A global map of human impact on marine ecosystems

Abstract

The management and conservation of the world's oceans require synthesis of spatial data on the distribution and intensity of human activities and the overlap of their impacts on marine ecosystems. We developed an ecosystem-specific, multiscale spatial model to synthesize 17 global data sets of anthropogenic drivers of ecological change for 20 marine ecosystems. Our analysis indicates that no area is unaffected by human influence and that a large fraction (41%) is strongly affected by multiple drivers. However, large areas of relatively little human impact remain, particularly near the poles. The analytical process and resulting maps provide flexible tools for regional and global efforts to allocate conservation resources; to implement ecosystem-based management; and to inform marine spatial planning, education, and basic research.

Full text

Page 1

DOI: 10.1126/science.1149345
, 948 (2008);
319
Science
et al.Benjamin S. Halpern,
Ecosystems
A Global Map of Human Impact on Marine

www.sciencemag.org (this information is current as of February 15, 2008 ):
The following resources related to this article are available online at
http://www.sciencemag.org/cgi/content/full/319/5865/948
version of this article at:
including high-resolution figures, can be found in the online
Updated information and services,
http://www.sciencemag.org/cgi/content/full/319/5865/948/DC1
can be found at:
Supporting Online Material
http://www.sciencemag.org/cgi/content/full/319/5865/948#otherarticles
, 10 of which can be accessed for free:
cites 28 articles
This article
http://www.sciencemag.org/cgi/collection/ecology
Ecology
:
subject collections
This article appears in the following
http://www.sciencemag.org/about/permissions.dtl
in whole or in part can be found at:
this article
permission to reproduce
of this article or about obtaining
reprints
Information about obtaining
registered trademark of AAAS.
is aScience2008 by the American Association for the Advancement of Science; all rights reserved. The title
CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience
on February 15, 2008
www.sciencemag.org
Downloaded from

Page 2

MBCDH1 and MBCDH2 lack the CCD and the
M.brevicollisgenomelacksab-cateninortholog.
MetazoanE-cadherinsandflamingocadherins
are bound by pathogenic bacteria which exploit
them as extracellular tethers during host cell in-
vasion(31–33).Itispossiblethatchoanoflagellate
cadherins fill an equivalent role in binding bacte-
rial prey for recognition or capture, functions con-
sistent with the enrichment of MBCDH1 and
MBCDH2 on the feeding collar (Fig. 2). If an-
cient cadherins bound bacteria in the unicellular
progenitor of choanoflagellates and metazoans,
cadherin-mediated cell adhesion in metazoans
may reflect the co-option of a class of proteins
whose earliest function was to interpret and re-
spond to cues from the extracellular milieu. In-
deed,thetransitiontomulticellularitylikelyrested
on the co-option of diverse transmembrane and
secreted proteins to new functions in intercellular
signaling and adhesion.
References and Notes
1. S. Tyler, Integr. Comp. Biol. 43, 55 (2003).
2. N. M. Brooke, P. W. Holland, Curr. Opin. Genet. Dev. 13,
599 (2003).
3. N. King, Dev. Cell 7, 313 (2004).
4. M. J. Wheelock, K. R. Johnson, Curr. Opin. Cell Biol. 15,
509 (2003).
5. R. A. Foty, M. S. Steinberg, Int. J. Dev. Biol. 48, 397 (2004).
6. W. J. Nelson, R. Nusse, Science 303, 1483 (2004).
7. B. M. Gumbiner, Nat. Rev. Mol. Cell Biol. 6, 622 (2005).
8. J. M. Halbleib, W. J. Nelson, Genes Dev. 20, 3199 (2006).
9. N. King, C. T. Hittinger, S. B. Carroll, Science 301, 361
(2003).
10. E. T. Steenkamp, J. Wright, S. L. Baldauf, Mol. Biol. Evol.
23, 93 (2006).
11. H. James-Clark, Annu. Mag. Natl. Hist. 1, 133–142;
188–215; 250–264 (1868).
12. D. J. Hibbert, J. Cell Sci. 17, 191 (1975).
13. G. Burger, L. Forget, Y. Zhu, M. W. Gray, B. F. Lang,
Proc. Natl. Acad. Sci. U.S.A. 100, 892 (2003).
14. D. V. Lavrov, L. Forget, M. Kelly, B. F. Lang, Mol. Biol. Evol.
22, 1231 (2005).
15. A. Rokas, D. Kruger, S. B. Carroll, Science 310, 1933 (2005).
16. E. Jimenez-Guri, H. Philippe, B. Okamura, P. W. Holland,
Science 317, 116 (2007).
17. Materials and methods are available as supporting
material on Science Online.
18. T. Tanoue, M. Takeichi, J. Cell Sci. 118, 2347 (2005).
19. M. Adamska et al., Curr. Biol. 17, R836 (2007).
20. L. E. Marengere, T. Pawson, J. Cell Sci. Suppl. 18, 97 (1994).
21. J. Gerhart, Teratology 60, 226 (1999).
22. C. Jamora, E. Fuchs, Nat. Cell Biol. 4, E101 (2002).
23. R. Kemler, Trends Genet. 9, 317 (1993).
24. B. M. Gumbiner, J. Cell Biol. 148, 399 (2000).
25. O. Sakarya et al., PLoS ONE 2, e506 (2007).
26. N. Jones et al., Nature 440, 818 (2006).
27. G. M. Rivera et al., Proc. Natl. Acad. Sci. U.S.A. 103,
9536 (2006).
28. Y. Segawa et al., Proc. Natl. Acad. Sci. U.S.A. 103, 12021
(2006).
29. M. Perez-Moreno, C. Jamora, E. Fuchs, Cell 112, 535
(2003).
30. R. W. Carthew, Curr. Opin. Genet. Dev. 15, 358 (2005).
31. J. Mengaud, H. Ohayon, P. Gounon, R. M. Mege, P. Cossart,
Cell 84, 923 (1996).
32. E. C. Boyle, B. B. Finlay, Curr. Opin. Cell Biol. 15, 633
(2003).
33. K. Blau et al., J. Infect. Dis. 195, 1828 (2007).
34. M. J. Telford, Curr. Biol. 16, R981 (2006).
35. Tree of Life Web Project, www.tolweb.org/tree
[accessed 10 December 2007].
36. I. Letunic et al., Nucleic Acids Res. 34, D257 (2006).
37. A. Bateman et al., Nucleic Acids Res. 32, D138 (2004).
38. We thank S. Nichols, J. Nelson, A. Rokas, N. Patel, C. Tabin,
and J. Reiter for critical reading of the manuscript and
J. Chapman, A. Morris, and our laboratory for technical
support and advice. Supported by the Gordon and
Betty Moore Foundation Marine Microbiology Initiative and
the Pew Scholars Program. N.K. is a Scholar in the Canadian
Institute for Advanced Research. The M. brevicollis genome
assembly and annotation data are deposited at DBJ/EMBL/
GenBank under the project accession ABFJ00000000. See
tables S2, S3, and S4 and notes S1 and S2 for nucleotides,
proteins, accession numbers, and sequences.
Supporting Online Material
www.sciencemag.org/cgi/content/full/319/5865/946/DC1
Materials and Methods
SOM Text
Figs. S1 to S4
Tables S1 to S5
References
28 September 2007; accepted 13 December 2007
10.1126/science.1151084
A Global Map of Human Impact on
Marine Ecosystems
Benjamin S. Halpern,1‡ Shaun Walbridge,1* Kimberly A. Selkoe,1,2*‡ Carrie V. Kappel,1
Fiorenza Micheli,3Caterina D’Agrosa,4† John F. Bruno,5Kenneth S. Casey,6Colin Ebert,1
Helen E. Fox,7Rod Fujita,8Dennis Heinemann,9Hunter S. Lenihan,10Elizabeth M. P. Madin,11
Matthew T. Perry,1Elizabeth R. Selig,6,12Mark Spalding,13Robert Steneck,14Reg Watson15
The management and conservation of the world’s oceans require synthesis of spatial data on the
distribution and intensity of human activities and the overlap of their impacts on marine
ecosystems. We developed an ecosystem-specific, multiscale spatial model to synthesize 17 global
data sets of anthropogenic drivers of ecological change for 20 marine ecosystems. Our analysis
indicates that no area is unaffected by human influence and that a large fraction (41%) is strongly
affected by multiple drivers. However, large areas of relatively little human impact remain,
particularly near the poles. The analytical process and resulting maps provide flexible tools for
regional and global efforts to allocate conservation resources; to implement ecosystem-based
management; and to inform marine spatial planning, education, and basic research.
H
the oceans through direct and indirect means
(1–5). Land-based activities affect the runoff of
pollutants and nutrients into coastal waters (6, 7)
and remove, alter, or destroy natural habitat.
Ocean-based activities extract resources, add
pollution, and change species composition (8).
These human activities vary in their intensity of
impact on the ecological condition of commu-
nities (9) and in their spatial distribution across
the seascape. Understanding and quantifying,
i.e., mapping, the spatial distribution of human
impacts is needed for the evaluation of trade-
offs (or compatibility) between human uses of
the oceans and protection of ecosystems and the
umans depend on ocean ecosystems for
important and valuable goods and ser-
vices, but human use has also altered
services they provide (1, 2, 10). Such mapping
will help improve and rationalize spatial man-
agement of human activities (11).
Determining the ecological impact of hu-
man activities on the oceans requires a method
for translating human activities into ecosystem-
specific impacts and spatial data for the activi-
ties and ecosystems. Past efforts to map human
impactsonterrestrialecosystems(12),coralreefs
(13), and coastal regions (14–16) used either
coarse categorical or ad hoc methods to translate
human activities into impacts. We developed a
standardized,quantitativemethod,onthebasisof
expert judgment, to estimate ecosystem-specific
differencesinimpactof17anthropogenicdrivers
of ecological change (table S1) (9). The results
provided impact weights (table S2) used to
Fig. 2. Subcellular local-
ization of MBCDH1 and
MBCDH2 (A and F),
compared with polymer-
ized actin stained with
rhodamine-phalloidin (B
and G), or antibodies
against b-tubulin (C and
H). Cells were exposed
to antibodies against
MBCDH1 after (A to E)
or before (F to J) perme-
abilization. Overlay of
MBCDH1 and MBCDH2
(green), actin (red), and b-tubulin (blue) reveals colocalization of MBCDH1 and MBCDH2 with actin
(yellow) on the collar and at the basal pole (D and I). Differential interference contrast microscopy shows
cell morphology (E and J). Brackets, collar of microvilli; arrow, apical organelle; arrowhead, basal pole;
asterisk, cluster of autofluorescent bacterial detritus.
JI
H
GF
****
EDBAC
15 FEBRUARY 2008VOL 319
SCIENCE
www.sciencemag.org
948
REPORTS
on February 15, 2008
www.sciencemag.org
Downloaded from

Page 3

combine multiple drivers into a single compara-
ble estimate of cumulative human impact on 20
ecosystem types (17). We focused on the current
estimated impact of humans on marine ecosys-
tems (within the last decade), as past impacts and
future scenarios of human impacts are less
tractable, though also important (17).
Predicted cumulative impact scores (IC)
were calculated for each 1 km2cell of ocean
as follows:IC¼ ∑
n
i¼1
∑
m
j¼1Di∗Ej∗mi;jwhere Diis
thelog-transformedandnormalizedvalue[scaled
between 0 and 1 (17)] of an anthropogenic driver
at location i, Ejis the presence or absence of
ecosystemj(either1or0,respectively),andmi,jis
the impact weight for the anthropogenic driver
i and ecosystem j [range 0 to 4 (table S2)], given
n = 17 drivers and m = 20 ecosystems (fig. S1).
We modeled the distribution of several inter-
tidal and shallow coastal ecosystems lacking
global data (17). Weighting anthropogenic
drivers by their estimated ecological impact in
this way resulted in a different picture of ocean
condition compared with simply mapping the
footprints of human activities or drivers (fig.
S1). Summing across ecosystems allows cells
withmultipleecosystemstohavehigherpotential
scores than areas with fewer ecosystems; sensi-
tivity analyses showed that summing or averag-
ing across ecosystems within cells resulted in
similar global pictures of human impacts on
marine ecosystems (17). Theglobal impact of a
particulardriver(ID)isID¼ ∑
n
i¼1Di∗Ej∗mi;jand
of all drivers on a particular ecosystem type (IE)
isIE¼ ∑
m
j¼1Di∗Ej∗mi;j. This method produced
ICscores ranging from 0.01 to 90.1. The IC
scores were significantly correlated with inde-
pendent estimates of ecological condition in 16
mixed-ecosystem regions containing coral reefs
(17, 18). The linear equation relating the two
scores [R2= 0.63, P = 0.001 (fig. S5)] was then
used to divide ICscores into six categories of
human impact ranging from very low impact
(IC< 1.4) to very high impact (IC> 15.5) (17).
Predicted human impact on the oceans shows
strong spatial heterogeneity (Fig. 1) with a
roughly bimodal distribution of per-cell ICscores
(Fig. 2), but with every square kilometer affected
by some anthropogenic driver of ecological
change. Over a third (41%) of the world’s oceans
have medium high to very high ICscores [>8.5
(17)], with a small fraction (0.5%) but relatively
large area (~2.2 million km2) experiencing very
high impact (IC> 15.5). Most of the highest
predicted cumulative impact is in areas of con-
tinentalshelfandslope,which are subjecttoboth
land- and ocean-based anthropogenic drivers.
Large areas of high predicted impact occur in the
NorthandNorwegianseas,SouthandEastChina
seas, Eastern Caribbean, North American eastern
seaboard, Mediterranean, Persian Gulf, Bering
Sea, and the waters around Sri Lanka (Fig. 1).
1National Center for Ecological Analysis and Synthesis, 735
State Street, Santa Barbara, CA 93101, USA.2Hawai’i Institute
of Marine Biology, Post Office Box 1346, Kane‘ohe, HI 96744,
USA.3Hopkins Marine Station, Stanford University, Oceanview
Boulevard, Pacific Grove, CA 93950–3094, USA.
Conservation Society, 2300 Southern Boulevard, Bronx, NY
10460, USA.5Department of Marine Sciences, University of
North Carolina at Chapel Hill, Chapel Hill, NC 27599–3300,
USA.6National Oceanographic Data Center, National Oceanic
and Atmospheric Administration (NOAA), 1315 East-West
Highway,SilverSpring,MD20910,USA.7ConservationScience
Program, World Wildlife Fund—United States, 1250 24th
Street NW, Washington, DC 20037, USA.
Defense,5655CollegeAvenue,Suite304,Oakland,CA,94618,
USA.9OceanConservancy,130019th Street,NW,Washington,
DC 20006, USA.10Bren School of Environmental Science and
Management, University of California, Santa Barbara, CA
93106, USA.11Department of Ecology, Evolution, and Marine
Biology, University of California, Santa Barbara, CA 93106,
USA.12Curriculum in Ecology, University of North Carolina at
ChapelHill,ChapelHill,NC27599–3275,USA.13Conservation
Strategies Division, the Nature Conservancy, 93 Centre Drive,
Newmarket, CB8 8AW, UK.
University of Maine, Darling Marine Center, Walpole, ME
04353, USA.15Fisheries Center, 2202 Main Mall, University of
British Columbia, Vancouver, V6T 1Z4, Canada.
*These authors contributed equally to this work.
†Present address: School of Life Sciences, Arizona State
University, Tempe, AZ 85287–4501, USA.
‡To whom correspondence should be addressed. E-mail:
halpern@nceas.ucsb.edu, selkoe@nceas.ucsb.edu
4Wildlife
8Environmental
14School of Marine Sciences,
Fig. 1. Global map (A)
of cumulative human
impact across 20 ocean
ecosystem types. (Insets)
Highly impacted regions
intheEasternCaribbean
(B), the North Sea (C),
and the Japanese waters
(D) and one of the least
impacted regions, in
northern Australia and
the Torres Strait (E).
www.sciencemag.org
SCIENCE
VOL 31915 FEBRUARY 2008
949
REPORTS
on February 15, 2008
www.sciencemag.org
Downloaded from

Page 4

Ecoregions, a classification of coastal (<200 m
depth) areas based on species composition and
biogeography (19), also showed variation in
scores indicating differential risks to unique
marine assemblages (table S3).
The majority of very low impact areas (3.7%
of the oceans) occurs in the high-latitude Arctic
and Antarctic poles (Fig. 1), in areas with sea-
sonal or permanent ice that limits human access.
However,ouranalysesdidnotaccountforillegal,
unregulated, and unreported (IUU) fishing,
which may be extensive in the Southern Ocean
(20), or atmospheric pollution, which may be
particularly high in the Arctic (21). Furthermore,
projections of future polar ice loss (22) suggest
that the impact on these regions will increase
substantially.In general, small human population
and coastal watershed size predict light human
impact(Fig.1E)butdonotensureit,asshipping,
fishing, and climate change affect even remote
locations—e.g., impact scores are relatively high
in the international waters of the Patagonian
shelf. In some places, predicted impact scores
may be higher than anticipated because many
anthropogenic drivers are not readily observable.
Conversely, impact scores may seem unexpect-
edly low in other locations because a more
abundant but less-sensitive ecosystem (e.g., soft
sediment) surrounds a sensitive, but rare, eco-
system (e.g., coral reefs).
Ecosystems with the highest predicted cumu-
lative impact scores include hard and soft con-
tinental shelves and rocky reefs (Fig. 3). Coral
reefs, seagrass beds, mangroves, rocky reefs and
shelves, and seamounts have few to no areas
remaining anywhere in the world with IC< 1.5
(Fig. 3). Indeed, our data suggest that almost half
of all coral reefs experience medium high to very
high impact (13, 17, 23). Shallow soft-bottom
and pelagic deep-water ecosystems had the
lowest scores (>50% of these ecosystems have
IC< 1.1 and 1.2, respectively), partly because of
the lower vulnerability of these ecosystems to
most anthropogenic drivers (table S2). Overall,
these results highlight the greater cumulative im-
pact of human activities on coastal ecosystems.
Perhaps not surprisingly, anthropogenic driv-
ers associated with global climate change are
distributed widely (Fig. 4A) and are an important
component of global cumulative impact scores,
particularly for offshore ecosystems. Other driv-
ers, in particular commercial fishing, are also
globally widespread but have smaller cumulative
impacts because of their uneven distribution.
Land-basedanthropogenicdrivershaverelatively
small spatial extents and predicted cumulative
impacts (Fig. 4A), but their cumulative impact
scores approach those of other more widespread
drivers within coastal areas where they occur
(Fig. 4B). The spatial distribution of land-based
impacts is highly heterogeneous but positively
spatially correlated. Therefore, management of
coastal waters must contend with multiple drivers
in concert. Coordination with regulating agencies
for urban and agricultural runoff is warranted,
although such efforts can be challenging when
watershedscrossjurisdictionalboundaries.Where
anthropogenic drivers tend to be spatially distinct
(uncorrelated), as with commercial shipping
versus pelagic high-bycatch fishing, management
will require independent regulation and conser-
vation tools. Assessing positive and negative spa-
tialcorrelations among drivers can help anticipate
potential interactions (24) and provides guidance
in adjusting spatial management accordingly.
Our approach may be used to identify re-
gions where better management of human activ-
itiescouldachieveahigherreturn-on-investment,
e.g., by reducing or eliminating anthropogenic
drivers with high impact scores (fig. S2). It may
also be used to assess whether or how human
activitiescanbespatiallymanagedtoreducetheir
negative impacts on ecosystems. For example,
fishing zones have been shifted to decrease im-
pacts on sensitive ecosystems (25), and naviga-
tion lanes have been rerouted to protect sensitive
areas of the ocean (26). Wide-ranging fish stocks
Fig. 2. Histogram of cumulative impact scores depicting the fraction of global area that falls within
each impact category. There are no zeros; histogram bars are in bins of 0.1. Categories described in
(17). (Insets) Expanded views of the tail of values.
Fig. 3. The distribution of
cumulativeimpactscoresforeach
ecosysteminouranalyses(means
in parentheses). Individual eco-
system scores have a smaller
range of values than cumulative
impactscores(Fig.2)becausethe
latter sum all ecosystem-specific
scores within a cell. Ground-
truthed estimates of coral reef
condition (17) were used to
identify IEvalues at which coral
reefs experience medium high to
very high impact, as indicated on
the coral reef histogram. Note
differences in y-axis scales.
15 FEBRUARY 2008VOL 319
SCIENCE
www.sciencemag.org
950
REPORTS
on February 15, 2008
www.sciencemag.org
Downloaded from

Page 5

and those that occur primarily in international
waters present challenges in determining who
must take responsibility for management. If
ecosystem-specific weighting values (mi,j) are
excluded, we can also evaluate the distribution,
or footprint, of summed anthropogenic drivers of
ecosystem change. This global footprint of driv-
ers correlates with the distribution of cumulative
impact scores (R2= 0.83), but ignores the
importantsmall-scalespatialpatternsthatemerge
when accounting for ecosystem vulnerability
(fig. S1) (17).
Our results represent the current best esti-
mate of the spatial variation in anthropogenic
impacts. Although these estimates are conserva-
tive and incomplete for most of the ocean, they
potentially inflate human impacts on coastal
areas because we used an additive model (17).
Averaging impacts across ecosystems produced
highly correlated results, very similar to those
from the additive model (17), which suggests
such inflation is limited, if it exists. Furthermore,
the large extent of the ocean that our model
predicts to be negatively affected by human
activities will likely increase once additional
drivers, their historical effects, and possible
synergisms are incorporated into the model.
Key activities with significant impacts on marine
ecosystems but without global data include
recreational fishing (27), aquaculture (28, 29),
disease (30), coastal engineering (habitat altera-
tion), and point-source pollution (31). Most of
these activities primarily affect intertidal and
nearshore ecosystems rather than offshore eco-
systems, which suggests that our estimates for
nearshore areas are particularly conservative. In
addition, the spatial data for many anthropogenic
drivers were derived from valid but inexact
modeling approaches (17). Ecosystem data were
highlyvariableinquality,bothwithinandamong
ecosystem types, and in many cases, we may
have underestimated the full extent of these
ecosystemsand,therefore,thecumulativeimpact
scores. Furthermore, many changes occurred in
the past with lasting negative effects, but the
drivers no longer occur at a particular location,
e.g., historical overfishing (4) or past coastal
habitat destruction (32). Although we used a
conservative, additive model, some drivers may
have synergistic effects (24). Despite these
limitations, this analysis provides a framework
and baseline that can be built upon with future
incorporation or refinement of data. It is note-
worthy that the data gaps emphasize the need for
research on the most basic information, such as
distributionofhabitattypesandwhetherandhow
different anthropogenic drivers interact.
Humans depend heavily on goods and ser-
vices from the oceans, and these needs will
likely increase with a growing human popula-
tion (10). Our approach provides a structured
framework for quantifying the ecological trade-
offs associated with different human uses of
marine ecosystems and for identifying locations
and strategies to minimize ecological impact
and maintain sustainable use. In some places,
such strategies can benefit both humans and
ecosystems, for example, using shellfish aqua-
culture both to provide food and to improve
water quality. Our analytical framework can
easily be applied to local- and regional-scale
planning where better data are available and can
be extended by incorporating other types of in-
formation, such as species distribution or diver-
sity data (33–35) to identify hot spots with
both high diversity and high cumulative hu-
man impacts that perhaps deserve conserva-
tion priority. A key next research step will be
Fig. 4. Total area affected (square kilometers, gray bars) and summed threat scores (rescaled units, black bars) for each anthropogenic driver (A)
globally and (B) for all coastal regions <200 m in depth. Values for each bar are reported in millions.
www.sciencemag.org
SCIENCE
VOL 31915 FEBRUARY 2008
951
REPORTS
on February 15, 2008
www.sciencemag.org
Downloaded from

Page 6

to compile regional and global databases of
empirical measurements of ecosystem con-
dition to further validate the efficacy of our
approach.
References and Notes
1. Pew Oceans Commission (POC), America's Living Oceans:
Charting a Course for Sea Change (POC, Arlington, VA,
2003).
2. U.S. Commission on Ocean Policy, An Ocean Blueprint for
the 21st Century: Final Report to the President and
Congress (U.S. Commission on Ocean Policy, Washington,
DC, 2004).
3. R. A. Myers, B. Worm, Nature 423, 280 (2003).
4. J. B. C. Jackson et al., Science 293, 629 (2001).
5. H. K. Lotze et al., Science 312, 1806 (2006).
6. J. P. M. Syvitski, C. J. Vorosmarty, A. J. Kettner, P. Green,
Science 308, 376 (2005).
7. P. M. Vitousek et al., Ecol. Appl. 7, 737 (1997).
8. D. Pauly, R. Watson, J. Alder, Philos. Trans. R. Soc.
London Ser. B 360, 5 (2005).
9. B. S. Halpern, K. A. Selkoe, F. Micheli, C. V. Kappel,
Conserv. Biol. 21, 1301 (2007).
10. W. V. Reid et al., Millennium Ecosystem Assessment:
Ecosystems and Human Well-Being—Synthesis Report
(World Resources Institute, Washington, DC, 2005).
11. L. B. Crowder et al., Science 313, 617 (2006).
12. E. W. Sanderson et al., Bioscience 52, 891 (2002).
13. D. Bryant, L. Burke, J. McManus, M. Spalding, Reefs at
Risk: A Map-Based Indicator of Threats to the World's
Coral Reefs (World Resources Institute, Washington,
DC, 1998).
14. N. Ban, J. Alder, Aquat. Conserv.: Mar. Freshwat. Ecosyst.
16, 10.1002/aqc (2007).
15. M. W. Beck, M. Odaya, Aquat. Conserv.: Mar. Freshwat.
Ecosyst. 11, 235 (2001).
16. D. Vander Schaaf et al., A Conservation Assessment of the
Pacific Northwest Coast Ecoregion (The Nature
Conservancy of Canada, Victoria, BC, and Washington
Department of Fish and Wildlife, Olympia, WA, 2006).
17. Materials and methods are available as supporting
material on Science Online.
18. J. M. Pandolfi et al., Science 307, 1725 (2005).
19. M. D. Spalding et al., Bioscience 57, 573 (2007).
20. U. R. Sumaila, J. Alder, H. Keith, Mar. Policy 30, 696(2006).
21. A. Clarke, C. M. Harris, Environ. Conserv. 30, 1 (2003).
22. J. E. Overland, M. Y. Wang, Geophys. Res. Lett. 34,
L17705 (2007).
23. J. M. Pandolfi et al., Science 301, 955 (2003).
24. B. S. Halpern, K. L. McLeod, A. A. Rosenberg, L. B. Crowder,
Ocean Coast. Manage. 10.1016/j.ocecoaman.2007.08.002
(2008).
25. D. Witherell, C. Pautzke, D. Fluharty, ICES J. Mar. Sci. 57,
771 (2000).
26. S. D. Kraus et al., Science 309, 561 (2005).
27. F. C. Coleman, W. F. Figueira, J. S. Ueland, L. B. Crowder,
Science 305, 1958 (2004).
28. R. L. Naylor et al., Nature 405, 1017 (2000).
29. R. Dalton, Nature 431, 502 (2004).
30. K. D. Lafferty, J. W. Porter, S. E. Ford, Annu. Rev. Ecol.
Evol. Systemat. 35, 31 (2004).
31. M. Shahidul Islam, M. Tanaka, Mar. Pollut. Bull. 48, 624
(2004).
32. L. Airoldi, M. W. Beck, Annu. Rev. Oceanogr. Mar. Biol.
45, 347 (2007).
33. A. P. Kerswell, Ecology 87, 2479 (2006).
34. A. R. G. Price, Mar. Ecol. Prog. Ser. 241, 23 (2002).
35. C. M. Roberts et al., Science 295, 1280 (2002).
36. This work was funded by the National Center for
Ecological Analysis and Synthesis (NCEAS) and supported
by the National Science Foundation and a grant from the
David and Lucile Packard Foundation to NCEAS to
evaluate ecosystem-based management in coastal oceans.
Thanks to J. Hutton at U.N. Environmental Programme
World Conservation Monitoring Centre for sharing coral
reef, mangrove, and seagrass ecosystem data, J. Guinotte
and the Marine Conservation Biology Institute for providing
ocean acidification data, E. Sanderson for comments on
earlier drafts, and support from the Pew Charitable Trusts to
the Sea Around Us Project for development of mapped
fisheries data. R. Myers participated in this project but
passed away before it was completed; we are grateful for his
contributions.
Supporting Online Material
www.sciencemag.org/cgi/content/full/319/5865/948/DC1
Materials and Methods
Figs. S1 to S6
Tables S1 to S6
References
16 August 2007; accepted 11 December 2007
10.1126/science.1149345
Effects of Predator Hunting Mode on
Grassland Ecosystem Function
Oswald J. Schmitz
The way predators control their prey populations is determined by the interplay between predator
hunting mode and prey antipredator behavior. It is uncertain, however, how the effects of such
interplay control ecosystem function. A 3-year experiment in grassland mesocosms revealed that
actively hunting spiders reduced plant species diversity and enhanced aboveground net primary
production and nitrogen mineralization rate, whereas sit-and-wait ambush spiders had opposite
effects. These effects arise from the different responses to the two different predators by their
grasshopper prey—the dominant herbivore species that controls plant species composition and
accordingly ecosystem functioning. Predator hunting mode is thus a key functional trait that can
help to explain variation in the nature of top-down control of ecosystems.
S
such as production, decomposition, and nitrogen
mineralization (1). These effects can be direct,
as when selectively feeding herbivores alter
plant community composition and hence alter
the quality and quantity of plant material en-
tering the soil organic matter pool to be de-
composed and mineralized (1–6); or indirect, as
when predators alter the way in which herbi-
vores affect plant community composition (7–10).
Theexactnatureofaspecies’effectwill,however,
depend on traits that determine the way it
pecies are most likely to have strong ef-
fects on ecosystems when they alter fac-
tors that regulate key ecosystem functions
functions (1, 11). Explaining such trait dependen-
cy is an important hurdle to overcome in
developing predictive theories of species effects
on ecosystem function (1). This endeavor is
currently hampered by a limited understanding
of what kinds of species’ traits control function-
ing (11–15).
Here I report on a 3-year experiment quan-
tifying the effect of one important functional trait
of top predator species—their hunting mode—
on the nature of indirect effects emerging at the
ecosystem level (Fig. 1). Predators can propa-
gate indirect effects down trophic chains in at
least two ways (16). They can alter the numer-
ical abundance of herbivore prey by capturing
and consuming them. Alternatively, their mere
presence in a system can trigger herbivore prey
to modify foraging activity in a manner that
reduces predation risk. A general rule, derived
from empirical synthesis, is that these different
kinds of effect are related to predator hunting
mode, irrespective of taxonomic identity (17).
Sit-and-wait ambush predators cause largely
behavioral responses in their prey because prey
species respond strongly to persistent point-
source cues of predator presence. Widely roam-
ing, actively hunting predators may reduce prey
density, but they produce highly variable pre-
dation risk cues and are thus unlikely to cause
chronic behavioral responses in their prey.
These hunting mode–dependent herbivore re-
sponses should lead to different cascading ef-
fects on the composition and abundance of
plant species within ecosystems (9, 18) that
should further cascade to affect ecosystem func-
tion (10). Predator effects do indeed cascade to
influence ecosystem functions, and they vary
withpredatorspecies(10,19–22).Butthebasis
for variation in predator species effectsremains
unresolved.
This study was carried out in a grassland
ecosystem in northeastern Connecticut. The im-
portant plant species in this ecosystem (deter-
mined by their interaction strengths) may be
effectively represented within three functional
groups of plants: (i) the grass Poa pratensis,
(ii) the competitively dominant herb Solidago
rugosa, and (iii) a variety of other herb spe-
cies, including Trifolium repens, Potentilla
simplex, Rudbekia hirta, Crysanthemum leu-
canthemum, and Daucus carota. The important
animal species are the generalist grasshopper
herbivore Melanopuls femurrubrum and the
spider predators Pisaurina mira and Phidippus
rimator (23). Pisaurina mira is a sit-and-wait
predator in the upper canopy of the meadow.
School of Forestry and Environmental Studies and De-
partment of Ecology and Evolutionary Biology, Yale Uni-
versity, New Haven, CT 06511, USA. E-mail: oswald.
schmitz@yale.edu
15 FEBRUARY 2008VOL 319
SCIENCE
www.sciencemag.org
952
REPORTS
on February 15, 2008
www.sciencemag.org
Downloaded from