[tt] Science: Diaz and Rosenberg: Spreading Dead Zones and Consequences for Marine Ecosystems

[Premise Checker]

Diaz and Rosenberg: Spreading Dead Zones and Consequences for 
Marine Ecosystems
Science 15 August 2008
Vol. 321. no. 5891, pp. 926 - 929
[Here's the Science article itself. Big Heat continues.]

Review

Robert J. Diaz1* and Rutger Rosenberg2

Dead zones in the coastal oceans have spread exponentially since
the 1960s and have serious consequences for ecosystem functioning.
The formation of dead zones has been exacerbated by the increase in
primary production and consequent worldwide coastal eutrophication
fueled by riverine runoff of fertilizers and the burning of fossil
fuels. Enhanced primary production results in an accumulation of
particulate organic matter, which encourages microbial activity and
the consumption of dissolved oxygen in bottom waters. Dead zones
have now been reported from more than 400 systems, affecting a
total area of more than 245,000 square kilometers, and are probably
a key stressor on marine ecosystems.

1 Virginia Institute of Marine Science, College of William and
Mary, Gloucester Point, VA 23062, USA.
2 Department of Marine Ecology, University of Gothenburg,
Kristineberg 566, 450 34 Fiskeb c ckskil, Sweden.

* To whom correspondence should be addressed. E-mail:
diaz{at}vims.edu

The visible ecosystem response to eutrophication is the greening of
the water column as the algae and vegetation in coastal areas grow
in direct response to nutrient enrichment. The most serious threat
from eutrophication is the unseen decrease in dissolved oxygen (DO)
levels in bottom waters, created as planktonic algae die and add to
the flow of organic matter to the seabed to fuel microbial
respiration (1). Hypoxia occurs when DO falls below <= 2 ml of
O[2]/liter, at which point benthic fauna show aberrant
behavior--for example, abandoning burrows for exposure at the
sediment-water interface, culminating in mass mortality when DO
declines below 0.5 ml of O[2]/liter (2). In most cases, hypoxia is
associated with a semi-enclosed hydrogeomorphology that, combined
with water-column stratification, restricts water exchange. More
recently, dead zones have developed in continental seas, such as
the Baltic, Kattegat, Black Sea, Gulf of Mexico, and East China
Sea, all of which are major fishery areas.

Although the anthropogenic fertilization of marine systems by
excess nitrogen has been linked to many ecosystem-level changes,
there are natural processes that can lead to nutrient enrichment
along continental margins that produce similar ecosystem responses.
Coastal upwelling zones associated with the western boundary of
continental landmasses are highly productive but are associated
with severe hypoxia (<0.5 ml O[2]/liter). These oxygen minimum
zones (OMZs) occur primarily in the eastern Pacific Ocean, south
Atlantic west of Africa, Arabian Sea, and Bay of Bengal, and are
persistent oceanic features occurring in the water column at
intermediate depths (typically 200 to 1000 m) (3). Where they
extend to the bottom, the benthic fauna is adapted to DO
concentrations as low as 0.1 ml of O[2]/liter. This is in stark
contrast to the faunal responses seen during recent
eutrophication-induced hypoxic events in coastal and estuarine
areas where DO concentrations this low led to mass mortality and
major changes in community structure (2).

Global Nature of Eutrophication-Induced Hypoxia

The worldwide distribution of coastal oxygen depletion is
associatedwith major population centers and watersheds that deliver
large quantities of nutrients (Fig. 1 and table S1). Most of these
systems were not hypoxic when first studied, but it appears that
from the middle of the past century, the DO concentrations of many
coastal ecosystems have been adversely affected by eutrophication.
The observed declines in DO have lagged about 10 years behind the
increased use of industrially produced nitrogen fertilizer that
began in the late 1940s, with explosive growth in the 1960s to
1970s (4). For marine systems with data from the first half of the
20th century, declines in oxygen concentrations were first observed
in the 1950s in the northern Adriatic Sea (5), between the 1940s
and 1960s in the northwestern continental shelf of the Black Sea
(6), and in the 1980s in the Kattegat (7). Localized declines of DO
levels were noted in the Baltic Sea as early as the 1930s, but it
wasn't until the 1960s that hypoxia became widespread (7).
Localized hypoxia had also been observed since the 1930s in the
Chesapeake Bay (8) and since the 1970s in the northern Gulf of
Mexico (9) and many Scandinavian coastal systems (7).
Paleo-indicators (foraminifera ratios and organic and inorganic
compounds) show that hypoxia had not been a naturally recurring
event in these ecosystems (10, 8). The number of dead zones has
approximately doubled each decade since the 1960s (fig. S1 and
table S1).

Figure 1 Fig. 1. Global distribution of 400-plus systems that have
scientifically reported accounts of being eutrophication-associated
dead zones. Their distribution matches the global human footprint
[the normalized human influence is expressed as a percent (41)] in
the Northern Hemisphere. For the Southern Hemisphere, the occurrence
of dead zones is only recently being reported. Details on each
system are in tables S1 and S2. [View Larger Version of this Image
(59K GIF file)]

Hypoxia tends to be overlooked until higher-level ecosystem effects
are manifested. For example, hypoxia did not become a prominent
environmental issue in the Kattegat until several years after
hypoxic bottom waters were first reported and fish mortality and
the collapse of the Norway lobster fishery attracted attention
(11). Although hypoxia in the northern Gulf of Mexico has affected
benthic communities over the past several decades, there is no
clear signal of hypoxia in fishery landings statistics (9).

Ecosystem-level change is rarely the result of a single factor, and
several forms of stress typically act in concert to cause change.
The shallow northwest continental shelf of the Black Sea provides
an example of a system stressed by eutrophication-driven hypoxia in
combination with other stressors, including overfishing and the
introduction of invasive species, all of which led to drastic
reductions in demersal fisheries. Nutrient inputs declined in the
1990s, hypoxia disappeared, and ecosystem services recovered;
however, nutrient inputs are again rising as agriculture expands
and a return to hypoxic conditions may be imminent (12). The key to
reducing dead zones will be to keep fertilizers on the land and out
of the sea. For agricultural systems in general, methods need to be
developed that close the nutrient cycle from soil to crop and back
to agricultural soil (13).

Degrees of Hypoxia

The most common form of eutrophication-induced hypoxia, responsible
for about half the known dead zones, generally occurs once per
year, in the summer after spring blooms--when the water is warmest
and stratification is strongest--and lasts until autumn (table S1).
The usual ecosystem response to seasonal oxygen depletion is
mortality of benthic organisms followed by some level of
recolonization with the return of normal oxygen conditions.
Higher-level trophic transfer from the benthos is limited by
seasonal hypoxia and can occur only when normal DO conditions
prevail (2).

Periodic oxygen depletion has been observed in about a quarter of
systems reported as hypoxic and may occur more often than
seasonally, but this tends to be less severe, lasting from days to
weeks. Many smaller systems, such as the York River in the
Chesapeake Bay (2), are vulnerable to periodic hypoxia because
local weather events and spring neap-tidal cycles influence
stratification intensity. Diel cycles that influence production and
respiration can also cause hypoxia that lasts only hours but has a
daily reoccurrence (14). The margins of seasonal dead zones may
also experience periodic hypoxic events influenced by wind and
tides (15).

Another 17% of the systems reported as hypoxic experience
infrequentepisodic oxygen depletion, with less than one event per
year, sometimes with years elapsing between events. Episodic
oxygendepletion is the first signal that a system has reached a
critical point of eutrophication, which, in combination with
physical processes that stratify the water column, tips the system
into hypoxia. In 1976, a single hypoxic event in the New York
Bightthat covered about 1000 km2 caused mass mortality of
demersalfishes and benthos and blocked the northward migration of
bluefish (Pomatomus saltatrix) (16). Many systems experience
episodic hypoxia before the onset of seasonal hypoxia, such as in
the northern Adriatic, Pomeranian Bay, and the German Bight.
Paleo-indicators and models from the northern Gulf of Mexico also
support this pattern of occurrence.

Because eutrophication increases the volume of organic matter that
reaches the sediments, there is a tendency for hypoxia to increase
in time and space. In systems prone to persistent stratification,
oxygen depletion may also persist. This type of persistent hypoxia
accounts for 8% of dead zones, including the Baltic Sea, the
largest dead zone in the world, as well as many fjordic systems.

Progression of Hypoxia

Coastal hypoxia seems to follow a predictable pattern of
eutrophication first enhancing the deposition of organic matter,
which in turn promotes microbial growth and respiration and
produces a greater demand for oxygen. DO levels become depleted if
the water column stratifies. In the second phase, hypoxia occurs
transiently, accompanied by mass mortalities of benthic animals.
With time and further buildup of nutrients and organic matter in
the sediments, a third phase is initiated, and hypoxia becomes
seasonal or periodic, characterized by boom-and-bust cycles of
animal populations. If hypoxia persists for years and organic
matter and nutrients accumulate in the sediments, a fourth phase is
entered, during which the hypoxic zone expands, and as the
concentration of DO continues to fall, anoxia is established and
microbially generated H[2]S is released. This type of threshold
response has been documented in the Gulf of Mexico (17), Chesapeake
Bay (8), and Danish waters (18).

The critical point in the response trajectory of an ecosystem to
eutrophication is the appearance of severe seasonal hypoxia.
Although some level of nutrient enrichment is a positive force in
enhancing an ecosystem's secondary productivity and, to a point,
fishery yields (19), eutrophication and seasonal hypoxia favor only
benthic species with opportunistic life histories, shorter life
spans, and smaller body sizes (2).

Ecosystem Responses

The effect of seasonal hypoxia on biomass and annual secondary
production is well documented (2, 9). What is not well understood
is how hypoxia affects the habitat requirements of different
species or the resilience of an ecosystem. Pelagic species will
experience habitat compression when hypoxia makes deeper, cooler
water unavailable in the summer (15) or overlaps with nursery
habitat (9). For example, the spawning success of cod in the
central Baltic is hindered by hypoxic water at the halocline (70 to
80 m), the depth where salinity is high enough to provide buoyancy
for cod eggs (20). Similar habitat compression occurs when sulphide
is generated in sediments. In this case, as the redox potential
discontinuity layer is compressed close to the sediment-water
interface, deeper-dwelling species, including the key bioturbators
that control pore-water chemistry (21), are eliminated. The
presence of Fe3+ and Mn4+ in the sediment may buffer the system
and reduce the formation of poisonous H[2]S. Reduced bioturbation
associated with hypoxia also alters sedimentary habitats by
disrupting nitrification and denitrification. Hence, under hypoxic
conditions, instead of nitrogen being removed as N[2] by
denitrification, ammonia and ammonium together with phosphorus are
the main fluxes out of reduced sediments (8, 22) and may stimulate
further primary production.

Habitat compression and the loss of fauna as a result of hypoxia
have profound effects on ecosystem energetics and function as
organisms die and are decomposed by microbes. Ecosystem models for
the Neuse River estuary (23), Chesapeake Bay (24), and Kattegat
(25) all show hypoxia-enhanced diversion of energy flows into
microbial pathways to the detriment of higher trophic levels (Fig.
2). Only under certain circumstances will demersal fish predators
be able to consume stressed benthic prey, because their tolerance
to low oxygen concentration tends to be less ( ~ 3 to 4 ml of
O[2]/liter) than that of the benthic fauna. Thus, it is only within
a narrow range of conditions that hypoxia facilitates upward
trophic transfer. As the benthos die, microbial pathways quickly
dominate energy flows. Ecologically important places, such as
nursery and recruitment areas, suffer most from energy diversion
into microbial pathways because hypoxia tends to occur in summer,
when growth and predator energy demands are high.

Figure 2 Fig. 2. Conceptual view of how hypoxia alters ecosystem
energy flow. The green area indicates the range of energy
transferred from the benthos to higher-level predators under
normoxia, typically 25 to 75% of macrobenthic carbon. As a system
experiences mild or periodic hypoxia, there can be a pulse of
benthic energy to predators. This "windfall" is typically
short-lived and does not always occur. With declining oxygen,
higher-level predation is suspended, benthic predation may continue,
and the proportion of benthic energy transferred to microbes rapidly
increases (orange). Under persistent hypoxia, some energy is still
processed by tolerant benthos. Microbes process all benthic energy
as hydrogen sulphide, and anoxia develops (red). [View Larger
Version of this Image (19K GIF file)]

Missing Biomass

Areas within ecosystems exposed to long periods of hypoxia have low
annual secondary production and typically no benthic fauna.
Estimates of the missing biomass in Baltic dead zones that are now
persistently hypoxic are ~ 264,000 metric tons of carbon (MT C)
annually (7) and represent ~ 30% of total Baltic secondary
production (26). Similarly, estimates for the Chesapeake Bay
indicate that ~ 10,000 MT C is lost because of hypoxia each year,
representing ~ 5% of the Bay'stotal secondary production (27). If
we estimate that ~ 40% of benthic energy should be passed up the
food chain in the Baltic (28) and 60% in the Chesapeake Bay (26),
when hypoxic conditions prevail, 106,000 MT C of potential food
energy for fisheries is lost in the Baltic and 6,000 MT C in the
Chesapeake Bay, respectively. In areas of the Gulf of Mexico that
experience severe seasonal hypoxia, benthic biomass is reduced by
as much as 1.4 MT C/km2 (9); assuming a 60% transfer efficiency,
this is equivalent to approximately 17,000 MT C of lost prey to
demersal fisheries.

Is the production lost during periodic hypoxia made up by the
ecosystem during normal conditions, or partly compensated for by
higher secondary production outside the dead zone? The latter seems
to occur in the Baltic, where secondary production outside the dead
zones has doubled as a result of eutrophication (26); but if the
dead zones were eliminated, the Baltic would be more productive by
at least a third to a half, assuming that organic matter was
processed through benthos instead of by microbes. In Chesapeake
Bay, because hypoxia dissipates after about 3 months, the entire
area affected is returned to production by recruitment (27). Aerial
estimates of missing biomass for about a third of the world's dead
zones (table S1) indicate that as much as 343,000 to 734,000 MT C
is displaced over a total area of 245,000 km2 as a result of
hypoxia.

The duration of seasonal hypoxia then becomes the primary factor
affecting ecosystem energy flows. Within most systems that have
strong seasonal cycles, increases in populations are related to
recruitment events timed to take advantage of the input of new
organic matter, usually a spring or autumn bloom; populations
normally decline from a combination of resource depletion and
predation (29). Thus, the shorter the interval between recruitment
and the onset of hypoxia, the greater the negative effect on the
upward flow of energy in the food chain. During persistent hypoxia,
there is a drastic reduction in secondary production, and microbes
remineralize virtually all organic matter.

Recovery

By the end of the 20th century, oxygen depletion of marine systems
had become a major worldwide environmental problem, with only a
small fraction (4%) of the 400-plus systems that had developed
hypoxia exhibiting any signs of improvement (table S1). These
improvements in DO were related to reductions in three factors:
organic and nutrient loadings, stratification strength, and
freshwater runoff.

>From 1973 to 1990, the hypoxic zone on the northwestern continental
shelf of the Black Sea had expanded to 40,000 km2; however, since
1989, the loss of fertilizer subsidies from the former Soviet Union
reduced nutrient loading by a factor of 2 to 4, with the result
that, by 1995, the hypoxic zone had gone (12). As oxygen levels
normalized, ecosystem function improved, and the benthic fauna
started to recolonize but have not recovered to prehypoxic levels.
In the Gulf of Finland, a decrease in water-column stratification
occurred between 1987 and 1994, which improved DO conditions and
facilitated the return of benthic fauna (7); however, with the
return of stratification, conditions have again deteriorated.

In the northern Gulf of Mexico, the occurrence and extent of the
dead zone are tightly coupled with freshwater discharge from the
Mississippi River, which delivers large quantities of nutrients
from U.S. agricultural activities. During years with low river
flow, the area of hypoxia shrinks to <5000 km2, only to increase
to >15,000 km2 when river flow is high (30).

The management of nutrients and carbon inputs has virtually
eliminated dead zones from several systems, including the Hudson
and East Rivers in the United States and the Mersey and Thames
Estuaries in England (31, 32). However, in other systems, such as
the Chesapeake Bay, the management of nutrient input has not
improved DO. Nevertheless, the management of sewage and pulp mill
effluents has led to many small-scale reversals in hypoxia (table
S1).

The key factors in determining the degree of ecosystem degradation
are the duration of exposure and DO concentration. It may take
years to recover from severe hypoxia and, moreover, the tolerance
to oxygen depletion of mature community species may not mirror that
of the opportunistic species that are the first colonizers. The
benthos of many coastal areas may be reestablished by larval
recruitment and succession, as described in the Pearson-Rosenberg
model (33); however, the pattern of species that establish during
recovery will not be the same as the pattern of species loss during
DO deterioration, and consequently a hysteresis-like response will
be observed (Fig. 3). A pronounced hysteresis-like response was
documented in Gullmarsfjord, Sweden, which suffered hypoxia for
about half a year, during which time the fauna was eliminated in
deeper areas and diversity and abundance were reduced to less than
one-third at medium depths (34). Within 2 years, the benthic
community had recovered to the same community composition and
density that had existed before the hypoxic event (Fig. 3). In this
fjord, sedimentary redox conditions had not become intensely
reducing, and rapid colonization occurred by larvae from benthic
communities in adjacent undisturbed areas. Should hypoxia prevail
for more than 5 years, recovery would be prolonged (35) and the
hysteresis-like response exaggerated, as was recently observed in
the Black Sea, where recovery of the benthos after hypoxia in 1994
was still not complete in 2004 (36).

Figure 3 Fig. 3. Generalized pattern of benthic community response
to hypoxia (34). As DO declines to <0.7 ml of O[2]/liter and extends
through time, mass mortality of both equilibrium (stage III) and
opportunistic (stage I) species occurs (red). If anoxia is reached,
benthos are eliminated. The recovery path from severe hypoxia is
different than the decline path because of the hysteresis-like
progression of successional dynamics. When exposed to mild hypoxia,
mortality is moderate, and the recovery path is closer to the
response path (blue) as fauna restart from midsuccessional stage II.
When exposed to intermediate oxygen conditions, the response is
minor (green) and not hysteresis-like. [View Larger Version of this
Image (33K GIF file)]

Prospects for Change

Further expansion of dead zones will depend on how climate change
affects water-column stratification and how nutrient runoff affects
organic-matter production. General circulation models predict that
climate change alone will deplete oceanic oxygen by increasing
stratification and warming as well as by causing large changes in
rainfall patterns (37), enhancing discharges of fresh water and
agricultural nutrients to coastal ecosystems. For example, climate
predictions for the Mississippi River basin indicate a 20% increase
in river discharge, which will elevate nutrient loading and lead to
a 50% increase in primary production, a 30 to 60% decrease in
subpycnocline DO, and an expansion of the oxygen-depleted area
(38). Conversely, if the climate becomes stormier and
stratification decreases because of increased mixing, the risk of
oxygen depletion will decline. Tropical storms and hurricanes
influence the duration, distribution, and size of the Gulf of
Mexico dead zone in a complex way. In 2005, four hurricanes (Cindy,
Dennis, Katrina, and Rita) disrupted stratification and aerated
bottom waters. After the first two storms, stratification was
reestablished and hypoxia reoccurred, but the total area was a
fourth less than predicted from spring nitrogen flux. The other two
hurricanes occurred later in the season and dissipated hypoxia for
the year (30).

Climate change also has the potential to expand naturally occurring
OMZs into shallower coastal waters (3), damaging fisheries and
affecting energy flows in the same way that eutrophication-driven
hypoxia does. There is currently about 1,148,000 km2 of seabed
covered by OMZs (<0.5 ml of O[2]/liter), and a small change in
oceanographic processes could lead to a major expansion of these
zones. Areas at greatest risk for expanding OMZs encompass the
western continental shelves of South America, Africa, and the
Indian subcontinent, where extensive OMZ and upwelling areas
already exist. The development of dead zones along the western
coast of other countries is highly likely if wind patterns shift
and cause stronger upwelling. This effect might explain the recent
development of a dead zone off the coast of Oregon (39).
Furthermore, there is a possibility that increased loadings of
terrestrial nutrients have contributed to an expansion of the OMZ
on the western Indian continental shelf (40).

The weight of evidence indicates that in pre-industrialized times,
most coastal and offshore ecosystems never became hypoxic except in
natural upwellings. However, measuring the effects of hypoxia on
ecosystems is complicated by many factors, not least of which is
the inadequate data on historic trends in DO concentrations and
faunal populations, as well as the combined effects of multiple
stressors, including fishing and habitat loss. It is the recurring
nature of hypoxia that alters an ecosystem's state and prevents
full recovery of function.

Currently, hypoxia and anoxia are among the most widespread
deleterious anthropogenic influences on estuarine and marine
environments, and now rank with overfishing, habitat loss, and
harmful algal blooms as major global environmental problems. There
is no other variable of such ecological importance to coastal
marine ecosystems that has changed so drastically over such a short
time as DO. We believe it would be unrealistic to return to
preindustrial levels of nutrient input, but an appropriate
management goal would be to reduce nutrient inputs to levels that
occurred in the middle of the past century, before eutrophication
began to spread dead zones globally.

References and Notes

* 1. N. N. Rabalais, R. E. Turner, W. J. Wiseman, Annu. Rev. Ecol.
  Syst. 33, 235 (2002).
* 2. R. J. Diaz, R. Rosenberg, Oceanogr. Mar. Biol. Annu. Rev. 33,
  245 (1995).
* 3. J. J. Helly, L. A. Levin, Deep Sea Res. Part I Oceanogr. Res.
  Pap. 51, 1159 (2004).
* 4. J. N. Galloway et al., Science 320, 889
  (2008).
* 5. D. Justi c , T. Legovi c , L. Rottini-Sandrini, Estuar.
  Coast. Shelf Sci. 25, 435 (1987).
* 6. L. D. Mee, Ambio 21, 278 (1992).
* 7. K. Karlson, R. Rosenberg, E. Bonsdorff, Oceanogr. Mar. Biol.
  Annu. Rev. 40, 427 (2002).
* 8. W. M. Kemp et al., Mar. Ecol. Prog. Ser. 303, 1 (2005).
* 9. N. N. Rabalais, R. E. Turner, Eds., Coastal Hypoxia:
  Consequences for Living Resources and Ecosystems (American
  Geophysical Union, Washington, DC, 2001).
* 10. B. K. Sen Gupta, R. E. Turner, N. N. Rabalais, Geology 24,
  227 (1996).
* 11. R. Rosenberg, Mar. Pollut. Bull. 16, 227 (1985).
* 12. L. Mee, Sci. Am. 295, 78 (2006).
* 13. D. Tilman et al., Science 292, 281
  (2001).
* 14. R. M. Tyler, T. E. Targett, Mar. Ecol. Prog. Ser. 333, 257
  (2007).
* 15. D. Breitburg, Estuaries 25, 767 (2002).
* 16. E. V. Garlo, C. B. Milstein, A. E. Jahn, Estuar. Coast. Mar.
  Sci. 8, 421 (1979).
* 17. R. E. Turner, N. N. Rabalais, D. Justi c , Environ. Sci.
  Technol. 42, 2323 (2008).
* 18. D. J. Conley et al., Ecol. Appl. 17, S165 (2007).
* 19. S. W. Nixon, B. A. Buckley, Estuaries 25, 782 (2002).
* 20. M. Cardinale, J. Modin, Fish. Res. 41, 285 (1999).
* 21. R. C. Aller, Chem. Geol. 114, 331 (1994).
* 22. R. A. Duce et al., Science 320, 893
  (2008).
* 23. D. Baird, R. R. Christian, C. H. Peterson, G. A. Johnson,
  Ecol. Appl. 14, 805 (2004).
* 24. D. Baird, R. E. Ulanowicz, Ecol. Monogr. 59, 329 (1989).
* 25. T. Pearson, R. Rosenberg, Neth. J. Sea Res. 28, 317 (1992).
* 26. R. Elmgren, Ambio 18, 326 (1989).
* 27. R. J. Diaz, L. C. Schaffner, in Perspectives in the
  Chesapeake Bay: Advances in Estuarine Sciences, M. Haire, E. C.
  Krome, Eds. (Chesapeake Research Consortium, Gloucester Point,
  VA, 1990), pp. 25-56.
* 28. P. Möller, L. Pihl, R. Rosenberg, Mar. Ecol. Prog. Ser. 27,
  109 (1985).
* 29. G. Graf, Oceanogr. Mar. Biol. Annu. Rev. 30, 149 (1992).
* 30. N. N. Rabalais et al., Estuar. Coasts 30, 753 (2007).
* 31. C. A. Parker, J. E. O'Reilly, Estuaries 14, 248 (1991).
* 32. P. D. Jones, Mar. Pollut. Bull. 53, 144 (2006).
* 33. T. H. Pearson, R. Rosenberg, Oceanogr. Mar. Biol. Annu. Rev.
  16, 229 (1978).
* 34. R. Rosenberg, S. Agrenius, B. Hellman, H. C. Nilsson, K.
  Norling, Mar. Ecol. Prog. Ser. 234, 43 (2002).
* 35. R. Rosenberg, Oikos 27, 414 (1976).
* 36. L. D. Mee, J. Friedrich, M. T. Gomoiu, Oceanography
  (Washington D.C.) 18, 100 (2005).
* 37. Intergovernmental Panel on Climate Change, Climate Change
  2007: the Physical Science Basis, Contribution of Working Group
  I to the Fourth Assessment Report of the Intergovernmental Panel
  on Climate Change (Cambridge Univ. Press, New York, 2007).
* 38. D. Justi c , N. N. Rabalais, R. E. Turner, Limnol. Oceanogr.
  41, 992 (1996).
* 39. F. Chan et al., Science 319, 920  (2008).
* 40. S. W. A. Naqvi et al., Nature 408, 346 (2000).
* 41. E. W. Sanderson et al., Bioscience 52, 891 (2002).
* 42. This work was supported in part by National Oceanic and
  Atmospheric Administration Coastal Hypoxia Research Program
  grant NA05NOS4781202 to R.J.D. We thank N. Rabalais and L.
  Schaffner for discussions and critical comments; A. Puente, K.
  Sturdivant, and S. Lake for helpful discussions; H. Burrell for
  artwork on the figures; and H. Berquist for producing the global
  map. This is contribution 2942 of the Virginia Institute of
  Marine Science.