A. 114,
be a robust response to multicentury warming. ing northward lateral nutrient transfer. Southern E1441–E1449 (2017).
All CMIP5 models project southward shift and in- Ocean dynamics will dominate this climate- 33. J. Caesar et al., J. Clim. 26, 3275–3284 (2013).
tensification of the westerlies over the 21st century cooling time scale, because the Southern Ocean 34. M. A. Giorgetta et al., J. Adv. Model. Earth Syst. 5, 572–597
(2013).
under the RCP8.5 scenario, despite considerable will dominate ocean uptake of CO2 by 2300 (figs. 35. S. Solomon, G. K. Plattner, R. Knutti, P. Friedlingstein,
differences in the structure and performance of S17 and S18 and supplementary materials). The Proc. Natl. Acad. Sci. U.S.A. 106, 1704–1709 (2009).
atmospheric models (5). Most CMIP5 models pre- long time scales associated with ocean uptake
dict increasing Southern Ocean export by 2100, and storage of anthropogenic CO2 (35) and the AC KNOWLED GME NTS

despite large differences in plankton and bio- subsequent time necessary for the circulation We received support from the Reducing Uncertainty in
Biogeochemical Interactions through Synthesis and Computation
geochemical models (4, 6–8, 24, 28). We found to return depleted nutrients to the upper ocean (RUBISCO) Scientific Focus Area (SFA) in the Regional and
that the same climate-driven nutrient redistri- (12) ensure that NPP will be depressed for a Global Climate Modeling Program in the Climate and Environmental
bution occurs in two other ESMs [the Hadley thousand years or more. This puts the climate Sciences Division of the Biological and Environmental Research
Centre Global Environmental Model, version 2 change impacts on marine biogeochemistry and (BER) Division of the U.S. Department of Energy (DOE) Office of
Science (as well as DOE BER Earth System Modeling Program grants
(HadGEM2) (33), and the Max-Planck-Institute productivity on the same time scale as continen- ER65358 and DE-SC0016539 to J.K.M. and F.P.). Some authors
Earth System Model (MPI-ESM) (34)] that con- tal ice sheets, with cumulative, catastrophic effects received additional support from the NSF. The Coupled Model
ducted RCP8.5 simulations to the year 2300 (fig. that will be increasingly difficult to avoid with de- Intercomparison Project received support from the World Climate
S16). Both models show nutrient increases in layed reductions in greenhouse gas emissions (26). Research Programme and the DOE Program for Climate Model
Diagnosis and Intercomparison. The National Center for Atmospheric
the Southern Ocean and in the deep ocean, ac- Research (NCAR) provided computational and other support.
companied by nutrient reductions in the upper RE FERENCES AND NOTES NCAR is sponsored by the NSF.
ocean. Thus, our results are not dependent on 1. J. L. Russell, K. W. Dixon, A. Gnanadesikan, R. J. Stouffer,
the model details of atmosphere-ocean circula- J. R. Toggweiler, J. Clim. 19, 6382–6390 (2006). SUPPLEMENTARY MATERIALS
tion, plankton dynamics, or biogeochemistry, but 2. T. L. Frölicher et al., J. Clim. 28, 862–886 (2015). www.sciencemag.org/content/359/6380/1139/suppl/DC1
3. T. Ito et al., Geophys. Res. Lett. 42, 4516–4522 (2015). Materials and Methods
rather seem to represent a robust Earth system 4. J. Hauck et al., Global Biogeochem. Cycles 29, 1451–1470 (2015). Supplementary Text
response to multicentury climate warming. 5. T. J. Bracegirdle et al., J. Geophys. Res. Atmos. 118, 547–562 Figs. S1 to S18
Relatively modest increases in export are suf- (2013). Tables S1 to S5
6. L. Bopp et al., Biogeosciences 10, 6225–6245 (2013). References (36–47)
ficient to induce nutrient trapping if they occur 7. A. Cabré, I. Marinov, S. Leung, Clim. Dyn. 45, 1253–1280 (2015).
in the critical location above the Antarctic Diver- 8. W. W. Fu, J. T. Randerson, J. K. Moore, Biogeosciences 13, 10 August 2017; accepted 5 February 2018
gence. Currently, heavy sea ice cover, cold tem- 5151–5170 (2016). 10.1126/science.aao6379

Moore et al., Science 359, 1139–1143 (2018) 9 March 2018 4 of 4

Sustained climate warming drives declining marine biological productivity
J. Keith Moore, Weiwei Fu, Francois Primeau, Gregory L. Britten, Keith Lindsay, Matthew Long, Scott C. Doney, Natalie
Mahowald, Forrest Hoffman and James T. Randerson

Science 359 (6380), 1139-1143.

DOI: 10.1126/science.aao6379

Starving ocean productivity

Projected increases in greenhouse gas emissions could suppress marine biological productivity for a thousand
years or more. As the climate warms, westerly winds in the Southern Hemisphere will strengthen and shift poleward,
surface waters will warm, and sea ice will disappear. Moore et al. suggest that one effect of these changes will be a
dramatic decrease in marine biological productivity (see the Perspective by Laufkötter and Gruber). This decrease will

Downloaded from http://science.sciencemag.org/ on March 8, 2018

result from a global-scale redistribution of nutrients, with a net transfer to the deep ocean. By 2300, this could drive
declines in fisheries yields by more than 20% globally and by nearly 60% in the North Atlantic.
Science, this issue p. 1139; see also p. 1103

ARTICLE TOOLS http://science.sciencemag.org/content/359/6380/1139

SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2018/03/07/359.6380.1139.DC1
MATERIALS

RELATED http://science.sciencemag.org/content/sci/359/6380/1103.full
CONTENT

REFERENCES This article cites 46 articles, 5 of which you can access for free
http://science.sciencemag.org/content/359/6380/1139#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of
Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive
licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. The title
Science is a registered trademark of AAAS.